Bank Voles Show More Impulsivity in IntelliCage Learning Tasks than Wood Mice

—Impulsivity is a personality trait of healthy individuals, but in extreme forms common in mental disorders. Previous behavioral testing of wild-caught bank voles and wood mice suggested impulsiveness in bank voles. Here, we compared behavioral performance of bank voles and wood mice in tests for response control in the IntelliCage. In the reaction time task, a test similar to the ﬁve-choice serial-reaction time task (5CSRTT), bank voles made more premature responses. Impulsivity in the reaction time task was associated with smaller medial habenular nucleus in bank voles. Additional tests revealed reduced behavioral ﬂexibility in the self-paced ﬂexibility task in bank voles, but equal spatial and reversal learning in the chaining/reversal task in both species. Expression of immediate early gene Arc after behavioral testing was low in medial prefrontal cortex, but high in hypothalamic supraoptic and paraventricular nucleus in bank voles. Wood mice showed the opposite pattern. Numbers of Arc-positive cells in the dorsal hippocampus were higher in bank voles than wood mice. Due to continuous behavioral testing (24/7), associations between behavioral performance and Arc were rare. Corticosterone measurements at the end of experiments suggested that IntelliCage testing did not elicit a stress response in these wild rodents. In summary, habenular size diﬀerences and altered activation of brain areas after testing might indicate diﬀerently balanced activations of cortico-limbic and cortico-hypothalamic circuits in bank voles compared to wood mice. Behavioral performance of bank voles suggest that these rodents could be a natural animal model for investigating impulsive and perseverative behaviors. (cid:1) 2022 The Author(s


INTRODUCTION
Deficient response control, apparent as impulsive, compulsive or perseverative behaviors are characteristic for psychiatric conditions such as obsessive-compulsive disorder (OCD), attention deficit hyperactivity disorder (ADHD), or addiction (Fineberg et al., 2010;Milad and Rauch, 2012;Bari and Robbins, 2013). While an excessive lack of inhibitory control is disadvantageous, impulsiveness is part of the normal behavioral repertoire of healthy individuals and the presence of impulsive individ-uals within a population might be advantageous for the community (Dickman, 1990;Harnishfeger and Bjorklund, 1994). In previous experiments, we investigated aspects of learning and memory in wild and laboratory rodents (Galsworthy et al., 2005;van Dijk et al., 2019). In one of the wild rodents, the bank vole, we noticed impulsive behavior in many individuals. The behavioral tests were conducted using an automated behavioral phenotyping system called IntelliCage (Galsworthy et al., 2005;Kiryk et al., 2020). The IntelliCage contains four operant chambers giving access to water. Nosepokes towards the water bottles are registered by the system, and we observed that bank voles made more nosepokes than other rodents. Repetitive nosepoking in the IntelliCage has been seen in laboratory rodents as an experimental outcome and was found to be an index of impulsive or perseverative behavior (Endo et al., 2012;Oizumi et al., https://doi.org/10.1016/j.neuroscience.2022.11.011 0306-4522/Ó 2022 The Author(s). Published by Elsevier Ltd on behalf of IBRO. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). 2020). The repetitive nosepoking in bank voles in our previous experiments led to the hypothesis that impulsive or perseverative behavior might be a natural behavioral phenotype of bank voles. To investigate this hypothesis, we behaviorally tested wild-trapped bank voles in comparison to wood mice in the IntelliCage in tasks requiring response control, thus testing aspects of impulsive action, inflexibility and compulsivity/perseveration (Fineberg et al., 2010;Bari and Robbins, 2013;Herman et al., 2018). Behavioral experiments included the reaction time task (Kobayashi et al., 2013), a test conceptually similar to the five-choice serial-reaction time task (5CSRTT), where a waiting period between a first nosepoke and a light stimulus signaling water access was used to measure response inhibition (action restraint). In the selfpaced behavioral flexibility task (Endo et al., 2011) and chaining/reversal task (Kobayashi et al., 2013), the operant chambers had to be visited in specific spatial or spatio-temporal sequences. In both tests, the sequence of correct chamber visits changed during the test, thus repetitive entries into previously rewarded chambers were considered as perseverative responses reflecting behavioral inflexibility. In addition, an approach-avoidance test was run and general behavioral measures were assessed in the adaptation phases. As possible neuroanatomical correlates of the behavioral patterns, we measured the volume of the medial habenular nucleus (mHb), as previous studies have shown impaired response control in the reaction time task and 5CSRTT after ablation of mHb (Lecourtier and Kelly, 2005;Kobayashi et al., 2013). Furthermore, endocannabinoid receptors in the habenula are involved in impulsive behavior in the 5CSRTT (Zapata and Lupica, 2021), and neonatal lesions of the habenula leads to an ADHD-like phenotype with increased impulsivity (Lee and Goto, 2011;Lee et al., 2021). The role of the habenula in motor control and cognitive behaviors is well documented (Hikosaka et al., 2008). The lateral habenula reciprocally links the ventral striatum (Nucleus accumbens), lateral hypothalamic and limbic areas with monoaminergic structures such as raphe nuclei and ventral tegmental area (Lecourtier and Kelly, 2007) while the mHb has a more targeted input from the limbic system (septal nuclei) and projects to the interpeduncular nucleus (Kobayashi et al., 2013;Viswanath et al., 2014). In addition, we performed immunohistochemistry for the immediate early gene Arc to identify brain regions activated by behavioral testing. Arc signal analysis focused on brain areas that have been implicated in the top-town circuit model of impulse control, that is medial prefrontal cortex (mPFC), hippocampus and hypothalamus (Dalley et al., 2011;Munakata et al., 2011;Noble et al., 2019). Lastly, we measured blood corticosterone levels as a general readout for potential stress in wild rodents imposed by the experiments.

EXPERIMENTAL PROCEDURES Animals and ethics statement
Forty-six animals (23 bank voles (Myodes glareolus, formerly Clethrionomys glareolus, family Cricetidae, Fig. 1A), and 23 wood mice (Apodemus sylvaticus, family Muridae, Fig. 1B) were included in the study (Table 1). Two wood mice and eight bank voles were either not tested behaviorally or excluded during the course of experiments (exclusively bank voles) because they did not reach a minimum drinking activity considered compatible with animal welfare (100 licks/day). These ten animals were used as controls for neuroanatomical and hormonal analysis. Animals were either trapped wild (30 animals) or were the offspring of wild trapped pregnant females (seven female and nine male bank voles). Only adult animals with body weights >18 g were used for behavioral testing. Trapping and animal experiments were conducted under the Canton of Zurich veterinary office permit #27034.

Trapping, housing and habituation
Animals were trapped within a four-week period of a summer at different locations in the Canton of Zurich using Sherman live traps (detailed description see Wiget et al., 2017). Animals were weighed, treated for ectoparasites (FrontlineÒ), sexed and aged according to body weight and season. In the laboratory, animals were given one to two months to habituate to the new environment and for pregnant females to give birth and raise pups before behavioral experiments started. Animals were kept in cages either alone (pregnant females) or in same-sex groups of two to four animals. Cages were equipped with bedding, grass, hay and paper tissues. Water and commercial mouse food containing Ivermectin (IvomecÒ) were given ad libitum, small quantities of fruits, nuts and seeds were given daily when checking the animals. Habituation took place in the same room as behavioral testing, with a minimum room temperature of 17°C and controlled light cycle (light on from 7:00 to 19:00). Three weeks before the start of experiments, animals were injected with a radio-frequency identification (RFID) transponder (Planet IDÒ GmbH, Germany) under isoflurane inhalation anesthesia (5% isoflurane, 0.7 l/min oxygen). Six days before experiments, animals were assigned into five groups of six to twelve animals each according to species, sex and approximate age and group-housed in a multicage system connected by sealable acrylic glass tubes (van Dijk et al., 2019). Before the start of experiments, we waited until all animals of a group gathered in one cage (called extension cage) and quickly sealed off the other cages of the multi-cage system. The extension cage with the animals was then connected via a tube to an IntelliCage (Fig. 1C), and behavioral experiments started simultaneously for all groups by opening the connecting tubes to the individual IntelliCages (van Dijk et al., 2019). The extension cage remained connected to the IntelliCage for the entire experimental phase.

Sequence of behavioral experiments
Behavioral testing was conducted in the IntelliCage system (TSE systems, Germany, Fig. 1C), which is a high-throughput automated device to test group-housed small rodents. Assessment of individual performance of each animal is transponder-based and requires no human interaction, providing test conditions as stress-free as possible (for detailed description see Galsworthy et al., 2002;Endo et al., 2011;Kiryk et al., 2020). In short, the IntelliCage consists of four operant chambers fitted into the corners of a T2000 rat cage (592 Â 405 Â 295 (H) mm). Chambers can be accessed through a ring antenna registering visits. Inside each chamber (henceforth called corners), two water bottles are placed behind motorized doors. Pokes to the doors are registered as (C) Schematic drawing shows the experimental setup, each group was housed in an IntelliCage with 4 operant chambers fitted into the corners were the animals had access to two water bottles. An extension cage was connected to the IntelliCage. (D) Sequence of experimental phases and duration, test phases are indicated in bold, habituation and adaptation phases indicated in italic. After trapping, animals were given a habituation phase of 40-70 days and were then tested in the IntelliCage as follows: LED: Approachavoidance test using bright LED light; RTT: Reaction time task. This task was run twice; SP-FLEX: Self-paced behavioral flexibility task; Place-time task Chaining/Reversal. Adaptation phases in the Intellicage were: FA: Free adaptation; NPA: Nosepoke adaptation; DSA: Drinking session adaptation. (E) Activity measured as mean corner visits per hour in bank voles and wood mice during the first 3 days of FA in the IntelliCage system. After the initial novelty driven hyperactivity peak, both species habituated fast to the environment and returned to their natural nocturnal (wood mice) and cathemeral (bank voles) activity patterns. (F) Bank voles performed significantly more nosepokes during each visit than wood mice. The first 3 days of FA are shown. (G) In the LED approach-avoidance test, bank voles increased their visit activity during the light phases (yellow bars) compared to the same time periods in the previous night (dark grey bars). nosepokes, drinking is registered as licks due to contact to the nipple. Sequence of testing phases are summarized in Figure 1D.

Adaptation phases
-Free adaptation (FA): All doors in all corners were open continuously and animals had free, unlimited access to water. Analysis focused on nosepokes per visits and total licks. -Nosepoke adaptation (NPA): All doors were closed by default, but could be opened with a nosepoke. Time to drink was limited to six seconds, after which the door closed again. -Drinking session adaptation (DSA, Krackow et al., 2010): Animals could open doors with a nosepoke as in NPA, but only during specific time windows, called drinking sessions. Four sessions of two hours each (0:00 to 2:00, 5:00 to 7:00, 11:00 to 13:00, 19:00 to 21:00) were scheduled corresponding to the natural activity peaks of the animals. Outside these drinking sessions, no water could be accessed.

Learning tasks
-Light-induced approach-avoidance task (LED): For two short periods during the dark phase, corners were brightly illuminated (390 LUX) by switching on six yellow LED lights in every corner regardless of the presence of an animal. The two time periods were chosen according to the animals' previous activity peaks in the dark phases, which was from 19:30 to 20:30 and from 5:00 to 6:00. Otherwise, the test conditions corresponded to the NPA, i.e. nosepokes on any door gave access to water. Activity (visits, nosepokes and licks) in the LED periods were compared with the activity of the animals during the same time periods in the previous night. -Reaction time task (RTT, Kobayashi et al., 2013): A waiting period between a first nosepoke and door opening was used to assess response inhibition (action restraint), because animals had to wait a certain time span without further nosepokes for their reward. The first nosepoke initiated the trial and started a random delay period of 0.5, 1.5 or 2.5 s until one LED light was switched on signaling correct response, simultaneously the door opened for drinking without further nosepoking. After the first nosepoke, additional nosepokes during the delay period were considered as premature nosepokes and were punished by terminating the trial. The animal had to leave the corner and start a new trial. The following parameters were calculated for each animal and analyzed for species differences: premature nosepokes (%), that is percentage of visits in which animals made premature nosepokes; rewarded visits (%), that is percentage of visits in which animals performed correctly and could drink. Associations with performance in the reaction time task and habenular size were tested with second poke latency (s), that is time between the first nosepoke initiating the trial and second nosepoke, and with average number of premature pokes per visit. Two training phases were run prior to this test: In the first training phase (three days), delay was set to 0 s (no waiting period). The second training phase (three days) was equal to the test, except that premature nosepokes had no consequences and animal could access water after the random delay even if they performed premature nosepokes. -Self-paced behavioral flexibility task (SP-FLEX, modified from Endo et al., 2011;Oizumi et al., 2020;Balan et al., 2021): Every animal was allowed to drink water in two out of four corners (rewarded corners), which were opposite (diagonal of the cage) to each other. The other two corners (unrewarded corners) did not provide water. The two rewarded corners were assigned at random to each individual, but corner assignment was balanced within each group. Repetitive entries into the same rewarded corner did not give access to water, therefore, each animal had to shuttle between the two rewarded corners in order to obtain water. The first (original) learning phase was finished when the upper level of successful diagonal visits was reached (see below), and rewarded and unrewarded corners were reversed (complete shift). Whenever an animal's performance reached the upper criterion again, rewarded corners changed again and the next learning phase started. This protocol was designed as a self-paced task, in which the timing of complete shift varied for each animal based on the sequential probability ratio test (SPRT) for an animal.
The SPRT was calculated after each rewarded or unrewarded corner visit of an animal. A visit without nosepokes was ignored for the calculation of the SPRT. The SPRT was then used to determine if the animal's probability of success significantly exceeded the upper criterion (30%) or if the performance was below the lower criterion, i.e below chance level (20% probability of success). Total number of trials (the sum of rewarded and unrewarded visits) required to reach the upper criterion was used to assess the efficiency of learning. The number of trials conducted while the animal's success rate was below the lower criterion was counted as perseverative responses, and % perseverative responses in each learning phase were calculated. -Place-time chaining acquisition and reversal task (Chaining/Reversal, Albuquerque et al., 2013;Kobayashi et al., 2013): Successful learning of the spatio-temporal rule of this task gave water rewards during drinking sessions (as in the DSA protocol). Visits outside drinking sessions were not rewarded. The rewarded corner rotated after every rewarded visit, for half the animals clockwise and for the other half counterclockwise. To avoid mimicking behavior, direction and starting position of the rewarded corner was assigned randomly to each animal, but with an even distribution across the animals in the cage. Eight hours before the end of behavioral testing at 8:00am, i.e. during the last two drinking sessions, the direction (clockwise or counterclockwise) switched (reversal phase).
Only half of the animals in each cage were assigned to the reversal task (reversal), the other half continued the chaining acquisition task (familiar). Correct responses (%) per animal were analyzed. For animals subjected to the reversal task, error frequencies (%) were calculated.

Collection of tissue and blood
One hour after the last drinking session (at 8:00), sample collection started and was completed within two hours the latest. Control animals were processed first. Experimental animals were confined to their extension cage and given 20 minutes to adapt to the new room and to equalize physiological responses between the first and the last animal to be caught. Animals were deeply anesthetized using isoflurane inhalation (5% isoflurane, 0.7 l/min oxygen), decapitated, and trunk blood was collected in EDTA-coated tubes (500 ml, MicrovetteÒ, Sarstedt, Germany). Blood samples were placed on ice and centrifuged within two hours for 15 minutes at 4°C and 3000 rpm, supernatant plasma was pipetted off and frozen at À80°C until analysis. Brains were extracted and the right hemisphere was immersion fixed in cold 4% paraformaldehyde solution for 30 hours and transferred to 30% sucrose in phosphate-buffered saline. After saturation, hemispheres were frozen and stored at À80°C until analysis.

Determination of plasma corticosterone levels using ELISA
Plasma corticosterone levels of the animals were determined using an enzyme-linked immunosorbent assay (ELISA) approach (Corticosterone ELISA kit, Enzo Life Sciences), procedure was conducted according to manufacturer instructions and as described previously (Spinelli et al., 2013). In an initial test with plasma of two animals per species, the optimal sample dilution was found to be 1:40. This dilution was used for all samples. Samples were run in duplicates and values, corrected for non-specific binding, were average for each animal. To correct for non-specific binding, the optical density of the non-specific binding probe was subtracted from the optical density of each individual. The obtained values were then divided by the maximum binding optical density (from which the non-specific binding optical density was subtracted as well) to obtain the percentage bound. Using standards of known corticosterone concentrations that were run on the same well plate as the samples, a standard curve of logit-transformed percentage bound vs log-transformed corticosterone concentration was fitted using polynomial regression. The corticosterone concentrations (ng/ml) of the samples were extrapolated from the regression equation for the standard curve.

Brain histology and immunohistochemistry
Frozen right hemispheres were cut on a sliding microtome equipped with a freezing stage. Sections were cut coronally with a section thickness of 40 mm. Sections were collected in six series, spanning the entire hemisphere. One series was mounted in the correct anatomical order and Giemsa stained as a reference. One complete series (section spacing 240 lm) per animal was used for free-floating immunohistochemical staining of the protein of the immediate early gene Arc. Sections were rinsed in Tris-Triton-NaCl (Tris buffered saline, pH = 7.4, 0.3% Triton, 1.5% NaCl) between steps. Endogenous peroxidase was blocked using 1% H 2 O 2 for 15 minutes at RT. One hour pre-incubation in 2% normal goat serum (NGS) in Tris-Triton-NaCl was followed by incubation with 1:1000 polyclonal rabbit anti-Arc (Synaptic Systems, Cat. No. 156003) over night at 4°C . Incubation with the secondary antibody (biotinylated goat anti-rabbit, Vector) was done in 1:300 dilution in 2% NGS + 0.1% BSA in TBS, followed by incubation in avidin-biotin-peroxidase complex solution (Vectastatin Elite kit, Vector laboratories, Burlingame, CA, USA) and finally staining with 3,3 0 -diaminobenzidine (SigmafastTM, D4418-50SET, Sigma-Aldrich, Steinheim, Germany). Sections were mounted on slides and cover-slipped.

Volumetric analysis of medial habenular nucleus
Animal identity was blinded prior to analysis. The volume of the medial habenular nucleus was determined using the Cavalieri method (Slomianka, 2021) in the Giemsa stained reference series (section spacing 240 lm) using StereoInvestigator Version 10 (MicroBrightField) software on a Zeiss Axio Imager. M2 microscope at 20Â and grid points separated by 40 lm along the xand y-axes. On average, six sections per animal contained the medial habenular nucleus. Two control animals (one bank vole and one wood mouse) were excluded from the analysis due to tissue damage in the mounted sections.

Analysis of anti-Arc stained cells
Arc-positive cells (Fig. 2) were assessed in every sixth coronal section spanning the entire rostro-caudal extent of the brain. Animal identity was blinded before data collection. Sections were analyzed on a Zeiss Axioplan at 20Â. Numbers of Arc-positive cells were assessed in the dentate gyrus of the hippocampus ( Fig. 2A, B) with respect to dorsal to ventral differences. Caudally, where the dorsal (septal) and ventral (temporal) poles merge, a horizontal line was drawn separating equally sized dorsal and ventral halves of the dentate gyrus. On average, 15 sections per animal contained the hippocampal formation and were analyzed. Arc-positive cells in the dentate gyrus were counted exhaustively focusing through the entire thickness of the sections, but omitting cells laying in the top-focal plane. Furthermore, Arc staining intensity was evaluated in all sections containing the medial prefrontal cortex (mPFC, Fig. 2C) that included anterior cingulate area ventral and dorsal (ACAv and ACAd, respectively) and secondary motor area (MOs) according the Allen Mouse Brain Atlas. On average, 10 sections per animal were investigated. Rating was done according the following criteria: 0 = no Arc-positive cells; 1 = few, weakly stained Arc-positive cells; 2 = moderate Arc-positive cells in few sections; 3 = many Arc-positive cells in all sections with the structure of interest. Arc-positive cells in the suprachiasmatic nucleus (SON, Fig. 2D, E) and paraventricular nucleus (PVN, Fig. 2D, F) were scored as present or absent.

Data analysis
Behavioral data were extracted with IntelliCage Analyzer software and compiled in Excel. Statistical analysis and graphs were conducted with R version 4.2.1. Behavioral variables were Box-Cox transformed prior analysis to ensure approximate normality. To test differences in behavior between species, one-way ANOVA was used. Where repeated measures per animal were analyzed, repeated measures ANOVA was applied, followed by Tukey's post-hoc test. Sex was initially included in the model, but it was excluded from the final model as sex effects were not significant. To test species differences in quantitative data of habenula size, Arc expression and Corticosterone levels, one-way ANOVA was applied. Lastly, the relationships between behavior and habenula, Arc, Corticosterone and species were tested using ANCOVA. Main effects and possible interactions between neuromorphological and hormonal level and species were included in the model. In graphs, untransformed data are plotted, in boxplots mean and interquartile range are indicated, scatterplots indicate SD or confidence intervals. Significance level was set to <0.05.

Spontaneous behavior in the free adaptation phase
As reported previously (van Dijk et al., 2019), bank voles showed a cathemeral activity pattern, while wood mice were nearly exclusively nocturnal (Fig. 1E). Drinking behavior of bank voles followed an ultradian rhythm with a period length of around 3 hours. After the initial novelty-induced activity peak, habituation took place after the first dark phase (Fig. 1E). Bank voles performed more nosepokes per visit than wood mice (F (1,41) = 38.46, p < 0.0001, Fig. 1F), yet bank voles only showed a tendency to drink more than wood mice (lick number: F (1,41) = 3.7, p = 0.06, data not shown).

Approach-avoidance behavior in the LED light test
Approach-avoidance behavior in response to a bright light stimulus was assessed by comparing the behavior during the two 1-hour periods of bright light in the corners relative to the behavior in the same (dark) periods in the previous night. Analysis was done within species. Bank voles showed approaching behavior by increasing their visits in the presence of the light stimulus (F (1,22) = 5.7, p = 0.02, Fig. 1G). Wood mice did not respond by altering the number of visits (F (1,19) = 2.36, p = 0.13), but showed an avoidance behavior for licks by decreasing their water intake during the bright light period (F (1,19) = 19.49, p < 0.0001, data not shown).

Impulsivity in the reaction time task
The reaction time task, measuring the animal's ability for response inhibition, was run twice. Each run included six days of training and two days of testing. Only performances during the testing days were analyzed. Percentage of visits with premature nosepokes was significantly higher in bank voles compared to wood mice in the first run (bank voles 87%, wood mice 72%, F (1,41) = 25.3, p < 0.001) and in the second run (bank voles 91%, wood mice 71%, F (1,35) = 28.6, p < 0.001), post-hoc analysis of delays per run revealed that in each of them, bank voles had higher percentages of premature responses than wood mice (all p < 0.01, Fig. 3A). The percentage of rewarded visits where animals could drink was significantly lower in bank voles in the first run (F (1,41) = 20.9, p < 0.0001) and in the second run (F (1,35) = 28.6, p < 0.001). Post-hoc analysis revealed that bank voles and wood mice improved from day one to day two in both runs (all p < 0.001, Fig. 3B).

Self-paced behavioral flexibility test
This test was run over five days, and each animal completed the learning phases in a self-paced manner, where the timing of complete shift of rewarded corners was determined based on the performance of individuals reaching the defined upper criterion. Over the entire test phase, bank voles made more visits (F (1,38) = 33.7, p < 0.0001) than wood mice. Thus, bank voles were more successful in this task by completing eight learning phases on average, while wood mice completed only three phases on average (Fig. 3C). For statistical analysis, only the first three phases were analyzed, and animals that finished less than three phases (nine wood mice) were excluded as they did not meet the analysis criteria. Two primary readouts assessed performance in the self-paced flexibility test: total number of trials required to reach the upper criterion for phase reversal, that is 30% of success responses, and percentage of perseverative responses in each phase. Bank voles made more trials to reach criterion than wood mice (F (1,27) = 6.9, p = 0.01, Fig. 3D). Percentage of perseverative responses was higher in bank voles than wood mice (F (1,27) = 9.69, p = 0.004, Fig. 3E).

Place-time learning in chaining acquisition and reversal
In this task, animals learned to visit corners in a clock-or counterclockwise order. Repeated entries in the same corner or omission of the next correct corner did not yield rewards. Performance was measured as the percentage of correct corner visits. Both species started at chance level (25%) for visiting the correct corners on day 1 of testing. Overall, there was a strong improvement in correct corner visits over days (F (2,66) = 279.2, p < 0.0001, Fig. 3F). We observed a weak, not significant species difference (F (1,33) = 3.1, p = 0.09), with bank voles tending to learn faster in the intermediate time (days 2-6) than wood mice. As a final task, the order of correct corners was reversed. Only half of the animals in each cage performed the reversal task, the other half continued with chaining acquisition (familiar task). Species did not differ in percentage of correct responses in the last task (F (1,31) = 0.09, p = 0.77), but animals in the familiar and reversal groups were different (F (1,31) = 9.4, p = 0.004, Fig. 3G), with no interaction between species and groups (p = 0.61), animals of both species subjected to the reversal task performed worse. A detailed look at the type of errors in the reversal group (Fig. 3H) revealed that most of the errors occurred due to perseverative visits (visits according the previous rule) or repeated Both species improved correct responses over days (p < 0.0001), the weak species effect (p = 0.09) was due to better performance of bank voles in days 2-6. Line at 25% indicates chance level, bars indicate SD. (G) Correct performance in the reversal group dropped in both species compared to the animals continuing the previous rule (familiar), no species differences were observed. (H) Error frequency (%) in animals subjected to the reversal task: most errors could be attributed to perseverative (Persev.) visits, that is, animals visited corners according to the previous rule, followed by repetitively entering the same corner (Rep.). Diagonal visits (Diagonal) or errors related to the new rules (New) such as visit abortions were rare. No species differences in terms of error frequency were found. entries into the same corner. Visits of diagonal corners or errors related to the new rules such as abandoned visits were low. We found no species differences with respect to the error type (F (1,16) = 0.8, p = 0.4, Fig. 3H).

Habenula size and impulsivity
Volumetric analysis of the mHb in wood mice (Fig. 4A) and bank voles (Fig. 4B) was performed using the Cavalieri method. The volume of the medial habenula (mHb) was smaller in bank voles than wood mice (F (1,41) = 7.68, p = 0.008, Fig. 4C, Table 1). Behavioral performance in the reaction time task was tested against mHb size. We found an effect of habenula size (F (1,31) = 6.5, p = 0.02, black line with confidence interval in Fig. 4D) and species (F (1,31) = 38.9, p < 0.001) on the latency between first and second nosepoke in the reaction time task, no interaction was seen here. The strong species effect in latency might be driven by other components in the circuitries mediating impulse control. The number of premature nosepoke repetitions in the reaction time task was correlated with habenula size (F (1,31) = 4.9, p = 0.03, black line with confidence interval in Fig. 4E), no species effect or interaction were found here.

Arc expression in the hippocampus
First, we tested in all animals if the number of Arc-positive cells in the dentate granule cell layer differed between species. Overall, there was no species difference in Arcpositive cells in the dentate gyrus (F (1,42) = 3.07, p = 0.087, Fig. 5A). Separated into dorsal and ventral hippocampus, wood mice had lower numbers of Arcpositive cells in the dorsal dentate gyrus than bank voles (F (1,42) = 4.73, p = 0.035), no species difference in numbers in the ventral part were found (F (1,42) = 1.58, p = 0.22, Fig. 5A, Table 1). Focusing on the dorsal dentate gyrus, we tested if Arc expression was related to the behavioral testing in the IntelliCage (Fig. 5B). There was a species effect (F (1,38) = 5.1, p = 0.03) and a group effect (F (2,38) = 3.6, p = 0.04), but no interaction. Post-hoc analysis revealed that animals performing the familiar task had higher numbers of Arc-positive cells in the dorsal dentate gyrus than those performing the reversal task (p = 0.03). Comparisons with the control group were not significant. The number of Arc-positive cells in the hippocampus and behavioral parameters assessed in the last testing phase did not correlate.

Arc expression in mPFC
Arc signal in the area of the medial prefrontal cortex (ACAv, ACAd and MOs, Fig. 2C) was scored between 0-3. Bank voles had markedly lower mPFC Arc signal than wood mice (F (1,39) = 21.5, p < 0.0001, Fig. 5C, Table 1), no effect of reversal vs familiar task groups could be seen (F (2,39) = 1.7, p = 0.2, data not shown). Improvement of correct responses from the second last to the last test session (performance gain in %) was correlated negatively with mPFC Arc expression (F (1,27) = 5.7, p = 0.02, black line with confidence interval in Fig. 5D), no species effect but a species Â group interaction was found (F (1,27) = 7.8, p = 0.009). Post-hoc analysis revealed that the significant interaction was due to differences in wood mice between the familiar and reversal group (p = 0.03).

Arc expression in paraventricular and supraoptic nuclei
All bank voles tested in the IntelliCage showed Arcpositive cells in the paraventricular nucleus (PVN, Fig. 2D, F, Table 1) and the supraoptic nucleus (SON, Fig. 2D, E, Table 1) of the hypothalamus. Only 1 out of 21 experimental wood mice (5%) showed Arc-positive cells in PVN or SON. We found that activation of SON in bank voles was specific to behavioral testing, as none of the control bank voles (not tested in IntelliCage) showed Arc-positive cells in the SON. 63% of the control bank voles did show Arc-positive cells in the PVN. Interestingly, although unrelated to the main questions addressed in this paper, immunohistochemistry against Arc revealed a cytoplasmatic expression in PVN and SON (Fig. 2D, E,   Fig. 4. Representative Giemsa-stained medial habenular nucleus (mHb) of wood mouse (A) and bank vole (B), scale bar = 50 lm. (C) Volume of mHb was smaller in bank voles than wood mice. (D, E) Performance in the reaction time task correlated with mHb volume. (D) Latency to second poke (s), that is time interval between the first nosepoke initiating the trial and the first premature nosepoke, correlated with mHb size (p = 0.02, black line, confidence interval = 0.95 indicated in grey) and shorter latency in bank voles (p < 0.001). (E) Premature poke repetitions (mean per animal) was correlated with mHb size (p = 0.025, black line with confidence interval = 0.95 in grey), species effect was not significant (ns). F), while Arc-positive cells in the mPFC and hippocampal dentate gyrus showed a nuclear pattern ( Fig. 2A, B, C).

Corticosterone blood level and behavioral task performance
Corticosterone blood levels as a measure for basal hypothalamic-pituitary-adrenal axis activity were compared between species and experimental groups (Table 1). One wood mouse required an unusually long time to capture, and its corticosterone level was more than 2 SD higher than the mean. It was therefore excluded from further analysis. Corticosterone levels showed no species effect (F (1,38) = 0.05, p = 0.81), and only a trend for an experimental group effect (F (2,38) = 3.2, p = 0.052, Fig.  5E). Behavioral performance in the last hours of testing was investigated for correlations with corticosterone levels, no significant associations were found.

DISCUSSION
This study aimed to explore response control in wild bank voles and wood mice using a series of behavioral tests in the IntelliCage system. Impulsivity was assessed using the reaction time task, a test conceptually similar to the 5CSRTT. After an initial nosepoke defining the target door, animals had to wait a random delay until a gosignal (light) indicated opening of the door giving access to water. Animals needed to restrain action, waiting for the go-signal until reward could be obtained. Anticipatory responses, here premature nosepokes, were punished by termination of the trial, meaning the animal had to leave the operant chamber and restart the trial. Bank voles showed more premature responses dentate gyrus (DG) of the hippocampus was not significantly different between species. Separated into dorsal and ventral DG, we found higher numbers of Arc-positive cells in the dorsal DG in bank voles than wood mice (p = 0.035). No difference was found in the ventral DG. All animals were included in this graph. (B) For dorsal Arc-positive cells, we found a group effect (p = 0.04), post hoc test revealed higher number of dorsal Arc-positive cells in the dentate gyrus in the familiar group (continuing the chaining acquisition task) compared to animals in the reversal group (performing the reversal task). Control animals (not tested behaviorally) did not differ, species effect remained (p = 0.026). (C, D) Arc expression in the mPFC. (D) In wood mice, Arc signal (ranked, O: no staining to 3: many Arc-positive cells, see Fig. 2C) in the mPFC was common, in bank voles rare (species effect p < 0.0001). All animals were included in this graph. (D) Arc signal in the mPFC was correlated with performance gain (%) in the last behavioral tests (p = 0.02), no species or group effect was found. (E) Blood corticosterone level was not significantly different between species, a possible group effect just missed significance level. than wood mice, indicating increased impulsivity. Premature responses in bank voles were also higher than the numbers reported for laboratory mice performing the same task (Kobayashi et al., 2013;Ma¨tlik et al., 2018). Bank voles were able to learn the task, as their performance improved from day 1 to day 2 in the first and second run of the test, but they never reached a percentage of rewarded visits of more than 19% and could also not improve when the test was run for the second time. In comparison, wood mice started with 17% correct visits and improved continuously to 34%. Perseverant behavior was tested in the SP-FLEX and chaining/reversal task. Both tasks required that animals visited corners at a defined spatial sequence. In the SP-FLEX task, rule changes were defined by the individual's performance, while in the chaining/reversal task, rule changes were simultaneous for all animals. Both repetitive visits into the same corner and visits according to the previous rule were considered as perseverative responses. Bank voles made more perseverative responses than wood mice in the SP-FLEX task, but were overall more successful in this task than wood mice, indicated by the completion of more learning phases. In the chaining/reversal task, both species performed equally in terms of acquisition learning and perseverative responses, confirming previous results (van Dijk et al., 2019). Currently, it is unclear what might have caused the discrepant findings for perseverant behavior in the two tests. Further studies would be required to eliminate external causes such as sequence of tests or test differences (e.g. using higher thresholds to reach criterion for rule changes in SP-FLEX as shown in Oizumi et al. (2020)). Furthermore, low visit numbers in wood mice in the SP-FLEX task resulted in low number of complete shifts, thus performance comparisons between species were skewed for animal numbers and overall performance. The approach-avoidance test indicated that the light stimulus led to an approach behavior in bank voles as they increased their visit numbers, wood mice however responded with an avoidance behavior as they lowered their drinking activity. In part, the species differences in response to the light stimulus may stem from their natural activity patterns (cathemeral bank voles versus nocturnal wood mice). Increased approach behavior of bank voles compared to wood mice corresponded however with our qualitative observations while working with these animals. Impulsivity as measured in the reaction time task was correlated with medial habenula (mHb) size. Bank voles had significantly smaller mHb volumes than wood mice. Notably, the relation between mHb size and performance in the reaction time task did not only hold on species level, but was also found within species, i.e. also wood mice with smaller mHb were more impulsive than those with larger mHb. Our findings in wild rodents are in line with previous studies in which ablation of the mHb or the entire habenular complex in mice and rats resulted in deficits in impulse control in the reaction time task or the 5CSRTT (Lecourtier and Kelly, 2005;Kobayashi et al., 2013). In contrast to the findings in mHb lesioned mice (Kobayashi et al., 2013), smaller mHb in bank voles were not associated with deficits in environmental adaptation and spatial memory, as their performance in the chaining/reversal task, previously shown to be hippocampus-dependent (Voikar et al., 2018), was not impaired (van Dijk et al., 2019 and data presented here). An extensive literature links response control to topographically organized pathways involving prefrontal cortices, basal ganglia, limbic structures such as the hippocampus and amygdala, and neuromodulatory systems (reviewed in Bari and Robbins, 2013;Dalley and Ersche, 2019;Kozak et al., 2019). In this study, the expression of the immediate early gene Arc in different brain areas was assessed after the chaining/reversal task, we therefore focused our investigations to mPFC and hippocampus. Higher Arc expression in the dorsal dentate gyrus of bank voles compared to wood mice might emphasize the importance of the dorsal hippocampus in learning processes in bank voles. The lack of an association between hippocampal arc expression and behavioral performance in both species could result from the 24/7 schedule of behavioral testing in the IntelliCage, compared to previous studies where Arc expression in the hippocampus was assessed after a brief exposure to a novel environment (Ramı´rez-Amaya et al., 2005). Response control in impulsivity and compulsivity/perseveration depends on mPFC functions (Pattij and Vanderschuren, 2008;Turner and Parkes, 2020). Strikingly, Arc expression in mPFC was rare in bank voles. A general low engagement of mPFC areas in bank voles might explain their shorter response latency in the impulsivity test that is not explained by habenula size. In contrast to the mPFC findings, in the lateral hypothalamic area, Arc-positive cells were found in PVN and SON in all behaviorally tested bank voles, while only one wood mouse showed Arc expression in these nuclei. Both nuclei are involved in the neuronal network regulating water balance by arginine vasopressin (AVP) secretion (Qin et al., 2018). Bank voles have been reported to have a higher water metabolism than wood mice (Shore et al., 1992). Thus, SON Arc expression in bank voles could be a response to water restriction in the experimental setup, as none of the control bank voles receiving water ad libitum showed Arc expression in this nucleus. However, PVN Arc expression was present in $60% of control bank voles, too, indicating that activation of this nucleus might depend on factors other than water intake regulation. The lateral hypothalamic area is part of the habenular circuitries (Lecourtier and Kelly, 2007), the PVN specifically receives projection from the lateral hypothalamic region (Larsen et al., 1994), and a recent study described a novel pathway regulating impulsivity via projection from lateral hypothalamic neurons to the ventral hippocampus (Noble et al., 2019). Further investigations would be needed to test if PVN activation in bank voles is an indicator of a compensatory effect in the habenular circuitry, directly or indirectly linked to high impulsivity in this species. Lastly, we assessed blood corticosterone to assess potential stress responses to behavioral testing. Corticosterone levels of the wild animals were in the range of baseline concentrations of laboratory-bred house mice or Kunming mice (Malisch et al., 2007;Gong et al., 2015). Other strains commonly used in laboratory settings have generally lower corticosterone levels (Pryce et al., 2011;Ramı´rez-Rosas et al., 2019). Bank voles and wood mice did not differ in their corticosterone levels, and behavioral testing did not significantly increase corticosterone, indicating that the IntelliCage testing environment itself imposed no additional stress on these wild rodents.

Bank voles as model animals
Bank voles have been and are still frequently investigated in ecology. As a model organism in basic science, bank voles have a marginal existence despite their early use as tuberculosis model due to their susceptibility to both bovine and human tuberculosis (Jespersen, 1954), and their key role as the reservoir host of hantavirus (Olsson et al., 2003;Strandin et al., 2020). Bank voles, both wild and captive, can suffer from hyperglycemia, thus developing diabetes-like features (Freimanis et al., 2003;Niklasson et al., 2003a;Niklasson et al., 2003b). They are susceptible to several prion strains of both animal and human origin (Cartoni et al., 2005;Nonno et al., 2006), and transgenic mice have been generated carrying the bank vole prion protein (BVPrP) (Larsen, 2016;Vorberg and Chiesa, 2019). Descriptions of the behavioral phenotype of bank voles included cathemeral activity (Niethammer and Krapp, 1982;van Dijk et al., 2019; this study), novelty-seeking, explorative and risk-taking behaviors (Korpela et al., 2011). Others reported increased latencies in novelty-induced explorative behaviors in bank voles compared to laboratory mice (van Dijk et al., 2019). Captivity-bred bank voles can spontaneously develop stereotypic behaviors under standard laboratory conditions, but less so in enriched environments (Ö dberg, 1987;Sørensen, 1987;Kapusta et al., 2022). Bank voles performed better than laboratory mice in hippocampus-dependent tests such as the Morris Water Maze (Pleskacheva et al., 2000) and the chaining/reversal tasks in the IntelliCage (van Dijk et al., 2019). Indeed, the hippocampus of bank voles is well developed (Vandebroek et al., 1999), and hippocampal dentate gyrus granule cells are more numerous in bank voles than in laboratory mice (Amrein et al., 2004) or even rats (West et al., 1991). The present study showed high impulsivity in wild-caught bank voles in the reaction time task, low perseverative responses in the chaining/reversal task but reduced behavioral flexibility in the SP-FLEX task. We could show that impaired impulse control in bank voles was correlated with small mHb size and might be linked to low mPFC activation, as assessment of Arc signal after behavioral testing was low in mPFC, but high in hypothalamic SON and PVN and dorsal hippocampus. We hypothesize that impulsivity in bank voles might be rooted in species-specific cortico-limbic, possibly also corticothalamic, regulatory mechanisms. Thus, bank voles could serve as a natural animal model to investigate alternative response control pathways in dissecting impulsivity from perseverant behavior.

AUTHOR CONTRIBUTIONS
MJ trapping animals, conducting behavioral experiments, performing histological processing and analysis, writing of manuscript; JM trapping animals, assisting in behavioral experiments; DPW analysis of behavioral experiments, writing of manuscript; CRP performing and analyzing corticosterone measures, writing of manuscript; TE and BS providing and analyzing IntelliCage protocols, writing of manuscript; IA study design, instructing trapping and behavioral experiments, analyzing, writing of manuscript.

CONFLICTS OF INTEREST STATEMENT
Author Toshihiro Endo is a representative of Phenovance LLC (Kashiwa, Japan). The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.