Gonadal Hormones Impart Male-Biased Behavioral Vulnerabilities to Immune Activation via Microglial Mitochondrial Function

SUMMARY

There is a strong male bias in the prevalence of many neurodevelopmental disorders such as autism spectrum disorder. However, the mechanisms underlying this sex bias remain elusive. Infection during the perinatal period is associated with an increased risk of neurodevelopmental disorder development. Here, we used a mouse model of early-life immune activation that reliably induces deficits in social behaviors only in males. We demonstrate that male-biased alterations in social behavior are dependent upon microglial immune signaling and are coupled to alterations in mitochondrial morphology, gene expression, and function specifically within microglia, the innate immune cells of the brain. Additionally, we show that this behavioral and microglial mitochondrial vulnerability to early-life immune activation is programmed by the male-typical perinatal gonadal hormone surge. These findings demonstrate that social behavior in males over the lifespan are regulated by microglia-specific mechanisms that are shaped by events that occur in early development.

INTRODUCTION

Neurodevelopmental disorders are a major public health challenge, affecting up to 18% of children worldwide [1]. There is a strong male bias in the prevalence of early-onset neurodevelopmental disorders such as Autism Spectrum Disorder (ASD) and childhood onset-schizophrenia (SZ) [2–6]. For instance, ~3–4 males are diagnosed with ASD to every female [7, 8]. Despite this strong male bias, the mechanisms that program male vulnerability, or female resilience, remain unknown. Transcriptomic profiling of typically-developing male and female brains demonstrated that genes more highly expressed in male brains are also significantly enriched in the brains of individuals diagnosed with ASD [9]. Interestingly, these same genes are enriched in microglia, the innate immune cells of the central nervous system [9]. Immune alterations are implicated in the etiology of many neurodevelopmental disorders such as ASD [10–12]. Indeed, maternal infection or fever during the perinatal period is associated with increased risk of offspring ASD diagnoses [13–15]. Transcriptomic profiling of postmortem brains from individuals diagnosed with ASD identified enrichment of neuroinflammatory genes and dysregulation of genes in microglia, although these analyses were not sex-stratified [16–21].

Importantly, there are sex differences in microglial development and function during physiological neurodevelopment [22–25]. These developmentally-programmed sex differences are largely organized by the surge of gonadal hormones that occurs only in males during a critical period of perinatal brain development and serves to masculinize the body and brain [26]. Importantly, microglial inhibition prevents hormone-induced masculinization, revealing the importance of microglial function to this male-typical neurodevelopment [27–29]. Intriguingly, the in utero hormonal milieu is hypothesized to play a role in the sex-biased risk profile of ASD [30]. Fetal exposure to high levels of the hormone estradiol – the aromatized form of gonadally-produced testosterone [31, 32] - is highly correlated with an increased odds ratio of developing ASD [33, 34]. However, other analyses have demonstrated that lower maternal serum levels of unconjugated estriol during the second trimester are also associated with offspring ASD diagnoses [35], demonstrating the necessity to understand how gonadal hormones contribute to the sex bias in neurodevelopmental disorder prevalence. Whether gonadal hormone-mediated masculinization of the male brain programs the male bias in neurodevelopmental disorder susceptibility, and the mechanisms through which brain masculinization induces these vulnerabilities, remain unresolved.

Alterations in social behavior are a core symptom of ASD, and microglia play a critical role in the organization of social behaviors in male rodents [36]. Microglia are involved in synaptic pruning and refinement, suggesting that microglial dysfunction may contribute to synaptic abnormalities associated with some neurodevelopmental disorders [37]. Numerous studies suggest an intimate link between microglial inflammatory processes and mitochondrial functions [38–40]. Importantly, mitochondrial respiratory function is critical for the regulation of microglial homeostatic and inflammatory processes and is impaired following lipopolysaccharide (LPS)-mediated proinflammatory activation [38, 39, 41, 42]. Social behavioral changes may also be regulated by dysfunctional mitochondrial metabolism [43–45]. Mitochondrial metabolism was identified as the key molecular mechanism underlying changes in social behaviors in the CYFIP1 knockout fly model of neurodevelopmental disorders [43]. Postmortem analyses of brain tissues from individuals diagnosed with ASD or SZ revealed decreased expression of mitochondrial electron transport chain (ETC) proteins as well as alterations in proteins regulating mitochondrial morphology [46, 47], and mitochondrial Complex I availability was lower in male individuals diagnosed with ASD compared to neurotypical controls [48], suggestive of potential deficits in mitochondrial respiratory capacity and function in these brain tissues. Similar to the sex differences in ASD prevalence, sex differences in mitochondrial function in both healthy tissues as well as following perinatal injury have been noted [49–51]. However, these studies often isolated mitochondria from bulk brain tissue instead of from specific cell types, potentially masking cell-specific sex-specific biology. Whether immune activation during the perinatal period impacts microglial mitochondria, and whether these cell type-specific organellar alterations impact brain function, including social behavior, in a sex-specific manner, is unknown.

Here, we used an established model of early-life immune activation with the bacterial endotoxin LPS [52–55] to investigate whether gonadal hormones present during perinatal brain development induce male vulnerability to immune activation through influencing microglial mitochondrial function. We found that early-life LPS challenge resulted in male-biased deficits in social behavior that were dependent upon microglial immune signaling. This male behavioral vulnerability was coupled to male-biased alterations in microglial mitochondrial morphology, gene expression, and respiratory function. We demonstrate that this vulnerability can be induced by masculinizing female mouse pups at birth with the gonadally-derived hormone estradiol prior to immune activation. These data uncover a fundamental mechanism through which physiological developmental programs may regulate sex- and cell-specific neurological susceptibilities.

METHODS

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Experimental animals

Wild-type (WT) C57BL/6J mice were purchased from Jackson Laboratories (Bar Harbor, ME: Stock # 000664). TLR4-floxed mice were purchased from Jackson Laboratories (Bar Harbor, ME: Stock # 024872). Cx3cr1-CreBT (MW126GSat) mice were generated and provided by L. Kus (GENSAT BAC Transgenic Project, Rockefeller University, NY) and backcrossed over 12 generations on a C57BL/6J (Charles River Laboratories) background. All mice were group housed in standard mouse cages under standard conditions (12 hour light/dark cycle, 23°C, 50% humidity) with same-sex littermates. All dams used for breeding in this manuscript were second-time breeders. All breedings occurred in-house and harem breeding strategies were used. All experiments were performed in accordance with the NIH Guide to the Care and Use of Laboratory Animals and approved by the Massachusetts General Hospital Institutional Animal Care and Use Committee. Mice were weaned on postnatal day 24 into new cages with a maximum of 4 mice per sex per cage post-weaning. See Supplementary Methods for detailed methods regarding determination of transgenic mouse genotypes.

METHOD DETAILS

Masculinization

100 ug estradiol benzoate (Sigma-Aldrich catalog # E8515, St. Louis, MO, USA) dissolved in sesame seed oil was administered subcutaneously on postnatal days 1 and 2 [56–58].

PN9 Saline or lipopolysaccharide injection

10 mg/kg lipopolysaccharide (from Escherichia coli serotype 0111:B4, Sigma-Aldrich, St. Louis, MO, USA) dissolved in 0.9% sterile saline or 0.9% sterile saline control was administered subcutaneously on postnatal day 9 (PN9). This dose and timing were chosen based on previous work demonstrating its ability to induce male-specific changes in social behavior [52] and data suggesting that PN9 is relatively equivalent to term in humans [59]. All animals within a litter received the same injection to prevent any indirect exposure to the other drug treatments. Litter effects were avoided by populating experimental groups with a maximum of 2 mice per sex from individual litters.

Behavioral assays

Behavioral assays were conducted at PN15 and PN30–45 in a separate cohort of animals from those used for molecular analyses. For all assays, testing took place in the first half of the light phase. Mice were moved to the behavioral testing room 1 hr prior to behavioral testing to habituate to the room each day and were additionally habituated to the testing room for 4 hr prior to the first day of testing. All behavior was video recorded and hand-scored using Solomon Coder by a blinded observer.

Juvenile social exploration

Social exploratory behavior was assessed at PN15, an age in which mouse pups have opened their eyes and are becoming more independent (Bilbo et al., 2018). Extra nesting material was placed into the home cages of age and treatment-matched litters at PN13. On the day of testing, litters are acclimated to testing room for 1 hr prior to testing. A quarter-sized amount of nesting material is taken from the cage of the subject mouse and placed on one side of the testing cage that contains clean bedding material. A quarter-sized amount of nesting material is also taken from the cage of an age- and treatment-matched litter and spread on the opposite side of the testing cage. PN15 mice are then placed in the center of the cage and allowed to explore for 3 min. For scoring purposes, the cage was divided into three areas: home nesting, center, and unfamiliar nesting. The amount of time spent in the home nesting side, and the latency to first enter the side containing unfamiliar litter nesting was quantified.

Sociability and social novelty preference tasks

A 3-chambered social preference task was used to test the preference of mice to investigate either a social stimulus or a non-social stimulus (sociability) or a novel social stimulus vs a familiar social stimulus (social novelty preference), as previously described (Smith et al., 2020). For the sociability task, novel age-, sex- and treatment-matched conspecific mice or a novel rubber duck were placed in small plexiglass containers on opposite sides of the 3-chambered box. Experimental mice were placed in the center chamber and allowed to freely explore for 5 min. Separate 3-chambered boxes were used for female and mice male, and sexes were tested on separate days. Sociability was calculated as the amount of time spent investigating the novel social mouse as a proportion of total time spent investigating any stimulus as a quantification of social preference. For the social novelty preference task, a similar procedure was used, with the two stimuli this time consisting of a new novel age-, sex- and treatment-matched conspecific mouse or a familiar cage mate sibling. Experimental mice were then placed in the center chamber and allowed to freely explore for 10 min. The following behavioral outcomes were manually scored using Solomon Coder: time spent investigating the novel mouse, time spent investigating the novel object (sociability task), time spent investigating the familiar cage mate sibling (social novelty preference task), and time spent in each chamber.

Elevated zero maze

The elevated zero maze was used to test for anxiety-like behavior as previously described [60]. We used a zero maze (Stoelting) with an elevated (50 cm height) circular lane (5 cm wide) that is divided into four quadrants. Two quadrants are enclosed by 15 cm tall walls with the remaining two quadrants left exposed. Test mice were placed into a closed arm, and time spent in the open or closed quadrant arms was scored over a 5 min period.

Light dark box

The light dark box task was used as a second test for anxiety-like behavior as previously described. We used a box with 35 cm tall walls that was split into two 20cm x 40cm chambers. The walls on one side were clear and brightly illuminated while the walls of the other chamber were opaque and the chamber itself was dark, with a small open space allowing for free passage between chambers. Test mice were placed into the dark chamber, and time spent in the light or dark chambers was scored over a 10 min period.

Marble burying

Standard mouse cages were prepared with 5 cm depth wood shavings (5 cm depth). 20 black and blue colored marbles were arranged on the top of the wood shaving in a 5×4 grid formation. Mice were then placed in the cage for 20 min, after which mice were removed from cage and a picture was taken. A blinded observer assessed number of buried marbles. Marbles were considered to be buried if > 2/3 of surface was no longer visible above the wood shavings.

Brain Endotoxin Level Determination

Endotoxin was measured in PN10 brain using a chromogenic kinetic limulus amebocyte lysate (LAL) assay according to our laboratory’s previously published protocol [61]. Following transcardial perfusion with ice-cold saline, whole forebrains from PN10 mice were extracted and frozen in the vapor phase of liquid nitrogen until homogenization in 500 uL endotoxin-free water using an oven-baked Dounce homogenizer. Samples were then diluted 6-fold in endotoxin-free water and a Kinetic-QCL Kinetic Chromogenic LAL Assay (Lonza catalog # 50–650U) was performed according to manufacturer instructions. 96 well plates were run on a temperature controlled kinetic SpectraMax ABS Plate Reader (Molecular Devices) and analyzed using SoftMax Pro Software with a baseline correction of O.D. = 0.2.

Blood Serum Multiplexed Cytokine Assay

Blood was collected by cardiac puncture from PN10 mice into BD Microtainer Blood Collection tubes (SKU: 365967), allowed to clot for 30 min, and centrifuged at 1250g for 15 min at room temperature to collect serum. Levels of CCL2, IL-1β, IL-6, and TNF-α in serum were then quantified using a mouse Simple Plex SPCKC-MP-004230 (Protein Simple) multiplexed cytokine array on an Ella ELISA machine (Protein Simple).

Microglia isolation for gene expression

Microglia isolations were performed as previously described [62]. CD11b+ cells (herein referred to as microglia) were isolated from anterior cingulate cortex (ACC), a brain region critical for the regulation of social behavior [63, 64], on PN30. Following transcardial perfusion with ice-cold saline, ACC was dissected from whole brain on ice using sterile forceps and minced on a petri dish set on ice using a sterile razor blade. Tissue homogenate was then placed in Hank’s Buffered Salt Solution (HBSS; ThermoFisher Scientific, NY, USA) with collagenase A (Roche, Indianapolis, IN, USA; Catalog # 11088793001) and 0.4 mg/mL DNase I (Roche, Indianapolis, IN, USA; Catalog # ) and incubated for 15 min at 37°C. Mechanical dissociation of tissue homogenate was performed by pipetting samples through successively smaller flame-polished Pasteur pipettes until a single-cell suspension was obtained. Samples were then filtered, rinsed with HBSS, and centrifuged. Samples were then incubated on ice for 15 min with CD11b-conjugated microbeads (Miltenyi Biotec, San Diego, CA, USA: Catalog # 130–093-634), and passed through a magnetic bead column (Miltenyi Biotec, San Diego, CA, USA: Catalog # 130–042-401) to separate CD11b+ and CD11b-cells. Cells were washed in 1X PBS and resuspended in Trizol (ThermoFisher Scientific, NY, USA: Catalog # 15596026), after which they were stored at −80°C until RNA extraction.

RNA extraction

Isolated microglia were homogenized in Trizol, mixed for 10 min at 2000 rpm on a MixMate (Eppendorf), followed by resting at room temperature for 15 min. Chloroform (1:5 with Trizol) was then added and samples were mixed for 2 min at 2000 rpm on a MixMate (Eppendorf). Tubes then sat at room temperature for 3 min followed by centrifugation for 15 min at 11,900 rpm at 4°C. Aqueous phase was aliquoted into a fresh tube and isopropanol was added (1:1 with aqueous phase volume) along with 2 uL of GlycoBlue coprecipitant (ThermoFisher Scientific: Catalog # AM9515). Tubes were then mixed for 1 min at 2000 rpm on a MixMate, allowed to incubate at RT for 10 min, and centrifuged at 11,900 rpm for 10 min at 4°C. Pellets were then rinsed twice in ice-cold ethanol (75%) and resuspended in 6–8 uL of nuclease-free water (ThermoFisher Scientific: Catalog # AM9932). RNA quantity and purity was assessed using a NanoDrop 2000 (ThermoFisher Scientific).

cDNA synthesis and qPCR

cDNA was synthesized from 200 ng of total RNA using the QuantiTect Reverse Transcription Kit (Qiagen, Hilden, Germany) according to manufacturer’s instructions. qPCR was subsequently performed using this cDNA and Taqman gene expression assays performed on a QuantStudio 5 Real-Time PCR System (ThermoFisher Scientific) as previously reported [65]. Gene expression was normalized to the reference gene 18S and calculating using 2-ΔΔCt method relative to the lowest sample on the plate.

cDNA synthesis and PCR Array

cDNA to be used for PCR Array was synthesized from 200 ng of total RNA using the RT² First Strand Kit (Qiagen: Catalog # 330404) according to manufacturer’s instructions. Mitochondrial gene expression was then assessed using the RT² Profiler^™ PCR Array Mouse Mitochondrial Energy Metabolism (Qiagen: Catalog # PAMM-008ZA-24) and RT² SYBR Green ROX qPCR Mastermix (Qiagen: Catalog # 330523) according to manufacturer’s instructions.

Immunohistochemistry and confocal microscopy

Following transcardial perfusion with ice-cold saline followed by 4% paraformaldehyde (PFA), brains were post-fixed in 4% PFA for 48 hr at 4°C. PFA-fixed brains were then cryoprotected in 30% sucrose for at least 48 hr at 4°C. Brains were then flash frozen in 2-methylbutane on dry ice and stored at −80°C before cryosectioning at 30 μm on a Leica cryostat. Brain sections were stored in cryoprotectant until staining. Tissue sections of interest were then removed from cryoprotectant and rinsed 3 × 5 min in 1X PBS to remove excess cryoprotectant. Immunostaining was performed on layer II/III of the dorsal ACC (area 24b, approximate bregma level 1.69mm, Figure 17 in [66]). Tissue was first permeabilized in 0.3% Triton-X100 in PBS for 30 min and rinsed 3 × 5 min in 1X PBS. Epitope retrieval was then performed by incubating at 80°C in 10 mM citric acid (pH 9.0) for 30 min followed by 3 × 10 min 1X PBS washes. Background fluorescence was then quenched by 60 min incubation in 1 mg/mL sodium tetraborate in 0.1M PB, followed by 6× 5 min PBS washes and incubation in 50% methanol for 60 min followed by 6× 5 min PBS washing. Sections were then blocked in 10% normal goat serum (Vector Laboratories: Catalog # S-100–20) in PBS for 60 min. Primary antibodies were applied sequentially for overnight incubations in 5% normal goat serum and 0.3% Tween20 in PBS at 4°C. Primary antibodies used were chicken anti-Iba1 (Synaptic Systems 234 006, 1:500) and rabbit anti-Tomm20 (Novus NBP2–67501, 1:500). Secondary antibodies were then applied for 2 hr at RT protected from light in 5% NGS with 0.3% Tween20 followed by 6 × 10 min PBS washes. Secondary antibodies used were goat anti-rabbit highly cross-adsorbed Alexa Fluor Plus 488 (ThermoFisher Scientific A32731, 1:200) and goat anti-chicken Alexa Fluor 568 (ThermoFisher Scientific A-11041, 1:200). Sections were then mounted on gelatin subbed slides, coverslipped with VECTASHIELD Antifade Mounting Medium with DAPI (Vector Laboratories, Burlingame, CA, USA: Catalog #H-1000), sealed with nail polish, and protected from light until imaging.

Imaging analyses

To quantify immunofluorescence of Iba1 and Tomm20 staining, 63x (x1.39) magnification Z-stacks were taken on a Nikon A1R confocal microscope. Z-stacks were taken every 0.2 μm. Mitochondrial length and volume within microglia were quantified using Imaris Versions 9.0 and 10.0 software (Bitplane). Individual channel backgrounds (microglia, mitochondria, nuclei) were first subtracted and then threshold values were recorded using Fiji software. Images were then uploaded into Imaris to create volumetric images using the Surface tool. Mitochondria were masked [36] inside of the microglial (Iba1+) signal prior to reconstruction to quantify mitochondrial morphology within microglia, and then mitochondrial length was measured using the BoundingBoxOO Length C setting (longest principal axis) as previously described [67]. All Tomm20 signal masked outside of microglia (Iba1-) was used to quantify non-microglial signal. Mitochondrial volume was also exported for statistical analysis.

Flow cytometry

Mice were transcardially perfused with ice-cold saline until perfusate was clear, after which ACC was dissected on ice and minced on a petri dish set using a sterile razor blade. Tissue homogenate was then placed in Hank’s Buffered Salt Solution (HBSS; ThermoFisher Scientific, NY, USA) with collagenase A (Roche, Indianapolis, IN, USA; Catalog # 11088793001) and 0.4 mg/mL DNase I (Roche, Indianapolis, IN, USA; Catalog # ) and incubated for 15 min at 37°C. Mechanical dissociation of tissue homogenate was performed by pipetting samples through successively smaller flame-polished Pasteur pipettes until a single-cell suspension was obtained. Samples were then filtered, rinsed with HBSS, and centrifuged. Cells were then cleaned using Debris Removal Solution (Miltenyi Biotec: Catalog # 130–109-398), washed with PBS, and then plated in a 96 well U-bottomed plate at 70,000 cells/well in Stain Buffer (BD Biosciences: Catalog # 554657). Cells were blocked with TruStain FcX (anti-CD16/32) antibody (1:100; BioLegend: Catalog # 101320) and stained with LIVE/DEAD Fixable Blue Dead Cell Stain Kit (ThermoFisher Scientific: Catalog # L23105). Cells were then labelled with anti-CD11b-BV510 (1:100; BD Biosciences: Catalog # 562950), anti-CD45-APC/Cy7(1:200; BioLegend: Catalog # 103116), anti-CCR2-APC (1:100; R&D Systems: Catalog # FAB5538A-100), and anti-Cx3cr1-FITC (1:100; BioLegend: Catalog # 149019) antibodies in Stain Buffer. Each experiment contained a fluorescence minus one F(FMO) control and was analyzed on a Beckman Coulter CytoFLEX flow cytometer. Compensation was performed using VersaComp Antibody Capture Kit (Beckman Coulter: Catalog # B22804) according to manufacturer’s instructions.

In this study, we chose to isolate microglial cells from the brain using a magnetic bead enrichment method [62] over isolation by fluorescence activated cell sorting (FACS) due to the known ability of FACS to upregulate immediate early gene expression [68–70], induce transcription factors associated with proinflammatory states [68], and to alter cellular redox state [71]. However, an obvious caveat of our CD11b-based enrichment strategy is the single-antigen approach, which is likely to pull down immune cells present in the brain, not only microglia. Our findings that there were not sex differences in the number of CD11b⁺CD45^hi cells infiltrating into the brain following LPS suggest that any sex difference found in isolated cell biology would be due to microglia, but this is a limitation of our study. Future studies of microglial biology could be performed using FACS and newly-published transcriptional inhibitors to dampen some of the negative impacts of cell sorting [72].

Microglial oxygen consumption and glucose uptake assessment

Microglia were isolated from the ACC using CD11b bead enrichment as described above. To assess cellular oxygen consumption, isolated microglia were plated in 96 well plates at 50K cells/well and assessed by the MitoXpress Xtra Oxygen Consumption Assay (Agilent). Cells were plated in 96 well plates in 90 μL of assay medium consisting of DMEM without phenol red (ThermoFisher), L-glutamine (SigmaAldrich), PenStrep (SigmaAldrich), N2 media supplement (ThermoFisher), sodium pyruvate (ThermoFisher), and forskolin (SigmaAldrich). 10 μL of reconstituted MitoXpress reagent are added to each well, except for blank wells. Cells were then treated with 3 μM FCCP to stimulate maximal oxygen consumption or with vehicle control. Wells were then sealed with pre-warmed HS Mineral Oil. Oxygen consumption was then read on a Synergy Neo2 plate reader (BioTek) using dual TRF mode. To assess glucose uptake, isolated microglia were plated in white 96 well plates at 50K cells/well and assessed by the Glucose Uptake-Glo assay (Promega). Cells were plated in 96 well plates in 100 μL of PBS (ThermoFisher). Cells were then treated with 50 μL of 1 mM 2-DG and incubated at room temp for 10 min. 25 μL of Stop Buffer followed by 25 μL of Neutralization Buffer were added, followed by 100 μL of 2DG6P Detection Reagent. Wells were then incubated at room temperature for 1 hr, and then luminescence was detected using a BioTek Synergy 2 plate reader.

QUANTIFICATION AND STATISTICAL ANALYSES

Statistical analysis

All statistics and data visualization were performed using GraphPad Prism Version 9. Exact statistical tests performed for each experiment are reported in the figure legends and included in Supplementary Table 1. Individual points on graphs represent individual biological samples. Statistical significance was determined at p < 0.05.

RESULTS

Early-life immune challenge induces male-specific alterations in social behavior

To assess whether early-life LPS challenge resulted in sex-specific changes in social behavior, we injected female and male mice with LPS at postnatal day 9 (PN9), a time point previously established to induce long-term behavioral changes in male mice [52]. LPS injection decreased body weights at PN10 in both sexes compared to saline-injected controls, but group differences were no longer detectable by PN30 (Supplemental Fig. 1A). We assessed social preference starting at PN30 using a modified Crawley’s 3-chambered arena task [73] in which mice were allowed to investigate a novel age- and sex-matched conspecific in one of the side chambers, or a novel object (i.e. small rubber duck) in the opposite chamber (Fig. 1A). Consistent with the juvenile drive to seek out novel social experiences, saline-treated female and male mice both preferentially investigated a novel social stimulus compared to a novel object (Fig. 1B-C, Supplemental Fig. 1C-E). LPS treatment did not alter the time that females spent investigating either the social stimulus or the novel object. However, male mice treated on PN9 with LPS showed decreased social preference at PN30, with decreased social investigation and increased object investigation, and no longer preferred to investigate a novel animal over a novel object (Fig. 1B-C, Supplemental Fig. 1C-E). To measure the preference of a mouse to investigate a novel mouse or a familiar mouse (novelty preference), we then performed a similar 3-chambered task in which the subject mouse is placed in a 3-chambered arena and given the option to investigate either a new novel age- and sex-matched conspecific or a familiar cage mate animal (Fig. 1D). Consistent with alterations in sociability, saline-treated mice preferred to investigate the novel mouse over their cage-mate sibling, and LPS treatment did not alter social novelty preference in female mice. Treatment with LPS at PN9 decreased novel investigation, increased investigation of the familiar animal, and decreased social novelty preference in ~PN30 males (Fig. 1E-F, Supplemental Fig 1F-H). While early-life immune challenge induced male-biased alterations in several social behavior tasks, no changes were observed in the elevated zero maze (Fig. 1G), light dark box test (Fig. 1H), or in marble burying (Supplemental Fig. 1B), suggesting that changes in anxiety-like behaviors or repetitive behaviors do not explain deficits in social behaviors.

Figure 1. Perinatal immune challenge induces deficits in social behavior in male but not female mice.

A. PN30 female (orange) and male (blue) mice challenged with saline (Sal: open bars) or LPS (striped bars) at PN9 were placed in a 3-chambered arena and given the choice to interact with a novel conspecific mouse or an inanimate object for 5 min.

B. Time investigating the novel animal compared to novel object (% sociability) was assessed. Differences across groups were assessed by 2-way ANOVA followed by Bonferonni’s post-hoc analyses. **** p < 0.0001. # single sample t-test against 50% (chance), p < 0.05. Bar graphs display mean ± standard error of the mean. N = 7–11 per group.

C. Social investigation time and object investigation time was compared within saline-treated and LPS-treated female and male mice. Differences between social and object investigation time were assessed by paired t-tests. **** p < 0.0001. Bar graphs display mean ± standard error of the mean. N = 7–11 per group.

D. PN30 female (orange) and male (blue) mice challenged with saline (Sal: open bars) or LPS (striped bars) at PN9 were placed in a 3-chambered arena and given the choice to interact with a sex- and age-matched novel conspecific or a cage mate for 10 min.

E. Time spent investigating the novel animal compared to cage mate (% novelty preference) was assessed. Differences across groups were assessed by 2-way ANOVA followed by Bonferonni’s post-hoc analyses. ** p < 0.01. # single sample t-test against 50% (chance), p < 0.05. Bar graphs display mean ± standard error of the mean. N = 7–11 per group.

F. Novel investigation time and familiar investigation time was compared within saline-treated and LPS-treated female and male mice. Differences between novel and familiar investigation time were assessed by paired t-tests. **** p < 0.0001. Bar graphs display mean ± standard error of the mean. N = 7–11 per group.

G. ~PN30 mice were placed in an elevated zero maze for 5 min and time spent in closed arms was assessed. Differences across groups were assessed by 2-way ANOVA. Bar graphs display mean ± standard error of the mean. Figure was created using Biorender. N = 9–14 per group.

H. ~PN30 mice were placed in a light dark box for 10 min and time spent in the light zone was assessed. Differences across groups were assessed by 2-way ANOVA. Bar graphs display mean ± standard error of the mean. Figure was created using Biorender. N = 3–5 per group.

Social behavior differences are not caused by sex-biased monocyte infiltration

Given the known ability of LPS injection to induce trafficking of peripheral leukocytes into the brain [74], we next tested the hypothesis that sex differences in social behavior in response to LPS challenge are due to elevated infiltration of peripheral monocytes into the brain in males. We collected ACC tissue, based on the relevance of this brain region to social behavior and its disruption in disorders such as ASD, from female and male mice at 24 and at 72 hr following saline or LPS challenge and performed flow cytometry to determine the extent of brain infiltration. We defined microglia in the ACC as CD11b⁺CD45^lo cells, and infiltrating monocytes as CD11b⁺CD45^hi cells [75](Supplemental Fig. 2A). CD11b⁺CD45^hi cells express the chemokine receptor CCR2, whereas CD11b⁺CD45^lo cells do not express CCR2 (Supplemental Fig. 2B). CD11b⁺CD45^lo microglia also expressed the fractalkine receptor Cx3cr1, whereas CD11b⁺CD45^hiCCR2⁺ infiltrating monocytes did not express Cx3cr1 (Supplemental Fig. 2C). We observed no sex differences in Cx3cr1 expression within either CD11b⁺CD45^lo microglia or CD11b⁺CD45^hi infiltrating cells (Supplemental Fig. 2D). Consistent with previous reports, we observed that LPS challenge induced infiltration of peripheral monocytes into the brain 72 hr following injection, although we found negligible infiltration 24 hr following LPS injection and no change in microglial number at either time point (Fig. 2A-B, Supplemental Fig. 2E-F). In sum, there were no sex differences in brain infiltration of CD11b⁺CD45^hi cells (Fig. 2B), suggesting that the sex differences observed in social behavior following LPS injection are not caused by sex differences in the number of peripheral monocytes that infiltrate into the brain.

Figure 2. Male-biased social behavior deficits are dependent upon microglial TLR4.

A. Anterior cingulate cortex (ACC) from female and male mice injected on PN9 with saline or LPS and collected either 24 hr or 72 hr post-injection. ACC tissue was dissociated, brought to a single cell suspension, and cleaned with Miltenyi Debris Removal Solution prior to flow cytometric analysis. Tissue was analyzed for expression of CD11b/CD45. Cells were gated on forward scatter (FSC)/side scatter (SSC), singlets, live, CD11b, and CD45.

B. Brain cells from PN10 and PN12 male and female mice injected on PN9 with saline or LPS were analyzed for CD11b positivity and CD45 hi/lo status. The percentage of cells assessed to be CD11b⁺CD45^hi is presented. Graphs display mean ± standard error of the mean.

C. Brains were collected from PN10 mice that had been injected on PN9 with LPS or saline and endotoxin levels were assessed by kinetic limulus amebocyte lysate assay. Orange dots indicate brains from female mice, whereas blue dots indicate brains from male mice. Differences between groups were assessed by unpaired t-test. **** p < 0.0001. N = 6–7 per group.

D. Microglia (CD11b+ cells) and non-microglia (CD11b^- cells) were isolated from the ACC of TLR4flox/flox (Ctrl) and Cx3cr1-CreBT:TLR4flox/flox (TLR4 cKO) mice. qPCR for Tlr4 relative to 18S was performed to confirm knockdown of Tlr4 expression in Cx3cr1+ cells. Differences across groups were assessed by 2-way ANOVA followed by Bonferonni’s post-hoc analyses. Bar graphs display mean ± standard error of the mean. **** p < 0.0001. N = 6–8 per group.

E. Serum was extracted from PN10 TLR4flox/flox (Ctrl) and Cx3cr1-CreBT:TLR4flox/flox (TLR4 cKO) mice that had been injected on PN9 with LPS or saline. Serum concentration of the inflammatory cytokines CCL2, IL-1β, IL-6, and TNF-α were assessed by multiplexed SimplePlex cytokine assay. Orange dots indicate brains from female mice, whereas blue dots indicate brains from male mice. Differences across groups were assessed by 2-way ANOVA followed by Bonferonni’s post-hoc analyses. Bar graphs display mean ± standard error of the mean. * p < 0.05, ** p < 0.01, **** p < 0.0001. N = 3–5 per group.

F. PN30 female (orange) and male (blue) Ctrl and TLR4 cKO mice challenged with saline (Sal: open bars) or LPS (striped bars) at PN9 were placed in a 3-chambered arena and given the choice to interact with a novel conspecific mouse or an inanimate object for 5 min. Time investigating the novel animal compared to novel object (% sociability) was assessed. Differences across groups were assessed by 3-way ANOVA followed by Bonferonni’s post-hoc analyses. *** p < 0.001. # single sample t-test against 50% (chance), p < 0.05. Bar graphs display mean ± standard error of the mean. N = 4–10 per group.

G. PN30 female (orange) and male (blue) Ctrl and TLR4 cKO mice challenged with saline (Sal: open bars) or LPS (striped bars) at PN9 were placed in a 3-chambered arena and given the choice to interact with a sex- and age-matched novel conspecific or a cage mate for 10 min. Time spent investigating the novel animal compared to cage mate (% novelty preference) was assessed. Differences across groups were assessed by 3-way ANOVA followed by Bonferonni’s post-hoc analyses. * p < 0.05. # single sample t-test against 50% (chance), p < 0.05. Bar graphs display mean ± standard error of the mean. N = 4–10 per group.

Microglial TLR4 signaling is required for male-biased social behavioral alterations in response to early-life immune challenge

Considering the lack of sex difference in peripheral monocyte infiltration into the brain following PN9 LPS challenge, we next tested whether LPS could be directly acting on the brain. As the blood brain barrier is not fully formed by PN9 [76], we tested whether LPS could directly accumulate in the brain. Indeed, we found robust upregulation of endotoxin in the brain of PN10 mice injected with LPS at PN9 (Fig. 2C). Importantly, PN10 is prior to any infiltration of peripheral monocytes into the brain. Therefore, we then tested whether early-life immune challenge led to male-specific alterations in social behavior through microglial immune signaling. To assess this question, we bred mice in which the gene for toll-like receptor 4 (TLR4), a pattern recognition receptor that initiates inflammatory cascades in response to LPS, was flanked by loxP excision sites (floxed) with BAC transgenic Cx3cr1-CreBT mice to ablate TLR4 signaling within microglia throughout the brain [77] (Fig. 2D). We then administered a PN9 LPS challenge and assessed social behaviors in these mice lacking microglial TLR4. LPS injection caused an increase in the concentration of the inflammatory factors CCL2, IL-1β, IL-6, and TNF-α in the serum of Ctrl mice on PN10, but importantly the ablation of TLR4 signaling within microglia (TLR4 conditional knockout: TLR4 cKO) did not prevent this LPS-induced peripheral serum cytokine increase (Fig. 2E). However, microglial TLR4 ablation prevented the LPS-induced deficit in sociability and social novelty preference in male mice (Fig. 2F-G, Supplemental Fig. 2G-J), indicating that microglial immune signaling is necessary for the male-specific deficits in social behaviors induced by early-life immune challenge. Ablation of TLR4 signaling within microglia had no effect on sociability or social novelty preference in saline-treated female or male mice nor LPS-treated female mice.

Perinatal sex hormones organize male behavioral vulnerability to early-life immune challenge

Upon establishing that microglial immune signaling is necessary for the male-biased alterations in social behavior in response to PN9 LPS, we next set out to determine the mechanisms by which this early-life sex difference in microglial susceptibility is programmed. One well-characterized sex-specific developmental program that may underlie this male bias is a surge of gonadal hormones that occurs only in male mice during a perinatal critical period of brain development and organization. This gonadal hormone surge in which the fetal testis produces androgens in the latter third of gestation in rodents [26, 78] masculinizes the male body and brain in a microglia-dependent manner [27–29]. During early postnatal life, the female rodent brain is sensitive to gonadal hormone exposure, and can be diverted towards a male-like state (“masculinized”) through exogenous administration of the gonadal hormone testosterone or its aromatized form estradiol [26, 57, 58, 79, 80]. We predicted that this hormone-driven organization of the male brain imparts vulnerabilities to subsequent immune challenges. To test this hypothesis, we masculinized the female brain by injecting female mouse pups with estradiol on PN1 and PN2 (Fig. 3A). This allows us to determine whether brain organizational changes induced by the perinatal gonadal hormone surge impart male vulnerability to immune challenge independent of the chromosomal milieu. Female mice masculinized with estradiol at birth (F+E₂) and then injected on PN9 with saline showed typical social drive, preferring the novel social stimulus compared to the novel object, similar to saline-treated females and males. Consistent with our previous data, PN9 LPS induced deficits in sociability and social novelty preference in male but not female mice (Fig. 3B-C, Supplemental Fig. 3D-I). Strikingly, female mice masculinized at birth with estradiol and then challenged at PN9 with LPS showed similar deficits in sociability and social novelty preference as LPS-treated males (Fig. 3B-C, Supplemental Fig. 3D-I). We also assessed an additional social behavior using a juvenile social exploration assay in which PN15 pups were presented with home-cage bedding or novel (age- and treatment-matched) bedding, a test in which neurotypically developing mice prefer to investigate novel bedding [81](Supplemental Fig. 3A). We found that PN9 LPS interrupted the exploratory drive, and that masculinizing females at birth introduced a similar susceptibility (Supplemental Fig. 3B-C). These data reveal unique immune challenge-induced alterations in social behaviors that are organized by the perinatal gonadal hormone surge typical of male neurodevelopment.

Figure 3. Perinatal sex hormones induce social behavior vulnerability to early life immune challenge.

A. Female mice were masculinized on PN1–2 by injection with estradiol benzoate (E₂). Male, female, and masculinized female (F+E₂) mice were injected on PN9 with saline or LPS.

B. PN30 female (orange), male (blue), and masculinized female (F+E₂: purple) mice challenged with saline (Sal: open bars) or LPS (striped bars) at PN9 were placed in a 3-chambered arena and given the choice to interact with a novel conspecific mouse or an inanimate object for 5 min. Time investigating the novel animal compared to novel object (% sociability) was assessed. Differences across groups were assessed by 2-way ANOVA followed by Bonferonni’s post-hoc analyses. **** p < 0.0001. # single sample t-test against 50% (chance), p < 0.05. Bar graphs display mean ± standard error of the mean. N = 7–12 per group.

C. PN30 female (orange), male (blue), and masculinized female (F+E₂: purple) mice challenged with saline (Sal: open bars) or LPS (striped bars) at PN9 were placed in a 3-chambered arena and given the choice to interact with a sex- and age-matched novel conspecific or a cage mate for 10 min. Time spent investigating the novel animal compared to cage mate (% novel investigation) was assessed. Differences across groups were assessed by 2-way ANOVA followed by Bonferonni’s post-hoc analyses. *** p < 0.001. **** p < 0.0001. # single sample t-test against 50% (chance), p < 0.05. Bar graphs display mean ± standard error of the mean. N = 7–12 per group.

Early-life immune challenge induces long-term male-biased alterations in expression of mitochondrial electron transport chain genes

Our data demonstrate that early-life immune challenge causes male-specific, microglial TLR4-dependent social behavior alterations that are dependent on the perinatal gonadal hormone surge. We therefore determined what microglial functions were altered by the perinatal gonadal hormone surge that imparts male behavioral vulnerability to immune challenge. Given the known neuroimmune and mitochondrial abnormalities present in early-onset neurodevelopmental disorders, and the recently appreciated link between microglial mitochondrial function and social behavior revealed in male mice, we analyzed our previously published [25] bulk RNA sequencing of isolated microglia treated acutely with LPS for mitochondrial ETC gene expression. In this experiment, PN60 female and male mice were injected with LPS and 2 hr later microglial isolations were performed followed by RNA sequencing (Supplemental Fig. 4A). Sequencing revealed robust male-specific downregulation of mitochondrial ETC gene expression within CD11b+ cells (microglia)(Supplemental Fig. 4B). We next assessed whether this robust sex-specific transcriptomic alteration was simply a kinetic difference in microglial mitochondrial gene expression between the sexes, or whether PN9 LPS treatment would induce lasting male-specific alterations in mitochondrial gene expression on a timescale consistent with the changes in social behavior (i.e. at PN30). Microglia were isolated from dissected ACC of PN30 female, male, and masculinized female mice that had been injected on PN9 with saline or LPS and mitochondrial ETC gene expression was assessed by PCR Array (Fig. 4A). Similar to what we observed following acute immune challenge, we observed no decrease in the expression of mitochondrial ETC genes in females treated with LPS compared to saline-treated females (Fig. 4B). PN9 LPS challenge resulted in dramatic decreases in mitochondrial ETC gene expression in microglia isolated from PN30 males compared to saline-treated males (Fig. 4B). Microglia from LPS-treated masculinized females also significantly downregulated mitochondrial gene expression compared to saline-treated mice (Fig. 4B). Interestingly, the majority of mitochondrial ETC genes that remained downregulated by PN30 encode proteins in Complex I, as 94% of Complex I genes were significantly downregulated in LPS-treated male mice, and 91% of these same genes were downregulated in LPS-treated masculinized female microglia (Fig. 4B). Importantly, this mitochondrial vulnerability to perinatal immune challenge appears to be specific to microglia, as there was no significant LPS-induced change in gene expression in CD11b-cells (Supplemental Fig. 4C-D).

Figure 4. Perinatal immune challenge leads to lasting changes in expression of microglial mitochondrial electron transport chain genes in male and masculinized female - but not in female - mice.

A. Experimental timeline. Female, male, or females masculinized on PN1–2 with estradiol benzoate were injected on PN9 with LPS or saline vehicle. At PN30, anterior cingulate cortex was dissected, and microglia were isolated by CD11b bead method. Extracted RNA from CD11b+ cells (microglia) was then assessed for mitochondrial ETC gene expression using Qiagen RT2 Mitochondrial Energy Metabolism PCR Array.

B. Log of mRNA expression of ETC subunits (Complexes I-V: CI-CV) from CD11b+ cells isolated from ACC of PN30 mice injected with LPS on PN9 relative to that gene’s expression from saline-treated mice (red = upregulation, white = no change, blue = downregulation). F+E₂ = masculinized female.

Differences across groups were assessed by 2-way ANOVA followed by Bonferonni’s post-hoc analyses. * p < 0.05 for male LPS compared to male saline. # p < 0.05 for F+E₂ LPS compared to F+E2 saline. There were no LPS-induced differences in female mice. N = 3–4 per group.

Perinatal immune challenge induces lasting male-biased alterations in microglial mitochondrial morphology and function

Given the striking downregulation of mitochondrial ETC genes in microglia from male and masculinized female mice in response to early-life immune challenge, we next assessed whether this early-life immune challenge would result in sex-biased alterations in microglial mitochondrial function. Changes in mitochondrial architecture through the opposing processes of fission and fusion impact mitochondrial bioenergetic functions [82–85]. We first assessed mitochondrial morphology within microglia in the ACC. To do so, we stained for the outer mitochondrial membrane protein Tomm20 and the microglial marker ionized calcium binding adaptor molecule 1 (Iba1) from PN30 mice challenged on PN9 with saline or LPS (Fig. 5A). We used volumetric reconstruction software to mask mitochondrial signal within and outside of microglia, and assessed morphology of these reconstructed mitochondria ([67] and Supplemental Fig. 5A). In females, microglial mitochondrial length did not differ regardless of early life treatment. However, PN9 LPS significantly decreased mitochondrial length within male microglia compared to saline-treated males (Fig. 5B-C, Supplemental Fig. Fig. 5B). Microglial mitochondria within the PN30 ACC were also shorter in masculinized females challenged with LPS at PN9 (Fig. 5B-C, Supplemental Fig. 5B). Similar results were found for mitochondrial volume measures, as PN9 LPS treatment led to smaller microglial mitochondria within males and masculinized females, but not in females (Fig. 5D-E, Supplemental Fig. 5C).

Figure 5. Sex differences in microglial mitochondrial morphometric response to perinatal immune challenge.

A. Representative images of microglia (Iba1; red), mitochondria (Tomm20; white) within microglia, and a merged image in the anterior cingulate cortex (ACC) of PN30 mice. Representative images for female, male, and masculinized female (F+E₂) mice treated at PN9 with saline or LPS are shown.

B. Cumulative frequency plots of microglial mitochondrial length analyzed in the ACC of female (orange), male (blue), and masculinized female (F+E₂: purple) mice. Differences across groups were assessed by Kruskal-Wallis test followed by Dunn’s post-hoc analyses. * p < 0.05.

C. Average mitochondrial length within microglia in the ACC of female (orange), male (blue), and masculinized female (F+E₂: purple) mice. Differences across groups were assessed by 2-way ANOVA followed by Bonferonni’s post-hoc analyses. **** p < 0.0001 for male LPS compared to male saline. # p < 0.05 compared to female saline. Bar graphs display mean ± standard error of the mean. N = 4–8 per group.

D. Cumulative frequency plots of microglial mitochondrial volume analyzed in the ACC of female (orange), male (blue), and masculinized female (F+E₂: purple) mice. Differences across groups were assessed by Kruskal-Wallis test followed by Dunn’s post-hoc analyses. * p < 0.05.

E. Average mitochondrial volume within microglia in the ACC of female (orange), male (blue), and masculinized female (F+E₂: purple) mice. Differences across groups were assessed by 2-way ANOVA followed by Bonferonni’s post-hoc analyses. * p < 0.05 ** p < 0.01 for male LPS compared to male saline. # p < 0.05 compared to female saline. Bar graphs display mean ± standard error of the mean. N = 4–8 per group.

Mitochondria form dynamic, interconnected networks that are vital to the energetic function of the cell [86, 87]. In addition to morphometric analyses of individual microglial mitochondrial morphologies, we also assessed the ability of early-life immune challenge to affect connectivity of microglial mitochondrial networks. Mitochondrial connectivity analyses led to similar observations as were seen in morphometric analyses, as PN9 LPS induced a decrease in the number of branches/cell and number of networks/cell in both males and masculinized females, while having no effect on these connectivity measures in female mice (Supplemental Fig. 5D-F). No group differences in mitochondrial endpoints were observed in non-microglia (Iba1 negative cells; Supplemental Fig. 5G-H).

Mitochondrial morphologies are often closely aligned with cellular bioenergetic function [38, 82–84]. To analyze microglial mitochondrial energetics, we first isolated microglia from the ACC of PN10 mice treated 24 hr before with saline or LPS and assessed their oxygen consumption and glucose uptake capacities (Fig. 6A). We assessed cellular oxygen consumption of isolated microglia at baseline as well as following treatment with the protonophore carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP) to stimulate maximal oxygen consumption. There were no significant differences observed in baseline (untreated) oxygen consumption rates (OCR) between microglia from saline or LPS-treated mice, regardless of sex (Fig. 6B). In all females as well as saline-treated males and F+E₂ mice, the FCCP-induced maximal OCR was significantly elevated from basal OCR (Fig. 6B). However, FCCP failed to stimulate a significant upregulation in LPS-treated male and F+E₂ microglia (Fig. 6B). Importantly, FCCP-stimulated maximal OCR was significantly lower in LPS-treated males and F+E₂ mice compared to saline-treated males, saline-treated F+E₂ mice, and all female groups (Fig. 6B). Accordingly, the spare respiratory capacity [88] is significantly impaired in LPS-treated male and F+E₂ mice compared to saline-treated males, saline-treated F+E₂ mice, and all female groups (Fig. 6C). Similarly, we observed a significant LPS-induced stimulation in glucose uptake from male and F+E₂ microglia compared to saline-treated microglia (Fig. 6D). To visualize microglial oxidative and glycolytic energetics, we plotted maximal OCR against glucose uptake, and noted that PN9 LPS treatment does not substantially alter the metabolic phenotype of ACC microglia isolated from females, whereas LPS shifted male and F+E₂ microglia towards a more glycolytic phenotype (Fig. 6E).

Figure 6. Perinatal immune challenge leads to lasting alterations in mitochondrial function in microglia from male - but not female- mice.

A. Experimental timeline. Female, male, or masculinized female (F+E₂) mice were injected on PN9 with LPS or saline vehicle. At PN10, anterior cingulate cortex (ACC) was dissected, and microglia were isolated by CD11b bead method. CD11b+ cells (microglia) were plated in 96 well plates and cellular oxygen consumption was assessed using Mito Xpress Xtra Oxygen Consumption Assay. Cells were either untreated or maximal oxygen consumption was stimulated by treatment with FCCP prior to sealing of the well. Extracellular acidification rate (ECAR) was assessed using the pH-Xtra Glycolysis Assay. Glucose uptake was assessed using Glucose Uptake-Glo assay.

B. Baseline and maximal oxygen consumption rates were assessed. Individual dots refer to oxygen consumption rate from distinct mice. Differences across groups were assessed by 3-way ANOVA followed by Bonferonni’s post-hoc analyses. ** p < 0.01, **** p < 0.0001. Bar graphs display mean ± standard error of the mean. N = 4–8 per group.

C. Spare respiratory capacity was calculated by subtracting the untreated oxygen consumption rate from the FCCP-treated oxygen consumption rate for an individual mouse. Individual dots refer to spare respiratory capacity calculated from distinct mice. Differences across groups were assessed by 2-way ANOVA followed by Bonferonni’s post-hoc analyses. *p < 0.05, *** p < 0.001. Bar graphs display mean ± standard error of the mean. N = 4–8 per group.

D. Glucose Uptake was assessed using the Glucose Uptake-Glo Assay and reported as mM of 2DG6P uptake. Individual dots refer to glucose uptake from distinct mice. Differences across groups were assessed by 2-way ANOVA followed by Bonferonni’s post-hoc analyses. **** p < 0.0001. Bar graphs display mean ± standard error of the mean. N = 5–10 per group.

E. Maximal oxygen consumption was plotted against Glucose uptake to depict cellular bioenergetic state. Data are presented as mean +/− Standard deviation. N = 5–8 per group.

DISCUSSION

Here we demonstrate the impact of sex on social behavior and microglial mitochondrial responses to early-life immune activation and point to the critical role of gonadal hormones during perinatal development in these alterations. We show that early life immune activation induces alterations in social behaviors in male but not female mice, and are dependent upon microglial TLR4 signaling. We also demonstrate that early life immune activation reveals male-biased vulnerabilities in mitochondrial gene expression, mitochondrial morphologies, mitochondrial network connectivity, and bioenergetic functions, specifically within microglia. These sex-biased alterations in microglial mitochondrial functions may explain the sex-biased vulnerabilities of social behaviors in response to perinatal immune challenge. Importantly, we induced these behavioral and mitochondrial vulnerabilities to early-life immune activation in chromosomally female mice by masculinizing female pups at birth with estradiol, the aromatized form of gonadally-derived testosterone that can masculinize female brain and behavior [29, 31, 79]. Collectively, these findings provide new evidence that sex-specific perinatal brain development is heavily influenced by gonadal hormones, and that this hormone-dependent development organizes male-biased susceptibility to inflammatory perturbations in the perinatal environment.

We initially hypothesized that early-life immune activation induces male-biased deficits in social behavior through male-biased recruitment of peripheral inflammatory monocytes (CD11b⁺CD45^hi cells) into the brain. Indeed, numerous studies have demonstrated that perturbations such as LPS injection or stress induce infiltration of peripheral immune cells into the brain, where these inflammatory cells act to induce behavioral changes [89–92]. However, these studies often fail to examine whether sex differences in infiltration may occur. In contrast to our hypothesis, we demonstrated no sex difference in the infiltration of peripherally derived monocytes into the brain, as PN9 LPS injection induced a robust increase in infiltration into both male and female brains. Importantly, we observed this infiltration at 72 hr, but not at 24 hr, the same PN10 time point at which we observed mitochondrial oxygen consumption and glucose uptake alterations in isolated microglia. As the blood brain barrier is not yet fully formed at the PN9 injection timepoint used in this study, we next asked whether injected LPS could reach the brain, thereby directly acting on microglia. Indeed, we found a robust increase in brain endotoxin levels at PN10 following PN9 LPS injection, prior to any brain infiltration of peripheral monocytes. The lack of observed sex differences in peripheral immune infiltration and the robust increase in brain endotoxin levels next led us to assess whether immune signaling within microglia was necessary for the sex-biased behavioral alterations. To do so, we used mice in which TLR4, a pattern recognition receptor to which LPS binds, was ablated within microglia [77]. Importantly, ablation of TLR4 signaling within microglia did not impact sociability or social novelty preference in the absence of an immune challenge. Knocking out TLR4 signaling within Cx3cr1+ cells prevented LPS-induced alterations, demonstrating the necessity of direct microglial immune signaling in this sex-biased behavioral response. The Cre-LoxP transgenic technology used in this study ablated TLR4 in all Cx3cr1-expressing cells, i.e. tissue resident macrophages. That the infiltrating monocytes observed following early life immune activation do not express Cx3cr1 further suggests that infiltrating monocytes were not contributing to the male-biased behavioral vulnerability. Importantly, we have previously demonstrated that the Cx3cr1-CreBT line that we used to ablate microglial TLR4 labels tissue resident microglia but not inflammatory monocytes [93]. Additionally, multiplexed cytokine arrays suggested that TLR4 cKO did not alter the peripheral cytokine response to PN9 LPS injection. However, future studies using more precise genetic ablation strategies will be important to further discern the impact of peripheral vs central macrophage immune signaling in the regulation of social behaviors. Given the in vitro findings that microglia respond differently to bacterial vs viral infection [94], whether our findings of cell-specific sex-biased mitochondrial alterations and behavioral changes are specific to bacterial infection or are more generalizable to maternal immune activation via viral infections remains an important unanswered question.

Activation of TLR4 by LPS results in robust changes in the morphology and function of mitochondria within microglia, and microglial TLR4-mediated alterations in neuronal functions require alterations in microglial mitochondrial metabolism [38, 39]. We therefore next assessed whether early-life immune activation could be leading to the male bias in social behavioral deficits through sex-biased alterations in microglial mitochondria. Consistent with a growing body of literature demonstrating critical links between microglial inflammatory processes and mitochondrial functions [38–40], we observed robust male-specific alterations in microglial mitochondria. For instance, we demonstrated that LPS treatment induced a strong downregulation of mitochondrial ETC genes specifically in microglia from male mice. LPS treatment did not similarly downregulate ETC genes in female microglia, although masculinizing female mice at birth with male-typical gonadal hormones conferred a similar susceptibility to immune activation in ETC genes. This programming of vulnerability appears to be cell-type specific, as CD11b^- cells did not show alterations in mitochondrial ETC gene expression in response to LPS treatment. Importantly, several RNA sequencing databases demonstrate that mouse brain microglia express estrogen receptors such as Esr1 and Gper1/Gpr30 [25, 95, 96], demonstrating that gonadally-derived hormones indeed have the potential for direct action on microglia. In addition to gene expression, we demonstrate robust sex- and microglia-specific alterations in mitochondrial morphology that are induced by PN9 LPS challenge. Importantly, we demonstrate that the developmental gonadal hormone surge that is necessary for typical male masculinization of the body and brain also results in male microglial mitochondrial morphological vulnerability to early-life immune activation. Together, these results suggest that mitochondria specifically within microglia are particularly vulnerable to immune activation in male mice, and that this vulnerability is organized by the perinatal surge in gonadal hormones that is necessary to typically masculinize the male brain and body. However, it is important to note that we have not depleted the perinatal testosterone surge in male mice as it would require a technically difficult embryonic depletion. Future experiments aimed at further dissecting testosterone’s role in male mice, such as utilization of the four-core genotypes model [97] will further refine our understanding of gonadal hormone interplay with male vulnerability.

Our findings identify microglia and mitochondria as potentially important therapeutic targets in the treatment of social alterations in neurodevelopmental disorders. This aligns with previous work in other mouse models as well as postmortem analyses of brains obtained from patients diagnosed with neurodevelopmental disorders such as ASD and SZ that demonstrate altered microglial morphologies and mitochondrial alterations [98–101]. Of particular interest is our finding that the sex-biased downregulation of mitochondrial ETC genes induced by LPS was somewhat specific for genes encoding Complex I of the mitochondrial ETC. Intriguingly, a positron emission tomography (PET) study of a radioligand that binds to Complex I demonstrated decreased Complex I availability specifically in the ACC of males diagnosed with ASD [48]. Complex I is especially vulnerable to impairment by oxidative stress and is demonstrated to be dysfunctional in a wide variety of diseases. Due to the relative susceptibility of Complex I, particular attention has been paid to the development or repurposing of treatments that target Complex I functions. Our findings raise the possibility that more targeted mitochondrial drugs that can bypass dysfunction in Complex I, such as the electron donor idebenone [102], may be particularly useful for the treatment of neurodevelopmental disorders. Indeed, establishing whether social behavioral deficits induced by PN9 LPS challenge can be reversed by the rescue of microglial mitochondrial function remains an outstanding question and will surely be tackled in future studies. Conversely, definitively demonstrating that induction of microglia-specific mitochondrial dysfunction can lead to social behavioral deficits similar to those induced by PN9 LPS challenge will prove highly informative. The recent success of mitochondrial transplantation to restore mitochondrial function across disease and injury models may prove vital to the treatment of sex-biased neurodevelopmental disorders with mitochondrial dysfunction. However, brain region-and cell type-specific targeting of mitochondrial function, through either direct mitochondrial transplantation or other therapeutic regimens directly in the brain, will surely remain difficult in human patients for the foreseeable future. Other indirect avenues to impact brain mitochondrial function, particularly within microglia, will likely prove vital to treatment and prevention of behavioral symptoms associated with neurodevelopmental disorders such as ASD.

While our study demonstrates potential risk factors inducing male vulnerability and/or female protection, it must be noted that not all males subjected to perinatal immune challenges such as infections go on to develop neurodevelopmental disorders [103, 104], and that some females indeed go on to develop neurodevelopmental disorders [105]. These realities suggest that potential male vulnerability factors may act similarly to ‘priming events’, thereby increasing susceptibility to other risk factors [106, 107]. For example, rodent ‘two-hit’ models of perinatal immune challenge (such as maternal immune activation) that are coupled with perinatal or postnatal stressors often induce robust behavioral alterations in offspring [108–110].

In summary, perinatal hormone exposure has been associated with increased risk of offspring neurodevelopmental disorder diagnosis [33, 34], but the mechanisms through which hormone exposure may program male behavioral vulnerability, or provide protection in females, are scarce. We demonstrate that the male-specific perinatal hormone surge that is essential for masculinizing the male brain introduces vulnerability to social behavior deficits in response to early-life immune activation via microglial TLR4 signaling. Within microglia, we demonstrate that the mitochondrial functional vulnerability to immune challenge, consistent with previous findings, is also particularly vulnerable and highly specific to male mice. This male-biased behavioral and microglial mitochondrial vulnerability is programmed by the male-typical perinatal gonadal hormone surge.

Limitations of the study

We report that PN9 LPS challenge induced a robust increase in the infiltration of peripheral immune cells into the brains of both male and female mice. The lack of sex difference in infiltrating cell number does not rule out the possibility that the infiltrating cells may act differently between males and females. For instance, infiltrating cells may induce behavioral deficits in male brains whereas infiltrating immune cells could be providing a female protective factor against immune challenge. Additionally, we did not assess whether there were sex differences in the infiltration of other immune cells such as T cells, which have previously been shown to impact microglial function as well as mitochondrial function following their migration into the brain [111]. Therefore, a deeper understanding of immune dysregulation in the brain, both in terms of type of infiltrating cell as well as functionality of these cells following perinatal perturbations is needed in both females and males.

Interestingly, recent studies using a mouse model in which the mitochondrial protein UCP2 was ablated specifically within microglia demonstrated male-specific alterations in synaptic architecture and neural functions [112], suggesting a link between microglial mitochondrial function and synaptic regulation. These data build upon other findings that mitochondrial dysfunction and altered metabolism play a critical role in the regulation of social behaviors [43–45], suggesting that altered microglial mitochondrial function may regulate the establishment of complex social behaviors through alterations in synaptic regulation. Our previous studies have demonstrated that microglial phagocytosis regulates the development of synaptic circuitry important for social behavior in the ACC during the time period surrounding our PN9 injection [108]. Given the importance of the ACC to regulation of social behaviors, and it’s implication in the etiology of various neurodevelopmental disorders [63, 64, 108], we assessed microglial mitochondrial function within the ACC. While our studies revealed robust alterations in microglial mitochondrial function in the ACC, this study did not comprehensively characterize mitochondrial function from every region that may control these complex behaviors. Future studies characterizing immunometabolic responses to inflammatory challenges, and whether these microglial mitochondrial alterations influence synaptic pruning/phagocytosis by microglia, across all brain regions, will be highly informative.

Highlights.

• Early-life immune activation induces male-biased alterations in social behavior

• Male-biased behavioral alterations are dependent on microglial TLR4 signaling

• Immune activation induces male-biased microglia-specific changes in mitochondria

• Vulnerability to early-life immune activation is programmed by gonadal hormones