Homeobox C4 Transcription Factor Promotes Adipose Tissue Thermogenesis

More than 1 billion people worldwide have obesity (1), posing a significant challenge to public health. Adipose thermogenesis releases chemical energy in the form of heat primarily through the uncoupling protein 1 (UCP1), which elevates energy expenditure and can be used to correct positive energy balance and treat obesity. Emerging evidence supports the contribution of homeobox (HOX) genes to adipose thermogenesis. HOXC8 and HOXC10 act as negative regulators of white adipose tissue (WAT) thermogenesis and brown adipogenesis (2–4). HOXC4 expression in brown adipose tissue (BAT) notably surpasses that in WAT (5). Additionally, an increased frequency of a specific genetic variation in the HOXC4 gene has been observed in the population adapted to extreme cold environments (6), suggesting the potential role of HOXC4 in BAT function and cold acclimation. Despite these findings, the exact function of HOXC4 in modulating adipose thermogenesis and energy metabolism remains largely unexplored.

Our study bridges this gap by identifying HOXC4 as a novel transcriptional regulator fueling adipose thermogenesis. Our bioinformatic analyses suggest the potential involvement of HOXC4 in BAT function and obesity development, which was experimentally validated in a genetically modified animal model and primary adipocytes from both mice and humans. We further delineate a novel mechanism by which HOXC4 cooperates with nuclear receptor coactivator 1 (NCOA1) to promote Ucp1 transcription and thermogenesis, enhancing our understanding of adipose energy metabolism and providing potential therapeutic strategies for obesity and its related metabolic disorders.

C57BL/6J male mice were sourced from Vital River Laboratory. HOXC4 fl/fl and adiponectin-Cre mice were constructed by Cyagen Biosciences, Jiangsu, China. The adiponectin-Cre mouse model was generated by the transgenic insertion of the Cre recombinase gene under the control of the adiponectin promoter. HOXC4 knockout mice using adiponectin (Adipoq)-Cre–mediated deletion of floxed HOXC4 alleles, referred to as CKO mice, were produced by breeding HOXC4 fl/fl mice with Adipoq-Cre mice. PCR genotyping was performed using genomic DNA from the mouse tail, using primer sequences provided in Supplementary Table 1. Animals were housed in a specific pathogen-free facility. Throughout the duration of the study, all mice were maintained at an ambient temperature of 22°C, except during the cold exposure experiment. All animal experiments received approval from the Capital Medical University Institutional Animal Care and Use Committee.

For the cold exposure experiment, male mice housed at 22°C were transferred to the 4°C environment for 6 h, with rectal temperature measured by the microprobe thermometer (Uni-Trend Technology, UT323). The BAT and tail temperatures were recorded by the FOTRIC infrared camera detection system and processed using AnalyzIR 4.3.1.15 software.

For the high-fat diet (HFD) experiment, 8-week-old male mice were switched from normal chow to the HFD (MP Biomedicals, LLC). The adipose volume of mice was measured using the micro-CT scanner (SkyScan, Bruker), with tissue volume subjected to Gaussian smoothing and analyzed via CTAn software (Bruker), manually delineating areas of interest.

A glucose tolerance test (GTT) was conducted on male mice that were fasted for 16 hours, with glucose administered intraperitoneally at 2 g/kg. An insulin tolerance test (ITT) was conducted on mice that were fasted for 6 hours, with insulin administered intraperitoneally at 0.5 units/kg. Blood glucose levels were assessed from the mouse tail vein at 30-min intervals using the Contour TS glucose meter (Ascensia Diabetes Care Holdings AG).

Adeno-associated virus (AAV) 2/9 serotype particles were directly injected into BAT or inguinal WAT (iWAT) of 8-week-old C57BL/6J male mice (50 μL of 1.3 × 10¹² viral genomes/mL; HANBIO, Shanghai, China) to each side.

Fresh tissues were preserved in 4% formaldehyde (Servicebio, G1101) overnight before being transferred to 70% ethanol and embedded in paraffin for sectioning. The hematoxylin and eosin (H&E) stained slides were visualized by an Axioscope A1 microscope equipped with Zen lite software (Zeiss). Adipocyte sizes were quantified using the Adiposoft plugin for ImageJ 1.53q software (National Institutes of Health).

Rehydrated tissue sections were antigen retrieved in modified citrate buffer, blocked using 5% goat serum blocking buffer, and then incubated with the anti-UCP1 (1:200; Protintech, 23673-1-AP) or anti-HOXC4 (1:200; Bioss, bs-12196R) antibody overnight at 4°C. The slides were then washed in PBS and subsequently treated with the secondary antibody (Proteintech, PR30009) and hematoxylin (Servicebio, G1004) staining for nuclear visualization.

Total RNA was isolated using the Trizol total RNA extraction kit (Absin, abs9331), quantified by NanoDrop Microvolume Spectrophotometer, and reverse transcribed by the cDNA synthesis kit (TIANGEN, KR116-02). Quantitative (q)PCR analysis was conducted using the Quant Studio 1 system (Thermo Fisher Scientific) and a SYBR Green-based detection kit (Absin, abs601511), with data normalized against the reference gene GAPDH via the 2^–ΔΔCT method. The qPCR primers and antibody information are detailed in Supplementary Tables 2 and 3.

HEK-293T cells and 3T3-L1 cells were cultured in DMEM (Servicebio, G4511) with 10% FBS (Gibco, 10099141). Primary preadipocytes were isolated from BAT and iWAT of mice aged 2–3 weeks using collagenase I (Gibco,17100-017) digestion and were cultured in DMEM/F12 (Gibco, C11330500BT) with 10% FBS. Adipocyte differentiation was induced following previously established protocols (7).

Human adipose-derived mesenchymal stem cells (ADMSC) were obtained from Cyagen Biosciences (HUXMD-01101). Cells were cultured in ADMSC complete medium (HUXMD-90011). For human ADMSCs differentiation, the cells were seeded at 10⁴ cells/cm² in tissue culture plates. Upon reaching confluence, the medium was switched to adipogenic induction medium using the differentiation kit (HUXMD-90031).

A mitochondrial stress test was conducted by the Seahorse Extracellular Flux Analyzer (Seahorse Bioscience, North Billerica, MA). Oligomycin (1.5 μmol/L), carbonylcyanide-4-(trifluoromethoxy)-phenylhydrazone (FCCP; 2 μmol/L), antimycin A (0.5 μmol/L), and rotenone (0.5 μmol/L) (Alicelligent, ALS22012) were sequentially injected to assess mitochondrial respiration.

Cells were plated to achieve 40–60% confluence in confocal dishes. Following PBS washes, the cells were fixed with 4% paraformaldehyde for 20 min and subsequently washed three times with PBS. The fixed cells were then blocked and subsequently incubated with HOXC4 primary antibody (Bioss, bs-12196R) or NCOA1 primary antibody (Bioss, bs-10603R) overnight at 4°C. Subsequently, the cells underwent three successive washes with PBS to ensure the removal of unbound components. They were then exposed to secondary antibodies tagged with Alexa Fluor 594 (Thermo Fisher, A-11012) and Alexa Fluor 488 (Thermo Fisher, A-21202) at a 1:1,000 dilution for 1 h. The nuclei were stained using DAPI (Servicebio, G1012). Microscopic images were captured on a Leica DMI6000 confocal microscope equipped with 63× oil immersion objectives. Fluorescence intensity was analyzed using ImageJ software.

HEK-293T cells were grown in 48-well culture plates and cotransfected with the expression plasmids (0.5 μg/well), luciferase reporter construct (1 μg/well), and Renilla luciferase vector (0.25 μg/well) with the Lipo293 kit (Beyotime, C0521). The quantity of transfected DNA was balanced with empty vector (pcDNA 3.1). The cells were collected 48 h following transfection to measure luciferase activity using the Dual-Luciferase Reporter Gene Kit (Beyotime, RG027). Luciferase activity was standardized against Renilla luciferase activity to correct for variations in transfection efficiency.

HEK-293T cells were transfected with HOXC4-hydroxyapatite (HA) and NCOA1-FLAG constructs. Cell lysates were extracted using immunoprecipitation lysis buffer 48 h after transfection and immunoprecipitated using anti-HA beads (MedChemExpress, HY-K0201) for HOXC4 and anti-FLAG beads (MedChemExpress, HY-K0207) for NCOA1. For assessing endogenous protein interactions, primary adipocyte lysates were immunoprecipitated using anti-HOXC4 antibody (Bioss, bs-12196R), which was preincubated with protein A/G beads (MedChemExpress, HY-K0202) at room temperature for 2 h before immunoprecipitation. Western blot analysis was conducted to identify HOXC4 and NCOA1 using anti-HOXC4 antibody (NOVUS, NBP2-56195) and anti-NCOA1 antibody (Bioss, bs-10603R).

Primary brown adipocytes overexpressing HOXC4 were fixed at room temperature with 1% formaldehyde for 10 min. To terminate the crosslinking reaction, cells were treated with 125 mmol/L glycine for 5 min. The cells were subjected to PBS washes, collected, and centrifuged at 900g for 3 min. The pellet was lysed to extract nucleus by centrifuging at 2,000g for 5 min. Genomic DNA was sonicated into fragments ranging from 200 to 500 base pairs. Of the sonicated chromatin, 10% was saved as the input sample, 80% underwent immunoprecipitation using anti-HA beads (MCE), and the remaining 10% was treated with normal rabbit IgG (Cell Signaling Technology) to serve as the negative control. For chromatin immunoprecipitation (CHIP)-qPCR, crosslinking was reversed and purified DNA was subjected to qPCR with the primers detailed in Supplementary Table 4. For ChIP-sequencing (seq) analysis, the sequencing libraries were constructed using the VAHTS Universal DNA Library Prep Kit (Vazyme, ND607). Library fragments ranging from 200 to 500 base pairs were selectively enriched and sequenced on the Illumina Novaseq 6000 platform using the PE150 sequencing mode.

The statistical analyses were conducted using GraphPad Prism 9.5.1 software (GraphPad Software). Data are expressed as means ± SD. Statistical significance was assessed using the Student t test, one-way or two-way ANOVA, and Pearson correlation analysis.

All data generated or analyzed during the current study have been included in this article and its supplementary information files. The ChIP-seq data have been deposited in the National Center for Biotechnology Information’s Gene Expression Omnibus (GEO) and are publicly available as of the date of publication.

We retrieved the top 2,000 coexpressed genes of HOXC4 from COXPRESdb (8), an enhanced animal gene coexpression database that integrates RNA-seq and microarray data for coexpression analysis and infers the functional roles of genes. Kyoto Encyclopedia of Genes and Genomes (9) enrichment analysis was performed to study the potential biological functions of HOXC4, which revealed that genes coexpressed with HOXC4 are enriched in metabolic-related pathways (Supplementary Fig. 1A), suggesting HOXC4’s involvement in metabolic regulation. We also studied the genetic associations between HOXC4 polymorphisms and metabolic traits using the ExPheWas platform (10), a gene-based phenome-wide association study tool, which linked HOXC4 polymorphisms to distribution of body fat (waist-to-hip ratio) and other metabolic traits (Supplementary Fig. 1B). The association between HOXC4 expression and obesity was experimentally demonstrated using the HFD-induced mouse obesity model, where HOXC4 mRNA and protein expression in BAT were significantly reduced after HFD feeding (Supplementary Figs. 1C and E and 2B) and positively correlated with Ucp1 expression (Supplementary Fig. 1D). The establishment of the HFD-induced obesity model is demonstrated in Supplementary Fig. 2A and C. These results suggest that HOXC4 may participate in adipose thermogenesis and potentially influence obesity development.

To elucidate the role of HOXC4 in promoting thermogenesis, we performed gain-of-function experiments by administering AAVs expressing HOXC4 or ZsGreen to the BAT or iWAT of 8-week-old male mice. Three weeks later, we validated AAV-mediated overexpression in mouse BAT and iWAT depots, where ZsGreen expression in both viral vectors enabled us to confirm successful viral transduction in both groups (Supplementary Fig. 3A). In both cold exposure and HFD-induced obesity models, we analyzed the overexpression of HOXC4 at both the mRNA (Supplementary Fig. 3B and C) and protein (Supplementary Fig. 3D and E) levels. In response to acute cold exposure, mice with overexpression of HOXC4 in BAT or iWAT better sustained rectal temperatures (Fig. 1A, C, and D) and BAT/iWAT surface temperatures (Fig. 1B, F, and G), rather than altering tail skin temperature (Fig. 1E and H). These findings indicate that BAT- and iWAT-specific overexpression of HOXC4 can increase adipose thermogenesis without affecting heat dissipation from the tail, thereby maintaining the core body temperature during cold stress.

Consistent with the enhanced thermogenic phenotype, the overexpression of HOXC4 upregulated Ucp1 and multiple genes associated with thermogenesis and mitochondrial function (Fig. 1I and J). This gene upregulation did not extend to adipogenesis markers, such as Adipoq (adiponectin) and Pparγ (peroxisome proliferator-activated receptor-γ) (Fig. 1I and J), suggesting the specific activation of the adipose thermogenic pathway without altering adipogenesis. In addition, compared with the control group, mice receiving AAV-HOXC4 showed elevated protein expression of UCP1 in BAT and iWAT (Fig. 1K–M), along with smaller lipid droplets and a more active thermogenic phenotype (Fig. 1N).

The reduced thermogenic capacity observed in obese animals prompted us to study whether HOXC4 overexpression could counteract this effect. We used a HFD-induced obesity mouse model (Fig. 2A) to examine the metabolic changes following HOXC4 overexpression in BAT and iWAT. Overexpression of HOXC4 slowed body weight gain in mice during the 16-week course of HFD feeding (Fig. 2B and C). Analysis of the average daily food consumption (Fig. 2D and E) showed that food intake did not contribute to the weight change. Further analysis of body composition changes (Fig. 2F and H) demonstrated that the weight reduction following HOXC4 overexpression was due to decreased fat mass rather than lean mass (Fig. 2G and I). To evaluate whether HOXC4 overexpression also enhanced adipose thermogenesis in the HFD model, we conducted H&E staining (Fig. 2K) to assess adipocyte distribution (Fig. 2L–P), which showed increased percentage of smaller adipocytes and reduced average size (Fig. 2N and Q) of adipocytes in both BAT and iWAT following HOXC4 overexpression.

Obesity is commonly associated with metabolic dysfunction, such as insulin resistance. Consistent with the improved obesity phenotype, the overexpression of HOXC4 improved systemic glucose tolerance (Fig. 3A and B) and insulin sensitivity (Fig. 3C and D) in mice. Additionally, HOXC4 overexpression increased basal BAT temperature (Fig. 3F) and both mRNA (Fig. 3E) and protein (Fig. 3G–I) expression of UCP1 in BAT and iWAT. These results suggest that HOXC4 overexpression activated the thermogenic capacity in obese mice and induced metabolic benefits, including reduced fat mass and improved insulin sensitivity.

To directly examine the function of HOXC4 in adipose thermogenesis, we generated HOXC4 CKO mice (Fig. 4A). The genetic alteration and its specificity were validated via PCR genotyping (Fig. 4B) and mRNA (Fig. 4C) and protein (Fig. 4D–H) expression analyses.

Following acute cold exposure, HOXC4 CKO mice exhibited reduced rectal, BAT, and iWAT temperatures (Fig. 4I–K and M) with unchanged tail temperature (Fig. 4L). This was accompanied by reduced thermogenic gene expression in BAT and iWAT (Fig. 4N and O), while adipogenesis-associated genes remained unaffected. The decrease in UCP1 protein expression in BAT further supports this effect (Fig. 4P). Furthermore, the adipose depots from HOXC4 CKO mice appeared enlarged, displaying hypertrophic cells with increased lipid accumulation compared with floxed controls (Fig. 4Q), indicating impaired thermogenic capacity.

To determine the susceptibility of HOXC4 CKO mice to HFD-induced obesity, we fed both CKO and control mice the HFD comprising 60% fat and investigated the metabolic consequence of HOXC4 deficiency. HOXC4 CKO mice exhibited increased body weight gain relative to control counterparts (Fig. 5A), mainly due to a fat mass gain (Fig. 5C and D), while food consumption remained similar across genotypes (Fig. 5B). HOXC4 ablation reduced the thermogenic capacity of HFD-fed mice, as evidenced by decreased expression of thermogenic genes (Fig. 5E and F), paler BAT appearance (Fig. 5G), and more hypertrophic BAT and iWAT depots (Fig. 5H). In line with the aggravated obesity phenotype, HOXC4 CKO mice displayed impaired glucose tolerance (Fig. 5I) and reduced insulin sensitivity (Fig. 5J) following HFD feeding, further supporting the function of HOXC4 in maintaining adipose thermogenesis and metabolic homeostasis.

To determine whether the thermogenic-promoting effect of HOXC4 occurs via cell-autonomous mechanisms, we harvested the stromal-vascular fraction from mouse BAT and iWAT, induced adipocyte differentiation in vitro, and assessed mitochondrial respiration by the Seahorse assay. Consistent with in vivo observations, HOXC4 overexpression increased basal respiration, proton leakage, and maximal respiratory capacity in primary brown adipocytes (Fig. 6A and B), accompanied by elevated protein expression of UCP1 (Fig. 6C). Conversely, HOXC4 ablation reduced mitochondrial respiration and UCP1 expression in primary adipocytes (Fig. 6D and E). To explore the translational potential of HOXC4-activated thermogenesis, we infected primary human subcutaneous adipocytes with control adenovirus (Adv-EGFP) or adenovirus overexpressing HOXC4 (Adv-HOXC4). Mirroring the findings in mice, HOXC4 overexpression in human adipocytes also induced thermogenesis, as evidenced by increased mitochondrial respiration (Fig. 6F and G) and the upregulation of UCP1 (Fig. 6H).

To elucidate how HOXC4 regulates adipose thermogenesis at the transcriptional level, we applied ChIP-seq to map genome-wide HOXC4 binding sites and uncover potential HOXC4 transcription targets. ChIP-seq analysis revealed that ∼7% of the peaks were located in the promoter-transcription start site regions (Fig. 7A). Pathway enrichment analysis of the ChIP-seq data showed that genes associated with HOXC4 binding sites are enriched in several metabolic pathways (Fig. 7B). Motif analysis of the peaks identified the classic HOX binding motif (Fig. 7C).

Specifically, ChIP-seq data identified enriched HOXC4 binding peaks in the Ucp1 promoter (Fig. 7D). A similar result was obtained through ChIP-qPCR, showing clear evidence of HOXC4 enrichment to the proximal Ucp1 promoter region in primary adipocytes (Fig. 7E). Because our qPCR and Western blot analyses demonstrated an increase in UCP1 expression with HOXC4 overexpression and a decrease with HOXC4 ablation, we proposed UCP1 as a thermogenic transcriptional target of HOXC4.

Integrated analysis of ChIP-seq data and the JASPAR database (11) identified the potential HOXC4 binding sites within the Ucp1 promoter. To test the regulatory function of the HOXC4 binding site, we cloned the wild-type, truncated, and mutated Ucp1 promoter upstream of the luciferase reporter gene. HOXC4 induced the luciferase reporter gene driven by the wild-type Ucp1 promoter, a response attenuated by the deletion mutant lacking HOXC4 binding sites or when the HOXC4 binding site with the highest score was mutated (Fig. 7F and G). These data suggest that HOXC4 can bind to the Ucp1 promoter and activate its expression.

Next, we asked whether HOXC4 interacts with specific cofactors in the transcriptional activation of Ucp1, focusing on the known cofactors that have been implicated in adipose thermogenesis. Coimmunoprecipitation assay showed that NCOA1 interacted with both overexpressed and endogenous HOXC4 (Fig. 8A–C), which was verified by immunofluorescence experiment (Fig. 8D). We also used HDOCK (12) for molecular docking to simulate the interaction of the HOXC4-NCOA1 complex (Fig. 8E and F). Molecular docking identified that the hexapeptide-related core sequence in HOXC4 (amino acids 136–139) is important for NCOA1 binding (Fig. 8G) and that mutation of this motif (YPWM → AAAA) impaired the interaction between HOXC4 and NCOA1 (Fig. 8H and J). Moreover, this physical interaction could synergistically promote the transcriptional activity of Ucp1, as demonstrated by luciferase reporter assay (Fig. 8I). NCOA1 enhanced the luciferase activity driven by the Ucp1 promoter when coexpressed with wild-type HOXC4, but this effect was abolished with the hexapeptide motif mutant HOXC4 construct (Fig. 8I). These findings demonstrate the importance of the HOXC4-NCOA1 interaction, mediated by the hexapeptide motif, in regulating Ucp1 transcription.

HOXC4 is a member of the HOX transcription factor family, which is crucial for tissue patterning and development (13). Recent research suggests the potential functions of HOX genes in temperature perception, cold adaptation, and adipose thermogenesis. HOXA9 regulates the expression of the cold-activated transient receptor potential channel, potentially impacting temperature sensing in bats during the repeated torpor-arousal cycles of their hibernation (14). The Fuegians, a population with remarkable cold resistance, exhibited a higher frequency of the HOXC4 gene variant, which is predicted to increase HOXC4 expression (6), suggesting the evolutionary importance of HOXC4 in cold adaptation. Consistently, our study identified that the overexpression of HOXC4 enhanced cold tolerance in mouse models and increased thermogenic gene expression and activity in both primary mouse and human adipocytes, reinforcing the importance of HOXC4 in adipose thermogenesis. The role of evolutionary pressures in selecting the cold-resistant HOXC4 variant in frigid environments and its relation to HOXC4's function in thermogenesis warrant further investigation.

HOXC8 and HOXC10 have been reported to act as negative regulators of WAT browning and brown adipogenesis by targeting C/EBPβ and PRDM16, respectively (3,4), while Myf5-Cre mediated deletion of HOXC10 did not affect the expression of thermogenic genes in BAT (2), indicating its effects are mainly confined to WAT. On the other hand, HOXC4 is more abundant in BAT than in WAT and is upregulated in mature brown adipocytes (5,15,16), suggesting a more active role of HOXC4 in BAT function and thermogenesis. Accordingly, we have determined that overexpressing HOXC4 significantly enhanced thermogenic gene expression and activity in both WAT and BAT, while its deletion reduced thermogenic capabilities, indicating that HOXC4 has a broader impact in regulating thermogenesis across both types of adipose tissue. Our results further demonstrate that HOXC4-induced changes in BAT thermogenic activity significantly affect systemic glucose homeostasis and insulin sensitivity. Therefore, HOXC4 likely modulates these metabolic outcomes through its direct regulation of BAT thermogenesis.

Like other HOX transcription factors, HOXC4 comprises a highly conserved homeodomain that binds to AT-rich DNA regulatory elements, along with a short hexapeptide motif that facilitates protein interactions and DNA binding (17). However, the specific binding partner and transcription target of HOXC4, especially in relation to thermogenesis, remain unidentified. Using a combination of coimmunoprecipitation, ChIP, and luciferase assay, along with mutation or truncation analyses, we established that HOXC4 binds to coactivator NCOA1 through the hexapeptide motif and collaboratively induced Ucp1 transcription to promote thermogenesis.

In summary, our study demonstrates the important function of HOXC4 in adipose thermogenesis and reveals the mechanism by which HOXC4 activates Ucp1 transcription. Targeting HOXC4 may serve as a potential therapeutic approach to enhance adipose thermogenic capacity and treat obesity and its related metabolic disorders.

Acknowledgments. The authors thank Zhongxin Xiao and Jianfeng Lei, from the Core Facility at Capital Medical University, for their assistance with confocal immunofluorescence microscopy and micro-CT scanning.

Funding. This work was supported by the National Natural Science Foundation of China Grant 82270898 and the Youth Talent Training Program of Capital Medical University (B2404).

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

Author Contributions. T.Y., Y.W., H.L., F.S., S.X., Y.W., J.X., and Y.L. performed experiments and analyzed data. T.Y. and M.J. wrote and edited the manuscript, with final approval from all authors. M.J. designed the study and secured funding. T.Y. and M.J. are the guarantors of this work, and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy.