We first examined whether the ER stress response/UPR pathway was altered in the adipose tissues of animals with obesity. In C57BL/6 J mice fed high fat diet for 14 weeks, we observed that the mRNA levels of genes representing all three branches of UPR [spliced Xbp1 (Xbp1s) for IRE1α branch, Atf4 and Ddit3 for PERK branch, and Edem1 and Hspa5 for ATF6 branch] were up-regulated in both epididymal white adipose tissues (eWAT) (Figure 1A) and inguinal white adipose tissue (iWAT) (Figure 1—figure supplement 1A). As IRE1α activation is associated with the IRE1-dependent decay of mRNA (RIDD) activity in which IRE1 cleaves mRNAs (Hollien and Weissman, 2006), we analyzed the mRNA levels of two RIDD targets Blos1 and Col6a1 by qRT-PCR. We observed that Blos1 and Col6a1 mRNA levels were decreased in the eWAT (Figure 1B) while Col6a1 mRNA level was also down-regulated in the iWAT (Figure 1—figure supplement 1B) of HFD-fed mice. These results suggest the over-nutrition/obesity triggers ER stress and IRE1α activity in adipose tissue.

We then asked whether IRE1α activation contributes to obesity-induced insulin resistance and importantly whether pharmacological inhibition of IRE1α can reverse the pre-existing obesity-induced insulin resistance. We used STF-083010 (STF), a member of hydroxy-aryl-aldehyde class of IRE1α RNase inhibitors that engage the RNase-active site of IRE1a with high affinity and specificity by exploiting a shallow complementary pocket through pi-stacking interactions with His910 and Phe889 and an essential Schiff base interaction between the aldehyde moiety of the inhibitor and the side chain amino group of Lys907 (Sanches et al., 2014), to treat diet-induced obesity (DIO) mice with established insulin resistance. Male C57BL/6 J mice fed on HFD for 8 weeks gained significant glucose intolerance (Figure 1—figure supplement 1C) and were treated with STF at 10 mg/kg of BW once daily for 4 weeks (Figure 1—figure supplement 1D) via IP injection, a dose previously shown to significantly inhibits IRE1α RNase activity in vivo (Papandreou et al., 2011; Tufanli et al., 2017; Herlea-Pana et al., 2021), while still on HFD. We observed that STF treatment improved glucose tolerance with lower peak glucose level and decreased AOC (area of the curve) in response to a bolus of exogenous glucose compared to their vehicle-treated counterpart (Figure 1C–C’). STF treatment significantly improved insulin sensitivity in DIO mice (Figure 1D–D’). We also observed markedly decreased resting and glucose-stimulated plasma levels of insulin in DIO mice treated with STF (Figure 1E), indicative of increased insulin sensitivity in the STF-treated mice. To investigate whether the increased insulin sensitivity by STF is reflected in insulin signaling, we examined the phosphorylation status of Akt, an effector of insulin signaling pathway, in the adipose tissue. Contrary to the significant enhancement in the insulin-stimulated Akt phosphorylation at Ser473 observed in the epididymal white adipose tissue (eWAT) of non-HFD lean animals, the insulin-stimulated Akt phosphorylation was blunted in the eWAT of DIO mice. However, STF treatment significantly reversed the insulin-stimulated Akt phosphorylation (Figure 1F–F’). Collectively, these data indicate that STF improves insulin sensitivity and glucose control under HFD.

To investigate whether the STF amelioration of insulin resistance and obesity reflects on its inhibition of IRE1α activity in adipose tissue, we examined the status of IRE1α-mediated Xbp1 mRNA splicing in WAT from STF-treated DIO mice and observed that as assessed by RT-PCR followed by electrophoretic separation, the level of spliced Xbp1 (Xbp1s) mRNA was significantly reduced in both eWAT (Figure 1G–G’) and iWAT (Figure 1—figure supplement 1E-E’’) from STF-treated animals relative to vehicle group. Moreover, the mRNA levels of Xbp1 target genes Hspa5 and Edem1 were highly suppressed in both eWAT and iWAT from STF-treated mice (Figure 1H, Figure 1—figure supplement 1F). Interestingly, we also observed the decrease in the mRNA levels of Atf4 and Ddit3, key components of the PERK pathway in eWAT and iWAT of STF-treated DIO mice (Figure 1H, Figure 1—figure supplement 1F). The effect of STF treatment on PERK pathway in the adipose tissue probably reflects the positive feedback loops among the branches of UPR under in vivo conditions as previously reported (Tsuru et al., 2013; Márton et al., 2017; Kapuy et al., 2020).

We observed that while vehicle-treated mice continued to gain weight on HFD, 4 week STF treatment slightly decreased the body weight (Figure 2A and Figure 4—figure supplement 1A), with no apparent food intake difference between the vehicle and STF groups (Figure 4—figure supplement 1B). Body composition analysis using EchoMRI also revealed that STF treatment resulted in a significant decrease in the fat mass but without significant changes in the lean mass (Figure 2B and C) in the HFD-fed mice. We next analyzed the size of adipocytes in WAT sections and observed that adipocytes were significantly smaller from STF-treated DIO mice than that of vehicle-treated ones (Figure 2C and F) as shown in both average diameters (Figure 2D,G) and frequencies in large adipocytes (Figure 2E and H) in both eWAT and inguinal WAT (iWAT). The average area of adipocytes was similarly significantly shrunk in the eWAT (Figure 4—figure supplement 1D) and iWAT (Figure 4—figure supplement 1E) of DIO mice treated with STF relative to vehicle treatment. In addition, there was a significant reduction in lipid droplets in the BAT from STF-treated DIO mice (Figure 2I–J).

A common comorbidity of obesity and insulin resistance is metabolic dysfunction-associated steatotic liver disease (MASLD; Khan et al., 2019). We therefore investigated the potential effects of STF on liver steatosis in obesity. First, we examined the liver histology in haematoxylin and eosin-(H.E.) stained sections. We observed the presence of numerous lipid droplets in the liver, a hallmark of live steatosis, in the vehicle-treated DIO mice compared to that in the BL/6 mice fed normal chow (Figure 3A). Strikingly, STF treatment markedly reduced the number of the lipid droplets in the liver of DIO mice compared to vehicle treatment (Figure 3A and A’). We next examined the expression of genes involved in lipogenesis in the liver of the STF- and vehicle-treated DIO mice and found that the lipogenic genes FASN, stearoly-coA desaturase-1 (Scd1), and acetyl coA-carboxylase (Acc) were down-regulated in the liver of STF-treated DIO mice compared to vehicle-treated mice (Figure 3B). Furthermore, because MASLD is associated with an impaired suppression of hepatic glucose output (Marchesini et al., 2001), we assessed the expression of genes involved in gluconeogenesis, including glucose-6-phosphatase catalytic subunit (G6pc1) and phosphoenolpyruvate carboxykinase (Pepck). We found that the expression level of G6pc1 in the liver from STF-treated DIO mice was also significantly decreased compared to that of vehicle-treated ones while there was no effect on Pepck expression (Figure 3C). MASLD is also characterized by excessive inflammation, in particular macrophage accumulation, in addition to lipid deposition (Lefere and Tacke, 2019). We therefore investigated the effect of STF on macrophage population in the obese liver and observed that STF treatment suppressed the macrophage accumulation of DIO mice, as marked by the F4/80 immunofluorescent staining (Figure 3D–E). Furthermore, obesity is tightly associated with dyslipidemia. We therefore examined the effect of STF on dyslipidemia and observed that STF treatment decreased the plasma levels of triglyceride, free fatty acid (FFA), and total cholesterol (Figure 3F–H).

The observation that STF treatment protects against HFD-induced obesity (Figure 2A) suggests an increase in energy expenditure (EE) in these mice. We therefore investigated this in DIO mice treated with STF. We observed that average oxygen consumption (VO₂) was higher in DIO mice treated with STF than with vehicle, using indirect calorimetry (Figure 4A–B and Figure 4—figure supplement 1F-F’). Similarly, the average carbon dioxide production (VCO₂) was significantly increased in STF-treated DIO mice (Figure 4C–D and Figure 4—figure supplement 1G-G’). As a result, we found that STF treatment moderately but significantly increased mean (EE) compared to vehicle treatment (Figure 4E–F and Figure 4—figure supplement 1H-H’), without alterations in respiratory exchange ratio (RER) value or physical activities (Figure 4G–H).

To ask whether the energy expenditure increase by STF is attributable to thermogenic heat production, we investigated the effect of STF treatment on the activation status of brown adipose tissue (BAT) in DIO animals. BAT is an organ specialized for energy expenditure through the action of mitochondrial uncoupling protein 1 (UCP1). During obesity, lipid-droplet content is increased in BAT (whitening of BAT), which impairs its ability to dissipate the energy. We observed that STF treatment reversed the whitening of BAT, as reflected in much less lipid droplet intensity in BAT, in DIO mice, compared to vehicle treatment (Figure 2I). Immunofluorescent staining showed the elevated staining intensity of UCP1 protein in BAT per unit area in DIO mice compared to vehicle treatment (Figure 4I–K). Furthermore, STF-treated mice showed increased expression of genes involved in thermogenic and mitochondrial functions, including Prdm16, Ucp1, Cidea, Ppara, Dio2, Cox5b, and Ppargc1a (Pgc1α), in BAT (Figure 4L). These data thereby suggest that STF promotes BAT activation. Induction of beige fat in the WAT also plays important role in the adaptive thermogenesis and energy expenditure (Kajimura et al., 2015). We therefore examined the effect of STF on the status of beiging activation in the WAT. The immune-signal intensity of UCP1 protein was markedly increased in the eWAT (Figure 4M and Figure 4—figure supplement 2A) and iWAT (Figure 4—figure supplement 2B-C) of DIO mice treated with STF. In addition, the expression of thermogenic genes was significantly increased in both eWAT and iWAT of DIO mice subjected to STF treatment compared to vehicle (Figures 4N and 3D). These results suggest the involvement of beige fat activation in thermogenesis and energy expenditure in STF-treated mice. Moreover, we also investigated the effect of STF on the expression -adrenoceptors, which play important role in adaptive thermogenesis in response to cold stimulation, on adipose tissues and observed that STF treatment elevated the levels of both β2-adrenoceptor (Adrb2) and β3-adrenoceptor (Adrb3) in BAT (Figure 4O) and the levels of Adrb3 in iWAT and eWAT (Figure 4—figure supplement 2E-F’).

Since adipose tissue macrophage (ATM) accumulation plays important role in contributing to insulin resistance and obesity (Xu et al., 2003; Lumeng et al., 2007), we investigate whether STF treatment impacts ATM accumulation in DIO mice as IRE1α activity was observed to be heightened in obese macrophages (Shan et al., 2017). During obesity, the expansion of adipose tissues accompanies the recruitment and accumulation of macrophages in the adipose tissue, and crown-like structures (CLS) are formed upon the infiltration and residence of macrophages around dead adipocytes in the adipose tissues (Cinti et al., 2005). As shown in Figures 2C and 5A, the area of CLS was drastically reduced in the eWAT of DIO mice treated with STF relative to vehicle-treated mice, although CLS area showed no significant difference in iWAT between vehicle and STF groups (Figure 5—figure supplement 1A). Immunofluorescent staining of F4/80, a pan-macrophage marker, was drastically increased in the eWAT under HFD, but was reversed in STF-treated mice (Figure 5B–C). Moreover, we examined the expression levels of pan-macrophage marker genes (Cd68 and F4/80 (Adgre1)) and found that STF treatment significantly alleviated the mRNA levels of Cd68 and Adgre1 in the stromal vascular fraction (SVF where immune cells reside) of the eWAT (Figure 5D). Expectedly, STF treatment also inhibited the splicing of IRE1α target Xbp1 mRNA in the eWAT SVF (Figure 5—figure supplement 1B-D). Next, we used flow cytometry to analyze the changes of ATMs by STF treatment under HFD. We observed that the percentage of macrophage population (double positive for F4/80 and CD11b markers [F4/80⁺CD11b⁺]) did not alter substantially in the SVF fraction isolated from eWAT under HFD compared to normal chow (Figure 5E and Figure 5—figure supplement 2A); however, given the HFD-induced expansion of the eWAT, the total number of F4/80⁺CD11b⁺ macrophages has increased drastically by 49.6-fold: 0.29±0.25 × 10⁵ (normal chow, n=8) vs 1.43±0.086 × 10⁶ (14-week HFD, n=8) cells/eWAT fat pad/mouse (Pp0.0041) (Table 1a). The density of F4/80⁺CD11b⁺ macrophage population in eWAT also increased under HFD by 12.5-fold after normalizing the total number of macrophages to fat pad weight (0.62±0.45 × 10⁵ (normal chow) vs. 0.77±0.015 × 10⁶ cells (14 week HFD)/g of eWAT/mouse, p=0.0045, Table 1a). Excitingly, STF treatment significantly lessened the HFD-induced increases in the percentage, total number, and density of ATMs by 37.4%, 46%, and 54%, respectively (Figure 5E and Figure 5—figure supplement 2A and Table 1a). These results indicate that STF suppresses macrophage accumulation in the adipose tissue.

It was previously reported that macrophages that accumulated in adipose tissues undergo a ‘phenotypic switch’ from an anti-inflammatory M2 (alternatively activated) phenotype to a proinflammatory M1 (classically activated) state, a process believed to be critical for the development of systemic insulin resistance (Lumeng et al., 2007). We therefore investigated the effect of STF treatment on such an ATM switch. We observed that under HFD, the percentage (14.59% (ND) vs. 52.95% (HFD), p=0.007, Figure 5F and Figure 5—figure supplement 2A, Table 1b), total number/eWAT/mouse (0.25±0.16 × 10⁴ (ND, n=8) vs 0.75±0.11 × 10⁶ (HFD, n=8), 299.58-fold, p<0.0001, Table 1b), and density/g of eWAT/mouse (0.59±0.23 x 10⁴ (ND) vs. 4.11±0.96 x 10⁵ (HFD), 70-fold, p=0.0018, Table 1b) of F4/80⁺CD11b⁺CD11c⁺ triple positive cells, which are typically characterized as the classical ‘M1-like’ adipose tissue macrophages, were markedly increased in the eWAT. Unexpectedly, we found that the bulk majority of these CD11c⁺ ATMs were also positive for CD206, a marker thought for M2 macrophages, in the eWAT under HFD (Figure 5—figure supplement 2B), and that approximately half the population is positive for both CD11c and CD206 in DIO ATMs, whereas there are only 8.55% CD11c⁺CD206⁺ ATMs under ND when gated against CD11c and CD206 (Figure 5G and Figure 5—figure supplement 2A), suggesting the enrichment of a mixed M1/M2 phenotype, which is consistent with recent reports (Shaul et al., 2010; Wentworth et al., 2010; Sica and Mantovani, 2012). Moreover, we found that the total number and density of CD11c⁺CD206⁺ ATMs increased by 337- and 75-fold respectively under HFD (Table 1c; ND: n=8; HFD: n=8), a greater increase than those of CD11C-postive alone (CD11c⁺CD206^-) ATMs under HFD (Supplementary file 1; ND: n=8; HFD: n=8). On the other hand, for CD206⁺CD11c^- ATM subpopulation, a presumed M2-like ATM subpopulation which accounted for the majority of ATMs under ND as expected (Figure 5G and Table 1d), HFD drove the increase in the total cell number and density of this population by over 16- and 4-fold respectively despite the diminished percentage given the huge expansion of total ATMs (Table 1d; ND: n=8; HFD: n=8), although to a lesser degree than that for CD11c⁺CD206⁺ or CD11c⁺CD206^- macrophages (Table 1c and Supplementary file 1a). Moreover, CD206^-CD11c^- ATM subpopulation also exhibited a trend similar to that of CD206⁺CD11c^- ATMs (Supplementary file 1b; ND: n=8; HFD: n=8). Our findings confirm the notion that ATMs do not follow the classic M2 to M1 switch (Xu et al., 2013; Kratz et al., 2014; Hill et al., 2018).

It has been reported that the germline deletion of IRE1α in myeloid lineage reverses the switch in M1/M2 polarization with the increase in M2 population to induce BAT activation and WAT remodeling and protect mice against diet-induced obesity (Shan et al., 2017). However, in light of increasing evidence from others (Xu et al., 2013; Kratz et al., 2014; Hill et al., 2018) and our current findings that do not support a HFD-triggered classical M1/M2 polarization, we sought to interrogate the effect of IRE1α inhibition on ATM subpopulations, which by far is poorly understood. In DIO mice treated with STF for 4 weeks, we found that consistent with the overall reduction of ATM population (Figure 5E, Table 1a), the percentage, total number, and density of the posited ‘M1-like’ F4/80⁺CD11b⁺CD11c⁺ triple positive cells were substantially diminished in the SVF of eWAT (Figure 5F and Figure 5—figure supplement 2A, Table 1b), but to an even greater extent than the overall ATM population. The trend in the reduction of F4/80⁺CD11b⁺CD11c⁺ ATMs by STF remains similar regardless of whether CD206⁺ was positive (F4/80⁺CD11b⁺CD11c⁺CD206⁺) or negative (F4/80⁺CD11b⁺CD11c⁺CD206^-) (Figure 5G, Table 1c and Supplementary file 1a). In addition, the total number and density of M2-like CD206⁺CD11c ^– ATMs were also significantly reduced in the SVF of eWAT of STF-treated DIO animals (Table 1d), but to a lesser degree than CD11c⁺ ATMs. In contrast, STF treatment even slightly increased the total number and density of the CD206 CD11c double-negative ATMs (CD206^-CD11c^-F4/80⁺CD11b⁺) or non-macrophage (F4/80^-CD11b^-) populations (Supplementary file 1b and c), although none of them reaching statistically significance.

As the ‘M1-like’ CD11c⁺ and M1/M2 mixed ATMs were reported to be associated with pro-inflammation (Shaul et al., 2010; Wentworth et al., 2010), we examined the mRNA levels of proinflammatory cytokine genes —Tnfa, Il1b, Il6, and monocyte chemoattractant protein-1 (Ccl2)— and found that they were down-regulated in the SVFs of eWAT from STF-treated DIO mice compared to vehicle-treated counterparts (Figure 5H), which correlates with our observed decreased populations of ‘M1-like’ CD11c⁺ and M1/M2 mixed ATMs in the STF-treated DIO animals (Figure 5E–F). Similarly, STF treatment also inhibited the mRNA levels of these cytokine genes in the SVFs of iWAT (Figure 5—figure supplement 2C). Furthermore, in CD11c⁺ ATMs (F4/80⁺CD11b⁺CD11c⁺) sorted from the eWAT SVFs of DIO mice, which expectedly expressed higher mRNA levels of Tnfa, Il1b, and Ccl2 than CD11c^- ATMs (Figure 5I), STF treatment remarkably reduced the expression levels of Tnfa, Il1b, and Ccl2 (Figure 5J). STF inhibited the IRE1α RNase activity in these cells as reflected by the diminished Xbp1s mRNA levels (Figure 5—figure supplement 2D). These results reveal that STF treatment suppresses the adipose tissue inflammation and the accumulation of pro-inflammatory ATM without augmenting (suppressing instead) M2-like ATMs.

Recent studies have shown that instead of the classical proinflammatory M1 polarization, obesity drives the expansion of metabolically active heterogeneous ATM populations. These metabolically active ATMs possess heightened lysosome and lipid metabolism (Xu et al., 2013; Kratz et al., 2014; Coats et al., 2017), express cell surface markers CD9 or Trem2 at a higher frequency, and are pro-inflammatory (Hill et al., 2018; Jaitin et al., 2019). These studies reported that the CD9⁺ macrophages have high amounts of intracellular lipid in lysosome‐like structures, express genes related to lysosomal‐dependent lipid metabolism (Hill et al., 2018) and represent the majority of total adipose tissue macrophages in obesity (Jaitin et al., 2019). Given our findings on the significant inhibitory effect of STF on ATM accumulation and inflammation, we therefore investigated whether STF impacts these alterations in the adipose tissues of DIO mice. We first determined the effect of STF treatment on the metabolic activation of ATMs. Proteins involved in lipid metabolism including ABCA1, CD36, and PLIN2 were reported to be specifically up-regulated in metabolically active macrophages in obesity (Kratz et al., 2014). We observed that the mRNA levels of Abca1, Cd36, and Plin2 were elevated in the SVF portion of the eWAT in DIO mice compared to their normal diet counterparts, but STF treatment significantly suppressed the up-regulation of these genes (Figure 6A). Moreover, genes involved in lysosomal biogenesis and phagocytosis including Atp6v1b2, Atp6v0d2, Lipa, and Lamp2 were also reported to be induced in the ATMs in obesity (Xu et al., 2013; Coats et al., 2017). We therefore examined the effect of STF on the mRNA levels of Atp6v1b2, Atp6v0d2, Lipa, and Lamp2 and observed that their mRNA levels were elevated in the SVF portion of the eWAT in DIO mice and that STF treatment suppressed the up-regulation of these genes, although the suppression on Lipa does not reach statistical significance (Figure 6B). In addition, expression levels of other genes known to be involved in lipid metabolism and lysosomal biogenesis, including Fabp4, Fabp5, and Stsb1, were also suppressed in the SVF portion of the eWAT in DIO mice by the treatment of STF (Figure 6C). Lastly, we also observed a similar trend in the SVFs of the iWAT (Figure 6—figure supplement 1A-B).

We next investigated the effect of STF on the newly identified ATM subpopulations. First we observed that the mRNA levels of CD9 and Trem2 were elevated in the SVFs of the eWAT and iWAT in DIO mice relative to age-matched normal chow-fed mice but STF treatment suppressed the up-regulation of these genes (Figure 6D and Figure 6—figure supplement 1C). In addition, immunofluorescent staining of CD9 and Trem2 was drastically increased in the eWAT under HFD, relative to normal chow (Figure 6—figure supplement 1D-E), but was profoundly diminished in STF-treated DIO mice (Figure 6E–G). Similarly, we observed a similar trend in the effect of STF on the immunofluorescent staining of CD9 and Trem2 in BAT (Figure 6—figure supplement 1F) and iWAT (Figure 6—figure supplement 1G-H) slides from DIO mice treated with STF or Veh, although without reaching statistical significance. We then used flowcytometry to characterize the effect of STF on CD9- or Trem2-expressing ATM subpopulations. We initially examined the HFD-induced changes in these populations. As expected, the percentage (23.25 ± 8.75% (ND, n=8) vs 72.70 ± 2.60% (HFD, n=8), p=0.032, Figure 6H, Table 2a), total number (0.45±0.32 x 10⁴/eWAT/mouse (ND) vs. 1.04 ± 0.026 x 10^6. /eWAT/mouse (HFD), 228-fold, p=0.00062, Table 2a), and density (1.04±0.51 x 10⁴ /g of fat pad (ND) vs. 0.56±0.032 x 10⁶ /g of fat pad (HFD), 54-fold, p=0.0034, Table 2a) of the CD9⁺ ATMs (F4/80⁺CD11b⁺CD9⁺) in the eWAT were increased under HFD. It was previously reported that CD63 is enriched in CD9⁺ ATMs (Jaitin et al., 2019); indeed, we found that up to 98% of CD9⁺ ATMs also expressed CD63 in HFD-fed mice (Figure 6—figure supplement 2). As expected, the percentage, total number, and density of CD9⁺CD63⁺ ATMs increased significantly in the eWAT (Figure 6—figure supplement 2B and Supplementary file 2; ND: n=8; HFD: n=8). Similarly, Trem2⁺ ATM subpopulation expanded markedly in the percentage, total number/eWAT/mouse, and density (number/g of fat pad/mouse) under HFD (Figure 6I, Table 2b). In addition, we found that while only approximately 47% of the CD9⁺ ATMs (F4/80⁺CD11b⁺CD9⁺) express Trem2, an overwhelming majority (~86.7%) of Trem2⁺ ATMs (F4/80⁺CD11b⁺Trem2⁺) expresses CD9 (Figure 6—figure supplement 2C), corroborating a previous finding that Trem2⁺ expression emerges after CD9⁺ ATMs over the course of obesity (Jaitin et al., 2019). As expected, double-positive CD9⁺Trem2⁺ ATMs also expanded tremendously under HFD (Figure 6J, Table 2c). In DIO mice treated with STF, we observed a decrease in the percentages of CD9⁺, Trem2⁺, and CD9⁺Trem2⁺ ATMs relative to that of vehicle-treated mice (Figure 6H–J, Table 2a–c). Further, STF treatment significantly diminished the total number and density of CD9⁺, Trem2⁺, and CD9⁺Trem2⁺ ATM subpopulations all by over 60% (Table 2a–c).

The CD9-expressing ATM subpopulation under HFD was proposed to be pro-inflammatory (Hill et al., 2018; Jaitin et al., 2019). Indeed, we observed that compared to CD9^- ATMs sorted from the eWAT of DIO mice, the sorted CD9⁺ ATMs expressed significantly higher mRNA levels of pro-inflammatory genes Tnfa, Il1b, and Ccl2 (Figure 6K). Similarly, the mRNA levels of Tnfa, Il1b, and Ccl2 were also heightened in sorted Trem2⁺ ATMs relative to Trem2^- ATMs (Figure 6L). Notably, STF treatment of DIO mice markedly reduced the expression levels of Tnfa, Il1b, and Ccl2 in CD9⁺ ATMs sorted from the eWAT (Figure 6M). A similar trend was obtained in sorted Trem2⁺ ATMs from the obese eWAT (Figure 6N). These results together demonstrated that STF treatment suppresses the subpopulations of metabolically active CD9 and Trem2-positive ATMs and adipose inflammation.

To investigate whether IRE1α inhibition acts on ATMs directly, we examined the effect of STF on the macrophage population and inflammation in SVF cells (freshly isolated from obese eWATs) treated with STF. We observed that STF treatment for 20 hours slightly but statistically significantly diminished the percentage of ATMs (F4/80⁺CD11b⁺; DMSO 56.03% vs. STF 49.30%, p=0.0024) as analyzed by flow cytometry (Figure 6O). When further gated against Trem2, we observed a trend similar to that for the overall ATMs in that STF treatment moderately but significantly decreased the percentage of Trem2⁺ ATMs (F4/80⁺CD11b⁺Trem2⁺; DMSO 68.67% vs. STF 61.93%, p=0.0013; Figure 6P). We also observed that STF mildly lessened the percentages of CD9⁺ ATMs (F4/80⁺CD11b⁺CD9⁺; DMSO 76.73% vs. STF 74.80%, p=0.013; Figure 7—figure supplement 1A) and of CD11C⁺ ATMs (F4/80⁺CD11b⁺CD11C⁺; DMSO 81.23% vs. STF 77.63%, p=0.0040) (Figure 7—figure supplement 1B). We next investigated the effect of STF on the expression levels of pro-inflammatory genes in ATMs sorted from STF- or DMSO-treated SVF cells. As expected, STF treatment decreased spliced Xbp1 mRNA levels (Figure 6Q, Figure 7—figure supplement 1C, G, I), reflecting the inhibition of IRE1α RNase activity, and significantly down-regulated the expression levels of pro-inflammatory genes (Tnfa, Il1b, and Ccl2) in CD9⁺ ATMs (F4/80⁺CD11b⁺CD9⁺; Figure 6R) and in CD11C⁺ ATMs (F4/80⁺CD11b⁺CD11C⁺; Figure 6—figure supplement 1S), relative to DMSO control-treated counterparts. Similarly, in CD9^- ATMs (F4/80⁺CD11b⁺CD9^-) or CD11C^- ATMs (F4/80⁺CD11b⁺CD11C^-) sorted from STF- or DMSO-treated SVFs, which displayed lower mRNA expression levels of Tnfa, Il1b, and Ccl2 compared to their respective CD9⁺ or CD11C⁺ counterparts (Figure 7—figure supplement 1D-E), STF also down-regulated the mRNA levels of Tnfa, Il1b, and Ccl2 (Figure 7—figure supplement 1F, H). Together, these results indicate that STF likely acts directly on ATMs to regulate inflammation.

One important question on the obesity-associated ATM accumulation is the obscure relationship between the newly-identified CD9⁺ and Trem2⁺ ATM populations and the originally identified ‘M1-like’ CD11c⁺ ATMs (Hill et al., 2018; Wu and Ballantyne, 2020). To investigate this relationship, we analyzed these populations using flow cytometry (Figure 7—figure supplement 2). First, we asked whether the previously identified F4/80⁺CD11b⁺CD11c⁺ triple positive ATMs express signature markers of the newly identified ATM subpopulations (CD9 and/or Trem2) in DIO mice. Our results showed that approximately 80% (91.30 ± 0.40%) of CD11c⁺ ATMs (F4/80⁺CD11b⁺CD11c⁺) express CD9 (Figure 7A and Table 3a; ND: n=8; HFD: n=8). On the other hand, about 70. 4% of CD11c-negative ATMs (F4/80⁺CD11b⁺CD11c^-) express CD9 (Figure 7A, Table 3a). We also found that 23.95% of F4/80⁺CD11b⁺CD11c⁺ ATMs express CD9^-hi while only 3.185% of CD11c-negative ATMs are positive for CD9^-hi (Figure 7A). Similarly, the percentage of cells expressing Trem2 was higher in the CD11c⁺ ATMs than in CD11c^- ATMs (43.55 ± 8.65% for CD11c⁺ vs. 24.35 ± 4.05%, albeit not reaching statistical significance (p=0.074), Figure 7B, and Table 3a). We also observed a higher percentage of the CD11c⁺ ATMs (30.05 ± 5.55%) expressing both CD9 and Trem2 than that of CD11c^- ATMs (8.35 ± 1.86%), p=0.066, (Figure 7C and Table 3a). These results revealed that CD9 and Trem2 are highly enriched but not exclusively expressed in the CD11c⁺ ATMs under HFD.

In reciprocal studies, we analyzed the distribution of CD11c expression in CD9⁺ and/or Trem2⁺ ATMs under HFD. We observed heightened percentages of CD9⁺, Trem2⁺, and CD9⁺Trem2⁺ ATMs that co-express CD11c (63.25 ± 10.85%, 78.25 ± 12.35%, and 78.65 ± 13.05%, respectively) relative to the CD9^-, Trem2^-, and CD9^-Trem2^- ATMs that co-express CD11c (29.90 ± 9.10%, 49.55 ± 9.85%, or 24.60 ± 5.90%, respectively), in the eWAT of HFD-fed mice (Figure 7D–F and Table 3b; ND: n=8; HFD: n=8). We also observed that higher percentage of CD9^hi ATM subpopulation expresses CD11c than CD9⁺ ATMs (90.60 ± 1.4% vs 63.25 ± 10.25%, Figure 7D). These results showed that while CD11c expression is not exclusively expressed in CD9⁺, Trem2⁺, or CD9⁺Trem2⁺ ATMs, it is highly enriched in all of these subpopulations. Together, these results reveal that in the eWAT under HFD, the majority of CD11c⁺ or CD9⁺ ATMs expresses CD9 or CD11c, respectively. Indeed, up to half (45.45 ± 5.05%) of ATMs in the obese eWAT co-express CD9 and CD11c (F4/80⁺CD11b⁺CD9⁺CD11c⁺) (Figure 7G and Table 3c; ND: n=8; HFD: n=8). Notably, similar to its effects on CD9⁺ or CD11c⁺ ATMs (Figure 6H–I), STF treatment substantially decreased the percentage, total number /eWAT/mouse, and density (per gram of fat pad/mouse) of the CD9⁺CD11c⁺ ATMs in the eWAT of DIO mice (Figure 7G, Table 3c). As CD9⁺ and CD11c⁺ ATMs were reported to be pro-inflammatory, we next compared the expression levels of proinflammatory genes in these ATM subpopulations from HFD-fed mice and observed that the mRNA levels of Tnfa, Il1b, and Ccl2 were all up-regulated in sorted CD9⁺CD11c^-, CD9^-CD11c⁺, and CD9⁺CD11c⁺ ATMs compared to the CD9^-CD11c^- ATMs, with the CD9⁺CD11c⁺ ATMs expressing higher level of Ccl2 and the CD9⁺CD11c^- ATMs expressing lower level of Il1b (Figure 7H), suggesting that these ATM subpopulations are all proinflammatory but yet heterogeneous in nature.

C57BL/6 J mice were obtained from Jackson laboratory (Bar Harbor, ME). Mice were kept in an animal facility with a 12 hr light/dark cycle and temperature of 22 °C. All animals had access to diet and water ad libitum. All procedures involving animals were approved by the Institutional Animal Care and Use Committee of the University of Oklahoma Health Science Center. All experiments were performed with age-matched male mice. The C57BL/6 J male mice were fed a HFD (60% kcal fat, Bio-Serv, NJ, USA) at the age of 5–6 weeks. At the 8-week HFD feeding, mice were randomly divided into two groups and were administered with either vehicle (10% DMSO) or STF (10 mg/kg body weight in 10% DMSO) once daily through the route of intraperitoneal injection. Blood glucose levels were measured using the OneTouch Ultra2 glucometer after fasting for 6 hr. Body weights were measured weekly.

Intraperitoneal glucose tolerance test (ipGTT) was performed after 16 hr overnight fasting. Blood glucose levels were measured at 0, 15-, 30-, 60-, and 120 min glucometer (OneTouch Ultra 2 Meter) after an intraperitoneal administration of glucose at dose of 1.5 g/kg body weight. Intraperitoneal insulin tolerance test (ipITT) was performed after 6 hr fasting. Blood glucose levels were measured at 0, 15, 30, 60, and 120 min after an intraperitoneal administration of human insulin at dose of 1.2 IU/kg body weight.

Fasting serum insulin (ALPCO, NH, USA), serum cholesterol TG (Cayman Chem., MI, USA) and FFAs (Bioassay system, CA, USA) were determined by ELISA.

Body lean and fat composition and total body weight were determined by EchoMRI test.

The mice were individually housed in chambers of a Promethion Core Monitoring system (Sable Systems, Las Vegas, NV, USA). After a 1-day acclimation period, their oxygen consumption and carbon dioxide production were measured for the next 3 consecutive days. The respiratory exchange ratio and energy expenditure were calculated using standard equations.

Mice were sacrificed and the epididymal visceral white adipose tissue (eWAT) was immediately washed by PBS. The tissue was placed in 10 cm petri dish and excised into tiny bits with scissors at room temperature. Transfer the fine chopped tissue pieces in 10 mL plain DMEM containing Collagenase P (0.45 mg/mL) in 50 mL tubes. Keep the tubes in 37 °C incubator shaker at a speed of 200 rpm for 30–40 min. Add 5 mL of media containing FBS to each tube and pipette well and spin at 700 × g rcf for 10 min at 4 °C. Remove gently all the supernatants, and the remaining pellet is stromal vascular fraction (reddish brown –containing RBC).

For flow cytometry, cells were resuspended in sterile FACS buffer, then filter through 70μm strainer and centrifuge at 1000 rpm at 4 °C for 5 minutes. Resuspend the pellet in 2 mL of 1 X RBC lysis buffer solution (Sigma) and incubated on ice for 2–3 min. Add 10 mL of PBS and spin the cells at 1100 rpm for 5 min. The resulting cell suspension was incubated with Fc Block (BD Biosciences) prior to staining with conjugated antibodies for 20 min at 4 °C followed by two washes in FACS buffer. Samples were stained using the following antibodies: PerCP-conjugated anti-mouse CD45 (Cat# 110725, RRID:AB_893347), FITC-conjugated anti-mouse F4/80 (Cat# 123108, RRID:AB_893502), PE-conjugated anti-mouse/human CD11b (Cat# 101207, RRID:AB_312790), BV421-conjugated anti-mouse CD11c, (Cat# 117330, RRID:AB_11219593), Alexa Fluor 700-conjugated anti-mouse CD206 (Cat# 141734, RRID:AB_2629637), PE Dazzle594-conjugated anti-mouse CD9 (Cat# 124821, RRID:AB_2800601), and PE-Cy7-conjugated anti-mouse CD63 (Cat# 143909, RRID:AB_2565499), purchased from Biolegend. APC-conjugated Trem2 (Cat# FAB17291N, RRID:AB_3646995, R&D Systems). Fluorescence Minus One (FMO) controls were included in the flow experiments. Cells were analyzed using Aurora Spectral Flow Cytometer (Cytek) and FlowJo software (FlowJo 10.10, LLC). Sorting was performed using MoFloXDP (Beckman Coulter). Single cells were sorted into 1.5 mL tube containing 50 μL of DMEM medium. Immediately after sorting, tubes were spun down and removed the medium, snap-frozen on dry ice and stored at −80 °C until further processing.

Freshly isolated SVF portions were cultured in RPMI 1640, supplemented with 10% FBS and 1% penicillin/streptomycin (PS; RPMI/FBS/PS). 2 hr after seeding for a density at 2x10⁶ cells/10 cm-dish, SVF cells were treated with 0.01% DMSO control or 30μM STF for 20 hr. Cells were collected and labeled with the antibodies as above followed by FACS analysis or sorting.

Adipose tissue and liver were dissected and immediately fixed in 4% paraformaldehyde (Sigma–Aldrich), paraffin embedded, and stained with hematoxylin and eosin. For Immunofluorescence staining, the eWAT and iWAT paraffin sections (the section thickness is 5–10 µm) were blocked with 1% BSA, 0.2% Triton PBS, then incubated with primary antibody F4/80 (Cat#:70076, RRID:AB_2799771, Cell Signaling Technology), CD9 (Cat#:ab223052, RRID:AB_2922392, Abcam), Trem2 (Cat#: MAB17291, RRID:AB_2208679, R&D Systems) UCP1(Cat#:ab10983, RRID:AB_2241462, Abcam) overnight at 4 °C after deparaffinization. The slides were washed thrice with PBS with 0.2% Triton and then incubated in Alexa Fluor 488-, 555-, and 647- conjugated secondary antibodies (Jackson ImmunoResearch, PA), DAPI (0.5 mg/mL) for 1.5 hr at room temperature. Tissue sections were imaged on the Olympus Fluoview 1000 laser-scanning confocal microscope (Center Valley, PA) and quantified with Image-J histogram software.

Total RNA (2 μg) was isolated from tissues or cells using TRIzol reagents (Life Technologies) or total RNA (0.1–0.5 μg) from sorted cells using E.Z.N.A. total RNA kit (Omega Bio-Tek, Norcross, GA) isolated and reverse transcribed using oligo d(T) primers (New England Biosystems) and SuperScript IV reverse transcription kit (Applied Biosystems, Foster City, CA). qPCR was performed in a CFX96 Touch Real-Time PCR detection system using SYBR Green mix (Applied Biosystems, Foster City, CA). The amplification program was as follows: initial denaturation at 95 °C for 15 min, followed by 40 cycles of 95 °C for 15 s, 60 °C for 1 min, and 40 °C for 30 s. Relative mRNA expression was determined by the ∆∆Ct method normalized to Tbp mRNA. The sequences of primers used in this study are found in Supplementary file 3. Regular RT-PCR was performed and the products were resolved by agarose gel electrophoresis. The unspliced and spliced Xbp1 mRNA levels were quantified using ImageJ software (National Institutes of Health, Bethesda, MD).

Proteins were extracted with RIPA buffer containing protease and phosphatase inhibitors (Thermo) and then centrifuged for 10 min at 10,000 × g. Total protein concentration in the cell lysate was determined by Pierce BCA Protein Assays. Samples of ~30 μg protein prepared with (reducing) 5% β-mercaptoethanol were separated on a 4–12% Bis-Tris NuPAGE gels (Invitrogen) and transferred to polyvinylidene-fluoride membrane for 1 hr at 4 °C at 40 V. The membrane was blocked in TBS (10 mm Tris-HCl, pH 7.4, 150 mm NaCl) containing 3% BSA for 1 hr at room temperature and was probed with primary antibodies followed by the appropriate HRP-conjugated secondary antibodies (1:5000; Santa Cruz Biotechnology, CA, USA). The primary antibodies were: Phospho-Akt (Ser473; 4060 S, RRID:AB_2315049, Cell Signaling Technology), Akt (pan) (4685 S, RRID:AB_2225340, Cell Signaling Technology), GAPDH (sc-365062, RRID:AB_10847862, Santa Cruz Biotechnology).

All data are presented as the mean ± SEM of the indicated number of replicates. Data were analyzed using the unpaired two-tailed Student’s t-test or ANOVA with multiple comparisons (Prism 10.3.1) and p<0.05 was considered statistically significant. For energy expenditure experiments, the data was analyzed for group effect (ANOVA) along with the mass and interaction effects as necessary (generalized linear model, GLM; Mina et al., 2018).