Hypoxic microenvironment has been associated with cancer progression and stemness maintenance. Hypoxia changes the expression of several target genes, but the mechanism underlying the hypoxia-mediated pleiotropic effects on metabolic status, cell cycle, differentiation, angiogenesis, cell migration, and metastasis is unclear (Moon et al., 2015). Hypoxia-inducible factor (HIF) is a master transcription factor comprising HIF-α and HIF-1β (also known as aryl hydrocarbon nuclear translocator, ARNT) subunits. HIF-α subunit is sensitive to changes in O₂ concentration because its stability and trans-activity are regulated by two different HIF-α-specific O₂ and a-ketoglutarate-dependent dioxygenases, namely, prolyl hydroxylases (PHD) and asparaginyl hydroxylase (also known as factor-inhibiting HIF, FIH). HIF-1α is hydroxylated at proline 564 and/or 402 residues by PHD using O₂. The hydroxylated proline residues of HIF-1α are recognized by an E3 ubiquitin ligase, von Hippel Lindau protein, that initiates the ubiquitin-dependent degradation of HIF-1α. FIH hydroxylates the asparagine 803 residue of HIF-1α, and prevents the binding of the coactivators CBP/p300 to the C-terminal region of HIF-1α. These two enzymes are inactivated under hypoxia, leading to the stabilization of HIF-1α and its subsequent translocation into the nucleus to form a heterodimer with HIF-1β. HIF-1 heterodimer interacts with E-box-like hypoxia-responsive element to induce the expression of target genes (Gorlach et al., 2015).

Hypoxic regions in human body are often found where O₂ consumption exceeds O₂ supply. HIF-1α is activated at the sight of inflammation because inflammatory processes increase O₂ consumption owing to an increase in oxidation reactions catalyzed by NADPH oxidases and an increase in the production of reactive oxygen species, which are inhibitors of PHD and FIH (Lee et al., 2011). O₂ supply is limited in avascular regions such as the inner core of solid tumors, stem cell niche, and the obese adipose tissue. In the obese adipose tissue, hypertrophic adipocytes move further away from blood vessels and become hypoxic. Adipocytes have been re-evaluated for their endocrine functions because they secreted many cytokines and hormones such as inflammatory cytokines, leptin, and adiponectin (Makki et al., 2013). These secreted proteins from adipocytes are termed as adipokines. In comparison with lean adipocytes, obese adipocytes upregulate the expression and secretion of many inflammatory adipokines such as interleukin-6, leptin, and plasminogen activator inhibitor-1, most of which are target genes of HIF-1α (Lee et al., 2014). Beside adipocytes, the adipose tissue is made of other cells such as endothelial cells, macrophages, and mesenchymal stem cells. In our previous microarray analyses, we identified several novel target genes that are upregulated by hypoxia in human adipose-derived stem cells (hADSCs) (Lee et al., 2017). We first found that the expression of carboxypeptidase A4 (CPA4) and carboxypeptidase E (CPE) was highly upregulated in hypoxic hADSCs (Lee et al., 2017). The hypoxic environment of hypertrophic adipocytes increases the expression of CPA4 and CPE, which function by processing other secreted proteins and matrix proteins in the adipose tissue. This finding indicates that hypoxia changes the microenvironment of the obese adipose tissue by inducing the expression of not only adipokines but also peptidases such as CPA4 and CPE from adipose-derived mesenchymal stem cells.

Materials

Anti-HIF-1α (610958) and anti-CPE (610758) antibodies were purchased from BD Biosciences (USA). Anti-CPA4 (ab81543) and anti-14-3-3γ (MAB5700) antibodies were obtained from Abcam (USA) and R&D Systems (USA), respectively. Mouse IgG (sc-2025) and anti-H3K4me3 (9751S) were obtained from Santa Cruz Biotechnology (USA) and Cell Signaling Technology (USA), respectively. Ponceau S solution and Oil Red O stain were supplied by Sigma-Aldrich (USA).

Cell culture and adipogenesis

hADSCs (catalog No. 510070, lot No. 2033) were purchased from Invitrogen (USA) and maintained as previously described (Lee et al., 2017). IMR90 human fetal lung fibroblasts were obtained from the American Type Culture Collection (ATCC, USA) and maintained in an Eagle’s minimum essential medium (EMEM) supplemented with 10% (v/v) fetal bovine serum. Cells were exposed to hypoxia through incubation in a Forma anaerobic incubator (model 1029; Thermo Fisher Scientific, USA) in an atmosphere of 5% CO₂, 95% N2, and 0.5% O₂. HIF-1α expression was knocked down in hADSCs using a pLKO.1 lentiviral vector (Addgene, USA) encoding either control a short-hairpin RNA (shRNA) or a shRNA sequence against human HIF-1α (5′-CCA GTT ATG ATT GTG AAG TTA-3′). For adipocyte differentiation, hADSCs were cultured in an adipogenic medium (catalog No. A1007001; Invitrogen); the medium was replaced every 3 days for 12 days according to the manufacturer’s instructions. Mouse adipocyte progenitor cells were isolated from the subvascular fraction (SVF) of the subcutaneous white adipose tissue (sWAT) of an 8-week-old C57BL/6J strain using an adipose tissue progenitor isolation kit (Miltenyi Biotec, Germany) according to the manufacturer’s instructions. The SVF was obtained after collagenase digestion of the sWAT using a previously described method (Freshney, 2000). Conditioned medium (CM) was a spent medium harvested from hADSCs cultured under normoxia (21% O₂) or hypoxia (0.5% O₂) for 2 days (Park et al., 2013). CM was centrifuged to remove debris and its supernatant was used.

Reverse-transcription quantitative polymerase chain reaction (RT-qPCR)

Steady-state RNA expression was measured by real-time qPCR using Power SYBR Green PCR master mix on a QuantStudio 3 real-time PCR system (Applied Biosystems, USA). The Ct value of target mRNA was normalized against that of endogenous 18S rRNA (ΔCt = Ct target – Ct 18S). Ct is the threshold cycle of qPCR defined by the QuantStudio 3 real-time PCR system. The relative mRNA level of a target gene is obtained by 2_-ΔΔ_Ct; ^ΔΔCt = ΔC_{t treated value} – ΔC_{t untreated value}. Pairs of forward and reverse primer sequences for RT-qPCR were as follows: human CPA4, 5′-CAA TGA AGG GCA AGA ACG GAG C-3′ and 5′-GGT CAG GAA AGT CTG CGG CAA T-3′; human CPE, 5′-CTT GGC CCA GTA CCT ATG CAA-3′ and 5′-ACC AGT CCT TGA GTT CAC CAG-3′; human PPARγ2, 5′-GCG ATT CCT TCA CTG ATA C-3′ and 5′-CTT CCA TTA CGG AGA GAT CC-3′; human HIF-1 α, 5′-CAT CAG CTA TTT GCG TGT GAG GA-3′ and 5′-AGC AAT TCA TCT GTG CTT TCA TGT C-3′; mouse CPA4, 5′-CAG CTG CTG ATG TAT CCC TAT G-3′ and 5′-TAC TTA GTG CCC GAG AGA GAA-3′; mouse CPE, 5′-TGA GAA AGA AGG CGG TCC TAA C-3′ and 5′-GCA GAT TGG CAG AAA GCA CAA-3′.

Chromatin immunoprecipitation (ChIP)

ChIP analysis was performed as previously described (Lee et al., 2017). The Ct value of target gene from the immunoprecipitated DNA sample was normalized against the Ct value of target gene in input DNA (ΔCt = Ct _sample – Ct _input). The percentage of input indicated the value of 100 × 2Δ_Ct. Forward and reverse primer sequence pairs used for ChIP-qPCR were as follows: human CPA4 (–1.6 kb), 5′-ACA GAT TTG GGT CTC AGT TGT TCT-3′ and 5′-AGA GCC TTT GTT AGG AGC TGT G-3′; human CPA4 (–0.9 kb), 5′-GGC GTT ATG GAG ATA TGA GTG TCT TC-3′ and 5′-CTT CCT AGA CTG TGG GTC TGG T-3′; human CPA4 (transcription start site [TSS]), 5′- GAA TGG GAG AGG CAT TGG TAG G-3′ and 5′-CAG TGG CTG AGT CAA GCT GTA T-3′; human CPA4 (+0.7 kb), 5′- TGG TGC TTT CTT ACT GGC TTC T-3′ and 5′-CCT ACC ACC CAA ATG TCC TCT C-3′; human CPE (–0.9 kb), 5′-CAG CAC CTT ACC AAC CAT GTC T-3′ and 5′-TGG ACA CCA TGC GGG TAA TTT-3′; human CPE (TSS), 5′-GTG GGT CAT GGG TGT CTC AAT-3′ and 5′-AGG GCT GCT GGC GTT ATT-3′; human CPE (+0.9 kb), 5′-ACC TTA GGC CAC CCT CTA GTA A-3′ and 5′-ACT CTT GTC AGC AAG GAA GAC C-3′; human MT3 (–0.4 kb), 5′-CAG CAA CCC GTC TGA ACA-3′ and 5′-CCA TCT CCG AAA GTG GAG AAA-3′.

Statistical analysis

All quantitative measurements were performed through at least two independent experiments. For comparison between two groups, P values were calculated using the paired two-tailed Student’s t-test. A value of P < 0.05 was considered statistically significant.

Hypoxia increases the expression of CPA4 and CPE in hADSCs

The microarray data of hypoxic hADSCs revealed the hypoxia-mediated increase in the mRNA expression levels of several carboxypeptidases (Lee et al., 2017). Among the 21 carboxypeptidases, CPA4 and CPE mRNA expression was upregulated in response to hypoxia (Fig. 1A). RT-qPCR show that the hypoxia (0.5% O₂, 48 h) increased the mRNA levels of CPA4 and CPE by 13.5- and 3.9-fold, respectively (Figs. 1B and 1C). Anti-CPE antibody detected the hypoxic induction of CPE protein expression in hADSCs, while anti-CPA4 antibody detected the slight increase in the level of CPA4 protein in hypoxic hADSCs (Figs. 1D and 1E). Furthermore, western blot analysis using anti-CPE antibody showed that the secreted CPE protein was higher in the hypoxic CM of hADSCs than in the normoxic CM (Fig. 1F). However, we failed to detect any secreted CPA4 in CM using anti-CPA4 antibody (data not shown). These results indicate that hypoxia increases the expression of CPA4 and CPE in hADSCs.

We exposed IMR90 human fetal lung fibroblasts to hypoxia to investigate whether hypoxia induces the expression of CPA4 and CPE in other primary human cells. In comparison with IMR90 fibroblasts, hADSCs were more sensitive to hypoxia in terms of induction of CPA4 and CPE gene expression (Figs. 2A and 2B). To test whether hypoxia also induces CPA4 or CPE gene expression in mouse primary adipocyte progenitor cells, we isolated adipocyte progenitor cells from the sWAT and exposed them to hypoxia. In contrast to the observations reported for hADSCs, hypoxia induced neither CPA4 nor CPE expression in mouse adipocyte precursor cells (Figs. 2C and 2D). Further, we failed to detect upregulation in CPA4 and CPE gene expression in response to hypoxia in mouse embryo mesenchymal C3H10T1/2 cells and mouse 3T3-L1 preadipocytes (Figs. 2C and 2D). These results indicate that human and mouse adipocyte progenitor cells respond differently to hypoxia for the induction of CPA4 and CPE gene expression.

CPA4 and CPE genes maintain their sensitivity to hypoxia in mature adipocytes

We also tested whether hADSCs maintain their responsiveness to hypoxia for the induction of CPA4 and CPE gene expression after their differentiation into mature adipocytes. We treated hADSCs with an adipogenic medium for 12 days to induce adipogenesis (Fig. 3A). Oil Red O staining and RT-qPCR analysis revealed the accumulation of lipid droplets and induction of peroxisome proliferator-activated receptor (PPAR)γ-2, a master adipogenic transcription factor, respectively, confirming the successful differentiation of hADSCs into mature adipocytes (Figs. 3B and 3C). During adipogenesis, CPA4 mRNA expression was reduced but CPE mRNA level increased (Fig. 3C). After adipogenesis, mature adipocytes were exposed to hypoxia and their mRNA levels of CPA4 and CPE were measured (Fig. 3A). Hypoxia also increased mRNA of CPA4 and CPE even in mature human adipocytes (Figs. 3D and 3E). Western blot analyses detected the hypoxic increase of CPE protein in human mature adipocytes which were differentiated from hADSC (Fig. 3F). Thus, CPA4 and CPE genes maintained their sensitivity to hypoxia after the differentiation of hADSCs into mature adipocytes.

HIF-1α mediates hypoxia-induced CPA4 and CPE expression

To test whether HIF-1α is essential for mediating the hypoxia-induced expression of CPA4 and CPE genes, we knocked down the expression of HIF-1α in hADSCs by infecting them with a lentivirus encoding shRNA against HIF-1α. As control, hADSCs were infected with a lentivirus encoding a control shRNA. The knockdown of HIF-1α expression resulted in a decrease in the mRNA level of HIF-1α along with a decrease in the expression of its target gene, MT3 (Figs. 4A and 4B). In control hADSCs, hypoxia induced CPA4 and CPE gene expression by 14- and 3.5-fold, respectively. In HIF-1α knockdown hADSCs, hypoxia increased the mRNA levels of CPA4 and CPE only by 1.6-fold (Figs. 4C and 4D). These results indicate that HIF-1α is essential for mediating the hypoxia-induced upregulation in both CPA4 and CPE gene expression. Using JASPAR program, we found putative HIF-1α binding sites in the upstream regions of these two genes (Figs. 4E and 4F). ChIP-qPCR analysis showed that HIF-1α binding was detected on the promoters of CPA4, CPE, and MT3 genes (Figs. 4G-4I). These results suggest that HIF-1α binds to the promoters of both CPA4 and CPE genes and is essential for the hypoxic induction of these carboxypeptidase expression. To investigate the epigenetic changes in histone modification, we detected the hypoxia-mediated changes in a positive histone mark, H3K4me3, at the promoters of CPA4 and CPE genes. ChIP-qPCR analysis showed that H3K4me3 level significantly increased at the promoters of both CPA4 and CPE genes (Figs. 4J and 4K). Thus, hypoxia increased the transcription of CPA4 and CPE genes and this effect was accompanied by the recruitment of HIF-1α and increment in the active histone modification H3K4me3 at their promoter regions.

This study shows that hypoxic condition dramatically induced the expression of two carboxypeptidases CPA4 and CPE in an HIF-1α-dependent manner. The hypoxic responsiveness of CPA4 and CPE genes is restricted to human mesenchymal cells, and the adipocytes differentiated from these stem cells. In mouse adipocyte precursor cells, hypoxia failed to induce the expression of Cpa4 and Cpe genes (Figs. 2C and 2D). Although mice have been used for an animal model of human obesity studies, there is a lack of studies showing the similarities and differences in function and gene expression of adipose tissues between human and mouse (Chusyd et al., 2016; Lindgren et al., 2009; Vohl et al., 2004). A few comparative analyses of differentially expressed gene between WAT and brown adipose tissue revealed that human and mouse share less than ten percent of genes which are commonly regulated during browning process of adipose tissues (Baboota et al., 2015). Hypoxia increased expression of CPA4 and CPE genes in hADSC of human but not Cpa4 and Cpe genes in preadipocytes of mouse. It remains to investigate the different hypoxic signaling pathways to regulate the expressions of these genes between human and mouse. Carboxypeptidases cleave peptides from their C-termini, usually one residue at a time. CPE and CPA4 belong to the metallo-carboxypeptidase M14 family, which comprises 25 distinct genes (Sapio and Fricker, 2014). Carboxypeptidases have broad substrate specificity; the members of the A family cleave C-terminal aromatic/aliphatic residues, while those of the E subfamily remove C-terminal lysine and arginine residues without further hydrolyzing the peptide (Fricker and Snyder, 1983). CPA4 was also found to be upregulated in several tumors such as colorectal cancer, hepatocellular carcinoma, gastric cancer, esophageal squamous cell carcinoma, pancreatic cancer, and breast cancer (Handa et al., 2019; Pan et al., 2019; Sun et al., 2016a; 2016b; 2017; Zhang et al., 2019). CPA4 is closely associated with cancer metastasis. Patients with high levels of CPA4 had poorer prognosis. Serum levels of CPA4 may serve as a marker of poor prognosis and CPA4 could be a promising therapeutic target in several tumors with aggressive phenotypes. Similar to CPA4, CPE expression is detected in tissues that secrete bioactive peptides and is associated with peptide-containing secretory granules (Ji et al., 2017). CPE knockout mice show characteristics such as obesity, low bone mineral density, infertility, and neurodegeneration (Che et al., 2001; Fricker et al., 2000). Peptidomic analysis using brains of CPE knockout mice identified several CPE substrate neuropeptides, and their expression was severely dysregulated in mutant mice (Fricker et al., 2000). CPE is involved in the peptide secretion process. A membrane-associated form of CPE binds to proopiomelanocortin, a multivalent prohormone, to facilitate its trafficking from the trans-Golgi network into secretory granules (Cawley et al., 2016).

Considering that hypoxic regions are often found in the obese adipose tissue, our findings suggest the possibility that hypoxia-mediated induction of CPA4 and CPE expression in the obese adipose tissue may contribute to the upregulation in secreted CPA4 and CPE proteins in the serum, ultimately leading to increased tumor metastasis. Hypoxia is an important environmental condition that determines cell fate by reconstituting microenvironment through induction of target secretory proteins. This study suggests that hypoxic induction of secretory CPE and CPA4 proteins in hADSCs further changes the extracellular microenvironment by processing diverse extracellular substrates of CPE and CPA4.

This work was supported by the 2018 sabbatical year research grant of the University of Seoul for H Park.

Y.M., R.M., and H.P. conceived ideas, performed experiments and wrote the manuscript. H.R., S.C., and S.L. performed experiments.

The authors have no potential conflicts of interest to disclose.