Sex Differences in Adipose Tissue Distribution Determine Susceptibility to Neuroinflammation in Mice With Dietary Obesity

Adipocyte overload drives inflammation in obesity, and visceral fat from males is more susceptible to lipid overload and cellular stress (1–3). Adult females preferentially store energy in subcutaneous adipose tissue (SAT) and are less susceptible to metabolic and cardiovascular comorbidities in obesity (3–5). However, sexually dimorphic vulnerability to obesity-induced pathology is unlikely to occur as a function of SAT mass because SAT removal does not influence glycemic control in obese females (6,7). Sex differences are frequently attributed to estrogens, based on visceral adipose tissue (VAT) hypertrophy and susceptibility to insulin resistance after hysterectomy in women or ovariectomy in rodents (3–5). However, the consequences of rapid-onset hormonal deprivation do not always recapitulate the effects of gradual dysregulation in chronic diseases, and data on the relative progression of metabolic, reproductive, and immunological dysfunction are scarce in preclinical models of obesity.

Sexually dimorphic immunoregulation in obesity involves differential exposure to humoral factors, changes in precursor cell proliferation and differentiation, and local interactions between principal cells and tissue-resident lymphocytes (1–5,8). These multilayered signaling events govern local inflammation in VAT and SAT (5), but the degree of direct communication between the two depots remains uncertain. Communication does occur indirectly via the well-characterized regulation of cells in the central nervous system (CNS) by adipocyte-derived hormones and cytokines (9). Neural regulation of VAT and SAT occurs via sympathetic nerves, which are in turn controlled by neurons in the hypothalamus and brainstem (10). Neuronal activity in the CNS and periphery is locally modulated by nonneuronal cells, including macrophages at sympathetic terminals and microglia in the brain (10–12). Hypothalamic microglia in males, but not females, exhibit morphological activation with dietary obesity, and this dimorphism underlies female resistance to obesity-induced metabolic dysfunction (12–14). Outside of regulatory circuits, hippocampal microglia are also activated by adipose-derived cytokines in males with dietary obesity (13), but the impact of biological sex remains elusive.

We carried out parallel analyses of metabolic, immune, and reproductive functions in intact mice to address gaps in knowledge surrounding progressive impairment along each arm of the triad in obesity. After observing correlations between developmental differences in adipose tissue distribution and subsequent vulnerability to inflammation in obese mice, we surgically removed SAT before beginning the high-fat diet (HFD) to determine whether these were causal relationships. The outcome of this study implicates sex differences in adipose tissue distribution as lasting determinants of neuroimmune responses to obesity in females.

C57BL/6J mice were purchased from the Jackson Laboratory and were bred in house for this study. Mice were housed on a ventilated rack with ad libitum food and water on a 12-h light/dark schedule (lights on at 6:00 a.m.). For time course experiments, mice were pseudorandomized into weight-balanced groups and switched from Teklad chow (cat. no. 2918) to the low-fat diet (LFD) (10% fat; cat. no. D12450; Research Diets, Inc.) or the HFD (60% fat; cat. no. D12492; Research Diets, Inc.) at 8 weeks old. In follow-up experiments, 8-week-old mice underwent lipectomy (LPX) or sham surgery (SHAM). Animals were maintained on chow during 2-week postsurgical recovery before being switched to the LFD or HFD for 12 weeks. Body weights were determined weekly, and food intake was quantified on 2 consecutive days/week. All procedures followed the National Institutes of Health guidelines and were approved by the Institutional Animal Care and Use Committee of Augusta University.

Surgical removal of dorsolumbar SAT (dlSAT) was carried out as described (15) with modifications. Incisions were made bilaterally on each flank, rostral to the expected position of dlSAT, and shifted caudally for access. Sterile saline was applied inside of each incision, and the closed tip of a hemostat was used to separate dlSAT from superficial dermal fascia. The dlSAT pad was then grasped using blunt forceps and excised without damaging the underlying muscle or vasculature. Nerves innervating SAT were visible as white fiber tracts with the aid of a wearable loupe and fiber optic light. Innervation of inguinal SAT (iSAT) was intact based on visual inspection, and the inguinal lymph node was left undisturbed. SHAM involved identical incisions and manipulation of dlSAT before closure with Vicryl sutures and topical antibiotic application. Mice were treated with an analgesic carprofen (10 mg/kg s.c.) and monitored daily for 1 week after surgery for signs of distress. None of the mice met exclusion criteria (weight loss >15% below presurgical level or dehiscence at incision site).

Longitudinal analysis of adipose tissue was performed using T1-weighted MRI. Mice were anesthetized with 2.0% isoflurane and maintained at 36 ± 1°C throughout imaging. Respiratory signals were detected with a physiological monitoring system (model no. 1025; SA Instruments, Inc., Stony Brook, NY), and isoflurane was adjusted ±0.5% to maintain respiration at 45–70 breaths/min. Imaging was performed on a 7.0T/20-cm horizontal bore with a Bruker Avance spectrometer (Bruker BioSpin Corporation, Billerica, MA) using a Doty 63-mm quadrature volume resonator in the Small Animal Imaging Core at Augusta University. After scout scans for animal positioning, multislice axial images were acquired using a moderate T1-weighted, fast-spin, echo sequence. The MRI sequence included 15-/1,300-ms echo time/repetition time, rapid imaging with refocused echoes factor of 4, 10 averages, 36 × 36 mm field of view, 1.0-mm slice thickness, and 256 × 256 matrix size. A frequency-selective Gaussian saturation pulse of 1,050 Hz in bandwidth centered at the water carrier frequency was presented before the fast-spin echo train for water signal suppression to enhance contrast between adipose and other tissues (16). Images were imported into Reconstruct software (17) for manual segmentation and three-dimensional rendering of SAT and VAT. To control for sex differences in animal size, the volume of SAT and VAT was expressed relative to trunk volume (delineated by tracing the external border of the torso on each image). Adipose imaging by DXA was performed as described (18).

Vaginal cytology was determined by lavage twice daily for 7 days. To induce cycling, bedding from male mice (0.5 g) was introduced into female cages 3 days before starting lavage. The time of day was recorded at sample collection and used to estimate cycle duration. Samples were collected onto spot slides (Ted Pella) using a small, flexible transfer pipette (tip diameter 3.0 mm; Thermo Fisher Scientific). Slides were stained with Toluidine blue before imaging. Images were captured at 20× and analyzed independently by 2 blinded experimenters. Mice that were acyclic, defined by persistence in a single stage or irregular cycle sequences, were excluded from cycle-length estimates.

Intraperitoneal glucose (IPGTT) and insulin tolerance tests (IPITTs) were performed as described (13,19,20).

Serum was separated from whole blood and stored at −80°C until analysis, as described (19). Concentrations of progesterone, testosterone, and corticosterone were quantified by liquid chromatography–tandem mass spectroscopy (LC-MS/MS), with concentration determined relative to standard curve. Circulating estradiol was quantified using a commercial kit (Calbiotech, Inc.) according to manufacturer instructions. For ELISA, hemolysis was quantified by reading absorbance at 414 nm on a plate reader (Molecular Devices), and samples with absorbance within 1 SD of whole blood were excluded. Concentrations of proinflammatory cytokines were determined in serum samples using the Meso Scale Discovery platform. Each sample was assayed in duplicate by the Multiplex Immunoassay Core at Emory University using Mouse Proinflammatory Panel 1 (Meso Scale Discovery). Serum lipid profiles were also determined in duplicate by the University of Cincinnati Medical Center Mouse Metabolic Phenotyping Center using commercially available assays. Insulin (cat. no. 90080; Crystal Chem, Inc.), leptin (cat. no. 90030; Crystal Chem, Inc.), and adiponectin (cat. no. MRP300; R&D Systems) levels were quantified using commercial ELISA kits according to manufacturer instructions.

Forebrain mononuclear cells (FMCs) were isolated by density gradient centrifugation, as described (13,20). Separation of stromal-vascular fraction (SVF) and whole-blood lysis also followed published methods (13). For flow cytometry, cells were labeled with conjugated primary antibodies, and positive labeling was determined based on isotype controls, single-labeled tubes, and unlabeled cells (Supplementary Table 1 lists antibody information). Intensity thresholds for phenotypic antigens (e.g., CD45, Ly6C) were determined using FMCs spiked with freshly dissociated splenocytes, as reported (13,20). Data were acquired on a five-laser LSRII (BD Biosciences) and analyzed using FlowJo (version 10.0; Treestar).

For immunofluorescence, mice were sacrificed by transcardial perfusion with 4% paraformaldehyde, and serial sections were prepared as described (13,20). Immunodetection of IBA1 (cat. no. 019-19741; FUJIFILM Wako Chemicals) and CD169 (cat. no. MCA884; Bio-Rad Laboratories) followed published protocols, and nuclei were counterstained with DAPI (Thermo Fisher Scientific). Labeled cells were imaged in the mediobasal hypothalamus (MBH) or dentate gyrus (DG) on a Zeiss LSM780 confocal microscope. Reconstruction was carried out in Neurolucida 360 (MBF Bioscience, Williston, VT) on z stacks of 5–7 cells per mouse. Process lengths and Sholl profiles were calculated on a per-animal basis for statistical comparisons as reported (13,20).

Paraffin embedding, sectioning, and hematoxylin-eosin staining in adipose tissue were carried out with assistance from the Histology Core at Augusta University. Immunoperoxidase detection of F4/80 (cat. no. sc25830; Santa Cruz Biotechnology) and frequency-based analysis followed published methods (20).

For longitudinal experiments, data were analyzed using repeated-measures ANOVA with Tukey post hoc. For cross-sectional time course experiments, data were analyzed using 2 × 2 × 2 ANOVA (sex × diet × time point) followed by planned comparisons with separate 2 × 2 ANOVAs (sex × diet) and Tukey post hoc at each time point. The effect of diet on sex steroids was determined using 2 × 2 ANOVA (diet × cycle) in females or with bidirectional t tests in males. For the LPX experiments, data were analyzed using 2 × 2 × 2 ANOVA (sex × diet × surgery) followed post hoc Bonferroni-corrected t tests. Analysis was carried out in Graphpad Prism (version 9.0), with significance at P < 0.05.

Data sets generated in these studies are available from the corresponding author upon reasonable request. All the critical resources used in these experiments are commercially available, and detailed protocols are available upon request.

Reciprocal interactions between metabolism, immunity, and reproduction are widespread in the literature (1–5), but progressive dysfunction in each system remains incompletely characterized in obesity. We investigated these processes in a series of longitudinal and cross-sectional time course studies. For longitudinal analysis of adipose tissue expansion, 8-week-old male and female C57BL/6J mice were maintained on the HFD for 24 weeks, with MRI imaging every 6 weeks (Fig. 1A and B). Females had proportionally less VAT than males before starting the HFD (Fig. 1B and C), as quantified by manual segmentation and three-dimensional reconstruction of water-corrected MRI images (F_4,24 = 14.90; P < 0.01) (Fig. 1B). Sex differences in VAT persisted at 6 weeks but were no longer detectable after 12 weeks (Fig. 1C). The proportional volume of SAT did not differ between males and females (Fig. 1D), but VAT-to-SAT ratios followed sexually dimorphic trajectories over the course of the study (Fig. 1E). Males had higher VAT-to-SAT ratios before the HFD (F_4,24 = 16.61; P < 0.01) (Fig. 1E). However, females rapidly accrued additional VAT, based on significant increases after 18 and 24 weeks on the HFD (Fig. 1E and F).

Converging trajectories of VAT accumulation were upheld by dissection of SAT and perigonadal VAT at sacrifice (Fig. 1G). Although there was no significant difference in tissue weights (Fig. 1G), histological analysis revealed smaller adipocytes in VAT from females (t₁₉ = 4.69; P < 0.01) (Fig. 1H). Positive correlations between adipocyte size and adipose tissue weight indirectly reflect hyperplasia-driven expansion (21). VAT weight and adipocyte area were positively correlated in males (Fig. 1I), but not in females. This pattern was unique to VAT, because SAT weights were not associated with adipocyte diameter in either sex (Fig. 1J). Taken together, these patterns suggest that metabolically healthy adipose tissue expansion persists after the onset of male-typical adipose tissue distribution in females with dietary obesity.

To evaluate sex differences in glycemic control, IPGTT was performed after 12, 24, or 48 weeks on the LFD or the HFD (Fig. 2A). Males and females exhibited similar weight gain after normalization to starting body weight (Fig. 2B). There was no effect of sex or diet on glucose tolerance at 12 weeks (Fig. 2C). After 24 weeks on the HFD, males exhibited fasting hyperglycemia (F_1,50 = 3.90; P < 0.05) (Supplementary Fig. 1A), but females did not (Fig. 2D and Supplementary Fig. 1A). The opposite pattern was evident after 48 weeks on the HFD, with elevated peak glucose concentrations and slower recovery 60 min after glucose challenge in females (F_1,50 = 4.26; P < 0.01) (Fig. 2E). Females exhibited improved glucose tolerance after 24 weeks on the HFD, based on area under the curve (AUC) (F_1,50 = 3.59; P < 0.01) (Fig. 2F), but improvements were no longer evident after 48 weeks on the HFD (Fig. 2F). However, females on the HFD maintained lower-fasting insulin concentrations than males throughout the experiment (F_1,50 = 6.64; P < 0.01) (Supplementary Fig. 1B).

To examine sex differences in the progression of insulin resistance, nonoverlapping groups of mice received IPITTs, as shown (Fig. 2A). There were no differences after 12 weeks on the HFD (Fig. 2G). Females exhibited milder reductions in insulin sensitivity than males after 24 weeks on the HFD, as reflected by lower glucose concentrations 15 min after injection (F_1,26 = 7.04; P < 0.05) (Fig. 2H). After 48 weeks on the HFD, females were no longer significantly different from males (Fig. 2I), based on glucose concentrations and AUC (Fig. 2J). Overall, these data indicate that females exhibit delayed-onset insulin resistance and glucose intolerance on HFD.

Estrogen plays a critical role in female resistance to obesity-induced metabolic dysfunction (2–5,14). To measure cycling, mice received lavage twice daily for 1 week (Fig. 3A). Stage duration was unaffected by diet (Fig. 3B), and acyclicity was similar in females on theLFD and the HFD (n acyclic/n per condition = 2/24 and = 3/25 for the LFD and HFD, respectively). Quantification of sex steroids by LC-MS/MS at sacrifice revealed cyclic fluctuations in serum progesterone (Fig. 3C). Progesterone concentrations were elevated in metestrus, relative to other stages (Fig. 3C). Serum progesterone was significantly higher in females (F_1,67 = 4.59; P < 0.05) (Fig. 3C), with no effect of diet in either sex. The limit of detection for LC-MS/MS quantification of progesterone was 0.5 ng/mL, and 6 male samples fell below that threshold (LFD n = 2; HFD n = 4).

The HFD reduced testosterone in males (t₁₅ = 3.80; P < 0.01) (Fig. 3D). The opposite pattern was detected in females, where the HFD was accompanied by increased testosterone during estrus and metestrus (F_3,27 = 4.45; P < 0.05) (Fig. 3D). For testosterone, 9 LFD, female samples fell below the limit of detection (0.09 ng/mL) and were excluded. Circulating estradiol was not detectable in males, but quantitation in females revealed cyclic elevations during proestrus and estrus (Fig. 3E). There was no effect on diet, consistent with vaginal cytology and LC-MS/MS analysis of progesterone. Taken together, these observations indicate that females cycle normally despite increased testosterone concentrations after 24 weeks on the HFD.

After 48 weeks (Fig. 3F), most females were acyclic, as reflected by irregular staging and/or persistent epithelial-cell cornification on vaginal smears (n acyclic/n per condition = 19/25 and 20/24 for the LFD and the HFD, respectively). Although acyclicity prevented blood collection at specific stages, estradiol concentrations were numerically lower after 48 weeks (Fig. 3G–I), relative to the 24-week cohort (Fig. 3E). Serum progesterone levels after 48 weeks (Fig. 3G) were within the lower range observed after 24 weeks for each sex (Fig. 3C). Circulating testosterone trended higher (P = 0.06) in females on the HFD for 48 weeks, relative to sex- and age-matched LFD mice (Fig. 3H). By contrast, testosterone was significantly reduced after 48 weeks in HFD in males, relative to males on the LFD (t₁₈ = 3.25; P < 0.01) (Fig. 3H). The acyclicity evident after 48 weeks on experimental diets, when mice are 13 months old, is consistent with published reports (22), as is the confluence of lower estradiol, elevated testosterone, and lower progesterone levels (23). Taken together, these patterns indicate that females are in reproductive senescence at the 48-week time point. When interpreted relative to changes in glycemic control (Fig. 2), these results indicate that glucoregulatory deficits emerge after reproductive senescence in females on the HFD.

After observing sex differences in adipose tissue accumulation and sexually dimorphic loss of glycemic control, we sought additional insight into the progression and immunological characteristics of these processes. To this end, VAT and SAT were collected after 12, 24, or 48 weeks on the LFD or HFD. Fat pads were dissected and weighed before fixation and histological analysis or SVF isolation, as described (20). Males accumulated more VAT than females on the HFD (F_2,132 = 4.52; P < 0.01) (Fig. 4A) and had larger VAT adipocytes than females after 12 or 24 weeks on a HFD (F_1,60 = 14.80; P < 0.01) (Supplementary Fig. 2A–C). However, sex differences were no longer evident at 48 weeks (Supplementary Fig. 2A–C), when females had stopped cycling (Fig. 3G–I). Females accumulated more SAT than males after 12 weeks on the HFD, but differences at later time points were not statistically significant (F_1,132 = 20.83; P < 0.01) (Fig. 4B). Adipocyte sizes in SAT were similarly increased in males and females on the HFD (Supplementary Fig. 2D–F). VAT-to-SAT ratios were initially higher in males after 12 weeks on the HFD (F_1,132 = 5.63; P < 0.01) (Fig. 4C and D). Males on the LFD maintained higher VAT-to-SAT ratios than females, but this dimorphism was lost on the HFD, likely because of expansion of SAT in males at later time points (Fig. 4B–D).

We next quantified immune cell recruitment and polarization after 12, 24, or 48 weeks on the LFD or the HFD. Adipose tissue macrophages (ATMs) were identified as Ly6G⁻/CD11c⁺/F4/80⁺ events, as shown (Fig. 4E). Proinflammatory (M1) ATMs were identified as CD68^HI/CD206⁻, and anti-inflammatory (M2) ATMs were gated as CD68^LO/CD206⁺ (Fig. 4E). Males had more macrophages in VAT than females on the HFD at all time points (F_1,34 = 24.27; P < 0.01) (Fig. 4F). Visceral ATMs from male mice on the HFD also adopted a proinflammatory phenotype, indicated by increases in M1 (F_1,34 = 36.29; P < 0.01) (Fig. 4G) and stability of the M2 population relative to age- and sex-matched LFD mice (Fig. 4H). Proinflammatory polarization of VAT macrophages was delayed and attenuated in females on the HFD, based on the absence of within-sex differences after 12 weeks on the HFD and on reductions in M1 ATMs at later time points (F_1,34 = 6.28; P < 0.01) (Fig. 4G).

Obesity-induced macrophage accumulation occurred at lower levels in SAT, relative to VAT, but sex differences followed similar patterns (F_1,34 = 12.74; P < 0.01) (Fig. 4I). Macrophage polarization also occurred at lower rates in SAT, relative to VAT, but significant increases in M1 ATMs were evident in males on the HFD, relative to sex-matched LFD mice (F_1,34 = 6.84; P < 0.01) (Fig. 4J). Resistance to obesity-induced macrophage polarization in SAT from females was partially explained by lower frequency of M1 macrophages at 12 and 24 weeks (F_1,34 = 4.23; P < 0.05) (Fig. 4J). However, SAT from females also contained more M2 ATMs after 24 weeks on the HFD (F_1,34 = 6.02; P < 0.05) (Fig. 4K).

Consistent with previous reports (24), males had more circulating monocytes than females on HFDs (F_1,35 = 22.51; P < 0.01) (Fig. 4L) and exhibited earlier expansion of the M1 population (F_1,42 = 41.90; P < 0.01) (Fig. 4M). Males also exhibited earlier increases in M2 monocytes on the HFD (F_1,44 = 7.04; P < 0.05) (Fig. 4N). Differences in circulating monocytes and ATMs were associated with sexually dimorphic accumulation of F4/80⁺ crown-like structures (Supplementary Fig. 3A and B). Crown-like structures were less prevalent in VAT from females, as reflected by lower frequency on systemic random sampling images (F_1,30 = 16.29; P < 0.01) (Supplementary Fig. 3A and B). When interpreted relative to changes in adipose tissue weight and adipocyte morphology (Fig. 4A–C and Supplementary Fig. 2), these data suggest that sex differences in adipose tissue expansion are driven by changes in local inflammation.

Resident microglia both regulate and respond to obesity-associated pathophysiological signals in the brain (11–13). For insight into the progression of CNS inflammation, FMCs were isolated from males and females after increasing durations on the LFD or HFD. Bone marrow–derived macrophages gain access to brain parenchyma with obesity (13,20,25). Therefore, we used CD169 as a retained marker of peripheral origin and TMEM119 as a marker for microglia (Fig. 5A) (26,27). These analyses revealed sexually dimorphic vulnerability to macrophage accumulation in the CNS with obesity (Fig. 5B). After 12 weeks on the HFD, male mice exhibited increases in CD45^HI/Ly6C^HI/CD169⁺/TMEM119⁻ macrophages (F_1,36 = 9.89; P < 0.01), whereas females did not differ from sex-matched, LFD mice (Fig. 5B). Macrophage infiltration was delayed, but not eliminated, because males and females exhibited comparable increases in gated macrophages after 24 or 48 weeks on the HFD (Fig. 5B).

Entry of peripheral lymphocyte populations into the brain occurs in multiple neuropathological conditions, but there is also evidence that bone marrow–derived immune cells promote resolution after CNS injury (28). To determine whether macrophage accumulation was associated with microglial activation, we quantified TLR4 and MHCII in CD11b⁺/CD45^LO/Ly6C^LO/CD169⁻/TMEM119⁺ microglia. Microglia from obese males upregulated both TLR4 (F_1,36 = 18.30; P < 0.01) (Fig. 5C) and MHCII (F_1,36 = 33.13; P < 0.01) (Fig. 5D) at all time points, indicative of persistent proinflammatory activation. Microglial upregulation of TLR4 was delayed in females on the HFD and attenuated relative to males on the HFD at each time point (Fig. 5C). Similar trends were observed for MHCII, with delayed-onset induction in obese females, relative to age- and sex-matched LFD controls, and reductions relative to obese males (Fig. 5D).

Morphological activation of microglia has been reported in the MBH and hippocampal DG of males, but not females, on the HFD (11–13,20,25). We therefore performed colabeling for IBA1, expressed by microglia and macrophages, and CD169 on brain sections (Fig. 4E and G). In the MBH, IBA1 immunoreactivity was elevated in males and females on the HFD at all time points (F_1,60 = 24.80; P < 0.01) (Fig. 5E). Likewise, accumulation of CD169⁺/IBA1⁺ cells was detected at all time points in the MBH, independent of sex (F_1,60 = 43.93; P < 0.01) (Fig. 5F). In the hippocampus, females exhibited delayed onset microgliosis on HFD, relative to males (F_1,60 = 12.00; P < 0.05) (Fig. 5G). CD169⁺/IBA1⁺ colabeled cells were rare in the hippocampus, and there was no effect of diet, sex, or duration (Fig. 5H). Taken together, these patterns suggest that microgliosis is widespread, whereas macrophage infiltration is regionally heterogeneous, in males and females with dietary obesity.

The longitudinal (Fig. 1) and cross-sectional data sets (Figs. 2–4) revealed correlations between VAT-to-SAT ratios before the HFD and susceptibility to inflammation after extended periods on the HFD. We therefore hypothesized that pre-HFD differences in adipose tissue distribution confer-lasting protection against inflammation in obesity. To test this hypothesis, groups of female and male C57BL/6J mice underwent surgical LPX of SAT at 8 weeks old. Surgeries involved excision and removal of dlSAT, sparing iSAT (Supplementary Fig. 4A). This strategy was chosen because complete removal induces ectopic lipid deposition, and sparing iSAT would enable subsequent regeneration from local precursors (29,30). Body weights at surgery did not differ between conditions (Supplementary Fig. 4B), and similar amounts of SAT were removed from males and females (Supplementary Fig. 4C and D). After surgery, mice were fed chow for 2 weeks before being switched to the LFD or the HFD for 12 weeks (Fig. 6A). Weight gain was unaffected by LPX (Fig. 6A), and DXA revealed similar increases in total fat mass on the HFD(F_1,52 = 9.47; P < 0.01) (Fig. 6B). Accumulation of lean mass was also comparable in males and females on the HFDs (F_1,52 = 13.36; P < 0.01) (Fig. 6C). There were no changes in food intake in males (mean ± SEM: LFD/SHAM 10.54 ± 2.65, HFD/SHAM 13.12 ± 2.21, LFD/LPX 11.38 ± 3.05, and HFD/LPX 12.15 ± 3.38 kcal/day) or females (LFD/SHAM 9.14 ± 2.73, HFD/SHAM 10.55 ± 2.42, LFD/LPX 9.72 ± 2.13, and HFD/LPX 10.72 ± 2.66 kcal/day).

Complete removal of SAT impairs glucose tolerance in male mice on the HFD (29,31). We carried out IPGTT 10 weeks after surgery to determine whether partial LPX would have similar effects. There were no differences in glycemic control based on circulating glucose and AUC (Fig. 6D), and there were no differences in fasting insulin (Fig. 6E). Given reported hyperlipidemia after complete removal of SAT in obese mice (29), we also measured circulating lipids after partial LPX (Fig. 6F–I). There was no effect of sex or surgery on diet-induced elevations in cholesterol (F_1,69 = 17.27; P < 0.01) (Fig. 6F) and total phospholipids (F_1,69 = 14.74; P < 0.01) (Fig. 6G). Diet-induced elevations in triglycerides were evident in males (F_1,69 = 10.70; P < 0.01), but not females, and sex differences were unaffected by LPX (Fig. 6H). Nonesterified fatty acid concentrations were significantly higher in males (F_1,69 = 11.76; P < 0.01), but there was no effect of LPX in either sex (Fig. 6I). Although we cannot rule out the possibility of tissue-specific alterations in lipid metabolism, these data indicate that the impact of sex and diet on circulating lipids was unaffected by LPX.

Because of the intimate relationship between adiposity and reproduction (2–5), we considered the possibility that LPX might alter estrus cycling and/or steroid hormones. Vaginal cytology was determined after 10 weeks on the HFD or the LFD, as shown (Fig. 5A). Blind inspection of vaginal cytology revealed no effect of diet or surgery on cycling (Supplementary Fig. 5A), indicating that differences in reproductive hormones were unlikely. Consistent with this interpretation, analysis of serum progesterone at sacrifice revealed sex differences (F_1,79 = 6.01; P < 0.05) (Supplementary Fig. 5B) but no effect of diet or surgery. Serum estradiol concentrations were also unaffected (Supplementary Fig. 5C). Males on the HFD had significantly lower testosterone levels, relative to male LFD/SHAM mice, irrespective of surgery (F_1,26 = 24.65; P < 0.01) (Supplementary Fig. 5D). Similar reductions in testosterone were evident after 24 weeks on the HFD in surgically naive mice (Fig. 3D), consistent with pervasive antiandrogenic effects of obesity in males. Testosterone concentrations were higher in HFD/SHAM and HFD/LPX females, but there were no statistically significant differences, relative to LFD/SHAM mice. For additional insight into steroidogenesis in males and females, we also quantified corticosterone levels. There was no significant effect of sex, diet, or surgery (Supplementary Fig. 5E), indicating that LPX does not alter the impact of the HFD on steroid hormones.

After observing recovery of total body fat by DXA (Fig. 6B), mice were sacrificed for dissection and analysis of SAT and VAT weights. Although there was no effect of LPX before the LFD, LPX reduced SAT weights in both sexes after the HFD (F_1,58 = 20.08; P < 0.01) (Fig. 7A). LPX induced diet-independent hypertrophy in VAT from females (F_1,54 = 5.12; P < 0.05) and eliminated sex differences on the HFD (Fig. 7B). There was no effect of LPX on adipocyte areas in SAT from mice on either diet (Supplementary Fig. 6A–C). By contrast, adipocyte areas were significantly larger in VAT from females that underwent LPX before the HFD, relative to female HFD/SHAM mice (F_1,27 = 13.25; P < 0.01) (Supplementary Fig. 6D–F). In males, HFD-induced adipocyte hypertrophy was unaffected by LPX (Supplementary Fig. 6D). Surgical removal of SAT also reduced serum leptin in males on the HFD (F_1,40 = 6.88; P < 0.05) but had no effect in females (Fig. 7C). Females had higher circulating adiponectin concentrations than males (F_1,29 = 11.57; P < 0.01), and there was no effect of diet or surgery (Fig. 7D). Taken together, these patterns suggest that surgical removal of SAT is followed by diet-dependent redistribution of energy storage in females, but not males.

To examine adipose tissue inflammation, we next analyzed the frequency of F4/80⁺ crown-like structures after LPX. There were no differences in SAT (Fig. 7E), but crown-like structures were more frequent in VAT from HFD/LPX females, relative to female HFD/SHAM mice (F_1,32 = 12.06; P < 0.01) (Fig. 7F). By contrast, the frequency of crown-like structures was similarly elevated in males on the HFD, irrespective of surgery (Fig. 7F). Sexually dimorphic accumulation of crown-like structures in VAT was accompanied by increased circulating tumor necrosis factor-α (TNF-α) concentrations in HFD/LPX females, relative to sex-matched HFD/SHAM mice (F_1,36 = 13.65; P < 0.01) (Fig. 7G). In males, LPX reduced but did not eliminate obesity-induced increases in TNF-α (Fig. 7G). Similar patterns were observed for interleukin-1β (IL-1β), which increased with dietary obesity in males but remained lower in HFD/LPX males, relative to HFD/SHAM mice (F_1,35 = 15.86; P < 0.01) (Fig. 7H). Obesity-induced elevations in IL-1β were absent in SHAM females (Fig. 7H) but were comparably elevated in HFD/LPX females, relative to HFD/SHAM males (Fig. 7H). These trends were not ubiquitous because there was no effect of sex, surgery, or diet on circulating IL-6 (Fig. 7I). We also quantified circulating IL-10 to examine potential changes in anti-inflammatory cytokines. Dietary obesity increased serum IL-10 in females, and LPX eliminated this effect (F_1,37 = 15.47; P < 0.01) (Fig. 7J), but there was no effect of diet or surgery in males. Overall, these outcomes suggest that redistribution of energy storage into VAT after LPX amplifies proinflammatory humoral responses in obesity.

To determine whether sex differences in adipose tissue distribution regulate neuroinflammation in obesity, FMCs were isolated and gated as shown (Fig. 5A). Similar to surgically naive males (Fig. 5B), HFD/SHAM males exhibited accumulation of CD11b⁺/CD45^HI/Ly6C^HI/TMEM119⁻/CD169⁺ macrophages (F_1,40 = 12.70; P < 0.01) (Fig. 8A). Macrophage accumulation was unaffected by LPX in male mice (Fig. 8A). Unlike males, SHAM females did not exhibit macrophage infiltration on the HFD (Fig. 8A). However, the size of the gated macrophage population was significantly increased in HFD/LPX females, relative to all other sex-matched conditions (F_1,40 = 6.25; P < 0.05) (Fig. 8A). In obese males, peripheral macrophage infiltration was accompanied by microglial TLR4 induction, irrespective of surgery (F_1,40 = 10.88; P < 0.01) (Fig. 8B and C). By contrast, microglial TLR4 expression in HFD/SHAM females was comparable to that in sex-matched LFD/SHAM mice (Fig. 8B and C). Protection against microglial TLR4 induction was eliminated by LPX in females (F_1,40 = 14.15; P < 0.01) (Fig. 8B and C), indicating that a priori sex differences in SAT exert lasting effects on susceptibility to neuroinflammation in obesity.

After observing sexually dimorphic changes in IL-10 after LPX (Fig. 7J), we investigated whether microglia from females also upregulate anti-inflammatory signaling. There was no effect of diet, sex, or surgery on microglial IL-10Rα (Fig. 8D), but upregulation of IL-4Rα was observed in HFD/SHAM females (F_1,40 = 9.97; P < 0.01) (Fig. 8E and F). Microglial IL-4Rα induction was eliminated by LPX in obese females (Fig. 8E and F), suggesting that sex differences in adipose tissue distribution may promote anti-inflammatory polarization. This interpretation was upheld by morphological analysis of IBA1⁺ cells in the hippocampus (Fig. 8G). In males, the HFD reduced morphological complexity, based on Sholl analysis and process length (F_1,23 = 8.38; P < 0.01) (Fig. 8H–J), with no effect of LPX. In females, reductions in complexity and process length were only evident after the HFD/LPX (F_1,23 = 11.90; P < 0.01) (Fig. 8I and J), indicating that protection was SAT dependent. Taken together, the divergent patterns observed using flow cytometry and immunofluorescence support the existence of heterogeneous programs governing neuroinflammation in males and females.

These results indicate that sex differences in adipose tissue distribution confer lasting protection against inflammation in obesity. Converting female-specific patterns of subcutaneous and visceral adiposity into male-typical patterns before the HFD challenge unmasked vulnerability to inflammation in females. These effects were independent of changes in estrogen or progesterone, but sexually dimorphic changes in testosterone were evident after extended durations on the HFD. Increased serum testosterone in females, and reduced levels in males, emerged as signatures of progressive metabolic dysfunction and inflammation, consistent with reciprocal interactions at each arm of the immuno-endocrine-metabolic triad with dietary obesity.

Sexually dimorphic organization of adipose tissue is driven by gonadal steroids, based on increased VAT accumulation after ovariectomy in rodents or hysterectomy in women (3,5). The effects of androgens are more complex, with tissue- and sex-specific regulation of white fat occurring indirectly after aromatization to estradiol or as a downstream consequence of changes in lean mass (3,5,32). Consistent with the reductions in endogenous testosterone observed in males on the HFD, castration or whole-body androgen receptor ablation also increases visceral adiposity in males (32–35). Androgens also directly regulate adipogenesis because administration of dihydrotestosterone, which is not converted to estradiol, suppresses HFD-associated adipocyte precursor proliferation in castrated males (35). Age-related reductions in testosterone are associated with increased rates of obesity and insulin resistance in intact males (3,5,32). The effects of obesity on circulating testosterone in females were also age dependent in this study, based on elevations after 24 weeks on the HFD in surgically naive females and on the lack of significant differences between LFD/SHAM and HFD/SHAM females after 12 weeks. Although functional analysis of reproduction was not performed in this study, cyclic variability in progesterone and estradiol was unaffected by diet. Other groups demonstrated intact fertility in female C57BL/6J mice after 20 weeks on the HFD (36), which contrasts with reports of subfertility in obese females against a mixed background (37,38). Taken together, these results suggest that adipose tissue–driven protection against inflammation in females is unlikely to occur as a simple linear function of circulating reproductive steroids. However, it is hoped that direct quantitation of reproductive hormones in the current report will move the field toward standard inclusion of these measures, with the goal of identifying circulating profiles that predict pathology in obesity.

The outcomes observed after LPX could reflect pre-HFD differences in adipose tissue distribution, alone or in concert with downstream changes in VAT or dysfunctional regeneration of SAT after a HFD. Although more work remains to be done to understand cellular regeneration in SAT after LPX, adipocyte sizes were unaffected in females on either diet, and crown-like structures in SAT were also unaffected by diet or surgery. This does not rule out changes in local inflammation over time; however, the LPX experiments did not use dynamic measures. Future studies using intravital microscopy or bioluminescent imaging will be required to understand changes in immune cell trafficking and polarization as functions of sex, diet, and LPX. These effects are likely pleiotropic and may involve multiple tissues in addition to SAT and VAT. However, the fact that changes in local, humoral, and CNS inflammation were initiated by surgical reductions in SAT before the HFD demonstrates that sex differences in adipose tissue distribution underlie these effects (which may be directly or indirectly mediated).

Consistent with previous work (2–5), lean females exhibited sexually dimorphic adipose tissue distribution, indicated by lower VAT-to-SAT ratios. Adipose tissue expansion eventually reached parity with males on the HFD, but females maintained a relatively immunoquiescent microenvironment, as reported by others (2,24). There was no clear role of sex steroids, but differential exposure to X-linked immunoregulatory genes could underlie reduced susceptibility to adipose inflammation. Silencing of duplicate alleles on the X chromosome is mediated by long noncoding RNAs and chromatin modification during embryogenesis in females, but inactivation occurs heterogeneously in different cell types and tissues (39). Leaky inactivation occurs prominently in brain and spleen, and breakthrough expression of previously inactivated alleles is a feature of pathological conditions (8,39,40). The potential role of sex chromosomes is indirectly supported by cell transfer studies in males and females on HFDs (2,24). Hematopoietic stem cells from obese males repopulate ATMs more readily than cells from obese females (24), and adipocyte precursors from males proliferate more rapidly after engraftment in SAT than from females on the HFD (2). The cell-autonomous nature of these effects, and their independence from circulating sex steroids (2,24), implicates a pervasive, sex-linked mechanism.

We and others have shown that VAT transplantation promotes neuroinflammation in male mice (13,41). Because LPX recapitulated features of male VAT in obese females, visceral adiposity–induced neuroinflammation may occur via similar mechanisms. Female mice were protected against increases in IL-1β and TNF-α on the HFD, and protection was eliminated by LPX. IL-1β and TNF-α enter the CNS via distinct transporters at the blood-brain barrier (42). Both transporters saturate at picomolar levels, but receptor activation triggers local amplification by parenchymal microglia (13,42). Although LPX also reduced IL-10 in these studies, kinetic studies revealed no influx of circulating IL-10 into the brain (43). Given the observed sex differences and SAT-dependent fluctuations in circulating cytokines, and the availability of transporters for blood-to-brain influx (42), IL-1β and TNF-α represent potential mechanisms linking inflammation in brain and adipose tissue. Interestingly, several genes in the IL-1β and TNF-α signaling pathways are located on the X chromosome and exhibit leaky inactivation in brain and spleen (40). Sex differences in exposure to positive and negative regulators of IL-1β and TNF-α could orchestrate dimorphic responses to obesity-induced inflammation, but more work is necessary to address this possibility.

The intracellular signaling domain of IL-1R1 is shared by TLRs, including TLR4 (44), which was upregulated on resident microglia in obese males, relative to females, in this study. Sexually dimorphic recruitment of TLR4 was SAT dependent because LPX resulted in male-specific induction of TLR4 in microglia and brain-penetrant macrophages. Enhanced recruitment of TLR4 in obese, male mice aligns with recent proteomic studies reporting greater enrichment of TLR proteins in microglia from lean, young adult males (45), and the anti-inflammatory effects of pre-HFD sex differences in SAT are also reminiscent of the persistent neuroprotection reported after transplantation of female microglia into male brains (46). The consequences of TLR and IL-1R1 activation on microglia and macrophages are broadly proinflammatory, whereas TNF-α exerts pro- and anti-inflammatory effects mediated by TNF receptor 1 (TNFR1) and TNFR2, respectively (47). The enrichment of TNFR isotypes and IL-1 coreceptors positions microglia (and other CNS-adjacent macrophages) (27) as sensors and potential amplifiers after low-level penetration of circulating TNF-α and IL-1β across the blood-brain barrier. Further work will be required to address whether the organizational effects of sex steroids govern local vulnerability to inflammation in specific brain regions, and to determine whether long-range immunological signaling between fat and brain underlies sexually dimorphic vulnerability to inflammation in obesity.

Acknowledgments. The authors thank Dr. Suneil Koliwad (University of California San Francisco) for critical reading of the manuscript.

Funding. This research was supported by grants R01DK110586 and R01DK110586S1 from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) to A.M.S. and P01AG036675 from the National Institute on Aging to C.M.I. Circulating lipid assays were performed by the University of Cincinnati Medical Center Mouse Metabolic Phenotyping Core with support from grant U2CDK059630 from NIDDK.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.M.S., D.-H.G., M.Y., C.M.H., H.K., B.B., W.Z., Y.L., and K.D. carried out experiments. Y.L., X.L., K.D., and C.M.I. provided essential input and resources. All authors provided input on the manuscript. A.M.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.