Effects of single cage housing in the rat and mouse pilocarpine models of epilepsy

Manouze H1,2, Ghestem A2, Poillerat V3, Bennis M1, Ba-M’hamed S1, Benoliel JJ3,4, Becker C3*, Bernard C2* 1Lab of Pharmacology, Neurobiology & Behavior (URAC-37), Cadi Ayyad University, Marrakech, Morocco 2Aix Marseille Univ, Inserm, INS, Institut de Neurosciences des Systèmes, Marseille, France Université Paris Descartes, Sorbonne Paris Cité, Faculté de Médecine, F-75006, Paris, France AP-HP, Hôpital de la Pitié-Salpêtrière, Service de Biochimie Endocrinienne et Oncologique, F-75013, Paris, France

The experimental protocol is shown in Figure 1A. Swiss male mice were housed in groups 92 of 3 per cage. In order to validate the stress state of same mouse throughout experimental 93 protocol, one mouse from each litter was randomly picked, ear-marked and returned to its 94 own cage. They were left together during one week without experimental intervention to 95 allow adaptation to a novel housing environment. After one week, we performed different 96 behavioral tests for each ear-marked mouse in order to obtain reference levels. Sucrose 97 preference was assessed every day during 1 week. At the end of the week, we evaluated 98 anxiety levels (Day-3: elevated plus maze -EPM) and performance in the novel object 99 recognition (NOR) test (Day -2-1). Animals were then randomly assigned to two groups to 100 receive a pilocarpine injection to trigger status epilepticus (pilo group-experimental 101 procedure described hereafter) or saline (non-pilo group). Both groups were further divided 102 into two groups: group-housing (social condition -SC, n =18, 3/cage) and one individual-103 housing condition (isolated condition -IC, n = 6, 1/cage). Individual housing used ear marked 104 mice. After 4 weeks exposure to different housing conditions, ear-marked mice from social 105 and isolated conditions were tested for anhedonia (Day 46-53), EPM (Day 54) and NOR (55-106 56). Behavioral tasks (EPM and NOR) were recorded and analyzed using 107 Ethovision®XTNoldus 8.5 video tracking program (Noldus, Netherlands) connected to a 108 video camera (JVC). At the end of each behavioral session apparatuses were cleaned with a 109 75% ethanol solution to remove any of odor or trace. The behavioral tests were performed 110 between 8:00 and 12:00 a.m. during the light cycle to avoid the circadian-related fluctuation 111 in the performance of the mice. Before the beginning of behavioral tests, the animals were 112 transferred to the testing room in their home cages and left there to habituate for 60 min. 113 At the end of the experimental protocol, we measured ACTH, CORT and BDNF levels as 114 in rats (described below). 115 116

In rats 117
The experimental protocol is shown in Figure 2A. Animals were received in sets of 4 from the 118 same litter from the vendor. Two main groups of animals were used: animals with 119 spontaneous seizures following pilocarpine-induced status epilepticus (pilo group -120 experimental procedure described hereafter) and control animals (non-pilo group). Pilo and 121 non-pilo animals were further divided into three groups: 122 -Isolated group: rats were singly housed and were not handled during the experimental 123 period except for cage cleaning and body weight measurement once a week. 124 -Handled group: rats were singly housed and were handled daily until the end of 125 experiment (see below). 126 -Paired group: rats were kept in buddy pairs with social interaction upon arrival in the 127 animal facility. 128 . CC-BY 4.0 International license available under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made A preliminary analysis made on isolated (n=4) and handled (n=4) pilo groups for seizure 142 frequency allowed us to perform a power analysis. We determined that n=3 per group was 143 enough. The experimental procedure in the pilo groups was conducted twice. The first and 144 second group of simultaneously recorded animals being composed of (n=4 isolated, n=4 145 handled, n=6 paired) and (n=3 isolated, n=4 handled, n=6 paired), respectively. In the paired 146 groups, only one animal was instrumented. Seizure analysis was thus performed in 6 animals. 147 Results were similar in the two sets; and data was pooled together. 148 For the mice experiments, the number of subjects was defined on the basis of the study 149 design. Preliminary analysis was conducted with n=3 for social condition and n=3 for isolated 150 condition where the results were non-significant. In this case, a second study was done with a 151 sample size of 3 per group. 152 153 . CC-BY 4.0 International license available under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which was this version posted March 29, 2019. ; https://doi.org/10.1101/214528 doi: bioRxiv preprint

Status epilepticus induction and electrode implantation 154
In rats, status epilepticus (SE) was induced by a single i.p injection of pilocarpine (pilo) 155 320 mg/kg, one week after receiving the animals from the vendor. In mice, we used repeated 156 100 mg/kg i.p. injections every 20 min until SE onset 17 days after receiving the animals (1 157 week for habituation and 10 days for baseline of behavioural tasks). To reduce peripheral 158 effects, animals were pre-treated with Methyl-scopolamine (1mg/kg) 30 min prior pilo 159 injection. SE was stopped by diazepam (10 mg/kg i.p., twice within a 15 min interval) after 60 160 min and 90 min of SE; respectively. At the end of these injections, mice and rats were 161 hydrated with saline (2ml i.p. twice within 2 h) and fed with a porridge made of soaked 162 pellets, until they resumed normal feeding behavior. All drugs were obtained from Sigma. 163 In rats, four weeks following SE, the telemetry implant was surgically inserted 164 intraperitoneally under anesthesia (ketamine [1 mg/ kg]/xylasine [0.5 mg/kg] i.p) and 165 connected to screws on the surface of the brain by two electrodes; one above the cortex (4.0 166 mm anteroposterior, 2.0 mm mediolaterally, compared with bregma), the second, above the 167 cerebellum as reference. The EEG signal was transmitted by radio to a data acquisition and 168 processing system (DSI). In the paired group, both animals developed epilepsy but only one 169 rat was equipped with the telemetry system as two animals cannot be recorded simultaneously 170 with EEG transmitters in the same cage, while the other was monitored with video only. 171 Animals were left to recover during one week before switching on the transmitter. 172 173

Monitoring of spontaneous recurrent seizures 174
Continuous (24/7) video-EEG recordings started in the 6 th week after SE, a period sufficient 175 to reach stability in seizure frequency (Williams et al., 2009), and were stopped at week 10. 176 We verified that animals displayed stable seizure activity, quantifying seizure frequency 177 during each successive week. Spontaneous recurrent seizures were detected and quantified 178 . CC-BY 4.0 International license available under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made In mice, the baseline for sucrose consumption was assessed every day for 1 week, 7 191 days after their arrival in laboratory. The sucrose preference was calculated at the 5 th week 192 following the separation and treatment. In rats, sucrose consumption was assessed at week 4 193 and week 8 following animal reception. Sucrose and water intakes were measured daily at ZT 194 2. 195 Briefly, animals were given a free choice between two bottles, one with 1% sucrose 196 solution and another with tap water. The location of the bottles was alternated every day to 197 prevent possible effects of side preference in drinking behaviour. The consumption of water 198 and sucrose solution was estimated by weighting the bottles. For the paired, the volume of 199 sweet water consumed by rat was taken as the total consumed volume divided by 2. For the 200 social groups, the volume of sweet water consumed by mouse was taken as the total 201 consumed volume divided by the total number of animals at each time. Sweet water 202 . CC-BY 4.0 International license available under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which was this version posted March 29, 2019. ; https://doi.org/10.1101/214528 doi: bioRxiv preprint consumption corresponds to sucrose preference, which is calculated as a percentage of the 203 volume of sucrose intake over the total volume of fluid intake using the following equation: Isolated, handled and paired rats in the non-pilo group were killed by decapitation after light 237 anesthesia with isoflurane at the beginning of the 8 th week following the separation at ZT4 238 and at the 6 th weeks after the separation and treatment at ZT4 for pilo and non-pilo mice. 239 Animals were decapitated in a quiet separate room, one by one, with the bench cleaned 240 between animals. Trunk blood was collected in less than 5 sec after decapitation in dry tube 241 and EDTA tube in order to obtain serum and plasma samples, respectively. The plasma was 242 prepared by a 15 min centrifugation at 1600g, 4°C. The ACTH concentration was determined 243 according the manufacturer's instructions (Clinisciences, France). For corticosterone and 244 BDNF levels, blood was centrifuged at 3500g for 10 min at 4°C and the serum was stored at 245 80°C until used. Corticosterone and BDNF concentrations were determined according to the 246 manufacturer's instructions (Coger, Promega, France). 247 248

Statistical analysis 249
Statistical analysis was performed using Sigma Plot 11.0 software. All data is presented as 250 mean ± SEM. The effect on body weight and anhedonia (rat) was measured using the 251 ANOVA with repeated measures followed by Bonferroni post-hoc analysis. The biological 252 . CC-BY 4.0 International license available under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which was this version posted March 29, 2019. ; https://doi.org/10.1101/214528 doi: bioRxiv preprint and seizures parameters were tested according to the one-way analysis of variance (ANOVA) 253 followed by Bonferroni multiple comparison test. The effects of group and pilo treatment on 254 anxiety index, discrimination index, ACTH, BDNF and corticosterone were measured using 255 the two-way ANOVA followed by Bonferroni post-hoc analysis. The effects on anhedonia 256 was measured using the non-parametric Mann-Whitney U. The significance threshold was set 257 at p<0.05. 258 259 260 261 . CC-BY 4.0 International license available under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which was this version posted March 29, 2019. ; https://doi.org/10.1101/214528 doi: bioRxiv preprint

RESULTS 262
We first tested the consequences of social isolation on behavioral parameters and stress 263 markers in adult pilo and non-pilo mice ( Figure 1A). 264 Repeated measures ANOVA showed a significant effect of time (F (5,100) =128.08, 265 p<0.0001 a ) on body weight evolution; while the experimental group (F (3,20) =2.23, p=0.11 a ) 266 and the interaction of group and age (F (15,100) =1.18, p=0.29 a ) had no effect (Figure 1B). At 267 the 4 th and 5 th week post-social isolation, post-hoc analysis showed that both pilo-SC and pilo-268 IC groups gained significantly less weight than the non pilo-SC and the non pilo-IC group. 269 There was no significant difference between all groups until the 3 rd week. Thus, the pilo group 270 gained less weight than the non-pilo group, but housing conditions had no effect in both 271

groups. 272
We used the sucrose preference test to assess anhedonia (lack of interest in rewarding 273 stimuli). The typical behaviour of animals is a bias toward the sweetened drink. Lack of 274 preference for the sweetened drink indicates anhedonia, which signs stress-related disorders 275 (Gold, 2015). At baseline (before changing housing conditions), animals from different 276 groups (different cages) did not show differences in sucrose preference ( Figure 1C). Four 277 weeks later, our results demonstrated that mice injected by pilo consumed less of sweet water 278 than mice injected with vehicle. In addition, mice in isolated condition had sweet water 279 consumption lower than those in social condition ( Figure 1C). From day 1 until day 7, the 280 non-parametric Mann-Whitney U test, applied on sweet water consumption, revealed that the 281 pilo-SC (day 1: U=0.00; Z=-10.89; p<0.01; day 7: U=0.00; Z=-9.35; p<0.01 B-1 ) and pilo-IC 282 (day1: U=0.00; Z=-10.23; p<0.01; day 7: U=0.00; Z=-10.09; p<0.01 B-1 ) groups consumed 283 significantly less of sweet water than the non pilo-SC and non pilo-IC groups. Additionally, 284 the Mann-Whitney U showed that both non pilo-IC (day1: U=2.00, Z= -4.87, p<0.05; day 7: 285 U=1.50; Z= -4.72; p<0.01 B-1 ) and pilo-IC (day 1: U=1.00; Z=-4.42; p<0.01; day 7: U=5.00; 286 . CC-BY 4.0 International license available under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which was this version posted March 29, 2019. ; https://doi.org/10.1101/214528 doi: bioRxiv preprint Z= -5.00; p<0.05 B-1 ) groups had a significant decrease in sucrose preference as compared to 287 non pilo-SC and pilo-SC groups, respectively. Moreover, the Mann-Whitney U test did not 288 show any effect of day on sweet consumption (p>0.05). Therefore, social isolation induces 289 anhedonia in the control (non pilo) group. The pilo group socially housed also displayed 290 anhedonia as compared to the control group, but the anhedonia phenotype was increased in 291 the socially isolated pilo group. 292 We then tested animals for anxiety-like behavior. All groups tested before separation 293 had similar anxiety levels ( Figure 1D) p<0.0001 b ). Post-hoc comparisons confirmed that the pilo-treated groups showed a significant 297 increase in anxiety (p<0.01 and p<0.001) with respect to the non pilo groups ( Figure 1D). In 298 the isolated condition, the anxiety index was significantly larger in both pilo and non pilo 299 mice as compared to mice in social housing (p<0.05). These results show that social isolation 300 readily increases anxiety levels in the control group. The pilo group maintaining social 301 contact displayed increased anxiety as compared to the control group (epilepsy effect), but 302 anxiety was further exacerbated in the pilo isolated group (combining the effects of epilepsy 303 and social isolation). 304 Next, we assessed the effect of social isolation on memory performance using the NOR 305 test. All groups performed similarly at baseline before separation and treatment ( Figure 1E). 306 The two-way ANOVA test showed significant difference in the discrimination index among not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which was this version posted March 29, 2019. ; https://doi.org/10.1101/214528 doi: bioRxiv preprint prevented us to test the effect of social isolation on epilepsy severity. This issue was 337 investigated in the pilocarpine rat model. 338 Because of their larger size, it was not technically possible to maintain rats in small 339 colonies while performing wireless 24/7 recordings. For the social group, animals were kept 340 as pairs (only one was equipped with wireless transmitter, but both received pilocarpine 341 injection). We added a third group (handled) of singly housed animals but which could 342 interact with the experimenter everyday (see methods). The rationale for this group is that 343 some EEG studies are performed with wired systems, which precludes group housing, as 344 animals tend to severe the wires. We first analysed general stress markers in control (non pilo) 345 animals to determine whether social isolation had similar effects to those described in mice. 346 The ANOVA repeated measures showed a significant effect of social condition (F (2,17) =41.58, 347 p<0.0001 g ) and time (F (6,102) =266.83, p<0.0001 g ) on the gain of body weight (Figure 2B) p<0.05, p<0.01 and p<0.001). Thus, just two weeks of social 358 isolation is sufficient to produce a state of anhedonia. 359 We then tested stress-related ACTH and corticosterone hormones. One-way ANOVA 360 revealed a significant effect of social isolation on hormone levels (ACTH: F (2,17) =5.05, 361 . CC-BY 4.0 International license available under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which was this version posted March 29, 2019. ; https://doi.org/10.1101/214528 doi: bioRxiv preprint p<0.02 j ; corticosterone: F (2,17) =8.42, p<0.01 j ; Figure 2D). The post-hoc analysis showed that 362 ACTH and corticosterone levels were significantly increased in the isolated group as 363 compared to the handled (t=2.69, p<0.05 and t=3.57, p<0.01; respectively) and paired (t=2.93, 364 p<0.05; t=3.70, p<0.01; respectively) groups. 365 Finally, we assessed serum BDNF levels. One-way ANOVA revealed a significant 366 effect of social isolation on BDNF levels (F (2,17) = 6.18, p<0.01 l ). BDNF levels were 367 significantly lower in the isolated group as compared to the handled (t=2.38, p<0.05) and 368 paired (t=3.47, p<0.01) groups ( Figure 1D). There was no significant difference between the 369 handled and paired groups in all tested biological parameters (ACTH: t=0.18; corticosterone: 370 t=0.42 and BDNF: t=1.02, p>0.05). 371 We conclude that social isolation in the strains of control rats and mice used here 372 produces aphenotype associated with stress. We then compared how different housing 373 conditions influence the epilepsy phenotype. 374 One-way ANOVA analysis showed a significant difference between groups for seizure 375 parameters (frequency, duration and severity) (F (2,26) =10.73, p<0.001; F (2,26) =30.90, p<0.0001; 376 F (2,26) =4.07, p<0.05; respectively m,n,o ) (Figure 3). showed that the seizure frequency and duration were considerably increased in the isolated 379 group as compared to the handled and paired groups (handled: t=4.29 and t=7.82, p<0.001; 380 paired: t= 3.45 and t=4.07, p<0.01; respectively). The handled group significantly 381 demonstrated lower Racine scores (seizure severity) as compared to the isolated group 382 (t=2.72, p<0.05); whereas, there was no significant difference between the paired and isolated 383 groups (t=1.94, p>0.05). We only assessed the severity of the seizures during the light phase 384 in the animal facility. Finally, there was no significant difference between the handled and 385 paired groups for seizure frequency and seizure severity. Thus, lack of social interaction 386 . CC-BY 4.0 International license available under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Swiss mice in the pilocarpine experimental model of epilepsy. It is important to note that 394 daily social interaction with the experimenter was sufficient to prevent the development of a 395 severe epilepsy phenotype in rats although they were singly housed. It was as efficient as 396 keeping animals in pairs. Although we could not quantify it, we noted that isolated pilo 397 animals were very aggressive and displayed escaping behaviour when cages were changed 398 every week. Their cages were close to one another, suggesting that visual contact was not 399 sufficient to prevent the effect of social isolation. Abnormal reaction to handling is a direct 400 consequence of single housing (Hatch et al., 1965). In contrast, rats and mice socially housed 401 (or in daily contact with an experimenter) remained calm when handled. In the laboratory, we 402 are also performing electrophysiological recordings with high density silicon probes in 403 experimental rat models of epilepsy. The equipment is protected with a copper mesh hat, 404 which precludes social housing. We are using large cages, divided in two, separated by a grid We demonstrate that epilepsy in socially housed mice produces an activation of the HPA axis 423 and anhedonia, responses which are amplified in isolated mice with epilepsy. We found the 424 same exacerbation in the rat experimental model between isolated and paired rats. A strong 425 activation of the HPA axis in isolated animals with epilepsy may contribute to the expression 426 of co-morbidities such as anxiety, cognitive deficits and depression-like behaviour. 427 In rats with epilepsy, we found that spontaneous seizures were 16 times more frequent 428 in isolated animals as compared to animals kept in pairs or singly housed animals but with 429 daily interactions with experimenter. During the 3 weeks of continuous recordings (starting 6 430 weeks after pilocarpine-induced status epilepticus), we did not find an increase in seizure 431 frequency (Figure 3-1 Finding the mechanisms underlying the difference in phenotypes between the different 450 housing conditions was beyond the scope of this study. We do not claim that the strong 451 seizure phenotype found in the isolated group is solely due to the stronger activation of the 452 HPA axis, although it may contribute to it. We may tentatively propose that social isolation 453 produces a vulnerability phenotype, as suggested by the low levels of serum BDNF (Becker et  454 al., 2015). We suggest that epileptogenesis occurring in networks "weakened" by social 455 isolation-induced stress would result in a strong phenotype. 456 The field of epilepsy research makes use of many different experimental models, 457 strains and species. Results are clearly model-, strain-and species-specific, which may 458 explain some conflicting results. We propose that another factor to consider is housing 459 conditions. Some patients with epilepsy may experience social isolation, which correlates 460 . CC-BY 4.0 International license available under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which was this version posted March 29, 2019. ; https://doi.org/10.1101/214528 doi: bioRxiv preprint However, their conclusion that antioxidant treatment is strongly disease modifying in the 486 context of social isolation is not only valid but also highly clinically relevant to socially 487 isolated patients. Finally, many (often unpublished) preclinical studies performed in isolated 488 animals did not evidence any significant effect on seizure frequency. Since the social isolation 489 stress component could act as a confounding factor producing a strong phenotype, there may 490 be false negative results; i.e. the tested drugs may be very efficient in experimental animals 491 maintaining social interaction. 492 493 Although we cannot claim that maintaining social interaction between rodents in 494 laboratory conditions corresponds to "normality" (as in the wild), single housing does 495 produce strong biological alterations that render data interpretation more complex, in both 496 physiological and pathological conditions. We recommend that material and methods should 497 systematically report housing conditions, and that all efforts should be made to maintain 498 social interaction. Finally, these results were obtained in one experimental model 499 (pilocarpine), in two species, one strain for each species and only in males. Similar studies 500 should be performed in females, other strains and models such kainic acid, kindling, tetanus 501 toxin etc. Our results and conclusions should not be seen as a criticism of previous works. 502 Rather varying housing conditions provide a unique opportunity to study, with the same 503 experimental model, different situations found in patients with high and low stress levels. 504 . CC-BY 4.0 International license available under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which was this version posted March 29, 2019. shown for isolated (n=11), handled (n=12) and paired rats (n=6). No significant difference 656 was shown between weeks in seizure frequency for all experimental groups. Other panels 657 show the average seizure frequency per week for each animal. Two animals (out of 11) in the 658 isolated group had a seizure frequency similar to that found in paired or handled animals 659 (0.03/h). 660 Data are mean ± SEM. ANOVA two way followed by Bonferroni post-hoc, *p<0.05 and 661 **p<0.01 refers to handled or paired groups vs. isolated group comparison. 662 663 . CC-BY 4.0 International license available under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which was this version posted March 29, 2019. ; https://doi.org/10.1101/214528 doi: bioRxiv preprint A-Statistical analysis of the effects of mice groups (grouped/isolated) and treatment (Veh and pilo) on behavioral, biological and biochemical parameters: . CC-BY 4.0 International license available under a not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which was this version posted March 29, 2019. ; https://doi.org/10.1101/214528 doi: bioRxiv preprint