MicroRNA-29 Ameliorates Fibro-Inflammation and Insulin Resistance in HIF1α-Deficient Obese Adipose Tissue by Inhibiting Endotrophin Generation

Within obese adipose tissue (AT), adipocyte hypertrophy and insufficient vascularization lead to oxygen shortage and hypoxia (1). AT hypoxia then promotes tissue fibrosis, inflammation, and systemic insulin resistance (2). The hypoxic response is mediated via hypoxia-inducible factors (HIFs), which are heterodimeric transcription factors consisting of HIF1α and HIF1β subunits (3). HIF1α is oxygen-sensitive and undergoes proteolytic degradation under normoxia, whereas HIF1β (ARNT) is constitutively expressed (3). Under hypoxia, HIF1α is stabilized and translocated into the nucleus, forming a heterodimeric complex with HIF1β. This complex binds to hypoxia response elements (HREs) within promoters, transactivating a wide range of target genes involved in angiogenesis, cell survival, development, and stem cell function. The central roles of HIF1α in metabolism have been investigated by using in vivo mouse models. In aP2 promoter-driven Hif1α knock-out (KO) mice, high-fat diet (HFD)-induced obesity alleviated chronic inflammation, fibrosis, and insulin sensitivity (4,5). In addition, macrophage-specific deletion of Hif1α in Lys-CRE mice had no significant impact on inflammation in the early stages of obesity (8 weeks of HFD) (6), whereas protective effects against obesity-induced chronic inflammation and systemic insulin resistance following 18 weeks of HFD feeding were observed (7). Further, both adiponectin promoter-driven overexpression of dominant-negative Hif1a and pharmacological inhibition of HIF1α via PX-478 (S-2-amino-3-3[4′-N,N,-bis(2-chloroethyl)amino]phenyl propionic acid N-oxide dihydrochloride) in HFD-fed obese mice markedly reduced AT fibrosis and improved insulin sensitivity (8). Notably, HIF1α upregulation induces extracellular matrix (ECM) component expression, leading to extensive fibrosis in various tissues, including ATs and tumors (2,9). Within hypoxic AT, the “fibrotic response” is an early event that induces inflammation in a sequential pathological cascade that is directly linked to obesity-related metabolic dysfunction (2).

Type VI collagen (COL6) is a major ECM constituent within AT (2). In obese AT, the levels of COL6 chains and their cleavage product endotrophin (ETP) are significantly elevated, influencing obesity-related metabolic disease onset (10). ETP plays a crucial role in the pathogenesis of various metabolic conditions, such as obesity, diabetes, chronic liver diseases, and cancer, exerting profibrotic, proinflammatory, and angiogenic effects within the microenvironment (10–15). Clinical studies have also explored serum ETP as a prognostic biomarker for predicting chronic fibrotic disease progression (16,17) and thiazolidinedione response in patients with type 2 diabetes (18,19). These have suggested that ETP release from COL6 fibrils is tightly controlled in a context-dependent manner, with its levels only increased under certain pathological conditions. However, details on the proteinase-mediated cleavage of the COL6A3 C-terminus for ETP generations are scarce. Studies have suggested that ETP-containing fragments of different sizes are released within connective tissues, while ETP is being generated through multistep processing (20). The extracellular metalloproteinase, bone morphogenetic protein 1, and matrix metalloproteinase 14 (MMP14) have been suggested to cleave COL6A3 by targeting distinct cleavage sites (14,20,21). While different forms of ETP and ETP-containing peptides have been characterized, whether the latter share biological activities with ETP or are processing intermediates remains unclear.

We previously reported that COL6A3 and ETP levels were upregulated in white AT (WAT) under obesity and diabetes (10). ETP is also detected in muscle and the skin, with lower levels reported in other connective tissues, including the cornea and cartilage (20). This is probably because the activity of specific proteinases generating ETP is regulated in a cell context-dependent manner. However, the specific mechanisms regulating ETP levels remain unclear. MMPs and tissue inhibitors of metallopeptidase (TIMPs) regulate homeostasis within the extracellular environment by facilitating the turnover of ECM proteins, such as collagens, aggrecan, proteoglycans, and fibronectin (22). In humans, >20 MMPs have been identified, structurally subdivided into basic MMPs, MMPs with fibronectin domain inserts, membrane-type MMPs, and minimal domain MMPs (22,23). Studies have reported important roles for MMPs and TIMPs in the progression of various diseases, highlighting their potential as drug targets (24). Hence, the aim of this study was to investigate the molecular mechanisms of COL6A3 processing to release ETP and develop an efficient therapeutics targeting ETP-induced pathologies in obese AT.

To generate adipocyte-specific HIF1α-KO mice (APN-HIF1αKO), Adipoq-Cre transgenic mice expressing Cre recombinase under the control of the mouse adiponectin (Adipoq) promoter/enhancer region within the bacterial artificial chromosome (BAC) transgene were crossed with HIF1α^flox mice (The Jackson Laboratory, no. 007561) having loxP sites flanking exon 2 of Hif1α. Mice were housed in cages under a 12-h dark-light cycle, with free access to water and standard diet. For the HFD feeding experiments, mice were fed a diet that provided 60% of calories from fat (Research Diets, catalog no. D12492). All experiments were conducted using littermate control male mice at 8 or 9 weeks old. Primer sequences for mouse genotyping are listed in Supplementary Table 1. All animal protocols were approved by the Ulsan National Institute of Science and Technology Institutional Animal Care and Use Committee (UNISTIACUC-19-35).

For the oral glucose tolerance test (GTT), mice were fasted for 4 h before glucose administration (1.5 g/kg by oral gavage; VWR, cat no. 50-99-7). For the insulin tolerance test (ITT), mice were fasted for 4 h before insulin injection (1 units/kg by intraperitoneal injection; Humulin R, Lilly, cat no. HI0210). At the indicated times, blood glucose levels were determined using a blood glucose meter (Accu-Chek) from tail vein clipping.

To isolate primary mouse adipocytes, epididymal WAT (eWAT) was harvested, washed with sterile PBS, and filtered through a 0.2-μm filter. Tissues were minced and digested with filtered working digestion buffer (1.5% BSA, 100 mmol/L HEPES [pH 7.4], 120 mmol/L NaCl, 50 mmol/L KCl, 5 mmol/L glucose, 1 mmol/L CaCl₂, and 0.75 mg/mL collagenase type IA [Sigma-Aldrich, cat no. C9891]). After digestion for 40 min at 37°C, samples were passed through a 100-μm strainer (SPL Life Sciences, Pocheon, South Korea, cat no. 93100). After centrifugation at 200g for 3 min, floating adipocytes or stromal vascular cells were harvested and seeded onto 24-well plates with medium (10% FBS and 1% penicillin/streptomycin in DMEM/F12).

Tumor necrosis factor α (TNFα) recombinant protein (rTNFα), free fatty acids (FFAs), and thapsigargin were used to treat 3T3-L1 adipocytes. microRNA (miR) mimics (miR-29s) and control were purchased from Genolution (Seoul, Korea) and transiently transfected into 3T3-L1 adipocytes using G-fectin (Genolution), following the manufacturer’s protocol. The chemicals and reagents used in this study are listed in Supplementary Table 2.

We used the Gaussia luciferase (GLuc)-based reporter system to determine ETP cleavage activity. For generating the ETP cleavage biosensor, a mouse interleukin (IL)-1β DNA fragment was cloned into the pRA vector with a prolactin signal sequence (PRL) at the N-terminus and a FLAG-tag at its C-terminus. Then, 7, 12, or 23 aa containing the ETP cleavage site (LMVSTEP, SSTINLMVSTEP, and PPPPQPARSASSSTINLMVSTEP, respectively) were inserted in the middle of the IL-1β DNA fragment. GLuc was cloned from pMCS-Gaussia Luc (Thermo Fisher Scientific) (25) and fused into the IL-1β C-terminus via a linker sequence. The DNA sequence of the ETP cleavage biosensor is given in Supplementary Table 3. ETP cleavage activities of MMPs were determined using a luciferase assay system (Promega) with conditioned media obtained from HEK293T cells transiently transfected with the biosensor, indicated MMPs, and a β-galactosidase (β-gal)–expressing vector using Lipofectamine 2000 (Invitrogen, Carlsbad, CA). β-Gal activity in cell lysates was used for normalization.

Conditioned media of C2C5-FLAG–overexpressing HEK293T cells was harvested after 24-h incubation and concentrated using Amicon Ultra-4 3K (Merck, Burlington, MA) at 13,000 rpm for 45 min at 4°C. Samples were subjected to 15% SDS-PAGE and transferred onto polyvinylidene fluoride membranes. Proteins at the 12-kDa and 25-kDa regions of the membranes were subjected to Edman N-terminal sequencing (Emass) to identify the five amino acid residues.

Pro-MMPs (MMPs) and catalytic MMPs (caMMPs) for the MMP2, 9, 11, 14, and 16 genes were cloned into pRA-green fluorescent protein (GFP) (26) vectors with a PRL signal sequence at the N-terminal region and a FLAG-tag at the C-terminal region. ETP cleavage site mutants (C2C5-L3097A, -M3098A, -V3099A, -S3100A, -T3101A, -LM3102AA, -VS3104AA, and -ST3105AA) were generated using the site-directed mutagenesis kit (Invitrogen) according to the manufacturer’s instructions. Primer sequences used for PCR are listed in the Supplementary Table 4.

The mRNA-miRNA paired gene expression data of WAT were obtained from Gene Expression Omnibus (mRNA: GSE25401; miRNA: GSE25470), which contains data from patients with obesity (n = 30) and lean controls (n = 26). Raw microarray intensity values were normalized via Robust Multiarray Average using the justRMA function in the affy R package. Briefly, raw intensity values were background-corrected, log₂-transformed, and quantile-normalized. For correlation analysis of two genes (or gene/miRNA), the Pearson correlation coefficient was determined.

Hypoxia response elements (HREs, gtatgtgc) within the Col6a3 (90845030 ∼90845037) and miR-29b1 promoters (31064055 ∼31064061, 31063314 ∼31063314, and 31063179 ∼31063179) were predicted using JASPAR. Col6a3 (90843939 ∼90845938) and miR-29b1 (31065058 ∼31063121) promoters were amplified via PCR and cloned into the pGL4.10 vector (Promega, Madison, WI). The 3′-untranslated region (UTR) of Col6a3 (90767411 ∼90767433), MMP2 (55505726 ∼55505748), MMP9 (46016406 ∼46016494), and MMP16 (88034873 ∼88034895, 88032639 ∼88032661, 88032666 ∼88032688) were predicted as complementary to the miR-29–binding sequence using miRTarBase and TargetScan 7.2. The 3′-UTR sequences were amplified via PCR and cloned into the pGL3-control vector (Promega). A luciferase-based reporter system (Promega) was used to measure promoter activity following the manufacturer’s protocol. β-Gal activity in cell lysates was determined using ortho-nitrophenyl–β-gal and used for normalization. Primer sequences used for PCR are listed in Supplementary Table 4.

All results are presented as the mean ± SE. Statistical significance between groups was determined via the two-tailed Student t test or one-way ANOVA. Differences with P values <0.05 were considered significant. Statistical analysis and graphs were generated using GraphPad Prism 7 software.

Data sets generated from current study are available from the corresponding author

To characterize human ETP and pre-ETP, we cloned cDNA sequences containing the C2–C5 domains of human COL6A3 (C2C5) with the prolactin signal sequence and FLAG-tag fused at the N- and C-termini, respectively (Fig. 1A). These were transiently transfected into HEK293T cells to evaluate ETP and pre-ETP secretion in conditioned media. Cleavage of C2C5 by endogenous MMPs (or other peptidases) resulted in ETP release. Conditioned media collected from control and C2C5-overexpressing HEK293T cells were resolved using SDS-PAGE. Various ETP cleavage peptides were detected using immunoblotting with the anti-FLAG antibody (Fig. 1B). We excised gel regions corresponding to ETP-containing proteins (20–37 kDa and 10–15 kDa [square boxes in Fig. 1B] predicted ETP size shifted due to the FLAG-tag) for N-terminal sequencing to determine cleavage sites. ARPAA was identified from a band corresponding to 20–37 kDa; MVSTE and TEPLA were identified from a smaller-sized band (10–15 kDa) (Fig. 1C). Consistent with the molecular mass detected in Fig. 1B, three predicted PAA sites within the C3 domain and two cleavage sites near C5 domain are predicted as involved in ETP release (Fig. 1D). We previously used recombinant human ETP (11), which has a 77-aa-long C5 domain (highlighted in yellow in Fig. 1D). While two different forms of ETP (77- and 80-aa) have been described, the major cleavage form is unknown (Fig. 1E). To identify cleavage sites involved in ETP release, we generated site-directed mutants at the C5 cleavage sites and analyzed ETP release in media of mutant-overexpressing HEK293T cells (Fig. 1F). ETP release was significantly reduced in media from cells expressing the LM mutant (LM3097AA), which contains a cleavage site for 80-aa-long ETP generation. The ST mutant (ST3100AA) has lesser impact on ETP release (Fig. 1F), suggesting that the L/MV cleavage site was crucial.

To identify the peptidases involved in ETP release, we first analyzed human MMP and TIMP levels in obese AT, comparing their levels to those in lean control AT (Fig. 2A). We selected MMP2, MMP9, MMP14, and MMP16 (Fig. 2B–E), showing a positive correlation with COL6A3 levels in the obese AT. Importantly, MMP2, MMP9, MMP14, and MMP16 levels were positively correlated with obesity (Fig. 2F–I). MMP2 and MMP9 are gelatinases, whereas MMP14 and MMP16 are membrane-type MMPs tethered to the plasma membrane (22). To quantify MMP enzymatic activity, we developed an ETP cleavage biosensor based on the defined cleavage sites of COL6A3 (Fig. 1D and E). Briefly, the G-luciferase reporter-based ETP cleavage biosensor is activated when the 12-aa peptide containing two cleavage sites (–SSTINL/MVS/TEP–) of COL6A3 gets cleaved in the presence of coelenterazine (Fig. 2J). Tissue extracts were isolated from the eWAT of mice fed the HFD for 11 weeks and analyzed for the ETP-12AA cleavage biosensor. ETP cleavage activity was significantly higher than in lean controls (Fig. 2K). Consistently, the degree of C2C5 cleavage for pre-ETP generation was also significantly elevated greater, as determined via Western blot analysis following incubation of conditioned media from C2C5 construct-overexpressing HEK293T cells with indicated tissue extracts (Fig. 2L and M), suggesting that MMP (or proteinase) activity that targets either pre-ETP or ETP is significantly upregulated in obese AT.

MMP14 was considered to generate the 91-aa ETP in human plasma through a putative MMP14 recognition site (PPQP) present in 14 aa ahead of human recombinant ETP (Fig. 1D) (14). To determine the activities of MMP candidates targeting ETP, we generated the ETP-23AA cleavage biosensor (–PPPPQPARSASSSTINL/MVST/EP–) containing an additional 11-aa putative MMP14 recognition site (14,21) and analyzed supernatants from HEK293T cells overexpressing each MMP candidate (Fig. 3A). MMP11, putatively involved in COL6 degradation, was included in this analysis as a control (27). Strikingly, ETP-23AA cleavage activity was significantly increased by the catalytic domain of MMP2 and MMP9 (caMMP2 and caMMP9, active forms) as well as MMP16, whereas this was not seen with MMP14 or MMP11 (Fig. 3A). The protein expression for each MMP used in this assay was confirmed via immunoblots (Fig. 3B). A shorter version of the ETP-cleavage biosensor, containing a 7-aa residue cleavage site (–L/MVST/EP–), yielded the same results (Supplementary Fig. 1A), suggesting that MMP2, MMP9, and MMP16 are directly involved in mature ETP generation.

In HEK293T cells, COL6A3-C2C5 was most efficiently cleaved by caMMP9 and caMMP16, as determined using immunoblots for ETP in cell lysates (Fig. 3C). Levels of ETP released in conditioned media were consistently increased by caMMP9 and caMMP16, and caMMP14 seemed responsible for the generation of pre-ETP ranging from 20 kDa to 30 kDa, as determined via immunoblots (Fig. 3D). Consistent results were obtained in the human hepatocyte cell line SK-HEP1 (Supplementary Fig. 2). To further confirm these results, we analyzed the direct interactions between ETP and MMPs using immunoprecipitation. FLAG-tagged candidate MMPs were coexpressed with or without ETP in HEK293T cells, and expressions were confirmed via immunoblotting (Fig. 3E). MMP2, MMP9, and MMP16 directly interacted with ETP, whereas MMP14 and MMP11 showed no interaction (Fig. 3F). Moreover, pharmacological or genetic ablation of MMP9 via AG-L-66085 or shRNA, respectively, significantly attenuated ETP release (Fig. 3G and H). shRNA-mediated MMP16 inhibition had the same effect (Fig. 3I). Thus, MMP2, MMP9, and MMP16 are the primary candidates involved in mature ETP generation, whereas MMP14 potentially targets different regions of the COL6A3 chain for generating pre-ETP.

Next, we investigated whether preprocessing of COL6A3 is required or mature ETP can be directly processed from COL6A3. To assess mature ETP generation from proETPs compared with the longer-sized COL6A3 precursor (C2C5), we generated a construct encoding pre-ETP (29 kDa, COL6A3 cleavage peptide IV, as indicated in Fig. 1B) and determined the levels of mature ETP cleaved by caMMP9. Indeed, mature ETP could be directly generated from C2C5 without preprocessing. Nevertheless, mature ETP was more efficiently generated from pre-ETP than from C2C5 (Fig. 3J). Thus, stepwise processing of COL6A3 via pre-ETP intermediates significantly facilitated mature ETP generation (Fig. 3K).

To understand how ETP release increases in adipocytes during obesity, we challenged 3T3-L1 adipocytes with various obesity-associated stimuli, such as hypoxia, inflammation, FFAs, and endoplasmic reticulum stress and then evaluated the ETP-12AA cleavage biosensor. All stimuli significantly elevated ETP cleavage activity in adipocytes (Fig. 4A), suggesting that the enzymatic activities of various MMPs (or other peptidases) associated with ETP release are increased under metabolic stress. To investigate the impact of hypoxia on ETP generation, total RNA was extracted from 3T3-L1 adipocytes exposed to low oxygen conditions (1% O₂) for 36 h. Col6a3 and Mmp mRNA was significantly increased under hypoxia (Fig. 4B and C). This was further verified via immunoblots for COL6A1, COL6A2, and COL6A3 as well, as HIF1α was used as a positive control for hypoxia-inducible protein. COL6A3 increased, and both COL6A3 and ETP release in adipocyte media consistently increased under hypoxia (Fig. 4D and E). Dimethyloxalylglycine (DMOG)-induced hypoxia in adipocytes led to identical results, with increases in COL6A3 expression and cleavage (Supplementary Fig. 3A–C). These results indicate that hypoxia increases the transcription and posttranslational proteolytic cleavage of COL6A3 in adipocytes, which leads to higher circulating ETP.

To determine whether HIF1α is involved in the regulation of hypoxia-inducible COL6A3 expression and MMP-mediated ETP release, we identified a putative HRE in the mouse Col6a3 promoter, 1.0 kb upstream of the transcription initiation site (Supplementary Fig. 3D). Chromatin immunoprecipitation-quantitative PCR analysis showed that HIF1α directly binds to the HRE within the mouse Col6a3 promoter in 3T3-L1 adipocytes under hypoxia (1% O₂) as opposed to normal conditions, with the Vegfa promoter used as a positive control (Fig. 4F). Consistently, luciferase activity at the mouse Col6a3 promoter (−1.3-kb wild-type) was significantly elevated by HIF1α-PPN, a DNA construct containing the HIF1α cDNA with alanine substitutions at Pro402, Pro564, and Asn803, thereby allowing constitutive HIF1α expression (28). This increase was abrogated by introducing mutations in the HRE (HRE mut) (Supplementary Fig. 3D), suggesting that Col6a3 transcription is induced by HIF1α. To directly assess the effects of HIF1α, we generated Hif1α knock-down 3T3-L1 adipocytes via shRNA. The hypoxia-induced upregulation of Hif1α and its target genes expression, such as Pdk1 and Pai1, was abrogated in shHIF1α-treated adipocytes in response to hypoxia (Fig. 4G and H). Mmp2, Mmp9, Mmp14, and Mmp16 were significantly elevated under hypoxia (1% O₂), but not in Hif1α knock-down adipocytes (Fig. 4I). Consistently, the expression of all COL6 chains, including Col6a1, Col6a2, and Col6a3, increased under hypoxia, but not in Hif1α knock-down adipocytes (Fig. 4J). Among the three α-chains of COL6, the protein levels of COL6A3 were most significantly increased under hypoxia and also abrogated under Hif1α knock-down (Fig. 4K). Hypoxia increased COL6A3 and ETP release by Ctrl adipocytes as opposed to shHIF1α-transduced adipocytes (Fig. 4L). These results highlighted the crucial role of HIF1α in COL6A3 expression and cleavage to ETP in adipocytes under hypoxia.

To check whether HIF1α in adipocytes is sufficient to control ETP-related pathological changes in the hypoxic AT during obesity, we generated an adipocyte-specific Hif1α KO mouse model (APN-HIF1αKO) by crossing adiponectin promoter-driven CRE mice with HIF1α-floxed mice (Fig. 5A and Supplementary Fig. 4A and B). We confirmed adipocyte-specific Hif1α KO efficiency within the eWAT of APN-HIF1αKO mice compared with their littermate controls (HIF1α^f/f) (Fig. 5B). The levels of Hif1α and its target genes, such as Glut1, Pdk1, and Pai1, were also significantly decreased in eWAT from APN-HIF1αKO mice compared with those in controls (Fig. 5C). To assess the metabolic phenotypes of APN-HIF1αKO mice compared with controls, we challenged a longer duration of HFD feeding (20 weeks). During the 20-week HFD feeding, body weight was comparable between the two groups (Fig. 5D). To evaluate systemic insulin sensitivity, we measured glucose tolerance and insulin response between APN-HIF1αKO and control mice at 16 weeks of HFD feeding, observing that HIF1α deficiency in adipocytes significantly alleviated HFD-induced systemic insulin resistance, as determined via the GTT and ITT (Fig. 5E and F). The gene expression of collagens and MMPs as well as inflammation and fibrosis marker genes was consistently downregulated in the eWAT of APN-HIF1αKO mice compared with n controls (Fig. 5G–J). Hematoxylin and eosin staining for eWAT revealed that “crown-like” structures in the AT of control mice were significantly decreased in APN-HIF1αKO mice (Fig. 5K). In addition, ETP accumulation and fibro-inflammation, as determined via Sirus Red and F4/80 staining, were dramatically decreased in the eWAT of APN-HIF1αKO mice (Fig. 5L–N). Hepatic steatosis was also ameliorated under adipocyte-specific HIF1α deficiency (Fig. 5O). Consistent with the decreased levels of ETP in the AT of obese APN-HIF1αKO mice, ETP cleavage activity was significantly lower, as determined based on the ETP-12AA cleavage biosensor (Fig. 5P). Furthermore, preprocessing of C2C5 to generate pre-ETPs was increased in the eWAT of obese mice and ameliorated in that of APN-HIF1αKO mice (Fig. 5Q and R). Lower ETP in the AT potentially accounts for the lower levels in circulation observed in APN-HIF1αKO mice compared with controls (Fig. 5S).

Of note, APN-HIF1αKO mice exhibited no significant changes in fibro-inflammation and systemic insulin sensitivity in the 12-weeks’ HFD-fed obesity (Supplementary Fig. 4C–G), similarly to the macrophage-specific Hif1α KO mouse model (6). These modest metabolic phenotypes in APN-HIF1αKO mice may be due to a significant contribution of stromal vascular cells at this stage as Col6 and Mmp genes were specifically downregulated in adipocytes, whereas their expression in stromal vascular cells was comparable to that in controls (Supplementary Fig. 4H). Importantly, COL6 chains, including COL6A1, COL6A2, and COL6A3, were mainly expressed by adipocytes, while MMP2, MMP9, and MMP16 were produced by stromal vascular cells in obese AT (Supplementary Fig. 4I and J), suggesting that adipocyte-specific suppression of COL6A3 and MMPs could not efficiently inhibit ETP generation.

In order to determine better ways for efficiently inhibiting ETP generation in obese AT in combination with adipocyte-specific HIF1α depletion, we analyzed hypoxia-responsive miRNAs as potential regulators of obesity-related AT alterations. miRNAs are small noncoding RNAs of ∼22 nucleotides that regulate transcription in various signaling networks, including cell differentiation, development, apoptosis, and tumorigenesis (29). In particular, miRNAs within AT have been recognized as crucial regulators of adipocyte function in obesity and related metabolic diseases (30–33). To determine whether miRNAs are involved in ETP release within the obese AT, we used hypoxia-linked global miRNA expression profiling by using a public database and identified 259 miRNAs. These were further evaluated based on their likelihood for targeting the 3′-UTR of human COL6A3 (BIMIR, TargetScan 7.2) (Fig. 6A). We identified seven candidate miRNAs (Supplementary Fig. 5A), and focused on the miR-29 family, consisting of miR-29a, miR-29b, and miR-29c, which target critical regulators involved in ECM synthesis and degradation (34–38). miR-29a transgenic mice previously exhibited the amelioration of HFD-induced obesity, hepatic steatosis, and liver fibrosis (39).

We examined miR-29 expression in 3T3-L1 adipocytes under hypoxia. miR-29a, -b, and -c rapidly decreased within 1 h, and this decrease was maintained over the course of hypoxia (Fig. 6B), suggesting that hypoxia-linked miR-29 suppression is associated with initial obesity-related pathological changes, such as fibrosis and inflammation. To determine whether HIF1α and HIF2α are involved in hypoxia-inducible miR-29 repression, we identified three HREs within the mmu–miR-29a/miR-29b1 proximal promoter and verified the functionality of this element by generating a luciferase reporter fusion (Supplementary Fig. 5B). miR-29 subtypes, including miR-29a, -29b, and -29c, are located on two separated chromosomes, with miR-29a/miR-29b1 being on chromosome 7 in humans and 6 in mice, whereas miR-29b2/miR-29c are on chromosome 1 in both species (40,41). Notably, there are no HRE sites in the mmu–miR-29b2/miR-29c promoter regions, as determined via JASPAR analysis. The miR-29b1–luc reporters were activated through cotransfection with HIF1α or HIF2α, whereas the reporter containing deleted or mutated HRE exhibited lesser activation (Supplementary Fig. 5B). These findings revealed that neither HIF1α nor HIF2α are involved in hypoxia-induced miR-29 repression. Nevertheless, the miR-29b1 promoter exhibited strong inhibition of luciferase activity under hypoxia, confirming the hypoxia-sensitive transcriptional repression of miR-29 (Supplementary Fig. 5C). We examined miR-29 levels in the eWAT of wild-type mice following either acute (7-day) or chronic (11-week) exposure to the HFD. miR-29 expression was consistently decreased in eWAT at the early stages of obesity, which continued into the advanced stages of obesity (Fig. 6C and D). In human visceral AT, miR-29a and miR-29b levels were negatively correlated with those of COL6A3 in both lean and obese tissue (Fig. 6E and F), indicating an association with COL6A3 expression. However, miR-29c levels did not correlate with COL6A3 (Fig. 6G). Consistently, we analyzed the correlation between miR-29b expression and MMP2, MMP9, as well as MMP16. MMP9 and MMP16, but not MMP2, showed a strong negative correlation with miR-29b in human visceral AT (Fig. 6H–J). These results suggest that the hypoxia-induced miR-29 repression regulates COL6A3 and MMP expression in obese AT.

We identified COL6A3 and associated MMPs as predicted targets of miR-29, which is known to target multiple genes associated with fibrosis, including collagens and ECM modifiers (42,43). Analysis of the 3′-UTRs of COL6A3, MMP2, MMP9, and MMP16 revealed an evolutionarily conserved seed region for miR-29 in both human and mouse genes (Supplementary Fig. 5G). To further confirm the functionality of these seed regions, we generated reporter constructs in which the luciferase-encoding sequence was fused to the 3′-UTRs. miR-29 markedly suppressed the luciferase activity of Col6a3, MMP2, MMP9, and MMP16 constructs (Fig. 6K–N), whereas mutating the seed regions within the Col6a3 3′-UTR reporter construct abrogated the inhibitory effects of miR-29 (Fig. 6K), indicating direct involvement of miR-29 in COL6A3 and MMP transcriptional repression. To analyze the effect of miR-29 on ETP-associated adipocyte function during hypoxia, we analyzed the mRNA levels of collagens, MMPs, and fibro-inflammation marker genes. Both Col6a3 and Mmp genes were significantly upregulated under hypoxia, which was abrogated under miR-29 treatment (Fig. 6O and P). In addition, ETP-related fibrosis and inflammation were significantly alleviated following treatment with miR-29 mimetics under hypoxia (Fig. 6Q and R). To further verify the effects of miR-29 on COL6A3 levels and ETP release by adipocytes, we treated 3T3-L1 adipocytes with miR-29 mimics, observing a decrease in both endogenous COL6A3 levels (Fig. 6S) and ETP release into the conditioned media (Fig. 6T).

Compared with the therapeutic effects of HIF1α-targeting drugs, such as PX-478, on HFD-induced obesity (8), the metabolic benefits of HIF1α genetic ablation in adipocytes seem limited during the early stage of obesity, partly due to a stromal contribution to ETP generation (Supplementary Fig. 4C–G), Nevertheless, HIF1α genetic ablation exerted clear beneficial effects in mice subjected to the chronic HFD challenge (Fig. 5). Thus, we evaluated whether miR-29 confers synergistic or additive therapeutic benefits into HFD-fed obese APN-HIF1αKO mice with regard to the suppression of ETP generation. APN-HIF1αKO mice and their littermate controls were challenged with the HFD for 8 weeks to induce a relatively early phase of obesity, and adenoviral miR-29 or adenovirus (Ad)-GFP control was ectopically injected into the subcutaneous WAT (sWAT) at 4 and 6 weeks of HFD feeding (Fig. 7A). Body weights for both groups of mice were not altered under Ad–miR-29 overexpression compared with Ad-GFP (Supplementary Fig. 6A). HIF1α immunostaining of obese AT indicated that HIF1α levels were significantly decreased in APN-HIF1αKO mice compared with controls, whereas blood vessel density, as determined via endomucin staining, was comparable between the two groups (Fig. 7B). Ad–miR-29 was efficiently overexpressed in sWAT and eWAT, while not detected in other tissues, including the liver and skeletal muscle (Fig. 7C and Supplementary Fig. 6C and D). The mRNA levels of Col6 and Mmp genes as well as inflammation and fibrosis markers in the eWAT of APN-HIF1αKO mice was significantly decreased under miR-29 overexpression, suggesting that HIF1α deficiency in adipocytes plus miR-29 overexpression in AT effectively suppressed ETP generation and fibro-inflammation even in the 8-week HFD challenge condition (Fig. 7D–G). Histological analysis of AT via hematoxylin and eosin staining revealed that the levels of “crown-like” structures were significantly decreased in APN-HIF1αKO control mice; further, this effect was more prominent under conditions of miR-29 overexpression. Hepatic steatosis was also significantly ameliorated after miR-29 overexpression in APN-HIF1αKO mice (Fig. 7H), indicating that miR-29 was beneficial for the suppression of obesity-related chronic inflammation and hepatic steatosis in HIF1α-deficient mice. Importantly, overexpression of miR-29 efficiently decreased ETP levels in the eWAT of control mice, and these effects were more pronounced in APN-HIF1αKO mice (Fig. 7I). Consistently, the combination of miR-29 with HIF1α deficiency alleviated local fibro-inflammation in AT to the greatest extent, as determined via Picro Sirius Red staining and F4/80 immunohistochemistry, respectively (Fig. 7J and K). Notably, ectopic injection of Ad–miR-29 into the AT of mice from both groups had no effect on the mRNA levels of Col6a3 and MMPs or on ETP levels in liver or skeletal muscle (Supplementary Fig. 6E–G). Consistently, ETP-12AA cleavage biosensor analysis revealed that local injection of Ad–miR-29 significantly suppressed the ETP cleavage activity within the AT of both APN-HIF1αKO and control mice (Fig. 7L and Supplementary Fig. 6H). Thus, miR-29 overexpression in AT efficiently suppressed both COL6A3 and MMPs, inhibiting ETP generation. Circulating ETP was most significantly decreased in Ad–miR-29–overexpressing APN-HIF1αKO (Fig. 7M).

GTTs were performed to explore the roles of miR-29 in APN-HIF1αKO mice and controls under obesity-induced insulin resistance. In control mice, there were no significant differences in GTTs between Ad-GFP– and Ad–miR-29–injected groups, whereas insulin sensitivity was significantly ameliorated by miR-29 in APN-HIF1αKO mice under 8 weeks of HFD feeding (Fig. 7N and O). The levels of circulating triglycerides, cholesterol, and FFAs were significantly decreased by miR-29 in APN-HIF1αKO mice, whereas these effects were not observed in littermate controls (Fig. 7P–R). Local injection of miR-29 in AT caused no hepatic toxicity based on circulating ALT and AST, with the ALT levels showing even further improvement in APN-HIF1αKO mice (Supplementary Fig. 6I and J). This was further confirmed with RNA sequencing analysis for eWAT of each group of mice, revealing that biological functions associated with fibrosis, inflammation, and insulin resistance were most significantly decreased by miR-29 overexpression in AT in the APN-HIF1αKO mice (Fig. 8A). Thus, miR-29 in combination with HIF1α deficiency within AT significantly ameliorated obesity-induced systemic insulin resistance, hepatic steatosis, and metabolic indexes when compared with either condition alone (Fig. 8B).

The rapid expansion of AT with significant ECM remodeling observed in obesity leads to the development of a hypoxic microenvironment. Although persistent hypoxia can trigger pathological alterations in AT, such as tissue fibrosis, chronic inflammation, and systemic insulin resistance, the specific mechanisms linking hypoxia to these pathological alterations remain elusive. ETP, a signaling peptide cleaved from the COL6A3 chain, has emerged as a target mediator of pathological processes within the obese AT. Clinical data linking high ETP in obese AT to diabetes development suggest that both COL6A3 levels and proteinases activity drive these processes.

We identified three specific proteinases that target the consensus cleavage site (–L/MVST/EP–) on COL6A3. MMP2, MMP9, and MMP16 directly interacted with ETP and predominantly cleaved –L/M– sites at its C5 domain, generating 80-aa-long human ETP. Additional proteinases regulated in a cell context-dependent manner may be involved in COL6A3 processing for the generation of mature ETP, which requires further investigation. Previously, the nature of locally acting and circulating ETP in humans has been studied by immunoprecipitating it from human plasma with a monoclonal antibody against the C-terminal region of what we predicted to be human ETP, detecting a 91-aa-long protein (14). While MMP14 was predicted as the protease-generating 91-aa-long ETP, this site was not recognized and cleaved by MMP14 based on the ETP-23AA cleavage biosensor analysis (Fig. 3A and F). Furthermore, ETP is efficiently generated through stepwise processing of COL6A3. COL6A3 preprocessing generated multisized pre-ETPs (ranging from 22 kDa to 29 kDa, containing partial C3 domain to C5 domain of COL6A3), which were then sequentially processed by MMP2, MMP9, and MMP16 to yield mature ETP. This stepwise processing significantly enhanced mature ETP generation from COL6A3 (Fig. 3J and K). Of note, MMP14 seems to be indirectly involved in the process by activating MMP2 and MMP9 (44–46) or generating pre-ETPs. Nevertheless, the roles of MMP14 in COL6A3 processing require further study.

Mechanistically, hypoxia promotes ETP release through an upregulation of HIF1α transcriptional activity and miR-29 repression, which upregulated Col6a3 as well as Mmp2, Mmp9, and Mmp16. Ad-mediated AT overexpression of miR-29 in APN-HIF1αKO mice resulted in greater systemic insulin sensitivity compared with mice receiving Ad-GFP or miR-29–overexpressing control mice, indicating that the miR-29 pathway synergistically enhances systemic metabolism under HIF1α deficiency conditions in obese AT. Strikingly, ETP accumulation was also significantly ameliorated in APN-HIF1αKO under miR-29 overexpression. Furthermore, miR-29 alleviated fibro-inflammation, but not hepatic steatosis and systemic insulin sensitivity in the AT of HFD-fed control mice, implying that ectopic miR-29 overexpression in AT synergistically ameliorated systemic insulin resistance under HIF1α inhibition. As opposed to its downregulation in AT, miR-29a was increased in the liver and skeletal muscle (Supplementary Fig. 5D–F), suggestive of tissue-specific regulatory mechanisms for miR-29 in subjects with obesity. Further, it has been suggested that increased levels of miR-29 in pancreatic β-cells, muscle, and the liver are associated with characteristics of diabetes, including lower insulin secretion (47) and sensitivity (48,49). In the metabolism perspective, distinctive roles of miR-29 in various tissues remain to be addressed in future research.

Compared with aP2 promoter-driven Hif1α KO mice (4,5), either macrophage-specific Hif1a KO mice (7) or adipocyte-specific adiponectin promoter-driven HIF1α KO mice, as shown in this study, exhibited more favorable metabolic phenotypes following longer HFD feeding (18- to 20-weeks of HFD) relative to controls, indicating that Hif1a in both stromal cells and adipocytes plays crucial roles in the hypoxic AT during obesity-related pathologies (Supplementary Fig. 7A). In the same line, pharmacological inhibition conferred greater beneficial effects than genetic ablation in AT, implying that HIF1α in other metabolic tissues could play a role in HFD-induced insulin resistance. In addition, drug targets other than HIF1α are likely involved. Although HIF1α represents an attractive therapeutic target, only a limited number of drugs are available for clinical trials due to toxicity. Extensive efforts have been made to target HIF1α for cancer therapy and obesity-related diabetes (50–52). Based on the current findings, hypoxialinked miR-29 signaling within obese AT represents a potential therapeutic target, which, in combination with HIF1α-targeted therapy, may efficiently suppress ETP generation, thereby decreasing fibro-inflammation in AT and ameliorating insulin sensitivity (Fig. 8B). Thus, our results can be used to develop therapeutic strategies combining low dosages of HIF1α-targeting drugs with miR-29 to treat metabolic diseases associated with elevated ETP in circulation or local AT. This therapeutic strategy may also be used in cancer and various fibrotic conditions characterized by ETP upregulation. Our data suggest that ETP generation is tightly controlled in a cell context-dependent manner. Thus, we are considering that ETP cleavage-associated miR pathways during hypoxia could be subject to similar variation, which needs to be further explored.

Acknowledgments. The authors thank Ulsan National Institute of Science and Technology (UNIST) Central Research Facilities (UCRF) for help with in vivo animal studies and S.M. Lee and H. Lee (UNIST, Ulsan, Korea) for the technical support.

Funding. This work was supported by grants from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (HI14C1277) and the National Research Foundation (NRF) through the Bio-Synergy Research Project of the Ministry of Science, ICT and Future Planning (2021M3A9C4000963) and the Basic Science Research Program (NRF-2018R1A2B6003878, NRF-2018R1A5A1024340, 2021R1A2C2005499) to J.P and (NRF-2020R1A2C4001503) to M.K.

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

Author Contributions. W.J. conceived the project, performed most of the experiments, analyzed data, and wrote the manuscript, M.K. provided critical comments, performed in vivo and in vitro experiments, and analyzed data. J.O., C.P., C.L., and S.K. performed experiments. C.-S.K. performed HIF1αKO animal studies and analyzed data. S.Y. and D.N. performed human data analysis. All authors approved the manuscript. J.P. designed the study, analyzed data, supervised the projects, and wrote the manuscript. J.P. is the guarantor of this work and, as such, has 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.