Inhibiting adult neurogenesis differentially affects spatial learning in females and males

Adult hippocampal neurogenesis has been implicated in the spatial processing functions of the hippocampus but ablating neurogenesis does not consistently lead to behavioral deficits in spatial tasks. Parallel studies have shown that adult-born neurons also regulate behavioral responses to stressful and aversive stimuli. We therefore hypothesized that spatial functions of adult-born neurons may be more prominent under conditions of stress, and may differ between males and females given established sex differences in stress responding. To test this we trained intact and neurogenesis-deficient rats in the spatial water maze at temperatures that vary in their degree of aversiveness. At standard temperatures (25°C) ablating neurogenesis did not alter learning and memory in either sex, consistent with prior work. However, in cold water (16°C), ablating neurogenesis had divergent sex-dependent effects: relative to intact rats, male neurogenesis-deficient rats were slower to escape and female neurogenesis-deficient rats were faster. Neurogenesis promoted temperature-related changes in search strategy in females, but it promoted search strategy stability in males. Females displayed greater recruitment of the dorsal hippocampus than males, particularly at 16°C. However, blocking neurogenesis did not alter activity-dependent immediate-early gene expression in either sex. Finally, morphological analyses of retrovirally-labelled neurons revealed greater experience-dependent plasticity in new neurons in males. Neurons had comparable morphology in untrained rats but 16°C training increased spine density, and 25°C training caused shrinkage of mossy fiber presynaptic terminals, specifically in males. Collectively, these findings indicate that neurogenesis functions in memory are prominent under conditions of stress, they provide the first evidence for sex differences in the behavioral function of newborn neurons, and they suggest possibly distinct roles for neurogenesis in cognition and mental health in males and females.

Adult hippocampal neurogenesis has been implicated in the spatial processing 23 functions of the hippocampus but ablating neurogenesis does not consistently lead to 24 behavioral deficits in spatial tasks. Parallel studies have shown that adult-born neurons 25 also regulate behavioral responses to stressful and aversive stimuli. We therefore 26 hypothesized that spatial functions of adult-born neurons may be more prominent 27 under conditions of stress, and may differ between males and females given 28 established sex differences in stress responding. To test this we trained intact and 29 neurogenesis-deficient rats in the spatial water maze at temperatures that vary in their 30 degree of aversiveness. At standard temperatures (25°C) ablating neurogenesis did not 31 alter learning and memory in either sex, consistent with prior work. However, in cold 32 water (16°C), ablating neurogenesis had divergent sex-dependent effects: relative to 33 intact rats, male neurogenesis-deficient rats were slower to escape and female Adult hippocampal neurogenesis has been implicated in many of the mnemonic functions of 50 the hippocampus, including memory for temporal events (1-3), locations (4), contexts (5, 6), 51 objects (7, 8) and conspecifics (9), as well as the consolidation (10, 11) and forgetting (12) of 52 memory. While spatial memory functions may be particularly apparent in conditions that 53 maximize conflict or interference, such as when a goal changes location (13-16), it is notable 54 that many of studies have failed to find a role for new neurons in learning and short-term 55 reference memory in the spatial water maze, a task that is highly sensitive to hippocampal 56 disruption (2, 6, 7, 17-22). 57 A relatively independent body of work has focused on the role of neurogenesis in 58 emotional and stress-related behavior, finding that neurogenesis buffers the endocrine 59 response to acute stressors and reduces depressive-and anxiety-like behavior (23-29). Since 60 stress and emotion potently modulate learning and memory (30, 31), here we hypothesized 61 that a role for neurogenesis in spatial learning may become particularly apparent in more 62 aversive conditions. Consistent with this possibility, a small number of studies have found that 63 neurogenesis does alter behavior in memory tasks depending on the aversiveness of 64 conditioned and unconditioned stimuli that are present (3, 32, 33). 65 Stress-related disorders such as anxiety, PTSD and depression impact a substantial 66 fraction of the population. Critically, these disorders affect females to a greater extent than 67 males, suggesting that neurogenesis functions in stress may be particularly relevant for female 68 cognition and mental health (34). Indeed, there are known sex differences in the rates of 69 addition (35), maturation (36) and activation of adult-born neurons (37). Furthermore, sex 70 modulates hippocampal plasticity (38-41) and behavioral responses to acute and chronic stress 71 (42-44). However, as is the case in neuroscience more broadly (45), the majority of 72 neurogenesis studies have focused on males (46). In a quantitative survey of the neurogenesis 73 literature we find that males are studied twice as often as females, less than 10% of studies 74 have reported data by sex, and more than 20% of studies do not report the sex of their 75 subjects ( Fig. 1). To our knowledge, no sex differences in behavior have been reported in 76 animals that have specific reductions in adult neurogenesis. One study has found reduced 77 neurogenesis is associated with female-specific impairments in adult learning (47). However, 78 this was in response to neurogenesis ablation beginning in infancy, raising the question of 79 whether sex differences may also occur in response to neurogenesis ablation in adulthood.

80
To address these outstanding issues we used a pharmacogenetic GFAP-TK rat model to 81 block adult neurogenesis (48), and tested male and female rats in the water maze at warm 82 (25°C, standard) or cold (16°C, more aversive/stressful) temperatures. Consistent with previous 83 work, neurogenesis-deficient rats were unimpaired at standard water maze temperatures. 84 However, cold water testing revealed striking sex differences in the behavioral function of adult 85 neurogenesis, and also elicited distinct dorsoventral patterns of hippocampal recruitment and 86 new neuron plasticity in males and females. 87 88 89

91
Subjects 92 This study used male and female transgenic GFAP-TK ("TK") and wild-type ("WT") rats on a 93 Long-Evans background (48). Here, a GFAP promoter drives expression of herpes simplex virus 94 thymidine kinase in radial-glial precursor cells, enabling these cells to be killed when rats are 95 treated with valganciclovir and the cells attempt mitosis. Rats were bred in-house, by crossing 96 heterozygous transgenic females with WT males. After weaning (postnatal day 21) rats were 97 housed in same-sex groups of 2-3 in polyurethane cages (48 cm x 27 cm x 20 cm), with aspen 98 chip bedding, a polycarbonate tube for enrichment, and ad-libitum access to food and water. 99 Animals were housed under a 12-hour light:dark cycle, and all testing was completed during 100 the light phase. Rats were genotyped via PCR after weaning and, therefore, housed randomly 101 with respect to genotype. Prior to all experiments, animals were handled 5 min/day for 5 days.

102
Experimental procedures were approved by the University of British Columbia Animal Care 103 Committee and followed guidelines from the Canadian Council of Animal Care on the ethical 104 treatment of animals.

106
Valganciclovir treatment 107 For experiments with neurogenesis ablation, animals were treated orally with pellets of 108 valganciclovir (4 mg) in a 1:1 peanut butter and rodent chow mix (0.5 g). Drug pellets were 109 given directly to each animal to ensure accurate dosing. Animals began treatment at 6-7 weeks 110 of age, and were treated twice a week (3-4 day interval) for 6-7 weeks before behavioral testing 111 began. Valganciclovir treatment stopped immediately prior to behavioral testing. In control 112 experiments without neurogenesis ablation, rats received neither valganciclovir nor peanut 113 butter and rodent chow mix.

115
Spatial water maze testing 116 The water maze consisted of a white circular pool (180 cm diameter), with 60 cm high walls. 117 The pool was filled with water to a 32 cm depth, and the water was made opaque with addition 118 of white non-toxic liquid tempera paint (Schola). Training contexts of high-or moderate-stress 119 were created by using either 16°C or 25°C water, respectively, similar to previous work (49, 50). 120 The pool was located in a room (~4m x 6m in size) with diffuse lighting, and contained extra-121 maze visual cues along the room's walls and distributed within the room (desk, computer, 122 cabinets). A circular escape platform (12 cm diameter) was placed in the NE quadrant of the 123 pool, and was positioned 2 cm below the water surface. Rats received 3 days of acquisition 124 training with 4 trials per day. Rats were tested in groups of 2-3, and during daily training 125 sessions were placed into individual holding cages filled with aspen chip bedding and paper 126 towel. 127 For each trial, rats were placed into the pool at one of four possible release locations 128 (pseudo-random order), with each release location occurring once on each day. Rats were 129 given a maximum of 60 sec to locate the escape platform, after which they were guided to the 130 escape platform by the experimenter. Following each trial, rats remained on the escape 131 platform for ~10 sec, and were gently dried with a towel before being returned to their holding 132 cage for the inter-trial interval (30-90 sec). The rats' trajectory was recorded with an Ethovision 133 (Noldus) tracking system, and performance was assessed via latency to locate the escape 134 platform and swim speed. Ideal path error (conceptually similar to cumulative search error / 135 proximity metrics (51)), which can detect spatial performance differences between trials that 136 have similar latencies and distances (52), was calculated with Pathfinder software as follows: the 137 distance from the platform was summed over all samples to obtain a cumulative distance 138 metric. To control for different release locations, the cumulative distance for the optimal path 139 was also calculated based on a direct escape path from the release location and the average 140 swim speed. The ideal path error was then calculated by subtracting the cumulative optimal 141 path from the cumulative actual path. On the day following acquisition training the platform 142 was removed from the pool and rats completed a 60 sec probe trial to assess memory. Spatial 143 memory was measured as the time spent in a 36 cm zone surrounding the former escape 144 platform location, and the corresponding 36 cm zones in each of the non-target quadrants. 145 Rats were euthanized 60 minutes after the probe trial in order to capture experience-146 dependent Fos expression in activated neurons (see Immunohistochemistry, below).

148
Search strategy analyses 149 Navigational search strategies employed in the water maze were detected using Pathfinder 150 software ( To visualize the change in usage of a given strategy, S, caused by reducing 164 neurogenesis, weighted difference scores were calculated as: 165 (% trials STK -% trials SWT)/% trials SWT) × (# trials SWT+TK)/(# trials totalWT+TK) × 100. In other words, 166 strategy difference scores were weighted against their relative frequency, to prevent 167 overrepresentation of differences that occurred on only a small proportion of the total trials. 168 169 Retrovirus injections 170 Moloney Murine Leukemia-Virus retrovirus, produced as recently described (53), was use to 171 express eGFP in adult-born neurons. Viral titers ranged from 1 to 8 x 10 6 colony forming 172 units/ml. Eight-week-old male and female rats were bilaterally injected with 1 µl of retrovirus 173 into the dorsal dentate gyrus (anteroposterior = -4.0 mm; mediolateral = ±3.0 mm; 174 dorsoventral = -3.5 mm from bregma). Thirty days later, rats either remained in their home 175 cage or were trained and tested for 4 days in the 16°C or 25°C water maze, as above. Rats 176 were perfused the next day, when cells were 35 days old.

178
Blood sampling and radioimmunoassays 179 In one group of rats, different from those used to generate the main behavioral data in Figures  180 2-3, blood samples were obtained 30 min following testing sessions on days 1 and 3 of 181 acquisition training and after the probe trial on day 4. After the last trial of a training session 182 was completed, rats remained in the testing room for 5 min, before being returned to their 183 home-cage and colony room for the remaining 25 min. Rats were then quickly brought into the 184 hallway adjacent to the colony room, restrained, and blood was collected via tail vein puncture. 185 For baseline circadian measurements, home cage control rats were sampled directly from their 186 cage without transport. Blood was left at room temperature for 30-45 min, centrifuged, and 187 serum supernatant was collected and stored at -80°C until analyzed by radioimmunoassay 188 (RIA). RIAs were completed using a I 125 corticosterone competitive binding assay (MP 189 Biomedical). In a subset of animals, body temperature was also obtained immediately following 190 blood sampling using a rectal thermometer.

192
Vaginal lavage and estrous staging 193 Vaginal lavages were completed on female rats within 1-6 hours of completing the probe trial.

194
Rats were gently wrapped in a towel and rotated so that the vagina was clearly visible. The 195 vagina was then flushed with tap water using a glass transfer pipette with a smooth, curved tip.

196
The water was then aspirated into the pipette and collected on a glass slide. The samples were 197 left to dry for at least 24 hours before being stained in cresyl violet (0.1% for 1 min). For 198 animals that were used in Figures 3-6, lavages were performed immediately prior to euthanasia 199 and perfusion, to prevent any effects of lavage on water maze behavior or experience-200 dependent Fos expression. Additionally, only a portion of the animals that were used for these 201 figures were lavaged. For animals that were used for corticosterone measurements, lavage was 202 performed at the same time blood was collected. Identification of estrous cycle stage was 203 completed based on the cytology of lavages, as described (54)  (3 x 5min), and incubated in DAPI (1:1000) for 10 min. Lastly, sections were washed for (3 x 250 5min) in PBS, mounted onto glass slides, and coverslipped using PVA-Dabco mounting 251 medium.

253
Quantification of immunolabelling 254 Quantification of all immunolabelling was completed by an experimenter blind to the 255 experimental conditions. For DCX, the number of immuno-positive cells was counted within 256 the granule cell layer of the DG, using an Olympus CX41 bright-field microscope with a 40x 257 objective. The number of immuno-positive cells were counted from 1 section of the 258 septal/dorsal hippocampus (Bregma, -2.92 to -4.0 mm). Counts of DCX cells were also 259 obtained from hippocampal sections which contained temporal/ventral hippocampus, although 260 counts were not separated between the supra-and infra-pyramidal blades (Bregma, -5.76 to -261 6.2 mm). Intermediate and ventral DG was delineated at 4.5 mm relative to the interaural line. 262 All counts of DCX positive cells were converted into densities based on the volume of the DG 263 subregions. 264 For quantification of Fos immunoreactivity, a confocal microscope (Leica, SP8) was used 265 to obtain representative z-stacks (40x objective), through the entire infrapyramidal and 266 suprapyramidal blades of the DG, the medial and lateral blades of the ventral DG, and dorsal 267 and ventral CA3. For each animal, an entire dorsal and ventral section was analyzed. Cells were 268 counted as Fos-positive when the intensity of immunolabelling was more than twice that of 269 neighboring, non-nuclei-containing, tissue in the hilus. To determine the percentage of GAD 270 cells that also expressed Fos, Gad immunopositive cells were examined throughout the entire 271 DG-CA3 and the proportion that expressed Fos at twice background levels was quantified. 272 Analyses of dendritic spine density were performed from z-stack images acquired with a 273 63x glycerol-immersion objective (NA 1.3). Images were 1024x1024 pixels in size, taken at 5x 274 zoom, a speed of 400 Hz, and a z-height of 0.5 µm. For each neuron, images were acquired 275 from the outer molecular layer (where lateral perforant path axons terminate), middle molecular 276 layer (where medial perforant path axons terminate), and inner molecular layer (where mossy 277 cell / commissural fibers terminate). All protrusions were counted as spines and mushroom 278 spines were defined as having a head diameter ≥ 0.6 µm. A total of 14-37 cells per group, 279 distributed equally across 3-5 animals/group, were analyzed. 280 Analyses of mossy fiber terminals were performed from z-stack images acquired with a 281 40x oil-immersion objective (NA 1.3). Images were 1024x1024 pixels in size, taken at 2x zoom, 282 a speed of 400 Hz, and a z-height of 0.5 µm. The area of the large mossy terminal was 283 measured from maximum intensity projections and the number of terminal-associated 284 filopodia, more than 1µm in length, was also quantified as a proxy for GABAergic interneuron 285 innervation (55, 56). Large mossy terminals and filopodia were categorized according to their 286 position along the proximodistal CA3 axis, where CA3a is the curved distal portion of CA3, 287 CA3c is proximal and enclosed within the blades of the DG, and CA3b is the intermediate CA3 288 region. A total of 59-122 large mossy terminals per group, distributed equally across 3-5 289 animals/group, were analyzed. 290 291 Statistical Analysis 292 Analysis of water maze acquisition performance was performed using mixed-design repeated 293 measures ANOVA with sex and genotype as between-subject factors and day of training as a 294 within subject factor. The distribution of search strategies in WT and TK rats was analyzed by a 295 Chi squared test with Bonferroni correction for multiple comparisons. Probe trial performance 296 was analyzed with between-subject ANOVAs (sex x genotype). For behavioral experiments, 297 16°C and 25°C groups were typically analyzed and presented separately; in some cases we 298 directly compared 16°C and 25°C groups to explore temperature effects. Cell densities were 299 analyzed by mixed design repeated measures ANOVA with sex and genotype as between 300 subject factors and dorsoventral subregion as a within subjects factor. Neuronal morphology 301 (spines, boutons, filopodia) was analyzed by ANOVA with sex and treatment as between-302 subjects factors. For all ANOVAs, where significant interactions were detected, post-hoc 303 comparisons were analyzed with Holm-Sidak tests. The significance level, , was set at p=0.05 304 for all tests. In most cases, statistical results are presented in the figure legends alongside their 305 respective data; for data that is not presented in figures, statistical results are presented in the 306 results text. 307 308 Sex and neurogenesis literature summary 309 To assess the degree to which sex has been included as a variable in studies of adult 310 neurogenesis (Fig. 1), a Pubmed search was performed using the search terms "neurogenesis" 311 and "dentate gyrus". Results were binned into 5-year increments from 2001 to 2020 and  112 studies/bin (mean=98, distributed equally over the 5 years of a bin) were examined for 313 whether they studied male, female or both male and female subjects, whether they formally 314 analyzed their data by sex, or whether they did not report the sex of their subjects.

323
Inhibition of neurogenesis in male and female TK rats. 324 To establish that neurogenesis was effectively inhibited along the dorsoventral axis of the DG 325 in both male and female TK rats, we quantified the density of cells expressing the immature 326 neuronal marker, doublecortin (DCX). As expected, in WT rats DCX + cells were observed at the 327 border of the granule cell layer and the hilus, in the subgranular zone ( Fig. 2A). DCX + cell 328 density was dramatically reduced in both male and female TK rats, to less than 15% of levels 329 found in WT littermates, consistent with recent work (48, 53, 57). This reduction was observed 330 in the dorsal and ventral hippocampus, and there were no sex differences in the extent of 331 neurogenesis reduction (Fig. 2B). 332 333 In cold water, ablation of neurogenesis impairs spatial learning in male rats and improves 334 spatial learning in female rats 335 Ablating neurogenesis typically does not impair learning a single spatial location in the water 336 maze (2, 6, 7, 17-22). Since adult-born neurons regulate unconditioned responses to stressors 337 (3, 23, 32), we hypothesized that stress or aversiveness may also reveal a role for new neurons 338 in spatial learning. We therefore tested WT and TK rats in the spatial water maze at standard 339 temperatures (25°C) or colder, more aversive temperatures (16°C). 340 In standard 25°C water, WT and TK rats learned to escape from the pool with similar 341 latencies ( Fig. 3A-C) and, in the probe trial, WT and TK rats displayed equivalent memory (Fig.  342 3D-F). We also observed sex differences in performance, where males escaped faster and 343 spent more time in the target zone than females. We explored whether estrous stages 344 influenced probe trial performance (but not after training trials, to avoid lavage impacts on 345 subsequent behavior). The distribution of WT and TK rats across the 4 stages of the estrous 346 cycle did not differ (χ 2 =1.3, P=0.7) but rats in proestrus displayed better memory than rats that 347 were not in proestrus, an effect that was comparable for both WT and TK rats (Fig. 3G). 348 At 16°C, blocking neurogenesis altered learning in both males and females, but in 349 opposite directions: male TK rats located the platform faster but female TK rats located it 350 faster, compared to their WT counterparts ( Fig. 4A-C). As in 25°C water, WT male rats located 351 the platform faster than WT females. These effects could not be explained by differences in 352 swim speed (Supplementary Fig. 1). A similar pattern was observed when we analyzed ideal 353 path error, a measure of the cumulative positional error relative to the platform that is not 354 influenced by differences in swim speed or path length (52): at 16°C female TK rats had a lower 355 path error and male TK rats had a greater path error, relative to WT controls (Supplementary 356 Fig. 2). 357 On the probe trial, TK rats spent less time searching in the target zone ( Fig. 4D-F). This 358 pattern was stronger in male TK rats but the ANOVA interaction (sex x time spent in target 359 zone) was not significant. Following the 16°C probe trial, the estrous distribution of female WT 360 and TK rats did not differ (χ 2 =2.7, P=0.4) and there was no effect of estrous stage on probe 361 trial performance (Fig. 4G).

362
To rule out the possibility that behavioral differences were due to nonspecific 363 physiological effects caused by cold water, we measured body temperature in a separate 364 group of rats. At both 16°C and 25°C, body temperature was lowest after day 1 training, was 365 lower on day 1 in females than in males, but not different between WT and TK rats 366 ( Supplementary Fig. 3 learning and memory performance at 16°C or 25°C, suggesting that water temperature did not 373 differentially impact sexes or genotypes due to hypothermic effects (Supplementary Tables 1 &  374 2). Finally, to rule out the possibility that TK impairments and enhancements in learning are due 375 to nonspecific effects of the GFAP-TK transgene, we trained additional WT and TK rats that did 376 not receive valganciclovir treatment. Here, no genotype differences were observed at 16°C or 377 25°C water temperatures ( Supplementary Fig. 4).

378
To gain insight into navigational strategies employed during learning, we analyzed 379 search patterns with Pathfinder software (52). Generally, rats displayed increasing use of 380 spatially-specific search strategies over days of testing (Fig. 5). Specifically, they shifted from 381 thigmotaxic and random searches, or searches that covered multiple areas of the pool equally, 382 to searches that were biased towards the escape platform with increasing precision. Male TK 383 rats relied less on spatially-specific search strategies than their WT counterparts. Consistent 384 with their faster escape latency, female TK rats tended to display more spatially-specific 385 searches than their WT counterparts but this difference was not statistically significant.

386
Consistent with the latency and path error data, search strategies did not differ between WT 387 and TK rats tested at 25°C (Supplementary Fig. 5).

388
Behavioral sex differences often reflect differences in strategy (58, 59). We therefore 389 explored whether maze aversiveness caused males and females to employ different 390 navigational strategies in the water maze. Female WT rats responded strongly to cold 391 temperature, and spent less time searching randomly and at the edge of the pool, and more 392 time performing spatial searches in the center of the pool and near the platform. Temperature-393 dependent changes in search strategy were absent in female TK that lacked neurogenesis 394 ( Supplementary Fig. 6). In contrast, male WT rats employed similar strategies at both 16°C and 395 25°C, but blocking neurogenesis led to temperature-dependent differences, where TK males 396 performed fewer spatially precise searches in 16°C water. Thus, neurogenesis promotes 397 aversiveness-related changes in search strategy in females but it promotes consistent search 398 strategies in males.

400
Blocking neurogenesis did not alter the HPA response. 401 Neurogenesis regulates the HPA axis in mice (23) and cold temperatures can enhance water 402 maze learning via glucocorticoid-dependent mechanisms (49). We therefore explored whether 403 neurogenesis regulates HPA axis function in rats at baseline and after learning. Consistent with 404 previous work in mice (23), we found no neurogenesis-related changes in baseline circadian 405 HPA function. Corticosterone levels were highest at the onset of darkness, they were higher in 406 females, but they did not differ between WT and TK rats ( Supplementary Fig. 7). When 407 corticosterone was measured 30 min after the first day of acquisition training, both WT and TK 408 rats displayed high levels of corticosterone, which did not differ between genotypes. 409 Corticosterone levels also did not differ between rats trained at 16°C vs 25°C. When 410 normalized to escape latency, i.e. time spent in the water, there was a tendency for greater 411 corticosterone levels at 16°C but this did not reach statistical significance. A subset of rats that 412 were subjected to the full 4 days of testing displayed HPA habituation, but no corticosterone 413 differences were observed between genotypes or temperatures. Thus, females elicit a stronger 414 HPA response than males, but neurogenesis-associated behavioral differences at 16°C are not 415 due to differences in HPA output.

417
Activity-induced Fos expression is modulated by sex and dorsoventral location, but not 418 neurogenesis. 419 Behaviorally-relevant DG neuronal populations express the activity-dependent immediate-early 420 gene, c-Fos (60-62). To determine whether blocking neurogenesis alters neuronal population 421 activity in males and females, we quantified Fos expression in excitatory principal cell 422 populations in DG-CA3, in both WT and TK rats (Fig. 6). Notably, Fos activation was never 423 different between WT and TK rats. However, more dentate granule neurons were active in 424 females than in males, particularly at 16°C (74% more at 16°C, 24% more at 25°C ventral hippocampus with performance on the acquisition and retrieval stages of testing, 438 however, no significant correlations were observed (data not shown).

439
Since adult-born neurons can influence DG-CA3 activity via efferent connections with 440 inhibitory interneurons (56, 64), we quantified Fos + inhibitory, GAD67-expressing neurons in 441 DG-CA3 (Fig. 6D,E). In rats trained at 16°C, there was a strong dorsoventral gradient of activity 442 in GAD67 + cells, with greater activity in the ventral DG than in the dorsal DG. There was also 443 significantly greater activation of GAD67 + cells in females than in males, but no differences due 444 to loss of adult neurogenesis. Sparse activation precluded a robust subregional analysis but, 445 when analyzed by DG-CA3 subregion, sex differences in GAD67 + cell activation were observed 446 in the molecular layer, granule cell layer and hilus+CA3 (female > male), but the dorsoventral 447 difference was specific to the granule cell layer (ventral > dorsal; Supplementary Fig. 8A-C). At 448 25°C, fewer GAD67 + cells were activated (mixed effects analysis; effect of temperature: 449 F1,91=8.2, P=0.005) and the dorsoventral gradient (V > D) was weaker. In contrast to rats trained 450 at 16°C, there were no sex differences in activation of GAD67 + cells in rats trained at 25°C. 451 Finally, at 25°C there also were no differences between genotypes. A similar pattern was 452 observed in the granule cell layer and hilus+CA3 region ( Supplementary Fig. 8D-F).

454
Training-and sex-dependent morphological plasticity in adult-born neurons. 455 Functionally-relevant morphological features of adult-born neurons develop during the weeks 456 and months post-mitosis (65-67), and can be modified by spatial learning (68, 69). To examine 457 sex differences in experience-dependent plasticity we labelled adult-born neurons with 458 retrovirus and analyzed GFP + spines and presynaptic terminals as morphological proxies for 459 afferent and efferent connectivity (Fig. 7). At baseline, in naïve home cage rats, there were no 460 differences in spine density between adult-born neurons from male and female rats. However, 461 in male rats, training at 16°C specifically elevated spine density compared to rats that were 462 untrained or trained at 25°C, and compared to female rats trained at 16°C. This effect was 463 observed throughout the molecular layer. In both males and females, regardless of treatment, 464 spine density increased with distance from the cell soma as previously observed (65)  465 ( Supplementary Fig. 10). The density of large, mushroom spines was not altered by training 466 (Fig. 7C). 467 Finally, we examined the large mossy fiber terminals that excite CA3 pyramidal neurons.

468
No sex differences were observed between naïve, home cage control rats. However, only in 469 males, training decreased mossy fiber terminal size, an effect that was greatest in the 25°C 470 group (Fig. 7D). In both females and males, 25°C training also reduced the number of 471 filopodial extensions that protrude off of mossy fiber boutons, putative synapses onto 472 inhibitory interneurons (Fig. 7E). 473 474 475 476

478
Sex modulates hippocampal memory, plasticity and physiology (70). And while there is also 479 evidence that sex regulates the addition and activation of new neurons (71), relatively few 480 studies have formally investigated sex and none have identified sex differences in the 481 behavioral consequences of manipulating adult neurogenesis (Fig. 1). Here we report that 482 blocking neurogenesis caused female rats to learn faster and male rats to learn slower, relative 483 to intact rats in a spatial water maze at aversive 16°C temperatures. These findings were not 484 confounded by genotype differences in swim speed, body weight or body temperature, and 485 they were not present in TK rats that were not treated with valganciclovir (and therefore had 486 intact neurogenesis). Whereas new neurons were morphologically equivalent at baseline, 487 learning evoked distinct patterns of pre-and post-synaptic plasticity depending on sex. Our 488 study therefore provides new evidence that adult-born neurons make unique sex-dependent 489 contributions to spatial learning under stress and have distinct plasticity profiles in male and 490 female rats.

492
Temperature-dependent spatial functions of newborn neurons.

493
While some have reported acquisition and short term reference memory deficits in the spatial 494 water maze in neurogenesis-deficient animals (14, 69, 72), a majority of studies have found 495 intact spatial learning (2, 6, 7, 15, 17-22), raising questions about the necessity of adult 496 neurogenesis for spatial learning. Our findings indicate that the degree of stress and/or 497 aversiveness present at the time of learning is critical (as suggested by (73)). Indeed, there is 498 ample evidence that neurogenesis regulates innate fear and anxiety-like behaviors in response 499 to stressful and aversive stimuli (23-25, 27-29). And while stress is known to potently modulate 500 hippocampal memory, few studies have examined a role for neurogenesis in learning as a 501 function of stress: one study found that neurogenesis is critical for context fear memory when 502 mice receive a single, but not multiple, footshocks (33); another found that TK rats made more 503 errors in a dry spatial maze only when an aversive odor was present (32). That we found no 504 learning differences at 25°C suggests that neurogenesis may be particularly important for 505 spatial learning under conditions of higher stress. Notably, we did not find differences in HPA 506 activation between rats trained at 16°C and 25°C. However, 16°C training did lead to greater 507 hippocampal recruitment (in females), greater dorsoventral differences in hippocampal 508 activation, differences in strategy usage, and it caused a greater reduction in body 509 temperature. Thus, 16°C water evoked physiological changes and behaviors that are consistent 510 with the concept of a stressor as a threatening stimulus that perturbs an organism from 511 baseline, necessitating an adaptive or homeostatic response (74). 512 513 Sex differences in the behavioral function of adult-born neurons. 514 We found that blocking neurogenesis led to opposite behavioral outcomes in females and 515 males which, to our knowledge, is the first report of sex differences in the behavioral function 516 of neurogenesis. To date, sex differences in function have gone undetected because few 517 studies have compared male and female animals that have altered neurogenesis ("functional" 518 studies). In our attempt to comprehensively survey the literature (Fig. 1), we counted only 4 519 functional studies that have reported data by sex or included sex as a variable in their statistical 520 analyses (9, 57, 75, 76). 521 It is typically understood that neurogenesis benefits cognition and so it may seem 522 paradoxical that blocking neurogenesis improved water maze learning in females. However, it 523 has been repeatedly demonstrated that males and females can display opposite patterns of 524 hippocampal-dependent learning, with manipulations facilitating performance in males in some 525 paradigms and facilitating performance in females in others (42-44). Our findings also may 526 seem paradoxical if it is assumed that "faster is better" in the water maze. It is increasingly well-527 documented that sex differences in learning tasks can reflect strategy differences rather than 528 frank differences in learning ability (58, 77) and escape latencies cannot reveal differences in 529 strategy and navigational choice that may be highly adaptive (15). Here we found that male 530 rats that lacked neurogenesis performed more general searches, but female neurogenesis-531 deficient rats tended to (nonsignificantly) perform more spatially specific searches. While it is 532 common to view spatially-specific searches as "better", generalized search has clear 533 advantages in cases where a spatial goal moves to a new or unexpected location (78, 79). Thus, 534 one possibility is that neurogenesis adjusts search/memory specificity differently, increasing it 535 in males and perhaps decreasing it in females. That females trained at 16°C had significantly 536 higher levels of Fos in the dorsal DG indicates that there are clear sex differences in regional 537 hippocampal recruitment, which could impact the adoption of precise search strategies (63). 538 Another possibility, related to the fact that neurogenesis effects were selectively 539 observed in 16°C water, is that emotional functions of neurogenesis were differentially 540 engaged by stress. In other studies, stress impairs spatial learning in males and is either without 541 effect, or actually improves learning, in females (42, 43). These divergent effects may reflect 542 differential effects of stress on cognition (males) and hyperarousal (females) (80). Since 543 neurogenesis ablation mimics some features of the stressed brain (e.g. structural atrophy) (81, 544 82), possibly male learning was impaired by dysregulated integration of stress and learning, 545 and females learned faster due to heightened arousal and attention effects. A role for 546 attentional processes is also suggested by recent work showing that blocking neurogenesis 547 reduces orienting responses to distractor stimuli (83), an effect that may explain why TK rats are 548 faster to navigate a dry spatial maze in the presence of an aversive, but irrelevant, mint odor 549 (32). Given sex differences in processing object arrays and configurations (70), blocking 550 neurogenesis may differentially alter water maze cue processing such that females are less 551 susceptible to distraction from irrelevant cues (leading to faster escape) but males are less 552 attentive to relevant cues (leading to slower escape). 553 Finally, insights into the potential adaptive significance of neurogenesis also come from 554 our analyses across temperatures ( Supplementary Fig. 6). Intact females were highly sensitive 555 to temperature: 16°C shifted females away from random and wall-focused search, toward the 556 center of the pool and the specific area of the platform. In contrast, TK females were not 557 different at 16°C and 25°C. Thus, in females, neurogenesis promotes changes in strategy 558 according to the aversiveness of the situation. In males, neurogenesis promoted equivalent 559 strategy usage 16°C and 25°C, which could also be adaptive in cases where performance 560 needs to remain stable despite perturbations from external forces. 561 562 Sex differences in hippocampal subregional activation. 563 To investigate possible subregional and cellular mechanisms we examined activity-dependent 564 Fos expression along the dorsoventral axis in male and female rats that did, or did not, have 565 adult neurogenesis. While previous studies have reported that ablating neurogenesis can 566 increase (13, 27, 64) or decrease (53, 84) activity in the hippocampus, here we found no effect 567 on global activity amongst dentate granule cells. Newborn neurons also target inhibitory 568 interneurons (56, 64), whose activity regulates the precision of hippocampal-dependent 569 memory (85, 86). However, we also observed no changes in inhibitory recruitment in TK rats 570 relative to WT rats. While these findings suggest that neurogenesis ablation did not affect 571 behavior by altering hippocampal activity, it is possible that activity differences were present 572 early in training, when behavioral sex differences were more prominent.

573
Little is known about how dorsoventral subregions of the hippocampus are activated in 574 males and females by training in the standard spatial water maze. Here, we found that females 575 consistently had greater levels of DG activity than males, particularly at 16°C. This was largely 576 driven by elevated Fos levels in the dorsal hippocampus, a finding that builds on previous 577 evidence that the spatial water maze recruits dorsal more than ventral DG (60). However, 578 whereas that study only included males, here we find that the dorsoventral gradient is 579 significantly stronger in females. Elevated Fos in dorsal vs ventral DG was mirrored by an 580 opposite gradient of Fos in GAD67 + inhibitory cells, suggesting that regional activity is 581 controlled by local inhibitory circuits. Since the temporal progression of water maze learning 582 strategies involves sequential recruitment of ventral to dorsal hippocampus (63) we explored 583 relationships between water maze performance (latency, path error, strategy specificity on 584 acquisition and probe trials) and activity in the dorsal and ventral DG. However, we found no 585 consistent correlations, suggesting that other forms of activity and plasticity may be more 586 tightly linked to performance. 587 588 Sex differences in morphological plasticity of adult-born neurons. 589 To our knowledge, this is the first study to examine functionally-relevant morphological 590 features of adult-born neurons in males and females. At baseline, we observed no differences 591 in spine density or mossy fiber terminal size between the sexes. However, water maze training 592 induced plasticity of excitatory synaptic structures but only in males.   organized by degree of spatial specificity relative to the target. B) Strategies employed by 688 female WT and TK rats. The distribution of strategies in female TK rats was not significantly 689 different from female WT rats (χ 2 = 7, P = 0.5). Right-most graph shows weighted strategy 690 changes in TK rats relative to WT rats, quantified as % changes in strategy usage multiplied by 691 the total fraction of trials where rats employed that strategy (genotypes pooled). The 692 magnitude of the bars therefore reflects changes in strategy but prevents misleading 693 perceptual artefacts caused by large % changes for strategies that were rarely used. (C) 694 Reducing neurogenesis significantly altered the distribution of strategies used by male rats, 695 demonstrated by the greater proportion of spatially non-specific trials and the smaller 696 proportion of spatially-specific trials (right; χ 2 = 17, P = 0.02).