human superior colliculus activity relating to spatial working memory

Posted comment on ´Spatially specific working memory activity in the human superior colliculus` by M. Rahmati, K. DeSimone, C.E.Curtis, K.K.Sreenivasan and published in Journal of Neuroscience 2020 vol 40 (49) p. 9487 doi 10.1523/JNEUROSCI.2016-20.2020

SUMMARY

   In their article, Rahmati and colleagues describe their findings relating to the neural activity of the human superior colliculus (SC) representing location used in the performance of spatially specific working memory tasks.

  Rahmati and colleagues` study involved generated visual stimuli linked to a MRI scanner, a button box and eyetracker. The 6 human subjects viewed visual images on a screen with trigger pulses from the scanner so that the onsets of stimulus presentations and image acquisition were synchronised. Eye gaze was constantly monitored and the data was automatically pre-processed and scored. Various quantifications were made during the course of the experiments, eg. accuracy was quantified using absolute Euclidian distance between the saccade landing point and the true target location; precision was quantified using the average SD of tangential and radial components of the saccade landing points; and the response times of the visual (VGSs) and memory guided saccades (MGSs) were quantified and time courses and trajectories of saccades were plotted. Fixation during the delay was found to be stable with fixation breaks in 0.5-2% of the trials for five subjects with the sixth making quick saccades away from and then back in 7% of the trials.

   MRI data was acquired from 14 coronal slices and 20 functional series of 120 volumes were collected for the retinotopic mapping task and 20 functional runs of 232 volumes for the spatial working memory task. Functional scans were matched to anatomic images by collecting single whole brain coverage functional images with the same spatial resolution as the partial brain coverage functional images and then motion correcting the data and co-registering the functional data with the anatomic images.

  Population receptive fields (pRF) in the SC voxels were modelled according to established procedures. Subjects were presented with a black and white bar with reversed contrast at a frequency of 8HZ that orientated either vertically or horizontally on the screen and swept across it perpendicular to the bar orientation and passing through the central fixation. Four 30 second sweeps made up each scanning run. For the fixation task, subjects were asked to detect and map the colour of a fixation cross (red, green, blue, yellow) to one of 4 button presses. For the mapping experiment, each voxel was modelled in terms of Gaussian pRF which gave each voxel`s BOLD response in terms of retinotopic location and extent. The delay in hemodynamic response function (HRF) (time to peak and time to undershoot of the HRF) and the baseline of the BOLD signal (ensures that modelled and measured BOLD signals vary about the single global mean) were also modelled.

  A ROI for the SC was drawn for each subject based on anatomy and the pRF model. The fovea (FOV) of the SC map was estimated from the full pRF model. It was represented in the full SC map in visual space by 2D Gaussians whose positions within the visual field and widths were determined by each voxel`s pRF centre and size parameters and whose maximum value equalled 1. The pRF parameters were aggregated across the left and right SC of all subjects. Since many points in the visual field were covered by several pRFs, combining the pRFs meant that each visual field coordinate was mapped to the maximum pRF value.

   For the spatial working memory representations, the SCs of the subjects were imaged whilst the subjects maintained the stimuli locations in working memory during a long memory retention interval. The subjects were presented with a brief visual stimulus at the periphery with its colour indicating the transformation required to remap its location to the goal of the later MGS, eg. white – no transformation, green – MGS to the location mirrored across the horizontal meridian etc. This stimulus was followed by a black dot so that the subjects made a visually guided saccade (VGS) to this target followed immediately by an MGS to the transformed location guided by memory. The stimulus was then re-presented at the correct original location for performance feedback. This corrective saccade was followed by an interval of 9.8secs and then the process was repeated. Each scanning session lasted nearly an hour.

  The authors used two task manipulations in order to eliminate simple visual or motor components while focusing in on working memory representations of visual space. The first manipulation was that  intermediate VGS prevented subjects from being able to plan and potentially maintain the metrics of the MGS during the delay. The second manipulation was that various transformations moved the task-relevant location, the goal of the MGS, to a position in visual space that was independent of the retinal position of the visual stimulus. In order to reconstruct a representation of spatial working memory from the pattern of SC activity during the delay, authors used the inverted encoding model (IEM). They modelled each voxel`s response during trials in which the stimulus and MGS target were at the same location. This was then used to estimate the contribution of each of the nine information channels which were then all averaged to reconstruct the populations activity. The trials were combined to increase signal-to-noise ratio and repeated IEM training and reconstruction procedures were used to eliminate bias. Appropriate statistical analyses were carried out on all experiments including a modified version of the representational fidelity metric.

   Rahmati and colleagues modelled the pRF properties of voxels in the SC by overlaying the pRF model parameters on the T1 anatomic image for a representative stimulus and individual subjects to obtain a detailed topographic organisation. They found orthogonal polar angle and eccentricity representations of the visual field along the SC. There was a graded upper-to-lower visual field representation along the medial-to-lateral axis of the SC and a graded foveal-to-peripheral visual field representation along the anterior to posterior axis.  The pRFs mainly tiled the contralateral visual field. The shape of the distribution of polar angles suggested an under-representation of angles near the vertical meridian although a complete RF model shows that visual stimulation at most retinal locations drives the SC. The team found a positive correlation between their human SC topographic structures to those of cat and macaque.  

   Further examination of the images showed that neural activity of the SC persisted throughout a retention interval of several types of modified memory guided saccade tasks. It was found that the accuracy, precision and latency of the visual and MGSs were similar to other studies of this type. Performance did not differ across task conditions although visual guided saccades (VGSs) were slightly longer in latency than those recorded in previous reports. The average BOLD signal in the retinotopically defined human SC persisted above pre-trial baseline during the memory period. This suggested that the SC may play an important role in working memory. However, the MGS target was not lateralised with the pattern incongruous with the clear lateralisation of the SC pRFs. Reasons for this lack of lateralisation were given by the authors as: contamination of the delay activity by the earlier visual stimulation; that it may reflect the subject`s covert attention to the entire visual field in anticipation of the intervening VGS target; that it may result from the complex transformations required by the task; or that the averaging over many voxels may result in a measure that is too inaccurate to capture the population dynamics by which the SC encodes the working memory.

   Therefore, because of the lack of lateralisation and the pRFs observed, Rahmati and colleagues used a multivoxel encoding model where the multivoxel population response is mapped into the coordinates of visual space with the assumption that the neural architecture of the SC is based on retinotopic organisation. So that spatial working memory could be modelled, Rahmati and colleagues first trained the model with no transformation of the visual cued target to give the estimated voxels preferred polar angle. Then, using circular correlation, the pRF and IEM polar angle parameters were confirmed as being similar. This suggested that the two forward modelling approaches converged on very similar polar angle parameters.

  Following verification, the model was then trained with trials requiring transformation. The authors found that if the SC population delay activity encoded the spatial information of the working memory representation then the model could accurately reconstruct the transformed location of the MGS. The locations stored in the working memory were computed from spatial transformations of the visual targets and were not locations that were retinally stimulated earlier in the trial. The models trained on the location of the visual target or the VGS location were unable to reconstruct these locations which indicated that the SC delay activity encoded the abstract representation of the location from memory rather than the visually presented targets. The SC population activity during the delay was spatially tuned for only the location of the MGS target. Therefore, Rahmati and colleagues concluded that their results showed that the topographic pattern of activity in human SC represents locations stored in working memory.

   Rahmati and colleagues concluded their article by discussing the contribution of human SC population activity to working memory. The SC is known to belong to a wide number of cortical and subcortical brain areas that form distributed networks supporting this particular type of cognitive function. The human SC visual field map has been identified with the neural representations of the upper part of the visual field found in the medial part of the SC map, the lower parts of the visual fields in the lateral part, the foveal representations in the anterior part of the SC map and the peripheral parts in the posterior area of the SC map. The size of the estimated RFs of the voxels correlates to the eccentricity where smaller RFs are nearer the fovea. According to Rahmati, the mapping of visual stimulation and saccades is likely to originate from the same map. There is the same anisotropic distribution of angles as other studies with lesser representation along the upper and lower vertical meridians compared with the horizontal meridian. The fovea is estimated as covering the whole visual field.

   The encoding model used by the team leads to the clear neural mechanism by which the topographically organised SC encodes working memory representations. The activity in the retinotopic SC persisted during working memory retention intervals since low, persistent BOLD activity was observed. This was in agreement with the electrophysiological recordings from macaque SC neurons that typically show a slow, but increased rate of discharge before saccades including MSGs. However, the delay activity was not contra-lateralised with respect to location of working memory targets as expected and a number of reasons were given by the authors as to why this might occur. This suggested that although the delay activity points to SC playing some role in spatial working memory, it did not indicate precisely what that role is. The multivoxel encoding model was used to elucidate this. The authors found that the spatial location of the working memory representation was encoded at the population level and that the patterns of delay period activity did not encode the retinal positions of past visual stimuli or planned future saccades. The locations held in the working memory were abstract transformations of visual stimulated locations and therefore, the pattern of SC neural population activity encoded the abstract, cognitively defined locations in the absence of visual stimulation or motor commands.

  The authors then went on to discuss the nature of the SC signalling in relation to working memory functioning. They indicated that their model could not confirm whether the SC signals originated from the intermediate and/or deep layers of the SC where more cognitive processing is thought to occur, even though these neurons show persistent activity during MGS delays. Rahmati and colleagues continued with a discussion of whether the SC encoding is initiated by feedback signals from the cortex, or not. There is support for this view since if this is the case, population activity would resemble spatial attention effects as reported for the macaque SC. For example, the visually evoked SC responses are larger when the stimulus is behaviourally relevant and the goal of a saccade; SC neurons with RF matching an attended target also show enhanced discharge rates when task-related saccades are dissociated from the locus of attention; macaque SC plays a role in covert attention; and attention causes enhanced SC neural responses in humans. If this is valid then likely sources of these top-down influences are the lateral PFC, FFF, lateral intraparietal area and V1 which are all areas known to support spatial working memory and have direct connections to the macaque SC. However, Rahmati and team continued by saying that it is unlikely that the SC simply integrates cortical commands and relays them to the brainstem oculomotor areas. This is because: representations of visual priority emerge more rapidly in the SC than in the V1 indicating that feedback signals from the SC may influence the gain of responses in the cortex as in the case of macaques; the SC has more ascending projections through the pulvinar and mediodorsal thalamus that could influence the cortex than descending projections arriving; and lesions of the SC impair behaviours that depend on covert attention, but do not affect  the typical attentional enhancement of neuronal activity in the extrastriate cortex. Therefore, Rahmati and colleagues concluded that the SC may play critical roles in spatial cognition such as attention and working memory through circuits that both interact with, but at the same time are independent of the cortex.

    The discussion was concluded by looking at the interaction of visual and motor information which the SC collectively encodes. As reported for the visual cortex, the authors suggested that the neural populations of the SC may encode a probability distribution where the population response may encode the probability of a prioritised (including remembered) location.  This was suggested as an area for future research of the SC along with how cortical and subcortical areas differ in their support of working memory.

COMMENT

What makes this article interesting is that it looks at a brain area, the superior colliculus (SC), that normally receives little attention when talking about cognition, but as Rahmati has shown plays a role in spatial working memory. The SC is located in the dorsal midbrain sitting below the thalamus and above the inferior colliculus, both of which share connectivity with it. It has a branch, the superior brachium, that extends laterally from it to the lateral geniculate body of the thalamus and partly into the optic tract. This connectivity indicates its role in the visual pathway.

  The SC has a layered (superficial and intermediate/deep) and columnar structure which aids incoming information from different sources, integration of signals and specificity of outgoing message. Each layer contains a topographic map with each point having retinotopic coordinates. Therefore, activation of neurons at any point in the map will evoke a response directed towards a corresponding point in visual space. This allows visual information and top-down instruction to be changed into visually related behavioural response (eg. eye movement). Therefore, specificity of input to the SC will correlate to specificity of output whether movement-related or responding cortical activity.

  The SC layers are categorised into superficial, intermediate or deep. The superficial layers receive input from the retina (and layer lamina III from the optic tract) and the cortex, primarily the V1 (Brodmann 17), secondary visual areas 18 and 19 and frontal eye fields. This indicates that this layer is under control of the retina and certain cortical areas and supports the role of the SC in visual input, working memory and eye movement. Output of the superficial layers is to the pulvinar and lateral intermediate areas of the thalamus which then project to the cortical areas controlling eye movements. The superficial layers also project to the lateral geniculate nucleus, the pretectal nuclei (both having direct connections from the optic tract with the pretectum controlling the size of the pupil and certain types of eye movement) and the parabigeminal nucleus.

  The intermediate layers receive information and respond to other sensory modalities. The deeper layers have a population of motor-related neurons capable of activating eye movements. They also have input from most of the cortical areas (top-down control), as well as inhibitory action via GABA functioning from the basal ganglia areas (substantia nigra and pars reticulata – involved in motor movement control) and input from the spinal trigeminal nucleus (conveys information from the face), the hypothalamus, thalamus and the inferior colliculus (auditory pathway role). Again, like the superficial layers, output is to the pulvinar and lateral intermediate areas of the thalamus which would then project to the cortical areas controlling eye movements. This means that different input into the various layers can lead to the same effect. There is also direct output from the intermediate/deep layers to the brain stem, pons and reticular formation so that unconscious processing can lead to faster responses than those via the cortex.

  The ultimate effect of the SC is in directing eye movements (gaze shifts) so that the visual subject becomes or remains in the fovea, the area of the retina with the highest visual sensitivity. This is important in terms of relating visual awareness to arm movements for example, attention or in the case of memory where repeated or sustained activation of visual cortical areas is needed to convert short term sensory stores to long term memories. The type of eye movement discussed in Rahmati`s article is that of saccadic eye movement which is where the eyes move rapidly from one location to another one and this movement is commonly demanded in spatial working memory type tasks. Imaging of neuronal activity by MRI shows that neurons fire according to known topographic organisation. It is known that in the case of saccades, the neurons are activated in a region that represents the points to which the saccade will be directed (ie. the target location) with activity increasing rapidly just prior to the saccade. The area of neuronal activity is wide, described as a ´hill` with the ´peak` being the target location. The firing in the SC then translates into eye movements via its output projections whether cortical, pretectum or quicker via the brain stem for example.

   Rahmati and team investigated neuronal firing in the SC during a working memory task demanding location awareness, spatial memory and transformation to target location. Therefore, it is clear that neuronal activity in this particular SC area would correlate to the demands placed upon it and the study shows that the neural population activity of the human SC encodes the spatial location of the working memory representation. If we look at Rahmati`s experiment in more detail we can see which cognitive skills are required for each stage and where the SC fits in. Since the team`s experiment was a spatial working memory task as usual the first stage is the giving of instructions and describing what the demands and goals of the task are. Therefore, the subjects have to know and remember using short term memory and procedural memory skills the task itself as a whole and in particular the relevance of the colour of the dot and the specific action that the presentation of this colour demands. From an attention point of view, the initiation of the task will lead to a general state of readiness and this is seen in Rahmati`s experiments where the visual field is likely to extend to represent the whole computer screen. The general state of readiness allows more focus of the visual pathway on relevant rather than irrelevant information.

  The next stage of the experiment requires activation of the visual pathway and this is where the first dot is presented at the periphery. The visually guided saccade (transformation) leads to a shift gaze to go from the natural point of fixation up to that time to the presented coloured dot at the periphery, ie. from one location to another, from one area to a target location, with information encoded in the colour, but not the shape. The location and shift in location requires activity in the SC and this is seen in Rahmati and team`s images with input coming from the retinal ganglion cells (about 10% of overall activity), as well as possibly via the optic tract-pretectum (controlling the size of the pupil and certain types of movement). Therefore, stage 3 is the instigation of the sensory store (and even short term memory store) of the visual information and location information inputted.

   It is the next stage, stage 4, where working memory state is activated. This is where the incoming visual information is processed in order to carry out the required action. In stage 4, the visual information passes up through the system from the retina to the thalamus (likely lateral geniculate nucleus) and up to the cortex where the working memory state is activated. Therefore, memory recall mechanisms are activated and the instructions given at the start of the experiment recalled. Working memory state has to process immediate location and calculate the new location based on the instructions given. This new location is stored in short term memory and employs visual imagery mechanisms. At the information processing stage it is likely that the SC is not required, but once the working memory state has calculated where the new location should be and the interval gone through top-down instructions from the cortex to the SC occur so that the required gaze shift from the interval black dot stimulus to the new location is achieved when required. This forms the memory guided saccade (MGS) and is observed in the MRI scans of the SC taken by Rahmati and team. It is likely that this MGS occurs prior to perception if the neuronal activity of the saccade precedes action. The final stage of the working memory task in Rahmati`s experiment is the feedback stage and again working memory state is activated to process whether there is a match between new location and original location. Successful processing means no target shift whereas a gaze shift is required if the calculated location is inaccurate.

   Therefore, by going through the cognitive requirements needed for successful completion of this task it can be seen that although SC activity is required for this type of task, its role is in providing information about transformations of location if the input is from the retina or carrying out gaze shifts according to top-down demands following information processing. The SC then provides a general state of readiness to location shifts as given in the instructions, VGS from the natural point of gaze to the target when the dot is presented, VGS when the interval stimulus is presented and MGS in moving from this stimulus to the required new location according to working memory state instructions. The signal to the SC for this MGS is likely to be from the dorsolateral prefrontal cortex (DLPFC) which is known to share connectivity with the SC area (Johnston). The view is supported by the results of other studies similar to Rahmati`s experiments, but less complex where monkeys perform memory guided saccade tasks where they are required to generate saccades to remembered stimulus locations. Lesions of the right posterior parietal and right DLPFCare known to cause deficits in spatial working memory task performance (Asselen) and activity of the DLPFC area requires GABA (Yoon) and glutamate receptor activity (van Vugt). In humans there also appears to be a requirement for the precentral sulcus as well for working memory representations and MGS where a delay is part of the experimental set-up (Mackey).

   Although difficult to visualise, the SC is not responsible for actual visual characteristics of the stimulus. This comes from other areas of the brain which are also active in the working memory state. For example, it is known that separate senses have different working memory locations (Pasternak) such as MT, IT, V4 for visual information, S1 and S2 for tactile information and auditory information split with time, frequency and intensity of sound all having separate brain area modules.  Also, behavioural studies show that SC is not needed for object recognition as given above, but it plays a critical role in the direction of actions towards specific objects even in the absence of the cerebral cortex, eg. cats with lesions to the visual cortex cannot recognise objects, but the animals can follow and orient toward moving stimuli even if at a slower speed than normal.

  It is likely that due to its columnar and layered structure the SC can integrate information from multiple sources and therefore, auditory information coming via the inferior colliculi can be associated with gaze shifts via the connectivity of the area to the SC intermediate/deep layers. This forms a multisensory integrated neuronal population which relays information to the necessary areas. It has been reported that there is multisensory integration of audio and visual information regardless of temporal asynchrony, but at different time points (Liu). However, we can ask the question whether multiple neural representations from the same sensory pathway can exist simultaneously within the human SC, ie. is it possible to have awareness of 2 gaze shifts at any one time? The answer to this is likely to be no although it might be possible to have one gaze shift, but with multiple neuronal populations alight at any one time representing multiple object locations moving in unison within the same visual field. The resulting neuronal population activity would likely be broadened with no single hill peak and no targets discernible since imaging techniques normally require averaging of active voxels The neuronal population activity therefore, would represent the whole visual field.

   Two observations support this view. The first is that there is in the SC competition between successive visual stimuli even when the overlap is very small since the stimuli are seen as separate representations. This is shown by a study where human subjects are given the task of making saccades to targets that are separated by distance and whose presentation are separated by time. The subjects were asked if they thought the stimuli overlapped or if there was a gap. Although the saccades were similar, subjects misjudged the targets as being separate if they even overlapped by less than 100msec (Pretegiani). This means that the neural representations in the SC relating to location are distinct even if the presentation of the stimuli are fleeting. Delay between presentation can also have an effect on population activity observed. Although a second representation lags behind the first by an enforced delay as occurs in Rahmati`s experiments (MGS comes after VGS and working memory intervention and the interval given is according to experimental set-up) then the second representation is not as great as one where the saccade occurs immediately, ie. another VGS (Sadeh). Spatial coding however, remains the same (Sadeh).

  Support for single gaze shifts also comes from the lack of lateralisation seen with human SC. There are normally two SC (left and right) with each representing half the visual field with the fovea being represented at the front edge of the map and the periphery at the back. However, each human SC represents only the contralateral half of the visual field up to the midline with the ipsilateral half excluded due to the lack of connectivity between the retinal ganglion cells of the temporal half of the retina and the contralateral SC. Rahmati and team also observed that only the contralateral half of the visual field was represented in their scanned images and also observed no lateralisation in their spatial working memory task for the memory guided saccade. This they put down to contamination of the neuronal population activity during the delay which was maintained by the earlier VGS, covert attention to the entire visual field, the complex transformations required by the task (eg. VGS, VGS  followed by MGS followed by VGS) or by experimental failure to capture population dynamics because of the averaging process across many voxels. All of these are valid reasons why lateralisation is not observed with human SG, but it can also be that not all visual input from one eye is transmitted to its opposite counterpart further up the visual system and therefore, for example, activity in the right SC may reflect mainly input from the left retina, but there may also be visual information incoming from the right eye. It is also possible that lateralisation relates only to input to the SC from the retina and not top-down from the cortex, as would be the case for working memory state and the calculated and enforced MGS. This would explain why lateralisation is not observed in Rahmati and team`s MGS experiment. 

   A single gaze shift is also supported by the attentional system which shows that visually evoked SC responses are larger when the stimulus is behaviourally relevant and the goal of the saccade (Soto). Lesions in the SC can result in increased distractibility or relevancy/irrelevancy disruption. Selective attention to a single gaze shift would be the case in the spatial working memory task where there is an overlap of attention on the VGS and working memory which would lead to working memory benefits (Soto). Firing characteristics of these neurons during the early parts of the task means that they would be critical for selecting the MGS (Lovejoy). Attention allows the persistence of stable neural representations that would hold the information temporarily and would be necessary for working memory state activity and resulting action (Spaak).

   Therefore, we can conclude that the midbrain superior colliculus is a neuronal relay centre responsible for instigating neural population activity representing visual location shifts. It does this by having a columnar, layered structure that allows input from the visual pathway in a bottom-up process and top-down from various cortical areas including the dorsolateral prefrontal cortex, part of the working memory state network. Firing of populations prior to saccades correlates to shifts in eye gaze and connectivity to other modalities allows eye movements to be associated to other incoming sensory information (eg. auditory, touch) or top-down influences (eg. visual memory, working memory). Although not a major player in thinking and information processing mechanisms, the capability to shift gaze so that the visual event is in the foveal part of the visual field is advantageous to visual perception and to memory formation where sustained neuronal firing is required. Therefore, the SC should not be ignored and should be regarded as a contributing factor to visual system and network performance required for visual input and behavioural action based upon it.

Since we`re talking about the topic…………………………..

                ……working memory training is seen to improve working memory performance in children (Astle). Would training where eye gaze is controlled improve working memory performance in cases where distractibility is an issue? Would similar results be achieved with mindfulness meditation?

                …certain drugs have effects on working memory performance (eg. MDMA increases spatial awareness, decreases visuospatial working memory – McCann). Can we assume that MDMA administration for example would have a negative effect on spatial working memory performance if Rahmati`s experiments are repeated and would there be noticeable differences in eye gaze shifts and SC neuronal population activity?

                …..positive emotions are known to enhance verbal memory and impair spatial working memory whereas negative mood enhances spatial memory and impairs verbal working memory (Storbeck). If Rahmati and team`s experiments are repeated, can we assume that negative mood or anxiety would affect the spatial working memory performance in general and possibly this could be equated to poorer capability regarding eye gaze fixations and saccadic eye movements as measured by SC neural activity?

Posted in superior colliculus, working memory | Tagged , | Comments Off on human superior colliculus activity relating to spatial working memory

ultrasound effect via mechanosensitive ion channel Piezo1 of neurons

Posted comment on ´The mechanosensitive ion channel Piezo1 significantly mediates In Vitro ultrasonic stimulations of neurons` written by Z.Qui, J. Guo, S.Kala, J.Zhu, Q.Xian, G.Li, T. Zhu, L.Meng, R.Zhang, H.ChangChan and H.Zheng and published in IScience vol 21 November 2019 p. 448 doi 10.1016/j.sci.2019.10.037

SUMMARY

   Qui and colleagues report in their article the results of their investigation into the role that piezo-type mechanosensitive ion channel component 1 (piezo1) has on mediating the effects of ultrasound in mouse primary cortical neurons and a neuronal cell line, CLU199.

   For their experiments, Qui and colleagues used mouse primary cortical neurons. The mouse piezo1 channel was transfected into HEK293T cells. The expression of functional piezo1 in piezo1-transfected cells (piezo1) compared to vector control (ctrl) was verified by qRT-PCR and Western Blot analyses. Piezo1 functionality was verified by stimulating the cells with the 1uM piezo1 agonist Yoda1 and performing calcium imaging. The effect of ultrasound was investigated by looking at calcium imaging compared to ultrasound stimulation at different intensities of piezo1-transfected cells compared to the control. The ultrasound was applied using Qui and team`s customised ultrasound stimulation system incorporated with a calcium imaging system. This meant that the authors could generate acoustic maps of controllable ultrasound.

  For primary neurons obtained from embryonic mice cortex, piezo1 expression was examined using immunocytochemical staining MAP2 and c-FOS staining (a molecular marker for neuron activation that is responsive to calcium influx). Imaging studies were carried out on cells treated with 0.3MPa ultrasound with or without pre-treatment with 20uM piezo1 blocker, GsMTx-4. The functionality of piezo1 in primary neurons was examined by calcium imaging of cells stimulated with Yoda1 with or without the pre-treatment with GsMTx-4. Ultrasound was applied at 0.45MPa at varying intensities. The levels of piezo1 in the mouse neuronal cell line CLU199 were verified by PCR and Western Blots. Western Blots were also carried out to determine the expression levels of p-CAMKII, p-CREB and c-Fos in CLU199 cells treated with varying ultrasound intensities.

   Piezo1 was also knocked down in the CLU199 cells using non-targeting (Ctrl) or Piezo1 siRNA (piezo1 KD). Expression was verified using qRT-PCR. Again, calcium imaging on Yoda1 treatment was performed. Western blots were also carried out to determine the expression levels of p-CAMKII, p-CREB and c-Fos in the CLU199 cells treated with siRNA and varying ultrasound intensities.

  The authors did report some limitations and problems with their study which led to their set-up being used. For example, a common problem also reported by others was that patch clamping could not be used to demonstrate piezo1 channel opening because the recording pipette of the system was found to be incompatible with the low-frequency ultrasound; and vibration caused by ultrasound in the base of the hard-surfaced lysine-coated glass culture dish could lead to an underestimation of the acoustic pressure required to active the piezo1 channels. Appropriate statistical analyses were carried out on results obtained, eg. unpaired two-tailed T tests (eg. calcium imaging of primary neurons stimulated with Yoda1) and one way ANOVA with post-hoc Tukey test (eg. for primary neurons stimulated with ultrasound and Western blot images in KO-CLU199 cells).

   Qui and colleagues found in their experiments that ultrasound alone at low intensity could activate heterologous expressed and endogenous expressed piezo1 and that the effect of this ultrasound exposure was the initiation of calcium influx. Although HEK293T cells normally exhibit no mechanical stimulation, heterologous overexpressed mouse piezo1 in the HEK293T cells (transfected cells with a pcDNA3.1-mPiezo-IRES-GFP plasmid) showed significant induced calcium ion influx when treated with the piezo1 specific agonist, Yoda1 following transfection. Increasing ultrasound intensity led to increased calcium influx compared to the control where no influx was observed.  Pre-treatment with the piezo1 specific inhibitor, GsMTx-4 prevented the influx. Therefore, it was concluded that ultrasound stimulation at low intensity could successfully activate piezo1 containing cells and calcium influx was the result.

   The authors continued their study by looking at the effects of ultrasound on primary cortical neurons. Neurons harvested from embryonic mice and mouse pups at P3 also gave the same results as the transfected cells (eg. calcium influx on Yoda1 exposure, but pre-treatment with GsMTx-4 leading to a significant reduction). Therefore, it was concluded that these cells had endogenous piezo1 expression. Ultrasound stimulation resulted in dose-dependent calcium ion influx into the neurons which was prevented by pre-treatment with GsMTx-4. An experiment to investigate c-Fos expression (identified by MAP2 staining) in the nuclei of the primary neurons led to the c-Fos level being found to be significantly increased with ultrasound treatment and this reduced significantly with GsMTx-4 pre-treatment. Therefore, endogenous piezo1 channels found on mouse primary cortical neurons were found to be activated by low-intensity ultrasound and this led to calcium ion influx and c-FOS expression increase.

   The study continued with an investigation into the downstream effects of calcium ion influx induced by the ultrasound exposure. For this group of experiments the authors used a mouse hippocampal cell line mHippoE-18 (CLU199) and looked at the levels of phosphorylated activated form of Ca2+/calmodulin-dependent protein kinase type II (p-CaMKII) and the phosphorylated active form of the transcription factor CREB (p-CREB). Ultrasound treatment produced dose-dependent increased levels of p-CaMKII and p-CREB. A repeat of the experiment to see the effect on c-Fos expression in the cell line by ultrasound produced the same increased effect as that seen in the primary neurons. The dependence of the ultrasound effect on piezo1 channels was confirmed by looking at expression in piezo1 knockdown CLU199 mice. These showed no significant up-regulation. Therefore, the up-regulation of the p-CaMKII, p-CREB and c-Fos can be attributed to the ultrasound action on the piezo1 channels.

   From their results, Qui and colleagues concluded that there was endogenous piezo-type mechanosensitive ion channel component 1 (piezo1) expression in mouse primary neurons and in their chosen neuronal cell line, CLU199. The activity of the piezo1 channels was stimulated by the agonist, Yoda1 and by low frequency low intensity ultrasound and this activation led to calcium ion influx. Therefore, piezo1 activity and calcium ion influx were determined as being important for the mediation of the intracellular effects induced by the ultrasound exposure. The ultrasound was also found to affect the levels of downstream calcium signalling proteins involved in neuronal function. This was indicated by the increased expression on ultrasound stimulation of the activated forms of CaMKII and CREB, both of which play roles in neuronal plasticity and cognitive functioning such as learning and memory. C-Fos involved in neuronal maturation was also found to be involved.

   Therefore, Qui and colleagues proposed that piezo1 activation is plays a major role in the mechanical transduction of ultrasound and stimulus of this mechanosensitive transmembrane protein is capable of significantly affecting function and activation of neurons in vivo via its effect on intracellular calcium concentration and calcium-dependent protein expression. These observations support the idea of ultrasound treatment having therapeutic effects in the brain and that it may be used to advantage in treating certain conditions. This aspect of the piezo1 functioning according to the authors requires further study and future research may include manipulation of up- and down-regulation of appropriate genes and brain area specificity and distribution.

COMMENT

  What makes this article interesting is it demonstrates the presence of another group of receptors that can influence intracellular functioning of neurons. The use of ultrasound gives an indication of the functioning of the mechanosensitive transmembrane proteins, piezo1 and this comment focuses on how and why the activation of this receptor type is important to the modulation of neuronal cell firing. 

   Mechanoreceptors are transmembrane proteins that act as channels. Piezo1 is one example of the Piezo group and these transmembrane proteins are highly conserved and likely to be around 2500 amino acids long, have between 24–32 transmembrane regions and be assembled into tetramers. For experimental purposes, the piezo1 protein is activated by the agonist, Yoda1 and blocked by the extracellular inhibitor peptide, GsMTx-4, both compounds used by Qui in their experiments and described in the article above.  However, in normal circumstances in vivo and in vitro, mechanoreceptors of this type respond to local mechanical pressure or distortion of the membrane and piezo1 is one of the most sensitive of the piezo group to these types of endogenous stimuli.

  The piezo1 reacts to the agonist or endogenous stimuli of extracellular membrane pressure or distortion by changing its structural conformation and interactions with the membrane around it. Membrane lipids appear to be essential for its function (Cox) and it has been shown that in particular the organisation of membrane contained cholesterol is important for piezo1 responses (Ridone). The piezo1 channels are thought to be associated with one another in the membrane as clusters and this community-like structure, therefore fits in with membrane lipid rafts hypothesis (cholesterol rich areas) for embedded proteins. This is supported by observations that other areas also highly sensitive to mechanical forces eg. hearing, touch, have the same structure. The phosphoinositides (PIP1, PIP2 and PIP3) are also thought to be important for channel functioning since they appear to be highly enriched around the channels. There also appears to be a binding site consisting of 4 lysine residues (K2166-K2169) around the channel pore domains in humans, but this is more likely to be important in switching the channel off since deletion mutation experiments remove channel inactivation. This appears to be involved in cases of repeated stimulation.

   The result of the activation of the piezo1 of neurons by endogenous or applied means is the initiation of calcium ion influx into the cell and corresponding increases of levels of c-Fos and down-stream calcium signalling proteins (CaMKII and CREB) involved in neuronal function are observed.  Therefore, piezo1 can translate mechanical/physical signals from the external environment into intracellular neurochemical ones and it does so by using the common calcium ion messaging system. 

   We are used to in the neurochemical world of linking ion channel functioning with physical receptor binding of endogenous ligands (agonists and antagonists) and this binding action inducing chemical effects down the line (cascade signalling). Therefore, the action of mechanoreceptors or to give them their proper name these mechanosensitive transmembrane proteins introduces another aspect of neuronal modulation. If piezo1 is responsive to local changes in pressure or distortion of the membrane in which it is located, the question is: where does this pressure or distortion come from in the neuronal area?

  The first possible cause of mechanical pressure and distortion on the membrane could come from the membrane`s own local environment and may be instigated by changes in the bilayer structure and its naturally occurring components, eg. phospholipids. This is an obvious cause since we have already seen that piezo1 efficiency is affected by the lipid rafts in which the clusters of proteins sit. This also ties in with the effect of applied ultrasound on piezo1 functioning. In this case, ultrasound acts as a ´mechanical stimulus`. High intensity ultrasound can lead to membrane permeabilization by ´breaking` the bilayer structure by either an indirect mechanism involving the membrane surface/interface or by direct interaction with the bilayer by causing expansion and contraction inducing conformational changes of the bilayer structure itself. These would provide the necessary stimulus to induce piezo1 action since conformational changes of the piezo channel structure especially since it is tetrameric would occur in response to membrane structural changes.

   Therefore, since the ´force` comes from the external environment it is implied that mechanical stimuli of this nature ´push` at the membrane. However, the current view is that piezo channels are relatively insensitive to ´pushing` whereas ´pulling` at the membranes (ie. from inside the cell and as a result of intracellular events) activate the piezo channel effectively. It appears that this type of sensitivity is elicited through collagen IV which is a component of the basal lamina forming a cohesive network and mechanical connection between cells. This implies that collagen V may also be required for piezo1 functioning. The observations that ultrasound and external administration of ligands as well as unknown internal stimuli can both lead to calcium influx supports the view that the conformational structure of the piezo1 protein can be modulated by the membrane structure around it and that the lipid bilayer organisation is important for its functioning.

   Another cell characteristic which has both internal and external components is the cytoskeleton and the piezo1 channel can be stimulated via changes to the cytoskeleton and extracellular matrix which exists in its immediate external environment. In the case of the cytoskeleton, this again links to local mechanical pressure modulating embedded proteins. In other mammalian cells integrins form adhesions on the membrane surface which allow mechanical stimuli to be focused on different cytoskeleton components which indirectly transmit the signal to membrane proteins/ion channels. This links to the view that the cytoskeleton works as a ´tension producing` structure that pre-stresses cells to keep the cell shape stabilised and it does this by a network of opposing tension and compression components. Therefore, according to the ´force for lipids` hypothesis (Syeda) the function of the cytoskeleton (and extracellular matrix) is to mechanically protect the lipid bilayer by absorbing mechanical stresses so that the embedded proteins and channels are less likely to suffer changes. Some cytoskeleton proteins in particular, eg. STOML3 and tropomyosin 4.2, appear to pre-stress the membrane and this leads to increased mechano-sensitivity of piezo1 to membrane tension whereas filamin A is shown to reduce it.

   However, the relationship between piezo1 and the cytoskeleton is not clear cut since piezo1 is constitutively active in liposomes, indicating that no cytoskeletal elements are required for its function and the application of CytochalasinD which leads to the disruption of the cytoskeleton component, F-actin results in the loss of whole cell currents in HEK293 cells, but piezo activity is unaffected. The action of ultrasound on the cytoskeleton also provides support that it is not involved in the piezo1 action in neurons since under normal circumstances, ultrasound application results in the rearrangement and disassembly of the cytoskeleton internally via both components, actin and tubulin. Therefore, if the cytoskeleton is involved in piezo functioning then results for both piezo1 ligand and ultrasound should not be the same. One explanation for the discrepancy is that it may indicate differences between the intracellular and extracellular located cytoskeleton. However, since both are composed of the same constituents and ultrasound acts on these components this argument for inaction may not be valid. Therefore, it is unlikely that the endogenous piezo1 responds to mechanical pressure in a mechanism that involves the cytoskeleton and may imply that processes that rely on cytoskeleton for internal transport such as exocytosis and endocytosis are not susceptible to piezo1 channel activation. 

  The other component of the external environment of the neuron that could affect the mechano-sensitive transmembrane proteins is the extracellular matrix (EM) itself and mechanical stimulation of piezo1 has been found to be dependent on channel/interactions with EM proteins(Gaub). The EM constitutes about 20% of the total volume of the adult brain and is a 3D network composed of molecules such as collagen, enzymes, glycoproteins. It surrounds the cells providing structure and biochemical support. Therefore, changes in the EM can manifest as stimuli for mechanosensitive transmembrane proteins action which then would take this external influence and translate it into an intracellular chemical one.

  One factor that may cause a change in the mechanical pressure from the EM on the cell membrane of neurons is increasing water volume of the EM and the effect of swelling. This link between piezo1 and water volume comes from the group of constituents of the EM that play a role in water management. These are the proteoglycans which are made up of the carbohydrate polymers, the glycosaminoglycans (GAGS) which are attached to the EM proteins. Proteoglycans because of their net negative charge attract positively charged sodium ions which attract water molecules via osmosis. This leads to the ´trapping` of water so that the EM and cells remain hydrated. Hyaluronic acid, another proteoglycan, found in the extracellular space absorbs significant amounts of water and confers upon the cells the ability to resist compression by providing a counteracting swelling force. With regards to piezo1 action, the excitability of the neuron and depolarisation causes changes in ion concentrations across the membrane. Sodium ions flow out which will bind to the net negatively charged proteoglycans and promote water trapping in the EM. This swelling of the EM may therefore be the stimulus required to initiate piezo1 activity and calcium ion influx, which coincides with the calcium ion influx through neurotransmitter linked channels at the later part of neuronal depolarisation. Therefore, the action of the piezo1 supplements the normal calcium influx from the depolarised active neuronal cell leading to cell recovery.

  The involvement with depolarisation and sodium ions indicates another way in mechanosensitive transmembrane protein piezo1 activity can influence neuronal functioning and this relates not to its mechanical pressure sensitivity, but to its calcium influx function. Piezo1 activity appears to be modulated by voltage and this implies that this modulation unlike the signal from mechanical pressure coming from outside in is in reverse and the stimulus is inside out. The channels can activate and deactivate in the range of milliseconds and have a reversal potential around 0 mV and can show voltage dependent inactivation, but also can exhibit non-inactivation (ie. permanently switched on) as under conditions of LTP. Normally the intracellular concentration of calcium is low since calcium phosphorylated proteins are insoluble and the level is maintained like this by calcium pumps, intracellular calcium binding proteins and organelles such as the mitochondria. However, neuronal firing produces a change in intracellular concentration.

   If we assume that the piezo1 protein acts like other calcium ion channels in the neurons then the action potential induces calcium influx into the pre-synapse which leads to neurotransmitter release into the synaptic cleft via exocytotic mechanisms. The neurotransmitter binds to the appropriate post-synaptic receptors and initiates the mechanisms for the transmission of the signal. For example, binding of receptors linked to G proteins leads to the secondary messenger cAMP production resulting in PIP3 production, further release of calcium and also, CaMKII and PKC activation. Therefore, in these circumstances the influx of calcium in the pre-synapse would lead to firing and signal transmission at the post-synapse. Piezo1 proteins may aid this process instigated by other calcium ion channels or perhaps be sole instigator of the calcium influx. If the piezo1 proteins are located post-synaptically instead, then again the calcium influx from the extracellular space or synaptic cleft by the action of the piezo1 into the post-synaptic area may aid the simultaneous neurotransmitter-receptor associated calcium ion influx (eg. by NMDAR and linked SK channels or some AMPAR linked to calcium ion channels). This would support the initiation of mechanisms dependent on post-synaptic calcium ion concentrations, eg. CaMKII activation by phosphorylation. The characteristic of having the reversal potential at 0mV would mean that the mechanosensitive transmembrane protein, piezo1 works like other neuronal ion channels and that firing would lead to ion influx until the point when the membrane potential reaches 0mV. However, it is possible that at 0mV instead of changing flow direction or ion specificity (piezo1 only prefers calcium ions it can cause flow of other ions like sodium, too), the channel then inactivates (possibly through the lysine residues at K2166-K2169) and this supplementary calcium flow mechanism (or even other ions) stops. The consequence of this is that from this point the concentration of calcium in the pre- and post-synaptic areas from influx is purely dependent on firing status.

   The situation, however, changes with repeated or continual stimulation which would normally lead to long-term potentiation (LTP). LTP is advantageous in these circumstances and forms the basis of the neurochemical mechanisms relating to memory because it provides the brain area with localised increased synaptic strength and increased susceptibility of the neurons to depolarisation. LTP is achieved through the trafficking and clustering of AMPARs to the post-synaptic membrane surface. A turnover of PIP3 is required which is also involved in the membrane bilayer structure around the piezo1 protein. One type of AMPAR is linked to sodium or calcium ion channels (L type channel) and therefore, calcium influx will result on glutamate binding and this leads to the post-synaptic mechanisms that are dependent on calcium ions, eg. PKC activation. In the case of the piezo1 action, LTP appears to be linked to non-inactivation of the transmembrane protein, ie. that the channel remains open independent of voltage. The consequences of this would mean that the down-stream signalling associated with high calcium concentration would be independent of actual firing status. This would be an advantage since continual firing of neuronal cells is not possible because the cells need to recover (the refractory period) and perform ´housekeeping duties` in order to maintain firing or fire again.  Therefore, uninhibited piezo1 action would lead to continuity of pre- and post-synaptic calcium influx and associated calcium-dependent mechanisms, such as neurotransmitter release by exocytosis since the presence of calcium can briefly increase vesicle fusion rates one millionfold above spontaneous rates in (Schneggenburger).

   Another interesting point about piezo channels and neuronal firing refers to their role in brain  functioning as a whole and combines voltage dependency and mechanical pressure sensitivity. If the piezo channels are low-pressure sensors (Wang) reacting to gentle mechanical pressure then it is suggested that they may also react to the low-grade mechanical changes mediated by changes in blood pressure and breathing. Therefore, there is a suggestion that in neurons the piezo mechanosensitive transmembrane proteins provide specific brain neurons with an ´intrinsic resonance` that acts to synchronise their firing with the normal pulsed-like nature of changes in intracranial pressure associated with breathing and cardiac cycles. This ´resonance` would likely come via the extracellular matrix/space, and hence, transmit to the intracellular environment via the mechanoreceptors. The view is that these associations would manifest as a ´global brain rhythm` which would begin with the mechanosensitive neurons within the olfactory epithelium reacting to nasal airflow leading to synchronisation via direct neuronal signal connectivity to the activity of higher neural networks in the cortex and hippocampus for example (Wang). Most attempts at proposing a ´global rhythm` come via firing changes, eg. brain waves and therefore, it would be unlikely that such a rhythm is purely dependent on mechanical pressure changes in higher animal species, but instead that this form of signal transmission supplements the normal neurotransmitter/ion channel one.

   The role of piezo mechanosensitive transmembrane proteins in neuronal firing modulation is supported by the reported effects of ultrasound on neuronal firing. Firstly, piezo mechanosensitive transmembrane proteins are found in only certain cell types and only in a few brain areas, eg. pyramidal neurons in the neocortex and hippocampus, Purkinje cells in the cerebellar cortex and mitral cells in the olfactory bulb, but at different densities. This would mean that there would be different levels of impact from external mechanical influences dependent on area and that the intracellular effects would also be variable. This is supported by the effects of ultrasound on the different brain areas where ultrasound is reported to affect areas such as the cortex, hippocampus and thalamus, but appears to have no effect on the striatum or amygdala.

  The neurochemical effects of the piezo mechanosensitive transmembrane protein stimulated by endogenous means can also be compared to the neurochemical effects observed on ultrasound exposure. Qui reported in their article that ultrasound at low-intensity could activate endogenous piezo1 channels, initiating calcium influx and increasing levels of c-Fos and downstream calcium signalling proteins such as phosphorylated CaMKII and CREB. This correlated to the effects seen with the applied agonist Yoda1. Coadministration of the piezo1 antagonist GsMTx-4 with ultrasound elicited no response indicating that ultrasound was the stimulus at the piezo1 proteins in the experimental samples. The preference of piezo1 for calcium ions does not exclude that influx of other ions is also possible and this is comparable to that seen with ultrasound exposure, eg. sodium ion channels activation and potassium channels mixed effects dependent on area. As given above the neurochemical effects can be the result of induced structural changes of the membrane bilayer and ultrasound is known to cause membrane perturbation and pore formation. In this case, then we can interpret the ultrasound as a ´pull signal` with the implication that collagen IV is involved in its action.  

   Therefore, to conclude, Qui and team`s article on the mechanosensitive transmembrane protein, piezo1 adds another dimension to modulation of neuronal firing and brain area functioning. This modulation comes from two properties of the protein, ie. the ability to react to mechanical pressure, and/or the ability to react to depolarisation. The former is largely ignored when discussing neuronal functioning, but there are still questions to be answered about how and what controls firing of neuronal cells in particular brain areas, eg. the reason why the hippocampus exhibits hyperexcitability and therefore, further studies of this type of modulation are important. 

Since we`re talking about the topic……….

                …….since piezo protein expression can be examined using anti-PIEZO2 antibody and staining techniques, can we assume that if neurons are put under stress this technique will give an indication of the role of piezo1 proteins in the cell`s response? Would the administration of a calcium ionophore (A23187 or ionomycin) aid in confirming the piezo1 role?

…… interactions with membrane lipids are essential for piezo1 function (Cox), therefore would the enrichment of membrane cholesterol, depletion of membrane cholesterol (by methyl-beta-cyclodextrin) or disruption of membrane cholesterol organisation (by dynasore) lead to expected changes in piezo1 function when the agonist, Yoda1 is used?

 …..ruthenium red is known to inhibit piezo currents in cells, can we assume therefore, that administration of ruthenium red simultaneously with ultrasound exposure would prevent calcium influx and support the link between ultrasound action and piezo1 function?

Posted in calcium ions, mechanosensitive ion channels, ultrasound | Tagged , , | Comments Off on ultrasound effect via mechanosensitive ion channel Piezo1 of neurons

neuron-derived estrogen role in astrocyte activation and neuroprotection in ischemic injured hippocampus

Posted comment on ´Neuron-derived estrogen is critical for astrocyte activation and neuroprotection of the ischemic brain` written by Y. Lu, G.R. Sareddy, J. Wang, Q. Zhang, F.-L. Tang, U.P. Pratap, R.R. Tekmal, R.K. Vadlamudi and D. W. Brann and published in Journal of Neuroscience 2020 40 (38) p. 7355 doi 10.1523/JNEUROSCI.0115-20.2020

SUMMARY

   The authors, Lu and colleagues, describe in the article their investigation into the role of neuron-derived 17 beta estradiol (E2) in astrocyte activation in the hippocampus CA1 region following global cerebral ischemia. They found that E2 is part of the reactive astrocyte neuroprotective response and is linked to the suppression of FGF2 signalling.

   In their experiments, Lu and team used the FBN-ARO-KO mouse model where the forebrain aromatase is deficient and hence, E2 production in the brain is lacking. Intact males and bilateral ovariectomised females plus sham animals were used of the FBN-ARO-KO species compared to FLOX mice having normal forebrain aromatase activity. The global ischaemia condition was set-up by using the two-vessel occlusion global cerebral ischemia (GCI) model (ie. cardiac arrest and hypotensive shock). The mice were subjected to GCI and for most experiments the hippocampal CA1 tissue lysates and slides were prepared 7 days later. Various experiments were carried out. Western Blots, RNA sequencing, fluorescent imaging, GFAP staining and confocal single-plane imaging (for astrocyte 3D images) were performed on the different mouse groups and appropriate image processing and RNA sequencing analyses were carried out on the results obtained. E2 levels were measured using high-sensitivity ELISA kits. Experiments where the level of E2 was rescued by the administration of exogenous E2 were undertaken using 3 month old ovariectomised female mice and these were divided into 4 groups (FLOX+GCI, KO+GCI, KO+GCI+E2 and KO+GCI+placebo) with the control being the placebo group. The endogenous E2 was given at a dose known to effectively restore forebrain E2 levels in the FBN-ARO-KO mice. In the experiments where the effect of neuronal FGF2 was assessed, 4 groups of mice were constructed (FLOX+GCI, KO+GCI, KO+GCI+FGFR3-neutralising antibody and KO+GCI+vehicle). Bilateral intracerebroventricular microinjections of FGFR3 antibody at the time of the GCI were given where necessary. The experiments to measure astrocyte activation and astrocyte phenotypes were performed at 7 days after GCI. This was also the time frame used for the assessment of cognitive capability which was assessed using three different types of test: the Barnes Maze test for spatial reference learning and memory; the Novel Object Recognition Test for hippocampal dependent recognition memory; and the Open Field Test for testing locomotor function. Appropriate statistical analyses were carried out where necessary.

   Lu and team`s experiments showed that ovariectomised female FBN-ARO-KO mice that had GCI had significantly increased astrocyte activation in the hippocampal CA1 region after 3, 7 and 14 days as shown by GFAP intensity. The level at 7 days was higher than both 3 days and 14 days suggesting that this was the highest point of activation. Compared to the FLOX mice after GCI, the FBN-ARO-KO mice showed lower levels of astrocytic activation in general (140% to 200%) which suggested to the authors that the loss of neuron-derived E2 causes astrocytes to be less activated after GCI compared to the control, FLOX+GCI mice.

   With regards to aromatase, the aromatase double staining experiment and the GFAP measurement showed that aromatase was specifically located in the hippocampal CA1 neurons of the FLOX+SHAM mice. The expression of aromatase of neurons was strongly decreased in the FBN-ARO-KO+SHAM mice compared to the control, FLOX+SHAM mice. It was not detected in the resting astrocytes of either the FBN-ARO-KO+SHAM mice or the control FLOX+SHAM mice. However, GCI strongly induced aromatase in the astrocytes of the FLOX+GCI mice (approx. 250% of control) whereas in the astrocytes in the FBN-ARO-KO+GCI mice the level was significantly decreased (approx. 50% of the control value). Using the ELISA kits, E2 levels of the different groups were measured. It was found that hippocampal E2 levels in ovariectomised female FLOX-GCI mice were significantly increased 7 days later than FLOX-SHAM (approx. 120% of the control value).  However, FBN-ARO-KO+SHAM mice had only approx. 33% of the FLOX+SHAM mice and there was no significant increase in E2 production either after the GCI.

   The authors produced 3D images of the astrocytes of the hippocampus CA1 and showed that there were no differences in the volumes of the astrocyte cell bodies between the FLOX+SHAM and FBN-ARO-KO+SHAM mice. This confirmed that the resting states of the two species were the same. However, GCI was found to strongly increase the volume of the astrocytes of the FLOX mice indicating that the GCI produced a robust induction of reactive astrocytes. The volume of the astrocytes of the FBN-ARO-KO+GCI mice was found not to be significantly increased compared to the SHAM control and it was significantly lower than the FLOX+GCI equivalent. This indicated that the induction of reactive astrocytes in the FBN-ARO-KO mice after GCI did not occur. The results were confirmed with the investigation of the expression levels of the typical astrocyte markers, GFAP and vimentin. Both of these were robustly increased in the FLOX mice 7 days after GCI, but the increases were not observed in the FBN-ARO-KO mice. (Although increases were seen these were said to be insignificant – 20% GFAP, 18% for vimentin.) Therefore, reactive astrogliosis was said not occur in the FBN-ARO-KO model. By using immunostaining techniques for GFAP and S100beta cells where no differences were observed between the various models, the authors could relate the decrease seen in the FBN -ARO-KO mice to decreased astrocyte activation and not to astrocyte loss or proliferation.

   Lu and team then continued their investigation by looking at the level of neuronal damage following GCI. Using double staining for NeuN (marker for neurons) and F-Jade C (marker for neuronal degeneration), the authors found that in ovariectomised female FBN-ARO-KO+GCI mice there was significantly increased levels of F-Jade C positive pyramidal hippocampal neurons 7 days after GCI compared to the control FLOX+GCI mice. This indicated a higher level of neuronal degeneration in the FBN-ARO-KO mice.  There was no F-Jade C staining in either FLOX+SHAM or FBN-ARO-KO-SHAM mice showing that neither SHAM models exhibited neuronal damage. A study of MAP2 staining showed that the neuronal structural integrity of the FBN-ARO-KO+GCI mice was poorer for these mice than the control FLOX+GCI mice.

  The authors continued their investigation of the role of neuron-derived E2 in hippocampal function following ischemic injury by looking at behavioural cognitive capability. This was carried out by using the following behavioural tests: Barnes Maze test for spatial memory; Novel Object Recognition Test for object recognition; and Open Field Test for locomotor function. The results of the Barnes Maze Test showed that the FLOX+GCI mice displayed a significant increase in escape latency, an increase in exploring errors and a decrease in quadrant occupancy compared to the FLOX+SHAM mice. This indicated impaired spatial reference memory recall following the ischemic event. The FLOX+SHAM mice performed in general better than the FBN-ARO-KO+SHAM mice showing that they too exhibited cognitive deficits. However, there were even greater increases in escape latency and in exploring errors and greater decreases in quadrant occupancy in the FBN-ARO-KO+GCI mice compared to the FLOX+GCI mice. The differences could not be attributed to speed variations since the escape velocities for all groups were identical. In the case of the Open Field Test, no differences were shown between the subject groups showing that loss of forebrain neuronal E2 does not affect locomotor function under normal or GCI conditions. Therefore, the authors concluded that spatial memory capability was decreased with their FBN-ARO-KO model and this decrease was amplified by GCI.

  In order to explore the nature of the astrocytic reactivity observed following GCI, Lu and team performed RNA sequence analyses on the hippocampus CA1 tissue of ovariectomised female FLOX+GCI mice and FBN-ARO-KO+GCI mice. They found that the FBN-ARO-KO+GCI mice exhibited differences in key pathways regulating astrocytic reactivity (eg. RhoA signalling, actin-based motility by Rho signalling, signalling by Rho family GTPases and NRF2-mediated oxidative stress response). It was found that there was significant down regulation of top pan-reactive astrocyte transcripts and strong down regulation of most of the astrocyte A2 specific transcripts in the FBN-ARO-KO+GCI mice. In the case of astrocyte A1 specific transcription, several areas of transcription showed significant decrease in the FBN-ARO-KO+GCI mice whereas a few others showed significant increase. Using qRT-PCR it was possible to show that there was significant up regulation of genes involved in neuroinflammation and apoptosis, but significant down regulation of genes involved in regulating the astrocyte A2 phenotype and significant down regulation of genes involved in synapse maturation. From the combination of the RNA sequencing results, it was possible to conclude that A2 astrocytes were reduced in the FBN-ARO-GCI mice and neuron-derived E2 regulates transcription of genes and pathways involved in astrocyte activation and neuroprotection after global brain ischemia.

   The study on astrocyte phenotype continued with imaging experiments using the markers GFAP, vimentin and S100beta. Lu and team found that 7 days after GCI, FLOX+GCI mice demonstrated significantly increased GFAP, vimentin and S100beta compared to the FLOX+SHAM mice. This indicated a robust astrocyte activation following the GCI event. However, in the case of the FBN-ARO-KO+GCI mice, these astrocytes showed a pronounced decrease in GFAP, vimentin and S100beta compared to the FLOX+GCI astrocytes indicating attenuated astrocyte reactivity after loss of neuron-derived E2. The level of the astrocyte A1 phenotype was measured by using the three A1 selective markers C3D, FKBP5 and GBP2. No level of expression was found in the FLOX+GCI and FBN-ARO-KO+GCI showing that A1 phenotype was not induced in the CA1 with ischemic injury. In the case of the A2 phenotype, the markers S100A10, PTX3 and TGM1 were used. These were found not to be expressed in the FLOX-SHAM astrocytes and the FBN-ARO-KO+SHAM astrocytes, but all three were strongly induced in the FLOX+GCI astrocytes and only S100A10 and PTX3 were robustly down regulated (approx. 30% of FLOX control for S100A10 and approx 20% for PTX3) in the FBN-ARO-KO+GCI mice. This indicated to the authors that the FBN-ARO-KO+GCI mice have strongly induced A2 astrocyte polarisation compared to FLOX+GCI mice. These results were supported by the use of IHC analysis. C3D, the selected marker for A1 astrocytes was found not to be detected in any of the groups of ovariectomised female mice at 7 days after GCI whereas S100A10 (the marker for A2 astrocytes) showed strong induction and colocalization with GFAP in the FLOX+GCI hippocampal astrocytes. The FBN-ARO-KO+GCI exhibited significant decrease of S100A10 level in the hippocampal CA1 compared to the FLOX-GCI mice. Hence, this indicated to the authors that A2 astrocyte induction following GCI is suppressed in mice deficient in forebrain neuron-derived E2.

  The authors continued their study by looking at the expression of BDNF and IGF-1 in hippocampal CA1 tissue. Using both double staining and Western Blot analysis, the expression of BDNF was strongly increased in FLOX+GCI mice at 7 days after GCI (approx. 450%) compared to the FLOX-SHAM whereas in the less activated astrocytes of the FBN-ARO-KO+GCI mice the astrocytic BDNF levels were significantly lower (approx. 200%) compared to the FLOX+GCI mice. There was no difference between BDNF levels in the resting astrocytes of FLOX-SHAM and FBN-ARO-KO+SHAM. In the case of IGF-1, the astrocytic expression of IGF-1 in FLOX+GCI mice was strongly upregulated (approx. 450%) at 7 days after GCI compared to the FBN-ARO-KO+GCI mice (approx. 150%). Again, there was no difference between the FBN-ARO-KO+SHAM and FLOX+SHAM. The levels of GLT-1 and GFAP showed that they were robustly up-regulated in the FLOX-GCI reactive astrocytes (GLT-1 – approx. 450%) compared to the FLOX-SHAM whereas the level was markedly reduced in the FBN-ARO-KO+GCI mice (approx. 150%) compared to the FLOX+GCI mice. This confirmed to the authors that neuron-derived E2 is critical for hippocampal CA1 astrocyte activation and up-regulation of the neuroprotective astrocyte-derived neurotrophic factors and GLT-1 after GCI.

  Further studies were carried out to see whether time between GCI and sampling made a difference to the BDNF, IGF-1 and GLT-1 levels. Samples were taken at 3 days, 7 days (the normal sampling time) and 14 days and analysis showed that expression of all three in the FBN-ARO-KO+GCI mice was robustly decreased at both 3 day and 14 day time points compared to the FLOX+GCI mice (approx. 50% of control for BDNF, approx. 65% of IGF-1 and approx. 40% of GLT-1 at 3 days whereas approx. 70% for all at 14 days).  This indicated to the authors that the astrocyte dysfunction occurs at an early stage of GCI injury along with diminished astrocyte activation and may cause ischemic brain damage. Therefore, they examined the levels of F-Jade C and NeuN and found increased F-Jade C staining at 3 days in FBN-ARO-KO+GCI mice compared to FLOX+GCI mice and even higher values at 14 days. This confirmed the view that compromised astrocyte function after loss of neuron-derived E2 contributes to enhanced ischemic brain injury.

  Since most of the experiments had been performed with female mice, the authors repeated their studies with male mice in order to address whether neuron-derived E2 has similar functions in males. In the case of aromatase, this was found to be specifically localised in hippocampal CA1 neurons in male FLOX-SHAM mice and the level was markedly decreased in the FBN-ARO-KO+SHAM male mice. No aromatase was found in astrocytes from either group. GCI led to a strong induction of aromatase in the FLOX+GCI male mice and in the FBN-ARO-KO+GCI male mice it was also strongly decreased compared to the FLOX+GCI mice. Therefore, it was concluded that there is diminished astrocyte activation and aromatisation in male FBN-ARO-KO mice after GCI in the same way as the ovariectomised female mice.

  Similar findings for the male mice to the female mice were also found for the astrocyte phenotypes. No A1 astrocytes were induced in either the FLOX-GCI males or the FBN-ARO-KO+GCI males since there was no detectable C3D expression in either the FLOX-GCI or the FBN-ARO-KO+GCI hippocampus 7 days after GCI. In the case of A2 astrocytes, the A2 marker S100A10 was found to be markedly decreased in the FBN-ARO-KO+GCI male mice compared to the FLOX-GCI males. BDNF and GLT-1 were also found to be strongly reduced in the FBN-ARO-KO+GCI males compared to the FLOX-GCI mice whereas the level of neuronal damage shown by F-Jade C staining was shown to be significantly increased in the FBN-ARO-KO+GCI males compared to the control, FLOX-GCI mice.  Behavioural studies also showed that the male FBN-ARO-KO+GCI mice exhibited greater memory deficits (eg. increased escape latency, decreased exploring time) compared with the FLOX+GCI mice 7 days after the GCI administration. The level of GFAP expression and aromatase induction in astrocytes was found to be strongly decreased in FBN-ARO-KO+GCI male mice at both 3 days and 14 days after GCI indicative of diminished astrocyte activation and aromatisation. The levels of BDNF, IGF-1 and GLT-1 were also down-regulated in male FBN-ARO-KO+GCI mice at both 3 days and 14 days after GCI whereas F-Jade C intensity was enhanced showing neuronal damage was increased. Therefore, the authors concluded that there were no differences between male and female mice subjects.

   The study continued with an exploration of whether FGF2 signalling is a mechanism employed by neuron-derived E2 in the regulation of reactive astrocyte induction after GCI. Investigation of neuronal FGF2 signalling showed that the FGF2 transcript was strongly increased in the hippocampus CA1 region of the FBN-ARO-KO+GCI mice compared to the FLOX+GCI mice. This indicated to the authors a role of FGF2 in the mediation of suppression of reactive astrogliosis in FBN-ARO-KO mice following GCI. FGF2 levels were found to be increased in ovariectomised female FBN-ARO-KO+GCI mice compared to the other groups especially FLOX+GCI mice and levels in FLOX-SHAM were higher than FLOX-GCI. This supported the view that reduced neuronal FGF2 facilitates reactive astrogliosis.

  An investigation of FGF2´s major receptor FGFR3 in hippocampal astrocytes showed weak intensity of FGFR3 in the reactive astrocytes of the FLOX+GCI mice corresponding to the low levels of FGF2. However, the FGFR3 was strongly expressed in both SHAM groups and the less activated astrocytes of FBN-ARO-KO+GCI mice suggesting that increased FGF2 signalling in the FBN-ARO-KO+GCI might contribute to the lower astrogliosis levels in the FBN-ARO-KO+GCI mice following the GCI event. Blocking cerebral FGF2 signalling in ovariectomised female FBN-ARO-KO+GCI mice by bilaterally infusing simultaneously with FGFR3 antibody led to a restoration of the various factors to approximately FLOX levels, eg. GFAP levels elevated to approx. 92% showing that there was an almost complete rescue of reactive astrogliosis; approx. 95% induction of aromatase also occurred; pSTAT3 levels (marker for neuroprotective astrocyte phenotype) also showed increases to approx. 91%; BDNF and GLT-1 levels were also rescued; and Jade C intensity was found to be significantly attenuated indicating that decreased neuronal degeneration had occurred. These observations indicated to the authors that the neuroprotective astrocyte activation and functions were significantly restored after FGFR3 neuralisation in the FBN-ARO-KO+GCI mice. Therefore, neuron-derived E2 regulation of neuroprotective astrogliosis after GCI is due, at least in part to the suppression of neuronal FGF2 signalling.

  The authors continued to verify their findings by performing re-instatement of E2 levels by the administration of exogenous E2 in vivo to FBN-ARO-KO+GCI mice. The administration of E2 to the ovariectomised FBN-ARO-KO+GCI mice at a level known to fully restore E2 in FBN-ARO-KO mice showed that there was significant repression of the strongly increased neuronal FGF2 expression in the FBN-ARO-KO+GCI mice and that the FGFR3 levels were also now significantly repressed having been strongly increased in the FBN-ARO-KO+GCI mice. This confirmed the view that FGF2 signalling is inhibited by E2. Exogenous E2 replacement was also able to rescue the astrocyte derived BDNF level, IGF-1 level and GLT-1 level in the FBN-ARO-KO+GCI mice to almost FLOX+GCI levels.  It was also able to increase GFAP levels to 92% in FBN-ARO-KO+GCI mice compared to 65% in FLOX+GCI mice and robustly decrease the neuronal damage as shown by F-Jade C intensity measurements and neuronal structural changes as given by MAP2 staining.  Exogenous E2 replacement also significantly improved the performance of FBN-ARO-KO mice at the behavioural test of Barnes Maze test (ie. strongly decreased primary escape latency and exploring errors and increased quadrant occupancy). In FBN-ARO-KO+GCI mice the E2 levels which were not significantly different to those observed for FLOX+GCI mice led to improved performance at the Novel Object Recognition Test.  Exogenous E2 replacement strongly reversed the deficits as shown by increased exploring time of novel object and elevated discrimination index. Therefore, cognitive capability appeared to be restored following exogenous E2 administration to the FBN-ARO-KO mouse model where neuron-derived E2 is absent.

   The authors concluded their article by discussing their observations that neuron-derived E2 plays several key roles in the ischemic brain. The first role is that it is critical for astrocyte activation in the hippocampal CA1 region. Reactive astrocytes after ischemia become transiently hypertrophic and highly express the intermediate filaments, GFAP and vimentin. Mice deficient in GFAP and vimentin display attenuated astrocyte hypertrophy and reactive astrogliosis after brain injury. In this study FBN-ARO-KO mice displayed a significant decrease in astrocyte hypertropy and a significant decrease in GFAP and vimentin in the hippocampal CA1 region after GCI. There were also significant alterations in RhoA signalling in the FBN-ARO-KO mice which is shown to constrain actin motility and astrocyte reactivity and is tightly controlled by STAT3, a factor whose expression and activation are known to be critical for induction of the reactive astrocyte phenotype. The authors also explained their observed significant loss of elevated astrocyte aromatase and hippocampal E2 in the  FBN-ARO-KO+GCI mice model as being due to the decreased activation of astrocytes in this mouse species. Aromatase induction is known to occur only in activated astrocytes and not in the resting state and therefore, reinstating the astrocyte activation in the FBN-ARO-KO mice by blocking the FGF2 signalling rescued the aromatase levels in the hippocampal astrocytes after GCI treatment.

  The second role of the neuron-derived E2 was described by the authors as suppression of neuronal FGF2 signalling. The authors found that there was increased FGF2 signalling (increased mRNA levels and protein levels) in the FBN-ARO-KO mice after GCI. Neuron-derived FGF2 exerts an inhibitory effect on astrocyte activation by suppressing GFAP expression in astrocytes. Therefore, blocking the FGF2 receptor in astrocytes invokes reactive astrogliosis basally and strongly enhances it after injury. This indicated to the authors that FGF2 is an important signal to restrain induction of reactive astrocytes. In the FBN-ARO-KO mouse model, FGF2 signalling played an important role in the decreased astrocyte activation. Blocking the FGF2 signalling by administering its receptor`s inhibitor, a FGFR3 neutralising antibody, fully restored astrocyte activation after GCI. This is also supported by the observations that neuron-derived E2 also enhanced the expression of astrocyte derived neuroprotective neurotrophic factors, BDNF and IGF-1 as well as the glutamate transporter, GLT-1.

  The third role of the neuron-derived E2 described by the authors was that it exerts neuroprotection and preserves hippocampus dependent cognitive functions following ischemic injury. The FBN-ARO-KO mice used in the study meant that the lack of aromatase resulted in E2 depletion in neurons causing the cognitive deficits observed after GCI. This was supported by the exogenous E2 replacement being able to rescue these deficits. The FBN-ARO-KO mice exhibited more serious neuronal damage in the hippocampus following GCI manifesting as greater cognitive function deficits. They also exhibited attenuated reactive astrogliosis after GCI which suggested that the neuroprotective effect of neuron-derived E2 could be mediated by reactive astrocytes after GCI. This was supported by the observation that the FBN-ARO-KO+GCI model produced only mild injury and no scar formation which is indicative of a beneficial effect of reactive astrocytes rather than the detrimental situation of where reactive astrocytes cause proliferation and scar formation leading to the inhibition of axonal regeneration after injury.

   The FBN-ARO-KO+GCI model also showed strong attenuation of GFAP and vimentin and reduced astrocytic GLT-1 levels in addition to the increased neuronal damage and attenuated reactive astrogliosis. This was supported by the evidence that reactive astrocytes can protect neurons through several mechanisms including the release of the neuroprotective factors such as BDNF, IGF-1 as well as increased uptake of excess glutamate. Most of the glutamate uptake (90%) in the brain is carried out by the astrocyte specific Na+ dependent glutamate transporter GLT-1. Up-regulation of GLT-1 in astrocytes is reported as protecting hippocampal CA1 neurons after GCI since glutamate excitotoxicity is known to be a major mechanism of neuronal damage and death after ischemia. Therefore, it was suggested that up regulation of GLT-1 in astrocytes as well as up regulation of BDNF and IGF-1 could be mechanisms by which neuron-derived E2 could cause neuroprotection after GCI. The authors suggested that the attenuated astrocyte activation and reduced aromatase/E2 induction in astrocytes after GCI could contribute to the enhanced neuronal damage seen in the FBN-ARO-KO mouse model used.

   The authors also considered whether E2 itself provides a neuroprotective effect since E2 added exogenously in vitro can show a positive protective effect. However, the effect appeared to be instigated via the reactive astrocytes since aromatase and E2 increases in reactive astrocytes after GCI. The reinstatement of astrocyte activation in the FBN-ARO-KO mice after GCI by blocking the FGF2 signalling rescued the attenuated aromatase levels in the astrocytes as well as reinstating the pSTAT3, BDNF and GLT-1 levels. Neuronal damage was also strongly reduced.

  Therefore, the authors in their study showed that neuron-derived E2 is likely to play a beneficial role in astrocyte activation in the CA1 area of the mouse hippocampus following global ischemia and it does this by suppressing FGF2 signalling. This leads to neuroprotection and improved behavioural cognitive function following ischemic injury in the brain.

COMMENT

    What makes this article interesting is that it describes the roles of estradiol (E2) as a modulatory influence on hippocampal firing during normal functioning and as an influence in response to damage of this area caused by an ischemic event. Therefore, this comment focuses on two main areas: the role of E2 in hippocampus CA1 functioning relating to memory capability; and the role of E2 and the astrocytic response to ischemic assault of this area.

   With regards to E2 and hippocampus firing, the E2 is probably under normal circumstances in females sourced from both local neurons and the ovaries under hypothalamus control. It is assumed that independent of source the neurochemical action of E2 comes from binding of the ligand to the cell membrane receptors, ERs (both alpha and beta varieties) or GPR30 receptors with all types of receptors having the same affinity for estradiol. This binding is followed by inclusion and binding to the nuclear DNA binding sites (either the unique DNA sequences, estrogen response elements ERE or interaction with other DNA bound transcription factors such as AP-1). The response to the DNA binding in the case of the hippocampus is the excitatory effect on the overall firing of the cell as well as synaptic modulation which supports and aids that firing excitability. Estradiol binding (as well as other estrogens) can alter the excitability of neurons with latencies of a few seconds independent of whether ER or GPR30 binding and this can stimulate Akt, MAPK and cAMP production within minutes. Estradiol binding to GPR30 receptors in particular can stimulate not only the production of cAMP, but also leads to mobilization of calcium ions as well as activating the growth factor signalling pathways.

  This leads onto the second function of E2 in the hippocampal area which is the positive effects on synaptic and dendritic density via advantageous growth of neurites and dendritic spine density. Estrogens can also promote the growth of more excitatory synapses. The new spines have been shown to have more NMDA Rs which can increase the long-term plasticity of the hippocampus which would have an overall beneficial effect on the firing of the area manifesting in positive cognitive behaviour performances, eg. spatial memory. The LTP role of E2 in the hippocampus appears to be via cholinergic input into this area and increased NMDA neurotransmission mediated in particular by NR2B containing receptors (Smith). It also has a direct effect on excitation by depressing inherent synaptic inhibition. The binding of E2 to ERs on the inhibitory interneurons leads to the inhibition of spine growth and therefore, the E2 causes these inhibitory cells to produce less GABA resulting in decreased inhibition and by default greater neural activity which somehow triggers an increase in spines and excitatory synapses on the excitatory pyramidal cells. Therefore, the overall effect of E2 in the hippocampus is a general aid to the excitation of firing via neuronal changes and advantageous modulation of the synapse which supports the general firing excitability in the area.

    (Before we continue, we should mention the male equivalent to E2, the androgens. In fact, E2 is produced from testosterone and so it should be assumed that there is an equivalent aid to neuronal firing and synaptic modulation of the hippocampus in the male. This is found to be correct with the androgens being seen to act on brain cells to modify their functions and ultimately, behaviour. In the case of androgens there are also three categories of receptors. One receptor type preferentially binds testosterone and a second one preferentially binds dehydroxytestosterone (DHT). Both of these receptors are in equilibrium between the nucleus and cytoplasm and can be activated and transformed by the binding of the androgen steroid leading to concentration of the complex just like E2 in the nucleus. The third receptor type however, can bind both steroids with the same relative affinity, but can only be activated and is not transformed by the DNA binding and hence, does not concentrate in the nucleus.  The androgens are reported to have profound effects on hippocampal structure and function and like estrogens their effects include the induction of spines and synapses on the dendritic spines of the CA1 pyramidal neurons, as well as alterations in LTP resulting in effects on hippocampally dependent cognitive behaviours (Atwi). However, the method for this is different to the estrogens. The effects of androgens appear to be carried out via modulation of brain-derived neurotrophic factor (BDNF) in the mossy fibre (MF) system of the hippocampus which is suppressed in the presence of testosterone. This is supported by observations that orchidectomy of male rats increases synaptic transmission and excitability in the MF pathway with exogenous testosterone reversing these effects. This suggests that testosterone exerts a tonic suppression on MF BDNF levels and that loss of testosterone as in age for example leads to an increase in BDNF dependent MF plasticity. Therefore, unlike E2 androgens appear not to affect the firing of the neurons, but more the synaptic modulation. From a behavioural perspective, definitive findings are problematic since some researchers report no change in cognitive capability when androgens are deficient and others like in the case of estradiol report loss of verbal and spatial memory performance.)

  However, in this comment in response to Lu and team`s article we are concentrating on the effects of E2 on the mouse hippocampal CA1 region and the results of what happens when E2 is deficient in particular. Therefore, as expected using the aromatase knock-out mouse model, FBN-ARO-KO, the level of E2 in the hippocampal CA1 area of these females was lower than in the FLOX mouse control. Ovariectomy of the FBN-ARO-KO female meant that the ovarian contribution of E2 via the hypothalamus was also missing and therefore, the level of E2 in these mice was only 33% of the control FLOX. The physical consequences of the deficiency as reported by Lu and team were as expected in that the structural integrity of the CA1 neurons was lower in the ovariectomised FBN-ARO-KO mice than the control. This manifested as expected as deficits in cognitive functional capability with spatial memory reported as being lower in the ovariectomised female FBN-ARO-KO mice than the control.

   Several things should be pointed out at this stage relating to the findings. The first is there is little or no E2 in glials under normal circumstances. This implies that the reduction in structural integrity observed with E2 deficiency in the ovariectomised FBN-ARO-KO mouse model is not seen by the neuronal cells as an ´assault` on its functionality and therefore, a response of increasing glial functionality is not instigated. This may mean that there is redundancy in the neuronal system under normal circumstances. The second point is that an increase in FGF2 level is observed in the ovariectomised FBN-ARO-KO mouse compared to the FLOX control. Since applied FGF2 to rat cortical neurons is reported to enhance neurite growth, axonal branching and LTP in hippocampal neurons it is possible that the E2 deficiency may be compensated for under normal circumstances to some extent by the increase in FGF2 signalling. E2 loss of structural integrity may then relate to factors other than neurite growth, eg. microfilament organisation. The third point relates to the natural human situations of estrogen and androgen deficiencies as seen in menopause and male ageing for example. The physical and behavioural effects observed in these situations are not always observed and therefore, may reflect the physiological complexity of the human brain in relation to the mouse brain. One factor for the differences may be the dominance of the mouse behaviour on spatial information input, processing and memory compared to the human`s preference for visual information. Therefore, loss of estrogens and androgens in humans may indicate the same types of changes, but not to the same extent and direct transfer of knowledge of one system to another may not be fully valid.

   Now we have established the effects of E2 on excitatory firing and synaptic modulation we can look at how E2 in the hippocampal CA1 can affect the astrocytic response to an ischemic assault of that area. Lu and team`s study found that in mice E2 deficiency produced no expected astrocyte increase in response to this type of damage. Ischemia is a condition in which the blood flow and hence, oxygen is reduced so that metabolic demands cannot be met. With regards to neurons, the interruption of blood flow for as little as 20 seconds can result in local cortical activity ceasing with this cessation spreading outwards from the point of assault (spreading depression). Neurons which do not receive enough blood to communicate result in an area called a penumbra, but they do receive sufficient levels of oxygen to avoid cell death at this time and ischemia induced in brains for up to an hour may be at least partially recovered.

   The shut-down of firing observed with ischemia is induced by the efflux of potassium ions from the neurons via initially the opening of voltage dependent potassium channels and later by ATP dependent potassium channels leading to a transient plasma membrane hyperpolarisation. This cascades into a redistribution of ions across the plasma membrane (influx of sodium, calcium and chloride ions) which results in the excessive release of neurotransmitters. In particular, the dramatic release of glutamate is the major cause of destruction elicited by the ischemic event at the synapse. In this glutamate excitatotoxicity, the glutamate receptors are overstimulated leading to sodium ion influx and potassium ion efflux via the glutamate receptor-activated membrane channels and calcium influx via the NMDA receptor–gated ion channels. This adds to the firing challenges faced by the cells with further spread of the neuronal depolarisation as well as depletion of energy stores and initiation of calcium ion dependent detrimental cascades, eg. release of zinc ions. Further responses to the ischemic assault at the intracellular level involve enzyme activation, eg. protein kinase rapid activation that can lead to enhancement of the neuronal excitotoxicity by increasing the release of glutamate, or DNA transcription and gene expression changes, eg. MAP kinases and p38 kinases activation. The outcome is enhanced neuronal damage and cell death.

  Tissue damage as a result of ischemic assault is likely to be mediated via the formation of several reactive oxygen species (ROS) or by the activation of catabolic enzymes. Both of these stem from the increased intracellular calcium concentration. Research shows that nitric oxide radicals can cause DNA single strand breaks and when produced by inducible NO synthase expressed in macrophages neutrophils and microglia may contribute to late tissue damage whereas in other cases NO may be part of the anti-inflammatory response. Activation of calcium-dependent catabolic enzymes such as phospholipase A2 and C following NMDA R stimulation promote membrane phospholipid breakdown thus ameliorating the formation of free radicals and inflammation processes. Other calcium activated proteases or calpains contribute to the destruction of the structural and regulatory proteins adding to the cellular damage.

   Tissue damage by the ischemia also induces an inflammatory response with the elevation of mRNA levels for the pro-inflammatory cytokines TNFalpha and IL-1beta observed as early as I hour after the GCI. There is also an induction of adhesion molecules on the endothelial cell surface (eg. CAM-1, P- and E-selectins) which enhance the adhesion of neutrophils and passage through the vascular wall into the brain parenchyma leading to the invasion of macrophages and monocytes. Detrimental damage can be visualised by some cellular swelling and at some level apoptosis is instigated with the translocation of cytochrome c from the mitochondria to cytosol and the activation of caspase 3 and the apoptotic pathway.

   The responses to ischemia are not all negative though with the cell instigating processes to protect itself from the assault and to remove cell debris which is itself destructive. One response is to enhance vascular blood flow in order to limit the reduction in blood supply and oxygen caused by the ischemia and one method for this NO-mediated vasodilation. This is where one form of NO synthase present in endothelial cells leads to the relaxation of vascular smooth muscle cells and helps to preserve blood flow. At the cellular level, there are neuroprotective mechanisms activated in order to reduce the excessive excitation. For example, the activation of interneurons leading to release of the inhibitory neurotransmitter GABA; the down regulation of NMDA receptor function by blocking the zinc binding site; and the depletion of extracellular calcium and sodium ions in order to reduce the membrane concentration gradient which favours influx. The glutamate excitotoxicity is also counteracted by the action of several other neurotransmitters that are in ischemia neuroprotective, eg. serotonin, GABA and adenosine. In the case of adenosine, this accumulates rapidly in ischemia due to the breakdown of ATP. The beneficial effects come from the ability of the neuronal adenosine A1 receptor to reduce neurotransmitter release membrane excitability and from the A2 receptors where activation on vascular smooth muscle cells enhances blood flow and activation on neutrophils decreases inflammation.

   Lu and team observed massive upregulation of the growth factors BDNF, IGF-1 and the glutamate transport molecule, GLT-1 after GCI and these too are part of the neuroprotective response to ischemia. Others include (in rat) nerve growth factor (NGF), neurotrophins 4 and 5 (NT-4/5) and basic fibroblast growth factor (FGF2). The aim of the up regulation of these factors is to minimise damage, block apoptosis and even enhance functional recovery (eg. via enhancing nerve fibre sprouting and synapse formation). However, the situation is not clear since some, eg. BDNF, NT-3, NT4/5 are also reported as enhancing the neuronal vulnerability to excitotoxic and cell death induced by the free radicals. This negative effect is explained by some as being caused by the enhanced NMDAR mediated calcium influx, increased production of free radicals and even by acute pro-excitatory effects that could increase the ischemia induced excitotoxicity.

   However, the main neuroprotective mechanism to ischemia and with relation to the comment on this article, is the role of astrocytes and the formation of reactive astrocytes. Astrocytes appear to possess a high potential for regeneration and neuroprotection following ischemia, but it should also be noted that, just like the neurotrophic factors, there is also evidence that astrocytes can exacerbate ischemic injury. This is because the excessive accumulation of ions and glutamate can overload the buffering role of the astrocytes leading to the activation of the catabolic enzymes and eventually cell death.

  In Lu and team`s article, the response of the CA1 to the ischemic injury is the massive up regulation of astrocytes and their functional capability. Astrocytes are highly responsive and dramatically change their characteristics to the damage and therefore, these ´injury-activated` cells are termed reactive astrocytes. There are two subtypes, A1 and A2, depending on the assault faced. For example, A1 reactive astrocytes are induced by inflammatory agents such as lipopolysaccharide and are proinflammatory and neurotoxic; whereas A2 are induced following cerebral ischemia and are neuroprotective. (In Lu and team`s study it is the A2 subtype which appear to be dominant.)

  Ischemia leads to the up regulation of astrocytes and reactive protoplasmic astrocytes in the cortex began to proliferate (reactive astrogliosis) within 3 to 5 days after injury. This is mirrored by the increase in glial fibrillary acidic protein (GFAP) observed and whereas initial morphology shows a clear cytoplasm with many mitochondria and some processes, after 3 days there are many phagocytic inclusions visible within the cytoplasm. From a biochemical point of view, there are two advantages to astrocytes, one is its energy supply and the second is its ability to up regulate specific proteins beneficial to the synapse. In the case of energy supply, astrocytes can respond to an oxygen deficient situation by changing their energy supply in comparison to neurons. The inhibition of mitochondrial respiration induced by NO leads to astrocytes increasing their glucose metabolism via the glycolytic pathway which limits the fall in ATP levels and prevents apoptosis. In contrast, neurons cannot and a large ATP depletion results which can lead to apoptosis.

   The other biochemical response of the astrocytes to ischemia is the change in gene expression and specific protein levels. The intermediate filament proteins such as vimentin (part of the Lu study) and nestin, synemin are involved in the reactivity of astrocytes and are highly expressed in ischemic injury as well as in response to oxidative stress. Another factor up regulated under these conditions is cyclin-dependent kinase 5 (CDK5) of astrocytes which is involved in a number of beneficial processes such as synthesis of neurites, synapse formation and synaptic transmission and axonal cytoskeleton (assembly, organization and stabilization with substrates mainly neurofilaments and MAPs) and apoptosis. It is also involved in the elongation and reactivity of astrocytes themselves. GFAP, CDK5 and p35 (activated when the astrocyte is subjected to stress and generating active CDK5) forms immunocomplexes. The biochemical changes elicited by the ischemic injury allow the astrocytes to carry out their various neuroprotective functions.

   One such function is to reduce the level of extracellular glutamate increased by the excessive excitation of the neurons and released into the extracellular space and also by the microglials. This function is achieved by the astrocytes expressing a much larger amount of membrane localized EAATs excitatory amino acid transporters (EAATs) of which GLT-1 is one so that there is increased uptake of the glutamate. Lu and team saw in their experiment that there were much higher levels of this transporter in their experimental mouse models when subjected to GCI.

   Another function of the reactive astrocytes following ischemia is in aiding the defence of the neurons against oxidative stress. There are three ways in which this is carried out: indirectly via the  high abundancy of NADPH in astrocytes with NADPH being used as an electron donor for the regeneration of reduced glutathione, the ROS scavenger; directly via the astrocytes releasing in response to excess glutaminergic activity ascorbic acid which is an antioxidant (The ascorbic acid released is taken up by the neurons and modifies the local energy metabolism by inhibition of glucose consumption and increasing the uptake of lactate.); and thirdly, by activating the expression in the astrocytes of Nrf2 which is a redox-sensitive transcription factor which acts as in neuroprotection (This coordinates the expression of the cytoprotective enzymes that can also scavenge any ROS produced.).

   However, a major function of reactive astrocytes in the response to ischemic injury is the removal from the local environment of debris caused by the oxygen and blood flow depletion. Reactive astrocytes are capable of phagocytizing multiple cellular components because they are observed to engulf myelin-like structures, synaptophysin1plus synapses and other unidentified components. Although traditionally linked to A1 subtype, A2 astrocytes are also believed to demonstrate the same phagocytic function in response to ischemic injury as the A1 and they assist microglia in the removal of debris even if at different spatial temporal characteristics. These phagocytic reactive astrocytes found in the penumbra region are enriched with genes involved in engulfment pathways, which includes phagocytic receptors and intracellular molecules. Phagocytic astrocytes demonstrate upregulated ABCA1 and its pathway molecules, MEGF10 and GULP1, which are required for phagocytosis. The reactive astrocytes also express several other phagocytic receptors, including BAI1 and integrin αvβ3 or 5, which appear to function as upstream signals of another beneficial pathway. The onset of the astrocytic phagocytosis begins after 3 days (reactive astrocytes exhibit many phagocytic inclusions within their cytoplasm at this time) and persists for 14 days. The spatiotemporal pattern of the phagocytosis suggests a relationship to neuronal remodelling in the penumbra region of the ischemic injury with substantial axonal, dendritic and synaptic losses and debris removal within the first week following the ischemic injury followed by an increased number of synapses and axonal connections as part of the remodelling.

   Reactive astrocytes also play a role in the remodelling of the neuronal area following the damage and therefore, the phagocytic function of astrocytes may be seen positively since it removes the debris so that renewal of the synapses can take place. For the remodelling function, the reactive astrocytes release neurotrophic factors and synaptogenic factors which promote neuronal survival, synapse formation and plasticity. For example, astrocytes are known to release thrombospondin-1, which is a major regulator of synaptic maturation and tissue plasminogen activator which may be required for recovering neurons to remodel their dendritic arbors. Reactive astrocytes are also linked to the development of new neurons. One mechanism identified for this following ischemic injury is that specific astrocytes (possibly NPC-astroglial cells and Olig2PC-astroglial cells) can acquire stem cell traits. This is observed by the induction of neurogenesis from a GFAP-expressing progenitor cell in the SVZ. These give rise to a reactive astrocytes subpopulation in the cortex that contribute to astrogliosis and scar formation. These astrocytes in the SVZ can also be converted to neurons in vivo by forced expression of Ascl1 and hence, lead to the formation of and migration of newly born neurons into a unique neurovascular niche. This is of particular importance for the hippocampal area and cognitive functioning since the area relies on newly formed neurons for memory capability. It is also found that in the hippocampus, reactive astrocytes preserve the function of the hippocampal neural niches which play a part in adult neurogenesis. It is found that astrocytes can allow the synaptic integration of adult-born hippocampal neurons, allowing local dendritic spine maturation and NMDAR functional integration. Astrocytes also secrete certain proteins, eg. beta-arrestin-1 in the dentate gyrus area of the hippocampus which aids in neurogenesis via expansion of the neural precursors in this region.

   Part of the astrocyte response to ischemic injury and characteristic of it is glial scar formation which can be both positive and negative. On the positive side glial scar formation in areas surrounding severe ischemic lesions is characterized by astrocyte proliferation and a considerable extension of astrocytic processes beyond the previous domains of individual astrocytes. Upregulation of GFAP and other genes and pronounced hypertrophy of cell bodies and processes and interaction with other types of glial cells all occur.  However, on the negative side, glial scar formation has been considered an inhibitor of axon regeneration and a factor that can cause neurotoxicity, inflammation, or chronic pain and this is supported by the observed increases in inflammatory factors such as interleukins 1β, 6, 10, IFN, TGFbeta as well as inflammatory molecules such as ROS, NO, glutamate and calcium-binding protein B.  The immune response instigated is indicated by the expression of neurotrophic factors and by the expression of class II major histocompatibility complex (MHC) molecules which play a critical role in the induction of immune responses through the presentation of processed antigens to CD41 T-helper cells. This molecule is normally expressed on the known antigen presenting cells (APCs), such as B cells, macrophages, dendritic cells, and other cell types which includes the more unusual astrocytes. Reactive astrocytes also release chemokines after ischemia. In this case, chemokines in vascular endothelial cells increase adhesion molecules levels, attracting immune cells. The reactive astrocytes also are capable of expressing pattern-recognition receptors (PRRs), such as TLRs, scavenger receptors, and complement proteins. They are also resistant to apoptosis induced after inflammation by activation of the death receptors (eg. apoptosis antigen 1 and tumor necrosis factor (TNF)-related apoptosis-inducing ligand (FAS, TRAIL). Therefore, the cells themselves are capable of surviving inflammatory insults as well as playing a complex role in the local regulation of immune reactivity.

  The final function of reactive astrocytes and important because of the type of damage occurring in ischemic injury is that they can play a role in the restoration of the vascular system. This is important since ischemia causes disruption and damage to blood flow and oxygen supply. Under normal circumstances, astrocytes have very close interactions via membrane proteins such as ion channels and growth factor and cytokine receptors. They and endothelial cells can also release neurotrophins, vascular growth factors, glucose and amino acids in order to generate stability and maintenance of the blood brain barrier for example. In ischemic injury the signalling between the astrocytes, pericytes and endothelium becomes disrupted and hence, repair to the gliovascular system is required. Astrocytes following ischemic injury modulate the cytotoxic response and induces the necessary angiogenesis in order to re-establish blood flow. This astrocytic mechanism requires secretion of a number of factors including TGFbeta, glial-derived neurotrophic factor (GDNF), FGF2 and angiopoietin 1 (Ang1). These factors can stimulate the production of new blood vessels and the proliferation of endothelial progenitor cells. Therefore, another function of reactive astrocytes is the re-establishment of blood flow.

  With reference to Lu and team`s article, their research supports the other observations reported that GCI leads to the activation of astrocytes in their control FLOX mouse model (200% max, as shown by GFAP/vimentin) with no change in astrocyte number, but an increase in cell body volume which could indicate phagocytic functionality. The astrocytic response showed the expected temporal characteristics following the GCI administration with reports at 3 days, the best response at 7 days and possible recovery at 14 days. However, their research looked at the question as to the role of E2 in response to ischemia particularly in relation to astrocytes. The role of E2 was deduced by using a mouse model where the neuronal derived contribution was absent (FBN-ARO-KO mouse) and even the ovarian contribution was removed (ovariectomised FBN-ARO-KO mice). 

  So, what did Lu and team find and how can we interpret their findings? They found that under normal circumstances the control mouse (FLOX) on GCI exhibited strong activation of astrocytes (A2 variety) and a slight increase in E2 level (FLOX to FLOX+GCI –  100 to 120%). However, in the ovariectomised FBN-ARO-KO mouse with GCI, there was little astrocyte activation (approx. 20% max) although what was observed was time sensitive even if low and the temporal changes matched that of the FLOX controls. Again, there was no change in astrocyte number, but there was also no change in astrocyte volume implying non-involvement of astrocyte reactivity in the FBN-ARO-KO response to GCI. There was also no change in E2 level (FBN to FBN+GCI – no change – 33% of FLOX level). Therefore, from the aspect of E2 level, in normal mice the GCI stimulates E2 production since aromatase is produced in the reactive astrocytes so E2 is present in the astrocytes following ischemic assault (aromatase production increases time sensitive 7 days 250%, 14 days 70). Therefore, aromatase appears to be produced in response to this type of adverse condition. However, in the case of the FBN-ARO-KO mice, no increase in E2 is seen so in the FBN mice no stimulation is observed with GCI or if stimulated then no production is possible because of the lack of the aromatase enzyme. The increase in ER in response to ischemia is also likely to be lacking in the FBN-ARO-KO mice model.

   From the astrocyte aspect, the control FLOX mouse model produced the expected proliferation and reactivity astrocytic response identified for ischemic injury (200% rise and predominantly A2 subtype). It also appeared time sensitive as described above (time sensitive aromatase production increase 7 days 250%, 14 days 70; astrocyte reactivity – best at 7 days). However, there was little astrocyte reactivity in the ovariectomised FBN-ARO-KO mice which implies that the absence of E2 suppresses the induction of the astrocyte response following ischemic injury in the mouse model.

   How does the absence of E2 do this in the mouse model? The first explanation is that the removal of the excitotoxic glutamate released as a result of the ischemic injury is lower in the FBN-ARO-KO model and hence, there is more damage. Lu saw in the experiments that the level of glutamate transporter GLT-1 was lower in the ovariectomised FBN-ARO-KO mice than the control FLOX mice following GCI. This would result in higher levels of glutamate and hence, increased neuroinflammation and apoptosis through other non-E2 controlled means resulting in higher levels of damage seen. The astrocyte proliferation and shift to reactivity occurs even in the FBN-ARO-KO mouse model, but like most biochemical processes it acts as an interacting cascade and is self-proliferating and therefore, one process not activated at an early stage can have major consequences for the initiation of dependent processes both simultaneously and later on.

   The second explanation also relates to damage instigated through the ischemic assault. Structural integrity is already lower in the FBN-ARO-KO mouse model since E2 is deficient which would normally aid in the mouse model in the synaptic modelling and regeneration processes. This is important in the interpretation of Lu`s experiments because of the dynamic nature of the hippocampus CA1 area corresponding to its memory and information gathering functions. Therefore, lower structural integrity would imply a greater propensity to damage and cell death during destructive assaults and decreased capability for regeneration and renewal not only when under assault, but also under normal circumstances. Without the reactive astrocytes removing the cell debris and instigating remodelling processes following the GCI, the FBN-ARO-KO mice in particular are more likely to exhibit damage which is in itself, a signal for higher levels of apoptosis.

   The third explanation again links back to the cascade like processes that occur in the neurons, glials and synaptic spaces. Neuronal and glial cells function is dependent on neurotrophic factor actions. Two in particular, IGF-1 and BDNF, are cited in the role of E2 and its aid to reactive astrocyte functioning in the areas of neuronal synapse growth and renewal following ischemic injury. Therefore, lack of E2 as in the case of the FBN-ARO-KO mice means that these important neurotrophic factors are lacking and hence, their functions are not elicited. In the case of IGF-1, this acts in the early post-ischemic period to decrease apoptosis cell death and to promote vascular remodelling. When E2 is present, the IGF-1 and ERalpha bound with E2 forms a complex which interacts at the DNA sites to promote ERK/MAPK and CREB signalling aiding neuronal survival. Ischemia promotes rapid dephosphorylation and inactivation of the ERK/MAPK and CREB site and therefore, E2 when present performs a ´protective` effect on this particular site. When E2 is absent this protective function is not possible and therefore, ischemia can lead to the switch off of the ERK/MAPK and CREB sites, hence reducing neuronal survival. Although pre-treatment with E2 in situations where E2 is absent can lead to a reversal of the ischemic injury responses, a blockade of the ERK/MAPK site does not prevent the ischemic down regulation of Bcl-2, one of the anti-apoptotic proteins. Bcl-2 is a target of CREB and hence, a target of the ERalpha-E2 complex and would be expected to be at normal levels following E2 pre-treatment. However, its continual down regulation in the presence of E2 means that another factor or factors must also control Bcl-2 levels in response to ischemia. This is highly likely and indicates the complexity of the neurotrophic factor functioning. With regards to Lu and team`s experiments, they found that IGF-1 greatly increased in the control mouse model following GCI compared to the FBN-ARO-KO model where there was a smaller increase. Therefore, the response to ischemia in the presence of E2 was as expected with the IGF-1 performing a neuroprotective partnership and IGF-1 being induced probably due to the reactive astrocyte induction. In the absence of E2 only a slight rise in IGF-1 was observed, probably attributed to the induction by the lower level of reactive astrocytes present in the FBN-ARO-KO model. 

   BDNF followed the same pattern as IGF-1 in Lu and team`s experiments with a massive upregulation in BDNF in the control FLOX model and a smaller increase in the E2 deficient FBN-ARO-KO model. This again, can probably be attributed to induction of BDNF by the reactive astrocytes and represents the level of reactive astrocytes present not only in the control, but also the FBN-ARO-KO mice. Therefore, the results confirm that the lack of astrocyte reactivity in response to the glutamate increase and ROS from the neural damage caused by the ischemic assault means that not only was the astrocyte dependent debris removal and remodelling processes not put into play, but the synaptic remodelling aided by the E2 processes was also not available.

   However, Lu and team`s experiments did show a different effect with the neurotrophic factor, FGF2 and indicated a relationship between E2 and reactive astrocyte response to GCI assault. Lu and team saw in their experiments that in the absence of E2 there was a large increase in FGF2 levels in the ovariectomised FBN-ARO-KO model after GCI (220%) than with the control model, FLOX+GSI+ova (60) which decreases or stays relatively the same. The results were mirrored by the experiments looking at the ligand binding FGF3 receptors (FGF3Rs) which also confirmed that endogenous E2 would remedy the situation (ie. levels would decrease in the FBN-ARO-KO mouse model). (For the sake of the debate we have to assume that FGF2 level reflects overall FGF levels. This is because FGF signalling is interchangeable and that factors FGF1 to 4 can bind to more than one receptor. Therefore just because FGF2 signalling is decreased or increased does not mean that overall function of the FGF neurotrophic factors is increased or decreased.) One explanation for this rise is that it is a result of the increase in damage observed which is worsened by the lack of E2 rather than a direct effect of E2 itself. FGF2 signalling is normally suppressed by E2 in the mouse model or alternatively, in the presence of E2 there is less damage to the CA1 after ischemic assault. This is because FGF2 is reported as being secreted by neurons that are damaged by glutamate as in the case of glutamate excitotoxicity in ischemic injury (Noda). In this case, FGF2 enhances the microglial migration and phagocytosis of the cellular debris and is neuroprotective via the FGF3R extracellular signal regulated kinase ERK pathway which is directly controlled by the Wnt signalling in the microglia. Therefore, the FGF2 effect in ischemia is via microglial activation and therefore, in the case of the control mice FGF2 increases may not be so high where there is a high level of astrocyte functioning for the debris and re-modelling situation as in the FBN-ARO-KO mice where the level of reactive astrocytes is low.

  However, FGF2 signalling is linked to astrocytes directly and therefore, the low level activation of astrocytes in the ovariectomised FBN-ARO-KO mouse model following GCI means that this contribution is reduced. This probably occurs by reduced levels of appropriate migration of newly formed neurons. This view comes from the observation that in the subventricular zone (SVZ) of the hippocampus DG region, FGF2 is highly expressed normally by GFAP-positive cells (ie. glials and astrocytes) which suggests a pro-proliferation role for the astrocytes in this area. It is found that acute stress leads to enhanced neurogenesis in the SGZ and astrocytic FGF2 plays a role. FGF2 deficient mice however, show normal neuronal progenitor proliferation during development, but the progenitor cells fail to colonise their target layers in the cortex. This implies that FGF2 controls fate, migration and development of neuronal progenitor cells rather than proliferation during development. In addition, FGF2 is believed to modulate synaptic plasticity (enhances neurite growth, promotes LTP) and axonal branching in the hippocampus. Therefore, in situations where remodelling is required and total astrocytic guidance of this is not possible, FGF2 signalling would compensate.

   Therefore, to conclude it has been shown ischemic assault in the given mouse models has a detrimental physiological effect that manifests as structural and functional changes of the hippocampal CA1 area. These physiological changes lead to cognitive functional impairments. In response, there is reactive astrocyte proliferation which carries out a number of neuroprotective processes including neurite growth and synaptic remodelling in order to promote neuronal survival and restore cognitive behaviour. When this is not possible to the degree required, then neuronal degeneration and apoptosis follows. The reactive astrocytes are aided in their neuroprotective effects by the action of E2. The E2 help comes in the form of specific structural synaptic changes in the mouse models indicated, eg. synaptic growth and spine density. The suggested link between absence of E2 and superior FGF2 signalling as suggested by Lu and team may be only indirect. It is likely that the high levels of FGF2 observed come from the increased microglial activation observed due to the lower structural integrity of the neurons and synaptic area not only in the control mouse model, but also increased after the damage caused by the ischemic injury.

   For this reason, two things should be noted. The first is that it cannot be assumed that there is a relationship between two factors in one condition just because both things change when that condition changes and secondly, that in the case of E2 (and probably other hormones) the situation in mouse models may not directly be translated to the situation in humans. This refers to the lack of E2 and estrogens reported in human female menopause having reported effects on physiology and cognitive functioning compared to the same lack in the mouse model. For example, lots of research in humans relates to E2 loss and pre-treatment of E2 before ischemic events, but there are disputed effects. Estradiol administered at levels used for hormone therapy in postmenopausal women is reported as intervening in apoptotic cascades via increased caspase 3 activity in male gerbils which leads to strong protection against the ischemic injury induced cell death in the CA1. However, not all human studies show the same cognitive advantage when missing hormones are replaced. Therefore, mouse models can provide an idea of what may happen, but the demands on brain functioning in mice are different to humans, eg. spatial memory is strongly required in mice whereas the dominant sensory system in humans is visual. Transfer of knowledge about structure and function from one species to another should always be carefully scrutinised for validity.

Since we`re talking about the topic………………………………………………

                …..it has been shown that the hippocampal synapses of aged female rats respond differently to estradiol than the synapses of younger mice and that there is an age-related decline in IGF-1. Can we assume that if the above experiments were repeated with aged mice, the FLOX mice would show similar results to the FBN-ARO-KO mice since estradiol levels would be naturally reduced?

                ….it has been shown that ischemic tolerance develops in heart tissue and in the gerbil brain when the subject is subjected to a series of brief sublethal ischemic insults. If the experiment`s of Lu and team are repeated, but the GCI at a lower level and spaced over a number of days, would the CA1 area of the FLOX mice demonstrate this ischemic tolerance, but the FBN-ARO-KO mice still show extensive damage?

                …if E2 has a neuroprotective effect, would the administration of the ER antagonist ICI 182 780 to the FLOX+GCI mice lead to a reduction in astrocyte reactivity and neuroprotection following GCI that is in line with that shown by the FBN-ARO-KO mice?

Posted in Uncategorized | Leave a comment

computer model discrepancy to animal model in the performance of a spatial alternation task

Posted comment on ´Memory alone does not account for the way rats learn a simple spatial alternation task` by D.B. Kastner, A.K. Gillespie, P. Dayan and L.M. Frank and published in Journal of Neuroscience 2020 40 (38) p. 7311 doi.org 10.1523/JENEUROSCI.0972-20.2020

SUMMARY

   Kastner and colleagues in their article describe their investigation into using a particular computer-based memory model as a tool to indicate animal cognitive behaviour involving carrying out spatial alternation tasks. Models are chosen to reflect specific cognitive capabilities, eg.  Morris Water maze and spatial memory, spatial alteration task and working memory. The authors say that assumptions, sometimes based on straightforward intuition, that the model reflects only the cognitive capability it is designed for may lead the way to misinterpretation and misunderstanding of cognitive processes. Kastner and team state that it is unlikely that single cognitive capabilities are responsible for the way in which animals perform tasks and other factors that bring about behavioural decisions and movements must be identified.

   In their experiments, Kastner and team used the spatial alteration task as model to indicate working memory capability and overall functioning of the hippocampus, striatum, and prefrontal cortex areas. They compared the simulated reinforcement learning (RL) agent equipped with perfect memory process against the rat in the performance of a three-arm spatial alternation W-maze task. For the animal behavioural study, 10 Long Evans male rats were used and the W-maze carried out consisted of three 76cm long arms with a reward well at the end monitored by infrared beams. Pretraining was carried out by placing rewards at alternate arms so that the rats became familiar with movement linked to reward. The learnt task was then performed during 15 sessions over 5 days. At the start of the session the subject rat was placed at the base of arm 2 facing the reward well. The rat received the reward if it went towards the well. In further tests, rewards were given if the rat was at any arm other than arm 2 and then went to arm 2. This was followed by going to the least recently visited arm (either 1 or 2) independent of whether the movement to that arm was previously rewarded.  Therefore, the correct sequence of arm visits was determined to be 2-3-2-1-2. The same protocol was applied to the RL agent.

  For the memory model, Kastner and team used an adapted working memory Todd 2009 computer based model which specifies rules governing propensities that contain preferences of the RL agent of choosing arm A when the state is S. The state was defined as the combination of the current arm location of the RL agent and the location immediately preceding it. This information was maintained in the memory unit. In the adapted version, the agent always updated the information held in its memory unit whereas in the original version of Todd 2009 there was a choice to update or maintain. The adapted version, therefore, allowed the agent to have perfect memory for the task at hand. This resulted in each state having its own propensities which determine the probabilities of making a transition to the other arms updated for each trial. Therefore, if an action gave a reward when a reward was not expected, the propensity was increased. Two parameters that governed the performance of the RL agent through changing the propensities were the learning rate and the temporal discounting factor.

   In the second set of studies, Kastner and team added components to the model for particular arm transition propensities and these were: arm preference which is independent of the current state of the animal; and the preference for visiting arms that are close to the current arm. The probability for visiting each of the arms from a given state gave the propensity and this was calculated by passing it through a softmax. The agent`s visit was determined from this distribution and the choice of arm determined the reward R based on the algorithm that governed the spatial alternation task. The probability of revisiting the current arm was set to zero and then the probabilities for going to the remaining arms set to 1.  Various mathematical calculations were carried out on the model and these were; to update the propensities according to the presence or absence of reward so that the contribution of future rewards were taken into account of the current state; and calculating the learning rate which determined the amount by which all components of the propensities changed based on the new information. The model RL agent was appropriately fitted to the average behaviour of the rats using standard computation methods and setting parameters such as one that minimised the average root mean square difference between the average performance of the rats and the average performance of 200 repeats of the agent (´best fit to reward`), or maximised the total rewards by the model (´max reward`), or minimised the average rms difference between the inbound and outbound errors averaged across all animal and of the average of 200 different repeats of the model.  For all fitting categories only the first 1012 well visits of the animals were counted.

   At the start of the investigation, Kastner and team measured the performance of rats who had to learn to visit alternate arms after each visit to the centre arm with successful performance bringing them reward. For the RL agent, choice of which arm to visit next was based on its current state, S, which was defined by the current arm location of the agent and the previous arm visited by it. Both were maintained in the memory unit. The model had to be matched to the behaviour of the rats at the beginning of the session in order to capture the initial biases of the rats. To do this, the initial conditions were approximated by measuring the probability of the rats visiting each of the arms within the first 10 well visits. Since they exhibited a strong initial preference for arms 1 and 3 then the model was set with the initial propensities to match the initial average error rate of the whole test sample of rats. Results showed that with perfect memory the model was unable to reach the asymptotic performance as quickly as the rats. The RL agent learning was 1.4 times slower. With maximised reward, the model still demonstrated a slower learning rate than the rats, but the parameters provided a closer match to the initial learning phase. The major difference between the output of the model in this case where reward was maximised and averaged for the rats was that the model had a higher asymptotic performance level. Therefore, it was concluded that the model could more closely replicate part of the behaviour (ie. the initial part of learning), but it failed to match the whole learning curve.

  In order to define where the differences lie, the authors then looked at the error pattern of the rats and model. The rules that had to be learnt to successfully carry out the task were used by the authors to define two trial types: inbound trials where the rat was at an outer arm and had to go to the centre arm to get the reward with inbound error defined as the rat failing to do this; and outbound trials where the animal once at the centre had to go to the less recently visited arm in order to get the reward. This leads to outbound error where the animal goes to the ´wrong` arm, ie. the one it just came from. This latter type of trial requires working memory.  Rats learnt the inbound rule much faster and more completely than the outbound rule and this difference resulted, according to the authors, from the rules being differentially sensitive to hippocampal manipulations. The model used by Kastner and team did not capture this difference since after fitting, the errors of the memory model were distributed very differently to that of the rats. Even after manipulation of the fitting, the memory model showed that the inbound trials were learnt more slowly than the outbound ones which was opposite to the results of the rat behaviour. This result provided evidence that memory alone does not govern the way in which rats learn this type of task.

  In order to investigate this discrepancy, the authors investigated preference. They modified the RL agent by making two assumptions: the first, that animals do not randomly visit locations and instead form preferences for certain locations over others; and secondly, that animals in general do not randomly transition between locations but develop preferences for transitioning to neighbour locations. These dynamic preferences were incorporated into the memory model by adding contributions to the propensities so that the choices of the agent now reflected memory and the dynamic preferences. The results of the memory model then showed that the model learnt as quickly as the rats, fitted well the average performance of the rats, closely mimicked the learning rate of the rats and fitted the average inbound and outbound errors of the rats. A study when the model was fitted to the average reward rate showed that it was very similar to that found when fitting to the error rates. This indicated to the authors that the RL agent fit to the overall rewards of the rats and that it made similar types of errors as the rats with reference to inbound and outbound errors. 

   The study was then continued by looking at the individual preferences. When the memory model added only the independent arm preference to the memory unit, then it was found that the model learnt more quickly than the original model even though it was still slower than the rats. The addition of the other component that of neighbour transition preference to the memory showed that it too learnt quicker than memory alone, but not at the learning rate of the full model. This part of the study supported the view that both computations are required and interact synergistically to increase learning to the level of the rats.

  A further study was carried out to see if the model could fit how the rats learnt the task. The enhanced agent (memory, independent arm preference plus neighbour transition preference) was fitted to the animal model by minimising the difference of the likelihood of error between the inbound and outbound trials and by adding a third parameter to reflect the initial conditions. The results then showed that this enhanced agent then represented the inbound and outbound errors of the rats, matched the different time courses to learn the inbound and outbound trials, as well matching the different asymptotic levels of these two error types. The match was reflected in more similar values of the learning rates with the model for the inbound errors overlapping and the outbound reflecting the slower learning rates even though it was still faster than the average rat. This supported the view that there may be additional components to memory that might still underlie the behaviour of animals.

   The authors concluded their article with a discussion about the implications of their results. They stated that the core assumption that memory alone determines spatial alternation behaviour was likely to be incorrect. This came from the results that the RL agent learnt the simple spatial alteration task more slowly and made different errors than a group of male rats. Therefore, the MF RL memory-model could not be said to account for the way in which rats learn a simple alternation task. The results coming from incorporating spatial preference biases enabled the model to fit rodent behaviour more accurately. Therefore, the results demonstrated that it is not appropriate to suggest that memory alone is responsible for animal behaviour in this type of task. The authors also said that there might be other contributions that underlie the rapid learning observed in animals and therefore, including other biases to the model-based RL should be the basis of future research.  The authors concluded their article by saying that since even simple spatial alteration behaviours reflect multiple cognitive processes that need to be taken into account when studying animal behaviour, more quantitative models of behaviour reflecting the complexity of behaviour are required. This is important to consider when this type of tool is used to understand how dysfunctional neural processing occurs particularly when it underlies neuropsychiatric disease.

COMMENT

What makes this article interesting is the indication that the conclusions of computer-based modelling should not be taken as definitive without scrutinising and questioning the validity of the premises and mathematics on which the model is based.  Research carried out by what can be termed as traditional means, eg. laboratory experiments, animal studies, single cell studies have ´controls` built into the experimental set-up and these are the standard measures by which the experimental test conditions are compared.  Researchers aim to satisfy two conditions: firstly, to have a few as possible variables between the controls and the experimental test subjects and usually and hopefully the variable is the one being tested; and secondly, to have both control sample sizes and experimental condition sample sizes as large as possible so that individual variability is accounted for.

  This brings us on to the first questions about the use of computer-based modelling for cognitive studies. In Kastner and team`s experimental set-up the animal study part involved 10 rats with each rat performing a large number of trials. Therefore, the individual cognitive capabilities between the rats were accounted for by the treatment of the results so what remains is the ´average` performance of the whole 10 in the conduction of the task they were asked to perform.  The large number of trials that each performed enhanced the validity of the results, eg. if one rat is slightly more explorative than another this would be ´evened out` when each rat performs over 100 trials and there are 10 individual subjects. With regards to the computer-based model, essentially there is only one ´test subject` because the model is only programmed in a single way. Therefore, the equivalent would be one ´animal` performing thousands of trials. In this case, if the ´animal` displays unusual explorative behaviour, then the results obtained would be ´skewed` towards this if this factor plays a role in the performance of the task at hand. Kastner and colleagues` results indicate that the computer-based model they used in their initial experiments even after adaptation from the original model did not mirror the results they obtained from using their animal test subjects. It was only when further adaptations were made, eg. the addition of further preferences in decision-making that the computer model performed more as expected. This discrepancy is not a problem when researchers can perform the tests in parallel or can use a computer model system that has already been confirmed as mirroring real-life situations and conditions, but it does indicate that ´untested` or ´unverified` computer models may provide results and even lead to conclusions that are not strictly valid.  

   Bearing that in mind, if we accept that the computer-based model system used by Kastner and team in the end mirrored their animal study results, it can be seen that the performance of the spontaneous alternation task is more than just a simple memory recall task. Two parts of the task are in play: the first is the actual movement part which determines where the rat goes; and the second, is the disregarding of whether reward is given or not. The actual informational processes involved in the movement part once the learning phase is completed are: where am I? Look around/remember eg. arm 3; have to return to arm 2; where should I go now? Employ memory recall – where was I before arm 2? Answer arm 3. Task says to go to least recently visited arm, therefore, not arm 3 so have to go to arm 1; repeat stages. The second part of the task is that at all stages, the delivery or non-delivery of reward other than at arm 2 has to be ignored in order that in the future reward is obtained. This is because whether or not reward was issued on arm 1 or 3 was not location dependent per se, but previous position dependent. It is this problem-solving stage where the rats appeared to have the initial advantage. The animals were faster than the original computer model at learning that the reward was given according to previous position rather than actual location.

   As Kastner and team concluded the discrepancy was not dependent on visual capability and simple recall of short-term memory as it would be expected that the computer-model capability would be superior to the animal and therefore, computer performance would be higher. Instead, the discrepancy lies in how the rat learns from its actions compared to the computer model and this implies differences in decision-making during the learning phase. The task demands learning that can be divided into two parts: the easier first one which demands that the subject has to move and always return to arm 2 (Kastner`s inbound task); and the second, more difficult one where which arm the subject visits after this return to arm 2 has to be determined (Kastner`s outbound task). 

  If we consider the first part of Kastner`s task, the rat learnt faster than the computer model that wherever it was it had to go back to arm 2 where it was guaranteed a reward. From a cognitive point of view this is understandable since once learnt, there is no variation in action pattern so no updating is necessary and probably no employment of information processing capability and working memory. This speeds up the overall performance of this part of the task as it can be carried out unconsciously and this is seen when both consciously and unconsciously acquired memories guide decisions (Ruch). In the case of the memory model, the inbound stage was slower than the outbound probably because the original computer model is still choosing between arm 2 and the opposite arm and hence, ´cognitive load` is greater. It is only when the computer model adaptation to visit the arm that is close to the current arm is employed then the inbound task is performed quicker. This is because by default the next action has to be to arm 2 since it sits in the middle of the structured W-maze set-up. Hence, the adaptation mirrors the rat`s unconscious processing and the task becomes straightforward memory recall rather than an example of using working memory capability.  

   The second part of the task, whether to go to arm 1 or 3 independent of reward presentation requires decision-making in the learning phase. The rat achieves this by using trial and error type learning and again performance is faster than the computer model. There are several possible explanations for this.

   One explanation is that the rat knows why it has to perform the task correctly and this influences the speed at which the action is carried out. The rat learns he has to perform the task in the correct order to gain maximum reward. This is achieved by always returning to arm 2 as discussed above and by moving to one arm or the other independent of reward presentation. He learns by trial and error that the task involves moving to the arms in a specific order rather than their actual location. This mere act of knowing why he is doing this order of movements can speed up the action as shown in published research. Model based patterns of reward placement and the ability of the rat to determine that not all cases of receiving the reward are linked positively to future cases of receiving the reward and may speed up the decision-making process (Doll) in comparison to the computer model which is built on logical patterns. A lack of imagining the consequences of not choosing the correct arm and hence, not receiving the reward by the computer model may also cause the discrepancy between it its animal counterparts. It has been shown that prior experience is important for decision making because this allows the representation of options, but in the absence of prior experience or in addition to, imagining the consequences will also achieve the same effect (Barron). The awareness of why an action is taken can make the decision performance more rapid than if the decision is carried out by unconscious processing and decisions made by trial and error are the slowest (Mealor). If the decision has a time deadline, then awareness of judgement is still fast and accurate whereas the accuracy of unconscious processing is lower. Under time pressure, guessing is more likely than conscious processing using knowledge (Mealor). Monitoring and feedback reduces the time pressure and hence, allows faster decision making.

  Part of the monitoring and feedback relates to the placement of reward. Reward is known to influence later decision making (Elliot) and forms the basis of some behaviour. This is why it is used as a stimulus for the required behaviour in the experimental set-up of Kastner and team. It also can be said to stimulate trial and error type learning since it has been shown that high-reward objects (or in the case actions) given as primes increases risk taking (Elliot). Therefore, the reward would encourage the rat to explore. It is also possible that in the same vein as the rat knowing he has to perform the series of movements in the correct order by looking at the reward, the reward should not be taken from the point of view of its value, but by viewing it as the ´risk` of not getting it. This could mean that other brain areas are active which could speed up the decision-making and learning processes. Fear learning is quicker than learning under positive circumstances and therefore, by viewing the reward as ´risk`, the speed of the learning process and decision-making stages are increased relative to the computer model which would be programmed only to see the reward as ´reward`. 

   Another reason that the rat model is faster than the computer model at learning the task is that the rat may use more efficient feedback mechanisms. Monitoring of self-generated decisions and implementation of adaptation is carried out by the prefrontal cortex (Tsiyomoto). One consequence of feedback is that irrelevant information may be ignored by the rat model quicker than the computer model. This involves changing attention to task relevant information and it is known that attention dynamically alters the visual representation to optimize the behaviourally relevant objects (in this case, location) during natural vision (Cukor) and that the anterior thalamus plays a critical role in the direction of attention to task relevant stimuli (Wright). The repetition of events in the training phase and the learning phase also reinforce attention to be concentrated on relevancy as seen in the use of cognitive training programmes not influencing executive decision-making per se, but by improving attentional filtering (Kable, Schmicker). In the same vein, it may also be possible that the rat requires less information as it continues through its learning process in order to make the decision required. It was found that short-term repetition (as seen in the case of the task at hand during the trial and error learning phase) requires complete view and partial views for attended objects, but requires only complete same-views when unattended (Gosling). This would mean that in the case of the rat, the attention to task relevant material would probably lead to a reduced level of characteristics of the cue being required as the learning phase continued.

  Another possible explanation as to why the rat is faster at learning than the computer model is the use of intuition or insight which would be attributed to successful feedback processing. Experiencing insight during problem solving can improve memory formation for both the problem and its solution (Kizilirmak). This links back to understanding the problem and why a solution is required and indicates that the animal model may use this in its processing in order to speed up its decision-making process. This induced insight in relation to words relies on efficient connectivity between the higher order medial prefrontal cortex (responds to problem relevant material) with the rostral anterior cingulate cortex and the left hippocampus (responds to novelty and binds the characteristics together) whereas activation in the left striatum and parts of the left amygdala encodes the solution (Kizilirmak). It is likely that the same areas would be active with regards to location. This insight or intuition may not be conscious since it has been shown that even metacognition allows adaptive modification to ongoing behaviour when external feedback is not possible (Wokke). In this case the information used in insight is not the same as that on which the deliberate, conscious decisions are based (Wokke). Therefore, this could indicate that alternative factors are taken into account by the rat which may not be logically identified and encoded in the computer model and this could explain why the animal model demonstrates faster learning.

  These are only a few explanations as to why the animal model is faster than the original computer model in the performance of Kastner and team`s task. There are likely to be many more and this is why it is important that every experiment has the correct controls in place. The work of Kastner and team shows that the use of computer models as ´control` may not be ideal if they do not represent the biological situation or if the differences or discrepancies to that biological situation are not known. Kastner and team`s article is a reminder that experimental set-ups need to be carefully planned and controlled and that the Method sections of neuroscience articles are just as important sometimes as the Results sections.

Since we`re talking about the topic…………………….

……left basal ganglia lesions result in working memory susceptible to irrelevant information (Baier). If Kastner and team´s experiments were repeated, would the rat model show definite decreases in performance compared to the computer model? In the same vein, age is also linked to higher levels of task irrelevant information because of a reduction in its suppression (Blair). Again, would the use of older animals also show a definitive difference to the computer model in the performance of this spatial alternation task?

…..interference by the use of multiple items or noise has a negative impact on future memory performance (Steinwascher). Can we assume that if Kastner and team`s experiments were repeated, but with 4 arms and not 3 and with 1 arm not participating in the task at all but there as interference, performance of the task in the learning phase would be slower for both models, but is unlikely to affect performance of either once the order is learnt?

…various drugs alter working memory performance (eg. decreases seen with both dizalipine Tsakadi and ketamine – Ma whereas amphetamine leads to a dose-dependent U-shaped change – Lapish). If Kastner and team´s experiments were repeated with the animals administered the various drugs indicated, would the rat model show definite decreases in performance compared to the computer model resulting from poor working memory performance in the decision-making task? Would only part of the experiment, ie. the decision to visit arm 1 or 3 be affected rather than the simple memory recall part involving the return to arm 2?

Posted in learning | Tagged , | Comments Off on computer model discrepancy to animal model in the performance of a spatial alternation task

the importance of forgetting

Posted comment on ´The importance of forgetting` written by L. Gravitz and published in Nature 2019 vol 571 25th July 2019 S12-S14 doi 10.1038/d41586-019-02211-5

SUMMARY

   Gravitz introduces her article by describing the development of the idea that forgetting is not a passive process of memory decay and recall failure, but an active functional process. Therefore, it was said that the standard state of the brain is not to remember, but to forget. The forgetting process is one largely overlooked by researchers and it is possible that a better understanding of it may lead to breakthroughs in the treatments of conditions such as anxiety, post-traumatic stress disorder and Alzheimer`s disease where dysfunctional forgetting play a role.

   The article begins with a general summary of the memory processes and its requirement for neuronal plasticity which strengthens the connectivity and firing of the active neurons representing the event characteristics. This event representation (referred to in the article as an engram) may consist of synaptic connections across several areas of brain and each neuron and synapse of the network can be involved in multiple engrams.  Recall strengthens the neural network of firing cells and Gravitz describes the consistent recall as encoding the memory in both hippocampus and cortex areas until eventually the long-term storage reflects only cortex activity. The article continues by looking at the various aspects of the memory mechanisms from the viewpoint of forgetting and not memory formation.

   The discussion begins by looking at the association of neurotransmitters with forgetting. Gravitz describes the first reports of active forgetting as work reported in 2012 from Davis. Davis used fruit flies in his studies and looked at the action of dopamine released from other neurons on the mushroom bodies, which are dense networks of neurons in insect brains that store olfactory and other sensory information. Conditioning experiments were performed with flies avoiding odours associated with electric shocks. It was found that dopamine blocked the conditioned avoidance response, but inhibition of the same neurons preserved the memory. Therefore, Davis concluded that dopamine regulated how memories could be expressed. Hence, dopamine was said to provide the ´forget` signal. It was also found that dopamine neurons were active for long periods and this observation was interpreted as the brain always trying to forget the information it has learnt. An association between forgetting and neurotransmitters was also described for rats. In this case, the active process of forgetting was shown by Hardt to involve glutamatergic AMPA receptors. AMPA receptors are known to be part of the memory storage process by being trafficked to the post-synapse membrane and forgetting was described as the destruction of the connections at the synapse. Again, the premise that forgetting is not a failure of memory, but a function of it was stated.

   Gravitz`s article then went on to describe the association between forgetting and neurogenesis, which is a process normally associated with memory formation. In this case, the research by Frankland and neurogenesis of hippocampal cells in mice was cited. Using the knowledge that memory formation requires newly formed neurons, Frankland looked at whether increasing neurogenesis would aid memory recall in adult mice. The team found instead that under these conditions increased neurogenesis which would normally provide a greater neuronal capacity actually led to increased memory loss. This was explained by the new neurons in the hippocampus integrating into already established neuronal network actually causing ´disruption` of the network and making information harder to access. Although not always beneficial, it is used to advantage because of the constantly changing environment that the individual lives in and the need for continual learning of new information and adaptation of learnt material by over-writing.

   The article then continues by looking at studies on human memory. Gravitz cites Richards who states that the ability to generalise new experiences is due to the brain`s ability to carry out controlled forgetting. This Richards says prevents ´overfitting` which relates to artificial intelligence where the mathematical model is so good at matching programmed information that the AI is unable to predict what comes next. Richards explains that memory of gist rather than detail allows generalisation of experiences and use of that information in novel situations. This view is supported by Levine with reference to real-life situations where people can have highly superior autobiographical memory (HSAM) or severely deficient autobiographical memory (SDAM). In HSAM, an extraordinary level of detail is remembered, but there is limited informational processing capability and an increased tendency to obsessiveness. These characteristics are attributed to a person´s lack of ability to extract themselves from the pure data. In SDAM the individual is unable to vividly recall specific events and has a problem with projection into future events, but they are good at problem solving and abstract thinking.

   Work by Anderson and team on how active forgetting occurs is cited in the article. They found that in situations where thoughts are actively suppressed then high GABA levels are linked with greater suppression of the prefrontal cortex on the hippocampus. Gravitz says in her article that work on GABA and this type of suppression may explain the action of anti-anxiety benzodiazepines (eg. diazepam) which are known to enhance GABA receptors. The action of benzodiazepines is explained by the prefrontal cortex commanding the hippocampus to inhibit the thought. However, if the hippocampus does not have enough GABA then it cannot carry out the command and therefore, the increase in GABA receptors by anti-anxiety drugs will increase the hippocampus response. Gravitz goes on to say that GABA´s role in suppressing unwanted thoughts hence may be part of the mechanisms linked to phobias, schizophrenia and depression. These conditions present with various symptoms (eg. flashbacks, obsessive and/or depressive thoughts and difficulty in controlling thoughts) which may be linked to an overactive hippocampus. Anderson`s observations and the switch off mechanism orchestrated by GABA may play a role in removing these cognitive problems. Anderson also suggests that this might have implications for treatment of post-traumatic distress syndrome which may be failure to forget. It has been found that people who report more traumatic experiences are particularly good at repressing specific memories. The hypothesis may also apply to Alzheimer´s disease where memories are lost. Hardt thinks that the malfunction of forgetting lies with an overactive forgetting mechanism erasing more than it should may be at play rather than with a dysfunctional recall mechanism.  

  Gravitz concluded her article with a description of the shift of focus of research from the brain´s ability to form memories to the brain`s ability to forget, hence indicating this aspect`s rise in importance. The importance is said to be understandable since in a changing environment adaptive memory with the capability of updating knowledge enables individuals to move forwards and for adaptive memory to exist, forgetting is required.

COMMENT

   What makes this article so interesting is that it discusses ´forgetting` which is one of the many mechanisms involved in memory. Unlike learning, forgetting cannot be thought of as a single type of process because it applies to at least two different situations. The first is where forgetting is ´failure to remember`, a situation such as that occurring with Alzheimer disease and is essentially regarded as negative. This situation refers to when information is no longer available for recall even under prompting due to destruction of the neuronal network required for its representation. And the second situation is where ´forgetting` means that stored information cannot be remembered because it has been ´overwritten`. In this case, the original information has been updated because perhaps it is no longer relevant or needed. This type of situation occurs in for example, extinction in conditioning experiments or more simply knowing how to operate your first mobile of over 20 years ago. In this case, ´forgetting` may mean that the new information supercedes the old information and that it is truly gone or the new information has higher priority in the recall process. Here, the old information is still there, but only comes to mind when prompted to a high degree or when deliberately and specifically searched for. So, here is the first point about using the word ´forgetting`. Forgetting refers to how we, the individual, refers to the recall of the information and not how the brain as a physiological structure views the recall of the information. This indicates how we should look at forgetting in terms of neurochemical mechanisms.

If we look at the first situation, where forgetting is ´failure to remember` this occurs because the information is no longer there or accessible to us due to physiological destruction. This is a different situation to not being able to remember something because it was never there (ie. long-term memories were never formed from the real-time neural representations independent of the reason). Therefore, for this forgotten information we can assume that the physiological mechanisms required to form long-term memories were at some point carried out. This also means that the long-term memories are located in the relevant cortical areas or cerebellum and are in the form of a network of cells (neuronal assembly) that represent the event. The cells of the network fire together in response to a particular stimulus leading to the recall of the characteristics they represent.  Firing of the network is ameliorated (´fire together, wire together`) by the physiological changes that occur to its constituent cells on event storage, eg. increased AMPA receptors (long-term potentiation), increased connectivity through dendrite formation as well as gene modulation that aids neurotransmitter production, vesicle recycling, energy supply and glial cell functioning for example. Two things are assumed here: the first is that the neuronal assembly stays relatively stable for a period of time from its conception to its first use; and secondly, that the hippocampus is purely a relay station in charge of the recall stimulus and timing of event recall, real-time input and real-time informational processing. (Its role in binding of event characteristics and synchronising firing of different areas in the formation of the neuronal assembly is no longer applicable when the information is recalled.) Therefore, in the case of forgetting where forgetting is ´failure to remember` the overlying physiological cause is the decay of the ´silent` neuronal assembly so that a stimulus no longer initiates the firing of the neurons that made it up and that represented the event experienced in the past. The event can be forgotten in its entirety or only part as in the forgetting of certain characteristics.

   If we assume that memory retrieval involves the reinstatement of the neuronal firing patterns, then memory decay involves the physiological destruction or dysfunction of the relevant neurons and the connectivity between them. There are a number of different causes for such failure. The obvious one is the large scale physical destruction of the neuronal cells and this is seen in situations such as injuries, lesions and Alzheimer`s disease. It is clear therefore, that if the cells are no longer present then the network cannot be reactivated and the event cannot be remembered.

  Another important cause of memory failure is the destruction of neuronal connectivity which leads to synchronisation and timing issues. The cells representing the event act together and to a large extent retrieval of information relies on the reactivation of the same end cells that fired during encoding. The connectivity of the network relies on the activity of a number of brain areas, eg. greater similarity between patterns of firing of the dorsal lateral prefrontal cortex during encoding and retrieval relate to better memory recall performance (Javadi) and the increased connectivity between the perirhinal cortex and other areas relate to object recognition performance (Staresinaln). The specificity of this brain area connectivity is sometimes associated with particular information content of the memory, eg. autobiographical memory preferentially activates the areas of the default mode network (Chen) whereas visual memories activate parietal and frontoparietal areas (Chen and Ye) with both memory types having functional connectivity to the hippocampus (Westphal). Therefore, destruction of the connectivity or dysfunctional connectivity can lead to decay of neuronal networks specific for the event representation with the result that recall of the specific information fails. Dysfunctional connectivity also links to the failure of the instigation of appropriate brain waves consistent with memory reenactment, eg. beta waves between visual regions and parahippocampal cortex being required for the reinstatement of neural patterns matching the retrieval of visual information (Staudigl).

   Failure to remember can also be due to micro-scale dysfunction at the neuron level since firing of each relevant cell is a necessary process for event characteristic retrieval. Since there are many components and processes required for satisfactory neuronal function in both encoding and recall of event characteristics there are many points at which dysfunction or deficit can prohibit the firing process necessary for re-enactment. For example, loss of memory reported in traumatic brain injury is related to loss of dendritic spines (Sen) which would diminish the capacity to receive the transmitted signal whereas forgetting of spatial memories is said to be caused by the removal of the AMPA receptors at the neuronal cell membrane (Migues). These are normally required for long-term plasticity of the neurons required for the event representation.

  It is clear therefore, that forgetting can be due to failure to re-instigate the event representation due to dysfunction and deficits of the neurons and networks that contribute to it. However, forgetting can also be due to failure to adapt existing memory when new information is presented. In this situation, the old information is remembered, but the failure to update with the new information means that this will be ´forgotten` at the next demand for retrieval. Updating event representations is an important part of the memory process and requires a number of correctly operating processes in the appropriate order. For example, it requires retrieval of past information (event representation re-enactment), simultaneous input of the ´new` information, ´comparison to or addition to` type decision mechanisms, and physiological processes for long-term storage as part of the re-consolidation of the ´old` information and binding and storage of the ´new` information.

   The correct functioning of two brain areas in particular, the hippocampus and the prefrontal cortex, appears to be required if ´overwriting` is successful. In the case of the hippocampus, this area plays a vital role in the relay of information and its binding during the event`s encoding. With regards to ´overwriting`, it is likely that it plays equivalent roles not only in the relay of the new information to the upper cortical areas, but also maintains that firing in order for the ´comparison` mechanisms to be carried out. This view is supported by research that shows the hippocampus`s requirement in matching of ´old` information to ´new` information as seen in face-diagnosis memory with connectivity with the left middle temporal gyrus (Brod) and contextual fear conditioning where learning is required if the recalled information does not match the new input (Bernier). Theta gamma rhythms involving the hippocampus at the time of retrieval appear to be important as to whether ´new` information overwrites the ´old` in fear memory consolidation (Radiske). Therefore, failure to overwrite manifesting as ´forgetting` may be caused by a dysfunctional hippocampus activity during the encoding and recall processes.

  The other important brain area relating to failure to overwrite the old information with new information is the functioning of the prefrontal cortex. This is a key area in decision-making processes which relates to the question whether the ´new` information is assessed as being ´valuable` and hence, an overwrite is necessary or irrelevant and classed as ´interference` and ignored. Research shows that the prefrontal cortex performs this role in the case of reward prediction errors with connectivity to the ventral tegmental area with the decision whether to maintain or deviate from previously learned cue-reward interactions (Ellwood). It is likely that both the dorsolateral prefrontal cortex (known for its role in strategic control and working memory – value assessment of incoming stimuli) and the ventromedial prefrontal cortex (known for its role in value assessment and comparison, decision-making awareness of choice  and switching of attention) are involved. Therefore, failure to overwrite can be caused by the failure of the prefrontal cortex to bring about the decision to update.

   Apart from dysfunctioning at one or both of these areas, another reason for failure is that the neuronal firing attributed to the new information is not strong enough to overcome the recalled information if it is contrasting, or not strong enough to form part of the event representation if the characteristic is to be added. Similarity to the ´old` information may be an advantage with greater similarity to the stored information being shown to be more easily encoded so leading to an improvement in memory performance with time (seen with face-diagnosis pairs – Brod). This may be due to the sharing of cells already exhibiting strong firing from the similar characteristic. However, the lack of strength of the firing for the new event characteristics may be due to input deficiencies, eg. poor visual input due to changing gaze, interference of visual details by flashing lights. Two factors may play roles here and these are emotional status and age which are both attributed as causing changes in memory performance. For example, negative emotional status such as from fear or stress is known not only to negatively impact informational input by changing attentional performance in general, but can also affect specific content. In the case of stress, there is interference with long-term memory for associated material (Trammel) and impaired memory selective for content, eg. memory for items is deficient, but not for background information (Steinmetz). The highest state of anxiety is shown to cause the inhibition of retrieval for both threatening and non-threatening informational categories (Nunez) and therefore, this will have an impact on whether event representations are updated or not. This appears to be the case also with avoidance strategy where an ability to suppress unwanted, upsetting memories. This is shown to have effects on the recall not only of the distressing event, but also of other details (Quang) and low arousal emotion can facilitate the recall for peripheral information if directly relating to it (Davidson). Therefore, if information is not being recalled, the capability to ´overwrite` it is disrupted. Positive emotions can also affect memory retrieval as shown in the case of episodic memory where performance is lower following a highly positive event even though the executive control performance is unaffected (Lagner). Therefore, as described, memory enhancement or impairment by emotion depends on the nature of the information to be retrieved and the circumstance (McKenzie).

   Another factor that can affect the performance of updating memories appears to be the age of the individual. Ageing has been shown to have an effect on memory performance attributed to physiological changes of the neuronal network, eg. aged rats exhibit defective recognition memory and alterations in hippocampal synaptic plasticity through defective LTP and enhanced LTD (Arias-Cavieres) and episodic memory declines with age, an observation that correlates to regional connectivity of the default mode network plus the medial prefrontal cortex (Huo). However, the situation is not clear cut, since age is not the reason for changes in specificity of detail retrieved for autobiographical memories since both young and older age groups show deficits (Aizpurua) and the frequency of involuntary autobiographical remembering appears not to decline with age (Berntsen). Reduced capability levels may be attributed to an inability to inhibit the incoming and storage of irrelevant material which is shown to increase with age rather than the more severe long-term physiological deficiencies. This view is supported by the reported positive effects of cognitive training which may increase older individuals` abilities to inhibit irrelevant material whereas for younger individuals it leads to an improvement in their cue-integration capability (Cappelleti). Therefore, forgetting may be due to selective failure to store new information to overwrite the old.

  Therefore, this comment shows that the topic of forgetting cannot be regarded as a simple failure to remember. It may involve different mechanisms depending on the circumstances in which it occurs or is demanded. Gravitz in her article supports the view of Davis that the brain is always trying to forget, but this is not strictly correct. The brain tries to remember but only remembers what it can when given the right conditions to do so, or what it is demanded to remember through conscious control. However, the transient physiological nature of neuronal cells means that the memory system is always in flux and cells are destroyed and formed to try to maintain what is important to the individual and recalled information normally means ´important`. This demand is not always fulfilled and success at memory retrieval is not always possible with information being lost or not being updated so effectively ´lost`. Therefore, as Hardt says, this latter type of forgetting is a function of memory and not a failure of it. It is possible that a through a greater understanding of the mechanisms that contribute to desired forgetting, solutions to the unwanted forgetting as seen in dementia for example may come to light.

Since we`re talking about the topic………………………

                ….a single electroconvulsive therapy (ECT) application is known to disrupt reactivated, but not non-reactivated memory recall for an emotional event in patients suffering from depression (Kroes). Would controlled staged recall of a highly emotional event interrupted with ECT applications lead to forgetting of the event in its entirety over time or would the repetition of the recall at each state consolidate the memory even more?

                ….–recent memories are generally recalled from a first person perspective and older memories from a third person perspective (Butler). The repeated retrieval of visual details in the first person is shown to lead to better retention of material and slowing of the shift from first person to third (Butler). Can we assume that by directing retrieval with imagery or cues (Harris) and instructing recounting in the first person that recall performance may show an improvement?

                ….the capability to update information into beliefs relies to some extent on whether the information is desirable or undesirable, with the former greater than the latter (Garrett). This biasness does not exist when there is perceived threat to the environment (Garrett). Therefore, would ensuring that the way information is ´phrased` is positive lead to an improvement of event characteristic recall?

Posted in memory recall | Tagged | Comments Off on the importance of forgetting

synaptic plasticity of proximal and distal dendrites of CA1 pyramidal neurons

Posted comment on ´Synaptic plasticity depends on the fine-scale input pattern in thin dendrites of CA1 pyramidal neurons` written by A. Magó, J.P: Weber, B.B. Ujfalussy and J.K. Makara and published in Journal of Neuroscience 40 (13) p. 2593 doi 10.1523/JNEUROSCI.2071-19.2020

SUMMARY

In this article, Magó and colleagues describe the findings of their investigation into the plasticity of glutamate excitatory input patterns located in the peri-somatic dendrites of CA1 pyramidal neurons of the rat hippocampus. Their aim was to explore the spatially structured forms of synaptic potentiation in those dendrites since this activity and long-term plasticity are considered as important for shaping experienced-related information processing.

    In their experiments, Magó and team used suitably prepared transverse slices from the CA1 area of the hippocampus. Current clamp whole cell recordings were taken from the somata of hippocampal pyramidal cells (CA1PCs) and only cells whose resting potentials were more negative than minus 55mV were used. The neurons were imaged with a dual galvanometer based two-photon scanning system which was also used to uncage glutamate (2PGU) at the individual dendritic spines. Two ultrafast pulsed laser beams were used for imaging the fluorophore and to photo-lyse the MNI caged glutamate and the emitted light was collected by multi-alkali or GaAsP photomultipliers.  The neurons used in the experiment had complete apical oblique dendritic arbors (35%, proximal stratum radiatum) and basal dendritic arbors (65%, stratum oriens). Stimulation via uncaging glutamate occurred laterally to the heads of the visually identified spines which were separated by at least 1.1micrometers for 0.5msecs. The time interval between stimulation of spines was 200ms or 0.1 for the LTP (long-term potentiation) protocol. Repeated excitatory post-synaptic potentials (EPSPs) were measured. The LTP experiments involved recording the EPSPs evoked by 2PGU in the whole cell voltage clamp experiments so that photo-damage could be taken into account (damaged or unusually acting cells were excluded) and EPSPs were evoked regardless of the depth of the spines used. To avoid the problem of washout of the intracellular components by whole cell dialysis, the method used with the LTP protocol started within 10 minutes from the time of the measurement of the control whole cell configuration. Four test spines were used as the sample and the laser power was kept the same for all homo-synaptic experiments, but it was increased 15% for some hetero-synaptic neurons so that d-spikes were evoked. The presence of d-spikes by at least one stimulus was confirmed also by visual inspection and these were most clearly detected at the first stimulus of the LTP induction protocol whereas later the results were more ambiguous and therefore, not taken into account.

   Magó and colleagues performed data analyses on all their results using various methods. Several conditions were used. For example, the mean normalised change in EPSP amplitude of all test spines averaged between 30 and 40 minutes after the LTP protocol indicated the magnitude of plasticity; the potentiation of an individual spine was said to occur via a normalised EPSP greater than 1.3 after LTP; and spines that were retracted, that failed to produce the normalised EPSP, that exhibited other unusual firing behaviour, or moved closer to other spines were not included in the analysis. Other analytical methods used included Image J which was used to measure the cell morphology and distance using dye-loaded neurons. A detailed biophysical CA1PC model was also used to reproduce the dendritic processing. Default passive parameters were used plus adjustments for ion channel activation (eg. voltage gated sodium, potassium) and AMPA and NMDA excitation. Therefore, the somatic and dendritic properties of the sample cells were determined including: the generation and propogation of Na+ action potentials at the soma and along the dendritic trunk; the generation of local Na+ spikes in thin dendritic branches; the amplitude distribution of synaptic responses; nonlinear integration of inputs via NMDA receptors; the similar voltage threshold for Na+ and NMDA nonlinearities; and the major role of A type K+ channels in limiting dendritic excitability. In vivo, stimulation produced the same parameters, but with place-selective activity.

   Magó and colleagues found with their experimental set-up that regenerative d-spikes are required for efficient cooperative LTP at proximal dendritic locations. The EPSP amplitudes of four test spines at either proximal or distal locations along individual branches were measured and then short PGU pulses (the LTP protocol) were applied. The d-spikes if present were measured in comparison to the amplitude and kinetics of the expected EPSPs. Previous experiments had shown that no subthreshold LTP was evoked with the co-activation of only 4 test synapses located at proximal sites along the peri-somatic dendrites and therefore, the synapse cluster was increased to 12-16 by including neighbouring spines. In this case, synchronous stimulation evoked substantial somatic EPSPs (3.9) with small peak EPSP non-linearity (0.7), but no regenerative d-spikes occurred. This was explained as being likely due to low impedance of proximal dendritic segments. However, no LTP was observed at the test synapses and the mean EPSP amplitude of test spines decreased.

  Magó and colleagues repeated the experiment, but changed the locations of the spines and in this case, d-spikes were evoked. For this to occur, 4 proximal test spines were co-activated during the LTP protocol with another 11 spines located more distally, but on the same dendrite. The d-spikes evoked were probably due to the extension of higher impedance of the dendritic segments. They were observed via transient increases in the rate of compound EPSP rise (indicative of dendritic Na+ spikes) and/or peak somatic non-linearity greater than 2mV (indicative of NMDA spikes). Somatic APs were excluded since they were prevented by slight hyperpolarisation during the LTP protocol. The result of the evoked d-spikes was the induced robust long-lasting increase in the mean EPSP amplitude of the 4 proximal test spines (1.98). When APs were also evoked by at least 1 of the 50 LTP stimulus pulses then EPSP amplitude increased to 215, similar LTP was recorded and potentiated synapses were found in all experiments. This suggested to the authors that large depolarisation involving regenerative d-spikes is required for the cooperative LTP induction of synapses located in their sample of proximal segments of peri-somatic dendrites.

   Previous investigations by the authors showed that repeated co-activation of 4 clustered test spines located instead distally increased their EPSP amplitude (1.32) and potentiation occurred in at least 1 spine. However, d-spikes were not detected during the LTP protocol. Experiments described here showed subthreshold LTP observed occurred with at least 3 synapses (increased EPSP amplitude – 1.35) dependent on location. No subthreshold LTP was produced when 4 co-activated test spines were distributed evenly on longer dendritic stretches (EPSP amplitude 0.91), but tight clusters of more than 3 coactive distal inputs could be strengthened by subthreshold LTP even in the absence of regenerative d-spikes. This demonstrated spatial selectivity. This led to further investigation as to whether subthreshold LTP changed in the presence of local d-spikes in distal dendritic segments. Therefore, during the LTP protocol then additional neighbouring spines were co-activated together with the 4 clustered test inputs to trigger d-spikes. Eight synchronous synapses evoked in general at least 1 d-spike during the LTP protocol without somatic AP and induced LTP in at least 1 test spine. There was no significant difference between these results in terms of magnitude of LTP or ratio of potentiated synapses to those obtained with only 3-4 clustered inputs. Magó and team also found that LTP with d-spikes depends on the spatial arrangement of the inputs with additional proximal synapses. In this case, they distributed the 4 test spines and co-activated them with 4 more additional proximal synapses to trigger the d-spikes. In 4-5 experiments at least one synapse was potentiated with an average LTP of 1.52 and therefore, the effect was significantly stronger than with only 4 distributed sub-threshold synapses. Therefore, it was concluded that the more extended propagation of d-spikes especially towards the sealed tip allows more distributed input patterns to potentiate.

   Magó and colleagues then continued their investigation by exploring the results of previous research that reports that using electrical stimulation d-spikes can trigger synaptic potentiation with fewer stimulus repetitions than other LTP-inducing activity patterns. The authors therefore, tested with a short LTP protocol consisting of only 5 co-activations of 4 subthreshold for d-spikes, or 8 suprathreshold clustered spines. The team found that suprathreshold clustered inputs developed robust LTP (1.5) whereas 5 asynchronous activations of subthreshold clustered inputs (0.79) did not. Therefore, d-spikes although not necessarily required for LTP at distal dendritic segments they can alleviate the tight spatial clustering requirement and reduce the number of synchronous events needed to induce LTP.

  The authors continued their investigation by looking at the crosstalk involved in local plasticity. Co-activation of a group of 8 spines during LTP induction which triggered d-spikes produced variable effects on a set of 4 test spines nearby. The long-term change in test spine EPSP amplitude was smaller than that for homo-synaptic LTP with d-spikes (1.43), but still signs of potentiation in the test spines were detected. In the majority of experiments, EPSP increased more than 30% in at least 1 test spine as well as in 12 out of 30 experiments the mean EPSP amplitude changes in the test spines were larger than the mean control with no LTP protocol. These changes were independent of test spine location (ie. proximal or distal) from the LTP induction spines. The crosstalk potentiation was not observed when the test spine and LTP induction spine groups were located at short distances on different dendrites of the cell and the EPSP amplitude was decreased when compared to the spines all being on the same dendrite. These observations were interpreted as the crosstalk potentiation affecting only the activated dendritic segment and involving intracellular rather than extracellular signals.  

   The next set of experiments described by Magó and colleagues was carried out to investigate whether the observations recorded could be attributed to a general change in electrical properties of the test cells due to the repeated stimulations in the LTP protocol. The strength of the dendritic sodium ion spikes showed no change in value and therefore, it was concluded that there was no change to dendritic excitability during the course of the experiments. Using the biophysical CA1PC model, strong stimulations elicited local dendritic sodium ion and NMDA spikes in the model which were visualised as small fast spikelets and slow plateaus in the soma. Increasing the local excitability of the branch (eg. by changing the passive parameters such as increasing local membrane resistivity and decreasing axial resistivity) led to increased amplitude of the individual EPSPs and increased the dV/dt of the somatic spikelets. When the local excitability was changed by locally eliminating potassium ion channels the EPSPs did not increase, but the dV/dt of the somatic spikelets did. Increasing the AMPA conductance of the synapses by 40% led to increases in the amplitude of the EPSPs, but the dV/dt did not. This showed the authors that crosstalk was likely to have occurred and could not be explained by changes in local dendritic excitability evoked by the LTP induction protocol.

  Crosstalk was further investigated since previous reports showed that LTP at a single spine could lower the LTP induction threshold at nearby spines. Therefore, Magó and colleagues tested whether crosstalk plasticity may be related to the weak test stimuli applied to monitor the EPSPs. It was found that the number of pre-LTP stimulation events did not affect the magnitude of the LTP by individual spines, but alternative analysis of groups of cells showed that the spine group receiving less than 15 pre-LTP stimuli produced a smaller EPSP amplitude change (1.02) than those receiving more (1.43). This indicated to the authors that the activation of synapses before LTP induction by other synapses may facilitate crosstalk potentiation.

   The last set of experiments carried out related to the neurochemical mechanisms of the crosstalk observed. The absence of potentiation in the presence of d-AP5 indicated that the mechanism appears to be NMDA receptor dependent. Using VGCC inhibitors, it was found that there was an initial increase in EPSP amplitude, but this was followed by a gradual decline. Therefore, although NMDA receptors are required the calcium ion channels VGSCCs are not. They were however, suggested as being beneficial for the stabilisation of the neurochemical process. Another requirement appeared to be the MEK/ERK pathway since crosstalk was eliminated by the inhibitor U0126.

   Magó and colleagues concluded that several mechanisms can lead to local, spatially structured LTP and found that there are rules about cooperation were based on dendritic location. Even low numbers of clustered inputs on distal segments of thin dendrites, providing they were close, cooperated and co-potentiated without generating d-spikes. However, local d-spikes play a role to induce LTP at proximal dendritic segments and it appears that the potentiation is most likely when input patterns are distributed throughout the dendrite.  Therefore, it was suggested that subthreshold and suprathreshold LTP may be hierarchically organised so that initial gradual potentiation of repeated activated small distal input clusters would help to reach d-spike threshold leading to a second spatially less constrained and faster mechanism evoking somatic action potential firing leading to neuronal cell assembly plasticity. Hetero-synaptic potentiation of inputs in the vicinity of synapses evoking d-spikes, although less prominent, was observed with both firing potentiation and depression. This crosstalk plasticity was suggested as being NMDA receptor dependent and mediated via the MEK/ERK pathway. The function of the crosstalk was suggested as binding together temporally separate informational characteristics of a common event onto the dendritic segment so that subsequent synchronous recurrence of those characteristics would be achieved more efficiently. In this way, the d-spikes and dependent LTP cooperation would form part of the neuronal tuning required in the formation of the event representation and its memory. 

COMMENT

  What makes Magó and team´s article interesting is that it describes dendritic firing which is an aspect of neuronal post-synaptic firing that is normally overlooked in the rush to investigate the more exciting topics of changes to glutamate receptors and ion channels. However, measurements of dendritic firing is an indication of neuronal firing and cell status and can as reported in Magó and team`s article be influenced by the cell`s environment and functional demands.

  Magó and colleagues looked at dendritic firing of one type of cell, the pyramidal cell in the area of CA1 of the hippocampus. This type of cell is ideal for studying dendritic firing because the cells can be described as a ´tree-like` with ´branches` (the apical arbor, proximal since away from the soma, but widely linked to the ´branches` of its neighbouring cells), the ´trunk` (long, with the possibility of discerning segments experimentally) and ´roots` as distal dendrites (closer to the soma, but also widely connected to neighbouring cells). Although pyramidal cells are found in the cerebral cortex, cerebellum and hippocampus, it is the CA1 area which the authors used as their tissue source. The advantage of using this particular area is the uniqueness of neuronal firing mechanisms, the connectivity of the area (within the hippocampus itself and externally) and the consequences of these two aspects on cognitive function (ie. information input, processing and memory). Therefore, this comment explores the influences of dendritic firing in relation to CA1 cognitive functionality.

   The comment begins with a general look at dendritic firing and activity at the synapse and it centres on firing initiated by the action potential signal travelling in the direction of the post-synaptic membrane to the soma. It should be noted that back propogation firing can also occur in the dendrite with firing in the opposite direction, but this is not discussed here (see Role of action back propogation in pyramid cell apical tuft dendrites – April 2020).

  In general, firing of the dendrite as part of the signal transmission process begins at the post-synaptic membrane with the binding of the released neurotransmitter (normally glutamate in the case of the CA1) to the glutamate receptors at the membrane surface. A simplified explanation of the process is that binding to NMDA type receptors or AMPA type receptors leads to the opening of the associated sodium ion channels so that there is an influx of sodium ions into the post-synaptic area. This depolarisation results in calcium ion channel opening and the instigation of the post-synaptic cascade of mechanisms that make up signal transmission propogation and regulation of the synaptic receptor number (eg. IP3 production, CaMKII activation and protein kinase C activation). Binding of glutamate can also occur to NMDA receptors that are linked to SK channels which on depolarisation lead to potassium ion channel opening, release of bound magnesium ions on the outer membrane surface which then bind to the attached NMDA receptor and blocks its activity. Therefore, there is a forward motion of the action potential signal from the post-synaptic membrane to the soma. The strength of the signal depends amongst other factors on the amount of neurotransmitter released and the number of glutamate receptors present on the surface. One method of regulating dendritic firing with relation to synaptic physiology is the presence of SK channel linked NMDA receptors at the synapse (Ngo-Anh).  This is shown to be the case in Schaffer collateral-CA1 synapses where a difference in neuronal firing between dorsal and ventral placed pyramidal cells is linked to the number of SK channels present (Babiec).

  The dendritic firing is observed as burst firing, ie. d-spikes and characteristics of this type of firing is linked to long-term potentiation (LTP) which is indicative of plasticity, are described in the article by Magó and colleagues. If it is assumed that dendritic firing mirror that observed by Magó and colleagues, then location and number of dendritic spines are important for the appearance of the d-spikes. In the case of proximal apical dendrites then according to Magó and team more than 4 test spines are required to evoke the d-spikes. The induction of d-spikes in their experiments requires 4 proximal test spines and a further 11 spines which are more distally located on the same dendrite.  This strong depolarisation achieved by spatial summation of firing from the proximal and distal located spines of the same neuron causes a transient increase in the rate of compound EPSPs rise indicative of a sodium ion spike. (The calcium ion effect appears not to be involved since PKA blockers significantly inhibit NMDAR-mediated calcium ion rises in activated dendritic spines but have no significant effect on synaptic current.) Therefore, the d-spikes evoked, indicative of the transmission of the signal in a forward direction, require cooperative behaviour from spines in proximal locations which are likely to receive input from the directly connected neurons from the CA3 hippocampal area supplemented by firing of those spines located distally. These may also receive input from the CA3 area or perhaps from other brain areas. In this case, the firing pattern representative of the current event may have characteristics from many sources.  

  Magó and team also found in their study that between there has to be synchronous stimulation of between 12 and 16 test spines if these are located on neighbouring spines rather than the same one before there is substantial EPSPs of the cells. However, no d-spikes are observed. This indicates a difference in functionality since the d-spikes are associated with the forward transmission of the action potential signal and when relevant, the induction of LTP. It is likely that in this case the contribution of the cells to the firing pattern lies in the strengthening of the local firing for a purpose other than direct signal transmission, eg. induced ion level changes for vesicle transport or enzyme activity regulation.

   When only distally located spines are taken then according to Magó and team the spines have to be clustered tightly in order that the increase in EPSPs translates into d-spikes, forward sweep and induced LTP. This may indicate that the firing response is specific rather than random and that input comes from one specific area and has to be strong.

   Linked to the forward transmission of the signal is the association of the dendritic d-spikes with the adaptive firing response, that of LTP (Magó). Repeated stimulation strengthens firing patterns and allows encoding of them via long-term physiological changes so that they can be reactivated at a later date. The increased firing sensitivity allows lower levels of firing to induce event representation of previously experience events and allows past events to influence the present. The dendritic d-spikes are associated with LTP (Magó) and hence, provide a means of experimentally investigating neuronal plasticity in response to repetitive stimulation. As described before, the induction of LTP requires a long-lasting rise in EPSP amplitude in the case of proximal cells as long as the spines are on the same dendrite plus preferably the supplementation of the signal with firing from more distal spines. The distally located spines need to be located in tight clusters in order to induce both d-spikes and LTP (Magó). This spatial summation and spine location specificity may be linked to the brain waves observed since dendritic firing is associated with gamma wave burst firing and theta brain wave oscillations (Tominga). These two brain waves appear to be associated with two different LTP systems in the CA1 pyramidal cells (Zhu). The gamma waves are reported to require a system involving adenosine A2 receptors, PKA activity and actin polymerisation whereas the theta brain waves with calpain-1-mediated suprachiasmatic nucleus circadian oscillatory protein degradation, ERK activation and actin polymerisation (Zhu). Therefore, the differences in LTP systems induced may reflect the contributions of dendritic firing from the distal and proximal pyramidal areas to the overall event representation.  

   The LTP observed in pyramidal cells of CA1 is likely to follow the mechanisms employed to induce plasticity in the mossy fibres of the CA3 (Nicoll), ie. released glutamate activates the post-synaptic NMDARs and AMPARs. Calcium permeable AMPARs are associated with calcium ion influx through L-type channels. This type of LTP is associated with the induction of PKA (Park) and to one form of NMDAR dependent LTP. This PKA dependent form requires multiple firing events linked to gamma burst firing or burst firing in the theta range. Both lead to the insertion of calcium-permeable AMPARs into the post-synaptic membrane (Ehringer, Koener, Plant) and the induced plasticity via this method results in strengthened firing.

   The LTP observed in addition to the different levels of SK channels and the distal or proximal contributions of different signal sources described above can also be influenced by the number of NMDARs and AMPARs present since these would have effects on firing, ion influx and plasticity. In an extreme example it is possible that there are no AMPARs and only NMDARs on the spines, the so-called ´silent synapses`. Repeated stimulation of this type of neuron leads to LTP by inserting AMPARs into the membranes just like with normal LTP (Arendt). Such an insertion has the effect of ´switching on` the synapse. Therefore, d-spikes may occur via NMDAR binding in a cell with no AMPARs which then promotes LTP and the insertion of AMPARs into the membrane. This results in the action potential threshold being in subsequent events achieved and the cell firing. The event representation in this case will include these characteristics and hence, be expanded.  This therefore, changes the contribution that a particular cell can make to the overall firing pattern. Hence, a dendrite that is previously subthreshold will become a dendrite that is part of the transmitted signal due to the insertion of AMPARs and spatial summation. The resulting firing pattern will be expanded to include this information in the event representation.

   The CA1 also experiences plasticity in the form of long-term depression (LTD) which is the natural progression from LTP in memory formation (consolidation of fear memories, but not acquisition – Liu). However, there are no specific reports of d-spikes linked to LTD in the CA1 area. This is unexpected since LTD is thought to be associated with extracellular calcium ion levels and NMDAR calcium ion channels (Babiec). The NMDAR receptor mediated activation of p38-MAPK and dephosphorylation of the GluA1 subunit of the AMPAR and subsequent internalization of the receptor implicated in LTD is also shown to be involved. Therefore, the dendritic firing and action potential induced because of spatial summation would be expected to cause the shift from LTP to LTD as expected. The observation that LTD prevents subsequent LTP of AMPAR transmission (Bhouri) may explain why. The aim of dendritic firing is to promote signal transfer and event representation and therefore, the calcium input is high, LTP is welcomed, but the switch off mechanism of LTD is not. If LTD does not occur then the forward signal is continued. The alternative view that LTD is said to be caused by the selective activation of a group of presynaptic metabotropic glutamate receptors (mGlu) instead at the Schaffer collateral-CA1 synapses (Rosenberg) may also provide an explanation as to why LTD is not associated with dendritic firing in this area.

  Therefore, d-spikes and LTP are linked to CA1 cognitive function capability and this can be used to advantage by specific input and output connectivity of the area. Stimulation of the CA1 leads to excitation of the entorhinal cortex, to the dentate gyrus (DG) of the hippocampus plus the prefrontal cortex and other areas and these all play roles in the event representation, memory and information processing. Different connectivity patterns support the signal transmission of one type of information over another. For example, there is greater connectivity between the amygdala and hippocampus in emotionally arousing material (Fastenrath); input from the lateral entorhinal cortex into the distal CA1 is important for information about objects, whereas the input into proximal pyramidal cells from the medial entorhinal cortex provides information about space (Hartzell); the inputs from the medial entorhinal cortex to the CA1 are also required for the temporal organization of the hippocampal firing patterns and therefore, are important for sequences of information (Schlesiger); and synchronisation of the theta wave oscillations of the CA1 and beta wave oscillations of the ventral striatum indicate reward expectancy (Lansink).

   The combined effect of firing mechanisms and connectivity allow the hippocampus to be a major player in the neuronal representation of event information, processing of that information and memory. Therefore, as expected any physiological changes to the particular areas can have wide-ranging effects on a cognitive scale, eg. epilepsy induces unusual resting connectivity between the anterior and posterior hippocampal areas which is associated with memory decline (Voets). In the examples given above it can be seen that specific areas of the hippocampus have different functions relating to event features, eg. distal CA1 receives input from the lateral entorhinal cortex and is important for information about objects, whereas the proximal pyramidal cells receive input from the medial entorhinal cortex and provide information about space (Hartzell); and there is differential encoding of spatial location and numeric object characteristics in the CA1 and CA3 areas (Opris).  The temporal interval between neuronal firing is important because this indicates the boundary of the event representation. This allows binding of appropriate characteristics together which are represented by simultaneously firing neurons and allows sequential events to be recorded. Dendritic firing with proportions of proximal and distal firing allows spatial summation to occur for features that under normal circumstances would be too weak to be part of the event representation or makes certain features stronger by increasing their firing contribution to the overall neuronal firing pattern in preference to others. Two examples come to mind: events where both object characteristics and subtleties of location are required, eg. in spatial encoding and memory dependent on three dimensional interpretation of viewpoint, placement and direction (Kim) where it has already been shown that there is a scale of spatial information between the dorsal and ventral poles of the CA1 with the threshold for LTP induction higher in the dorsal CA1 neurons (Malik); and in event representation where emotional information plays an important role, eg. in fear situations. Working memory may also benefit from the distal and proximal specificity of the dendritic firing since this requires real-time processing, maintenance and possible matching of multiple simultaneous event representations whose origins of material can be both real-time and reactivated. This hippocampal function (Vila-Ballo) requires both glutamate and NMDAR activation (Takadi, Gage). 

   The concepts of sequences and temporal order may also be elicited via CA1 pyramidal firing patterns (Allen) which may or may not involve dendritic firing contributions of the distal and proximal spines. It is known that there is a general link between hippocampus cell firing and timing (Jacobs). Sequence coding of non-spatial events has been associated with CA1 and gamma wave oscillations observed there are modulated more strongly by temporal context than theta waves (Allen, Salz) with theta wave activity attributed to clock-like activity of small groups of CA1 neurons (Zhang).

   The contributions of the distal and proximal areas in order to elicit forward firing of the signal or to widen the informational value of the event representation may have relevance when the mechanisms relating to memory formation on the basis of real-time neuronal firing patterns is considered. The hippocampus is known to be important for encoding of the short term event representations (mice create what-when-where memories – Fellini) into longer term memories and this process relies on the LTP changes at the synapse described above (Winocur, Hawkins). The CA1 plays a role in this by aiding the encoding of new information in the face of retrieved information stored from previous experiences (Douchanps, Kuhl). Two mechanisms involving the CA1 have been suggested for this priority. The first is that acetycholine released in firing relating to the input of new information selectively suppresses the excitatory projections from the CA3 which is involved with the reactivation of the past experience to the CA1. It has no effect on the entorhinal inputs in to the CA1 which is involved with the new material (Douchanps). The second mechanism involves the preferential encoding of the new material at the pyramidal layer associated with the theta brain wave oscillation peak relating to the input from the entorhinal cortex whereas retrieval occurs at the trough of the theta wave coincident with the input from the CA3. Therefore, there is a shift towards encoding of new information. This is not the case with reactivated memories where familiar events lead to bigger responses (Fried). In this case, the antagonism of acetylcholine firing shifts the theta firing phase to the theta wave oscillation trough associated to retrieval and involves both CA3 and DG areas (Douchanps, Reigh). Reactivation in the hippocampus also leads to patterns in the cortex and stabilizes the connections between the hippocampus and the cortex (Frankland).

   The shift in CA3, CA1 firing dominance according to whether the input relates to reactivated or novel material may correspond to the contributions of the dendritic firing action potential from the distal and proximal located sources as well as spatial summation. For example, a higher proportion of distal firing may reflect a shift to the higher proportion of reactivated features whereas a high proportion of proximally sourced action potential may represent the real-time novel input.

   This balance of novel information and reactivated information elicited through the different contributions of distal and proximal located spine activity may be part of the explanation of how the hippocampus undertakes its varying cognitive functions relying on memory. For example in object recognition there is a requirement for real-time input of the object presented and memory reactivation in order to identify it. The hippocampus, as part of the firing pattern of entorhinal cortex to hippocampus following on to fornix, mammalian bodies and anterior thalamus, plays the role of comparator capable of individualizing representations of overlapping units of familiar versus novel events (Zeamer, Jeneson, Smith).  In the case of emotionally charged events, the incoming information is judged in accordance to past experiences and therefore, forward sweep gamma firing may be supplemented by distal firing representing the emotional component. In anxiety, there is neural oscillation connectivity between the hippocampus and medial prefrontal cortex (Khmeka) with theta brain wave modulating the level of threat probability, but not its magnitude indicating a fear memory effect where incoming information is processed along with reactivated memory information (Khmeka). In the case of spatial memory, LTP of the hippocampus plays an important role with impaired AMPAR subunits leading to defective spatial working memory task performance (Schmidt) and LTP deficits (Rowland).

   One possible explanation as to the function of specific localization of dendritic firing contributions relating to memory can be ruled out however and that is that dendritic firing of the CA1 is unlikely to be related to neurogenesis that is required for memory storage. Memory formation is known to require newly formed cells ((Bischofburger, Kee, Toni), but this is likely to involve the DG area and the CA3. Newly formed synapses between the two can compete with and displace the established synapses demonstrating plasticity. Therefore, the neurogenesis destabilizes the past memory reactivation and shifts the dominance to the new input (Martinez). Therefore, the dendritic firing of the CA1 would be indirectly involved, but neurogenesis of the CA1 is not required.

   However, the dendritic firing of the CA1 could play a role in memory consolidation undertaken during sleep. The reactivations of hippocampal place cell assemblies during sleep (De Lavilion) lead to a shift from the hippocampus to cortex memories according to Oswald`s sleep theory (Mehta, Ji) as well as memory consolidation (Maingret) and better memory performance (Mednick). During the NREM stages there is functional global spike-based connectivity between the hippocampus and cortex whereas in REM there are increases at the neuronal level with increases in cAMP, MAPK and CREB phosphorylation (Luo). The hippocampus demonstrates sharp-wave/ripple complexes which are short periods of increased activity with high frequency oscillations (Maingret, Jahnke). These occur simultaneously with spike based firing representing the replay of previous experiences (Jahnke). Both are the product of dendritic firing (Jahnke) and sodium spikes in the CA1 and CA3. The hippocampus also demonstrates theta oscillations and spindle firing (Sullivan) with hippocampal sub-regions demonstrating more synchronized firing in the latter than the former. The spindle firing of the CA1 is phase locked to the medial entorhinal cortex in sleep stages 2 and 3 (Sullivan) and this modulates the strength of gamma oscillations. It is the sleep spindles that are linked to the transfer of local information to the cortex and memory consolidation (Staresina).

   Therefore, dendritic firing of the hippocampus is a requirement for neurotransmitter based signal transmission. The flexibility in sources of firing whether proximal via forward signal transmission or distal means that spatial summation allows only strong signals to be propogated further. Strong depolarisation can come from firing of multiple spines located either proximally and distally or distally alone, but only if clustered. This flexibility from an information point of view means that only dominant features continue or a wider range of strong features are bound together so that the event representation may be made up of a number of different features, eg. object characteristics, location, timing even emotional value. The adaptive response of the CA1 to action potential and post-synaptic firing of a repetitive or continued stimulatory nature involves the strengthening of the post-synaptic physiological area to future firing. The result is a priority for forward signaling and dendritic firing. Although the post-synaptic physiological changes often dominate research, dendritic firing should also be considered, not as a secondary feature, but as being a factor responsible for the continuation and modulation of forward signal transmission.  

Since we`re talking about the topic……………………

                ….since ageing is known to be associated with deficits in spatial memory in mice, can we assume that if Magó and team`s experiments were repeated, but with the CA1 pyramidal cells of 24 month or 29 month old mice (aged) we may see a difference in dendrite firing and even a change in the contributions of the distal and proximal located spines?

                ….adult rats undergoing locomotive stress (ie. elevated platform) exhibit an increase in gamma burst activity in the hippocampus of whereas the administration of diazepam decreases the gamma burst activity due to its inhibiting interaction with GABA receptors Takillah). If Magó and team`s experiments were repeated with tissues from 8month old rats, can we assume that the level of dendritic firing will be affected in general by the administration of diazepam?

                ….sodium ion channel activity is associated with dendritic firing (Magó). Would the administration of tetradoxin, a known sodium channel inhibitor, confirm the association between sodium ion entry and the evoking of d-spikes? Would the administration of specific inhibitors of the R-type calcium ion channels and T-type calcium ion channels both linked to lower voltage evoked calcium influx confirm that calcium ion transport is not required for d-spike activity in the same way as the higher voltage VGCCs?

Posted in hippocampus, LTP, neuronal firing, Uncategorized | Tagged , , | Comments Off on synaptic plasticity of proximal and distal dendrites of CA1 pyramidal neurons

near-death experience memories compared to flashbulb memories

Posted comment on ´Near-death experience memories include more episodic components than flashbulb memories` written by H. Cassol, E.A.C. Bonin, C. Bastin, N. Puttaert, V. Charland-Verville, S. Laureys and C. Martial  and published in Frontiers in Psychology 13th May 2020 doi 10.3389/fpsyg.2020.0088

SUMMARY

   Cassol and colleagues investigated whether memories of near-death experiences (NDEs) are comparable to aspects of flashbulb memories (FBs). In their investigation they identified and compared episodic and non-episodic information, phenomenological characteristics and the centrality of memories of participant`s verbal recollections of their NDE, flashbulb and control autobiographical memories.

  The experiment was set up so that the 25 participants who had lived through a life-threatening situation (eg. anoxia, trauma) and who had met the accepted criteria of having experienced a NDE, first took part in a screening task.  This was to detect and exclude memories biases relating to negative mood (according to Positive and Negative Affect Schedule) and mild cognitive impairments (Montreal Cognitive Assessment). The participants then performed the Autobiographical Memory Interview (AMI).  Each participant was asked to describe in detail three target memories that occurred in the same time period of their life. These were: the memory of their NDE; a flashbulb memory (these followed particular criteria such as that they were either very surprising and consequential, or emotionally arousing, eg. Man`s first steps on the moon, September 11th terrorist attacks); and an autobiographical memory. The interviewers were allowed to probe when required to encourage a full description of the events. The narratives were recorded, transcribed and analysed using an established manual scoring procedure so that the episodic details could be separated from the non-episodic details. Episodic details (termed ´internal details`) related to descriptions of the events such as the event details describing the unfolding of the story (eg. listed as the happenings, persons involved, reactions/emotions of oneself and other people, one`s clothing and the weather), the time (eg. life epoch as well as defined physical time) and place plus the sensory or mental state details (eg. thoughts and emotions) specific for the event. The non-episodic details (termed the ´external details` related to semantic or factual information not specific for the event). Based on the final classification of reliability between scores from independent scorers, the details generated were summed across the memory types.  The data of the AMI was analysed for each type of memory using appropriate software and statistical methods and significance was defined.

  After each memory recall the participants were given a short version of the Memory Characteristics Questionnaire (sMCQ – 16 items) and the Centrality of Event Scale (CES – 20 items) to fill in. The sMCQ assessed memory clarity, sensory details, self-referential and emotional information, reactivation frequency and confidence in memory on a Likert scale of 1 to 7. Data analyses were carried out using appropriate statistical software. The CES assessed how central the event was to the person`s identity and life story using agreement (scale 1-5) to the following statements of the event as: considered a reference point for generation of expectations to other events in life story; considered as a central component of a person`s identity; or considered as a turning point in the person`s life story. Both sets of results were analysed using appropriate statistical methods.

   The results of Cassol and team`s experiments showed that the level of negative mood was no higher for the NDE group than the control and was not related to the amount of internal or external details of the NDE, flashbulb and autobiographical memories. The age of the interviewee and the time of the interview in relation to the NDE event also did not correlate to the amount of internal or external details of the three memory types. However, the overall amount of details given for the NDE memory was higher than for both the flashbulb and autographical memories which were the same.

   The results also showed a significant interaction between the type of detail and the type of memory. The number of internal details reported for NDE memories was higher than the internal details reported for flashbulb and autographical memories. The number of internal details was also higher than the amount of external details reported for the NDE and again, this level was higher than for both the flashbulb and autobiographical memories.

   With regards to the results of the sMCQs, Cassol and colleagues found that NDE memories exhibited higher, but not statistically significant, ratings for internal perceptual details relating to emotional status compared to the autobiographical and flashbulb memories. They found that there was a memory difference in the feeling of mentally reliving the event and the sensation of feeling the emotions felt during the event while remembering. This feeling was higher in both NDE and autobiographical memories than with the flashbulb memories. The authors also reported a memory difference in the visual perspective taken while remembering. The NDE and autobiographical memories did not demonstrate statistical differences to each other, but they did show a difference to flashbulb memories. The results indicated that NDE and autobiographical memories were more likely to be remembered from the first person perspective. The emotions felt at the time of the event, ie. the valence, also produced a memory difference. Again, flashbulb memories gave lower scores than NDE and autobiographical memories and because NDE and autobiographical memories did not differ, this indicated to the authors that the latter were more positive in average. Two further observations were made that indicated a trend (but were not statistically significant) towards higher scores for NDE memories and autobiographical memories than flashbulb memories and these were: personal importance attached to the event; and frequency of sharing and reactivation. Scores also not reaching significance for the memory types were those relating to the amount of sensory details, clarity (time location and coherence), confidence in memory and the ability of the individuals to remember their own actions/words/thoughts.

   Cassol and colleagues found more significant differences with the results of their Centrality of Event Scale studies, ie. how central an event was to a person`s identity and life story. They found that the mean score for the NDE memory was significantly higher than for the autobiographical memory which was higher than that for the flashbulb memory. They also found a significant positive correlation between CES scores and a number of the reported internal details given above (eg. the feeling of re-experiencing the event, the feeling of re-experiencing the emotions felt at the time of the event , the importance of the event and the level of sharing of the event and reactivation frequency). However, there was no correlation between the CES and the external details. This indicated that the NDE memories are the most central memories to identity, more than the control autobiographical memories and with flashbulb memories the least.

   Cassol and colleagues concluded their article with a discussion of their findings. They found that NDE memories were not the same as flashbulb memories and presented explanations as to why this might be. The authors suggested several reasons why there is a higher overall amount of details and higher amount of internal/episodic details in the NDE memories compared to the autobiographical memories and flashbulb memories. The first reason related to the association between NDE and the unusual perceptions of the individuals at the time such as leaving the physical body, being in an unknown spatial-temporal dimension and the intense sensation of ´reality` reported with them. This, the authors suggested could lead to the vivid memories of the NDE in comparison to flashbulb memories.

   The second reason given suggested that the difference lies with the ´weakness` of flashbulb memories. The authors explained that flashbulb memories were more subject to forgetting in comparison to other emotionally arousing events. The severity of the emotions felt and the degree of involvement would maintain consistency and amplify the memories rather than for the flashbulb memories where the participants were not normally personally involved. This was supported by the strength of the NDE memories in terms of details. Cassol and colleagues found that the NDE narratives became richer in time compared to flashbulb memories which they said declined. They found that certain subtypes of recalled details were increased, eg. the event details describing the unfolding of the story and the perceptual details (eg. sensory and spatial-temporal information). This was attributed to the self-referential nature of the information which would have a positive effect on encoding, processing and storage of the memory.

   The authors also linked the self-referential nature with the differences relating to emotions and importance. They explained that the strength of the NDEs because of the event being highly emotional and the emotional relevance of autobiographical memories meant that the phenomenological characteristics for the NDEs and the autobiographical memories were higher than those for the flashbulb memories. The levels for autobiographical memories which were on a par with the NDEs were explained as due to the autobiographical memories being temporally linked in time in some way to the NDE which would strengthen them both. Again the self-referential aspect of both the NDEs and autobiographical memories would mean that they would be shared more often and communicated more often than the flashbulb memories and so would be strengthened because of it.

   With regards to the perspective taken during recollection, the authors also used the self-referential nature of the NDEs as being reflected by the individual using the first person compared to the third person when recalling the flashbulb memory. The autobiographical memories were shown as being recalled either in the first or third person perspective. Cassol and colleagues suggested that the use of the third person perspective for the NDE or autobiographical memories could be an avoidance strategy to set up distancing between the individual and the memory and hence, emotional intensity would be reduced. Therefore, the difference in the adopted visual perspective could be due to the emotions felt at the time of the event (eg. valence) that are overall very positive in NDE and autobiographical, but more negative in flashbulb. Cassol and colleagues also gave the results of the CES and centrality tests as support for the NDEs having a higher impact on the individual`s life story than the flashbulb memory.

   Therefore, Cassol and team`s experiments show that NDE memories have different characteristics to flashbulb memories with richer internal details and a propensity to self-reference and importance. They concluded their article with a few suggestions as to how to extend their experimentation such as increasing the number of subjects who had experienced NDEs, controlling for false memories and using autobiographical and flashbulb memories from different time periods.

COMMENT

   What makes this article interesting is that it describes two of the more unusual types of long-term memory: that of near-death experience memory (NDE) and flashbulb memory (FB). Both of these relate to memories formed of events experienced at particular points in an individual`s life and whereas the former is rare (hence, difficult to find experimental subjects) and extremely personal, the other is more common and relates to the individual and his/her social environment.  The NDE is an experienced situation of intense physical or emotional danger where some individuals report a dissociative consciousness state, vivid extraordinary perceptions (eg. out-of-body – OBE), or an intense feeling of peacefulness and calm. (This is similar to, but not the same as, near-death-like experiences where there are similar phenomenological experiences, but without the physical and/or emotional danger. NDE-like experiences are sometimes associated with meditation, or depression.) In comparison, flashbulb memories are experiences where the individual is given a ´piece of news`. This may be consequential to the individual, but it is unlikely that the individual is in physical danger at that time. Therefore, although the memories are autobiographical and relate to the individual at specific time points in their life story they are formed in different circumstances. Therefore, this comment looks at the similarities and differences between the two with reference to Cassol and team`s findings and looks at which neurochemical mechanisms may promote NDEs being formed.  

   The first difference that springs to mind between the two types of experience and memory from a neurochemical point of view is the level of acute stress and stress response experienced at the time of the event. The intense physical or emotional danger experienced by the NDEers (individuals who report a NDE) is likely to promote a physiological  ´defence` response in order to secure survival. If we look at Schauer and Elbert`s sequence of stages in response to traumatic stress (ie. Stages 1-6 -freeze, flight, fight, fright, flag, faint) the NDE is likely to begin to occur at the end of Stage 3 (the fight phase) when sympathetic activation has mobilised resources, but since the ´fight` is not successful the individual slips into the next stage, Stage 4, fright. Sympathetic activation brings about a range of physiological signs such as dizziness, light-headedness, palpitations, numbing and important for this topic, feelings of non-reality. Stage 4, although associated with unresponsive immobility but tachycardia and hypertension, consolidates the change in mental approach to the danger situation with signs of hyper-alertness, high emotional arousal and fear. This continues into the ´flag` phase where there is a physiological ´shut-down` instigated via the activation of the parasympathetic system and symptoms such as bradycardia, vasodilation and hypotension and relating to the NDE, a drop in arousal, feeling of surrender, cognitive failure and ´numbing` of all emotions (perhaps, the feeling of ´inner calm` experienced). Therefore, it appears that the reported experiences of the NDE event (eg. dissociative consciousness and reported increases in awareness leading to inner calm and peace) can be explained by the staged physiological responses to stress according to Schauer and Elbert . In comparison, flashbulb memories (FB) are associated with a minimal or temporarily raised stress level since the individual is not in physical danger.   

  The different stages of the stress response induce changes in the brain neurochemistry and these we assume bring about the cognitive changes observed with the NDE and highly emotional situations. These neurochemical changes are linked to cortisol (or glucocorticoid) production, which is known to cause effects in some cognitive mechanisms. Cortisol is eventually produced in response to the immediate threat to health. This manifests stimulation of a number of different routes, eg. vagus nerve to medulla to periventricular nucleus of the hypothalamus; somatosensory stressors via the tegmentum and the reticular formation to the paraventricular nucleus of the hypothalamus; painful stressors via the periacqueductal gray PAC to the paraventricular nucleus; locus coeruleus and changes to heart rate and blood pressure leading to increased release of noradrenaline leading to paraventricular nucleus; emotional stressors mediated via the raphe nucleus leading to the paraventricular nucleus; and finally and more importantly linked to cognitive function stimulation of the hippocampus, septum and amygdala. This stress pathway is involved in changes to emotional and cognitive input. The amygdala releases corticotrophin-releasing hormone (CRH) which activates the autonomic and endocrine systems which mobilise the energy for the fight or flight response.

   The next stage of the neurochemical response to stress is the activation of the sympathetic adrenomedullary axis (SAM) and more importantly for cognitive functions, the hypothalamic-pituitary adrenocortical axis (HPA). The SAM route involves the release of adrenaline, noradrenaline and stress hormones and increases arousal and vigilance. The HPA route involves actions of the hypothalamus via the paraventricular nucleus. This when activated leads to the release of corticotrophin releasing factor which enters the anterior pituitary gland where it binds to CRh R1 receptors and stimulates the production of adrenocorticotropic hormone (ACTH). ACTH stimulates the cortex of the adrenal gland leading to the synthesis and release of glucocorticoids, which spread through the body causing a number of different effects (eg. increased availability of blood glucose, decreased immune system and in the brain, inhibition of eating, sexual behaviour and growth). The main glucocorticoid in humans is cortisol and it binds to receptors found in the cytoplasm of many neurons. The activated receptor travels to the nucleus where it can stimulate gene transcription and appropriate protein synthesis. One role of cortisol is the increased influx of calcium ions via increased activation of voltage gated channels (shown by administration of GC R antagonist RU38486 – Karst). It is the influence on calcium ion concentrations on many neurochemical mechanisms within the neuron which cause the changes to neuronal functioning and ultimately can lead to cognitive effects associated with NDE. Therefore, whereas individuals that are experiencing NDE may have a physical stress response and release cortisol because of the threat to life which then has an effect on many cognitive functions, individuals that have FB memories are not under threat and therefore, the effects on their cognitive functions occur by different means.

   The first cognitive effect associated with NDE which may be linked to cortisol production and neuronal calcium ion effects is the informational content of the memories formed at the time of the event. Cassol and team found in their experiments that the amounts of information (total, external and internal) in NDE were greater than FB memories with both demonstrating a high level of context information not greatly relevant to the situation at the time. It is likely that this increase in informational content comes about by a change in attention, which is known to be a controlling factor of the quality and quantity of sensory information and this may be induced by the cortisol released as part of the response to the threat situation. Cortisol has been reported to increase attention (Banks), but what does this mean to the information being inputted and stored? We know that the fear attention state (and the corresponding fear emotional state since they are interrelated) can increase the quantity of informational input, but decrease its quality (ie. more gist rather than high level detail). This is possibly due to a change in the balance of task-relevant and task-irrelevant material where the latter under normal circumstances would not be learnt, but in NDE are included in a significant amount. This is also a possible explanation for FB memories too, since these memories are also known to include not only task-relevant information, but also a host of irrelevant information such as trivial sensory details (Brown and Kulik).

   Therefore, in both cases central and task relevant information is given the same weighting as irrelevant, non-central, peripheral information and this change in balance is likely to be achieved through alterations to the attentional system. For example, Cukor showed that attention dynamically alters visual representation to optimize the processing of behaviorally relevant objects during natural vision. The likely brain areas involved in such as shift would be: the intraparietal sulcus, precuneus and dorsolateral prefrontal cortex shown to be involved in bilateral attentional control in selecting the extent of relevancy and updating the representation (Niv); the anterior thalamus  which appears to direct attention to task-relevant stimuli that will bring reward (Wright); the amygdala involved in the fear state; and the hippocampus which is required for sustained activation of the relevant neurons without repetition and is important for memory formation. There is also likely to be an involvement of the NMDA receptors since the NMDA R antagonist, ketamine is reported to strongly impair the ability to ignore irrelevant task information (Stoets).

   The shift in informational content of the NDE and FB memories is also likely to include as well a shift in the balance of attended to unattended information (information of which the individual is aware to information that the individual has no awareness of). This is suggested with NDE memories since they are described by NDErs as being ´realer than real`.Such a shift could be explained by a widening of the limits of ´awareness` since the experience not only includes task relevant and attended information, but also, task irrelevant and what would normally be, unattended information. Hence, the experience would be described as ´realer than real`.

   The original models of Cherry, Broadbent etc. relating to levels of attended and unattended information describe them in terms of perception and processing. This leads onto factors affecting the balance and characteristics with for example, decreased processing of unattended information (Wojceulik) and greater processing of unattended information when the perceptual load is low (Lavie). Therefore, even if there is no awareness of unattended information there can still be processing (Wright, Nee and Norman). The situation with NDE is that this level of information would not normally be consciously experienced or be the first information recalled.  Therefore, the NDE condition changes the balance so that more information is attended. The question is how can this be achieved?

   One possible suggestion is that the stress response causes changes not only in the amount of information inputted as described above, but how the information is grouped or bound together. This has an effect on the capacity of working memory, which we know is limited to a certain number of event characteristics. Perception and Gestalt theories group visual elements such as lines, curves together into unitary objects such as forms and shapes which are then processed by working memory. By increasing the ´group size`, processing capability can therefore, be ´stretched` to more elements. We know of this through memory improvement techniques where individuals instead of trying to remember 6 different numbers group them together to represent years or months for example to increase recall efficiency. In the case of NDE experiences, the different sensory input may be grouped together as a single ´unit`. This has been shown in a similar manner by Jang who described event representation as containing not only the prime features of the event, but also more abstract, more internal features of the event such as attentional control states and category information. Such grouping was said to require the activation of the hippocampus and putamen as well as other regions representing visual feature-selective event information (primarily visual cortex), category-selective information (posterior frontal cortex) and control demand-selective event information (insula, caudate, anterior cingulate, and parietal cortex) event information.Therefore, brain areas affected by the stress response such as the hippocampus would, like the other attentional system brain areas, lead not only to changes in informational quality and quantity, but also awareness and processing capability. This would explain both the NDE event content and the content of FBs where both task relevant and task irrelevant information form part of the event representation.

   Although the stress response may then explain the change in conscious awareness relating to information and informational processing, can it also explain the dissociative consciousness state reported with NDE experiences? The mechanics of conscious awareness observed for visual input 170msec after presentation (Thierry) involves synchronized activation of many areas including the DMN, insular cortex, anterior cingulate cortex, medial prefrontal cortex and dorsal thalamic nuclei. However, conscious awareness under normal circumstances does not involve some of the transcendental, out-of-body type awareness reported by NDErs. In this case there is a dissociation from ´reality`, ie. the real-time sensory input information achieved via eyes, ears etc. One explanation for this could be the near-complete separation of top-down and bottom-up processing leading so that the top-down ´I, SELF` thinking part is distanced from the ´sensing, input` bottom-up part. This idea may be acceptable since imagining and dreaming are both examples where top-down processing and bottom-up input are ´distinct`. In the case of NDE, however the separation is likely to be achieved by the attentional system, which we have already seen is under the influence of cortisol released in the stress response. Top-down attentional modulation appears to suppress the incoming visual information via inhibition of firing at the V1 level (Jacob) and dissociation of attention and awareness is observed in the case of failure of applied chromatic flicker to reach awareness, but not alerting and orienting effects – also associated with V1 visual area effects (Lu). This separation is also apparent through the conscious awareness of an isolated object (or gist of a scene) in the near-absence of top-down attention (van Boxtel). (However, in this particular research example it appears that the explanation is against the proposed one for NDE since attention is described as top-down and ´thought` and consciousness is the information coming in and bottom-up. Possibly the way to approach the results of this study is to view conscious awareness as top-down and subject to top-down attention whereas sensory information is bottom up and subject to bottom-up attention with attentional modulation being the ´tool` and not the ´result`.)

   Therefore, one possible explanation for the more transcendental aspects of the dissociative consciousness reported by NDErs is the separation of top-down processing from bottom-up processing. The advantage of such a separation is protection of the higher brain areas from excessive information and high processing load. This may also be attributed to the emotional state at the time of the experience which is interrelated to attentional system functioning (Gregory). Cassol and team reported that the emotional stateduring the NDE experience was greater than for the autobiographical memory and both were greater than that recorded for the FB experience and it is clear that NDE are profound psychological events with highly emotional and self-related content. The increase in information is associated with this highly emotional state since it has been found that the link between episodic memory,  the feeling of ´being there` and the higher emotional experience is linked to better factual memory via an improved attentional focus on the stimuli, hence increasing memory encoding (Makowski). The same is reported with FB memories which also carry factual memory with strong emotions that can be recalled years later after the initial event (Brown, Finkenauer).

   The increased fear emotional state as the individual experiences the near-death event in Schauer and Elbert`s Stages 3 (fight) and 4 (fright) of the stress response is responsible for the increased levels of relevant and irrelevant material for the task (Newenhause, Stoets) and requires the activation of brain areas such as prefrontal cortex, amygdala, cingulate cortex and thalamus. Anxiety is shown to produce elevated cortisol levels linked to the hyperactivity of the HPA axis (Lenze) and there is a direct ventral connection between the amygdala and prefrontal cortex (Eden), both affected by anxiety.

   However, as the stress response progresses this fear emotional state shifts to a period of ´calm`. This relates to Schauer and Elbert`s Stage 5, ´flag` phase where there is a physiological ´shut-down` instigated via the activation of the parasympathetic system with symptoms such as drop in blood pressure and heart rate, but more importantly to the emotional experience of the NDE, a drop in arousal, a feeling of surrender, cognitive failure and a ´numbing` of all emotions, ie. the reported ´calm`. This can be brought about by the reduction in cortisol production and hence, cortisol effects on the brain areas involved in instigating the emotional state (eg. the Papez-Mclean limbic model involving activity of the amygdala, hippocampus and cingulate cortex and the two track Le-Doux theory with rapid emotional response based on information from the thalamus to the amygdala leading to autonomic and endocrine changes interpreted by the cortex and the slower, direct transmission of information from the thalamus to the cortex).  Reports show that the strength of the amygdala-hippocampus connectivity indicative of fear is enhanced with cortisol release (Hakamata) and cortisol production will re-instate fear after its extinction by activating this pathway (Kinner) and hence, reduced cortisol production will instigate a reduction in the emotional experience of fear. Also there is a possible switch off via the ventromedial prefrontal cortex in Schauer and Elbert`s Stages 5-6 since when stress is deemed controllable (ie. through the numbing of the emotions) then the stress induced activation of the dorsal raphe nucleus is inhibited by the ventromedial prefrontal cortex and the behavioural effects of stress are blocked because of the reduction in cortisol production (Amat). Therefore, the Stage 5, ´flag` stage is achieved and a reassessment of ´value` of the event occurs, eg. ´fear, threat` to ´calm, acceptance`.  

   What all of the Cassol`s observations and the above explanations show is that the functioning status of the brain during the NDE forming experience is not damaged to the extent that normal sensory input, processing and memory formation mechanisms do not occur. Therefore, there is support for the explanation of NDE from a psychological perspective of the ´biological/psychological` framework (Braithwaite). This framework supposes as supported here from looking at the memory characteristics and mechanism that the NDE memories are a consequence of the brain`s functional changes that can occur as a response to a perceived threat of death. Whether they form when death occurs for obvious reasons cannot be determined. Therefore, the sensory experience, memory characteristics and mechanism can be explained by normal cognitive physiology and functioning and even the OBEs and other more transcendental experiences can also be given explanations from normal neurochemical principles, eg. tunnels and lights caused by the disinhibition of the visual cortex and positive emotions and lack of pain attributed to the action of endorphins.

   Up to this point, we have only discussed the NDE from the perspective of the experience at the time of the event, but for both the NDE and FM, recall is important. Recall reflects the memory formed at the time and the level of forgetting between the event and the retrieval stage and since NDE and FM can be recalled years later this could have a significant effect if these memories follow the normal trajectories of other memory types. We have to assume that the actual physiological mechanisms for memory formation and storage are the same whether for NDE, FM or autobiographical memories, even if levels vary. This has been shown for the case of acute stress and cortisol which is relevant to the topic discussed here for the NDE and FM memories. Cortisol has been shown to be linked to memory changes due to its action on the hippocampus, an important area in the formation and recall of memories (Osborne). Acute stress (and even much later after the stress event) has been shown to cause rapid release of norepinephrine and glucocorticoids (cortisol) which bind to the hippocampal receptors. The resulting change in function of the hippocampus leads to the promotion of memory formation and consolidation, but appears to inhibit memory retrieval(Wais). This has been expanded in the case of cortisol increase by showing that the decreased memory level being observed is restricted to specific task relevant information whereas background memory is not affected (Steinmetz). This is contrary to what Cassol and team found where both internal and external information are remembered to a greater extent in NDE memories, but may support the ´change in grouping` hypothesis given above. 

   The demands for an ´undamaged` brain with NDE is reinforced by the need for connectivity of brain areas during the recall process. It is found that recall requires connectivity between the frontoparietal network, the default mode network and the hippocampus (Xiao, Westphal). Connectivity is demonstrated by brain wave activity and an increase in alpha brain wave activity is associated with NDE memory recall which is indicative of attentional control (van Schouwenburg). However, gamma waves normally associated with hippocampus activity during retrieval is not seen in NDE retrieval, but instead theta wave activity is observed which is linked to recall of emotional episodic events and the required hippocampal activity (Palmieri) and an increase in delta wave activity also associated with recall and personal feelings of reward (Knyazev). Therefore, the retrieval mechanism appears to be the same for the different memory types independent of quality or quantity of information recalled.

   Cassol and team however, expanded their exploration of the NDE memory retrieval by looking at the effect of emotional state during the recall session. We have already given an explanation as to how the actual quality and quantity of the memory formed was increased by the stress experienced by the individual during the NDE event, but Cassol and team looked into whether or not inducing the same contextual and emotional states during the recall as experienced during the event itselfwould strengthen the recall and reactivation of the NDE memory.  This exploration was indicated because it has been found that memory retrieval is increased when there is a match between the consciousness state at encoding and the consciousness state at retrieval (state-dependent memory – Overton). Researchers have also found that NDE memories are better retrieved when delta wave connectivity is greater because this appears to be the dominate wave pattern during the encoding NDE experience phase (Hartman). (Delta wave activity is also associated with transcendental type experiences (Hartman) which would correlate with the OBE and other dissociative consciousness events linked to the NDE). This appears to be supported by Cassol and team`s results since they found that the NDEr experienced the sensation of feeling the emotions felt during the event while they recalled the NDE and these feelings were greater for the NDE and autobiographical memories than for the FB memories. The retrieval of NDE induces the same emotional states of fear, ´realer than real` emotionality, but then calm, positive emotions.  Therefore, whereas fear can lead to memory impairment normally this is counteracted by the positive, calming emotional response which would lead to normalised recall (McKenzie). The expression of emotion during recall was found to aid both central and peripheral information retrieval (Davidson). In the case of the FB memories, which are more than just emotional memories,  emotions can help or impair recall just like with the NDE, eg.  people experiencing high levels of stress and distress show significantly higher levels of recall of context information than individuals with lower levels of stress after a period of 3 years (Bohannon). The emotional reactions however, to the event may differ in that the reaction could remain over the long-term, be short lived, or only appear after the event’s impact is fully appreciated by the individual in reference to his own consequentiality.

   This question of consequentiality and relevance to SELF and self-identity brings to light another factor that may or may not strengthen NDE memories and FB memories through recall. NDE experiences and memories are considered very important to the individual`s life story, greater than for autobiographical memories, which are more important than FB memories. These findings are understandable since NDEs are from living through a near-death, extremely personal and significant life threatening event and so should be more important than the other two memory types, but FB memories are also important to the individual otherwise why would the individual record it.

  The relationship between NDE memory and the SELF is that the NDE memory is characterised by the rich phenomenology recorded at the time which we have already described as being due to the stress response, increase in cortisol production and increased attentional and emotional systems affecting central and peripheral information input and storage. This richness of the NDE memory is associated with the SELF and self-identity as observed by the centrality scores (CES scores) in Cassol and team`s experiments and the effect of the NDE event may affect the SELF and self-identity for a much longer time and to a much greater degree than a FB memory. It has been reported that the prevalence of meaning attached to the NDE event may bring about changes in behavioural patterns such as for example a reduced fear of death, a greater level of compassion for others or a lower value placed on materialistic positions. Therefore, recall of the NDE memory strengthens the event details and if not changed, strengthens the relevance to the SELF and self-identity. The FB memories must also serve some role in SELF or self-identity since they are recorded for a particular reason. However, this is disputed by some who say that the FB memory does not serve personal consequentiality, but serves consequentiality for the community, hence shaping social identity rather than self-identity. However, the same event characteristics may be inputted and stored for a vast number of individuals, but the exact memory details may be different due to the incorporation of the internal, phenomenological information. Therefore, this renders only part of the memory having consequentiality for the community and all to the SELF and self-identity.

   This factor of consequentiality and association with the SELF and self-identity reinforces the personal nature of the NDE and FB memories and brings into discussion the topic of retrieval accuracy. Recall of memories can reinforce event characteristics through their reactivation and strengthening of firing of the neuronal assemblies that make up the event representation (internal and external features). Therefore, the thinking or re-telling of the experiences may change or strengthen the characteristics recorded. Cassol and team reported that in the case of the NDE that reactivation sharing and its frequency were found to be important and this is as expected. It was also found to influence the memory accuracy. The problem with judging whether the recall of NDE events is accurate is that the individual doing the recall is the only person who has experienced it and who has recorded the internal and external event characteristics. Therefore, the accuracy of the recall for at least the first reactivation cannot be verified totally by a third party. Subsequent retrievals can be judged to some extent via consistency of reporting and to this extent it has been reported that NDErs remember the event clearly and accounts are not modified over a period of two decades. In this way it appears that memory traces do not fade like other autobiographical memories and narratives remain consistent if no post-event processing, construction and embellishment takes place.

   In the case of FBs, judgement of accuracy would be deemed easier to establish since part of the event characteristics recorded are shared by a large number of people and therefore, there should be many accounts of these particular facts. However, the details may be affected by who and how events are recalled by others and reports have shown how media attention can affect the recall of these shared features. Again like NDE however, FB memory has a quantity of external peripheral facts that may not be common to all and a quantity of internal event characteristics which are entirely individual. Therefore, it would be difficult for a third person to check the accuracy of these features against the original encoding event and consistency accuracy is the only measurement possible. Researchers have found once an inconsistency emerges which usually occurs within the first year then it remains (Hirst) and like NDE events, FB memory recall may be years later after the event. Again, FB recall makes the memory subject to change via post-event processing, construction, embellishment and the influence of others.

   Therefore, this exploration of NDE and FB memories has shown that they are more examples of how emotional state can affect informational input and memory. The uniqueness of the memories with their increased level of peripheral, internal information can be attributed to the change in attentional state induced particularly in the case of NDE to a rise in cortisol production due to the perceived threat to life. Their value to memory research is via their strong association to the SELF and self-identity and this in the case of FB memories can be advantageous.

Since we`re talking about the topic……….

            …..dexamethasone is a synthetic glucocorticoid known to lead to the instigation of the stress response. Is it possible that the NDE-like responses can be induced by its use and these responses can be observed with neuroimaging and compared to those experienced by practiced meditators?

            …. If NDE-like memory can be induced by hypnosis, can neurochemical studies on cortisol production, calcium influx and neuroimaging of brain area activity confirm whether or not, cortisol is the cause of the memory changes observed?

            …. ketamine-induced anaesthesia is said to be similar to NDEs in the level of wakefulness and connectedness, could this be used to study the influence of external environmental changes such as loud noise and smells on memories if formed by experienced meditators?

Posted in attention, brain waves, consciousness, emotional system, flashbulb memories, memory recall, near-death experiences, stress | Tagged , , , , , , , | Comments Off on near-death experience memories compared to flashbulb memories

near-death experience memories compared to flashbulb memories

Posted comment on ´Near-death experience memories include more episodic components than flashbulb memories` written by H. Cassol, E.A.C. Bonin, C. Bastin, N. Puttaert, V. Charland-Verville, S. Laureys and C. Martial  and published in Frontiers in Psychology 13th May 2020 doi 10.3389/fpsyg.2020.0088

SUMMARY

   Cassol and colleagues investigated whether memories of near-death experiences (NDEs) are comparable to aspects of flashbulb memories (FBs). In their investigation they identified and compared episodic and non-episodic information, phenomenological characteristics and the centrality of memories of participant`s verbal recollections of their NDE, flashbulb and control autobiographical memories.

  The experiment was set up so that the 25 participants who had lived through a life-threatening situation (eg. anoxia, trauma) and who had met the accepted criteria of having experienced a NDE, first took part in a screening task.  This was to detect and exclude memories biases relating to negative mood (according to Positive and Negative Affect Schedule) and mild cognitive impairments (Montreal Cognitive Assessment). The participants then performed the Autobiographical Memory Interview (AMI).  Each participant was asked to describe in detail three target memories that occurred in the same time period of their life. These were: the memory of their NDE; a flashbulb memory (these followed particular criteria such as that they were either very surprising and consequential, or emotionally arousing, eg. Man`s first steps on the moon, September 11th terrorist attacks); and an autobiographical memory. The interviewers were allowed to probe when required to encourage a full description of the events. The narratives were recorded, transcribed and analysed using an established manual scoring procedure so that the episodic details could be separated from the non-episodic details. Episodic details (termed ´internal details`) related to descriptions of the events such as the event details describing the unfolding of the story (eg. listed as the happenings, persons involved, reactions/emotions of oneself and other people, one`s clothing and the weather), the time (eg. life epoch as well as defined physical time) and place plus the sensory or mental state details (eg. thoughts and emotions) specific for the event. The non-episodic details (termed the ´external details` related to semantic or factual information not specific for the event). Based on the final classification of reliability between scores from independent scorers, the details generated were summed across the memory types.  The data of the AMI was analysed for each type of memory using appropriate software and statistical methods and significance was defined.

  After each memory recall the participants were given a short version of the Memory Characteristics Questionnaire (sMCQ – 16 items) and the Centrality of Event Scale (CES – 20 items) to fill in. The sMCQ assessed memory clarity, sensory details, self-referential and emotional information, reactivation frequency and confidence in memory on a Likert scale of 1 to 7. Data analyses were carried out using appropriate statistical software. The CES assessed how central the event was to the person`s identity and life story using agreement (scale 1-5) to the following statements of the event as: considered a reference point for generation of expectations to other events in life story; considered as a central component of a person`s identity; or considered as a turning point in the person`s life story. Both sets of results were analysed using appropriate statistical methods.

   The results of Cassol and team`s experiments showed that the level of negative mood was no higher for the NDE group than the control and was not related to the amount of internal or external details of the NDE, flashbulb and autobiographical memories. The age of the interviewee and the time of the interview in relation to the NDE event also did not correlate to the amount of internal or external details of the three memory types. However, the overall amount of details given for the NDE memory was higher than for both the flashbulb and autographical memories which were the same.

   The results also showed a significant interaction between the type of detail and the type of memory. The number of internal details reported for NDE memories was higher than the internal details reported for flashbulb and autographical memories. The number of internal details was also higher than the amount of external details reported for the NDE and again, this level was higher than for both the flashbulb and autobiographical memories.

   With regards to the results of the sMCQs, Cassol and colleagues found that NDE memories exhibited higher, but not statistically significant, ratings for internal perceptual details relating to emotional status compared to the autobiographical and flashbulb memories. They found that there was a memory difference in the feeling of mentally reliving the event and the sensation of feeling the emotions felt during the event while remembering. This feeling was higher in both NDE and autobiographical memories than with the flashbulb memories. The authors also reported a memory difference in the visual perspective taken while remembering. The NDE and autobiographical memories did not demonstrate statistical differences to each other, but they did show a difference to flashbulb memories. The results indicated that NDE and autobiographical memories were more likely to be remembered from the first person perspective. The emotions felt at the time of the event, ie. the valence, also produced a memory difference. Again, flashbulb memories gave lower scores than NDE and autobiographical memories and because NDE and autobiographical memories did not differ, this indicated to the authors that the latter were more positive in average. Two further observations were made that indicated a trend (but were not statistically significant) towards higher scores for NDE memories and autobiographical memories than flashbulb memories and these were: personal importance attached to the event; and frequency of sharing and reactivation. Scores also not reaching significance for the memory types were those relating to the amount of sensory details, clarity (time location and coherence), confidence in memory and the ability of the individuals to remember their own actions/words/thoughts.

   Cassol and colleagues found more significant differences with the results of their Centrality of Event Scale studies, ie. how central an event was to a person`s identity and life story. They found that the mean score for the NDE memory was significantly higher than for the autobiographical memory which was higher than that for the flashbulb memory. They also found a significant positive correlation between CES scores and a number of the reported internal details given above (eg. the feeling of re-experiencing the event, the feeling of re-experiencing the emotions felt at the time of the event , the importance of the event and the level of sharing of the event and reactivation frequency). However, there was no correlation between the CES and the external details. This indicated that the NDE memories are the most central memories to identity, more than the control autobiographical memories and with flashbulb memories the least.

   Cassol and colleagues concluded their article with a discussion of their findings. They found that NDE memories were not the same as flashbulb memories and presented explanations as to why this might be. The authors suggested several reasons why there is a higher overall amount of details and higher amount of internal/episodic details in the NDE memories compared to the autobiographical memories and flashbulb memories. The first reason related to the association between NDE and the unusual perceptions of the individuals at the time such as leaving the physical body, being in an unknown spatial-temporal dimension and the intense sensation of ´reality` reported with them. This, the authors suggested could lead to the vivid memories of the NDE in comparison to flashbulb memories.

   The second reason given suggested that the difference lies with the ´weakness` of flashbulb memories. The authors explained that flashbulb memories were more subject to forgetting in comparison to other emotionally arousing events. The severity of the emotions felt and the degree of involvement would maintain consistency and amplify the memories rather than for the flashbulb memories where the participants were not normally personally involved. This was supported by the strength of the NDE memories in terms of details. Cassol and colleagues found that the NDE narratives became richer in time compared to flashbulb memories which they said declined. They found that certain subtypes of recalled details were increased, eg. the event details describing the unfolding of the story and the perceptual details (eg. sensory and spatial-temporal information). This was attributed to the self-referential nature of the information which would have a positive effect on encoding, processing and storage of the memory.

   The authors also linked the self-referential nature with the differences relating to emotions and importance. They explained that the strength of the NDEs because of the event being highly emotional and the emotional relevance of autobiographical memories meant that the phenomenological characteristics for the NDEs and the autobiographical memories were higher than those for the flashbulb memories. The levels for autobiographical memories which were on a par with the NDEs were explained as due to the autobiographical memories being temporally linked in time in some way to the NDE which would strengthen them both. Again the self-referential aspect of both the NDEs and autobiographical memories would mean that they would be shared more often and communicated more often than the flashbulb memories and so would be strengthened because of it.

   With regards to the perspective taken during recollection, the authors also used the self-referential nature of the NDEs as being reflected by the individual using the first person compared to the third person when recalling the flashbulb memory. The autobiographical memories were shown as being recalled either in the first or third person perspective. Cassol and colleagues suggested that the use of the third person perspective for the NDE or autobiographical memories could be an avoidance strategy to set up distancing between the individual and the memory and hence, emotional intensity would be reduced. Therefore, the difference in the adopted visual perspective could be due to the emotions felt at the time of the event (eg. valence) that are overall very positive in NDE and autobiographical, but more negative in flashbulb. Cassol and colleagues also gave the results of the CES and centrality tests as support for the NDEs having a higher impact on the individual`s life story than the flashbulb memory.

   Therefore, Cassol and team`s experiments show that NDE memories have different characteristics to flashbulb memories with richer internal details and a propensity to self-reference and importance. They concluded their article with a few suggestions as to how to extend their experimentation such as increasing the number of subjects who had experienced NDEs, controlling for false memories and using autobiographical and flashbulb memories from different time periods.

COMMENT

   What makes this article interesting is that it describes two of the more unusual types of long-term memory: that of near-death experience memory (NDE) and flashbulb memory (FB). Both of these relate to memories formed of events experienced at particular points in an individual`s life and whereas the former is rare (hence, difficult to find experimental subjects) and extremely personal, the other is more common and relates to the individual and his/her social environment.  The NDE is an experienced situation of intense physical or emotional danger where some individuals report a dissociative consciousness state, vivid extraordinary perceptions (eg. out-of-body – OBE), or an intense feeling of peacefulness and calm. (This is similar to, but not the same as, near-death-like experiences where there are similar phenomenological experiences, but without the physical and/or emotional danger. NDE-like experiences are sometimes associated with meditation, or depression.) In comparison, flashbulb memories are experiences where the individual is given a ´piece of news`. This may be consequential to the individual, but it is unlikely that the individual is in physical danger at that time. Therefore, although the memories are autobiographical and relate to the individual at specific time points in their life story they are formed in different circumstances. Therefore, this comment looks at the similarities and differences between the two with reference to Cassol and team`s findings and looks at which neurochemical mechanisms may promote NDEs being formed.  

   The first difference that springs to mind between the two types of experience and memory from a neurochemical point of view is the level of acute stress and stress response experienced at the time of the event. The intense physical or emotional danger experienced by the NDEers (individuals who report a NDE) is likely to promote a physiological  ´defence` response in order to secure survival. If we look at Schauer and Elbert`s sequence of stages in response to traumatic stress (ie. Stages 1-6 -freeze, flight, fight, fright, flag, faint) the NDE is likely to begin to occur at the end of Stage 3 (the fight phase) when sympathetic activation has mobilised resources, but since the ´fight` is not successful the individual slips into the next stage, Stage 4, fright. Sympathetic activation brings about a range of physiological signs such as dizziness, light-headedness, palpitations, numbing and important for this topic, feelings of non-reality. Stage 4, although associated with unresponsive immobility but tachycardia and hypertension, consolidates the change in mental approach to the danger situation with signs of hyper-alertness, high emotional arousal and fear. This continues into the ´flag` phase where there is a physiological ´shut-down` instigated via the activation of the parasympathetic system and symptoms such as bradycardia, vasodilation and hypotension and relating to the NDE, a drop in arousal, feeling of surrender, cognitive failure and ´numbing` of all emotions (perhaps, the feeling of ´inner calm` experienced). Therefore, it appears that the reported experiences of the NDE event (eg. dissociative consciousness and reported increases in awareness leading to inner calm and peace) can be explained by the staged physiological responses to stress according to Schauer and Elbert . In comparison, flashbulb memories (FB) are associated with a minimal or temporarily raised stress level since the individual is not in physical danger.   

  The different stages of the stress response induce changes in the brain neurochemistry and these we assume bring about the cognitive changes observed with the NDE and highly emotional situations. These neurochemical changes are linked to cortisol (or glucocorticoid) production, which is known to cause effects in some cognitive mechanisms. Cortisol is eventually produced in response to the immediate threat to health. This manifests stimulation of a number of different routes, eg. vagus nerve to medulla to periventricular nucleus of the hypothalamus; somatosensory stressors via the tegmentum and the reticular formation to the paraventricular nucleus of the hypothalamus; painful stressors via the periacqueductal gray PAC to the paraventricular nucleus; locus coeruleus and changes to heart rate and blood pressure leading to increased release of noradrenaline leading to paraventricular nucleus; emotional stressors mediated via the raphe nucleus leading to the paraventricular nucleus; and finally and more importantly linked to cognitive function stimulation of the hippocampus, septum and amygdala. This stress pathway is involved in changes to emotional and cognitive input. The amygdala releases corticotrophin-releasing hormone (CRH) which activates the autonomic and endocrine systems which mobilise the energy for the fight or flight response.

   The next stage of the neurochemical response to stress is the activation of the sympathetic adrenomedullary axis (SAM) and more importantly for cognitive functions, the hypothalamic-pituitary adrenocortical axis (HPA). The SAM route involves the release of adrenaline, noradrenaline and stress hormones and increases arousal and vigilance. The HPA route involves actions of the hypothalamus via the paraventricular nucleus. This when activated leads to the release of corticotrophin releasing factor which enters the anterior pituitary gland where it binds to CRh R1 receptors and stimulates the production of adrenocorticotropic hormone (ACTH). ACTH stimulates the cortex of the adrenal gland leading to the synthesis and release of glucocorticoids, which spread through the body causing a number of different effects (eg. increased availability of blood glucose, decreased immune system and in the brain, inhibition of eating, sexual behaviour and growth). The main glucocorticoid in humans is cortisol and it binds to receptors found in the cytoplasm of many neurons. The activated receptor travels to the nucleus where it can stimulate gene transcription and appropriate protein synthesis. One role of cortisol is the increased influx of calcium ions via increased activation of voltage gated channels (shown by administration of GC R antagonist RU38486 – Karst). It is the influence on calcium ion concentrations on many neurochemical mechanisms within the neuron which cause the changes to neuronal functioning and ultimately can lead to cognitive effects associated with NDE. Therefore, whereas individuals that are experiencing NDE may have a physical stress response and release cortisol because of the threat to life which then has an effect on many cognitive functions, individuals that have FB memories are not under threat and therefore, the effects on their cognitive functions occur by different means.

   The first cognitive effect associated with NDE which may be linked to cortisol production and neuronal calcium ion effects is the informational content of the memories formed at the time of the event. Cassol and team found in their experiments that the amounts of information (total, external and internal) in NDE were greater than FB memories with both demonstrating a high level of context information not greatly relevant to the situation at the time. It is likely that this increase in informational content comes about by a change in attention, which is known to be a controlling factor of the quality and quantity of sensory information and this may be induced by the cortisol released as part of the response to the threat situation. Cortisol has been reported to increase attention (Banks), but what does this mean to the information being inputted and stored? We know that the fear attention state (and the corresponding fear emotional state since they are interrelated) can increase the quantity of informational input, but decrease its quality (ie. more gist rather than high level detail). This is possibly due to a change in the balance of task-relevant and task-irrelevant material where the latter under normal circumstances would not be learnt, but in NDE are included in a significant amount. This is also a possible explanation for FB memories too, since these memories are also known to include not only task-relevant information, but also a host of irrelevant information such as trivial sensory details (Brown and Kulik).

   Therefore, in both cases central and task relevant information is given the same weighting as irrelevant, non-central, peripheral information and this change in balance is likely to be achieved through alterations to the attentional system. For example, Cukor showed that attention dynamically alters visual representation to optimize the processing of behaviorally relevant objects during natural vision. The likely brain areas involved in such as shift would be: the intraparietal sulcus, precuneus and dorsolateral prefrontal cortex shown to be involved in bilateral attentional control in selecting the extent of relevancy and updating the representation (Niv); the anterior thalamus  which appears to direct attention to task-relevant stimuli that will bring reward (Wright); the amygdala involved in the fear state; and the hippocampus which is required for sustained activation of the relevant neurons without repetition and is important for memory formation. There is also likely to be an involvement of the NMDA receptors since the NMDA R antagonist, ketamine is reported to strongly impair the ability to ignore irrelevant task information (Stoets).

   The shift in informational content of the NDE and FB memories is also likely to include as well a shift in the balance of attended to unattended information (information of which the individual is aware to information that the individual has no awareness of). This is suggested with NDE memories since they are described by NDErs as being ´realer than real`.Such a shift could be explained by a widening of the limits of ´awareness` since the experience not only includes task relevant and attended information, but also, task irrelevant and what would normally be, unattended information. Hence, the experience would be described as ´realer than real`.

   The original models of Cherry, Broadbent etc. relating to levels of attended and unattended information describe them in terms of perception and processing. This leads onto factors affecting the balance and characteristics with for example, decreased processing of unattended information (Wojceulik) and greater processing of unattended information when the perceptual load is low (Lavie). Therefore, even if there is no awareness of unattended information there can still be processing (Wright, Nee and Norman). The situation with NDE is that this level of information would not normally be consciously experienced or be the first information recalled.  Therefore, the NDE condition changes the balance so that more information is attended. The question is how can this be achieved?

   One possible suggestion is that the stress response causes changes not only in the amount of information inputted as described above, but how the information is grouped or bound together. This has an effect on the capacity of working memory, which we know is limited to a certain number of event characteristics. Perception and Gestalt theories group visual elements such as lines, curves together into unitary objects such as forms and shapes which are then processed by working memory. By increasing the ´group size`, processing capability can therefore, be ´stretched` to more elements. We know of this through memory improvement techniques where individuals instead of trying to remember 6 different numbers group them together to represent years or months for example to increase recall efficiency. In the case of NDE experiences, the different sensory input may be grouped together as a single ´unit`. This has been shown in a similar manner by Jang who described event representation as containing not only the prime features of the event, but also more abstract, more internal features of the event such as attentional control states and category information. Such grouping was said to require the activation of the hippocampus and putamen as well as other regions representing visual feature-selective event information (primarily visual cortex), category-selective information (posterior frontal cortex) and control demand-selective event information (insula, caudate, anterior cingulate, and parietal cortex) event information.Therefore, brain areas affected by the stress response such as the hippocampus would, like the other attentional system brain areas, lead not only to changes in informational quality and quantity, but also awareness and processing capability. This would explain both the NDE event content and the content of FBs where both task relevant and task irrelevant information form part of the event representation.

   Although the stress response may then explain the change in conscious awareness relating to information and informational processing, can it also explain the dissociative consciousness state reported with NDE experiences? The mechanics of conscious awareness observed for visual input 170msec after presentation (Thierry) involves synchronized activation of many areas including the DMN, insular cortex, anterior cingulate cortex, medial prefrontal cortex and dorsal thalamic nuclei. However, conscious awareness under normal circumstances does not involve some of the transcendental, out-of-body type awareness reported by NDErs. In this case there is a dissociation from ´reality`, ie. the real-time sensory input information achieved via eyes, ears etc. One explanation for this could be the near-complete separation of top-down and bottom-up processing leading so that the top-down ´I, SELF` thinking part is distanced from the ´sensing, input` bottom-up part. This idea may be acceptable since imagining and dreaming are both examples where top-down processing and bottom-up input are ´distinct`. In the case of NDE, however the separation is likely to be achieved by the attentional system, which we have already seen is under the influence of cortisol released in the stress response. Top-down attentional modulation appears to suppress the incoming visual information via inhibition of firing at the V1 level (Jacob) and dissociation of attention and awareness is observed in the case of failure of applied chromatic flicker to reach awareness, but not alerting and orienting effects – also associated with V1 visual area effects (Lu). This separation is also apparent through the conscious awareness of an isolated object (or gist of a scene) in the near-absence of top-down attention (van Boxtel). (However, in this particular research example it appears that the explanation is against the proposed one for NDE since attention is described as top-down and ´thought` and consciousness is the information coming in and bottom-up. Possibly the way to approach the results of this study is to view conscious awareness as top-down and subject to top-down attention whereas sensory information is bottom up and subject to bottom-up attention with attentional modulation being the ´tool` and not the ´result`.)

   Therefore, one possible explanation for the more transcendental aspects of the dissociative consciousness reported by NDErs is the separation of top-down processing from bottom-up processing. The advantage of such a separation is protection of the higher brain areas from excessive information and high processing load. This may also be attributed to the emotional state at the time of the experience which is interrelated to attentional system functioning (Gregory). Cassol and team reported that the emotional stateduring the NDE experience was greater than for the autobiographical memory and both were greater than that recorded for the FB experience and it is clear that NDE are profound psychological events with highly emotional and self-related content. The increase in information is associated with this highly emotional state since it has been found that the link between episodic memory,  the feeling of ´being there` and the higher emotional experience is linked to better factual memory via an improved attentional focus on the stimuli, hence increasing memory encoding (Makowski). The same is reported with FB memories which also carry factual memory with strong emotions that can be recalled years later after the initial event (Brown, Finkenauer).

   The increased fear emotional state as the individual experiences the near-death event in Schauer and Elbert`s Stages 3 (fight) and 4 (fright) of the stress response is responsible for the increased levels of relevant and irrelevant material for the task (Newenhause, Stoets) and requires the activation of brain areas such as prefrontal cortex, amygdala, cingulate cortex and thalamus. Anxiety is shown to produce elevated cortisol levels linked to the hyperactivity of the HPA axis (Lenze) and there is a direct ventral connection between the amygdala and prefrontal cortex (Eden), both affected by anxiety.

   However, as the stress response progresses this fear emotional state shifts to a period of ´calm`. This relates to Schauer and Elbert`s Stage 5, ´flag` phase where there is a physiological ´shut-down` instigated via the activation of the parasympathetic system with symptoms such as drop in blood pressure and heart rate, but more importantly to the emotional experience of the NDE, a drop in arousal, a feeling of surrender, cognitive failure and a ´numbing` of all emotions, ie. the reported ´calm`. This can be brought about by the reduction in cortisol production and hence, cortisol effects on the brain areas involved in instigating the emotional state (eg. the Papez-Mclean limbic model involving activity of the amygdala, hippocampus and cingulate cortex and the two track Le-Doux theory with rapid emotional response based on information from the thalamus to the amygdala leading to autonomic and endocrine changes interpreted by the cortex and the slower, direct transmission of information from the thalamus to the cortex).  Reports show that the strength of the amygdala-hippocampus connectivity indicative of fear is enhanced with cortisol release (Hakamata) and cortisol production will re-instate fear after its extinction by activating this pathway (Kinner) and hence, reduced cortisol production will instigate a reduction in the emotional experience of fear. Also there is a possible switch off via the ventromedial prefrontal cortex in Schauer and Elbert`s Stages 5-6 since when stress is deemed controllable (ie. through the numbing of the emotions) then the stress induced activation of the dorsal raphe nucleus is inhibited by the ventromedial prefrontal cortex and the behavioural effects of stress are blocked because of the reduction in cortisol production (Amat). Therefore, the Stage 5, ´flag` stage is achieved and a reassessment of ´value` of the event occurs, eg. ´fear, threat` to ´calm, acceptance`.  

   What all of the Cassol`s observations and the above explanations show is that the functioning status of the brain during the NDE forming experience is not damaged to the extent that normal sensory input, processing and memory formation mechanisms do not occur. Therefore, there is support for the explanation of NDE from a psychological perspective of the ´biological/psychological` framework (Braithwaite). This framework supposes as supported here from looking at the memory characteristics and mechanism that the NDE memories are a consequence of the brain`s functional changes that can occur as a response to a perceived threat of death. Whether they form when death occurs for obvious reasons cannot be determined. Therefore, the sensory experience, memory characteristics and mechanism can be explained by normal cognitive physiology and functioning and even the OBEs and other more transcendental experiences can also be given explanations from normal neurochemical principles, eg. tunnels and lights caused by the disinhibition of the visual cortex and positive emotions and lack of pain attributed to the action of endorphins.

   Up to this point, we have only discussed the NDE from the perspective of the experience at the time of the event, but for both the NDE and FM, recall is important. Recall reflects the memory formed at the time and the level of forgetting between the event and the retrieval stage and since NDE and FM can be recalled years later this could have a significant effect if these memories follow the normal trajectories of other memory types. We have to assume that the actual physiological mechanisms for memory formation and storage are the same whether for NDE, FM or autobiographical memories, even if levels vary. This has been shown for the case of acute stress and cortisol which is relevant to the topic discussed here for the NDE and FM memories. Cortisol has been shown to be linked to memory changes due to its action on the hippocampus, an important area in the formation and recall of memories (Osborne). Acute stress (and even much later after the stress event) has been shown to cause rapid release of norepinephrine and glucocorticoids (cortisol) which bind to the hippocampal receptors. The resulting change in function of the hippocampus leads to the promotion of memory formation and consolidation, but appears to inhibit memory retrieval(Wais). This has been expanded in the case of cortisol increase by showing that the decreased memory level being observed is restricted to specific task relevant information whereas background memory is not affected (Steinmetz). This is contrary to what Cassol and team found where both internal and external information are remembered to a greater extent in NDE memories, but may support the ´change in grouping` hypothesis given above. 

   The demands for an ´undamaged` brain with NDE is reinforced by the need for connectivity of brain areas during the recall process. It is found that recall requires connectivity between the frontoparietal network, the default mode network and the hippocampus (Xiao, Westphal). Connectivity is demonstrated by brain wave activity and an increase in alpha brain wave activity is associated with NDE memory recall which is indicative of attentional control (van Schouwenburg). However, gamma waves normally associated with hippocampus activity during retrieval is not seen in NDE retrieval, but instead theta wave activity is observed which is linked to recall of emotional episodic events and the required hippocampal activity (Palmieri) and an increase in delta wave activity also associated with recall and personal feelings of reward (Knyazev). Therefore, the retrieval mechanism appears to be the same for the different memory types independent of quality or quantity of information recalled.

   Cassol and team however, expanded their exploration of the NDE memory retrieval by looking at the effect of emotional state during the recall session. We have already given an explanation as to how the actual quality and quantity of the memory formed was increased by the stress experienced by the individual during the NDE event, but Cassol and team looked into whether or not inducing the same contextual and emotional states during the recall as experienced during the event itselfwould strengthen the recall and reactivation of the NDE memory.  This exploration was indicated because it has been found that memory retrieval is increased when there is a match between the consciousness state at encoding and the consciousness state at retrieval (state-dependent memory – Overton). Researchers have also found that NDE memories are better retrieved when delta wave connectivity is greater because this appears to be the dominate wave pattern during the encoding NDE experience phase (Hartman). (Delta wave activity is also associated with transcendental type experiences (Hartman) which would correlate with the OBE and other dissociative consciousness events linked to the NDE). This appears to be supported by Cassol and team`s results since they found that the NDEr experienced the sensation of feeling the emotions felt during the event while they recalled the NDE and these feelings were greater for the NDE and autobiographical memories than for the FB memories. The retrieval of NDE induces the same emotional states of fear, ´realer than real` emotionality, but then calm, positive emotions.  Therefore, whereas fear can lead to memory impairment normally this is counteracted by the positive, calming emotional response which would lead to normalised recall (McKenzie). The expression of emotion during recall was found to aid both central and peripheral information retrieval (Davidson). In the case of the FB memories, which are more than just emotional memories,  emotions can help or impair recall just like with the NDE, eg.  people experiencing high levels of stress and distress show significantly higher levels of recall of context information than individuals with lower levels of stress after a period of 3 years (Bohannon). The emotional reactions however, to the event may differ in that the reaction could remain over the long-term, be short lived, or only appear after the event’s impact is fully appreciated by the individual in reference to his own consequentiality.

   This question of consequentiality and relevance to SELF and self-identity brings to light another factor that may or may not strengthen NDE memories and FB memories through recall. NDE experiences and memories are considered very important to the individual`s life story, greater than for autobiographical memories, which are more important than FB memories. These findings are understandable since NDEs are from living through a near-death, extremely personal and significant life threatening event and so should be more important than the other two memory types, but FB memories are also important to the individual otherwise why would the individual record it.

  The relationship between NDE memory and the SELF is that the NDE memory is characterised by the rich phenomenology recorded at the time which we have already described as being due to the stress response, increase in cortisol production and increased attentional and emotional systems affecting central and peripheral information input and storage. This richness of the NDE memory is associated with the SELF and self-identity as observed by the centrality scores (CES scores) in Cassol and team`s experiments and the effect of the NDE event may affect the SELF and self-identity for a much longer time and to a much greater degree than a FB memory. It has been reported that the prevalence of meaning attached to the NDE event may bring about changes in behavioural patterns such as for example a reduced fear of death, a greater level of compassion for others or a lower value placed on materialistic positions. Therefore, recall of the NDE memory strengthens the event details and if not changed, strengthens the relevance to the SELF and self-identity. The FB memories must also serve some role in SELF or self-identity since they are recorded for a particular reason. However, this is disputed by some who say that the FB memory does not serve personal consequentiality, but serves consequentiality for the community, hence shaping social identity rather than self-identity. However, the same event characteristics may be inputted and stored for a vast number of individuals, but the exact memory details may be different due to the incorporation of the internal, phenomenological information. Therefore, this renders only part of the memory having consequentiality for the community and all to the SELF and self-identity.

   This factor of consequentiality and association with the SELF and self-identity reinforces the personal nature of the NDE and FB memories and brings into discussion the topic of retrieval accuracy. Recall of memories can reinforce event characteristics through their reactivation and strengthening of firing of the neuronal assemblies that make up the event representation (internal and external features). Therefore, the thinking or re-telling of the experiences may change or strengthen the characteristics recorded. Cassol and team reported that in the case of the NDE that reactivation sharing and its frequency were found to be important and this is as expected. It was also found to influence the memory accuracy. The problem with judging whether the recall of NDE events is accurate is that the individual doing the recall is the only person who has experienced it and who has recorded the internal and external event characteristics. Therefore, the accuracy of the recall for at least the first reactivation cannot be verified totally by a third party. Subsequent retrievals can be judged to some extent via consistency of reporting and to this extent it has been reported that NDErs remember the event clearly and accounts are not modified over a period of two decades. In this way it appears that memory traces do not fade like other autobiographical memories and narratives remain consistent if no post-event processing, construction and embellishment takes place.

   In the case of FBs, judgement of accuracy would be deemed easier to establish since part of the event characteristics recorded are shared by a large number of people and therefore, there should be many accounts of these particular facts. However, the details may be affected by who and how events are recalled by others and reports have shown how media attention can affect the recall of these shared features. Again like NDE however, FB memory has a quantity of external peripheral facts that may not be common to all and a quantity of internal event characteristics which are entirely individual. Therefore, it would be difficult for a third person to check the accuracy of these features against the original encoding event and consistency accuracy is the only measurement possible. Researchers have found once an inconsistency emerges which usually occurs within the first year then it remains (Hirst) and like NDE events, FB memory recall may be years later after the event. Again, FB recall makes the memory subject to change via post-event processing, construction, embellishment and the influence of others.

   Therefore, this exploration of NDE and FB memories has shown that they are more examples of how emotional state can affect informational input and memory. The uniqueness of the memories with their increased level of peripheral, internal information can be attributed to the change in attentional state induced particularly in the case of NDE to a rise in cortisol production due to the perceived threat to life. Their value to memory research is via their strong association to the SELF and self-identity and this in the case of FB memories can be advantageous.

Since we`re talking about the topic……….

            …..dexamethasone is a synthetic glucocorticoid known to lead to the instigation of the stress response. Is it possible that the NDE-like responses can be induced by its use and these responses can be observed with neuroimaging and compared to those experienced by practiced meditators?

            …. If NDE-like memory can be induced by hypnosis, can neurochemical studies on cortisol production, calcium influx and neuroimaging of brain area activity confirm whether or not, cortisol is the cause of the memory changes observed?

            …. ketamine-induced anaesthesia is said to be similar to NDEs in the level of wakefulness and connectedness, could this be used to study the influence of external environmental changes such as loud noise and smells on memories if formed by experienced meditators?

Posted in Uncategorized | Leave a comment

action video game experience linked to improved attention driven perceptual exploration in categorisation learning

Posted comment on ´Play to win: action video game experience and attention driven perceptual exploration in categorisation learning` written by S. Schenk, C. Bellebaum, R.K. Lech, R. Heinen and B. Suchan and published in Frontiers in Psychology 13th May 2020 doi 10.3389/fpsyg.2020.00933

SUMMARY

   Schenk and colleagues investigated the difference between action video gamers and non-gamers in visual exploration and attention driven perception during a categorisation learning task. Their subjects, 17 right-handed non-gamers (mean age nearly 23, nearly all female) and 16 action video gamers (mean age nearly 24, nearly all male) performed 490 random trials (5 blocks of 98) of a visual categorisation task. This consisted of distinguishing 14 ring stimuli which had the same structure, but different colour combinations. These had to be divided into 2 categories, seven in each and made up of a prototype, 5 typical stimuli and one exception. The typical stimuli shared 5 colour features with the prototype of the category, whereas the exception shared five colour features with the prototype of the other category.  The subjects did not know of the existence of the exceptions and had to first work out that they were present and then had to explicitly remember both exceptions in order to categorise all the stimuli presented. The stimuli were presented and the response made by pressing the left and right control keys designated for the two categories. The key press was followed by an immediate feedback of right or wrong. The subjects were required to react within 1.8 secs of stimulus presentation and failure to do so resulted in a verbal warning to quicken response time. Following the feedback the subjects were presented with a fixation cross for 1-2 secs before being allowed a short break before the next trial started.  EEG recordings were taken during the experiment and eye-tracking recordings of the subject`s right eye were also made during the stimulus presentation.  

   Data analyses were performed on the recordings. In the case of the EEG recordings, the datasets were analysed with software packages. Blink artefacts and vertical eye movements were removed from the EEG data and components reflecting eye movements were removed from the EEG signal. The recordings for the prototypes and typical stimuli were combined because of their similarity (prototypical stimuli) and the recordings were adjusted so that the segments for both stimulus types (prototypical stimuli and exceptions) were followed by a baseline correction relative to the 200ms preceding the stimulus presentation. In the case of the P150 ERP component, data from the parietal-occipital electrodes was recorded and the amplitude was taken as that within 120-180ms after stimulus presentation. The maximum negative peak amplitude within the time frame 150-190ms after stimulus presentation was taken and defined as the Peaks of the N170 amplitude. The analysis of the N170 component was carried out using different electrodes to those of the P150.

    The eye-tracking data was analysed using appropriate software. Fixations were visualised on the basis of heat maps with the ring stimuli divided into areas. Percentages of the fixation rates and number of fixations were calculated for all areas of the ring and for the centre and the differences between prototypical and exceptions calculated. Eye movements (taken as changes in velocity of the eye position exceeding 40 degrees per second and displacement of at least 1.5 degree visual angle) were also analysed with saccades detected within 800ms of the presentation period. Statistical analyses of the various sets of behaviour data were carried out using ANOVA with repeated measures (including Greenhouse-Geisser procedure and Bonferroni), paired T test and G Power.

   The results of Schenk and colleagues experiments showed that the learning process relied on the immediate feedback of right or wrong response. The ANOVA analysis carried out for the percentage of correct responses yielded significant main effects for both the factors ´stimulus` and ´block` as well as showing significant interaction. The pairwise comparisons revealed for both groups more correct categorisations in the last blocks, especially for the prototypical stimulus. Using the paired t-test it was shown that interaction was based on a high number of correctly categorised prototypical stimuli compared to a low number of correctly categorised exceptions at the early stage of the experiment. A high increase in correctly categorised exceptions in block 3 showed that the successful categorisation of the exceptions occurred later. The differences between the blocks between percentage correct responses for exceptions and prototypical stimuli were shown by paired t tests to yield significance between block 2 and 3. All other results showed no significant differences.

    The results from the ANOVA analysis of behavioural data of the two groups showed a significant interaction between ´stimulus` and ´group`. This showed that action video gamers had superior categorisation performance for the exceptions than the non-gamers. The interaction effect was particularly enhanced in block 2 which indicated that action video gamers perform correct categorisations of exceptions earlier than non-gamers. They also categorised the exceptions better than the non-gamers at the beginning of each of the five blocks.

   The studies of Schenk and colleagues also showed gender and reaction time differences. The female gamers categorised better than other female non-gamers and male gamers categorised better than other non-gamers. However, female gamers categorised better than male gamers and female non-gamers performed better than male non-gamers. In the case of reaction times, reaction times decreased during the course of the experiment for both stimulus types and the largest difference was observed in block 3 onwards.  The subjects categorised the prototypical stimuli in total faster than exceptions.

   The results of the eye-tracking studies showed that in general there were higher fixation rates at the stimulus centre. At the beginning of the learning process, that is at the beginning of the experiment, data showed scattered fixations over the whole stimulus with more focused eye movements for the exceptions than for the prototype stimuli. At the end of the learning process, eye movements were more concentrated for both stimulus types and focussed on specific colour segments. Action video gamers exhibited less scattered and more centre-focused eye movements compared to non-gamers and this was particularly apparent in the last block of the experiment. Significant interaction between stimulus and block began at block 3.

   The results of the ANOVA investigation into the saccadic eye movements supported the findings of the fixation rates. The mean numbers of saccades per trial were negatively affected by increasing block number with some participants having no saccades during the first 800ms of stimulus presentation toward the end of the experiment. Schenk and colleagues also found increases in saccade latencies during the course of experiment supporting the reduced number of saccades performed.

    The analysis of the EEG results of the P150 amplitude showed that action video gamers (both male and female) generally exhibited higher P150 amplitudes than non-gamers (both male and females) for both stimulus types. The analysis of N170 amplitude gave a significant between-subject effect with the amplitude more negative for non-gamers than for action video gamers. The analysis also yielded a main effect for the factors electrodes with the PO7 and PO8 amplitudes more negative than the others. Different profiles were shown for both groups. Only for the non-gamers was there a significant effect for the factors electrodes. For the video gamers there were no significant main effects or interactions. The analysis of the N170 latency showed that the action video gamers had higher latencies on the right side and non-gamers almost significant higher latencies on the left. For the exceptions, non-gamers showed significant higher PO7 latency whereas action video gamers had higher latencies at the PO8 position for prototypical stimuli. Both the PO7 and PO8 latencies were shorter for the prototypical stimuli and the P8 and PO8 latencies were shorter for exceptions.

    Schenk and colleagues concluded their article with a discussion on their findings. They concluded that a learning process took place for the stimulus material by both groups of subjects. Both groups categorised the prototypical stimuli in earlier blocks and with faster reaction times than the exceptions. A change appeared to occur in the third block. For example; in the third block the number of correctly categorised exceptions increased; the difference in accuracy only became significant between the second and third block; and the highest difference between reaction times for the two stimulus types was detected in the third block. These results suggested that the observations were due to different learning strategies for the stimulus types (not demonstrated) or a need to explicitly remember exceptions.

   Further analyses showed that there were differences between the two groups for exceptions. The action video gamers showed better categorisation performance at the beginning of each block throughout the experiment and also demonstrated superior categorisation performance especially in the second and third block, although it became balanced towards the end. The non-video gamers needed more trials to correctly categorise the exceptions. Analysis of the fixation rates showed that they were not dependent on stimulus construction, but were associated with central or peripheral location of the stimulus segments. In the first block there were more fixations on the stimulus segments and in the last block more on the stimulus centre independent of subject group. The fixation rates decreased with increasing number of blocks for both groups, as did the number of saccades supported by increases in saccade latencies from the beginning to the end of the experiment. In the case of the exceptions, one or two segments that were decisive for the correct categorisation showed higher fixation rates at the end of the experiment.

  Schenk and colleagues concluded that both groups instigated a learning process regarding the stimulus material. It was possible that in the early learning phase, both groups tried to learn the stimuli based on their different colour features (showed more fixations on stimulus segments) and in later blocks, showed more fixations on the stimulus centres. This meant that they had learnt the stimulus structure and did not need to explore anymore, going directly for the ´change` site. In the case of the exceptions, only action video gamers showed higher fixation rates on the stimulus centre possibly indicating covert peripheral processing. (And the authors noted that this was the advantage of their experimental design and that this advantage might disappear if the critical features were indeed placed in the centre.) However, it was also proposed that the action video gamers could be faster in capturing visual information per se, or more adept at processing complex stimuli. This was supported by the studies on the ERP components P150 and N170 which reflected differential processing for the stimulus material of the two subject groups.

   Therefore, it was concluded that action video gamers have enhanced attentional and visual processing capabilities and non-gamers normal. The action video gamers show different stimulus exploration (more centre vs peripheral) and use an enhanced early perceptual analysis of the stimulus material and hence, may detect changes in objects faster and learn the belonging of the stimuli to their categories earlier. Schenk and colleagues findings support the work by others (West) that action video gamers show enhanced counting and remembering of specific sequences, features and locations which could be an advantage when learning the exceptions in Schenk`s visual categorisation task. The authors concluded that there were two areas worthy of further exploration: the first, testing with other types of video games; and secondly, more in-depth study of differences in gender with this type of learning.

COMMENT

   What makes this article interesting is that supports the view that cognitive skills gained from performing one task can be applied advantageously to other tasks that are not direct imitations of the performed one. In Schenk and team`s experiments, expertise in action video gaming gained from long-term practice proved a benefit to individuals in their performance of a categorisation task which involved visual input, working memory, short-term memory and hand-eye coordination. Therefore, it appears that action video gaming which can be a solo activity and can be performed anywhere (provided the computer hardware and software are available) may give individuals who have limited access to training opportunities in social settings the chance to maintain or restore various cognitive skills.

   The experimental set-up of Schenk and colleagues involved general skills in attention, working memory, short-term memory and decision-making and more task-specific skills related to visual input (colour and location) and visual search.  Some cognitive skills were not required such as those relating to emotions (value, empathy), other senses and information-type factors such as temporal order and movement. In general, it was found that both subject groups demonstrated similar performance regarding short-term memory and decision-making. For example, both groups showed earlier responses to discount prototypical stimuli than exceptions which implies that the routine was learnt, the decision-making method decided upon and the categorisation grouping recalled independent of whether the individuals had prior gaming experience or not. The third block of tests appeared to be the turning point for both groups and therefore, it is likely that the advantages of prior gaming experience did not lie with these particular cognitive skills. The areas where the action video gamers gained an advantage from their training appeared to be directly and mainly related to the visual processing and perception areas (eg. capable of identifying the key colour segment characteristic as being more centrally located, higher fixations on the stimuli to support short-term memory demands), which lead to the indirect advantages of faster speeds at categorising and category learning. This comment focusses on these types of systems and mechanisms where the gaming experience appears to give an advantage.

  We begin by hypothesising whether the hierarchical physiological structure of the visual input pathway itself is a key point to the training improvement. Since the experiment was associated with colour segment recognition, it can be assumed that the forward sweep of the incoming visual information occurred in all subjects within 100msecs of the stimulus onset and retinal cone firing. This would be followed by the relevant firing of the retinal bipolar cells with their on/off centres and surrounds leading to greater discrimination of the colour at the next stage involving firing of cells in the retinal ganglion layer. Here, 95% of the cells are P-type ganglion cells, responsible for shape determination although some are sensitive to wavelength. The other 5% of cells, the nonM-nonP ganglion cells are sensitive to wavelength and have the on/off centre and surround type structure. The response to one wavelength in the centre can be cancelled by the response to another wavelength in the surround and this is where the colour signal becomes established in a firing pattern. This colour signal is passed further up the visual pathway to the complex structure of the lateral geniculate nucleus (LGN) located in the dorsal thalamus. Information about colour is transferred through the activation of appropriate nonM-nonP ganglion cells of the six layer LGN structure to the tiny neurons of the koniocellular layers, which lie just ventral to each parvocellular layer. The receptive fields of the cells of the koniocellular layers are also of a centre-surround type structure and have either light/dark or colour opponency and it is these cells that define colour (the blob pathway). It should be noted that input from both eyes is kept separate in this area through the layering structure, but as far as colour is concerned there appears to be no binocularity disparity. It may be that shadow/contrast and the other qualities of colour, brightness and saturation, may be eye-dependent though.

    Projections from the LGN lead to the V1 visual cortex and then colour determination seems to follow the psychologists WHAT pathway (P pathway or temporal pathway), which takes the ventral route into the infero-temporal lobe via the cortical regions V2, V4, (V8) and area IT. A quarter of the complex V1 structure deals with information from the retinal fovea, ie. 25% of incoming information is about colour and firing follows a characteristic retinotopic colour map which is maintained as the colour pathway continues upwards via the ventral stream to areas V2 and V4. Further discrimination of the colour signal occurs at the area IT (in the interior temporal lobe). This area is known to be stimulated by a wide variety of colours and abstract shapes and is said to be important for both visual perception and visual memory. A small patch of the area IT is particularly responsive to faces with some faces being more effective stimuli than others. Therefore, since the categorisation test was performed by all individuals successfully then it is likely that the hierarchical physiological structure of the visual pathways of the individuals were the same independent of whether they had gaming experience or not and the visual information and colours were inputted and perceived by the system described above. Deficits in eye sight, or particular brain area deficiencies (eg. lesions of the visual cortex) would have had an effect on categorisation performance and these reasons can be excluded since the individuals that took part in Schenk and colleagues experiments were all described as healthy and had perfect eyesight.

   Therefore, at which points of the visual input and processing could the gamers gain an advantage? One such area could be the determination and appropriation of colour to the segments. This relies on an increased capability to recognise colour and depends on the concept of colour constancy. The feature of colour appears to have debatable value when action is the individual`s response to the visual input. A degree of memory involvement occurs in the assessment of colour as colour is considered constantand the perception of it appears to be reliant on visual pathway detection and on previous experience and expectation. Colour constancy, attributed to certain cells in the V4 and hence, dependent on an individual`s V4 performance, is where there is a tendency for a surface or object to appear to have the same colour even when there is a change in the wavelengths contained in the lighting source, eg. an object does not appear ´redder` when seen in artificial light. This is because colour constancy is found to be influenced by top-down processes (Bloj and the retinex theory of Land) where individuals use their own knowledge to interpret incoming visual stimuli and so colour is defined according to an individual`s own interpretation according to their own perceptions and experience and not the true colour determined by the colour`s physical wavelength. It was found that the colour constancy factor was responsible for global (compared to the whole visual scene) and more relevant here to local contrast, which involves comparing the retinal cone responses from the target surface with those of the immediate background (Kraft). This is likely to apply to the experiments of Schenk and team where the target colours would be compared to the neighbouring colours. Therefore, colour constancy and the personal definition of colour may be an area where the gamers have an advantage. Action video games are fast and rely on the gamers identifying shapes quickly. Therefore, the gamers may have developed cognitive skills in fast assessment of colour according to their own definitions (their own individual measurement of colour constancy). This capability of fast perception could be applied to the categorisation task and lead on to more efficient and accurate formation of short-term memories of the target colours of the visual stimuli presented in the experiments. 

   Another skill that relates to the visual target that is likely to be increased in gamers theoretically and was actually shown by Schenk and team`s experiments is the increase in control of eye movement. This is shown by the change in fixation rates observed in Schenk and team`s experiments for those subjects with prior gaming experience and the observation supports work by others who show that eye movements can enhance sensitivity to the target (Ennis). The association comes from the coupling between eye movement and deliberate goal-directed focus (Walcher). In Schenk`s experiments this means that eye movement control is associated with the establishment of target colour and manifests as fewer saccades and longer fixations. The experiments of Schenk and team`s were made slightly easier for subjects since the target colours were in the centre of the shape and hence, focus of the visual field and not in the peripheral regions. Hence, their determination was maximised since most retinal cones of the first stage of the visual pathway are situated in the central point of the lens at the fovea where the light rays pass in a straight line through from the lens to the retinal cell layer. Therefore, gamers with their prior experience of keeping targets centralised in the focus of the visual field will optimise the firing patterns of the presented stimuli (Carrillo-Reid) whereas non-gamers are likely to be less focussed (searching central and peripheral) until they have learnt where to concentrate their focus. The advantage of keeping the target in the focus of the visual field may be even more specific in that targeting might be dependent on only a small proportion of the total segment available. This is supported by visual search hypotheses, which say in a large stimulus, parts of the stimulus are processed quicker than the whole (Kinchla). Therefore, the gamers could specifically target not only the central segment where they know that the target colour changes are located, but narrow it down to even a smaller proportion of that target area to reduce processing load.

 A topic linked to eye movement and one that may play a role in the categorisation task given by Schenk and team is that of visuomotor responses. It is possible that through their prior gaming experience, the gamers have developed faster eye-hand coordination. Once the visual process is stopped, recognition is made and the decision taken, the end result is a motor movement consisting of a key press action. Therefore, efficient and quick eye-hand coordination may extend the gamer`s visual advantage, ie. will process faster and carry out the required response faster. This improved eye-hand coordination may occur via strengthened connectivity between multiple cortico-cortical and cortico-subcortical frontal cortex networks (Brovelli) and the temporal locking of firing oscillations of the early visual processing with the early motor planning controlling the execution of the hand response (Tomassini). Therefore, training may give the gamers an advantage by speedier movements once the decision has been made.

   Gamers may also have an advantage in Schenk and team`s experiments by being able to maintain concentration and process relevant information faster during the learning phase and then shifting this to non-conscious processing at a later stage, albeit faster than the non-gamers. This type of capability would be gained by playing many hours of action video games where fast changing situations are constantly being presented and decisions have to be made. Gamers then when presented with a simpler categorisation task are likely to shift to non-conscious processing at an earlier stage than non-gamers and since non-conscious processing is faster than conscious, then response times of the gamers would be quicker. Conscious information processing which occurs early on in the trial blocks involves a number of skills relating to awareness of what is required and this conscious awareness is demonstrated by the N170 component (Thierry). Two cognitive processes come into play, working memory and attention, and these capabilities would have been honed to cope with the task demands by the prior gaming experience which could be considered as priming. From a physical perspective, priming would help the individual to target specifically (ie. only a proportion of the colour segment in the visual search part of the task) leading to improved cue utilisation for example. A greater level of cue utilisation leads to consistently greater response latencies consistent with strategies that maintain accuracy, but reduce the demands on cognitive resources (Brouwers). This means that priming would positively affect demands on attention and information processing relating to working memory performance.

   With regards to attention, the high level of conscious awareness particularly at the beginning of the trial blocks leads to higher levels of top-down attention which is likely to result in strong alpha brain wave activity in the fronto-parietal network (Van Schouwenburg). Various areas are responsible in the selection and maintenance of attention on stimuli eg. there is co-activation of neurons within 50-200ms across the anterior cingulate cortex and prefrontal cortex during stimulus selection in a spatial attention task (Oemisch); and firing in the ventral intraparietal sulcus maintains attention to a specific location (Capotosto). However, experience (ie. by prior gaming experience) would affect how targets are consciously selected and attentional selection would respond. This is supported by Corradi-Dell´Acqua who showed that the ability to select, within the complexity of sensory input, the information most relevant for the purpose of the task was influenced by both internal settings (ie. top-down control and conscious awareness) and the relevant features of the external stimuli (ie. bottom-up control and visual input). This capability to switch to relevant information comes from firing of the areas involved in top-down control (frontal, parietal and sensory cortices) and the lateral intraparietal neurons (Kumano). Lateral intraparietal neurons appear to accumulate relevant information depending on context to decide which eye movements to carry out to maximise it. The selection of the target would imply attended stimuli and this would enhance the evoked firing potentials for these stimuli in comparison to non-attended (Andersen). The effect of increased firing with reference to Schenk and team`s experiments is that colour (the target characteristic) would exhibit general sharpened selectivity by initial gain at the V1 level and within 100msec a sharper tuned profile in the posterior ventral cortex (Bartsch).

   This increased attention and selectivity of target afforded by conscious awareness and the ability of gamers to know faster what is required leads to working memory performance gain. Gaming experience, if considered as priming, would bring about benefits to working memory as seen through pre-stimulus alpha brain wave activity (Myers) and theta-band oscillations representing multiple predictions that are dynamically coordinated in time (Huang). The training of the individuals although unlikely to produce long-term changes in the physiological structure of the visual pathway are likely to induce changes to the physiological structures involved in working memory if they follow the pattern of other cases where training has been carried out prior to testing. For example, Caeyenberghs has reported that adaptive working memory training led to improvement on non-training related working memory tasks and tasks of reasoning and inhibition and these improvements were related to increased global integration within the fronto-parietal attention network observed by increased structural white matter network connectivity. Astle also reported that in children, after training, connectivity between the fronto-parietal networks and both lateral occipital complex and inferior temporal cortex areas exhibited increased strength of connectivity at rest. These were mirrored by an increased working memory performance.

    The problem with working memory is that it is of a limited capacity with regards to maintaining and manipulating the objects held in it (Myers) with resolution decreasing as the number of items increases above 3 (Anderson). This is in spite of the work load being spread over different areas (eg. specific brain areas have particular working memory functions such as post parietal cortex – manipulation of information and item maintenance; lateral occipital cortex – item maintenance and even different areas of the prefrontal cortex, with ventrolateral prefrontal cortex involved in face and vocal information; IT, V4 with visual information – Plakke). Therefore, conscious awareness and attention focused on relevant and targeted material aids the overall performance of working memory. The characteristics featured in object based attention are present in the working memory (Peters) and are strengthened (Soto) and sharpened by this overlapping (Lim). Therefore, attention selectively updates and maintains the relevant material (Blauracke, Heuer) required for the task at hand. This would mean that although both gamers and non-gamers would employ the same working memory processes the degree of efficiency of the former would be higher at least in the earlier trial blocks.

    Therefore, we can see that several areas of visual input and processing may be advantageously changed by training involving playing action video games. The value of playing action video games by increasing the selection of relevant material, maintaining it in working memory and then carrying out a response of a planned motor movement is significant for circumstances where these cognitive skills need to be maintained or improved and where the individual`s environment presents it as a valid learning option. The proviso is that the game used has to be challenging, capable of repetition without loss of motivation, set-up to give feedback preferably immediately and relevant to the required cognitive skill.

Since we`re talking about the topic…………………..

            …… achromatopsia sufferers have damage in the V2, V3 and V4 areas and therefore, can process information about colour implicitly but are unable to use the information explicitly when the judgement concerns colour. Therefore, if Schenk and team`s experiments were repeated would the advantages of the gamers be negated?

            …… serotonin administration causes a mainly multiplicative decrease of visual responses and a slight increase in stimulus-selective response latency (Seillier). Can we assume that if serotonin was administered prior to the test, both groups would suffer a loss in performance, but the gamers would still show the training advantage?

            …..……….mood is reported to have an effect on working memory performance with positive mood enhancing verbal and impairing spatial working memory, whereas negative mood enhances spatial and impairs verbal working memory (Storbeck). If Schenk and team`s experiments were repeated with anxiety levels increased for both groups, would we see any change in performance of the two groups in general particularly at block 3, the turning point between learning and automatic response?

Posted in attention, learning, perception | Tagged , , | Comments Off on action video game experience linked to improved attention driven perceptual exploration in categorisation learning

action video game experience linked to improved attention driven perceptual exploration in categorisation learning

Posted comment on ´Play to win: action video game experience and attention driven perceptual exploration in categorisation learning` written by S. Schenk, C. Bellebaum, R.K. Lech, R. Heinen and B. Suchan and published in Frontiers in Psychology 13th May 2020 doi 10.3389/fpsyg.2020.00933

SUMMARY

   Schenk and colleagues investigated the difference between action video gamers and non-gamers in visual exploration and attention driven perception during a categorisation learning task. Their subjects, 17 right-handed non-gamers (mean age nearly 23, nearly all female) and 16 action video gamers (mean age nearly 24, nearly all male) performed 490 random trials (5 blocks of 98) of a visual categorisation task. This consisted of distinguishing 14 ring stimuli which had the same structure, but different colour combinations. These had to be divided into 2 categories, seven in each and made up of a prototype, 5 typical stimuli and one exception. The typical stimuli shared 5 colour features with the prototype of the category, whereas the exception shared five colour features with the prototype of the other category.  The subjects did not know of the existence of the exceptions and had to first work out that they were present and then had to explicitly remember both exceptions in order to categorise all the stimuli presented. The stimuli were presented and the response made by pressing the left and right control keys designated for the two categories. The key press was followed by an immediate feedback of right or wrong. The subjects were required to react within 1.8 secs of stimulus presentation and failure to do so resulted in a verbal warning to quicken response time. Following the feedback the subjects were presented with a fixation cross for 1-2 secs before being allowed a short break before the next trial started.  EEG recordings were taken during the experiment and eye-tracking recordings of the subject`s right eye were also made during the stimulus presentation.  

   Data analyses were performed on the recordings. In the case of the EEG recordings, the datasets were analysed with software packages. Blink artefacts and vertical eye movements were removed from the EEG data and components reflecting eye movements were removed from the EEG signal. The recordings for the prototypes and typical stimuli were combined because of their similarity (prototypical stimuli) and the recordings were adjusted so that the segments for both stimulus types (prototypical stimuli and exceptions) were followed by a baseline correction relative to the 200ms preceding the stimulus presentation. In the case of the P150 ERP component, data from the parietal-occipital electrodes was recorded and the amplitude was taken as that within 120-180ms after stimulus presentation. The maximum negative peak amplitude within the time frame 150-190ms after stimulus presentation was taken and defined as the Peaks of the N170 amplitude. The analysis of the N170 component was carried out using different electrodes to those of the P150.

    The eye-tracking data was analysed using appropriate software. Fixations were visualised on the basis of heat maps with the ring stimuli divided into areas. Percentages of the fixation rates and number of fixations were calculated for all areas of the ring and for the centre and the differences between prototypical and exceptions calculated. Eye movements (taken as changes in velocity of the eye position exceeding 40 degrees per second and displacement of at least 1.5 degree visual angle) were also analysed with saccades detected within 800ms of the presentation period. Statistical analyses of the various sets of behaviour data were carried out using ANOVA with repeated measures (including Greenhouse-Geisser procedure and Bonferroni), paired T test and G Power.

   The results of Schenk and colleagues experiments showed that the learning process relied on the immediate feedback of right or wrong response. The ANOVA analysis carried out for the percentage of correct responses yielded significant main effects for both the factors ´stimulus` and ´block` as well as showing significant interaction. The pairwise comparisons revealed for both groups more correct categorisations in the last blocks, especially for the prototypical stimulus. Using the paired t-test it was shown that interaction was based on a high number of correctly categorised prototypical stimuli compared to a low number of correctly categorised exceptions at the early stage of the experiment. A high increase in correctly categorised exceptions in block 3 showed that the successful categorisation of the exceptions occurred later. The differences between the blocks between percentage correct responses for exceptions and prototypical stimuli were shown by paired t tests to yield significance between block 2 and 3. All other results showed no significant differences.

    The results from the ANOVA analysis of behavioural data of the two groups showed a significant interaction between ´stimulus` and ´group`. This showed that action video gamers had superior categorisation performance for the exceptions than the non-gamers. The interaction effect was particularly enhanced in block 2 which indicated that action video gamers perform correct categorisations of exceptions earlier than non-gamers. They also categorised the exceptions better than the non-gamers at the beginning of each of the five blocks.

   The studies of Schenk and colleagues also showed gender and reaction time differences. The female gamers categorised better than other female non-gamers and male gamers categorised better than other non-gamers. However, female gamers categorised better than male gamers and female non-gamers performed better than male non-gamers. In the case of reaction times, reaction times decreased during the course of the experiment for both stimulus types and the largest difference was observed in block 3 onwards.  The subjects categorised the prototypical stimuli in total faster than exceptions.

   The results of the eye-tracking studies showed that in general there were higher fixation rates at the stimulus centre. At the beginning of the learning process, that is at the beginning of the experiment, data showed scattered fixations over the whole stimulus with more focused eye movements for the exceptions than for the prototype stimuli. At the end of the learning process, eye movements were more concentrated for both stimulus types and focussed on specific colour segments. Action video gamers exhibited less scattered and more centre-focused eye movements compared to non-gamers and this was particularly apparent in the last block of the experiment. Significant interaction between stimulus and block began at block 3.

   The results of the ANOVA investigation into the saccadic eye movements supported the findings of the fixation rates. The mean numbers of saccades per trial were negatively affected by increasing block number with some participants having no saccades during the first 800ms of stimulus presentation toward the end of the experiment. Schenk and colleagues also found increases in saccade latencies during the course of experiment supporting the reduced number of saccades performed.

    The analysis of the EEG results of the P150 amplitude showed that action video gamers (both male and female) generally exhibited higher P150 amplitudes than non-gamers (both male and females) for both stimulus types. The analysis of N170 amplitude gave a significant between-subject effect with the amplitude more negative for non-gamers than for action video gamers. The analysis also yielded a main effect for the factors electrodes with the PO7 and PO8 amplitudes more negative than the others. Different profiles were shown for both groups. Only for the non-gamers was there a significant effect for the factors electrodes. For the video gamers there were no significant main effects or interactions. The analysis of the N170 latency showed that the action video gamers had higher latencies on the right side and non-gamers almost significant higher latencies on the left. For the exceptions, non-gamers showed significant higher PO7 latency whereas action video gamers had higher latencies at the PO8 position for prototypical stimuli. Both the PO7 and PO8 latencies were shorter for the prototypical stimuli and the P8 and PO8 latencies were shorter for exceptions.

    Schenk and colleagues concluded their article with a discussion on their findings. They concluded that a learning process took place for the stimulus material by both groups of subjects. Both groups categorised the prototypical stimuli in earlier blocks and with faster reaction times than the exceptions. A change appeared to occur in the third block. For example; in the third block the number of correctly categorised exceptions increased; the difference in accuracy only became significant between the second and third block; and the highest difference between reaction times for the two stimulus types was detected in the third block. These results suggested that the observations were due to different learning strategies for the stimulus types (not demonstrated) or a need to explicitly remember exceptions.

   Further analyses showed that there were differences between the two groups for exceptions. The action video gamers showed better categorisation performance at the beginning of each block throughout the experiment and also demonstrated superior categorisation performance especially in the second and third block, although it became balanced towards the end. The non-video gamers needed more trials to correctly categorise the exceptions. Analysis of the fixation rates showed that they were not dependent on stimulus construction, but were associated with central or peripheral location of the stimulus segments. In the first block there were more fixations on the stimulus segments and in the last block more on the stimulus centre independent of subject group. The fixation rates decreased with increasing number of blocks for both groups, as did the number of saccades supported by increases in saccade latencies from the beginning to the end of the experiment. In the case of the exceptions, one or two segments that were decisive for the correct categorisation showed higher fixation rates at the end of the experiment.

  Schenk and colleagues concluded that both groups instigated a learning process regarding the stimulus material. It was possible that in the early learning phase, both groups tried to learn the stimuli based on their different colour features (showed more fixations on stimulus segments) and in later blocks, showed more fixations on the stimulus centres. This meant that they had learnt the stimulus structure and did not need to explore anymore, going directly for the ´change` site. In the case of the exceptions, only action video gamers showed higher fixation rates on the stimulus centre possibly indicating covert peripheral processing. (And the authors noted that this was the advantage of their experimental design and that this advantage might disappear if the critical features were indeed placed in the centre.) However, it was also proposed that the action video gamers could be faster in capturing visual information per se, or more adept at processing complex stimuli. This was supported by the studies on the ERP components P150 and N170 which reflected differential processing for the stimulus material of the two subject groups.

   Therefore, it was concluded that action video gamers have enhanced attentional and visual processing capabilities and non-gamers normal. The action video gamers show different stimulus exploration (more centre vs peripheral) and use an enhanced early perceptual analysis of the stimulus material and hence, may detect changes in objects faster and learn the belonging of the stimuli to their categories earlier. Schenk and colleagues findings support the work by others (West) that action video gamers show enhanced counting and remembering of specific sequences, features and locations which could be an advantage when learning the exceptions in Schenk`s visual categorisation task. The authors concluded that there were two areas worthy of further exploration: the first, testing with other types of video games; and secondly, more in-depth study of differences in gender with this type of learning.

COMMENT

   What makes this article interesting is that supports the view that cognitive skills gained from performing one task can be applied advantageously to other tasks that are not direct imitations of the performed one. In Schenk and team`s experiments, expertise in action video gaming gained from long-term practice proved a benefit to individuals in their performance of a categorisation task which involved visual input, working memory, short-term memory and hand-eye coordination. Therefore, it appears that action video gaming which can be a solo activity and can be performed anywhere (provided the computer hardware and software are available) may give individuals who have limited access to training opportunities in social settings the chance to maintain or restore various cognitive skills.

   The experimental set-up of Schenk and colleagues involved general skills in attention, working memory, short-term memory and decision-making and more task-specific skills related to visual input (colour and location) and visual search.  Some cognitive skills were not required such as those relating to emotions (value, empathy), other senses and information-type factors such as temporal order and movement. In general, it was found that both subject groups demonstrated similar performance regarding short-term memory and decision-making. For example, both groups showed earlier responses to discount prototypical stimuli than exceptions which implies that the routine was learnt, the decision-making method decided upon and the categorisation grouping recalled independent of whether the individuals had prior gaming experience or not. The third block of tests appeared to be the turning point for both groups and therefore, it is likely that the advantages of prior gaming experience did not lie with these particular cognitive skills. The areas where the action video gamers gained an advantage from their training appeared to be directly and mainly related to the visual processing and perception areas (eg. capable of identifying the key colour segment characteristic as being more centrally located, higher fixations on the stimuli to support short-term memory demands), which lead to the indirect advantages of faster speeds at categorising and category learning. This comment focusses on these types of systems and mechanisms where the gaming experience appears to give an advantage.

  We begin by hypothesising whether the hierarchical physiological structure of the visual input pathway itself is a key point to the training improvement. Since the experiment was associated with colour segment recognition, it can be assumed that the forward sweep of the incoming visual information occurred in all subjects within 100msecs of the stimulus onset and retinal cone firing. This would be followed by the relevant firing of the retinal bipolar cells with their on/off centres and surrounds leading to greater discrimination of the colour at the next stage involving firing of cells in the retinal ganglion layer. Here, 95% of the cells are P-type ganglion cells, responsible for shape determination although some are sensitive to wavelength. The other 5% of cells, the nonM-nonP ganglion cells are sensitive to wavelength and have the on/off centre and surround type structure. The response to one wavelength in the centre can be cancelled by the response to another wavelength in the surround and this is where the colour signal becomes established in a firing pattern. This colour signal is passed further up the visual pathway to the complex structure of the lateral geniculate nucleus (LGN) located in the dorsal thalamus. Information about colour is transferred through the activation of appropriate nonM-nonP ganglion cells of the six layer LGN structure to the tiny neurons of the koniocellular layers, which lie just ventral to each parvocellular layer. The receptive fields of the cells of the koniocellular layers are also of a centre-surround type structure and have either light/dark or colour opponency and it is these cells that define colour (the blob pathway). It should be noted that input from both eyes is kept separate in this area through the layering structure, but as far as colour is concerned there appears to be no binocularity disparity. It may be that shadow/contrast and the other qualities of colour, brightness and saturation, may be eye-dependent though.

    Projections from the LGN lead to the V1 visual cortex and then colour determination seems to follow the psychologists WHAT pathway (P pathway or temporal pathway), which takes the ventral route into the infero-temporal lobe via the cortical regions V2, V4, (V8) and area IT. A quarter of the complex V1 structure deals with information from the retinal fovea, ie. 25% of incoming information is about colour and firing follows a characteristic retinotopic colour map which is maintained as the colour pathway continues upwards via the ventral stream to areas V2 and V4. Further discrimination of the colour signal occurs at the area IT (in the interior temporal lobe). This area is known to be stimulated by a wide variety of colours and abstract shapes and is said to be important for both visual perception and visual memory. A small patch of the area IT is particularly responsive to faces with some faces being more effective stimuli than others. Therefore, since the categorisation test was performed by all individuals successfully then it is likely that the hierarchical physiological structure of the visual pathways of the individuals were the same independent of whether they had gaming experience or not and the visual information and colours were inputted and perceived by the system described above. Deficits in eye sight, or particular brain area deficiencies (eg. lesions of the visual cortex) would have had an effect on categorisation performance and these reasons can be excluded since the individuals that took part in Schenk and colleagues experiments were all described as healthy and had perfect eyesight.

   Therefore, at which points of the visual input and processing could the gamers gain an advantage? One such area could be the determination and appropriation of colour to the segments. This relies on an increased capability to recognise colour and depends on the concept of colour constancy. The feature of colour appears to have debatable value when action is the individual`s response to the visual input. A degree of memory involvement occurs in the assessment of colour as colour is considered constantand the perception of it appears to be reliant on visual pathway detection and on previous experience and expectation. Colour constancy, attributed to certain cells in the V4 and hence, dependent on an individual`s V4 performance, is where there is a tendency for a surface or object to appear to have the same colour even when there is a change in the wavelengths contained in the lighting source, eg. an object does not appear ´redder` when seen in artificial light. This is because colour constancy is found to be influenced by top-down processes (Bloj and the retinex theory of Land) where individuals use their own knowledge to interpret incoming visual stimuli and so colour is defined according to an individual`s own interpretation according to their own perceptions and experience and not the true colour determined by the colour`s physical wavelength. It was found that the colour constancy factor was responsible for global (compared to the whole visual scene) and more relevant here to local contrast, which involves comparing the retinal cone responses from the target surface with those of the immediate background (Kraft). This is likely to apply to the experiments of Schenk and team where the target colours would be compared to the neighbouring colours. Therefore, colour constancy and the personal definition of colour may be an area where the gamers have an advantage. Action video games are fast and rely on the gamers identifying shapes quickly. Therefore, the gamers may have developed cognitive skills in fast assessment of colour according to their own definitions (their own individual measurement of colour constancy). This capability of fast perception could be applied to the categorisation task and lead on to more efficient and accurate formation of short-term memories of the target colours of the visual stimuli presented in the experiments. 

   Another skill that relates to the visual target that is likely to be increased in gamers theoretically and was actually shown by Schenk and team`s experiments is the increase in control of eye movement. This is shown by the change in fixation rates observed in Schenk and team`s experiments for those subjects with prior gaming experience and the observation supports work by others who show that eye movements can enhance sensitivity to the target (Ennis). The association comes from the coupling between eye movement and deliberate goal-directed focus (Walcher). In Schenk`s experiments this means that eye movement control is associated with the establishment of target colour and manifests as fewer saccades and longer fixations. The experiments of Schenk and team`s were made slightly easier for subjects since the target colours were in the centre of the shape and hence, focus of the visual field and not in the peripheral regions. Hence, their determination was maximised since most retinal cones of the first stage of the visual pathway are situated in the central point of the lens at the fovea where the light rays pass in a straight line through from the lens to the retinal cell layer. Therefore, gamers with their prior experience of keeping targets centralised in the focus of the visual field will optimise the firing patterns of the presented stimuli (Carrillo-Reid) whereas non-gamers are likely to be less focussed (searching central and peripheral) until they have learnt where to concentrate their focus. The advantage of keeping the target in the focus of the visual field may be even more specific in that targeting might be dependent on only a small proportion of the total segment available. This is supported by visual search hypotheses, which say in a large stimulus, parts of the stimulus are processed quicker than the whole (Kinchla). Therefore, the gamers could specifically target not only the central segment where they know that the target colour changes are located, but narrow it down to even a smaller proportion of that target area to reduce processing load.

 A topic linked to eye movement and one that may play a role in the categorisation task given by Schenk and team is that of visuomotor responses. It is possible that through their prior gaming experience, the gamers have developed faster eye-hand coordination. Once the visual process is stopped, recognition is made and the decision taken, the end result is a motor movement consisting of a key press action. Therefore, efficient and quick eye-hand coordination may extend the gamer`s visual advantage, ie. will process faster and carry out the required response faster. This improved eye-hand coordination may occur via strengthened connectivity between multiple cortico-cortical and cortico-subcortical frontal cortex networks (Brovelli) and the temporal locking of firing oscillations of the early visual processing with the early motor planning controlling the execution of the hand response (Tomassini). Therefore, training may give the gamers an advantage by speedier movements once the decision has been made.

   Gamers may also have an advantage in Schenk and team`s experiments by being able to maintain concentration and process relevant information faster during the learning phase and then shifting this to non-conscious processing at a later stage, albeit faster than the non-gamers. This type of capability would be gained by playing many hours of action video games where fast changing situations are constantly being presented and decisions have to be made. Gamers then when presented with a simpler categorisation task are likely to shift to non-conscious processing at an earlier stage than non-gamers and since non-conscious processing is faster than conscious, then response times of the gamers would be quicker. Conscious information processing which occurs early on in the trial blocks involves a number of skills relating to awareness of what is required and this conscious awareness is demonstrated by the N170 component (Thierry). Two cognitive processes come into play, working memory and attention, and these capabilities would have been honed to cope with the task demands by the prior gaming experience which could be considered as priming. From a physical perspective, priming would help the individual to target specifically (ie. only a proportion of the colour segment in the visual search part of the task) leading to improved cue utilisation for example. A greater level of cue utilisation leads to consistently greater response latencies consistent with strategies that maintain accuracy, but reduce the demands on cognitive resources (Brouwers). This means that priming would positively affect demands on attention and information processing relating to working memory performance.

   With regards to attention, the high level of conscious awareness particularly at the beginning of the trial blocks leads to higher levels of top-down attention which is likely to result in strong alpha brain wave activity in the fronto-parietal network (Van Schouwenburg). Various areas are responsible in the selection and maintenance of attention on stimuli eg. there is co-activation of neurons within 50-200ms across the anterior cingulate cortex and prefrontal cortex during stimulus selection in a spatial attention task (Oemisch); and firing in the ventral intraparietal sulcus maintains attention to a specific location (Capotosto). However, experience (ie. by prior gaming experience) would affect how targets are consciously selected and attentional selection would respond. This is supported by Corradi-Dell´Acqua who showed that the ability to select, within the complexity of sensory input, the information most relevant for the purpose of the task was influenced by both internal settings (ie. top-down control and conscious awareness) and the relevant features of the external stimuli (ie. bottom-up control and visual input). This capability to switch to relevant information comes from firing of the areas involved in top-down control (frontal, parietal and sensory cortices) and the lateral intraparietal neurons (Kumano). Lateral intraparietal neurons appear to accumulate relevant information depending on context to decide which eye movements to carry out to maximise it. The selection of the target would imply attended stimuli and this would enhance the evoked firing potentials for these stimuli in comparison to non-attended (Andersen). The effect of increased firing with reference to Schenk and team`s experiments is that colour (the target characteristic) would exhibit general sharpened selectivity by initial gain at the V1 level and within 100msec a sharper tuned profile in the posterior ventral cortex (Bartsch).

   This increased attention and selectivity of target afforded by conscious awareness and the ability of gamers to know faster what is required leads to working memory performance gain. Gaming experience, if considered as priming, would bring about benefits to working memory as seen through pre-stimulus alpha brain wave activity (Myers) and theta-band oscillations representing multiple predictions that are dynamically coordinated in time (Huang). The training of the individuals although unlikely to produce long-term changes in the physiological structure of the visual pathway are likely to induce changes to the physiological structures involved in working memory if they follow the pattern of other cases where training has been carried out prior to testing. For example, Caeyenberghs has reported that adaptive working memory training led to improvement on non-training related working memory tasks and tasks of reasoning and inhibition and these improvements were related to increased global integration within the fronto-parietal attention network observed by increased structural white matter network connectivity. Astle also reported that in children, after training, connectivity between the fronto-parietal networks and both lateral occipital complex and inferior temporal cortex areas exhibited increased strength of connectivity at rest. These were mirrored by an increased working memory performance.

    The problem with working memory is that it is of a limited capacity with regards to maintaining and manipulating the objects held in it (Myers) with resolution decreasing as the number of items increases above 3 (Anderson). This is in spite of the work load being spread over different areas (eg. specific brain areas have particular working memory functions such as post parietal cortex – manipulation of information and item maintenance; lateral occipital cortex – item maintenance and even different areas of the prefrontal cortex, with ventrolateral prefrontal cortex involved in face and vocal information; IT, V4 with visual information – Plakke). Therefore, conscious awareness and attention focused on relevant and targeted material aids the overall performance of working memory. The characteristics featured in object based attention are present in the working memory (Peters) and are strengthened (Soto) and sharpened by this overlapping (Lim). Therefore, attention selectively updates and maintains the relevant material (Blauracke, Heuer) required for the task at hand. This would mean that although both gamers and non-gamers would employ the same working memory processes the degree of efficiency of the former would be higher at least in the earlier trial blocks.

    Therefore, we can see that several areas of visual input and processing may be advantageously changed by training involving playing action video games. The value of playing action video games by increasing the selection of relevant material, maintaining it in working memory and then carrying out a response of a planned motor movement is significant for circumstances where these cognitive skills need to be maintained or improved and where the individual`s environment presents it as a valid learning option. The proviso is that the game used has to be challenging, capable of repetition without loss of motivation, set-up to give feedback preferably immediately and relevant to the required cognitive skill.

Since we`re talking about the topic…………………..

            …… achromatopsia sufferers have damage in the V2, V3 and V4 areas and therefore, can process information about colour implicitly but are unable to use the information explicitly when the judgement concerns colour. Therefore, if Schenk and team`s experiments were repeated would the advantages of the gamers be negated?

            …… serotonin administration causes a mainly multiplicative decrease of visual responses and a slight increase in stimulus-selective response latency (Seillier). Can we assume that if serotonin was administered prior to the test, both groups would suffer a loss in performance, but the gamers would still show the training advantage?

            …..……….mood is reported to have an effect on working memory performance with positive mood enhancing verbal and impairing spatial working memory, whereas negative mood enhances spatial and impairs verbal working memory (Storbeck). If Schenk and team`s experiments were repeated with anxiety levels increased for both groups, would we see any change in performance of the two groups in general particularly at block 3, the turning point between learning and automatic response?

Posted in Uncategorized | Leave a comment