Long-Term Aberrations To Cerebellar Endocannabinoids Induced By Early-Life Stress

Emerging evidence points to the role of the endocannabinoid system in long-term stress-induced neural remodeling with studies on stress-induced endocannabinoid dysregulation focusing on cerebral changes that are temporally proximal to stressors. Little is known about temporally distal and sex-specific effects, especially in cerebellum, which is vulnerable to early developmental stress and is dense with cannabinoid receptors. Following limited bedding at postnatal days 2–9, adult (postnatal day 70) cerebellar and hippocampal endocannabinoids, related lipids, and mRNA were assessed, and behavioral performance evaluated. Regional and sex-specific effects were present at baseline and following early-life stress. Limited bedding impaired peripherally-measured basal corticosterone in adult males only. In the CNS, early-life stress (1) decreased 2-arachidonoyl glycerol and arachidonic acid in the cerebellar interpositus nucleus in males only; (2) decreased 2-arachidonoyl glycerol in females only in cerebellar Crus I; and (3) increased dorsal hippocampus prostaglandins in males only. Cerebellar interpositus transcriptomics revealed substantial sex effects, with minimal stress effects. Stress did impair novel object recognition in both sexes and social preference in females. Accordingly, the cerebellar endocannabinoid system exhibits robust sex-specific differences, malleable through early-life stress, suggesting the role of endocannabinoids and stress to sexual differentiation of the brain and cerebellar-related dysfunctions.

sex differences. Stressed animals were also expected to exhibit impairments in key behavioral processes including recognition memory and social preference, based on previous studies 57 . Moreover, these behavioral differences were not expected to be due to an anxiety phenotype (assessed via elevated plus maze) or gross cerebellar dysfunction (assessed via rotarod). Overall, these predictions form a direct test of the model that early-life stress induces vulnerability of a portion of the cerebellar-endocannabinoid network, with implications for long-term, aberrant behavioral outcomes.

Methods
All experimental protocols involving rats were approved by the Indiana University Bloomington IACUC. All methods were carried out in accordance with relevant guidelines and regulations for care and use of laboratory animals.

Animals.
Timed-pregnant Long-Evans rats (Envigo, Indianapolis, IN) arrived at gestational day 14. Dams were individually housed in polypropylene cages (26.67 cm × 48.26 cm × 20.32 cm) in a 12:12-hour light-dark cycle (6:00 lights on, 18:00 lights off) and temperature-controlled (22.8 °C) vivarium. Food and water were provided ad libitum. Bedding was changed once per week. Rats were checked every 12 hours surrounding the expected date of birth. Day of birth was designated P0. At P2, animals were randomly cross-fostered, and cages were sex-balanced.
Limited Bedding Stressor. At P2, all cages were changed. Half of the cages were randomly selected to undergo the limited bedding manipulation. Limited bedding cages contained a wire mesh insert (Plastic-coated aluminum mesh, 0.4 cm × 0.9 cm, McNichols Co., Tampa, FL) that was fitted 2.5 cm above the cage floor (cf. 57 ). The mesh allowed the passage of excrement to the bedding material below the mesh. Additionally, limited bedding cages were given half of a paper towel square (13.97 cm × 27.94 cm) for the dam to use as nesting material. Normal rearing cages were given a full paper towel square and standard access to bedding material. All cages were left undisturbed from P2-9. On P10, all animals were transferred to clean, standard caging. All animals were reared normally from this point forward. At P21 pups were weaned into treatment-and sex-matched cages of 3-4 animals.

Plasma Corticosterone Quantification.
To assay acute and long-term changes in basal stress, plasma corticosterone levels were quantified using trunk blood. Animals were sacrificed at P8 (during the stressor) using rapid decapitation only and P70 (adulthood) using isoflurane then rapid decapitation immediately after anesthetization (1-2 minutes) as determined with a tail, foot, and ocular check. Procedures occurred between the hours of 9:00 (3 hours after the start of the light cycle) and 11:00. This is at the lower end of the circadian CORT cycle in both males and proestrus females 58 . After decapitation, trunk blood was collected in 1.5 mL Eppendorf tubes with 1 μL of liquid heparin (1 000 USP/mL) to prevent clotting and immediately frozen at −80 °C. Samples were shipped to Cayman Chemicals for quantification via ELISA (sensitivity 30 pg/mL; intra-assay variation 2-18%).
Endocannabinoid Quantification. Upon sacrifice at P70, neural tissue was harvested, frozen, and sectioned using a 1 mm coronal sectioning block (BrainTree Scientific). Two-millimeter round punches were taken across three adjacent sections of the interpositus (IP) nucleus and Crus I of the cerebellum and dorsal hippocampus. Punches were flash frozen using liquid nitrogen and stored at −80 °C. To extract and partially purify lipids from these tissues, samples were processed by the Bradshaw Lab of Lipid Neuroscience at Indiana University Bloomington as previously described (cf. 59 ).
26 lipids from the broader endocannabinoid lipidome, including AEA, its lipoamine structural analogs (e.g., N-palmitoyl ethanolamine), 2-AG, its 2-acyl glycerol structural analogs (e.g. 2-linoleoyl glycerol), associated free fatty acids, and prostaglandins were quantified in tissue extracts from a single hemisphere using high-performance liquid chromatography coupled with tandem mass spectrometry (HPLC/MS/MS). Lipid concentrations were normalized to sample mass prior to statistical analysis. Group comparisons were performed via one-way ANOVAs, correcting for multiple comparisons. mRnA Analysis. Two-millimeter round punches from the corresponding hemisphere used for lipid analysis were taken across three adjacent sections of the IP nucleus. Tissue samples were immediately homogenized in Trizol in preparation for mRNA isolation. Purification (QIAGEN RNeasy Plus Mini Kit) yielded 20 μL of RNA, which was quantified using Nanodrop and assessed for integrity via an Agilent 2100 Bioanalyzer. Three samples were eliminated due to low RNA integrity (RIN < 5). Samples were processed, run, and analyzed by the Indiana University School of Medicine Genomics Core Facility as described below.
A library was prepared using KAPA mRNA Hyperprep Kit (KK8581). Sequencing was performed using Illumina HiSeq 4000 sequencing and deemed good quality (95% Q30, 76% Cluster Density, 367 M reads/lane). Alignment and mapping also yielded acceptable quality (85.5% uniquely mapped reads, average 49.4% mapped onto gene). Differential expression analyses were performed using treatment and sex as primary factors and corrected for multiple comparisons.
Behavioral Assessment. Pups were assessed on four behavioral tasks in young adulthood. At P46, all rats underwent 3 consecutive days of handling and familiarization to the testing room. Behavioral testing continued from P50-70. Thirty rats underwent Novel Object Recognition (NOR), then Rotarod. 44 rats underwent NOR, followed by Social Preference, then Elevated Plus Maze (see Fig. 1C-F for group sizes). All testing occurred between 9:00-13:00.
Novel Object Recognition (NOR). Following handling, rats were familiarized to the NOR apparatus (15 minutes daily for 3 days) which included a clear Plexiglas box (50 × 50 cm base, 40 cm height; open top). During the testing day, rats were first given 10 minutes under red light to explore two identical objects located in adjacent corners of the testing apparatus. Rats were then returned to their home cage in the housing colony for a one-hour delay, after which they were placed back into the apparatus, under red light, for the test trial. One object from previously ("familiar") and one novel object were located in adjacent corners of the box. Rats were given 5 minutes to explore. Blinded raters scored the first 3 minutes of the test trial for time spent interacting with the familiar and novel objects. A discrimination ratio was calculated (time spent with novel object/time spent with novel and familiar object) to quantify recognition memory.
Social Preference. Rats were exposed to a novel sex-and weight-matched rat restrained within a wire mesh box (11 × 11 × 19 cm) and novel object covered by a wire mesh box in adjacent corners of the same apparatus as used for NOR (clear Plexiglas box; red light) and in the same testing room to which they were previously familiarized. Blinded raters scored 3 minutes of interaction for time spent engaging with the rat (i.e., social) or object (i.e., non-social). A social preference ratio was calculated (time spent with novel rat/time spent with novel object and rat).
Rotarod. Following 3 days of familiarization to the apparatus (IITC Life Sciences, Roto-Rod Series 8), rats were placed on the center rod and allowed to move freely as the velocity increased from a starting point of 4 RPMs by 7.2 RPMs. Testing occurred under low light (100 lux at equipment surface). Latency to fall was measured for each rat by a sensor within the apparatus. Statistical Analyses. Using SPSS (25.0, IBM Corporation), 2-Way ANOVAs were used to assess main and interaction effects of treatment and sex on corticosterone, lipid, and behavioral outcomes. ANOVAs were followed by exploratory post-hoc t-tests for trending findings (i.e., p > 0.050 to p < 0.075) to clarify the driving factors of such effects.

Plasma Corticosterone Quantification and
Endocannabinoid levels in Crus I of the cerebellum showed a baseline sex difference (Fig. 2, Supplementary Table 1): control females exhibited significant decreases in N-stearoyl ethanolamine, N-oleoyl ethanolamine, and N-linoleoyl ethanolamine (LEA) accompanied by increases in N-palmitoyl glycine (P-Gly), N-docosahexaenoyl glycine (D-Gly), 2-AG, and prostaglandin F2 α (PGF 2α ) compared to control males. Endocannabinoid levels in Crus I were significantly impacted by stress in females only, with stressed females exhibiting decreased D-Gly and 2-AG compared to normally-reared females. Accordingly, stressed females had significantly increased P-Gly and significantly decreased 2-linoleoyl glycerol (2-LG) and linoleic acid.
Striking sex and treatment differences were observed in cerebellar IP endocannabinoid levels ( Fig. 3, Supplementary Table 2). At baseline, females exhibited decreases in LEA, N-arachidonoyl taurine (A-Taur), linoleic acid, arachidonic acid (AA), and prostaglandin E 2 (PGE 2 ) compared to control males. Contrary to Crus I, treatment effects on endocannabinoids were only present in males. Stress significantly decreased 2-AG and AA in males compared to control males. In females, stress significantly increased PGE 2 . Consistent with these differences, stressed females had significantly decreased LEA and A-Taur and increased N-oleoyl glycine, 2-AG, and PGE 2 .
In dorsal hippocampus (Fig. 4, Supplementary Table 3), control females showed significantly decreased 2-oleoyl glycerol, 2-LG, and 2-AG and increased PGF 2α levels compared to control males. Stressed females had significantly elevated 2-palmitoyl glycerol and 2-LG compared to stressed males. Treatment effects were only present in males, with stressed males exhibiting significantly increased PGE 2 and PGF 2α compared to normally reared males. (  www.nature.com/scientificreports www.nature.com/scientificreports/ Of the 11 865 genome-wide genes evaluated, 3 009 (25.4%) had significantly different (FDR-corrected p < 0.05) mRNA expression between all males and females; specifically, 2 057 (17.3%) significantly differed between normally reared males and females and 260 (2.2%) differed between limited bedding males and limited bedding females, with 169 (1.4%) overlapping (see Fig. 5a). Minimal treatment differences were observed, with 2 (0.02%) genes differing between all limited bedding and normal rearing animals; specifically, 5 (0.04%) significantly differed between normally reared and limited bedding males and 4 (0.03%) differed between normally reared and limited bedding females, with no overlap. All genes significantly differing between normally reared and limited bedding males overlapped with genes significantly differing between normally reared males and females. Two genes significantly differing between normally reared and limited bedding females overlapped with genes significantly differing between limited bedding males and females.
Behavioral Assessment. Three blinded raters scored all behaviors, except rotarod performance, which was scored digitally via the apparatus. Mean scores from all three raters were computed and used in behavioral analyses. Inter-rater reliability was deemed acceptable, with an interclass correlation of >0.95 for each behavior scored.
Novel Object Recognition (Fig. 1C). A significant main effect of treatment group on discrimination ratio was observed, with limited bedding rats exhibiting decreased recognition memory (F(1,68) = 5.15, p = 0.026, ηp² = 0.070), though performance still indicated preference for the novel object. There was no main effect of sex www.nature.com/scientificreports www.nature.com/scientificreports/ and no sex-treatment interaction effect. There were no significant differences by sex or treatment group in exploration time (total time interacting with the objects).
Social Preference (Fig. 1D). A trending main effect of treatment was observed (F(1,40) = 3.04, p = 0.089, ηp² = 0.071), likely due to significantly decreased social preference in females that underwent early-life stress compared to normally reared females (t(20) = 2.52, p = 0.020, d = 1.074). There was no main effect of sex or sex-treatment interaction effect. There were no significant differences by sex or treatment group in exploration time (total time interacting with the object and/or social partner).
Rotarod (Fig. 1E). A significant main effect in latency to fall was observed between males and females (F(1,26) = 5.77, p = 0.024, ηp² = 0.182); specifically, males had a shorter latency to fall, likely attributed to weight differences making it more difficult for males to maintain balance on the dowels. No significant main effect was observed across treatment groups, as expected, and no sex by treatment interaction was present.
Elevated Plus Maze (Fig. 1F). No main effects of sex or treatment group and no interaction effects were observed in time spent in the center, open, or closed arms. No differences were seen in open or closed arm entries. Low overall exploration values may reflect testing occurring within the animal's light phase, which is when they are least active, but was consistent with other testing procedures in the current study 61 .

Discussion
The current study is the first to our knowledge to investigate the impact of early-life stress on long-term cerebellar endocannabinoid tone in male and female rats. Notable are three major findings: (1) clear baseline sex-specific differences were observed in endocannabinoids and related lipids in a region-specific manner as well as in mRNA expression; (2) developmental stress caused sustained sex-specific changes to CNS endocannabinoid lipid tone www.nature.com/scientificreports www.nature.com/scientificreports/ with minimal long-term effects on mRNA; and (3) such stress impaired performance on novel object recognition and social preference.
First, baseline sex and region differences in endocannabinoid tone and mRNA expression were found, consistent with previous studies 30,62 . Though some studies have found sex differences in cortical mRNA expression 29 or cerebellar regional protein differences in males only 22 , this study, for the first time to our knowledge, identified extensive baseline sex differences in cerebellum. 2-AG was decreased in males compared to females in cerebellar IP (Fig. 4) but increased in cerebellar cortex (Crus I; Fig. 3), whereas LEA was decreased in both regions in males. Within the IP specifically, mRNA expression was significantly different between males and females at baseline (no treatment groups) across a variety of gene systems, emphasizing the robust and broad effects of sex on neural systems.
Such baseline sex differences may cause exogenous cannabinoids to differentially affect the brain in males and females. This serves as a potential explanation for the key differences in outcomes of exogenous cannabinoid exposure in animal 63,64 and human 65-67 models and perhaps also reflects sex differences in use rates 25,66 and adverse side effects 68,69 in humans. Similarly, understanding regional differences in cannabinoids may help predict directional impacts on behaviors emerging from region-specific neural circuits. Thus, baseline differences in endocannabinoid tone, as well as previous work on receptor differences 25 , may together help make sense of differences in function, at the circuit and behavioral level.
Second, 2-AG was decreased in the cerebellar IP nucleus, but only in limited bedding males compared to normally reared males. A stress-induced rise in 2-AG during a critical developmental period may explain such a finding, resulting in a long-lasting downregulation of the endocannabinoid system to accommodate this new neuronal environment. In addition, a decrease in AA and a trending decrease in AEA was noted, suggesting that an upstream mechanism may be implicated in these effects. For example, down-regulation could be the function www.nature.com/scientificreports www.nature.com/scientificreports/ of a substrate imbalance, such as a decrease in phospholipids resulting in decreased overall endocannabinoid production. Although no mRNA changes for these enzymes were found, expressed protein or enzyme activity were not evaluated and, thus, this hypothesis cannot be ruled out.
No endocannabinoid changes were observed in stressed compared to non-stressed females in the cerebellar IP nucleus, though stressed females did exhibit a significant increase in prostaglandin E 2 (PGE 2 ). PGE 2 has been shown to increase cerebellar Purkinje cell arborization via estradiol production [70][71][72] . Estradiol-and PGE 2 -induced cerebellar Purkinje cell arborization may function as a neuroprotective mechanism in females, perhaps further explaining the limited behavioral impairments observed in females during stress paradigms [73][74][75] . Further, the increase in PGE 2 suggests activation of cyclooxygenase (COX) enzymes, which may be related to an increase in pro-inflammatory cytokines, identified as a neural response to stress to activate microglia 72,76 . Similarly, the increase in N-acyl glycines in stressed females suggests a pro-inflammatory response, as N-acyl glycines have been implicated in microglial migration [76][77][78] . In fact, microglia have been identified as having a critical role in the sexual differentiation of the brain 79 .
Early-life stress induced the opposite sex-specific profile in the cerebellar cortex (Fig. 3) compared to the cerebellar IP nucleus (Fig. 4): 2-AG was downregulated in stressed females compared to control females in Crus I of cerebellar cortex, whereas in cerebellar IP, 2-AG was downregulated in stressed males compared to control males. Cerebellar behaviors, such as associative learning, require carefully coordinated communication between the cerebellum's nuclei and cortex. Sex-specific differences in circuit coupling resulting from endocannabinoid regulation of neuronal activation may contribute to hypotheses of baseline and stress-effects on behavioral performance differences in cerebellar tasks, like eyeblink conditioning 47,80,81 . Understanding regional specificity within the cerebellum is critical, as extensive work has shown that distinct cerebellar cortical regions exhibit differential functional connectivity with cerebrum in humans and animals 52,82 . Though this study did not investigate www.nature.com/scientificreports www.nature.com/scientificreports/ differences across cortical regions, the question becomes increasingly important as we begin to understand more about how cerebellar signaling projects outwards through the deep nuclei to regulate cerebral signaling (cf. 51 ) and consider the importance of transmitter systems, such as endocannabinoids, in modulating these processes.
Many studies looking at stress-related endocannabinoid changes have investigated the early effects of stress, finding increased 2-AG and decreased AEA 20,83,84 . Thus, the current findings do not preclude the possibility that an acute change in endocannabinoid tone occurred immediately after the stressor; rather, the current study may suggest that any changes taking place close to the occurrence of the stressor were not enduring in these regions. Such changes may not have been maintained due to the early development of the rat hippocampus and lack of continual stressors throughout the study 85 .
Like 2-AG, PGE 2 showed a region-by-sex effect: upregulated in female cerebellar interpositus nucleus following stress but upregulated in stressed males in the hippocampus. As detailed above, such findings may relate to a potential pro-inflammatory response in male hippocampus following stress. Alternatively, these findings reinforce theories of foundational effects of stress on development and sexual differentiation of the brain 86 particularly for males, further reflected by baseline regional and sex differences in the endocannabinoid system and transcriptomics generally.
By adulthood (P70), male rats having undergone stress exhibited an increase in basal systemic CORT, suggesting the stressor carries long-term effects. These findings are in line with a growing literature that males are more sensitive to stress throughout development due to sex-differences in the uterine environment 86 . In females, increased CORT was not observed, perhaps due to a resilience or compensatory mechanism contributing to female pups overcoming insults to the broader HPA system, but this resilience is lacking in the still-developing endocannabinoid system. In humans, changes in cortisol have been linked to differences in cannabis use 87 ; the increased plasticity of the HPA system may explain sex-differences in exogenous cannabinoid use and effects.
Taken together, within a given neural region, stressed males exhibited a similar lipid profile to non-stressed females. Such relationships have been previously shown in the literature and lend support to changes in estrogen receptors, perhaps explaining this feminization of neural circuits susceptible to stress in the male brain 88 .
A third major finding is that early-life stress was associated with functional outcomes, particularly NOR performance, irrespective of sex. As shown in other studies of limited bedding 9 or maternal separation 89,90 and in the current study, rats that underwent limited bedding stress exhibited recognition memory deficits compared to controls. This was not the case for social preference, which was only impaired in stressed females in the current study. Other studies of early-life stress have shown that males are affected in social behavior later in life [91][92][93][94][95] . One reason for this discrepancy may be that Crus I stress-effects on lipids were specific to females. Interestingly, a recent study by Stoodley and colleagues 54 found that right cerebellar Crus I disruption was related to social impairments in mice. Thus, cerebellar signaling disruption in females may have contributed to this behavioral phenotype.
Limitations and future Directions. Some key limitations should be addressed. First, the current findings are specific to bioactive lipid concentrations and have not yet been contextualized regarding changes in endocannabinoid-related protein expression or activity (i.e., receptor or enzyme, though major differences at the transcriptional level were not observed), though the transcriptomic data presented here will provide key insights into these pathways. Thus, the mechanistic pathway by which these endocannabinoid and prostaglandin changes have occurred is unclear and is a promising path for future inquiry. Such a question could be addressed by further analysis of the RNA data presented in this study and is highly recommended.
Second, although no changes in gross cerebellar integrity were observed via the rotarod task, the impacts these cerebellar changes may have on more refined cerebellar-mediated behaviors, such as eyeblink conditioning, remain unclear. The well-defined neural circuit involved in eyeblink conditioning has been shown to be impaired by other early-life stress paradigms 47,49 , involves endocannabinoids at many steps 96 , and is highly translatable to humans 97 , making it an optimal task to interrogate cerebellar-and cannabinoid-specific impacts of stress 24,98 . In addition to eyeblink conditioning, the cerebellum is gaining traction as an important node in cognitive processes. Tasks of spatial navigation, working memory, and cognitive flexibility are most notable (e.g., spontaneous/learned alternation, Morris water maze, set-shifting, discrimination learning, among others) 52 . Specifically, the Crus I area investigated in the current study has been implicated in evidence-accumulation decision making 53 and autism spectrum-related behaviors such as social impairment 54 in mouse models.
Third, neither estrous phase nor sex hormones were measured in this study. Excluding such measures may obscure the specific neurobiological mechanisms by which endocannabinoid and behavioral sex-differences arose. Though it has been shown that cerebellar endocannabinoids in particular do not change across the estrous cycle 30 , there is evidence in CNS that the endocannabinoid system is regulated by estradiol 62 . Namely, estradiol has a modulatory role on AEA degradation through FAAH downregulation 99 , neither of which were significantly aberrant in the current study within our limits of detection. The interaction of sex hormone and endocannabinoid pathways suggests that estrous phase or individual differences in sex hormones, such as estrogens, may impact endocannabinoid tone or vice versa. Similarly, CORT was measured at a single timepoint, the lower end of the circadian phase, which may have contributed to variability in measurement due to individual differences and exact time of day during the sampling window. Relatedly, it should be noted that systemic CORT is not a direct reflection of cerebellar CORT levels.
Finally, it cannot be assumed that these lipid changes are solely due to the stressor; in fact, it may be that the stressor caused neuronal remodeling that altered rodent behavior, leading subsequent environmental, physiological, and social factors to impact adult lipid signaling. Thus, continued work on proximal lipid effects of stress are key. www.nature.com/scientificreports www.nature.com/scientificreports/

conclusions
The current study provides a necessary springboard to understand sex-specific, developmental vulnerability mechanisms conferred by stress, specifically those that provide long-lasting neural changes. With increased interest in the cerebellum's role as a node of cognitive processing 10,100 as well as its role in psychopathology, such as autism 101,102 and schizophrenia 103 , this line of inquiry is timely and impactful. Understanding these sustained vulnerabilities may help reveal mechanisms of risk, including two-hit models of the effects of early stress and adolescent or adult cannabis use 21,64 , and may also contribute to better-informed animal models of exogenous cannabinoid effects.