Validation of the Wistar-kyoto Rat Kept in Solitary Housing as an Animal Model for Endogenous Depression Using Voxel-based Morphometry


 Major depressive disorder is a common psychiatric condition that is often resistant to medication. The Wistar-Kyoto (WKY) rat has been suggested as an animal model of endogenous depression; however, it is challenging to translate results obtained in animal models into humans. Solitary housing is a mild stress paradigm that could simulate the environment of depressive patients with limited social activity due to symptoms. We used voxel-based morphometry to directly compare the solitary-housed WKY rat model with data from previous human studies, and validated our results with behavioural studies and correlation analyses. Atrophy in WKY rats was detected in the ventral hippocampus, caudate putamen, lateral septum, cerebellar vermis, and cerebellar nuclei (p < 0.05, corrected for family-wise error rate). Further, locomotor behaviour was negatively correlated with hippocampal atrophy and positively correlated with atrophy of the cerebellar vermis. The regions of brain atrophy validate WKY rats as an animal model for endogenous depression and can aid the translation of study results to humans. Our study also reveals the possibility of a cerebellar contribution to depression.

validation, and increased basal corticosterone levels, Adrenocorticotropic hormone (ACTH) levels 21,22 , and basal noradrenaline neurons activity 23 were reported. Reduced Brain-derived neurotrophic factor (BDNF) levels in the frontal cortex and hippocampus 24 , as well as reduced serum BDNF response to mild stress 25 have also been reported in WKY rats. Alterations in serotonergic and dopaminergic system markers have also been reported, which is similar to changes observed in MDD patients 26 . Further, WKY rats are considered a model of treatment-resistant depression because they are hyporesponsive to antidepressants 16,26,27 , and although predictive validities are still in discussion, this is important because of the similar response often observed in MDD patients.
Many depressed patients exhibit restricted social activity 28 , and there are signi cant associations between reduced social skills and depression 29 . We therefore suggest that an isolation paradigm in animal studies could mimic the typical environment of depressive patients. WKY rats within an isolation paradigm often exhibit enhanced depressive symptoms [30][31][32][33][34] .
The mechanisms underlying treatment resistance in MDD patients are unclear. Although several clinical studies have investigated this, the involvement of drug-naive patients presents challenges because the concepts of treatment resistance and drug naivety are incompatible. Conducting VBM in an animal model may be the rst step in connecting the ndings of animal and human studies.
We aim to validate the similarities in atrophic brain regions between WKY rats and humans with MDD with an animal VBM study using MRI and comparing results with clinical MDD studies. We also examined the correlations between behavioural and morphological changes to validate WKY rats as a model for depression.

Two open-eld tests (OFTs)
Behavioural changes were measured in two OFT trials. Strains × trials × parameters (F (1, 34) = 3.308, p < 0.001), strains × parameters (F = 187.4, p < 0.001), and parameters (F = 11.424, p < 0.001) exhibited signi cant interactions; however, the interactions of strains × trials (F = 3.663, p = 0.068), parameters × trials (F = 0.391, p = 0.999), and trials (F = 0.365, p = 0.552) were not signi cant. In addition, a signi cant difference was observed in body weight (Wistar: 350.24 ± 9.48 g; WKY: 241.37 ± 9.523 g; t = 11.06; p < 0.001) and TIV, which was calculated using the MRI data (Wistar: 2166.46 ± 153.40 mm 3 ; WKY: 2056.32 ± 97.34; t = 2.19; p = 0.041). All locomotion parameters were signi cantly reduced in WKY rats in both OFT trials ( Fig. 1b-h) and signi cant differences in anxiety behaviour were observed after isolation (second trial). In WKY rats, wall-side time was signi cantly higher (Fig. 1i) and the proportion of time and total time in the centre region were signi cantly lower in the second trial ( Fig. 1j- When comparing the two trials, we found that many locomotive parameters signi cantly decreased, including the total distance, total movement duration, total number of movement episodes, duration per movement, and average speed ( Fig. 1b-h). Wall-side time increased in WKY rats in the second trial ( Fig. 1i), and only moving speed signi cantly increased in Wistar rats. See Supplementary Table 2 for statistical results of the trial comparisons. Together, this indicates that the escape activity of WKY rats was disrupted from the initial phase and decreased more after an additional seven days in individual housing. These rats also showed enhanced anxiety parameters after the seven-day isolation period.
When comparing the time spent in each block between the two trials, we found that in the second trial, Wistar rats stayed signi cantly longer in block 25 and spent less time in block 3 (Fig. 1l), whereas WKY rats stayed signi cantly longer in block 1 and spent less time in block 6 (Supplementary Table 2; Fig. 1m). When comparing strains, Wistar rats stayed longer in distant blocks and the centre area in both trials (Supplementary Table 1; Supplementary Fig. 1b and 1c). These results demonstrate that WKY rats did not ee to distant blocks, and represents the despair of an endogenous depression model.

Freezing time and low locomotion during forced swim test (FST)
Freezing time was compared in both strains during the 6-min FST and was assessed as a measure of disrupted locomotion, a depressive trait. Freezing time, or the percent of time immobile, was signi cantly longer in WKY rats than Wistar rats (Wistar: 1.77% ± 0.59%; WKY: 24.45% ± 5.25%; t = − 4.297; p < 0.01; Fig. 1n). Total movement distance, a measure of locomotor activity, was also signi cantly less in WKY rats than Wistar rats (Wistar: 1211.84 ± 46.42 cm; WKY: 889.55 ± 56.56 cm; t = 4.404; p < 0.01; Fig. 1o). WKY rats thus exhibited reduced locomotor activity and more immobility compared to Wistar rats during the initial phase of this study. The data from two Wistar rats were excluded due to camera failure (Wistar: n = 11, WKY: n = 13).
We also performed a voxel-based correlation analysis between the parameters from the second OFT, except for time spent in each block and grey matter concentration, in WKY rats. We observed clusters in different brain areas with negative correlations between grey matter volume and certain behaviours typically studied in depression research. We observed negative correlations between total movement duration and total number of movement episodes in right habenula clusters ( Fig. 3a and 3b); between distance per movement and clusters in the left amygdala (Fig. 3c); total number of movement episodes and clusters in the left ventral hippocampus (Fig. 3b); and movement speed and clusters in the right lateral septum (Fig. 3d). On the other hand, clusters with positive correlations were detected in the cerebellar vermis and pituitary gland: a cluster in lobule 9 in the caudal vermis was positively correlated with distance per movement (Fig. 4a), duration per movement (Fig. 4b), and moving speed (Fig. 4c); a cluster in lobule 3 in the rostral vermis correlated with duration per movement (Fig. 4a); and a cluster in the pituitary gland correlated with duration per movement (Fig. 4b). However, no signi cant cluster was found to correlate grey matter with wall-side time, percent of time in the centre region, or total time in the centre region.

Discussion
VBM revealed signi cant atrophy in the following brain areas of WKY rats: the left ventral hippocampus, the right caudate putamen, the right lateral septum, the cerebellar vermis, and the cerebellar nuclei.
According to a recent clinical meta-analysis 15 , the VBM results comparing WKY and Wistar rats are similar to those in human clinical MDD studies, although the largest atrophic cluster was observed in the cerebellum. The results of the two OFT trials and FST indicates the behavioural characteristics of WKY rats as a depression model, namely locomotion disruption and reduced escape activity. Voxel-based correlation analysis revealed areas correlated with escape behaviour in the second OFT. We observed negative correlations in the hippocampus, which agree with the results of a previous animal study 35 , and the right habenula, as well as a positive correlation in the cerebellar vermis. Our VBM results indicate similar anatomical changes in WKY rats as in humans, validating their use as an endogenous depression model. Surprisingly, the most signi cant atrophic cluster was detected in the cerebellum with the moderate threshold. The clusters observed in the cerebellar nuclei and vermis in the 3rd and 9th lobules remained after the stricter threshold was applied. A recent meta-analysis reported an atrophic cluster in the cerebellum of MDD patients 12 , and other studies have suggested cerebellar involvement in modulating various aspects of mood and cognition 36-38 . Hypertrophy has been observed in the cerebellum of depressed patients after successful treatment with electroconvulsive therapy 39 as well as in the cerebellar vermis after the use of chronic antidepressant medication 40 . Another recent meta-analysis detected a hypertrophic cluster in the cerebellar vermis in BD patients relative to MDD patients 15 . The cerebellar atrophy in WKY rats thus re ects MDD but not BD. In addition, positive correlations between atrophic clusters and certain behaviours were detected in the vermis (9th lobule), although we could not con rm positive correlations with other atrophic brain regions, such as the hippocampus. The cerebellar vermis is included in the cerebrocerebellum and receives target projections from motor areas 41 . According to our OFT results, atrophy in the vermis may lead to disrupted escape activity. Cerebellar atrophy may trigger behavioural disturbances and induce an inability to perform even in the presence of motivation. The role of the cerebellum in emotional processing 42 is still being discussed, and the contribution of the cerebellum to depressive symptoms should be investigated further.
We observed a signi cant cluster extending from the hippocampus to the septal region, somewhat including the thalamus, with the moderate threshold, and a cluster in the left ventral hippocampus remained after the stricter threshold was applied. This demonstrates signi cant atrophy in the ventral hippocampus and an atrophic trend in the dorsal hippocampus ( Supplementary Fig. 3) A previous animal study reported negative correlations between locomotion and grey matter concentration in the ventral hippocampus 35 . Because the hippocampus is a common region of interest in MDD research, as supported by a recent meta-analysis 15 , the atrophy observed in the hippocampus and parahippocampal regions in our model indicates similarity to human studies. However, WKY rats have been reported to be vulnerable to stress 16 , and differential roles of the ventral (stress, emotion) and dorsal (cognition) hippocampi have been discussed 43,44 , so the atrophy we observed might be stress-related.
Our data demonstrates a negative correlation between escape behaviour and the right habenula as well as volume reduction. The relationship between the volume of the habenula and its effect on escape activity should be investigated further. It has been proposed that the lateral habenula could systematically learn to expect an adverse outcome, and frequent neural ring may lead to a state of continuous disappointment and hopelessness 45 . A signi cant volume reduction in the habenula of depressive patients has also been reported 46 , and the habenula is considered a critical region in MDD research 47 . Ketamine-induced metabolic reduction in the right habenula of treatment-resistant depression patients has been reported 48 , and the habenula may be a potential target for future research on treatment-resistant depression.
Brain atrophy observed in the right lateral septum neighbouring the mbria remained when the stricter threshold was applied, and a negative correlation with escape behaviour was detected. Although the septal area is rarely described in clinical depression studies using MRI, previous animal studies reported that the septal nuclei are involved in stress responses 49 . In addition, lesion studies of the septum reported aggressive behaviours 50,51 . Patients with clinical MDD exhibit symptoms of ambivalence, lost motivation, reduced locomotor activity, and increased impulsivity and irritation. Further study is warranted to clarify whether the atrophy that extends from the ventral hippocampus to the septum contributes to the complex symptoms observed in MDD.
Only two clusters including the amygdala were detected using the moderate threshold, although a negative correlation to movement duration was detected in the left amygdala. Some volumetric studies in MDD patients have reported increased amygdala volumes 52,53 , whereas others have reported reduced volumes 54,55 . A meta-regression analysis in rst-episode MDD patients showed negative correlations between the volume of the right amygdala and Hamilton Depression Rating Scale score 10 . In another recent study, patients with treatment-resistant MDD exhibited blunted amygdala activity during a facial recognition task 56 . Further research is thus warranted to determine the relationship between amygdala volume and treatment-resistant MDD.
Our results indicated atrophy from the secondary somatosensory cortex extending to the insular cortex, although this did not remain after the stricter threshold was applied. Therefore, we cannot determine any major contribution of this site in MDD. However, a human study using VBM analysis 57 and some metaanalyses 7,10,11 have reported insular cortex atrophy in MDD patients, and our VBM data seems to be consistent with these studies. Our voxel-based correlation analysis could not detect a signi cant cluster in the somatosensory or insular cortices, and no correlation between these areas and behaviour could be con rmed in WKY rats. Insular cortex atrophy is commonly observed in many psychiatric disorders 58 , so analysis of this region in post-mortem brain studies may help to elucidate genetic spectra across various psychiatric disorders 59 , including MDD.
To the best of our knowledge, a reduction in pituitary gland volume has not previously been reported in WKY rats. However, it has been demonstrated that isolation paradigms elevate plasma ACTH levels without changes in corticosterone levels 30 , and we cannot rule out the possibility that isolation exhausted pituitary ACTH in WKY rats, resulting in atrophy. However, our study showed that pituitary atrophy was positively correlated with locomotion behaviour, so future studies are required to determine whether this atrophy is a biomarker of treatment resistance.
In this study, we were unable to con rm any ACC atrophy in WKY rats. Although atrophy in the frontal regions containing the ACC has been reported in MDD patients 7,60 , a medication wash-out MDD study reported an increase in ACC volume 9 . Miniscule changes have been observed in medial prefrontal regions in WKY rats 16 , and we cannot completely rule out small anatomical variations corresponding with strain differences. In addition, deep brain stimulation of the prefrontal region in WKY rats has been shown to rescue disrupted locomotor activity 61 . The contribution of frontal brain atrophy to depression symptoms remains unclear.
Both the OFT and FST were used to validate depressive behaviour. Although we did not nd any literature showing the impact of a 10 min OFT on brain morphology, we cannot rule out the carry-over effect of the OFT upon the FST in this experimental paradigm. OFT has been criticized as an anxiety-detecting measure 62 , and it has been discussed that immobility during the FST is not indicative of despair, but a coping bene t from learning and memory to promote behavioural adaptation and survival 63 . We cannot rule this possibility out when interpreting the enhanced preference for passive survival coping strategies in WKY rats; however, our data simultaneously demonstrated low locomotor activity during both the FST and OFT. Disruption of exploratory locomotion during the OFT, where candidates could exhibit coping behaviours to escape from isolation more easily than in the FST, would not relate to survival. Although Wistar rats exhibited prolonged stays in distant blocks, WKY rats did not, although moderate stress is expected to induce this coping behaviour. Our behavioural results thus show reduced escape behaviour both in neutral and threatening situations, which seems to represent the presence of despair. We assume that these results indicate face validity for this model of endogenous depression.
In this study, we housed rats individually, and cannot completely rule out the effect of isolation on brain atrophy. However, there are few reports on microscopic morphological changes induced by social isolation (besides maternal separation) within the rst two weeks; only one report describes olfactory bulb atrophy induced by isolation 64 . Although it has been proposed that a social isolation paradigm might have adverse effects for recovery interventions, such inducing neurogenesis in the hippocampus following exercise 65,66 , no signi cant volumetric changes in the hippocampus were reported during two weeks of social isolation using the region of interest-volumetric MRI method 64 . In another endogenous depression model, the Flinders sensitive line, ve weeks of isolation induced erasure of depression-like behaviour 67 . Whether isolation can be dismissed as stress in animal endogenous depression models is still an interesting question. We suspect that the effect of isolation stress in the rst two weeks might be too small to affect brain morphology in animal depression models, although we cannot rule out the possibility that genetic differences might enhance environmental effects.
We used an ex vivo MRI protocol to achieve the quality necessary for VBM. However, ex vivo MRI measurements have been discussed as less preferable for structural brain studies due to potential displacement, disrupted brain tissue integrity, and deformations resulting in artifacts 68 . This method may cause a global shrinkage in brain volume (~ 10%) and damage is most notable in the cerebellum, olfactory bulb, and cortex 69,70 . We therefore cannot completely rule out any data noise caused by the xation method used. Spatially-normalized images in VBM analysis should have a relatively high resolution in humans, and grey matter segmentation is not excessively confounded by partial volume effects 4 , but these assumptions may not hold in small animals. For in vivo MRI in small animals, it is di cult to prevent motion artifacts and there are resolution disadvantages. In particular, it has been reported that the thalamus and hypothalamus can be more properly segmented using ex vivo MRI data 69 . There is also an MRI study reporting that the volume of the midbrain, hippocampus, thalamus, and cortex were relatively unaffected by perfuse-xation 70 . The signi cant data observed in parahippocampal regions and cerebellar nuclei in this study might have bene tted from using ex vivo MRI techniques.
This study shows that single-housed WKY rats, as an endogenous depression model, demonstrated morphological similarities to MDD in humans. Correlations were determined to exist between the habenula and cerebellar vermis with inhibited escape behaviour via voxel-based correlation analysis.
Although the contributions of the cerebellum 36 and habenula 47 to MDD have been discussed, there is insu cient research and therapies directly targeting these areas. We believe that WKY rats would be a good model for exploring the mechanisms of MDD and address therapeutic challenges.

Animals
Eight-week-old Wistar (n = 13) and WKY rats (n = 13) were obtained from Charles River Laboratories Japan. (Yokohama, Japan). Initially, all rats were housed individually, maintained in all air-conditioned rooms (23°C ± 1°C) on a 12 h light/dark cycle (6:00-18:00 light), and fed food and water ad libitum. All experiments were approved by the Committee on Animal Research at Kyoto Prefectural University of Medicine and were conducted according to the animal care guidelines of the National Institutes of Health and ARRIVE guidelines 71 . We minimised the number of animals used and their suffering.

Behavioural measurements and analysis
Two exposures to an open eld were done to compare the anxiety-related and locomotive behaviour of the two rat strains. The rst exposure trial provided a novel environment, while the second trial allowed us to observe exploratory motivation after the rats were kept in individual cages. The rst exposures and behavioural measurements were conducted three days after rats were delivered from the distributer. Rats were then subjected to a 6-min forced swim test (FST). After seven additional days of being housed individually, the second open-eld test (OFT) trial was performed (Fig. 1a).

Open eld test (OFT)
Animals were placed in an OFT apparatus (90 × 90 × 40 cm 3 ) to freely explore the eld for 10 min, and an overhead camera and tracking software recorded all movements (O'Hara and Co., Tokyo, Japan). The surface and walls of the apparatus were cleaned with ethanol before and after each session. To track movements, the apparatus was divided into 25 blocks and blocks 7 to 9, 12 to 14, and 17 to 19 were de ned as the centre region ( Supplementary Fig. 1a). The total distance (cm), total movement duration (s), total number of movement episodes, average speed (cm/s), moving speed (cm/s), distance per movement (cm), duration per movement (s), wall-side time (s), percent time in centre region (%), total time in centre region (s), and the time spent in each of the 25 blocks (s) were measured as behavioural parameters (hereafter simply referred to as parameters). Statistical analysis was performed with a repeated-measures general linear model (GLM) with a covariate of total intracranial volume (TIV) and a modi ed multiple comparisons test using Bonferroni's method and SPSS Statistics ver. 23.0.0.2 (IBM, Armonk, NY, USA).

Forced swim test (FST)
After the rst OFT measurements, animals were placed in a 60 cm × 20 cm acrylic resin cylinder (O'Hara and Co.) containing water (height: 30 cm; temperature: 23°C) for 6 min. The time spent immobile, which included oating and movements necessary for breathing, was assessed using video tracking software with two cameras (O'Hara and Co.). The total duration of swim movements was also measured. The freezing time was de ned as the time during which the two cameras simultaneously recorded no movement. Two-tailed two-sample t-tests (Welch's method) were applied for the statistical analysis.

Tissue preparation
After the behavioural measurements were taken, Wistar and WKY rats were deeply anesthetised with pentobarbital, then perfused and xed with 4% paraformaldehyde. Rat skulls, including the brain, were removed and immersed in the same xative overnight, then stored at 4°C until MRI.

Ex vivo brain MRI
We modi ed our previously established ex vivo brain MRI technique 72 to acquire images of four rats simultaneously using a cradle (Rehabitech, Kyoto, Japan; Supplementary Fig. 2). Brie y, a Helmholtz small-volume coil (probe dimension: 108/63) was used for both radiofrequency excitation and signal detection in a Voxel-based morphometry (VBM) and voxel-based correlation analysis MRI images were segmented to identify grey matter and spatially normalised with diffeomorphic anatomical registration using the exponentiated lie algebra (DARTEL) protocol in statistical parametric mapping software ver.12 (SPM12) and the in-house rat brain template and probabilistic maps created in our previous study 72 . We also followed the VBM methods detailed previously 72 . Brie y, experimental brain MR images were denoised with a 3D non-local means lter, co-registered, and segmented to identify grey matter before spatial normalisation 72 . Jacobian determinants from the DARTEL procedure were calculated and used for modifying the processed grey matter images. The images were resampled into 0.15 × 0.15 × 0.15 mm 3 voxels and smoothed with an isotropic Gaussian kernel of 0.8 mm in the original space, which was full-width at half-maximum. For the group-level statistical analysis, we performed voxel-by-voxel two-sample t-tests across the whole brain by applying the GLM. TIV was set as a covariate in all analyses. The signi cance level was rst set at a cluster height threshold of p < 0.001, with a determinant extent threshold of p < 0.05 corrected for family-wise error (FWE). We then used a more conservative peak level threshold of p < 0.05, corrected for multiple comparisons using the FWE correction method with a minimum cluster size of 19 voxels, based on the number of expected voxels per cluster (K = 18.563).
Voxel-based correlation analysis was performed to determine the association between parameters (except the time spent in each block and brain volume in WKY rats). We used only signi cant voxels, which were detected by comparing the WKY and Wistar rats using the signi cant voxel map (p < 0.05, cluster FWE-corrected) as a brain mask. An uncorrected p < 0.001 with a minimum cluster size of 19 voxels was considered signi cant. Pearson's r values were also calculated between parameters and mean grey matter voxel values of signi cant clusters. Tables   Due to technical limitations, table 1  test (FST). The rats were subjected to an OFT and FST on Day 4 and sacri ced on Day 11, after the second OFT. b-k) Locomotion parameter measurements of Wistar (n = 13) and Wistar-Kyoto (WKY; n = 13) rats from both OFT trials: b) total distance; c) total movement duration; d) total number of movement episodes; e) moving speed; f) duration per movement; g) distance per movement; h) average speed; i) wall-side time; j) percent of time spent in the centre region; and k) total amount of time spent in the centre region. *: p < 0.05; **: p < 0.01, ***: p < 0.001; analysis using repeated-measures general linear model (GLM) with Bonferroni correction with total intracranial volume (TIV) as a covariate. Blue box plots, Wistar rats; orange box plots, WKY rats. l, m) Statistical heat maps for block stay durations between the two OFT trials overlaid on a schematic of the apparatus. l) Wistar rats, n = 13. m) WKY rats, n = 13. Rats started to explore the open eld from block 1. Red shading indicates that block time during the second trial was less than the rst trial; tint indicates statistical signi cance (light red: p < 0.05; repeatedmeasures GLM with Bonferroni correction and TIV as a covariate). Blue shading indicates that time spent in the block during the rst trial was less than the second trial; tint indicates statistical signi cance (light blue: p < 0.05; dark blue: p < 0.001). n) Freezing time, de ned as the percent of time spent immobile during the FST, was signi cantly higher in WKY rats (n = 13) than in Wistar rats (n = 11; t = −4.297; p < 0.01; two-tailed Welch's t-test with TIV as covariate). o) Total movement distance (cm) as a measure of locomotor activity during the FST was signi cantly shorter in WKY rats (n = 13) than in Wistar rats (Wistar: n=11; t = 4.404; p < 0.01; two-tailed Welch's t-test with TIV as covariate). FWE-corrected; minimum voxel size: 19). With this threshold, atrophic clusters were present in the cerebellar nuclei (A1 and A3-6), cerebellar vermis (A2 and F1), and parahippocampus (B1-3). The letters used are aligned with the cluster identi cation labels from the moderate threshold results. Colour bar units represent t-scores.

Figure 3
Clusters with negative correlations in Wistar-Kyoto (WKY) rats (n = 13). A threshold level of p < 0.001, uncorrected for multiple comparisons, and a minimum of 19 voxels in a cluster were applied. a) Total movement duration. b) Total number of movement episodes. c) Duration per movement. d) moving speed. The negatively correlated clusters were in the right habenula, left ventral hippocampus, left amygdala, and right lateral septum. The adjacent graphs show the correlation between a given behaviour and the grey matter value inside the clusters detected by the voxel-based correlation analysis. Colour bar units represent t-scores. Figure 4