Microbial modulation prevents the effects of pervasive environmental stressors on microglia and social behavior, but not the dopamine system

Environmental toxicant exposure, including air pollution, is increasing worldwide. However, toxicant exposures are not equitably distributed. Rather, low-income and minority communities bear the greatest burden, along with higher levels of psychosocial stress. Both air pollution and maternal stress during pregnancy have been linked to neurodevelopmental disorders such as autism, but biological mechanisms and targets for therapeutic intervention remain poorly understood. We demonstrate that combined prenatal exposure to air pollution (diesel exhaust particles, DEP) and maternal stress (MS) in mice induces social behavior deficits only in male offspring, in line with the male bias in autism. These behavioral deficits are accompanied by changes in microglial morphology and gene expression as well as decreased dopamine receptor expression and dopaminergic fiber input in the nucleus accumbens (NAc). Importantly, the gut-brain axis has been implicated in ASD, and both microglia and the dopamine system are sensitive to the composition of the gut microbiome. In line with this, we find that the composition of the gut microbiome and the structure of the intestinal epithelium are significantly shifted in DEP/MS-exposed males. Excitingly, both the DEP/MS-induced social deficits and microglial alterations in males are prevented by shifting the gut microbiome at birth via a cross-fostering procedure. However, while social deficits in DEP/MS males can be reversed by chemogenetic activation of dopamine neurons in the ventral tegmental area, modulation of the gut microbiome does not impact dopamine endpoints. These findings demonstrate male-specific changes in the gut-brain axis following DEP/MS and suggest that the gut microbiome is an important modulator of both social behavior and microglia.


Introduction
Air pollution represents a global and ever-increasing burden to human health 1,2 . Importantly, there are pervasive socioeconomic disparities in air pollution exposure. Low-income and racial minority communities are exposed to higher levels of air pollution than high socioeconomic status, primarily white, communities 3 . These communities are also subjected to fewer resources and higher levels of psychosocial stress 4 . This convergence of toxicant and psychosocial burdens makes these populations more susceptible to the adverse health consequences of air pollution exposure 5 .
High levels of air pollution, particularly during development, have been linked to several diseases, including autism spectrum disorder (Odds ratios: 1.98-3.48, Hazard ratios: ~1.14) 6,7,8 . Like air pollution, gestational maternal stress is associated with an increased risk of having a child with ASD 9,10 . ASD is a neurodevelopmental disorder primarily characterized by de cits in social behavior/communication, as well as repetitive behaviors. ASD is also characterized by a strong male bias in prevalence (4:1) and males tend to be more sensitive than females to early life stressors 11,12 . Together, these ndings suggest that air pollution and maternal stress during pregnancy may synergize to increase ASD risk, particularly in male offspring. Many brain regions have been implicated as disrupted in function and/or connectivity in ASD, including the mesolimbic reward circuit 13,14,15 . Activity in this circuit is critical to reward/motivation and social interactions. Optogenetic stimulation of nucleus accumbens (NAc)-projecting dopamine neurons in the ventral tegmental area (VTA) increases social behavior in mice 16 and dopaminergic signaling in the NAc mediates social play behavior in rodents 17,18 . Microglia, the resident immune cells of the brain, respond potently to air pollution 19 and play a critical role in the developmental organization of the dopamine system 17,20 . However, little is known regarding the impact of prenatal environmental exposures on microglia or dopaminergic function within this pathway.
The gut-brain axis has emerged as key to the pathophysiology and potential treatment of ASD 21 .
Changes in the composition of the gut microbiome and gastrointestinal symptoms have been documented in individuals with ASD and long-lasting improvements in both gastrointestinal and behavioral symptoms are found following microbiota transfer therapy 22,23 . Changes in gut microbiota in ASD may be driven by the highly restrictive diets of individuals with ASD 24 , but regardless of the source of these changes, the gut microbiome represents an important site for therapeutic intervention in the amelioration of behavior impairments in this disorder.
Importantly, several studies suggest that microglial function and activity of the dopamine system are impacted by the composition of the gut microbiome. For instance, germ-free mice (lacking microbiota) have immature, hyper-rami ed microglia 25,26 . In mouse models of ASD, supplementation with the bacterial species Lactobacillus reuteri induced neuroplasticity in dopaminergic neurons and restored sociability 27,28 . While studies have shown that air pollution shifts the composition of the gut microbiome in adulthood 29,30 , nothing is known about how maternal exposure to air pollution during gestation might impact the gut microbiome of offspring, and how these shifts might impact the development of circuits underlying social behavior.
To examine these questions, our lab established a paradigm in which mouse dams are exposed to combined diesel exhaust particles (DEP) and maternal stress (MS) during pregnancy 31,32 . We report that maternal DEP/MS alters microglial morphology and gene expression and decreases dopaminergic tone in the NAc only in male offspring. We show for the rst time that prenatal DEP/MS alters the composition of the gut microbiome in male offspring only, and that shifting the colonization of the gut microbiome at birth via a cross-fostering procedure prevents both sociability de cits and changes in microglial morphology in DEP/MS-exposed males. Interestingly, while chemogenetic activation of the dopamine system is su cient to rescue social behavior de cits following DEP/MS in males, microbial modulation at birth does not appear to impact the dopamine system, suggesting that prevention of microglial phenotypes is more closely associated with the prevention of social behavior de cits following DEP/MS.

Animals
Wild-type (WT) C57Bl/6J mice were purchased from Jackson Laboratories (Stock # 000664). DAT-IRES-Cre mice (Jackson Laboratories, Stock # 006660) were obtained from Dr. Henry Yin at Duke University. Animals were group-housed under standard laboratory conditions (12- 19,31,32 ), or vehicle (CON) via oropharyngeal instillation during which females are suspended by their incisors from a plastic wire and the instillation is administered over the course of 30 sec. Females were monitored until they were awake before returning to their home cages.

Maternal Stress (MS)
To induce prenatal maternal stress, we adapted a nest restriction paradigm 19,33 . On E13, females were singly housed and placed into a cage containing a thin layer of AlphaDri bedding (AlphaDri; Shepherd Specialty Papers) covered by an elevated aluminum mesh platform (0.4 cm × 0.9 cm mesh; McNichols Co., Tampa, FL) and given 2/3 (~ 1.9 grams) of a cotton nestlet (MS condition) or placed into a cage with AlphaDri bedding and an entire nestlet (CON condition). On E18.5 (evening before delivery), dams were transferred to a clean cage of AlphaDri bedding with a full nestlet. Offspring were weaned into cages with same sex littermates at postnatal day (P)24, as recommended by our IACUC policy.

Behavioral Testing
Behavioral testing took place between postnatal days P27-P40, during the second half of the light phase (afternoon). For all assays, males and females were tested using separate apparatuses on different days. Animals were handled and habituated to the testing room and testing apparatus prior to testing.

Sociability and Social Novelty Preference
To assess sociability (preference to investigate a social vs a non-social stimulus) and social novelty preference (novel social stimulus vs. cage mate) a 3-chambered social preference test was used 34 . This test consists of a 3-chambered arena with openings allowing passage between the chambers. Stimuli were con ned within smaller containers (Plexiglass rod sides) in each of the side chambers. Subject animals were placed into the middle chamber and freely allowed to investigate each stimulus over the course of 5 min (sociability) or 10 min (social novelty preference; see Supplementary Methods for full details).

Marble Burying
To assess repetitive behavior, mice were placed into clean mouse cages with 5cms of wood shavings and 20 blue and black marbles arranged in a 4x5 grid. Each cage was photographed prior to introducing the subject mouse. 20 min. later, each mouse was removed, and a second picture was taken, from which the number of buried marbles was counted by a blind observer. Marbles were considered "buried" if more than 2/3 of their surface was no longer visible.

Open Field
To assess 'anxiety-like' behavior, mice were placed into an open arena (45cm x 45cm). A central square within the arena was outlined on the bottom of the test (15cm x 15cm). Ethovision software (Noldus) was used to quantify total distance traveled, velocity, and time in the central square for 10 min.

Microglial RNA sequencing
To isolate microglia for RNA sequencing, bilateral NAc was micro-dissected from saline perfused brains at P60 and microglial isolations were conducted using a CD11b antibody-based procedure, according to published methods 35,36 . RNA was extracted from both CD11b+ (microglia) and CD11b-cell populations. Samples were transferred to the MGH Next Generation Sequencing Core for sequencing on an Illumina HiSeq 2500 instrument and the resulting data was analyzed in house using DESeq2 39

Immunohistochemistry (IHC)
For all IHC, animals were euthanized via CO 2 inhalation during adolescence (P30-45) and brains were perfused with ice-cold saline followed by 4% paraformaldehyde. Brains were then removed and post-xed for 48 hrs in 4% paraformaldehyde, followed by 48 hrs in 30% sucrose with 0.1% sodium azide. Next, brains were ash frozen in 2-methylbutane and stored at -20°C until sectioning at 40µm on a cryostat (Leica Biosystems). All IHC methods were conducted according to our previous work 32 , and detailed methods, antibody concentrations, and imaging procedures can be found in Supplementary methods.

Tissue Punches for Gene Expression
Tissue Collection and Punches At P45, animals were euthanized and perfused with ice cold saline. Brains were removed and frozen in 2methylbutane. Punches were collected using a 1mm diameter core sampling tool (Electron Microscopy Sciences). Brains were mounted in a sterilized cryostat and punches were collected by inserting the sterilized core sampling tool to a depth of 1mm. Punches were immediately placed into Trizol and frozen at -80°C until RNA extraction. RNA was extracted, cDNA synthesized, and qPCR run on a Mastercycler ep realplex (Eppendorf). qPCR primers were designed in house and purchased from Integrated DNA technologies. Relative gene expression was calculated using the 2-∆∆CT method, relative to the housekeeping gene (18S) and the lowest sample on the plate (see Supplementary Methods).

Cross-fostering experiments
For cross-fostering experiments, WT C57Bl/6 dams were mated and exposed to CON or DEP/MS as described. On P0, pups were cross-fostered to dams of the same treatment (CON- . Entire litters were swapped by placing them in a small cup lled with bedding from the recipient mother. The entire litter was then placed into the recipient cage at once. Maternal care was assessed 3x daily for the rst 3 days of life (Supplementary Methods). Offspring were weaned at P24 into cages with same sex littermates. Offspring behavioral and neural endpoints were assessed according to the methods described above and in Supplementary Methods.

Statistics
Statistical analyses were conducted using GraphPad Prism 9 software, the Qiime2-2020.2 analysis platform, PAST (PAleonto-logical STatistics), the R environment, and Python. Detailed descriptions of all statistical methods are available in the gure legends and in Supplementary Table 2. Data are expressed as mean +/-SEM and statistical signi cance was set at p < 0.05.

Prenatal DEP/MS exposure induces social de cits in male offspring only
We characterized social behavior in offspring following prenatal exposure to either combined DEP/MS (diesel exhaust particles and maternal stress) or control (CON) conditions. No group differences were observed in maternal weight gain during pregnancy, litter size, sex ratio, or offspring body weight (Fig.  S1). For complete litters, animal numbers, and statistics throughout the manuscript, see gure legends and Supplementary Tables 1 & 2, respectively).
The adolescent period -~ postnatal day (P) 25-45 in mice -is one during which social interactions with peers are of heightened importance. Thus, we conducted behavioral testing during this period (Fig. 1a). In the sociability assay, we found that CON males showed a strong preference for a novel sex-, age-, and treatment-matched social stimulus as compared to an object, whereas DEP/MS males did not ( Fig. 1bd). This effect was male-speci c as no such difference was observed in females. In the social novelty preference test, CON males showed a strong preference for a novel social stimulus over a cage mate, while DEP/MS males did not ( Fig. 1e-g). To ascertain whether social de cits were driven primarily by one prenatal exposure or the other, we tested sociability in offspring following DEP or MS alone. Neither treatment on its own induced social de cits, indicating that synergism between the two is required (Fig. 1h-j). Finally, we observed no treatment effects on marble-burying (Fig. 1k&l) or anxiety-like behavior ( Fig. 1m&n), suggesting that the effects of DEP/MS are speci c to the social domain, at least during adolescence.
DEP/MS induces a hyper-rami ed phenotype in male but not female microglia Developmental insults have been shown to have a particularly potent impact on microglia, the resident immune cells of the brain. Microglia play a critical role in the organization of neural circuits via synaptic pruning 17 , trogocytosis 19 , and other neuronal contracts 19 . Gestational DEP increases microglial-neuronal interactions in cortex, and DEP/MS impairs microglial pruning of thalamocortical synapses in male offspring 31,32 . Therefore, we tested how DEP/MS impacts microglia in the NAc. Microglial morphology is often taken as an early indicator of alterations in microglial function. Using a MATLAB-based, semiautomatic program to quantify microglial rami cation (3DMorph 52 ), we found that NAc-microglia are hyper-rami ed following DEP/MS in males but not in females (Fig. 2a). No difference was observed in microglial density between CON and DEP/MS males or females, assessed using IHC for Iba1 (quanti ed by cell density; Fig. 2b). To further de ne changes in microglial morphology in males, we used Imaris 3D image reconstruction software which revealed increased total volume of NAc-microglia following DEP/MS (Fig. 2c). Sholl analysis revealed signi cantly more branch endpoints and Sholl intersections in DEP/MS male microglia as compared to CON (Fig. 2d-f). In females, no treatment effects were observed on microglial volume, branch endpoints, or Sholl intersections (Fig. S2). These ndings demonstrate, using multiple approaches, that male microglia are larger and hyper-rami ed in the NAc following DEP/MS.

DEP/MS alters dopamine circuitry in males but not females
Based on our microglial ndings, we hypothesized that NAc-microglia might be interacting more with neuronal populations in the NAc, potentially to alter social behavior. Indeed, microglia eliminate dopamine D1 receptors (D1Rs) in the NAc (with a speci c peak during adolescence [at P30]), and this D1R elimination is critical to the developmental trajectory of social play 18 . Dopamine, endogenous opioids, and oxytocin all mediate social motivation by acting within the NAc 18,53,54 . Thus, we performed tissue punching and qPCR for mRNA for dopamine receptors (Drd1 and Drd2), opioid receptors (Oprk1, Oprm1) and oxytocin receptor (Oxtr). We found that both Drd1 and Drd2 mRNA were decreased in the NAc of adolescent male offspring following DEP/MS exposure (Fig. 3a), but not in females (Fig. 3b). mRNA was not decreased for Oprm1 or Oxtr although, interestingly, Oprk1 mRNA was lower in males and higher in females (Fig. 3a&b). No changes in these receptors were observed in the amygdala (Fig. S3) -which is also a critical regulator of social behavior via connections with the NAc 55 .
We next asked whether decreases in D1R might be due to increased microglial phagocytosis of D1R at P30. However, we observed no differences in microglial engulfment of D1R using IMARIS 3D volumetric reconstructions of D1R and microglia (Iba1; Fig. S3). Alternatively, reduced dopaminergic input from the VTA might account for the lower expression of NAc-D1/D2Rs following DEP/MS in males. Therefore, we used immunohistochemistry to quantify tyrosine hydroxylase (Th) ber density within the NAc as a measure of dopaminergic input. We observed a signi cant interaction effect whereby Th mean grey value (a measure of density) was decreased in males but increased in females following DEP/MS (Fig. 3c&d). These data show that decreased dopamine receptor expression in DEP/MS males coincides with a decrease, albeit small, in dopamine input from the VTA into the NAc.
Chemogenetic activation of the dopamine system rescues male social de cits following DEP/MS Given the reduced Th-ber density in the NAc in male offspring following DEP/MS, we tested whether chemogenetic activation of the dopamine system would rescue social de cits in male offspring following DEP/MS. A Dat-Ires-Cre mouse line was used to generate offspring expressing Cre under the control of the dopamine transporter 1 (DAT) promotor (Fig. 3e). Cre + offspring were exposed to either CON or DEP/MS prenatally, and subsequently underwent stereotaxic viral transfection (P24-25) and social behavior testing (during adolescence; Fig. 3f-g). We found that CON males transfected with the control virus and treated with CNO showed a signi cant preference for the social stimulus, as we have previously observed (Fig. 3h). As predicted, DEP/MS males transduced with the control virus and treated with CNO displayed no social preference (Fig. 3h). In contrast, DEP/MS males transduced with the excitatory DREADD receptor showed a reinstatement of their preference for a social stimulus following CNO administration (Fig. 3h). Both CON males transduced with a control virus and DEP/MS males transduced with the excitatory DREADD virus spent signi cantly more time investigating the social stimulus as compared to an object, while DEP/MS males transduced with the control virus spent equal amounts of time investigating the animal and the object (Fig. 3i). Together, these ndings suggest that activating VTA-dopamine neurons is su cient to restore sociability following prenatal DEP/MS in males.

DEP/MS shifts the composition of the gut microbiome and epithelium
Our results so far demonstrate male-speci c changes in social behavior, microglial hyper-rami cation, and decreased dopaminergic tone within the NAc following DEP/MS. Importantly, changes in all these endpoints are observed following gut microbiome manipulation. In multiple ASD mouse models, supplementation with L. reuteri rescues social behavior de cits by modulating activity of the dopamine system 27,28 . Microglia are also exquisitely sensitive to gut microbiota 25,26 . Microglial hyper-rami cation, very similar to that observed in DEP/MS males, is reported in germ-free mice 25 .
Based on these ndings, we asked whether DEP/MS impacts the gut microbiome in offspring using bacterial 16S sequencing of cecal contents in offspring during early adulthood (P45). Measures of alpha diversity quantify community richness (how many bacterial taxa are present) or evenness (how evenly abundant the taxa are that form the community) within the gut microbiome of an individual animal. Community evenness was signi cantly increased in DEP/MS males as compared to CON (Pielou's evenness; Fig. 4a). Principal Coordinate Analysis (PCoA) of quantitative beta diversity indices revealed distinct clustering of microbiome pro les in CON and DEP/MS males (Fig. 4b). Signi cant differences were also observed in the phylogenetic relatedness and abundance of microbial communities between CON and DEP/MS males (Permutational multivariate analysis of variance [PERMANOVA]). Differences were not evident at the level of individual taxa. Notably, no changes in either alpha or beta diversity were found in female offspring (Fig. 4c&d).
Microbes within the gut interface directly with the intestinal epithelium and are important determinants of epithelial structure and immunity. The tight-junction proteins Occludin (Ocln) and Zonula occludins-1 (Zo1) stabilize the gap junctions between epithelial cells. Gastrointestinal dysfunction and evidence for disruption of the gut epithelial barrier -including changes in Ocln and Zo1 expression -are reported in patients with ASD 56,57 . We observed a sex-speci c effect of DEP/MS (decreased in males but increased in females) on Ocln and Zo1 mRNA in the ileum (Fig. 4e-g) and duodenum (Fig. S4), but not the colon (Fig. S4). Constipation and diarrhea are predominant components of GI dysfunction in ASD 56 . Oprm1 mRNA -a critical regulator of gut motility -was reduced following DEP/MS (Fig. 4h). Interestingly, we observed no changes in the proin ammatory genes Tlr4, Tnfα, or Il-1β in either the ileum (Fig. 4i-k) or colon (Fig. S4), suggesting that these effects are not due to current in ammation, per se. The structure of the intestinal epithelium itself is also sensitive to microbial composition 58,59,60 . Villi length (Fig. 4l-m) and mucosal thickness (Fig. 4n), but not crypt length (Fig. 4o), were increased following DEP/MS exposure. Together, these ndings demonstrate pervasive, male-biased changes in the microbiome and intestinal epithelium following DEP/MS -suggestive of decreased gut barrier function speci cally in male offspring.

DEP/MS shifts microglial gene expression towards a germfree phenotype
The DEP/MS-induced hyper-rami cation in microglia that we observe following DEP/MS is like that observed in germ-free mice 25 . This led us to ask whether microglial gene expression also changes in similar ways following DEP/MS and other microbial disruptions. We used strati ed Rank-Rank Hypergeometric Overlap (RRHO) analysis 61 to compare gene expression changes between two datasets: gene differentials in male microglia following DEP/MS vs. CON (see Fig. 2) to gene differentials in microglia from germ-free vs. conventionally housed male mice in a published dataset 26 . We observed signi cant concordance between genes that are differentially transcribed following DEP/MS and those that are altered in germ-free microglia (Fig. 5a). We also compared DEP/MS microglial gene differentials to gene differentials following acute immune activation (2h after lipopolysaccharide: LPS at P60 35 , Fig. 5b) and across typical development 35 (Fig. 5c). Interestingly, we found the opposite pattern (signi cant discordance) in both comparisons. These ndings are in line with the idea that germ-free microglia are immune-incompetent and immature and suggests that a similar phenotype is induced by DEP/MS exposure in male microglia.

Cross-fostering at birth prevents DEP/MS-induced social de cits in male offspring
Given the established link between gut microbiota and social behavior 62 we tested whether shifting the gut microbiome towards a CON-typical composition would prevent social de cits in DEP/MS male offspring. Cross-fostering on the day of birth shifts the composition of the offspring gut microbiome towards that of the foster mother 63-65 . DEP/MS exposed pups were fostered to either a different DEP/MS dam on the day of birth (D → fD) or to a CON dam (D→ fC). Similarly, CON exposed pups were fostered to a different CON dam (C → fC) or to a DEP/MS dam (C→ fD). Offspring were then tested on social behavior assays during adolescence prior to sacri ce and sample collection for microbiome and gut analyses (Fig. 5d). We also assessed maternal behavior to rule out that changes we observed were due to differences in maternal care. We found no differences between DEP/MS and CON dams in time spent on the nest nursing (Fig. 5e) or in licking and grooming behaviors ( Fig. 5f; Fig. S5).
To verify that cross-fostering of DEP/MS pups to a CON dam shifted the gut microbiome towards a CONtypical phenotype, we used 16S sequencing of cecal contents at P45. Indeed, alpha diversity differed signi cantly with foster condition (Pielou's Evenness, Fig. 5g). D→ fC males had signi cantly higher evenness than D → fD males (Fig. 5g) that did not differ from C→ fC males. Furthermore, PERMANOVA analysis revealed divergent microbial community structure between D → fD, D→ fC, and C→ fC males in all four beta diversity indices (Fig. 5h, Bray-Curtis, Jaccard dissimilarity, unweighted and weighted UniFrac). The microbiome of D → fD males differed signi cantly from that of C→ fC, and D→ fC males. Linear discrimination analysis effect size (LEfSe) identi ed several genera of bacteria that differed between D → fD and D→ fC male offspring (Fig. 5i-j). Among these, Helicobacter, Bacteroides and Parabacteroides were more abundant in D → fD male offspring, while Lachnospiraceae, and Oscillospira were more abundant in D→ fC male offspring. Helicobacter pylori (H. pylori) has been implicated in gut in ammation and acute gastritis 66,67 . Both Parabacteroides and Bacteroides are differentially abundant in the gut microbiome of human patients with ASD 68,69 . Neither tight junction protein mRNA nor villi length/mucosal thickness differed between D → fD and D→ fC males in the ileum at P45 (Fig. S6). We also conducted metabolomic analysis of short chain fatty acids (SCFA) to determine whether bacterial metabolites were in uenced by our cross-fostering manipulation. We found that 6 SCFAs including acetate and butyrate were increased in D → fD males as compared to C→ fC, and that this increase was abolished in D→ fC males (Fig. 5k).
Newborn offspring acquire microbes from their mother during rearing in the home cage. We hypothesized that DEP/MS-induced changes in the gut microbiome were due to differential transmission of maternal microbes. To our surprise, we found no differences between DEP/MS or CON dams at any timepoint (Fig.  S7). Furthermore, there were no differences in the vaginal or milk microbiomes (Fig. S7). These ndings point to the intriguing possibility that non-microbial constituents of maternal milk may carry the signal that leads to microbiome restoration in DEP/MS pups fostered to a CON dam, an exciting avenue for future studies.
Next, we tested whether cross-fostering to a CON dam could rescue sociability in DEP/MS-exposed male offspring. As expected, D → fD offspring displayed no preference for a social stimulus (Fig. 5l). However, D→fC males showed signi cantly higher sociability (Fig. 5m) and spent more time in social investigation as compared to D → fD offspring (Fig. 5l). In the social novelty preference test, D→fC males spent signi cantly more time in total social investigation as compared to D → fD males, but there was no signi cant effect on social novelty preference, per se (Fig. S8). This nding may suggest that shifting the gut microbiome increases social motivation across assays, rather than choice of social partner. We also compared the sociability of CON-exposed males fostered to DEP/MS dams (C→ fD) to that of CONexposed males fostered to CON dams (C→ fC). We found no difference in sociability between these groups, suggesting that cross fostering is insu cient to induce a DEP/MS behavioral phenotype on its own (Fig. S8). In sum, these ndings demonstrate that intervening at the level of the microbiome can ameliorate social de cits in male offspring following DEP/MS.
Cross-fostering at birth prevents DEP/MS-induced microglial hyper-rami cation, but does not affect the dopamine system, in male offspring Finally, we asked whether cross-fostering to a CON dam prevented the microglial and dopaminergic phenotypes we observed in DEP/MS-exposed male offspring. Using Imaris 3D image reconstruction, we observed signi cant main effects of treatment on microglial volume (Fig. 6a), branch endpoints (Fig. 6b) and Sholl intersections (Fig. 6c). D→ fC males had signi cantly smaller and less rami ed microglia as compared to D → fD males (Fig. 6d), indicating that microbial intervention at birth prevents microglial, as well as social, alterations. We also conducted RNA sequencing of NAc-microglia isolated from D → fD and D→ fC males. Gene expression differed dramatically between these two conditions, with many microglial genes both up-and down-regulated following cross-fostering to a CON dam at birth (Fig. 6e).
Interestingly, these genes were not the same gene sets that differed between CON and DEP/MS males in our previous assessment. Rather, genes such as Ccrl2, Cxcl9, Mpo, and Cstdc2 were the most up or down regulated (Fig. 6f). These changes suggest, not surprisingly, that a distinct transcriptional pro le is associated with returning microglia to a CON morphological phenotype following DEP/MS (D→ fC ) compared to those that differed between CON and DEP/MS independent of cross-fostering. Thus, crossfostering, on its own, likely shifted the microglial transcriptome. We also assessed D1 and D2 receptor mRNA within the NAc, as well as Th cell number within the VTA, to determine whether cross-fostering to a CON dam prevented the changes we observed in the dopamine system. Interestingly, we found no difference between D → fD and D→ fC males in D1R mRNA (Fig. 6g), D2R mRNA (Fig. 6h), or Th cell number (Fig. 6i&j).

Discussion
While epidemiological work has incontestably shown that toxicants and psychosocial stressors converge on vulnerable populations to increase disease burden, little mechanistic work has been done to understand the underlying biological substrates of these effects. Here, we show that DEP/MS alters the composition of the gut microbiome, induces microglial hyper-rami cation, and decreases dopaminergic tone within reward circuits in offspring, all in a male-biased manner. Furthermore, we nd novel evidence that social behavior de cits following DEP/MS can be rescued or prevented by activating the dopamine system or by shifting the composition of the gut microbiome. Moreover, modulation of the gut microbiome at birth prevents changes in microglial morphology, but not changes in the dopamine system.
Changes in social behavior, microglia, and the dopamine system have been reported in models of maternal immune activation (MIA) that use bacterial and viral mimetics (i.e. lipopolysaccharide [LPS] and Poly I:C, respectively) to directly activate the maternal immune system 71,72 . Our ndings suggest that gestational exposure to air pollution and maternal resource deprivation can elicit similar phenotypes in offspring -highlighting the harmfulness of these pervasive environmental exposures. This is especially interesting given that the diesel exhaust particles are given to the mother, and yet they have a potent, enduring impact on offspring. Our previous work showed that DEP/MS exposure leads to elevated maternal blood serum concentrations of cytokines such as IL-6, IL-17a, and TNFα, similar to what is observed in MIA 32 . This may suggest that DEP/MS alters the postnatal microbiome offspring as a secondary consequence of fetal reprogramming of the gut and brain by maternal immune factors.
Interestingly, while previous work in several ASD models has shown that modulation of the gut microbiome acts via the dopamine system to in uence behavior 27,28 , we nd that restoration of the gut microbiome does not alter the dopamine system. Rather, we nd that microglial morphology is robustly impacted by microbiome modi cation. This is in line with previous work showing that the microbiome is a potent organizer of microglial development, particularly during the perinatal period. For example, ceasarean-section delivery -which prevents offspring from receiving microbes typically acquired during passage through the vaginal canal -and represents a highly stressful birth event -leads to altered microglial colonization in the hippocampus and paraventricular nucleus of the hypothalamus 73 .
Importantly, brain region-speci c modi cation of dopaminergic function remains di cult in human patients. In contrast, the gut microbiome represents a more tractable therapeutic target. One strength of our work is that our cross-fostering procedure demonstrates that shifting the overall composition of the gut microbiome is effective at restoring sociability. There are several possible mechanisms by which this might be mediated, including changes in microbial metabolites such as short chain fatty acids (SCFAs) and direct vagal nerve activation 62 . Indeed, our ndings show that several SCFAs such as acetate and butyrate may be important mediators. Future studies are needed to determine the precise route by which changes in the gut microbiome following DEP/MS exposure leads to changes in the brain and, therefore, social behavior.
One important point is that while modi cation of the gut microbiome is su cient to prevent social behavior impairments in male offspring following DEP/MS, cross-fostering to a DEP/MS dam was insu cient to induce social behavior de cits in CON males. Excitingly, this suggests that the gut microbiome may be an important target for intervention, even if not responsible for the entire disease phenotype. Similarly, Sgritta et al., (2019) found that supplementation with L. reuteri rescued social behavior de cits in valproic acid-treated mice, despite L. reuteri abundance not being reduced by this exposure in the rst place. It also suggests that changes to the immune system and/or intestinal epithelium in utero may precede, and to some degree dictate, which bacterial taxa gain a foothold. This is especially likely given that our effects are observed in male but not female offspring, despite the same microbiome exposure. Indeed, numerous studies have reported sex differences in the gut microbiome and microglia, as well as their relationship to each other 26,74 . For example, Thion et al. characterized microglial brain colonization and gene expression in male and female germ-free vs. conventionally housed mice. They found that germ-free male mice had signi cantly altered microglial colonization and gene expression during embryonic development, while female germ-free mice had more microglial transcriptional changes in adulthood 26 . Elucidating the mechanisms underlying such sex differences is an important area for future investigation.
In closing, our results characterize male-speci c behavioral, microglial, and neural effects of synergistic exposure to both air pollution and maternal stress during pregnancy. Furthermore, we identify two potential sites for therapeutic intervention in the treatment of social behavior impairments in ASD.

Declarations
Con icts of Interest: The authors have no con icts of interest to report.   Cross-fostering prevents changes in microglial hyper-rami cation but does not alter dopaminergic endpoints. a, Imaris 3D reconstruction revealed that cross-fostering to a CON dam at birth reduced microglial volume following DEP/MS (a) as well as branchpoints (b) and sholl intersections (c; N=19-26 microglia from 6-8 animals/group, nested one-way ANOVA [C → fC vs. D → fD vs. D → fC], volume: p=0.014, branch endpoints: p=0.02, sholl intersections: p<0.001). In a and b, pale grey dots represent