Glutamine Production by Glul Promotes Thermogenic Adipocyte Differentiation Through Prdm9-Mediated H3K4me3 and Transcriptional Reprogramming

Obesity has become a global epidemic that contributes to the development of multiple metabolic comorbidities, including diabetes, fatty liver diseases, dyslipidemia, cardiovascular diseases, neuromuscular diseases, and some tumors. Although the underlying mechanisms of obesity are not yet clear, adipose tissue is known to play an essential role in this pathogenetic process through regulating insulin sensitivity, lipid metabolism, thermogenesis, and immune response (1,2). Classically, adipose tissue can be divided into white adipose tissue (WAT) and brown adipose tissue (BAT), as defined by morphology and function. While white adipocytes contain large unilocular lipid droplets and few mitochondria, brown adipocytes (BACs) feature numerous small multilocular lipid droplets and a high density of mitochondria, indicating their high capacity for fuel metabolism and thermogenesis (3). In the past few years, tremendous advances have been made in the identification and characterization of a new type of adipocyte named beige adipocyte (BeAC) (or brite adipocyte). BeACs show similarity with BACs in terms of morphologic and biochemical characteristics and scattered distributed in WAT (4,5). Both BAT and beige adipose tissue (BeAT) are capable of thermogenesis and are considered the primary sources of nonshivering thermogenesis in mammals. Physiologically, upon stimulation of specific stimuli, preadipocytes can be committed to differentiate into mature BACs and BeACs through de novo adipogenesis (6). In addition, BeACs can directly arise from mature white adipocytes through a process called transdifferentiation (7). When stimulated by cold exposure (8), catecholamines, or thiazolidinediones (9), these thermogenic adipocytes are activated, maximizing the expression of thermogenic genes, such as Ucp1, Prdm16, Dio2, and Cidea, to conduct the thermogenic and energy-consumption function. During the processes mentioned above, epigenetic mechanisms have been demonstrated to play key roles (10–13). Abnormal development and impaired thermogenic function of BATs and BeATs, resulting in dysregulation of whole-body energy expenditure, are important causes of obesity and insulin resistance (14–17). As such, the thermogenic adipocyte has been considered an attractive target for treating obesity and associated metabolic disorders.

Glutamine is well known as a carrier of nitrogen transport and a source of carbon and energy (18). Notably, glutamine has recently caught greater attention since its emergence as a pivotal player in immune and metabolic regulation. Glutamine metabolism is important for alternative activation of macrophages (19) and has distinct roles in promoting helper T-cell 17 (Th17) differentiation but constraining Th1 and cytotoxic T lymphocyte effector cell differentiation (20). Glutamate-ammonia ligase (Glul), also called glutamine synthetase, is the only enzyme that ligates glutamate and ammonia to form glutamine. In obese and diabetic conditions, the loss-of-function mutation of the Glul gene is positively correlated with an increased risk of coronary diseases (21), and decreased plasma glutamine levels are associated with the increased incidence of type 2 diabetes (22). Supplementation of glutamine in type 2 diabetic rats has been shown to enhance glucagon-like peptide 1 secretion (23) and prevents apoptosis of islet cells (24). Recent studies have demonstrated that posttranslational modifications and epigenetic mechanisms are involved in regulating biological processes by glutamine (25–29). It has been reported that obesity-associated reduction of glutamine in WAT increases O-GlcNAcylation of nuclear proteins, leading to transcriptional induction of proinflammatory genes and meta-inflammation in WAT (25). Local glutamine deficiency in the solid tumor core regions leads to cancer cell dedifferentiation and therapeutic resistance through histone hypermethylation (26). Glutamine metabolism has also been shown to regulate osteoblast and adipocyte specification in skeletal stem cells and bone formation via the intermediate product α-ketoglutarate (α-KG) (27). Furthermore, it has been demonstrated that intracellular glutamine-derived α-KG helps to maintain the pluripotency of embryonic stem cells by regulating multiple epigenetic modifications, including H3K27me3 and DNA demethylation (28). Intriguingly, a recent study demonstrated that glutamine catabolism–derived glutamate promotes microtubule glutamylation and cancer metastasis. However, the role and underlying mechanism of Glul-mediated glutamine metabolism in the regulation of thermogenic adipocyte differentiation and its function in the development of obesity and associated metabolic dysfunction remain to be defined.

In this study, through transcriptomic analysis of human subcutaneous WAT (scWAT), we found that GLUL gene expression was robustly downregulated in the scWAT from overweight human subjects compared with normal control subjects. This observation was further confirmed in thermogenic fat depots from obese and diabetic mouse models. Importantly, genetic, pharmacologic, or metabolic manipulations of Glul and glutamine levels in cell culture and mouse models revealed that glutamine is critical for thermogenic adipocyte differentiation and thermogenesis and systemic energy homeostasis through transcriptional reprogramming of gene networks associated with adipogenesis and thermogenesis. Further studies using a combination of RNA sequencing (RNA-seq), assay for transposase-accessible chromatin with sequencing (ATAC-seq), and cleavage under targets and tagmentation sequencing (CUT&Tag-seq) revealed that histone modification–mediated epigenetic mechanisms play a critical role in the transcriptional orchestration of adipogenic and thermogenic gene programs by glutamine. Specifically, we identified Prdm9, an H3K4 methyltransferase, as a key component in this regulatory process. Prdm9 is induced during BAC differentiation and glutamine supplementation. Importantly, Prdm9-mediated H3K4me3 and chromatin remodeling play a key role in glutamine-triggered transcriptional induction of adipogenic and thermogenic gene programs and adipocyte thermogenesis. Furthermore, we identified C/EBPb as a key transcription factor in glutamine-induced gene expression of Prdm9 through recruiting to its enhancer region.

The scWAT and visceral WAT (vWAT) from human subjects were investigated in this study. The studies on human scWAT and vWAT were approved by The Second Affiliated Hospital of Zhejiang University School of Medicine (approval no. 2020-528) and The Second Affiliated Hospital of Soochow University (approval no. JD-LK-2020-038–01), respectively. Only the subjects who provided informed written consent were included in the study. All subjects underwent medical history inquiries before hospitalization and clinical chemistry analyses after an overnight fast. Human subject scWAT samples were collected from the preperitoneal fat layer of patients who underwent surgery with oblique inguinal hernia, incisional hernia, or appendicitis at The Second Affiliated Hospital Zhejiang University School of Medicine during 2019–2020. The patients’ characteristics are described in Supplementary Fig. 1B. Human vWAT used in this study was obtained from the omental fat tissue of patients who underwent surgery for cholecystitis, gallstone, and oblique inguinal hernia at The Second Affiliated Hospital of Soochow University during 2019–2020. The patients’ characteristics are not shown and are readily available upon request. Patients with secondary or syndromic obesity and lean subjects with previously diagnosed diabetes or impaired glucose regulation were excluded. All the samples were dissected and immediately frozen in liquid nitrogen for RNA analyses and metabolite detection by liquid chromatography tandem mass spectrometry using a quadrupole time-of-flight mass spectrometer (Synapt XS; Waters) as previously described (30,31).

Wild-type (WT) mice on the C57BL/6J background were generated by GemPharmatech Co. Ltd. (Nanjing, China). The ob/ob (leptin-deficient) and db/db (leptin receptor-deficient) mice were purchased from The Jackson Laboratory. All animal studies were performed following the protocols approved by the committee on use and care of animals at Zhejiang University and conducted in accordance with the institutional guidelines on the care and use of laboratory animals. Mice were housed in 12/12-h light/dark cycles and fed a regular chow diet. Male mice were used in all the experiments unless otherwise indicated. Cl316,243 (1499; Tocris Bioscience) was used to treat WT mice for 7 days by intraperitoneal (IP) injection. WT mice received 1 g/kg daily l-alanyl-l-glutamine (Ala-Gln) (ST2041; Beyotime) by gavage. The Ala-Gln and Ala (A7627; Sigma-Aldrich) were used to treat WT mice 6% by drink with high-fat diet (HFD) feeding. L-methionine sulfoximine (MSO) (5 mg/kg, M5379; Sigma-Aldrich) was used to treat WT mice through IP injection three times a week. After experiments, mice were sacrificed, then tissues were harvested and immediately frozen in liquid nitrogen for later RNA and protein analyses and metabolite detection by liquid chromatography tandem mass spectrometry. Glucose tolerance test, immunoblotting analysis, cold exposure assay, hematoxylin-eosin staining of murine BAT tissue, immunohistochemistry of murine BAT tissue, and RNA isolation and quantitative RT-PCR (RT-qPCR) are described in the Supplementary Materials.

Brown preadipocytes were isolated and immortalized as previously described (32). C3H10T1/2 (10T1/2) cells and 3T3L1 cells were obtained from ATCC. Cells were cultured in DMEM (11995065; Gibco) with 10% (v/v) FBS (SE100-011; VISTECH) in a humidified incubator containing 5% CO₂ at 37°C. To establish BAC-10T1/2-Glul knockout (KO) cell lines, cells were transduced with lentivirus expressing control guide RNA (gRNA) or gRNA targeting Glul and subjected to puromycin selection. After reaching 100% confluency for 2–3 days, BAC and 10T1/2 differentiation were induced by replacing the culture medium with induction medium (DMEM containing 10% FBS, 0.5 mmol/L 3-isobutyl-l-methylxanthine, 125 μmol/L indomethacin, 1 μmol/L dexamethasone, 1 nmol/L T3, and 0.1 μg/mL insulin). Two days after induction, cells were cultured in a differentiation medium (DMEM containing 10% FBS plus 1 nmol/L T3 and 0.1 μg/mL insulin). 3T3L1 was induced by replacing the culture medium with an induction medium (DMEM containing 10% FBS, 0.5 mmol/L 3-isobutyl-l-methylxanthine, 0.25 μmol/L dexamethasone, and 5 μg/mL insulin). Two days after induction, cells were cultured in a differentiation medium (DMEM containing 10% FBS plus 1 μg/mL insulin). The cells were switched to a fresh differentiation medium every 2 days until fully differentiated. For conditioned media with glutamine deprivation during differentiation, glutamine-free DMEM (10313021; Thermo Fisher Scientific) was used; for media with glutamine, extra l-glutamine (25030081; Thermo Fisher Scientific) was added, and the concentration of glutamine was 4 mmol/L. MSO at 1 mmol/L, l-glutamate (G1251; Sigma-Aldrich) at 0–100 mg/L, α-KG (349631; Sigma-Aldrich), and MRK-740 (6803; Tocris Bioscience) at 10 μmol/L were used in cell differentiation. The plasmid construction, seahorse assay, luciferase reporter assay, siRNA transfection, and the procedures of RNA-seq, ATAC-seq and CUT&Tag-seq are described in the Supplementary Materials.

All statistical analyses were performed using GraphPad Prism 8 software. Data are represented mean ± SD or mean ± SEM. An unpaired, two-tailed Student t test was used to compare two groups. ANOVA and appropriate post hoc analyses were used for comparisons of more than two groups. For human studies, data are presented as mean ± SD. Pearson product moment correlation and Spearman rank correlation test were applied for the correlation analyses. Multiple linear regression was applied when BMI correlation was made. P < 0.05 was considered statistically significant.

The RNA-seq, ATAC-seq, and CUT&Tag-seq data sets are available for the data presented in Figs. 1A and C, 4A and F, 5C, and 6H. These data have been deposited in Mendeley Data (accession no. 10.17632/677hf6g3mm.3). All other data and image files are available from the corresponding authors upon reasonable request. All packages and codes used in this study are open-source and publicly available.

Thermogenic adipocytes in adult humans reside in at least six anatomic depots anatomically analogous to those in mice, including the abdominal scWAT (2,33). To explore the transcriptional signatures of thermogenic adipose tissues in obese humans, abdominal scWAT samples (preperitoneal fat layer of patients with oblique inguinal hernia, incisional hernia, or appendicitis) were collected from 46 human subjects for total RNA extraction and RNA-seq analysis. The samples were divided into two groups depending on the BMI values: normal weight group (BMI <24 kg/m², n = 29) and overweight group (BMI ≥24 kg/m², n = 17). A total of 1,995 genes (cutoff: P < 0.05 and |log₂ (fold change) | >0.3) were found to be significantly different between the control group and overweight group using DESeq2 analysis, in which 1,294 genes were upregulated and 701 genes were downregulated in the overweight group (Fig.1A). Kyoto Encyclopedia of Genes and Genomes pathway analyses of these differential genes revealed that the downregulated genes were mainly enriched in the PPAR signaling pathway, glucose-related (glucagon, insulin) signaling pathway, and fatty acid metabolism (Fig.1B, top), while the upregulated genes were mainly enriched in terms related to inflammation, such as cell adhesion molecules, complement and coagulation cascades, calcium signaling pathway, Th17 cell differentiation, and Th1 and Th2 cell differentiation (Supplementary Fig. 1A).

Interestingly, the pathway associated with the biosynthesis of amino acids was also enriched in the downregulated genes in the scWAT from overweight human subjects compared with control subjects. We next performed detailed analyses of pathways related to amino acid metabolism and found that these downregulated genes are mainly involved in valine, leucine, and isoleucine degradation; alanine, aspartate, and glutamate metabolism; arginine biosynthesis; as well as glycine, serine, and threonine metabolism (Fig.1B, bottom). In addition, the β-alanine metabolism pathway was enriched in the upregulated genes in the scWAT from overweight human subjects (data not shown). These data suggest that amino acid metabolism might play an important role in the dysregulation of thermogenic adipose tissue metabolism in obesity and associated metabolic diseases in humans. To test this, we further analyzed all the differential genes associated with amino acid metabolism in the scWAT between control and overweight human subjects and identified two sets of genes that are either significantly downregulated or upregulated in overweight human subjects (BMI ≥24 kg/m²) compared with control subjects (BMI <24 kg/m²) (Fig.1C). The tissue distribution patterns of those genes in major metabolic tissues of mice were also included to help to identify the potential role of the thermogenic adipose tissue–enriched genes in the regulation of thermogenic adipocyte development and function.

Notably, among all the differential genes related to amino acid metabolism, only two, N-acetyltransferase 8-like (NAT8L) and GLUL, are preferentially expressed in BAT (Fig.1C and Supplementary Fig. 1C). NAT8L catalyzes the formation of N-acetylaspartate from acetyl-CoA and aspartate. Interestingly, previous studies have demonstrated that the NAT8L gene is highly expressed in BAT and promotes lipid turnover and thermogenesis in BACs (34). GLUL encodes glutamine synthetase, which is the only endogenous enzyme to synthesize glutamine from glutamate and ammonia. We looked into other genes related to glutamine metabolism in RNA-seq data sets and found that in overweight human subjects, GLUL was the only gene that showed a significant difference in mRNA expression among all the other highly expressed genes in adipose tissue (Fig.1C and D). We noticed that GLUL is also expressed at a high level in WAT, though to a lesser extent compared with BAT. Therefore, another batch of human fat RNA-seq data sets was used to confirm the regulation of GLUL expression in obesity. In this case, vWAT, which represents the classical WAT, was collected from a total of 96 human subjects for RNA-seq analysis. Consistent with the result observed in scWAT, GLUL mRNA expression was also reduced in vWAT in overweight human subjects compared with normal control subjects (Supplementary Fig. 1D). It is worth noting that previous human studies have shown that high circulating levels of glutamine and glutamine-to-glutamate ratio are associated with a lower risk of obesity and associated metabolic defects (35–37), at least partially through attenuation of WAT inflammation in obesity (25). However, the role of GLUL in thermogenic adipocyte differentiation and energy expenditure has yet to be defined. As such, this study was designed to examine the role of GLUL in thermogenic adipocytes.

To this end, RT-qPCR was used to validate GLUL expression in human scWAT and revealed an ∼50% downregulation of GLUL mRNA expression in scWAT in the overweight human subjects compared with control subjects (Fig.1E). The protein level of GLUL was also markedly downregulated in scWAT in the overweight human subjects compared with control subjects (Fig.1F). In addition, the ratio of glutamine to glutamate, an indicator of Glul enzyme activity, was found to be negatively correlated with BMI value (r = –0.448, P = 0.001) in the scWAT (Fig.1G and Supplementary Fig. 1E). Furthermore, GLUL mRNA levels were inversely correlated with BMI (r = –0.53, P < 0.001), fasting blood glucose (FBG) (r = –0.401, P = 0.006), and plasma triglycerides (TGs) (r = –0.379, P = 0.011) (Fig.1H–J). These correlations suggest that the downregulation of GLUL gene expression in human scWAT is closely associated with overweight and its related metabolic dysfunction. To determine whether GLUL plays a role in thermogenic adipocyte differentiation and thermogenesis, correlation analyses between GLUL and adipogenic (FSP27, FABP4, ADIPOQ, PPARG) and thermogenic (CIDEA, ELOVL3) genes were performed. The results showed that GLUL is positively correlated with both categories of genes, even after BMI correlation (Fig.1K–P). These data suggest that GLUL may play a critical role in thermogenic adipocyte differentiation and function.

To confirm the expression and activity of Glul in BAT, inguinal WAT (iWAT), and epididymal WAT (eWAT) in mice, qPCR, Western blot, and mass spectrometry were performed, respectively, to measure both mRNA and protein levels of Glul and the concentrations of glutamine and glutamate. Consistent with our hypothesis that Glul-mediated glutamine production may play an essential role in the regulation of thermogenesis in adipose tissue, the mRNA and protein expression of Glul was much higher in BAT than in eWAT and iWAT (Fig.2A and B), accompanied by a significant elevation of glutamine level in BAT (Fig.2C). Impaired thermogenesis of adipocyte tissue is a critical cause of obesity and associated metabolic dysfunction (14,15). We examined whether Glul expression in adipose tissue is regulated in HFD-induced and genetically obese (ob/ob) and diabetic (db/db) mouse models. Interestingly, Glul mRNA expression was significantly downregulated in BAT (∼70%) from ob/ob mice compared with littermate controls (Fig.2D and Supplementary Fig. 2A). Similarly, Glul gene expression was also robustly downregulated in BAT (∼70%) from db/db mice compared with controls (Fig.2E and Supplementary Fig. 2B). Further immunoblotting analyses confirmed that Glul protein levels were lower in the BAT from ob/ob and db/db mice than their respective littermate controls (Fig.2F and G). In addition, HFD feeding for 1 month significantly increased mouse body weight as expected, accompanied by a significant downregulation of Glul mRNA expression in BAT (∼50%) compared with chow diet–fed control mice (Supplementary Fig. 2C and D). Taken together, these data demonstrate that Glul gene expression is inactivated in thermogenic adipose tissues in obese and diabetic mouse models.

Thermogenic adipose tissues are the major sites of nonshivering thermogenesis, which is important for the maintenance of body temperature and systemic energy homeostasis. During chronic cold exposure, BeACs in iWAT and eWAT are induced, together with classical BACs, to participate in thermogenic metabolism and body temperature maintenance. As such, we measured the expression of Glul in adipose tissues in response to chronic cold acclimation. As expected, upon chronic cold treatment, the mRNA expression of genes involved in thermogenesis, such as Ppargc1a, Ppargc1b, Ucp1, Prdm16, and Cidea, were significantly upregulated in eWAT, iWAT, and BAT tissues (Supplementary Fig. 2E–H). Intriguingly, Glul gene expression was also robustly elevated in all three types of adipose tissue upon chronic cold exposure (Fig.2H). In accordance, glutamine was significantly increased, whereas glutamate followed the opposite trend, under chronic cold exposure, leading to a marked elevation of glutamine-to-glutamate ratio (Fig.2I and J). We also examined the Glul protein expression in the BAT after exposure to different temperatures for 1 week from thermoneutrality (30°C) to cold (Fig.2K). We found that the protein levels of Glul in BAT were temperature dependently upregulated in response to cold exposure compared with thermoneutrality, similar to the alterations in Ucp1 protein levels (Fig.2K). Furthermore, IP injection of the β₃-adrenergic receptor agonist Cl316,243 for 7 days significantly increased the Glul mRNA levels in all three types of adipose tissue that mirrored Ucp1 expression (Fig.2L). In addition, the protein levels of Glul were also robustly upregulated in eWAT and iWAT in response to Cl316,243 treatment (Fig.2M). Together, these findings suggest that Glul may play an important role in thermogenic adipocyte induction and function.

To test this, the expression of Glul during immortalized BAC differentiation was examined. As expected, marker genes of thermogenic adipocytes, such as Pparg1 and Ucp1, were gradually induced along the course of differentiation (Fig.2N). Interestingly, the mRNA expression of Glul was also dramatically upregulated during the differentiation in a time-dependent manner (Fig.2N). Previous studies have shown that when the concentration of glutamine in the culture medium is ≥0.6 mmol/L, the glutamine-producing activity of Glul is negligible (38). However, the expression of Glul was continuously induced by >20-fold during the differentiation of BACs (∼27-fold), while the initial concentration of glutamine remained stable (4 mmol/L) in the differentiation medium replaced every 2 days, suggesting that the demand for Glul-mediated glutamine production was increasing during thermogenic adipocyte differentiation.

To test the role of Glul-mediated glutamine production in thermogenic adipocyte differentiation, lentiviruses expressing Cas9 and control or Glul targeting gRNA were used to transduce immortalized brown preadipocytes to generate Glul-KO stable cell lines. As shown in Fig.3A, Glul protein was eliminated in the undifferentiated brown preadipocytes. Since the concentration of glutamine in a standard adipocyte culture medium (4 mmol/L) is sufficient for normal adipocyte growth and activity, the differentiation medium was prepared with glutamine-containing or glutamine-free DMEM to better examine whether depletion of de novo Glul-mediated glutamine and/or exogenous glutamine supplementation would affect BAC differentiation (Fig.3B). We collected the cell lysates of control (CTR)-Gln (–), CTR-Gln (+), Glul-KO-Gln (–), and Glul-KO-Gln (+) BACs when cultured in Gln (–) and Gln (+) medium for 16 h to measure the cellular concentrations of glutamine and glutamate. The glutamine and glutamate were both dramatically downregulated in CTR-Gln (–) and Glul-KO-Gln (–) BACs compared with CTR-Gln (+) and Glul-KO-Gln (+) BACs, indicating that exogenous glutamine was much more than endogenous Glul-mediated glutamine production (Supplementary Fig. 3A). Notably, the cellular glutamine concentration in Glul-KO-Gln (+) BACs was much lower than that in CTR-Gln (+) BACs (Supplementary Fig. 3A, left), suggesting a critical role of Glul in the maintenance of glutamine abundance in BACs. Oil-red-O staining was used to evaluate the formation of large lipid droplets, which are typical features of fully differentiated adipocytes. As shown in Fig.3C, depletion of exogenous glutamine in the Glul-KO BACs [Glul-KO-Gln (–)] severely inhibited the formation of intracellular lipid droplets. However, when exogenous glutamine was supplemented [Glul-KO-Gln (+)], the impaired lipid droplet formation in Glul-KO BACs was largely rescued. Notably, in control BACs, exogenous glutamine supplementation [CTR-Gln (+)] had no apparent effect on lipid droplet formation compared with control BACs cultured without glutamine [CTR-Gln (–)] (Fig.3C), suggesting that Glul-mediated glutamine production is sufficient to support adipogenesis. To further evaluate the mitochondrial function, which is an important feature of adipocyte thermogenesis, seahorse studies were performed. As shown in Fig.3D and E, the basal, proton leak–linked, maximal respiration, and spare respiratory capacity–dependent oxygen consumption rates were all markedly decreased in Glul-KO-Gln (–) BACs compared with CTR-Gln (–) BACs. Interestingly, these mitochondrial activity indices were almost entirely restored by exogenous glutamine supplementation in Glul-KO-Gln (+) BACs. Consistently, the mRNA and protein levels of mitochondrial oxidative phosphorylation (OXPHOS)–related genes, including Atp5a1, Uqcrc2, Mtco1, Sdhb, Ndufa2, and Ndufb8, exhibited similar regulation patterns as the results shown in the seahorse studies (Fig.3F and G). These data suggest that transcriptional reprogramming of mitochondrial biogenesis is involved in the regulation of BAC thermogenesis by endogenous and exogenous glutamine.

To further validate our observations in clustered regularly interspaced short palindromic repeats (Crispr)/Cas9-mediated Glul-KO BACs, we used MSO, which is an irreversible Glul inhibitor (39), to deplete endogenous glutamine synthesis (Fig.3H). When the brown preadipocytes were not treated with MSO, the absence or presence of exogenous glutamine exhibited no noticeable effect on the normal differentiation of BAC, as evaluated by Oil-red-O staining (Fig.3I). However, in MSO-treated (1 mmol/L) brown preadipocytes, the absence of exogenous glutamine in the differentiation medium [MSO-Gln (–)] almost completely blocked the differentiation of brown preadipocytes into mature BACs. The supplementation of exogenous glutamine could significantly rescue the BAC differentiation [MSO-Gln (+)] (Fig.3I), consistent with the phenomenon using Crispr/Cas9-mediated Glul-KO (Fig.3C). As expected, further qPCR and Western blotting analyses confirmed the regulation of gene programs related to mitochondrial OXPHOS (Fig.3J and K). These results demonstrate that glutamine promotes BAC differentiation and function through transcriptional reprogramming of key genes related to adipogenesis and thermogenesis.

Because glutamine is synthesized from glutamate and ammonia, analysis of human RNA-seq data revealed that both GLUL and glutamic pyruvic transaminase (GPT) were downregulated in the scWAT of overweight human subjects compared with normal control subjects, leading to the elevation of glutamate concentration in the scWAT of overweight humans (Supplementary Fig. 1E). To exclude the effect of glutamate on the adipogenesis and thermogenesis of BACs, we treated BACs with different concentrations of glutamate from 0 to 100 mg/L and observed no significant changes in adipogenic and thermogenic genes (Supplementary Fig. 3B). Collectively, these results support the potentially important role of glutamine, instead of glutamate, in the adipocyte differentiation and thermogenic program.

To explore the mechanism underlying the regulation of BAC differentiation and thermogenesis by endogenous and exogenous glutamine, RNA-seq analysis was performed on CTR-Gln (–) BACs and Glul-KO BACs with or without glutamine in the culture medium. The analysis was divided into two parts. The first part was the comparison between Glul-KO-Gln (–) and CTR-Gln (–) to assess the effects of Glul-mediated glutamine production on the transcriptome. The second part was the comparison between Glul-KO-Gln (+) and Glul-KO-Gln (–) to examine the effects of exogenous glutamine supplementation on the transcriptome level. Using DESeq2 analysis, 9,698 differentially expressed genes were identified (cutoff: P < 0.05 and |log₂ (fold change) | >1) in the first comparison, and 10,355 differentially expressed genes (cutoff: P < 0.05 and |log₂ (fold change) | >1) were obtained in the second comparison. Moreover, combinatorial analysis of the three groups revealed that among the 9,698 differentially regulated genes due to the depletion of endogenous Glul-mediated glutamine production, 4,499 could be rescued by exogenous glutamine supplementation, which we call reversible genes. Among all the reversible genes, 2,718 were significantly downregulated in Glul-KO-Gln (–) BACs compared with CTR-Gln (–) and then upregulated in the Glul-KO-Gln (+) group compared with the Glul-KO-Gln (–) group (Fig.4A, set I); the rest of the 1,781 genes were upregulated in Glul-KO-Gln (–) BACs compared with CTR-Gln (–) and then downregulated in Glul-KO-Gln (+) BACs compared with Glul-KO-Gln (–) (Fig.4A, set II). Gene ontology (GO) enrichment analyses revealed that genes in set I are mainly enriched in biological processes critical for BAC differentiation and thermogenesis, such as fatty acid metabolic process, fat cell differentiation, and adaptive thermogenesis (Fig.4B, top). Conversely, genes in set II are enriched in biological processes that are not directly relevant to adipocyte differentiation and function, such as axonogenesis, negative regulation of cell adhesion, and actin filament organization (Fig.4B, bottom). As such, we mainly focused on genes in set I in this study. Notably, a set of key genes associated with BAC differentiation and thermogenesis were identified in set I, including Pparg, Fsp27, Fabp4, Adipoq, Ppargc1a, Ppargc1b, Prdm16, Cidea, Cox8b, Elovl3, and Ucp1 (Fig.4C). Additional RT-qPCR and immunoblotting analyses were used to confirm the mRNA and protein expression of these genes (Fig.4D and E).

We noticed that there is a considerable number of genes related to adipogenesis that are regulated by Glul depletion. It is possible that the effect of glutamine depletion on BAC thermogenesis might be secondary to BAC adipogenesis and differentiation. To rule out this possibility, we used MSO to treat 3T3L1 cells and mature BACs to assess whether glutamine regulates both adipogenesis and thermogenesis. The adipogenesis in 3T3L1 was blocked by MSO-Gln (–) compared with cells with glutamine (Supplementary Fig. 4A and B). After differentiation and maturation, BACs were treated with MSO in the culture medium with or without glutamine for 24 h, followed by norepinephrine stimulation for 6 h (Supplementary Fig. 4C). While the adipogenesis and lipid droplet formation were indistinguishable among the CTR-Gln (–), MSO-Gln (–), and MSO-Gln (+) groups (Supplementary Fig. 4D), the mRNA levels of mitochondrial OXPHOS and thermogenesis-related genes were significantly downregulated in MSO-Gln (–) BACs compared with CTR-Gln (–) BACs and were almost completely rescued by glutamine supplementation in MSO-Gln (+) BACs (Supplementary Fig. 4E and F). Consistently, similar alterations were observed in the protein level of Ucp1 (Supplementary Fig. 4G). These results suggest that glutamine is critical for both adipogenesis per se and thermogenesis, and the effect of glutamine on BAC thermogenesis is independent of its effect on adipogenesis.

Because of the importance of glutamine in adipogenesis and thermogenesis of white adipocytes and BACs, it also raised a question about whether glutamine is required for BeAC differentiation and function. Thus, we used 10T1/2 cells as a BeAC model and surprisingly found that Glul/glutamine exerted relatively modest effects on the adipogenesis and thermogenesis programs of BeACs (Supplementary Fig. 5A–C). Combined with the results that Glul was expressed at a higher mRNA level in eWAT and BAT than iWAT and was downregulated in both eWAT and BAT rather than iWAT in ob/ob and diet-induced obese mice, it suggests that the Glul-glutamine pathway may have distinct roles in WAT, BeAT, and BAT, which warrant further study.

Previous studies have implicated those posttranscriptional modifications and epigenetic mechanisms are involved in the transcriptional regulation by glutamine. To test this, ATAC-seq analysis was performed on fully differentiated CTR-Gln (–), Glul-KO-Gln (–), and Glul-KO-Gln (+) BACs to assess the changes in genome-wide chromatin accessibility among these groups (Supplementary Fig. 6A). Indeed, considerable differences were observed in comparisons among these groups. We obtained 19,692 reversible peaks that were regulated by depletion of glutamine and then rescued by glutamine resupplementation (cutoff: P < 0.05 and |log₂ (fold change)| >0.5), among which 13,216 peaks in set I were downregulated in Glul-KO-Gln (–) compared with CTR-Gln (–) and then restored in Glul-KO-Gln (+) compared with Glul-KO-Gln (–), and 6,476 peaks in set II were upregulated in Glul-KO-Gln (–) compared with CTR-Gln (–) and then reversed in Glul-KO-Gln (+) compared with Glul-KO-Gln (–) (Fig.4F and Supplementary Fig. 6B). Then peaks in set I were annotated to the 8,067 closest genes and peaks in set II annotated to the 4,264 closest genes shown in Fig.4G, and these reversible peaks are likely to be epigenetically regulated. GO analysis was performed on the genes annotated to these reversible peaks. Consistent with the results obtained at the transcriptional level by RNA-seq analysis, the annotated genes by reversible peaks in set I are mainly enriched in fat cell differentiation (Fig.4H). The annotated genes by reversible peaks in set II were mainly involved in biological processes that are not relevant to brown fat cell differentiation and thermogenesis, such as ossification and cell-substrate adhesion (Supplementary Fig. 6C). We next performed overlap analysis between genes in RNA-seq set I (cutoff: P < 0.05 and |log₂ (fold change) | >1) and the closest genes annotated to peaks in ATAC-seq set I (cutoff: P < 0.05 and |log₂ (fold change) | >1), and obtained 985 genes that are regulated in both data sets. Further GO analysis revealed that these 985 genes are enriched in biological processes involved in fatty acid metabolic process, adaptative thermogenesis, and brown fat cell differentiation (Fig.4I), suggesting that glutamine regulates adipogenic and thermogenic gene programs through orchestrating chromatin accessibility. However, how glutamine alters chromatin activity remains unclear.

Notably, previous studies have found that histone modifications and epigenetic mechanisms are involved in regulating adipocyte differentiation and function (40–42). In addition, other studies have demonstrated that intermediate metabolite–mediated histone modifications and chromatin remodeling play critical roles in glutamine-mediated regulation of cell identity and biological function (25–29). As such, it is possible that glutamine modulates chromatin accessibility through histone modifications, leading to the transcriptional orchestration of gene networks associated with adipogenesis and thermogenesis. Consistent with this hypothesis, GO analysis of the annotated genes by reversible peaks in set I demonstrated that besides brown fat cell differentiation, biological processes involved in epigenetic regulation, such as covalent chromatin modification and histone modification, are also enriched (Fig.4H). These data suggest that glutamine may directly regulate genes related to histone and chromatin modifications. To test this, we screened all the genes encoding enzymes responsible for either adding (writers, e.g., histone methyltransferases and histone acetyltransferases) or removing (erasers, e.g., histone demethylase and histone deacetylases) histone modifications in RNA-seq data sets. A cluster of genes encoding writers and erasers for histone acetylation and methylation that were regulated by glutamine was identified (Fig.5A). Further RT-qPCR was used to confirm these gene expressions. H3K4me and histone acetylation are well-known epigenetic markers associated with transcriptionally active chromatin. As shown in Fig.5B, several writer genes for H3K4me, including Kmt2b, Kmt2c, Kmt2e, and Prdm9, were markedly downregulated in Glul-KO-Gln (–) BACs compared with CTR-Gln (–) BACs and restored by glutamine supplementation in Glul-KO-Gln (+) BACs. In contrast, Kdm5b, an eraser gene for H3K4me, was elevated in Glul-KO-Gln (–) BACs compared with CTR-Gln (–) BACs and reversed in Glul-KO-Gln (+) BACs. Similarly, histone acetylation writer genes encoding histone acetyltransferases, such as Kat2b and Kat6b, were robustly downregulated in Glul-KO-Gln (–) BACs compared with CTR-Gln (–) BACs and restored by glutamine supplementation, while histone deacetylases, including Hdac4 and Sirt6, were upregulated in Glul-KO-Gln (–) BACs and attenuated by glutamine supplementation in Glul-KO-Gln (+) BACs. Similarly, genes involved in adipogenesis, thermogenesis, H3K4me, and histone acetylation were also robustly regulated by glutamine in MSO-treated BACs (Supplementary Fig. 7). Taken together, these findings suggest that glutamine regulates BAC differentiation and thermogenesis through transcriptional regulation of histone modification enzymes and chromatin remodeling, leading to transcriptional orchestration of adipogenic and thermogenic gene networks.

To confirm whether glutamine could regulate BAC differentiation through histone methylation and/or acetylation, CUT&Tag-seq was performed with antibodies against H3K4me3 and acetyl-H3 (Fig.5C). On the basis of RNA-seq and ATAC-seq results, we mainly focused on the reversible peaks, which were downregulated by deletion of glutamine and restored by glutamine resupplementation. Reversible peaks (cutoff: P < 0.05 and |log₂ (fold change) | >1) were identified by comparisons of CUT&Tag-seq data sets among CTR-Gln (–), Glul-KO-Gln (–), and Glul-KO-Gln (+). Overlapping peaks between 1,085 reversible peaks in CUT&Tag-seq using anti-H3K4me3 antibody and 7,161 reversible peaks downregulated in Glul-KO-Gln (–) compared with CTR-Gln (–) and then restored in Glul-KO-Gln (+) compared with Glul-KO-Gln (–) in ATAC-seq set I using a stricter screening criterion (cutoff: P < 0.05 and |log₂ (fold change) | >1) were annotated to the 768 closest genes. GO analysis revealed that these genes are mainly enriched in fat cell differentiation and regulation of lipid metabolic process (Fig.5D). Moreover, overlapping peaks between 640 reversible peaks in CUT&Tag-seq using anti–acetyl-H3 antibody and 7,161 reversible peaks in ATAC-seq set I (cutoff: P < 0.05 and |log₂ (fold change) | >1) were annotated to 377 genes enriched in fat cell differentiation, fatty acid metabolic process, and lipid storage (Fig.5E). These results indicate that glutamine regulates BAC differentiation and metabolism through histone modification–mediated remodeling of chromatin accessibility. Interestingly, genome browser track analysis of our sequencing data on the coding and flanking regions of representative adipogenic and lipid droplet–associated gene (Fsp27) and thermogenic genes (Cidea and Ucp1) demonstrated that the transcriptional signals in Glul-KO-Gln (–) BACs were markedly reduced compared with CTR-Gln (–) BACs. Similar trends were accompanied in ATAC-seq analyses. Moreover, glutamine resupplementation in Glul-KO-Gln (+) BACs almost completely reversed the ATAC-seq signals, leading to a large restoration of the transcription of these genes in Glul-KO-Gln (–) BACs (Fig.5F–H). Notably, for the Fsp27 gene, both H3K4me3 and acetyl-H3 binding peaks displayed the same trends as ATAC-seq signals (Fig.5F). However, for both thermogenic genes Cidea and Ucp1, only H3K4me3 binding peaks exhibited the same trends as ATAC-seq signals, whereas no obvious change in acetyl-H3 binding peaks was observed (Fig.5G and H). Other adipogenic genes, such as Adipoq and Fabp4, also showed similar patterns of binding peaks to Fsp27, and thermogenic genes, such as Ppargc1a, Ppargc1b, Cox8b, Dio2, and Elovl3, showed similar patterns of binding peaks to Cidea and Ucp1 (Supplementary Fig. 8). These data suggest that the H3K4me3-mediated epigenetic mechanism might play the most prominent role in the glutamine-mediated regulation of adipogenic and thermogenic gene programs in BACs.

Previous studies proposed that epigenetic regulation of gene expression by glutamine relies on its downstream product α-KG (19,27). α-KG has been demonstrated as functioning in brown fat adipogenesis through ten eleven translocation (TET)–mediated DNA demethylation of the Prdm16 promoter (43). To exclude the possible involvement of α-KG in glutamine-mediated regulation of adipocyte differentiation and thermogenesis, α-KG was used to replenish Glul-KO-Gln (–) BACs to see whether it can exert the same effect as glutamine. The α-KG gave a poor performance in improving the differentiation of Glul-KO-Gln (–) BACs (Supplementary Fig. 9). These results reveal that the effect of glutamine on histone modification–mediated BAC differentiation is relatively independent of α-KG.

Another speculation raised that the expression of the writer and eraser genes for histone methylation and acetylation might also be regulated during the BAC differentiation. We next examined the expression of the corresponding writer and eraser genes that were significantly altered in both Glul-KO–treated and MSO-treated BACs. Remarkably, only Prdm9 was robustly upregulated during the differentiation of BAC in a time-dependent manner (Fig.6A). Prdm9 (also known as Meisetz), is a PR domain containing protein that trimethylates histone 3 on lysine 4 (H3K4) and 36 (H3K36) (44). Prdm9 was originally found to be solely expressed in germ cells and identified as an essential regulator in meiotic prophase progression (44,45). Immunoblotting analyses were used to confirm the protein expression of Prdm9 and its mediated H3K4me3 during the differentiation of BACs (Fig.6B). Consistently, RNA-seq analysis of metabolic tissues from WT mice revealed that Prdm9 is highly expressed in adipose tissues, especially in BAT (Supplementary Fig. 10A and B). Similar to Glul, Prdm9 protein is predominantly expressed in BAT rather than in eWAT and iWAT (Fig.6C). Notably, this upregulation of Prdm9 during BAC differentiation was almost completely abolished by MSO treatment when BACs were differentiated in Gln (–) medium (Fig.6B and D), and Prdm9 was also significantly downregulated in mature BACs without glutamine [MSO-Gln (–)] compared with CTR-Gln (–) and MSO-Gln (+) (Supplementary Fig. 10C), suggesting that Prdm9 is regulated by glutamine and that Prdm9-mediated H3K4me3 may play an important role in BAC differentiation and thermogenesis. To test this, an shRNA targeting Prdm9 gene was designed. As shown in Fig.6E, Prdm9 mRNA expression was successfully knocked down by lenti-shRNA transduction in brown preadipocytes. Interestingly, shRNA-mediated knockdown (KD) of Prdm9 impaired the lipid droplet formation compared with control cells, as revealed by Oil-red-O staining (Fig.6F). In addition, KD of Prdm9 also significantly attenuated the genes related to adipogenesis, including Pparg1, Pparg2, and Fabp4, as well as genes involved in thermogenesis, such as Ppargc1b, Prdm16, Cidea, and Ucp1 (Fig.6G). We then performed H3K4me3 CUT&Tag-seq on Prdm9-KD BACs to see whether Prdm9-KD can alter H3K4me3 binding on adipogenic or thermogenic genes. We found that the downregulated peaks that annotated genes in Prdm9-KD BACs were enriched in adaptive thermogenesis, cold-induced thermogenesis, and temperature homeostasis biological processes compared with control BACs (Fig.6H). Significantly declined binding of H3K4me3 on thermogenic genes, such as Ucp1 and Cidea, was found in Prdm9-KD BACs compared with normal controls (Fig.6I). To verify this observation, a selective Prdm9 inhibitor, MRK-740, was used to inhibit H3K4 methylation (46). Accordingly, MRK-740 specifically inhibited the glutamine-induced elevation of H3K4me3, whereas the levels of H3K36me3, the other target of Prdm9, were not affected by MRK-740 treatments (Supplementary Fig. 11A). Consistently, pharmacologic inhibition of Prdm9 also impaired lipid droplet formation (Supplementary Fig. 11B), accompanied by a robust attenuation of the rescued expression of the adipogenic and thermogenic gene programs (Supplementary Fig. 11C). To further ensure the importance of Prdm9 in glutamine-mediated BAC induction and thermogenesis, we also performed the Prdm9 overexpression experiments in the Glul-KO BACs. As shown in Fig.6J–K, restoring the expression of Prdm9 could rescue the adipogenic and thermogenic defects in Glul-KO BACs in glutamine-free medium, similar to results obtained in glutamine supplementation studies. These results reveal Prdm9-mediated H3K4me3 as the critical regulator of glutamine-triggered BAC differentiation and thermogenesis.

There was still a question of how glutamine regulates Prdm9. Thus, we performed known motif enrichment analysis on the differentially accessible peaks by ATAC-seq (Fig.4F). Among all the differential peaks, six transcription factors, including C/EBPb, NFIL3, HLF, NF1, PPARG, and EBF2, were significantly enriched (Fig.7A). C/EBPb, PPARG, and EBF2 are well-known transcription factors involved in adipocyte adipogenesis and thermogenesis. The footprint results confirmed the important role of C/EBPb in the process of glutamine regulation of BAC differentiation and function (Fig.7B). C/EBPb is early and rapidly expressed following induction of adipocyte differentiation and is critical for triggering the transcription of Pparg and C/EBPa (47). Intriguingly, analysis of the GSE74189 data set showed that C/EBPb binds to the enhancer region, as supported by the enrichment of histone enhancer markers H3K4me1 and H3K27ac, of the Prdm9 gene on day 2 of BAC differentiation (Fig.7C). To test whether the binding of C/EBPb to this enhancer region of Prdm9 is responsible for its transcriptional regulation, we performed luciferase reporter assays by cloning this Prdm9 enhancer region into PGL3 luciferase reporter vector. The predicted that the C/EBPb binding site was also mutated in this enhancer region in this PGL3 luciferase reporter vector. We found that C/EBPb could robustly activate luciferase reporter activity of PGL3 vector carrying the WT Prdm9 enhancer region (Fig.7D), which was largely abolished by mutation of the C/EBPb binding site (Fig.7D). Furthermore, to confirm the effect of C/EBPb on Prdm9 transcription, we knocked down C/EBPb by siRNA in Glul-KO-Gln (+) cells and observed that the rescued expression of Prdm9 in Glul-KO-Gln (+) cells compared with Glul-KO-Gln (–) cells was significantly attenuated by C/EBPb KD (Fig.7E and F). These results suggest that glutamine regulates Prdm9 gene expression by recruiting C/EBPb to its enhancer region. However, the link between glutamine and C/EBPb recruitment still warrants further study.

Adaption of mice under thermoneutral (30°C) conditions leads to decreased multilocular structures and mitochondria in BAT (48). When switched to room temperature from thermoneutrality, lipid accumulation decreased and mitochondria increased in BAT, respectively. Previous studies have shown that there are dynamic interconversions of low- and high-thermogenic BACs in different ambient temperatures (49). In this study, we used water as control and Ala-Gln (1 g/kg) fluid, which is a more stable form of glutamine, to orally treat WT C57BL6/J mice daily in the gradual changing process from thermoneutrality to lower temperature to observe whether glutamine can improve the differentiation and function of BAT in vivo (Fig.8A). The blood kinetics of Ala-Gln in plasma were measured at 0, 15, 30, 60, and 120 min after gavage. The concentrations of glutamine and alanine both reached their peaks at 15 min, and glutamate reached its peak at 30 min after gavage (Supplementary Fig. 12). There was no significant difference in body weight between the Ala-Gln–treated group and the control group (Supplementary Fig. 13A). Intriguingly, the supplementation of Ala-Gln better maintained the body temperature of mice at 22°C and 16°C compared with the control group (Fig.8B). The transcription and protein levels of thermogenic genes in BAT were higher in Ala-Gln–treated mice compared with control mice (Fig.8C and D). Consistently, morphologic characteristics of BAT showed that the Ala-Gln–treated mice had much more obvious multilocular structures and Ucp1 expression and less lipid accumulation at both 22°C and 16°C compared with the control mice (Fig.8E and F and Supplementary Fig. 13B). However, thermogenic genes in WAT did not show any difference between control and Ala-Gln–treated mice (Supplementary Fig. 13C), indicating that glutamine has great potential in improving thermogenesis in BAT. These results suggest that elevation of glutamine production promotes adipose tissue thermogenesis.

Activation of BAT is considered an appealing approach to combat obesity and associated metabolic dysfunction. To further examine the potential beneficial effect of long-term elevation of glutamine in the circulation on systemic energy homeostasis, WT C57BL6/J mice were subjected to free access to regulator water or drink containing Ala-Gln and alanine (to exclude the effect of alanine), followed by HFD feeding. The Ala-Gln–containing drink significantly increased the plasma concentrations of glutamine compared with regulator water control (Supplementary Fig. 13D). As expected, HFD-induced body weight gain and elevation of blood glucose levels were significantly alleviated by the Ala-Gln treatment compared with the control and alanine treatment groups (Fig.8G–I). HFD-induced insulin resistance, as revealed by the HOMA of insulin resistance (HOMA-IR), was also significantly improved in the Ala-Gln–treated group compared with the control group (Fig.8J). Moreover, the proportions of eWAT, iWAT, and BAT in the whole body were significantly reduced after Ala-Gln treatment compared with the control and alanine groups (Fig.8K). Consistently, the BAT of Ala-Gln–treated mice showed better thermogenic function than control and alanine-treated mice (Fig.8L and M and Supplementary Fig. 13E). Together, these results reveal that Ala-Gln treatment promotes adipose tissue thermogenesis, thereby leading to attenuation of HFD-induced obesity and its associated metabolic disorders.

To further investigate the effects of Glul inactivation on HFD-induced obesity and metabolic impairments, HFD-fed WT C57BL6/J mice were subjected to MSO (5 mg/kg) treatment through IP injection three times a week. As shown in Supplementary Fig. 14I–C, MSO-treated mice gained significantly more body weight, accompanied by higher blood glucose levels and HOMA-IR. As expected, MSO-treated mice also exhibited lower body temperature compared with control mice (Supplementary Fig. 14C). The tissue weight and proportions of iWAT and BAT in the whole body were higher in MSO-treated mice compared with control mice (Supplementary Fig. 14D). Consistently, thermogenic genes in the BAT of MSO-treated mice were significantly downregulated compared with that from control mice (Supplementary Fig. 14E). These results indicate the indispensable role of Glul in metabolic health, likely through the regulation of adipocyte thermogenesis.

Thermogenic adipose tissues, including BAT and BeAT, are the main source of nonshivering thermogenesis in mammals. Loss or dysfunction of thermogenic adipose tissues is known to be an important cause of obesity and associated metabolic disorders. However, the underlying mechanisms remain incompletely defined. Here, we describe the critical role of Glul and glutamine (product of Glul) in BAC thermogenesis and systemic temperature homeostasis. The Prdm9-mediated H3K4me3 modification specifically induced by glutamine leads to transcriptional reprogramming of adipogenic and thermogenic gene programs.

Through transcriptomic analysis of human scWAT, which contains thermogenic adipocytes (4), we noticed that biological processes enriched in the downregulated genes in overweight subjects were much more remarkable than those in the upregulated ones. Among the downregulated genes, besides some well-known biological processes, such as fatty acid metabolism (50), insulin signaling, and PPAR signaling pathway (51), a tremendous change of amino acid metabolism was also observed. Indeed, previous studies have demonstrated the impact of some specific amino acids on obesity and its related metabolic disorders. For example, branched-chain amino acids, including valine, leucine, and isoleucine, contribute to insulin resistance in obesity (52), while glycine was proven to decrease in obesity and diabetes and positively correlated with insulin sensitivity (53). Considering the expression patterns and tissue-specific functions of amino acids, we mainly focused in this study on the enzymes that are preferentially expressed in thermogenic adipose tissues and dysregulated in obesity. As a result, GLUL and NAT8L were identified as the potential candidates in mediating the impairment of adipocyte thermogenesis in obese humans. Interestingly, NAT8L has been reported to participate in the regulation of lipid turnover and thermogenesis in BACs (34). As such, this study mainly focused on the role of GLUL and its product in thermogenic adipocyte differentiation and function.

It has been reported that the functions of Glul can be divided into two types, which are glutamine independent and glutamine dependent. The former mainly refers to its role in angiogenesis through regulating GTPase RHOJ palmitoylation and activation, with negligible glutamine synthesizing activity (38). The latter is the classical function of Glul, converting glutamate and NH4⁺ to glutamine. Glul is highly expressed in major metabolic tissues, such as adipose tissue, liver, skeletal muscle, and brain. A previous study using liver-specific Glul-deficient mice demonstrated that hepatic Glul-mediated ammonia detoxification plays an important role in the control of whole-body ammonia homeostasis (54). In skeletal muscle, although dispensable in fed mice, Glul was found not only contributing to ammonia detoxification and urea synthesis but also playing a key role in the adaptive response to fasting by transiently facilitating the production of glutamine (55). However, the role of Glul in adipocytes remains elusive. In this study, we found that the mRNA expression of GLUL in human scWAT is inversely associated with parameters of the metabolic disorder, including BMI, FBG, and plasma concentrations of TGs. In contrast, GLUL gene expression in human scWAT is positively associated with the expression of genes involved in adipogenesis and thermogenesis after BMI correction (Fig.1H–P). Similarly, the expression of Glul in BAT is also significantly downregulated in diet-induced, genetically obese, and diabetic mouse models. Furthermore, the expression of Glul is markedly elevated in adipose tissues upon cold stimulation, β₃-adrenergic receptor agonist in mice, and thermogenic adipocytes during differentiation. These results suggest that Glul and its product may be essential for the differentiation and thermogenesis of adipocytes. Indeed, genetic or pharmacologic inactivation of Glul-mediated glutamine production with exogenous glutamine deprivation robustly blocks BAC differentiation and thermogenic function.

Glutamine is the most abundant free amino acid in vivo, playing important roles in the nitro cycle, regulating cell proliferation, activating autophagy, and mediating cell signaling (18,27,56,57). Notably, several previous studies in humans have indicated that glutamine is a protective factor against obesity (35–37). This hypothesis is further confirmed in studies showing that extra supplementation of glutamine could improve glucose homeostasis in mice, restrain weight gain after HFD feeding in rats, and reduce waist circumference in overweight patients (58,59). However, the underlying mechanism and the role of adipose tissue in this process have yet to be explored. However, the role of glutamine metabolism in adipocytes per se, differentiation, and thermogenesis, in particular, has not been defined. In this study, through both genetic and pharmacologic deletion of Glul in premature BAC and mature BAC, we demonstrated the pivotal role of glutamine in BAC differentiation and thermogenesis. In vivo, glutamine and glutamate could be converted into each other through a series of enzymes. Interestingly, in this study, we found the downregulation of both GLUL and GPT in human scWAT, with an increase of glutamate in the scWAT of overweight human subjects. By treating BACs with different concentrations of glutamate, we observed mild differences in the expression of thermogenic genes and excluded the potential effect of glutamate on adipocyte differentiation and thermogenesis. Meanwhile, glutamine treatment in white adipocytes further confirmed the positive effect of glutamine on white adipocyte differentiation per se. Petrus et al. (25) previously observed that the glutamine released from WAT is decreased in obese patients, and extrinsic glutamine supplementation could inhibit the transcriptional activity of proinflammatory pathways in WAT. Together, these results reveal that glutamine is involved in the regulation of both differentiation per se and inflammation in WAT, suggesting a critical role of glutamine in WAT remodeling and function in the development of obesity. Consistently, supplementation with Ala-Gln promotes the differentiation and function of BAT in the process of thermoneutrality to lower temperature in vivo and ameliorates the weight gain and elevation of blood glucose levels induced by an HFD. Of note, the thermogenesis function of BAT and BeAT can be highly induced under cold stress or sympathetic stimulation. A persistent cold acclimation experiment might yield a more robust effect of the Glul-glutamine pathway on HFD-induced obesity and metabolic dysfunction and warrants further study.

To further explore the underlying mechanism, a cluster of reversible genes were detected by RNA-seq analysis of three differentiated cell lines, including CTR-Gln (–), Glul-KO-Gln (–), and Glul-KO-Gln (+) BACs. Consistently, GO enrichment analysis of the reversible genes revealed biological processes in which BAC differentiation is impaired and fatty acid metabolism, fat cell differentiation, and adaptive thermogenesis are downregulated. Previous studies have shown that epigenetic mechanisms, such as histone modification and DNA methylation, are involved in glutamine regulation of tumor progression and skeletal stem cell differentiation (26,27). Accumulating evidence suggests that epigenetics, including DNA methylation, histone modification, the regulatory effect of noncoding RNA, and chromatin remodeling on gene expression, plays a pivotal role in linking environmental factors to transcriptional reprogramming (60). In addition, several previous studies have demonstrated the important role of histone modification in regulating adipogenesis and thermogenesis in adipocytes. For example, Wang et al. (40) found that methylation of H3K9 could inhibit adipogenesis by inhibiting Pparg expression; Li et al. (41) found that acetylation of H3K27 could activate transcription of Ppargc1a and Ucp1, which promote brown adipogenesis; Roh et al. (42) provided evidence suggesting that H3K27 acetylation and H3K4 methylation participate in the temperature-associated beige fat whitening; Zhao et al. (61) demonstrated that KMT5c-mediated H4K20 methylation plays an important role in adipocyte thermogenesis by regulating Trp53 expression; and Shuai et al. (62) revealed that DOT1L-mediated H3K79 methylation is critically involved in the regulation adipocyte differentiation and thermogenesis. In this study, integrated analysis of ATAC-seq and RNA-seq suggests that glutamine may regulate BAC differentiation through histone modification–mediated remodeling of chromatin accessibility. Indeed, among the reversibly changed genes in response to the availability of glutamine, a series of histone methylation and acetylation–associated genes were identified that might contribute to the activation of genes associated with adipogenesis and thermogenesis. Through further combinational analysis of H3K4me3 CUT&Tag-seq, acetyl-H3 CUT&Tag-seq, and ATAC-seq of the representative genes involved in adipogenesis and thermogenesis, such as Fsp27, Cidea, and Ucp1, H3K4 methylation was revealed as the common mechanism in mediating the regulation of adipocyte differentiation and thermogenesis by glutamine. Accordingly, Prdm9, an H3K4me3 and H3K36me3 writer, was identified as a potential mediator of this regulation process. Here, we found that Prdm9 is highly expressed in adipose tissues, BAT in particular, and is robustly induced upon BAC differentiation. Consistent with our findings, Son et al. (63) also observed the upregulation of Prdm9 during BAC differentiation, but they did not study its exact role in BAC. Here, we show that inhibition of Prdm9 using genetic or pharmacologic approaches profoundly reduced the H3K4me3 binding on key genes associated with BAC adipogenesis and thermogenesis and decreased gene expression, leading to the impairment of BAC differentiation. Furthermore, to investigate the link between glutamine and Prdm9, using the combinational approaches of ATAC-seq motif enrichment analysis, footprint analysis, luciferase reporter assay, as well as C/EBPb KD, we identified C/EBPb as a key regulator in mediating the transcriptional induction of Prdm9 by glutamine.

To explore the therapeutic potential of glutamine, we used daily oral supplementation of Ala-Gln to treat WT mice. Intriguingly, Ala-Gln–treated mice better maintained body temperature during the gradually changing process from thermoneutrality to lower temperature by promoting thermogenesis in BAT. Furthermore, Ala-Gln alleviated the diet-induced obesity and related metabolic disorders in mice. Previous studies have shown that glutamine supplementation is beneficial for the recovery of the immune system, muscle, liver, and other tissues during catabolic circumstances and safe for use in clinical practice (64,65). Glutamine dipeptides like Ala-Gln are an advantageous treatment approach during the recovery period because of their stability (65). The beneficial effect of long-term oral supplementation of glutamine in metabolic disease development is worth further study.

Several limitations remain in this study. First, this study mainly focused on histone modification among all the epigenetic regulators, but we could not exclude the possibility that other epigenetic mechanisms or even epigenetic independent mechanisms, such as DNA methylation and noncoding RNA, might be involved in the regulation of BAC differentiation by glutamine. Additionally, Petrus et al. (25) demonstrated that a low glutamine level promotes WAT inflammation in obesity through nuclear UDP-GlcNAcylation. Our pilot experiments revealed that BACs grew poorly in the presence of GlcNAcylation inhibitors. Further study is needed to clarify their roles in glutamine regulation of BAC differentiation and thermogenesis. In any case, Prdm9-mediated H3K4me3 seems to be critical in this process, given that genetic or pharmacologic inhibition of Prdm9 abolished the induction of BAC differentiation and thermogenesis. Second, Okamatsu-Ogura et al. (66) found that cold exposure significantly increased circulating glutamine content in WT mice but not in Ucp1-KO mice, indicating the need for glutamine in the cold-induced BAT thermogenesis. Therefore, it is possible that during the thermogenic induction in response to cold exposure, glutamine might also serve as a fuel in addition to its epigenetic regulation of thermogenic gene expression. Further study using metabolic flux assay is needed to clarify this possibility. Third, oral supplementation of the Ala-Gln might lead to the elevation of glutamine in multiple organs, including the liver and muscle, besides adipose tissues. Glul depletion in mouse models through IP injection of MSO can also affect many organs. As such, we could not rule out the possible contributions of these metabolic tissues to systemic energy homeostasis. Further studies using adipose tissue–specific Glul overexpression or deficient mice will help to solve this issue.

In summary, this study reveals a key role of Glul and its product glutamine in thermogenic fat differentiation and thermogenesis. Glul-mediated glutamine metabolism is attenuated in thermogenic adipose tissues in obese humans and mice, thereby leading to the impairment of BAC thermogenesis and systemic energy homeostasis. We also uncovered an unexpected role of C/EBPb-Prdm9–mediated H3K4me3 in the transcriptional induction of thermogenic adipocyte differentiation. Our findings highlight the therapeutic and translational potential of targeting the Glul-glutamine-C/EBPb-Prdm9 pathway to fight obesity and associated metabolic diseases.

Acknowledgments. The authors thank Dr. Jiandie Lin, Dr. Siming Li, and Dr. Henry Kuang from the University of Michigan for critical reading of the manuscript. The authors also thank the Meng laboratory members for helpful discussion and technical support for this study. The authors thank the Core Facilities of Zhejiang University School of Medicine for technical support. Images used in Figs. 2K, 3 B and H, and 7 A and Supplementary Fig. 4C were created using BioRender.com.

Funding. This work was supported by the National Key Research and Development Program of China (grant 2018YFA0800403), Training Program of the Major Research Plan of the National Natural Science Foundation of China (grant 91857110), National Natural Science Fund for Excellent Young Scholars of China (grant 81722012), National Natural Science Foundation of China (grant 81670740), Zhejiang Provincial Natural Science Foundation of China (grants LZ21H070001 and LHDMD22H02001), Innovative Institute of Basic Medical Sciences of Zhejiang University, Fundamental Research Funds for the Central Universities, and the Construction Fund of Medical Key Disciplines of Hangzhou (to Z.X.M.). This study was also supported by the National Natural Science Foundation of China (grants 81870564 and 81670744 to P.S. and 82070838 to Y.F.) and Science Technology Department of Zhejiang Province of China (grant 2017C33037) (to P.S.). This study was supported by the “National Tutor System” Training Program for Youth Talents of Suzhou Health Care System (Qngg2021007) (to Y.F.). The authors gratefully acknowledge the support of the K.C. Wong Education Foundation.

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

Author Contributions. X.P., L.Y., X.G., W.W., J.H., J.X., Y.C., Y.W., and C.X. performed the experiments. X.P., L.Y., X.S., and Z.-X.M. wrote the manuscript with help from other authors. Z.Z. and Q.W. performed the bioinformatics analysis. Y.F. and P.S. provided critical reagents for the study. Z.-X.M. conceived and designed the research with input from other authors. Z.-X.M. 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.