Perineuronal nets are associated with decision making under conditions of uncertainty in female but not male mice

.


Introduction
Spatial learning and memory are a set of behaviours and processes through which information about our external environment is acquired, stored, organised and recalled to direct behaviour [1].Spatial assessments can be helpful when assessing neurodegenerative and neuropsychiatric diseases because impairments in spatial cognition tend to occur early and persist throughout the course of the disease [2,3].Therefore, to further understand the neuropathological disturbances and behavioural decline seen in patients with these diseases, a wide range of spatial learning and memory tasks have been developed for preclinical rodent models to investigate potential mechanisms and treatments to ameliorate behavioural outcomes.
In recent years, incorporating sex as a biological variable in study designs in basic and preclinical research has been requested by funding agencies due to the prevalent sex bias across scientific disciplines; wherein male subjects have been used as the default research subject.For example, in behavioural neuroscience, single-sex studies of male animals outnumber those using females 5.5 to 1 [4].It is a prevailing belief that using female subjects introduces too much variability in the data due to their oestrous cycle and the accompanying changes in hormone levels.Hodes and Kropp [5] contend this claim, suggesting the amount of variability that females exhibit across the cycle compared with males is non-significant and negligible.Moreover, they highlight as the field of medicine is moving to individualised treatment for stress and mood disorders in particular, incorporating sex as a biological variable is imperative to understand how behaviours may manifest differently in males and females [5].The existence of sex differences in brain biochemistry, physiology, structure, and function is widely accepted [6][7][8].By limiting the findings to one sex stymies our understanding of many aspects of neural function.
Despite this large persistent bias, there is compelling evidence in animal studies that there are many sex differences across a range of cognitive domains and decision-making behaviours.For example, in aversive tasks such as fear conditioning, many have shown reduced contextual fear conditioning in females, compared to males, as indicated by decreased freezing during the later retrieval of contextual fear memory (see review [9]).This reduction in contextual fear conditioning in females is in line with other evidence indicating reduced spatial-related learning and memory performance in comparison with males [10].
On spatial aversive tasks, such as the Morris water maze (MWM), Ciamadevilla and colleagues [11] found that while male and female Wistar rats both learn similarly to find the hidden platform over a four day learning period, female rats showed decreased memory retention compared to males on a probe trial, 6 and 12 h later.While this study indicated memory retention differences between the sexes, other studies show a difference in learning.A study by Rodriguez and colleagues [12] found subtle differences between males and females on learning in Trial 3, where males were faster to find the platform than females.However, they also had subsequent test trials with the location of the platform defined in terms of two sources of information, a landmark outside the pool and a particular corner of the pool.Testing revealed that females spent more time in an area of the pool that corresponded to the landmark, whereas males spent more time in the distinctive corner of the pool.The results suggest that males and females use different types of information in spatial navigation.This is in line with some literature suggesting that females more often use egocentric cues whereas males tend to use allocentric cues [13][14][15][16][17], yet this is often debated in the human and animal literature [18][19][20].On another aversive spatial task, the active place avoidance (APA) task, Cimadevilla and colleagues [21] found no differences on the APA task in adult rats but in a later article [22], in which they examined weanling rats on the APA task, they found that while both sexes learnt to avoid the foot shock over time; males had a significantly longer latency to the first shock earlier in development (postnatal day 23-24) compared to females (postnatal day [33][34].Furthermore, they found that males had a lower number of entrances to the shock zone earlier than females rats in development, with significant decreases on Day 3 compared to Day 1 and 2 while this was not seen until Day 5 in female rats [22].This suggest that males may have an advantage in spatial competence that develops earlier compared to females, however it is only transient on this task. Using aversive tasks is only one way to examine spatial learning and memory.Other tasks using appetitive stimuli have been developed and some tasks have investigated sex differences.One such task is the 8-arm radial maze, and studies have shown that males make fewer errors and receive more rewards compared to females [23,24].Groves and Burne [25] examined two appetitive tasks, the 5 choice-serial reaction task and the 5 choice-continuous performance task, in adult vitamin D (AVD) deficient BALB/c male and female mice.They found sex-dependent impairments in attentional processing revealing that male AVD-deficient mice were less accurate, took longer to respond when making a correct choice and were more likely to make an omission, without a change in the motivation to collect reward.By contrast, female AVD-deficient mice had a reduced latency to collect reward, but no changes on any other measures compared to controls.This work shows that dietary changes can influence performance in a sex-dependent manner.In an open-field tower maze, while male and female rats solved place and response learning at the same rate, female rats moved faster and had better long-term memory retrieval of place learning compared to male rats [26], and this difference was exaggerated when the rats were exposed to an acute stressor just prior to the task.This study strongly suggests stress impacts learning and memory ability on this task.
The appetitive touchscreen operant task can assess many types of memory including working memory and visual discrimination.For example, using the touchscreen reversal learning task in long-Evans rats, Gogos [27] showed that female wild type (WT) rats had a higher percentage correct responses compared to males.A study by Chen and colleagues [28] combined the touchscreen pairwise discrimination task with a two-armed multidimensional bandit task.They found that female mice acquired the correct image-value associations sooner than male mice [28].Chen et al. also examined the divergent strategies for learning in male and female mice.They found that female mice were more likely to constrain their decision-space early in learning by preferentially sampling the location rather than the type of image.Conversely, male mice were more likely to be inconsistent, changing their choice frequently and responding to the immediate experience of stochastic rewards.Moreover, they found that individual strategies were related to sex-biased changes in neuronal activation in early learning.However, sex differences specifically investigating spatial cognition using the touchscreens have not yet been investigated.
There are numerous sex differences reported in the morphology, plasticity and electrophysiological properties of neurons in the hippocampus and associated regions [10,[29][30][31], which could impact on spatial cognition.Studies have reported that there are sex differences in the morphology of granule neurons and CA3 pyramidal neurons [32][33][34].Juraska [33] found that male rats have greater dendritic intersections in granule neurons of the dentate gyrus compared to female rats, while in another study, Galea [34] found that female rats have greater branch points in the basal dendrites of CA3 pyramidal neurons compared to male rats.Other evidence suggests that long-term potentiation (LTP) can be different in males and females.For example, a recent study by Safari and colleagues [35], showed that male rats have greater LTP at perforant pathway-dentate gyrus (PP-DG) synapses in the hippocampus compared to females.Furthermore, other studies have shown that males exhibit enhanced early and late-LTP compared to females in the dentate gyrus, CA3 and CA1 regions [36][37][38].Other evidence suggests there are sex differences in adult neurogenesis.For example, a study by Yagi and colleagues [39] found that adult-born neurons matured faster in males compared with females.They also found that in males, there was a greater reduction in neurogenesis in the weeks following mitosis, whereas the levels were not altered in females.The authors suggest that this difference could result in an advantage in males to respond to external stimuli.While the hippocampus dominates the spatial memory literature, in a model of systems consolidation, other neocortical regions are also important in aiding the adequate encoding and storage of memories.The retrosplenial cortex (RSC) is becoming increasingly recognised for its role in supporting hippocampal-dependent spatial memory and the integration of allocentric and egocentric cues, making it a region of interest in neural pathways that underlie spatial cognition (see Reviews [40][41][42][43][44]).
Recently, perineuronal nets (PNNs) have become a focus of investigation as they are involved in modulating brain plasticity and learning and memory.PNNs are part of the extracellular matrix that ensheath the soma and proximal dendrites of neurons and are postulated to protect and stabilize afferent synapses [45,46].In the cortex, PNNs are predominantly associated with parvalbumin-expressing (PV+) fast-firing GABAergic interneurons especially in the rodent neocortex [47,48]; however, PNNs also surround glutamatergic neurons in the CA2 of the hippocampus, amygdala and deep cerebellar nucleus [49,50] and other neurons such as glycinergic output neurons in the medial nucleus of the trapezoid body (MNTB) at the calyx of Held synapse [51].Despite this varied composition, the notion that PNNs provide both structural and functional support and moderate synaptic plasticity is heavily supported, with examples showing that removal of PNNs alters several electrophysiological features of PV interneurons and nearby pyramidal neurons (see review [52]) which impacts LTP [53], LTD [54] and brain oscillations [52,53,55], ultimately affecting learning and memory processes (see reviews [56,57]).
Importantly, some studies have found evidence that the expression of PNNs may be sexually dimorphic.For example, Ciccarelli et al. [58] compared PNNs in the subcortical structures of male and female mice and revealed a significant sexually dimorphic expression, with expression found prominently on estrogen receptor expressing neurons in the medial amygdala.In addition, Griffiths [59] reported that there were more PNNs in the CA1 in males compared to females, yet this was not the case in the CA3 nor the adjacent neocortex.Taken together, these findings suggest that there may be a sex-dependent role of PNNs in regulating neural plasticity.
Therefore, the aim of this study was to test female and male mice on P. Mayne et al. aversive and appetitive spatial learning and memory tasks and examine the expression of PNNs and PV+ in regions associated with spatial memory.Based on the above literature, we expected that there would be a sex difference on the appetitive but not aversive task and this would be reflected in hippocampal PNN composition.

Subjects and housing
Behaviourally naïve adult (3 months) male and female BALB/c mice (N = 38) were used (Animal Resources Centre, WA).Mice were housed in groups of 2 or 4 in individually ventilated cages (Optimice) within the QBI PC2 animal facility with 12-h light/dark cycles (lights on 0700 h).Water and standard mouse pellets were available ad libitum and cages stored in a room maintained at 21 ± 2 • C and approximately 60% humidity.All procedures were performed with the approval from The University of Queensland Animal Ethics Committee (QBI/088/18), under the guidelines of the National Health and Medical Research Council of Australia.

Experimental design
Mice were randomly assigned into two behavioural cohorts with equivalent numbers of females and males.For the APA task, there was a total of 24 mice (female = 12, male = 12), and standard mouse pellets for food were available ad libitum.For the TUNL cohort there was a total of 14 mice (female = 7, male = 7) that were initially given access to mouse pellets ad libitum.However, the mice were then food restricted to maintain 85-95% of free feeding body weight from 2 weeks prior to behavioural testing, and this was maintained throughout the experiment.

Cognitive measures 2.3.1. Active place avoidance
2.3.1.1.Apparatus.The APA apparatus (Bio-Signal Group) consisted of an elevated arena with a metal grid floor (diameter 77 cm) fenced by a 32 cm-high transparent circular boundary.The stage rotated clockwise (1 rpm) and an electric shock could be delivered through the grid floor.This task required mice to learn to avoid a shock zone consisting of 60 • angle on the platform using four external cues hanging on the nearby walls.The shock zone was defined within a 60 • region of the stationary room and kept constant in relation to the room coordinates.The position of the mouse was tracked by PC-based software that analysed images from a camera installed on the top of the apparatus and delivered shocks as required (Tracker, Bio-Signal Group Corp., NY).Entrance into the shock zone directed the delivery of a brief constant mild electrical foot shock (0.5 mA, 60 Hz, 500 ms).If the mouse remained in the shock zone, it received additional shocks of the same intensity at 1.5 s intervals until the animal moved out of the area.

Task.
Mice were trained over a 6-day period.Mice were habituated for 5 min without the shock zone enabled on Day 1.Each subsequent day, separated by a 24-hour period, mice underwent one testing session, which lasted for 10 min.Using Track Analysis software (Bio-Signal Group), the data were analysed according to three main outcomes; "number of shocks received", "time to first entry", and "maximum time to avoid the shock zone".

Trial unique nonmatching-to-location
2.3.2.1.Apparatus.Touchscreen operant chambers (Campden Instruments Ltd.) were used and details have been described previously [60].The apparatus comprised of black Perspex sidewalls and a trapezoidal shaped floor with a touchscreen on one side and a reward delivery magazine positioned opposite.A black Perspex mask with five response windows (each comprised of a 4 ×4 cm square aperture, 1.5 cm above the grid floor) was positioned in front of the touchscreen to minimise incidental touches.Infra-red (IR) beams were positioned near the screen and near the magazine to capture motor activity.Strawberry-flavoured milk was provided as a liquid food reward (Breaka, QLD).The operant chamber was placed inside a sound-and light-attenuating box with a house light, a tone generator, a ventilating fan and an IR camera.ABET software and Whiskers by Campden Instruments Ltd. was used to control the system and collect the data.
2.3.2.2.Task.TUNL trials consisted of sample, delay and choice phases.In the sample phase, an initiation by making a head entry into the magazine triggered the illumination of one response window.Once the mouse responded with a nose poke, the stimulus was removed, and the delay period began.Following the delay, another head entry to the magazine triggered the presentation of the choice stimulione square in the same location as the sample phase and one square in a new location.A response to the new location (non-match, correct response) resulted in the delivery of reward, and the ITI began which was followed by the initiation of a new trial.A response to the old location (match, incorrect response) resulted in the repetition of the same trialreferred to as a reminder trialuntil a correct response was made.
Acquisition of the TUNL task occurred in two stages.In Stage 1, the sample and choice stimuli were only presented in non-centre locations.First, the choice stimuli were separated by three blank response windows (S3).Once criterion was reached, the level of separation was reduced to two blank response windows (S2), and then one window between the choice stimuli (S1).The task parameters were as follows: 2 s delay, 5 s ITI, 5 s correction trial ITI, reward given for every third sample stimulus touch, criterion for advancement was 70% accuracy for two consecutive days, and session completion either after 100 test trials (not including correction trials) or a maximum of 30 min Stage 2 differed in that the sample and choice stimuli could appear in the centre location.The choice stimuli were separated by one blank response window.Stage 2 training was conducted with the same parameters as Stage 1, except with no delay and there was no reward for sample touches.
There were two manipulations following task acquisition: Separation and Delay.In the Separation manipulation, the level of separation between the sample and choice stimuli was randomised and varied between 1 window (S1) and no window of separation (S0).In the Delay manipulation, the separation between the sample and choice stimuli remained constant at 1 blank window, while the time between the initiation nose poke to the display of the choice stimuli was randomised and varied between 0 s (D0) and 2 s (D2).
Outcome measures on each manipulation were the percentage of correct and incorrect trials, the number of correct and incorrect trials and the time to collect a reward for the correct trials.

Tissue processing 2.4.1. Perfusions and sectioning
Following behavioural testing, mice were anaesthetised by an intraperitoneal injection of sodium pentobarbital (100 mg/kg) and perfused transcardially with 0.1 M phosphate buffer solution (PBS), followed by a fixative containing 4% paraformaldehyde (PFA) in PBS (pH 7.4).The brain was dissected out and post-fixed in 4% PFA at 4 • C for 24 h.Brains were preserved at 4 • C in 0.1 M PBS solution containing 0.05% sodium azide followed by paraffin embedding.The brains were sectioned coronally at a thickness of 10 µm using the Lecia rotary microtome.
P. Mayne et al.

Image observation and cell segmentation
Fluorescent samples were viewed and captured with a Zeiss Axio Imager Z1 microscope using a 20x objective.Image preparation and brain regional determination was processed using several ImageJ custom macros.The agranular (RSA) and granular (RSG) regions of the RSC and subregions of the hippocampus were segmented according to the Allen Reference Atlas -Mouse Brain [brain atlas].Available from atlas.brain-map.org.Background subtraction was applied using a math correction algorithm in ImageJ: setMinAndMax(med, (mean+(x * std)y) where x is an integer value to raise the maximum intensity and y is a pixel value to subtract unwanted background.PNNs and PVs were segmented individually using a custom built deep neural network (DNN) using the Cellpose interface [61] and cell subtype analysis was analysed in CellProfiler (see Fig. 1).PNN and PV cells were only included in the analysis if they were also positive for DAPI (a nuclear marker).Cell number was then normalised to area (number/mm 2 ).

Data analysis
A repeated measures ANOVA was performed with appropriate within-and/or between-subject factors.Violation of sphericity assessed by Mauchly's test was corrected by the Greenhouse-Geisser method.When an interaction was found, simple main effects analyses were conducted for each factor with the Sidak correction.For the behavioural analysis on TUNL, the outcome of the trials (correct or incorrect) was coded.We analysed each outcome measure separately and used a general linear model with Separation/Delay (2 levels) as the within-subjects variable and Sex as the between-subjects variable.This was performed for the Percentage of responses, number of trials and time to collect the reward.To further characterise the relationship between PNN/PV expression and behaviour we used step-wise linear regression.
All statistical analyses were conducted using SPSS version 27.Graphs were created in Prism version 9 and graphical results are expressed as mean ± standard error of the mean (SEM).Values of p < 0.05 were considered statistically significant.

APA behaviour
Over the five days of APA learning trials, there was a significant main effect of Trial for the time to first entry (F (3, 60) = 9.1, p < 0.001; see Fig. 2B), such that the time to first entry increased across trials.Furthermore, there was a significant main effect of Trial on the maximum time to avoid the shock zone (F (3, 60) = 15.3, p < 0.001; see Fig. 2C), such that the maximum time to avoid the shock zone increased across trials.Finally, there was a significant main effect of Trial on the number of shocks (F (3, 60) = 20.7,p < 0.001, see Fig. 2D) such that the number of shocks decreased across trials.There was no main effect of Sex and no Trial x Sex interaction for any outcome measure, suggesting that all animals learnt the task and there were no differences between male and females on this task.

PNN and PV expression in the APA cohort
A two-way ANOVA was performed on the cell segmentation and revealed no main effects of Sex on the area of each region, or for percent of PNN cells with a PV (%), or percent of PV cells with a PNN (see Fig. 3).We also examined the numbers of PNN+PV+ , PNN+PV-, PNN-PV+ cells, and the intensity of PNN, PV or PNNs that surround a PV (see Fig. 3), however no main effects of Sex were revealed on any of these measures.

TUNL behaviour 3.2.1. Number of sessions to criteria
A two-way repeated measures ANOVA revealed that there was a main effect of Stage on the number of sessions to criteria (F (4,48) = 22.8, p < .0001),such that in Stage 1, the number of sessions to criteria was reduced in Separation 3 (M = 7.14) compared to Separation 1 (M = 11.36),p = 0.047, and in Separation 2 (M = 5.3) compared to Separation 1 (M = 11.4),p = 0.002, see Fig. 4A.However, the number of sessions to criteria was not significantly different between Separation 3 and Separation 2. In Stage 2, the number of sessions to criteria was significantly increased in Separation 1 (M = 17.6) compared to Separation 0 (M = 5.5), p < 0.001.There was no main effect of Sex nor a Stage x Sex interaction on the number of sessions to criteria at any stage.

Weight restriction
To rule out weight loss as a motivational factor to receive a reward, we also examined the percentage of free feeding weight (FFW) across all weeks on food restriction.A two-way ANOVA revealed no main effect of Sex on the percentage of FFW across all weeks of food restriction, see Fig. 4B.
We also found there was a significant main effect of Delay for the percent of Incorrect (F (1,12) = 97.6,p < 0.001) and Correct (F (1,12) = 89.9,p < 0.001) outcomes.Main comparisons revealed that there was a higher percentage of incorrect responses for D0 compared to D2 (mean difference = 19.5,SE = 2.0), and a higher percentage of correct responses for D0 compared to D2 (mean difference = 5.4,SE =.9), suggesting that D2 was perceived as more difficult.
For the reward collection latency, there were no main effects nor interactions at any level suggesting that there was no difference in time taken to retrieve the strawberry milk reward between sex across the delay trials (see Fig. 5G).

Separation manipulation: effect of separation on performance
Using the trial by trial data, there was a main effect of separation for the percentage of responses for Incorrect (F (1, 12) = 16.0,p = 0.002) and Correct trials (F (1, 13) = 14.4,p = .003),and main comparisons found that there was a higher percentage of incorrect responses for S0 compared to S1 (mean difference = 11.3,SE = 2.8), and a lower percentage of correct responses for S0 compared to S1 (mean difference = − 12, SE = 3.2).This suggests that the animals found it harder to distinguish the correct response in S0 trials compared to S1 trials.There were no other significant interactions.
There was no main effect of Sex nor a Separation x Sex interaction for any of the outcome responses.As a measure of motivation, we assessed the reward collection latency.There was a main effect of separation for reward collection only on the Reminder trials (F (1, 12) = 11.5, p = .005),such that mice were faster at collecting the reward on S0 trials compared to S1 trials (mean difference = − 0.07.SE = 0.02).Importantly, there was no main effect of Sex nor a separation x Sex interaction suggesting there was no difference between females and males to retrieve the strawberry milk reward between novel correct or repeated correct trials (see Fig. 5H).

PNN and PV expression in the TUNL cohort
PNN cell segmentation analysis, carried out by a two-way ANOVA, revealed no main effect of Sex on the area of each subregion (see Fig. 6A).However, there was a significant Sex x Region interaction for PNN mean grey value (F (1,12) = 3.3 p = <0.0001)and simple effects tests suggested that females had a higher PNN mean grey value compared to males in the RSA (p = 0.004) and RSG (p = 0.003) (see Fig. 6G).Similarly, there was no effect of Sex on the % of PV cells also positive for a PNN, but there was a Sex x Region interaction for PV mean grey value (F (5,60) = 6.3, p < 0.001; see Fig. 6H).Simple effects tests revealed that females had more intense PV cells in the RSA (p = 0.002) and RSG (p = 0.02) compared to males.
We also examined the subtypes of PNN cells, examining counts of PNN+PV+ , PNN+PV-, PNN-PV+ and the intensity of PNNs that surround a PV and intensity of PV cells that have a PNN.There were no significant differences between sexes on counts of PNN+PV+ , PNN+PV-or PNN-PV+ (See Fig. 6D,E,F), however, there was a significant Sex x Region interaction on the intensity of PNNs surrounding a PV cell F (5,58) = 5.62 p < 0.001 (Fig. 6I), such that females had more intense PNNs in the RSG and RSA compared to males.With respect to exploring the relationship between PNN and PV expression and behaviour we found that there were only two significant variables fitted in the step-wise linear regression model.The intensity of PNNs around a PV in the RSG Intensity predicted the percent correct responses at D0.The number of PNNs in the RSA predicted the percent correct at D2.These findings suggest that higher expression of PNNs in the RSC was associated with poor performance on the delay manipulation.There were no other significant variables fitted in the model.

Discussion
The aim of this study was to test female and male mice on aversive and appetitive spatial learning and memory tasks and examine the Fig. 3. Cell segmentation results for female and male mice in the APA cohort shown for hippocampal subfields and the RSC: A Area (mm 2 ) of each region, B Percent of PNN+ cells also positive for PV, C Percent of PV cells also positive for PNN, D PNN+PV+ count/mm 2 , E PNN+PV-count/mm 2 , F PNN-PV+ count/mm 2 , G PNN mean grey value for cells positive for PNN and PV, H PV mean grey value for cells positive for PNN and PV, I Intensity of PNN surrounding PV cells.All cell segmentation analyses were performed using a two-way ANOVA, and there were no main effects of Sex for PNN and PV expression.expression of PNNs and PV+ in regions correlated with spatial memory.Our main finding was that female and male BALB/c mice perform similarly on spatial learning and memory tasks.There were no sex differences on any outcome measure of the aversive APA task which supports previous findings in adult rats [21].By contrast, on the appetitive task (TUNL), female mice had a higher percentage of incorrect trials and a lower percentage of correct trials compared to male mice on the delay manipulation of the task, despite no sex differences in the separation manipulation of the task.We also found that females had significantly higher intensity of PNN and PV expression in the RSC than males, which was associated with incorrect responses on the delay manipulation.
Reports of sex differences in appetitive tasks are mixed.Mishima and colleagues found that adult male ddY mice performed better than females in two appetitive tasks (a lever press task and an 8-arm radial maze task), but this was only evident in the acquisition of the task [23].However, on appetitive touchscreen tasks, Gogos et al. [27] showed that females had higher percentage of correct responses compared to males and Chen et al. [28] found that females acquire the task faster than males.The current findings showed that while both sexes were able to acquire the task, males had more correct responses compared to females on a manipulation of the task with varied delay times before showing the choice stimuli.Other factors that could contribute to the observed differences between males and females could be due to non-cognitive factors such as stress or task difficulty.
Rodent research indicates that stress can impact cognitive ability.While the APA is an aversive task with a shock component, Lesburguères et al. [62] suggests that this task is no more stressful than exploring a familiar environment.Although touchscreen-based tasks are considered to be a source of environmental enrichment and also contain a reward component, they are complex, require several weeks of training prior to manipulations and can be stressful.A study in mice examining the effects of regular touchscreen training on adrenocortical activity found significantly increased corticosterone levels in anticipation of training [63].Another study from the same authors used a comprehensive testing battery and found a pronounced increase of faecal corticosterone and anxiety-like behaviour in touchscreen tested mice which persisted even after the termination of training [64].Yet as these authors noted, the touchscreen training mice were removed from their home cages for a longer time span than the comparison mice and were trained in another room, introducing a potential confounds that could explain the results.Importantly, some research indicates that females show cognitive resilience to chronic stressors which impair male cognitive function [65].Considering this, and the fact that the Mallien and Krakenberg studies only used male mice, it is plausible that female mice might respond differently to daily touchscreen testing; both in the production of corticosterone levels and to the cognitive demands, where it is possible that chronic stress might be cognitively protective.If this is the case, it is unlikely that chronic stress associated with daily testing explains the differences seen on the TUNL task.
An important suggestion proposed by Coluccia & Louse [19] is that sex differences tend to appear only when the task is difficult.This is also supported by data in rats with large cortical lesions that were able to solve a simple task in the Lashley Type 1 maze, but not the more complex tasks in the Lashley Type 3 maze [66].Spatial tasks high in cognitive demands are more often than not accompanied by sex differences, while orientation tasks low in cognitive demands are not [19].By contrast, a study in rats using a punishment risk task and an approach avoidance task demonstrated that sex differences may emerge in goal-directed behaviour when there is a threat of punishment, regardless of cognitive load [67].Although Chowdhury et al. [67] did not find sex difference during the acquisition or expression of reward motivated behaviour, sex differences differed if the punishment was unpredictable, in which females were more sensitive than males, or certain, whereby females were less sensitive than males.The present results provide some support for this hypothesis, which could also be a reason why we see differences in the delay manipulation of TUNL, under more uncertain conditions, but not in the separation manipulation, or in the APA task.
More specific to the TUNL task is the degree of uncertainty induced by the reward delay.Outcome uncertainty mediates attention, learning and decision-making [68].Our results showed that females have more incorrect and less correct trials on both the 0 s and 2 second delay.Studies specifically examining performance on tasks with varied delayed durations suggest that delay tasks may manifest a degree of uncertainty in the animals' responding.Park et al., [69] investigated the neural activity of the mPFC on mediating working memory depending on the predictability of the delay duration (fixed or random).They found that mPFC neurons conveyed higher working memory information under the random-delay (between 1-7 s) than fixed-delay (4 s) conditions, possibly due to a higher demand for stable working memory maintenance.They concluded that in random delay conditions, there was an increased level of uncertainty for when the delay will be terminated, which ultimately made it difficult for the animals to ready a behavioural response.In the current study, the delay periods (0 or 2 ss) were randomised throughout the trials.Therefore, it is possible that the mice experienced periods of uncertainty as to when to initiate a behavioural response which may have resulted in behavioural sex differences.
Moreover, when the uncertainty and unpredictability is high, the behavioural policy can switch from exploitation to exploration in order  to obtain a model of the environment [68].This results in an increase in exploration noise [70][71][72].Pisupati et al. [72] assessed a multisensory decision-making task in rodents, and used computational modelling to show that lapses in decision making occurs on conditions with higher perceptual uncertainty.Interestingly, Chen et al. [70] used computational modelling to characterize sex differences in a canonical explore/exploit touchscreen task (the restless two-armed bandit task).Although male and female mice had similar accuracy, they used different explore-exploit strategies during the task; female mice explored more than males but learned more quickly during exploration.Therefore, in situations where the outcome is uncertain, animals may continue to explore to gather more information even after they have gathered enough information to make a decision.In the current study it is possible that the delay increases the degree of uncertainty of reward delivery which results in this exploration/exploiting trade-off, with females exploring more to obtain a more accurate model of the environment.
We found that females exhibited more intense PNNs in the RSA and RSG, compared to males, and interestingly, this was only evident in the TUNL but not the APA cohort.We also found that the number of PNNs in the RSA predicted the percent correct responses after a 2 s delay (D0), and the intensity of PNNs around a PV predicted the percent correct responses after a 0 s delay (D0).There was no relationship between behaviour and PNN or PV intensity in any hippocampal subregion.The increased intensity of the PNN in these regions in females may serve as an innate biological trait or potential endophenotype that could predict behaviour of this task.The current hypothesis in the PNN field is that increased WFA intensity corresponds to a mature PNN with decreased capacity for plasticity [73].Furthermore, there is a rapid body of work showing that changes in PNN number and/or intensity can be modulated during adulthood as a result of physiological stimuli, such as stress [74], exercise [75], environmental enrichment [76], diet [77] and circadian rhythms [78] and can correlate with various behaviours [79].Moreover, some studies report that PNN intensity can occur in the absence of altered PNN number in both humans and rodents [80][81][82].Given the enriching nature of the TUNL task, it is quite possible that PNNs could be enhanced across multiple regions and there could be a sex-dependent effect.Future studies would need to examine behaviourally naïve controls to better understand these findings.Moreover, the histology of the two cohorts of mice were conducted in different immunohistological staining experiments which precludes direct comparisons between cohorts.Future studies should investigate whether exposure to the TUNL task increases the expression of PNNs and whether this is dependent on sex.
Females in the TUNL cohort also exhibited more intense PV cells in both the RSG and RSA compared to males.Extensive literature promotes a role for PV+ interneurons in learning and memory [83].Donato and colleagues [84] showed that shifts in immunoreactivity of PV+ cells can predict synaptic plasticity and memory retrieval in an adult rat.For example, they found that the low-PV-network configuration in the CA3 (PV+ cells with lower immunoreactivity, measured by intensity) had enhanced structural synaptic plasticity and increased memory consolidation and retrieval for the novel object recognition task, whereas these were reduced in the high-PV-network configuration.Moreover, Lupori et al. [85] showed a coupling mechanism between PV and PNN such that more intense PVs cells are more likely to have an intense PNN.As PNNs are also known to restrict plasticity [86], this could suggest two distinct network configurations; one permissive toward plasticity (low-PV-PNN) and another that limits plasticity (high-PV-PNN).The TUNL task is complex and previous studies have implicated the hippocampus and prefrontal cortex in this task.The RSC is becoming well known in the learning and memory literature in supporting hippocampal encoding and updating [41]; it receives both excitatory [87] and long-ranging GABAergic monosynaptic inputs from the hippocampus [88,89], which are both essential for memory formation, and there's evidence showing that the synchronous activity between the RSC and the hippocampus is essential in recall tasks [90].High-PV-PNN configurations in the RSC, could restrict plasticity and impair behaviour on the TUNL task.Future research should investigate whether PNN degradation in the RSC impacts learning and memory on this task.
Sex differences in spatial ability has been a point of controversy in the literature, with a manifest superiority of males, which has led to a poor representation of females in the field of decision-making neuroscience.The data from this study shows that female and male BALB/c mice perform at similar levels on both the APA and TUNL tasks.However, under conditions of uncertainty, during the delay manipulation, females have more lapses on the TUNL task compared to males.Furthermore, the expression of PNNs and PVs in females tested on the TUNL, but not the APA task, were enhanced compared to males in the RSA and RSG, subregions of the retrosplenial cortex.An open question that remains is whether PNNs might be involved in tuning exploration noise on this task.

Fig. 1 .
Fig. 1. A. Schematic example of region segmentation.B. Shows examples of cell types in the RSC and CA1 regions: grey arrowhead indicates PNN+PV+ cell; black arrowhead shows PNN + PV-cell and solid white arrowhead indicates PNN-PV+ cell.

Fig. 2 .
Fig. 2. Sex differences in behaviour on the APA task.A Schematic of the APA task.Data from female and male mice are shown for B time to first entry, C maximum time to avoid the shock zone, D number of shocks received.Behavioural analyses were performed using a repeated measures ANOVA.No shocks were delivered during the habituation (Hab) trial.

Fig. 5 .
Fig. 5. Sex differences in behaviour on the TUNL task.A Schematic representation of the delay manipulation and B separation manipulation.Data from male and female mice are shown for C the percentage of trials for the delay manipulation and D separation manipulation E the number of trials for the delay manipulation F and separation manipulation G time to collect the reward for the delay manipulation H and separation manipulation * indicates a main effect of sex, p < .05.

Fig. 6 .
Fig. 6.Cell segmentation results for male and female mice in the TUNL cohort are shown for hippocampal subfields and the RSC: A Area (mm 2 ) of each region, B Percent of PNN+ cells also positive for PV, C Percent of PV cells also positive for PNN, D PNN+PV+ count/mm 2 , E PNN+PV-count/mm 2 , F PNN-PV+ count/mm 2 , G PNN mean grey value for cells positive for PNN and PV, H PV mean grey value for cells positive for PNN and PV, I Intensity of PNN surrounding PV cells.All cell segmentation analyses were performed using a two-way ANOVA, and there were significant main effects of Sex for PNN and PV expression in the RSA/ RSG, * p < 0.05.