Adipose Tissue Hypoxia in Obesity and Its Impact on Adipocytokine Dysregulation

All animals were purchased from CLEA Japan, were housed in a room under controlled temperature (23 ± 1°C) and humidity (45–65%), and had free access to water and chow (Oriental Yeast Co.). We used male mice for the high-fat diet–feeding study and female mice for the KKAy mice study. For diet-induced obesity studies, male C57BL/6J mice were divided at random into two groups at 8 weeks of age. The first group was fed a high-fat diet containing 30% fat by weight (AIN93G), while the second group was fed normal chow containing 5.9% fat by weight (CRF-1; Oriental Yeast Co.) for 8 weeks. Normal chow-fed and high-fat–fed mice were killed at 16 weeks of age, and female C57BL/6J and female KKAy mice were killed at 10–11 weeks of age. Tissues were dissected and frozen in liquid nitrogen. Samples were stored at −80°C until use. For analysis of arterial blood gases, blood samples were collected from the left carotid artery and measured by a blood analyzer (i-STAT; Fuso Pharmaceutical Industries, Tokyo, Japan).

A Hypoxyprobe-1 Plus kit (Chemicon International, Temecula, CA) was used for the detection of tissue hypoxia. Mice were injected with 40 mg/kg pimonidazole intraperitoneally 1 h before they were killed. Organs were removed immediately, fixed in 10% neutral buffered formalin for 24–48 h, and then processed into paraffin blocks. The sections were prepared according to the instructions provided by the manufacturer, counterstained with hematoxylin, and analyzed in a standard fashion.

Mice were anesthetized by intraperitoneal injection of pentobarbital, and a polyethylene catheter was positioned in the aortic arch via the left carotid artery. Yellow dye microspheres (15.5 μm diameter, 4 × 10⁴ beads; Triton Technology, San Diego, CA) were injected and then flushed with saline. Tissues were dissolved in 4 mol/l KOH and filtered with polyester membrane filters (Triton Technology). The fluorescent dye was extracted with acidified cellosolve acetate, and fluorescence was measured with a plate reader and normalized by the weight of each tissue.

Tissue lactate concentrations were determined with an assay kit using the instructions provided by the manufacturer (Roche, Mannheim, Germany) and normalized by protein concentration. Protein was quantified by the bicinchoninic acid method using the BCA Protein Assay Reagent obtained from Pierce (Rockford, IL) and BSA as the standard.

3T3-L1 preadipocytes were grown to confluence and induced to differentiate into adipocytes, as described previously (25). The cells were then cultured for 12 h under hypoxia (1% O₂) or normoxia (21% O₂).

Total RNA from 3T3-L1 cells and tissues were prepared with an RNeasy Mini Kit (Qiagen, Valencia, CA). First-strand cDNA was synthesized from total RNA using the Transcriptor First Strand cDNA Synthesis Kit (Roche). Real-time PCR was performed on the ABI7900 using the SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA) according to the protocol provided by the manufacturer. Primer sets were the following: adiponectin, 5′-GTT CTA CTG CAA CAT TCC GG-3′ and 5′-TAC ACC TGG AGC CAG ACT TG-3′; PAI-1, 5′-TCA GCC CTT GCT TGC CTC AT-3′ and 5′-GCA TAG CCA GCA CCG AGG A-3′; peroxisome proliferator–activated receptor (PPAR)-γ, 5′-CCA GAG TCT GCT GAT CTG CG-3′ and 5′-GCC ACC TCT TTG CTC TGC TC-3′; GLUT1, 5′-CCA TCC ACC ACA CTC ACC AC-3′ and 5′-GCC CAG GAT CAG CAT CTC AA-3′; C/EBP homologous protein (CHOP), 5′-GTC CTG TCC TCA GAT GAA ATT GG-3′ and 5′-GCA GGG TCA AGA GTA GTG AAG GTT-3′; glucose-regulated protein, 78 kD (GRP78), 5′-ACC TAT TCC TGC GTC GGT GT-3′ and 5′-GCA TCG AAG ACC GTG TTC TC-3′; leptin, 5′-GAT GGA CCA GAC TCT GGC AG-3′ and 5′-AGA GTG AGG CTT CCA GGA CG-3′; matrix metalloprotease 2 (MMP2), 5′-GCA GGG AAT GAG TAC TGG GTC TAT-3′ and 5′-CAG TTA AAG GCA GCA TCT ACT TG-3′; adrenomedullin, 5′-AAG TGG AAT AAG TGG GCG CTA A-3′ and 5′-ACT GTC GTC TCA TCA GCG AGT C-3′; and 36B4, 5′-GCT CCA AGC AGA TGC AGC A-3′ and 5′- CCG GAT GTG AGG CAG CAG-3′. The levels of mRNA were normalized relative to the amount of 36B4 mRNA.

Total RNA from 3T3-L1 cells cultured in hypoxia (1% O₂) was reverse transcribed and amplified using the sense primer (5′-AAACAGAGTAGCAGCGCAGACTGC-3′) and the antisense primer (5′-GGATCTCTAAAACTAGAGGCTTGGTG-3′). This fragment was further digested by PstI as described previously (26). The cDNA fragments were resolved on a 2% agarose gel.

The luciferase reporter plasmids of human adiponectin promoter were generated by excising the promoter fragment from the genomic clone of adiponectin (generous gift from Dr. Junji Takeda) and inserting it into the KpnI and SacI sites of the pGL3 basic luciferase expression vector (Promega, Madison, WI). Expression plasmids encoding β-galactosidase (pCMX–β-gal) were generous gifts from Dr. David Mangelsdorf (University of Texas Southwestern Medical Center, Dallas, TX). Mouse CHOP (accession number BC013718) was cloned and inserted into pCMV to generate the expression plasmid for mouse CHOP (pCMV-CHOP).

On day 5 after induction of differentiation, the media of 3T3-L1 cells in six-well plates were changed to OPTI-MEM (Invitrogen, San Diego, CA), and the cells were transfected with plasmids using LipofectAMINE 2000 reagent (Invitrogen) according to the instructions provided by the manufacturer. At 48 h after plasmid transfection, luciferase reporter assays were performed using Luciferase Assay System (Promega). Luciferase values were normalized by an internal β-galactosidase control and expressed as the relative luciferase activity.

Western blot analysis was carried out by using anti-eIF2α (eukaryotic translation initiation factor 2α) antibodies (Santa Cruz Biotechnology, Santa Cruz, CA).

Two pairs of small interfering RNAs (siRNAs) were synthesized chemically by Qiagen, annealed, and transfected into 3T3-L1 adipocytes using Lipofectamine 2000 (Invitrogen) as described previously (27). Twenty-four hours after transfection, cells were cultured for 12 h under hypoxia (1% O₂) and total RNA extracted as described above.

All data are presented as means ± SEM and analyzed using unpaired Student's t test. For simultaneous multiple comparisons, differences between groups were analyzed by one-way ANOVA followed by Dunnet's multiple comparisons test. P values <0.05 were considered statistically significant. All study protocols described in this report were approved by the Ethics Review Committee for Animal Experimentation of Osaka University School of Medicine or Astellas Pharma.

Body weight of obese mice was significantly higher than that of the control group (male normal diet–fed C57BL/6J mice 29.07 ± 0.42 g (n = 11) and male high-fat diet–fed mice 36.32 ± 0.33 g (n = 9), P < 0.001; female C57BL/6J mice 21.12 ± 0.27 g (n = 9) and female KKAy mice 43.96 ± 0.4 g (n = 9), P < 0.001). Tissue hypoxia was evaluated by immunohistochemistry of pimonidazole, which is known to form adducts with thiol groups under hypoxia and has been used to quantify tissue oxygen concentration (28). To identify and quantitate oxygen concentrations of tissues, sections of liver and white adipose tissue (WAT) of control and obese mice were stained with pimonidazole and hematoxylin. Control experiments showed that pimonidazole binding in L6 cells was elevated under hypoxic conditions (supplementary Fig. 1A and B [available in an online appendix at http://dx.doi.org/10.2337/db06-0911]). Moreover, in accordance with a previous report (29), pimonidazole staining predominated in pericentral regions of liver lobule in all mice, where oxygen supply is naturally low (data not shown). Fig. 1 shows the patterns of pimonidazole binding (brown) against a hematoxylin counter stain (blue) in epididymal or parametrial WAT of control and obese mice. Weak pimonidazole signals were identified in WAT of normal diet–fed mice and control C57BL/6J mice. On the other hand, pimonidazole staining (cytoplasm as well as cytoplasmic membrane) was markedly increased in WAT of high-fat diet–fed mice and KKAy mice compared with that of each group of control mice, indicating low oxygen concentrations in fat depots in obese mice.

We also quantified tissue lactate concentrations as another indicator of tissue hypoxia (Fig. 2). The lactate concentrations in WAT of high-fat diet–fed mice and KKAy mice were 1.7- and 1.5-fold higher than those of normal diet–fed mice and control C57BL/6J mice, respectively. In contrast, the lactate concentration in muscle of obese mice was similar to that of control mice (Fig. 2).

To exclude possible accumulation of pimonidazole and lactate by systemic hypoxia in obese mice, we analyzed arterial blood gases. As shown in Table 1, systemic arterial PaCO₂ and PaO₂ as well as hemoglobin concentration and oxygenation (arterial O₂ saturation: SaO₂) were not different between control and obese mice.

We next examined the possible role of tissue perfusion using colored microsphere analysis (Fig. 3). The numbers of microspheres were significantly decreased in WAT of both types of obese mice (i.e., high-fat diet–fed and KKAy mice) compared with each group of controls (Fig. 3A and B), indicating that the WAT of obese mice had less perfusion than that of control mice. There were no significant differences in the number of microspheres in the muscle, lung, heart, or kidney between the control and obese mice (Fig. 3). Similar results were obtained after correction for protein amounts (supplementary Fig. 2). These results indicate that tissue hypoxia in WAT of obese mice is, at least in part, due to decreased perfusion in WAT.

To assess the hypoxic condition in WAT, mRNA levels of hypoxia-inducible genes such as leptin, PAI-1, MMP2, and adrenomedullin were measured. Expression levels of these genes were significantly elevated in WAT of high-fat diet–fed and KKAy mice compared with each group of control mice (Fig. 4). We measured mRNA levels of these hypoxia-inducible genes in liver and muscle of high-fat diet–fed and control mice. In liver, mRNA expression levels of PAI-1, MMP2, and adrenomedullin were not elevated in high-fat diet–fed compared with the control mice. In muscle, MMP2 and adrenomedullin mRNA levels were not altered, while only PAI-1 mRNA level was increased (supplementary Fig. 3).

Adiponectin mRNA levels as well as PPARγ mRNA level were significantly reduced in WAT of high-fat diet–fed and KKAy mice compared with those of normal diet–fed and control C57BL/6J mice. The mRNA levels of ER stress marker genes, such as GRP78 and CHOP, were significantly elevated in WAT of high-fat diet–fed and KKAy mice compared with those of normal diet–fed and control C57BL/6J mice (Fig. 4) as previously reported (23,24).

To examine the effects of hypoxia on adipocytokine mRNA expression, 3T3-L1 adipocytes were cultured in 1% O₂ for 12 h. mRNA levels of PPARγ and adiponectin were decreased, whereas mRNA level of PAI-1 was increased in hypoxic cells compared with that in normoxic control cells (Fig. 5A). We further determined the effect of hypoxia on expression of adipogenic marker genes. Exposure of differentiated 3T3-L1 adipocytes to hypoxic condition did not affect stearoyl-CoA desaturase 1 (SCD1) and preadipocyte factor-1 (pref1) mRNA expression levels (Fig. 5A). These results suggested that hypoxia did not alter the adipocyte differentiation stage itself. To elucidate whether hypoxia inactivates the adiponectin promoter, luciferase activity driven by −3.6 kb of the human adiponectin promoter was tested in 1 and 21% O₂. As shown in Fig. 5B, luciferase activity under hypoxia was significantly lower than under normoxia.

To determine whether hypoxia induces ER stress in 3T3-L1 adipocytes, we investigated the expression patterns of several molecular indicators of ER stress in these cells. The mRNA expression levels of CHOP and GRP78 in the 3T3-L1 adipocytes, incubated under hypoxia, were significantly higher than those under normoxia (Fig. 6A). PERK is an ER-resident type I transmembrane Ser/Thr protein kinase and phosphorylates eIF2α in response to ER stress. Immunoblot analysis demonstrated marked accumulation of phosphorylated eIF2α protein in cell lysates after 2–6 h hypoxia compared with the normoxic cells (Fig. 6B). ER stress also activates the IRE1-dependent pathway (26). Activated IRE1 cuts out 26 nucleotides of the unspliced X-box binding protein-1 (XBP1) mRNA to generate the spliced XBP1 mRNA. The RT-PCR products produced from the spliced and unspliced XBP1 mRNAs can be easily detected following PstI digestion of PCR products generated using primers flanking the XBP1 intron, since the unspliced product contains a PstI site that is lost in the spliced product. As shown in Fig. 6C, the PstI-undigested fragment increased after 6 h hypoxia and was still noted at 24 h. Taken together, these results indicate that hypoxia induces ER stress in 3T3-L1 adipocytes.

Treatment with tunicamycin, an ER stress inducer, decreased the mRNA expression levels of PPARγ and adiponectin in 3T3-L1 adipocytes with the induction of CHOP mRNA (Fig. 7A), suggesting that ER stress results in downregulation of PPARγ and adiponectin mRNA.

The CHOP protein is known to heterodimerize with other C/EBP proteins to transcriptionally inactive complexes (30). Previous reports showed that C/EBP is critical for the regulation of adiponectin expression (31). To confirm the negative regulatory effect of CHOP on the transcriptional activity of adiponectin, we measured luciferase activity driven by −3.6 kb adiponectin promoter in 3T3-L1 adipocytes. Simultaneous transfection of the CHOP vector significantly inhibited the transcriptional activity of adiponectin promoter in a dose-dependent manner (Fig. 7B).

To further confirm the role played by CHOP in adiponectin expression, we knocked down CHOP expression using a specific siRNA. Transfection of CHOP siRNA in 3T3-L1 adipocytes reduced the endogenous CHOP mRNA and protein levels (Supplementary Fig. 4A and B) and reduced mRNA levels of tribble 3 (TRB3), a target gene of CHOP (supplementary Fig. 4C). CHOP siRNA partially reversed the downregulation of adiponectin mRNA in 3T3-L1 adipocytes incubated under hypoxia (Fig. 7C). Under the same condition, interference with CHOP had no effect on downregulation of PPARγ by hypoxia (Fig. 7C). These findings suggest that hypoxia-induced reduction of adiponectin mRNA is mediated in part through induction of CHOP.

Because it is well established that HIF1α is a major regulator of cellular-adaptive responses to hypoxia (32), we next determined the role of HIF1α in hypoxia-induced downregulation of adiponectin mRNA. Transfection of 3T3-L1 adipocyte cells with HIF1α siRNA significantly reduced endogenous HIF1α mRNA and protein levels (supplementary Fig. 4A and B). The same treatment partially inhibited hypoxia-induced upregulation of GLUT1 mRNA, a known target of HIF1α (Fig. 7D). However, interference with HIF1α did not alter the gene expression levels of adiponectin under hypoxia (Fig. 7D) and reduced mRNA levels of PPARγ under normoxia and hypoxia (Fig. 7D).

Finally, we examined the effect of hypoxia on the stability of adiponectin mRNA. The remaining amounts of adiponectin mRNA under normoxia and hypoxia were determined at various time points after addition of the transcriptional inhibitor actinomycin D. The mRNA degradation of adiponectin was accelerated under hypoxia compared with normoxia (Fig. 8A). To examine the effect of ER stress on adiponectin mRNA stability, 3T3-L1 adipocytes cells were incubated for 12 h with medium alone or medium supplemented with tunicamycin under normoxia, and the rate of adiponectin mRNA decay was determined as above. In contrast to the effect of hypoxia on adiponectin mRNA decay, tunicamycin did not alter adiponectin mRNA stability, suggesting that hypoxia-induced degradation of adiponectin mRNA is not dependent on increased ER stress (Fig. 8B).

There is increasing evidence that obesity impairs adipocyte function and secretion of adipocytokines and that dysregulated release of adipocytokines contributes to insulin resistance and metabolic syndrome. One possible mechanism for the dysregulation is inflammatory reaction (33). In addition, recent studies revealed that obesity induces ER stress and that the latter in turn activates inflammatory response, thus contributing to insulin resistance in the liver and adipose tissue (23,24). However, it is still not clear how inflammation and ER stress begin in obesity. To answer this key question in the present study, we focused on oxygen availability in WAT of obese mice and investigated the effect of oxygen tension on the regulation of adipocytokines in adipocytes.

Although previous studies demonstrated changes in WAT blood flow and vasculature network in obesity (34–38), there is no evidence that the changes actually affect intracellular ambient oxygen tension. In the current study, we demonstrated that WAT of obese mice is under hypoxia as confirmed using pimonidazole hydrochloride adduction (physical evidence) as well as lactate concentration (physiological evidence). In this study, tissue hypoxia was detected in WAT of obese mice using pimonidazole staining, a 2-nitroimidazole, which is reductively activated at low oxygen concentrations (39–41) and adducts to cellular proteins. Hypoxia induces energy metabolism alteration and increases lactate concentration in tissue (42,43). We demonstrated that lactate concentration was markedly increased specifically in WAT of obese mice, providing physiological evidence of tissue hypoxia. Assessment of mRNA levels of hypoxia-inducible genes such as leptin, PAI-1, MMP2, and adrenomedullin was consistent with the above data, showing that WAT of obese mice was hypoxic and liver was not. In muscle, MMP2 and adrenomedullin mRNA levels were not altered, while only PAI-1 mRNA level was increased in high-fat diet–fed mice compared with normal diet–fed mice (supplementary Fig. 3). PAI-1 may be regulated by other mechanisms in muscle.

There are two possible explanations for the tissue hypoxia in the WAT of obese mice: 1) reduced oxygen pressure and oxygen content of the blood and 2) tissue hypoperfusion. The first possibility was ruled out by measurement of arterial blood gases, showing no difference in arterial PaO₂, hemoglobin concentration, or oxygenation. Moreover, pimonidazole staining and lactate concentration showed no significant differences in tissues other than WAT. Thus, local tissue hypoxia in WAT in obese mice is not due to reduced systemic oxygen supply.

The second possibility, i.e., tissue hypoperfusion in WAT of obese mice, was confirmed using colored microspheres (Fig. 3). West et al. (36) demonstrated that blood flow to adipose tissue, measured with radiolabeled microsphere, was reduced in an adipose-specific manner in Zucker obese rats. In humans, the levels of adipose tissue blood flow were measured with positron emission tomography using [¹⁵O]-labeled water (44) and the ¹³³Xe washout method (45) and were lower in obese compared with nonobese subjects. These reports are consistent with the current data, suggesting that a decrease of adipose tissue perfusion is a common feature in obesity. We measured the numbers of endothelial cell nuclei and adipocytes per sections of WAT in control, high-fat diet–fed, and KKAy mice. The ratio of endothelial cell number per adipocyte increased significantly in obese compared with control mice (data not shown). Considering these results, blood flow per vessels may decrease severely in WAT of obese mice.

Considering that oxygen diffusion is limited at most to 100 μm, hypertrophied adipocytes up to 140–180 μm in diameter are assumed to be in a relatively hypoxic state. However, in our study, pimonidazole staining was detected in small as well as large adipocytes of obese animals. These results suggest that reduced perfusion capacity rather than cell size seems the main culprit for tissue hypoxia.

We evaluated the effect of hypoxia on adipocytokine expression. Adiponectin and PPARγ mRNA expression levels were reduced while PAI-1 level was increased in hypoxic 3T3L1 adipocytes. In addition, hypoxia decreased adiponectin promoter activity, suggesting a suppressive effect of hypoxia on adiponectin transcription.

It has been reported that severe hypoxia (46,47) or anoxia (48) induces ER stress, and UPR plays a role in adaptation to hypoxic stress in an HIF1α-independent manner in other organs and cells (49). Our results establish the role of ER stress in altering adiponectin mRNA level in hypoxic adipocytes. This conclusion was based on the following findings: 1) hypoxia induced ER stress in adipocytes, concurrent with previous reports (20); 2) tunicamycin, an ER stress inducer, suppressed adiponectin mRNA expression; 3) CHOP dose-dependently inhibited adiponectin promoter activity; and 4) hypoxia-induced downregulation of adiponectin mRNA was inhibited partially by CHOP siRNA. CHOP is a member of the C/EBP family of bZIP transcription factors, and its expression is induced to high levels by ER stress (30) and mitochondrial reactive oxygen species (ROS) (47). Adiponectin transcription is regulated by C/EBP (31), thus CHOP should inhibit adiponectin transcription as a suppressor of C/EBP as the CHOP-C/EBP heterodimer cannot bind to the C/EBP binding site. Cycloheximide, a potent inhibitor of protein synthesis, blocked hypoxia-induced suppression of adiponectin mRNA (supplementary Fig. 5), indicating that de novo production of protein mediators was required. Production of CHOP via ER stress might be partly responsible for downregulation of adiponectin mRNA. These results suggest that tissue hypoxia in the WAT of obesity leads to ER stress and suppression of adiponectin transcription.

Posttranscriptional regulation of mRNA stability is another control of gene expression by hypoxia, as shown in several hypoxia-regulated genes including vascular endothelial growth factor, tyrosine hydroxylase, GLUT1, erythropoietin, and endothelial NO synthase (50,51). In our study, hypoxia increased adiponectin mRNA degradation in adipocytes, and this effect was independent of ER stress. Adiponectin mRNA, however, does not contain classical AU-rich motifs for mRNA stability (52), suggesting unknown sequence motifs involved in its stability. These data indicate that adiponectin mRNA transcription was regulated via ER stress and that mRNA stability was regulated posttranscriptionally via hypoxia.

Other mechanisms might also be involved in the hypoxia-induced downregulation of adiponectin mRNA such as HIF1α, ROS, redox, and inflammation. Hypoxia-induced downregulation of adiponectin mRNA was not affected by HIF1α siRNA, hence hypoxia-induced suppression of adiponectin expression should not be dependent on HIF1α. We also tested the effects of several antioxidants such as N-acetyl-cysteine, butylated hydroxyanisole, apocynin, catalase, and agents that alter the cellular redox status by affecting NAD(P)H-to-NAD(P) ratio, i.e., rotenone (an inhibitor of complex I [NADH-dehydrogenase] of the respiratory chain) and lactate (a product of the anaerobic energy metabolism, which induces an increase in NAD(P)H/NAD(P) ratio while oxidized to pyruvate), and protein kinase inhibitors such as SB203580 (a p38MAPK inhibitor), LY294002 (a PI3K inhibitor), PD98059 (a MEK inhibitor), and SP600125 (a JNK inhibitor). None of them attenuated the hypoxia-induced downregulation of adiponectin mRNA expression. This evidence may rule out roles of ROS, redox, and inflammation in reduced adiponectin production in hypoxic adipocytes.

In summary, we demonstrated that WAT of obese mice is hypoxic and that hypoxia dysregulates the production of adipocytokines in adipocytes. This effect is mediated by ER stress and posttranscriptional regulation. Collectively, our results suggest that local adipose tissue hypoxia is partly responsible for dysregulated adipocytokines production and metabolic syndrome in obesity. Hypoxia, per se, as well as signals mediated by hypoxia, could become a therapeutic target for the treatment of obesity-linked metabolic syndrome.

This work was supported by grants from the Ministry of Education, Culture, Sports, Science and Technology.