Chronic low-grade inflammation, often referred to as metainflammation, develops in response to overnutrition and is a major player in the regulation of insulin sensitivity. While many studies have investigated adipose tissue inflammation from the perspective of the immune cell compartment, little is known about how adipocytes intrinsically contribute to metainflammation and insulin resistance at the molecular level. In this study, we demonstrate a novel role for Jumonji C domain-containing protein 8 (JMJD8) as an adipocyte-intrinsic molecular nexus between inflammation and insulin resistance. We determined that JMJD8 was highly enriched in white adipose tissue, especially in the adipocyte fraction. Adipose JMJD8 levels were dramatically increased in obesity-associated insulin resistance models. Its levels were increased by feeding and insulin and inhibited by fasting. A JMJD8 gain-of-function was sufficient to drive insulin resistance, whereas loss-of-function improved insulin sensitivity in mouse and human adipocytes. Consistent with this, Jmjd8-ablated mice had increased whole-body and adipose insulin sensitivity and glucose tolerance on both chow and a high-fat diet, while adipocyte-specific Jmjd8-overexpressing mice displayed worsened whole-body metabolism on a high-fat diet. We found that JMJD8 affected the transcriptional regulation of inflammatory genes. In particular, it was required for lipopolysaccharide-mediated inflammation and insulin resistance in adipocytes. For this, JMJD8 required interferon regulatory factor 3 to mediate its actions in adipocytes. Together, our results demonstrate that JMJD8 acts as a novel molecular factor that drives adipocyte inflammation in conjunction with insulin sensitivity.
Chronic, low-grade tissue and systemic inflammation have been proposed as major mechanisms for insulin resistance in obesity (1–3). Adipose tissue stands at this nexus between inflammation and metabolism (2). In the lean state, leukocytes interact with macrophages within the adipose tissue to maintain low inflammatory tone and appropriate insulin sensitivity (3,4). However, during chronic overnutrition, there are fewer anti-inflammatory immune cells (e.g., eosinophils, innate lymphoid type 1 and 2 cells, and regulatory T cells) (5–9) and increased infiltration of proinflammatory immune cells (e.g., M1 macrophages, B cells, Th1 and 17 cells, dendritic cells, and neutrophils) (10–16). Concordantly, there is a concomitant decrease in local and circulating levels of anti-inflammatory cytokines, such as interleukin-4 (IL-4) and IL-13 (5–9), and increased levels of proinflammatory cytokines and chemokines, such as tumor necrosis factor-α (TNF-α) (17,18) and IL-6 (14), which trigger innate immune signaling and are commonly associated with insulin resistance.
It has been proposed that this inflammation is reacting to lipopolysaccharide (LPS), which is an outer membrane component of Gram-negative bacterial cell walls. High-fat diets (HFDs) may alter the gut microbiome and intestinal wall permeability, increasing enterobacterial production and LPS translocation into systemic circulation (19). Indeed, some groups of obese subjects and subjects with type 2 diabetes (T2D) (20,21) have elevated LPS plasma concentrations. In addition, an HFD increases the LPS concentrations in circulation in both healthy subjects and those with T2D (20,22), and LPS administration to healthy subjects rapidly induces insulin resistance (23). Moreover, disrupting LPS-mediated signaling at the receptor level protects mice from developing inflammation and insulin resistance in response to chronic changes in dietary fat (24). As such, the elevation in LPS concentration/signaling could play a key role in low-grade systemic inflammation and insulin resistance in obesity; however, the underlying mechanisms are not fully understood.
It has been proposed that adipocytes themselves sense inflammatory signals and stimulate innate immune pathways. Indeed, adipocytes express immune receptors, including Toll-like receptors (TLRs; e.g., TLR2 and 4), that can sense inflammatory signals such as free fatty acids and LPS (24–26). By the same token, adipocytes express MHC proteins and are thus capable of acting as antigen-presenting cells to activate T cells (27). Furthermore, immune transcription factors, such as nuclear factor-κB (NF-κB) and interferon regulatory factor 3 (IRF3), have immune functions in adipocytes (28,29). However, we lack insight into how adipocytes use these immune proteins to mediate innate immune signals and how they associate with insulin resistance.
JMJD proteins are a class of proteins containing the Jumonji C (JmjC) domain, which is essential for their enzymatic activity, and 33 family members have been identified so far (30). Many of the JMJD proteins containing the canonical JmjC domain regulate various signaling and transcriptional pathways, thus affecting a diverse set of biological activities, including activities in adipose cells. For example, JMJD1A (31–33) and JMJD2B/KDM4B (34) promote adaptive thermogenesis by demethylating H3 lysine 9 (H3K9) di- or trimethylation, which are repressive histone marks. JMJD1C, also acting as a H3K9me2 demethylase, is required for adipogenesis and the full response to insulin-stimulated glucose uptake in 3T3-L1 adipocytes (35). In contrast, 10 of the JMJD members, including JMJD8, are characterized by their relatively small molecular weights (27–71 kDa) (36) and comprise a subfamily that is evolutionarily distant from the other conventional JMJD proteins (36). Little is known about their biological function to date. Our gene expression profiling studies identified that JMJD8 expression is highly enriched in adipocytes and is regulated by metabolic and nutritional states, implying its metabolic function in adipocytes.
In this study, we show that the expression of JMJD8 is affected by obesity, fasting, feeding, and insulin. In vitro studies demonstrate that JMJD8 was both necessary and sufficient to mediate insulin resistance in both mouse and human adipocytes. Congruently, whole-body Jmjd8-knockout (KO) mice displayed improved insulin sensitivity on an HFD and even on a chow diet. Conversely, adipocyte-specific Jmjd8-overexpressing mice exhibited aggravated insulin resistance and glucose intolerance on an HFD. In adipocytes, JMJD8 expression affected the gene program of several proinflammatory genes, which can trigger inflammation through cell-to-cell communication with macrophages. Lastly, we found that JMJD8 functionally interacted with IRF3, a proinflammatory factor that orchestrates adipocyte inflammation and insulin sensitivity. Together, these demonstrate that JMJD8 is a novel molecular nexus linking adipocyte insulin sensitivity and inflammation.
Research Design and Methods
3T3-L1 preadipocytes were obtained from American Type Culture Collection, and these cells were authenticated by the ability to differentiate and confirmed to be mycoplasma negative. To generate lentivirus particles, lentiviral constructs were cotransfected with pM2DG- and psPAX-expressing plasmids into 293T cells. After 48 h, virus-containing supernatant was collected, filtered through 0.45-mm filters, and added to mature 3T3-L1 adipocytes for 24 h along with 8 mg/mL polybrene. Transduction efficiency was determined by comparing to cells transduced in parallel with a GFP-expressing lentivirus. For the ex vivo system, subcutaneous adipose tissue from wild-type (WT) C57BL/6 mice was fractionated with digestion buffer (10 mg/mL collagenase D, 2.4 units of dispase II, and 10 mmol/L CaCl2 in PBS). Simpson-Golabi-Behmel syndrome (SGBS) cells were maintained with DMEM/F12 and GlutaMAX. To differentiate SGBS cells, rosiglitazone (2 μmol/L), transferrin (0.01 mg/mL), cortisol (100 nmol/L), and triiodothyronine (0.2 nmol/L) were added in addition to isobutylmethylxanthine (250 μmol/L), insulin (20 nmol/L), and dexamethasone (25 nmol/L) from days 0–4.
Jmjd8-KO and Adi-Jmjd8-TG mice were generated at the Gene Targeting Facility at University of California Berkeley. Adipoq-Cre mice on the C57BL/6 background were obtained from The Jackson Laboratory (028020; Bar Harbor, ME). For chow and high-fat feeding studies, male C57BL/6J mice were put on the diet beginning at 9 weeks of age and continuing for 3 months. For histology, adipose tissues were fixed with neutral-buffered formalin and embedded in paraffin, and sections were stained with hematoxylin-eosin (H-E) at the Histology Core at Beth Israel Deaconess Medical Center. For the in vivo insulin signaling assay, after an overnight fast, insulin (10 units/kg for 10 min) or saline i.p. was given to WT and KO mice on HFD. After 10 min, various tissues were harvested and stored at –80°C until use. Tissue samples were homogenized in cell signaling lysis buffer containing protease inhibitors (Roche) and phosphatase inhibitors (Sigma-Aldrich) and subjected to Western blotting. All animal work was approved by the University of California Berkeley Animal Care and Use Committee.
Lentiviral overexpression vectors for JMJD8 and IRF3 were subcloned into pCDH at multicloning sites (MluI/NsiI), and hairpins targeting Jmjd8 were subcloned at AgeI/EcoRI. Single guide RNAs (sgRNAs) targeting Irf3 and human JMJD8 were cloned into lentiCRISPR v2 vectors. Briefly, lentiCRISPRv2 plasmid was digested and dephosphorylated with BsmBI. The sgRNA oligos were phosphorylated and annealed, followed by ligation into the digested lentiCRISPRv2 plasmid. Hairpin and sgRNA sequences are shown in Supplementary Table 2.
2-Deoxy-d-[2,6-3H]-glucose was purchased from PerkinElmer NEN Radiochemicals. Insulin, dexamethasone, isobutylmethylxanthine, rosiglitazone, thyroid hormone (triiodothyronine), glucose, 2-deoxyglucose (2-DG), and LPS came from Sigma-Aldrich. Mouse recombinant TNF-α was purchased from Millipore.
Antibodies were purchased from Cell Signaling Technology (9272, AKT; 3787, phosphorylated [p-]AKT [S473]; 2382, insulin receptor substrate 1 [IRS1]; 2118, GAPDH; 3033, p–NF-κB; 14269, histone H3; 4877, HSP90; 4302, IRF3; and 3025, insulin receptor [IR]), Santa Cruz Biotechnology (sc-515520, JMJD8; and sc-53566, Glut4), Sigma-Aldrich (F3165, Flag; and 4300653, phospho-IRS1 [S307]), Millipore (204359; NF-κB), Covance (MMS-101R, HA), GenScript (A01004; Ty1), and Invitrogen (44–816G, p-IRS1 [pY612]; and 44800G, p-IR [Y972]).
RNA Extraction and Quantitative PCR
Total RNA was extracted from tissues using TRIzol reagent according to the manufacturer’s instructions. cDNA was reverse transcribed from 1 μg of RNA using the cDNA High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Quantitative PCR (qPCR) was performed with SYBR Green qPCR Master Mix (AccuPower 2X; Bioneer Corporation) using a CFX96 Touch (Bio-Rad Laboratories) or QuantStudio 5 (Applied Biosystems). The relative amount of mRNA normalized to cyclophilin B was calculated using the delta–delta method. Primer sequences are listed in Supplementary Table 2.
Cell or tissue culture samples were lysed in RIPA lysis buffer containing 50 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 1% Triton X-100, 0.1% SDS, and protease inhibitor cocktail for 30 min on ice with rotating samples. Protein concentration was quantified by Bradford or BSA analysis. Proteins were size fractionated by SDS-PAGE and then transferred to polyvinylidene difluoride or nitrocellulose membrane and detected using the ECL Assay Kit (cat. no. NEL104001EA; PerkinElmer). Immunoblots were quantified by ImageJ.
Cells or tissue were homogenized with hypotonic lysis buffer (10 mmol/L HEPES [pH 7.9], 1.5 mmol/L MgCl2, 10 mmol/L KCl, 0.5 mmol/L dithiothreitol, and protease inhibitor cocktail) by 20 strokes of a loose-fitting pestle in a dounce homogenizer on ice. Supernatants were transferred into a new tube for cytosol-protein extraction after centrifugation at 2,000g for 10 min at 4°C, and the nuclear pellets were resuspended in 1 mL of hypotonic wash buffer and pelleted again by centrifugation at 2,000g for 1 min at 4°C. After the supernatants were removed, the washed nuclear pellets (retained as cleaner nuclei) were resuspended in 100 μL of RIPA lysis buffer and placed on ice for 30 min with occasional vortexing. The lysates were centrifuged at 15,000 rpm for 10 min at 4°C, and the supernatants were retained as high-purity nuclear proteins.
Cells or tissue were homogenized with hypotonic lysis buffer (50 mmol/L HEPES [pH 7.5], 1.5 mmol/L MgCl2, 10 mmol/L KCl, 1 mmol/L EDTA, 0.5 mmol/L dithiothreitol, and protease inhibitor cocktail) containing 0.04% Nonidet P-40 by 20 strokes of a loose-fitting pestle in dounce homogenizer on ice. Supernatants were transferred into a new tube for soluble cytosol-protein extraction after centrifugation at 2,000g for 10 min at 4°C, and the nuclear pellets were resuspended in 1 mL of hypotonic wash buffer and pelleted again by centrifugation at 2,000g for 1 min at 4°C. After the supernatants were removed, the washed nuclear pellets (retained as cleaner nuclei) were resuspended in 100 μL of RIPA lysis buffer and placed on ice for 40 min with occasional vortexing. The lysates were centrifuged at 15,000 rpm for 10 min at 4°C, and the supernatants were retained as chromatin fractionation.
Analysis of LPS and TNF-α–stimulated luciferase activity, driven by a synthetic NF-κB or Ccl5 promoter, was conducted. Cells were cultured overnight in 24-well plates and transfected with a liposome complex containing pGL3/luciferase reporter gene (50 ng), Renilla (5 ng), and other relevant plasmids (400 ng). After 48 h, cells were treated with LPS (1 μg/mL) or TNF-α (10 ng/mL) for 6 h, washed with PBS, and solubilized with lysis buffer. The luciferase activity of cell extracts was determined using the dual-luciferase reporter assay system from Promega. To determine transfection efficiency, luciferase activity was normalized to Renilla activity.
Adipose tissues of Jmjd8-KO and Adi-Jmjd8-TG mice were fixed with neutral-buffered formalin and embedded in paraffin. Fixed tissues were permeabilized with PBS containing 0.1% Triton X-100 for 10 min and then incubated with anti-MAC2 antibodies in PBS containing 10% (v/v) FBS for 18 h at 4°C temperature. The coverslide was washed and incubated with FITC-conjugated anti-rat IgG for 1 h. Staining was analyzed with a confocal LSM710 microscope (Carl Zeiss Microimaging Inc.).
After washing cells with ice-cold PBS, cells were lysed with 1 mL lysis buffer, containing 50 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 1% Triton X-100, and protease inhibitor cocktail, for 30 min on ice with rotating samples. The lysates were centrifuged at 15,000 rpm for 10 min at 4°C, and then supernatants were incubated with anti-FLAG antibody-conjugated beads (Sigma-Aldrich) at 4°C for 2 h. The beads were washed four times with lysis buffer. Bound proteins were eluted by boiling in SDS sample buffer and separated by SDS-PAGE. Immunoblotting was performed following the standard procedures of a Western blot.
Various types of adipocytes were incubated in serum-free DMEM for 4–6 h. Cells were then washed three times with KRH buffer (12 mmol/L HEPES [pH 7.4], 121 mmol/L NaCl, 5 mmol/L KCl, 0.33 mmol/L CaCl2, and 1.2 mmol/L MgSO4) and incubated for 20 min in KRH buffer in the absence or presence of 100 nmol/L insulin. Cells were treated with 2-deoxy-d-[2,6–3H]-glucose (0.33 mCi/mL) for another 10 min. Glucose uptake was stopped quickly by three rapid washes with KRH buffer containing 200 mmol/L glucose and 10 mmol/L cytochalasin B on ice. Cells were solubilized in 0.1% SDS for 30 min, and radioactivity was measured by liquid scintillation counting.
Glucose Tolerance Test or Insulin Tolerance Test
For the insulin tolerance test (ITT), mice were fasted for 6 h and i.p. injected with insulin (0.5 units/kg for chow diet and 0.8 units/kg for HFD). Blood glucose was measured at 0, 15, 30, 60, 90, and 120 min. For the glucose tolerance test (GTT), mice were fasted for 6 h and i.p. injected with glucose (1.5 g/kg). Blood glucose levels were measured at 0, 15, 30, 60, and 120 min using the Contour Next glucometer and glucose strips.
For the insulin ELISA, mice were fasted for 6 h. Plasma insulin was analyzed using an Ultrasensitive Mouse Insulin ELISA kit (cat. no. 90080; Crystal Chem). The end point calorimetric assays were performed using a Plate Reader SpectraMax i3. The IL-1β was analyzed using a mouse IL-1β uncoated ELISA kit (cat. no. 88-7013-22; Invitrogen). The end point calorimetric assays were performed using a Plate Reader SpectraMax i3.
Comprehensive Laboratory Animal Monitoring System Analysis
Metabolic rate was measured by indirect calorimetry in open-circuit Oxymax chambers, a component of the Comprehensive Laboratory Animal Monitoring System (CLAMS; Columbus Instruments). Mice were housed individually and maintained at 25°C under a 12-h light/12-h dark cycle. Activity was monitored in 1-min intervals of infrared beam breaks on the x-, y-, and z-axis and found not to be significantly different for any of the groups. Food and water were available ad libitum.
The adipocyte–macrophage coculture used the Transwell system (Corning, Corning, NY). 3T3-L1 adipocytes were cultured in six-well plates and differentiated at day 8, and macrophages (2 × 105 cells/well) were plated onto a Transwell insert containing a 0.4-μm polyethylene terephthalate membrane (Costar; Corning) in serum-free medium with or without the ligand (LPS). After incubating together for 24 h, the Transwell was removed, and 3T3-L1 adipocytes or macrophages were harvested for analysis.
RNA-Sequencing Library Generation and Analysis
RNA samples were extracted using the RNeasy Mini Kit (74104; Qiagen), and the quality of total RNA was assessed by the 2100 Bioanalyzer (Agilent Technologies) and agarose gel electrophoresis. Libraries were prepared using the BGI Library Preparation Kit, and sequencing was performed on the BGISEQ (Beijing Genomics Institute, Shenzhen, China). RNA-sequencing (RNA-seq) reads were aligned to the UCSC mm10 genome using STAR (37). rRNA and mitochondrial reads were removed. Differential expression analysis was done using edgeR (38), and differentially expressed genes upon KO were selected by fold change >1.5 and false discovery rate <0.05. Gene ontology analysis was performed using Enrichr (39). RNA-seq raw data are available in the Gene Expression Omnibus under accession number GSE164087.
Data from three independent experiments, carried out in triplicate, were combined, and the results were presented as the mean ± SE. Statistical differences were assessed using Prism 4 (GraphPad Software). Unpaired two-tailed Student t tests and two-way ANOVA were used. A P value <0.05 was considered statistically significant.
Data and Resource Availability
Data sets generated from current study are available from the corresponding author.
JMJD8 Is Most Abundant in White Adipose Tissues and Has Increased Expression Under Hyperinsulinemic Conditions
Many members of the JMJD protein family are active in adipose tissue. Therefore, we characterized the expression of the small members of JMJD family to investigate whether they are involved in the metabolic regulation of adipose tissue. Specifically, we characterized their expression in epididymal white adipose tissue (eWAT), a major visceral depot, from obesity-associated insulin-resistance models. Remarkably, Jmjd8 was the only member with differential expression, increasing by three- and fivefold in diet-induced obesity and genetically obese ob/ob mouse models, respectively, compared with its expression in lean WT littermate controls (Fig. 1A and B). We confirmed that JMJD8 protein levels also increased in eWAT and inguinal WAT (iWAT) in both insulin-resistant models (Fig. 1C–F) to a similar degree as the transcript levels. Next, we studied whether adipose Jmjd8 expression is regulated by nutritional state. Interestingly, 24 h of fasting resulted in a twofold reduction in Jmjd8 mRNA expression in eWAT, and subsequent refeeding caused upregulation (Fig. 1G). Since insulin is a major hormone that increases after feeding, and obesity is strongly associated with hyperinsulinemia, we investigated whether insulin regulated Jmjd8. Indeed, insulin injection significantly increased adipose Jmjd8 expression in C57BL/6 WT mice, especially when animals were challenged by an HFD (Fig. 1H). In addition, we asked whether inflammatory signals (i.e., TNF-α, LPS, and palmitate) increase JMJD8 levels, but none of them did (Supplementary Fig. 1A and B) Next, we studied the tissue distribution of JMJD8. Importantly, JMJD8 was expressed highest in adipose tissue, especially in white adipose depots, compared with other tissues, including liver and muscle (Fig. 1I). In addition to mature adipocytes, adipose tissue is comprised of multiple other cell types, including preadipocytes and macrophages, which are collectively termed the stromal vascular fraction (SVF). We thus compared Jmjd8 levels between fractionated SVF versus the adipocyte fraction of eWAT and iWAT from lean C57BL/6 WT mice. We detected approximately five times higher enrichment in the adipocyte fraction compared with the SVF isolated from both depots (Fig. 1J). Congruent with this, Jmjd8 expression was significantly induced at the differentiated stage of 3T3-L1, immortalized inguinal adipocytes, and brown adipocytes, compared with the preadipocyte stage of these models (Fig. 1K–M). Together, the enriched expression of JMJD8 in adipose tissue, especially in adipocytes, and the increased adipose expression under hyperinsulinemic conditions led us to hypothesize that JMJD8 regulates adipocyte metabolic function.
JMJD8 Inhibits Insulin Sensitivity in Adipocytes
Since JMJD8 levels increase during adipogenesis, we first examined whether JMJD8 regulates adipogenic potential by performing gain- and loss-of-function studies prior to differentiation. Knockdown of Jmjd8 slightly reduced lipid accumulation, while overexpression of Jmjd8 increased it, as assessed by Oil Red O staining (Supplementary Fig. 2A and B). Jmjd8 knockdown adipocytes had reduced expression of Pparg and Fabp4, but not Adipoq and Lep (Supplementary Fig. 2C). In contrast, Jmjd8 overexpressor adipocytes did not show marked changes in the expression of adipocyte markers (Supplementary Fig. 2A–C). Since Jmjd8 knockdown mildly affects adipogenesis, we sought to test the role of JMJD8 in the development of insulin resistance by knocking down Jmjd8 expression in fully differentiated 3T3-L1 adipocytes at adipogenic day 8, achieving a 73% knockdown (Supplementary Fig. 3A and B), and no dedifferentiation was noted. Remarkably, Jmjd8 knockdown increased insulin sensitivity by approximately twofold, as assessed by insulin-stimulated 2-deoxy-d-glucose uptake (Fig. 2A). Similar results were obtained when we knocked down Jmjd8 in adipocytes differentiated from immortalized preadipocytes from the inguinal depot of lean WT mice (Fig. 2B). To see if the function of JMJD8 is conserved in humans, we used SGBS cells, a widely used human preadipocyte cell line that was derived from a human patient with SGBS (40), and knocked down JMJD8 using gRNA, thus achieving 50% knockdown efficiency (Supplementary Fig. 3C). Similar to mouse adipocytes, JMJD8 knockdown significantly increased insulin-stimulated glucose uptake in SBGS adipocytes (Fig. 2C). Conversely, we conducted gain-of-function studies to determine if Jmjd8 expression is sufficient to cause insulin resistance. We overexpressed JMJD8 in fully mature 3T3-L1 adipocytes, primary inguinal adipocytes, and human SGBS adipocytes (Supplementary Fig. 3C–E), which resulted in a 30–50% reduction in insulin-stimulated glucose uptake compared with control cells (Fig. 2D–F). Together, our results indicate that JMJD8 plays a causal role in the development of insulin resistance in mouse and human adipocytes in a cell-autonomous manner.
Jmjd8 KO Mice Are Protected From Diet-Induced Insulin Resistance and Glucose Intolerance
To confirm that JMJD8 plays a role in adipose and whole-body insulin sensitivity in vivo, we generated whole-body Jmjd8 KO mice using the CRISPR/Cas9 system on the C57BL/6 background (Supplementary Fig. 4A). We validated complete Jmjd8 KO by qPCR and Western blot in multiple tissues, including adipose (Fig. 3A and B). The KO mice were viable and showed normal growth compared with their WT littermate controls. On a chow diet, glucose tolerance was not different between genotypes (Fig. 3C and D), but KO mice exhibited increased insulin sensitivity (Fig. 3E and F). In addition, KO mice had a dramatic reduction in both fasting and fed insulin levels (Fig. 3G) and a lower HOMA of insulin resistance (HOMA-IR) (Fig. 3H). This was without alterations in body weight (Supplementary Fig. 4B) or body composition, as assessed by EchoMRI and tissue weight (Supplementary Fig. 4C and D). Together, these data suggest that Jmjd8-KO mice have increased whole-body insulin sensitivity on a chow diet without changes in energy homeostasis.
Since we found increased JMJD8 expression in response to high-fat feeding, we tested whether Jmjd8 deficiency confers protection from diet-induced insulin resistance. For that, a cohort of mice were placed on an HFD (58% calories from fat), and we found a trend toward KO mice gaining less weight, but the difference did not reach statistical significance (Fig. 4A). Consistent with this, EchoMRI studies indicated that there were no marked differences in body composition (Fig. 4B) or the mass of various adipose tissues, including eWAT, iWAT, and mesenteric WAT (Fig. 4C). To further assess energy homeostasis, the indirect calorimetry of a cohort of high-fat–fed mice was analyzed using CLAMS. We confirmed that KO mice did not display a significant change in their energy balance, locomotor activity, or food intake (Supplementary Fig. 5). Despite no major changes in energy homeostasis, remarkably, Jmjd8-KO mice on an HFD exhibited a significant improvement in glucose tolerance (Fig. 4D and E) and insulin sensitivity compared with controls (Fig. 4F and G). Similar to the chow-fed cohort, KO mice had a dramatic reduction in serum insulin levels in both fed and fasting conditions and improved HOMA-IR compared with WT mice (Figs. 4H and I).
To better assess insulin responsiveness in various metabolic tissues, we performed in vivo insulin signaling assays on tissues from WT and Jmjd8-KO mice on an HFD. After fasting overnight, animals received a dose of bolus insulin (10 U/kg, 10 min) via i.p. injection, and then the tissues were collected and subjected to immunoblotting analysis for the phosphorylated and total protein fractions of insulin signaling components. We noted that phosphorylation levels associated with the active forms of IRS1, IR, and AKT were increased in the KO eWAT when normalized to total protein levels (Fig. 4J, K, N, and O). Interestingly, KO eWAT showed reduced phosphorylation of IRS1 at serine 307 (Fig. 4J and L), which was shown to be implicated in the counterregulatory feedback of insulin signaling by enhancing IRS1 protein degradation (41–43). We indeed noted a trend toward increased protein expression of IRS1 in the KO eWAT (Fig. 4M). By contrast, we did not observe significant changes in the active phosphorylation of AKT in other metabolic tissues (Supplementary Fig. 6A–F). Moreover, pyruvate tolerance was indistinguishable between genotypes, suggesting that there was no hepatic contribution to the improved glucose homeostasis of KO mice (Supplementary Fig. 6G). To verify improved insulin sensitivity in KO adipocytes, we conducted ex vivo insulinstimulated glucose uptake assays on WT versus KO mouse adipocytes fractioned using collagenase digestion and found that KO adipocytes exhibit an approximately twofold increase in insulin-stimulated glucose uptake (Fig. 4P). Lastly, we measured the mRNA and protein expression of insulin-dependent glucose transporter GLUT4 (Slc2a4) and noted, overall, increased expression in the KO eWAT (Fig. 4Q–S), which is likely to contribute to the improved insulin sensitivity. Together, our results suggest that Jmjd8 deficiency leads to improved adipose and systemic insulin sensitivity by involving improved insulin signal transduction and increased GLUT4 expression in adipose tissue.
Adipose-Specific Jmjd8 Overexpression Exacerbates Diet-Induced Glucose Intolerance and Insulin Resistance
Next, we performed JMJD8 gain-of-function studies in vivo. We generated Jmjd8 transgenic mice on a C57BL/6 background, in which Jmjd8 transgene expression was induced in a Cre-dependent manner (tagged by the Ty1 tag) (Supplementary Fig. 7). These mice were crossed to adiponectin Cre (44) mice (Supplementary Fig. 7) to generate adipocyte-specific overexpressor mice (Adi-Jmjd8-TG). As a result, we achieved a physiological level of overexpression (two- to fivefold) in iWAT, eWAT, and brown adipose tissue but not in other tissues (Fig. 5A and B). Adi-Jmjd8-TG mice were viable and showed normal growth compared with their littermate controls. On a chow diet, there were no marked differences in body weight, body composition, or tissue mass between genotypes, and no obvious differences were detected between genotypes using a GTT or ITT (not shown). On an HFD, there were no significant changes in body weight or body composition compared with controls (Fig. 5C–E) and no marked difference in energy homeostasis, assessed by CLAMS analysis, between genotypes (Supplementary Fig. 8). Without major advantages in body weight or adiposity, Adi-Jmjd8-TG mice on an HFD exhibited worsened glucose tolerance and insulin sensitivity (Fig. 5F–I), accompanied by higher insulin levels and higher HOMA-IR compared with control WT mice on an HFD (Fig. 5J and K). In addition, we measured the mRNA and protein expression of GLUT4 and found its expression level was decreased in the Adi-Jmjd8-TG eWAT (Fig. 5L–N). Consistent with in vivo results, we noted that Jmjd8 overexpression in 3T3-L1 adipocytes overall show reduced insulin signal transduction in adipocytes (Fig. 5O–T). Together, these results suggest that adipocyte-specific JMJD8 overexpression exacerbates systemic insulin resistance when challenged by HFD.
Jmjd8 Deficiency Improves Adipose Tissue Inflammation in an Adipocyte-Autonomous Manner
To understand the downstream molecular mechanisms by which JMJD8 mediates adipose insulin resistance, we profiled the adipose tissue transcriptome of WT and KO mice fed an HFD for 2 months. We identified 151 and 302 genes that were up- and downregulated in the KO tissues, respectively (Fig. 6A and Supplementary Table 1). To determine if specific biological processes and functions were affected by Jmjd8 deficiency, gene ontology analysis was performed. While no gene ontology terms were found to be significantly enriched in upregulated genes (not shown), inflammation and chemotaxis pathways were enriched in downregulated genes (Fig. 6B). Also, we confirmed reduced expression of several top-regulated genes and additional genes involved in inflammation, including Il1b, Ccl2, Il6, Tnf, and Lcn2, in the KO eWAT (Fig. 6C). Conversely, those inflammatory genes were coordinately upregulated in obese Jmjd8 transgenic eWAT compared with control tissue (Fig. 6D). A reduced serum level of IL-1b was also confirmed by ELISA in obese KO mice compared with WT mice (Fig. 6E). Histological studies found that macrophage infiltration was reduced in obese Jmjd8 KO eWAT compared to obese control tissue, as indicated by reductions in the numbers of crownlike structures and fewer cells stained with a macrophage marker (Mac-2), as detected by H-E and immunofluorescence staining (Fig. 6F and H). Conversely, macrophage infiltration was increased in obese Jmjd8-TG eWAT compared with obese control tissue (Fig. 6G and I). We also saw a concordant change in the expression of Cd11b, one of the surface markers for macrophages (Supplementary Fig. 9A). Macrophage polarity often shifts in obesity from alternative M2 to classical M1, thus promoting proinflammatory signaling via secreting proinflammatory cytokines. Obese Jmjd8-KO eWAT had increased expression of M2 markers (i.e., Tgfb1, ll10, Arg1, and Mgl2) (Supplementary Fig. 9A) in parallel with reduced expression of M1 markers (i.e., Il1b, Tnf, Ccl2, and Il6) compared with controls, implying that there might be a macrophage polarization toward M2. Although obese Jmjd8-TG eWAT had increased expression of M1 markers, there were no major differences in the expression of M2 markers (Supplementary Fig. 9B).
Since we used whole fat tissue for gene profiling and that the gene deletion occurs in all cells, we cannot exclude the possibility that some of the transcriptional changes came from immune cells such as macrophages. To dissect out the specific role of JMJD8 in inflammation in adipocytes versus macrophages, we conducted the following experiments. In order to determine the adipocyte autonomous role of JMJD8 in the regulation of inflammation, we measured a subset of inflammatory genes after altering JMJD8 levels in 3T3-L1 adipocytes. Jmjd8 knockdown suppressed the mRNA expression of several inflammatory genes (Fig. 6J), whereas Jmjd8 overexpression increased them (Fig. 6K). Next, we investigated if there is cross talk between these two cell types by conducting coculture experiments. First, we cocultured WT RAW264.7 macrophages in the upper chamber of the Transwell system and 3T3-L1 adipocytes with altered Jmjd8 expression in the lower chamber (Supplementary Fig. 10A). Importantly, WT macrophages cocultured with conditioned media from the JMJD8-overexpressing 3T3-L1 adipocytes had increased expression of inflammatory genes and reduced expression with Jmjd8 silenced adipocytes (Supplementary Fig. 10B and C). We also conducted the converse experiment by coculturing WT 3T3-L1 adipocytes with bone marrow–derived Jmjd8 KO and WT macrophages (Supplementary Fig. 10D). In this case, there were no discernable changes in the expression of inflammatory genes in adipocytes (Supplementary Fig. 10E) and no discernable change in insulin-stimulated glucose uptake (Supplementary Fig. 10F). Together, these results suggest that JMJD8’s regulatory role in inflammation and insulin resistance primarily stems from adipocytes, not macrophages.
JMJD8 Is Required for LPS-Mediated Inflammatory Gene Induction Through Functional Interaction With IRF3
Given JMJD8’s effect on inflammatory genes and insulin resistance, we tested whether Jmjd8 knockdown prevents insulin resistance mediated by inflammatory signals. Jmjd8 was knocked down in mature 3T3-L1 adipocytes using hairpins against Jmjd8 or control and treated with LPS or vehicle plus ATP. Interestingly, Jmjd8-knockdown adipocytes were selectively protected from the effects of LPS/ATP (Fig. 7A). An analogous experiment was performed using human SGBS adipocytes, and similar results were found (Fig. 7B). In line with these results, Jmjd8 KO mouse and human adipocytes showed attenuated LPS-mediated inflammatory gene stimulation, whereas JMJD8 overexpression enhanced the response (Fig. 7C and D). By contrast, Jmjd8 knockdown did not have a discernable effect on either TNF-α–mediated insulin resistance or its inflammatory gene stimulation (Supplementary Fig. 11). Together, our data indicate that JMJD8 is necessary in LPS-mediated, but not TNF-α–mediated, inflammation and insulin-stimulated glucose uptake.
Given that JMJD8 is necessary for LPS-mediated insulin resistance, we investigated whether JMJD8 alters downstream signaling pathways upon LPS stimulation. Adipocytes recognize LPS through TLR4 and subsequently initiate two arms of inflammatory pathways that depend on the MyD88 and TRIF signaling pathways (45). MyD88-dependent signaling activates NF-κB–dependent events, whereas TLR4/TRIF-dependent signaling results in TANK-binding kinase 1 (TBK1) and IκB kinase ε (IKKε) activation and triggers IRF3-dependent responses (45). Importantly, these two transcription factors activate distinct but overlapping sets of target genes and, in some cases, may cooperate functionally and physically to activate inflammatory genes (46,47). We did not detect any marked changes in the canonical regulation of NF-κB caused by JMJD8 alterations, as evidenced by the lack of clear changes in nuclear localization in both obese Jmjd8-KO and -TG eWAT compared with control tissues (Supplementary Fig. 12A). We did not detect changes in the levels of total or Ser536 phosphorylated NF-κB during Jmjd8 knockdown or overexpression in the presence or absence of LPS stimulation (Supplementary Fig. 12B–G), and there was no physical interaction between JMJD8 and the upstream regulators of NF-κB signaling, such as IκB or IKKβ (Supplementary Fig. 12H and I). Lastly, JMJD8 did not alter TNF-α– or LPS-driven transcriptional activity when assessed by NF-κB reporter assays (Supplementary Fig. 12J and K).
Next, we sought to determine whether JMJD8 is involved in IRF3-mediated signaling. Remarkably, we found that JMJD8 coimmunoprecipitated with IRF3 (Fig. 8A), and thus, we performed epistasis experiments to dissect the functional interaction between JMJD8 and IRF3. First, we determined whether JMJD8 requires IRF3 to mediate its actions by introducing Irf3 gRNA and/or WT Jmjd8 into mature L1 adipocytes. Strikingly, we found that JMJD8 was no longer able to inhibit insulin-stimulated glucose uptake without IRF3 (Fig. 8B). In line with this, JMJD8’s ability to enhance LPS-stimulated induction of inflammatory genes, especially Tnf and Il6, was completely abolished by losing IRF3 (Fig. 8C–E). Next, we asked whether IRF3 requires JMJD8 to promote inflammation and insulin resistance in the presence and absence of LPS by introducing the expression vectors for Jmjd8 hairpins and WT Irf3 in mature L1 adipocytes. We found that IRF3’s ability to inhibit insulin-stimulated glucose uptake or IRF3’s ability to induce inflammatory genes was marginally affected by losing JMJD8 (Supplementary Fig. 13A–D). Altered JMJD8 expression also did not affect the chromatin localization of IRF3 (Supplementary Fig. 13E) or the IRF3-driven transcriptional activity when assessed by IRF3 reporter assay (Supplementary Fig. 13F). Together, our data suggest JMJD8 requires IRF3 to mediate insulin resistance and inflammation in adipocytes and not the reverse.
We propose a model in which chronic overnutrition increases the adipocyte level of JMJD8, which then interacts with IRF3 to activate proinflammatory genes, thus causing the recruitment of proinflammatory immune cells and adipocyte insulin resistance.
Adipose tissue inflammation, characterized by immune cell infiltration and cytokine secretion, has been implicated as a major player in both the onset of metabolic syndrome and the associated pathophysiological consequences, including insulin resistance (1,48). During these processes, a dynamic interplay operates between adipocytes and immune cells. Although it has been suggested that adipocytes themselves play an important role in sensing and triggering inflammation and attracting immune cells (24,26,27,29), there is a paucity of studies that elucidate the molecular machinery that mediates inflammation during overnutrition. In this study, we demonstrate that JMJD8 acts as a molecular nexus that mediates adipocyte inflammation and insulin sensitivity.
We found that JMJD8 expression was highly enriched in adipose tissue, especially in the adipocyte fraction, relative to other tissues. The expression pattern of adipose JMJD8 in insulin resistance models and its dynamic regulation in response to nutritional states and insulin levels led us to explore the metabolic function of JMJD8 in adipose insulin sensitivity. We demonstrated that JMJD8 was both necessary and sufficient to mediate insulin resistance in both mouse and human adipocytes. Moreover, we proved the in vivo role of adipose JMJD8 using KO and transgenic mouse models. Our RNA-seq studies revealed that insulin resistance in Jmjd8-KO mice was accompanied by reduced inflammation. We then showed that JMJD8 gain- and loss-of-function coordinately regulated the transcriptional program of several inflammatory genes in adipocytes. Furthermore, we demonstrated that JMJD8 played a necessary role in LPS-stimulated inflammation and insulin resistance through a functional interaction with IRF3.
JMJD8 is a small-molecular-weight JMJD protein (36). While some JMJD proteins with large molecular weights possess histone demethylase activity, it does not seem to be the case with JMJD8. Its JmjC domain has a noncanonical HXHXnH motif compared with other JmjC proteins with HX(D/E)XnH (30). Moreover, JMJD8 does not contain any other functional domain, such as C2H2/C5H2 or AT-hook motifs, that are implicated in binding or interacting with DNA (49). By contrast, a recent study proposed that JMJD8 may act as a transcriptional repressor, as ectopic expression of an N-terminal truncation mutant localizes to gene promoters and represses them (50). However, so far, there is no report describing the existence of an endogenous form of JMJD8 with an N-terminal truncation, raising doubts about such JMJD8 nuclear activity. In support of this, our localization analysis found JMJD8 mainly in the cytosol under basal conditions and after stimulation by TNF-α and LPS (not shown). Therefore, we postulate that the repressive role for JMJD8 in the transcriptional program of proinflammatory cytokines is indirect and is likely via interfering with a signaling cascade in the cytosol. A recent proteomics study reported that JMJD8 contains N-terminal sequence that enables it to localize to the luminal endoplasmic reticulum (ER) (51), where it can interact with ER proteins involved in protein folding and complex formation. However, we found that JMJD8 does not colocalize with Sec61, a channel protein located in the ER. This discrepancy might be ascribed to differences in the mouse and human orthologs between the studies.
The early events in the development of adipose inflammation are not well understood. One of the proposed mechanisms is that the innate immune pathways are directly activated by nutritional signals such as free fatty acids (52,53) and endotoxin (54). Indeed, there are pattern recognition receptors like TLR2 and TLR4 in adipocytes (24,26). The stimulation of TLR4 by LPS binding activates multiple signaling components, including NF-κB, AP-1, and IRF3, and causes the subsequent production of proinflammatory cytokines (54). Recent studies suggest that JMJD8 regulates the NF-κB pathway, but the results seem to be inconsistent. While two studies suggest that JMJD8 acts as a positive regulator of TNF-α–stimulated NF-κB signaling in HEK293T and colorectal cancer cell lines such as H1299 (55,56), another study reported that JMJD8 negatively regulates NF-κB activity in H1299 cells (57). In our studies, we noted that JMJD8 does not alter the canonical regulation of NF-κB, as evidenced by: 1) the lack of clear changes in nuclear localization, 2) similar levels of total and Ser536-phosphorylated NF-κB among obese Jmjd8-KO, Jmjd8-TG, and control eWAT, 3) no physical interaction between JMJD8 and upstream regulators of NF-κB signaling, such as IκB or IKKβ, and 4) the lack of effect on NF-κB–driven transcriptional activity.
The other arm of LPS-stimulated TLR signaling leads to the activation of IRF3 by inducing localization to the nucleus, where it acts as a transcription factor. Thus, we also examined the signaling arm of IRF3, finding a physical interaction between IRF3 and JMJD8 in 293T cells. Interestingly, we noted that IRF3 is important for JMJD8 activity, as evidenced by JMJD8 dramatically losing the ability to activate LPS-mediated inflammatory gene regulation and insulin resistance after losing IRF3. On the contrary, IRF3’s actions were marginally affected when losing JMJD8. This arguably suggests that JMJD8 requires IRF3 to cause insulin resistance and inflammation, but not the reverse, at least in vitro. It is noteworthy that several functions for IRF3 have been reported in signaling within the cytoplasm, which are independent of its nuclear function. For example, activated IRF3 inhibits TGF-β/Smad signaling (58), thus preventing the differentiation of induced regulatory T cells in colons and the epithelial-to-mesenchymal transition of tumor cells. A recent study also reported that IRF3 associates with active β-catenin in the cytoplasm, thus inhibiting its nuclear translocation and cell proliferation of colorectal tumor cells (59). Thus, it is possible that IRF3 regulates JMJD8 signaling in the cytoplasm. Future studies will be required to determine how IRF3 affects JMJD8-mediated LPS signaling.
Inflammatory signals negatively impact insulin signaling, for instance, by causing inhibitory serine phosphorylation at IRS1, which can lead to an increased IRS1 protein degradation (41–43). Consistent with this notion, Jmjd8-KO eWAT in high-fat–fed mice has greatly improved insulin signal transduction, caused reduced S612 phosphorylation of IRS1 and increased IRS1 protein levels, and, conversely, JMJD8-overexpressing adipocytes showed reduced insulin signal transduction. It is worth nothing that recent studies have indicated that JMJD8 regulates AKT signaling in cancer cell lines (57,60), although the results are opposing among the studies. Nevertheless, it appears that JMJD8 plays an important role in regulation of phosphorylation of AKT and other components of insulin signaling cascade. Whether this regulation is direct or indirect remains to be determined.
In summary, we identified JMJD8 as an important molecular regulator of inflammation and insulin sensitivity in adipocytes. Identifying the function of JMJD8 in adipose metabolic health may provide novel therapeutic targets for metabolic disorders including obesity and T2D.
This article contains supplementary material online at https://doi.org/10.2337/figshare.16810945.
Acknowledgments. The authors thank Dr. Evan Rosen (Harvard Medical School) for sharing the IRF3 overexpression plasmid and Drs. Hei Sook Sul, Jen-Chywan Wally Wang (University of California Berkeley), and Evan Rosen for helpful conversations about the manuscript.
Funding. Work was funded by American Heart Association award number 19POST34380834 to D.Y. and National Institute of Diabetes and Digestive and Kidney Diseases grant R01 DK116008 and Stanford Diabetes Research Center Pilot and Feasibility Program grant P30DK116074 to S.K.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. S.K. supervised experiments and wrote the manuscript. D.Y. drafted the Research Design and Methods section and figure legends. Experiments were carried out by D.Y., B.C.J., S.D.V., and S.K. H.-W.L. analyzed RNA-seq. S.K. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.