Androgens Decrease Plasma Adiponectin, an Insulin-Sensitizing Adipocyte-Derived Protein

We studied 579 adult Japanese: 442 men (mean [±SD] age 52.6 ± 11.9 years; mean [±SD] BMI 23.3 ± 2.6 kg/m²) and 137 women (age 53.2 ± 12.0; BMI 22.8 ± 2.9). Because plasma adiponectin concentrations were reported to be low in obese subjects (19), the men and women enrolled in the present study were matched for age and BMI. These participants had had annual health checks. A signed informed consent was obtained from all subjects, and study protocols were approved by the Japanese Visceral Fat Syndrome study committee of the Ministry of Health and Welfare of Japan or the Medical Ethics Committee of Osaka University.

Blood was drawn after an overnight fast. Plasma adiponectin concentrations were measured by sandwich enzyme-linked immunosorbent assay (ELISA) as described previously (19). Plasma glucose was assayed by the glucose oxidase method, and insulin was assayed by immunoradiometric assay using Insulin Riabead (Dainabot, Osaka, Japan). Plasma concentrations of TNF-α were measured by ELISA using Quantikine HS (R&D Systems, Minneapolis, MN), and plasma leptin levels were measured by ELISA as described previously (27).

ICR mice and C57BL/6J mice were obtained from Clea Japan (Tokyo, Japan) and kept under a 12-h dark-light cycle (lights on 8:00 a.m. to 8:00 p.m.) and constant temperature (22°C) with free access to food and water. The composition of the diet was as follows: carbohydrate 54.0%, protein 23.8%, and fat 5.1%. Experiments of castration and ovariectomy were performed as follows: 20-week-old ICR male mice were castrated (CAST, n = 4) or sham-operated (SHAM, n = 4) and 20-week-old ICR female mice were ovariectomized (OVAR, n = 3) or sham-operated (SHAM, n = 4) under pentobarbital sodium anesthesia (50 mg/kg body wt). For castration, after midline abdominal incision, the testes were dissected out with epididymal fat pads, and testicular ducts and blood vessels supplying the testes were isolated and ligated. The testes were then excised through an incision, the epididymal fat pads were put back, and the incision was closed with suture. For the sham castration, after abdominal incision, the testes with epididymal fat pad were exposed and put back, and the incision was closed with suture. For ovariectomy, after abdominal incision, the uterus, oviducts, and ovaries were exposed with the fat tissues around them, the oviduct and blood vessels supplying the ovaries were tied up, and the ovaries were removed through an incision. The uterus and oviducts were put back with the fat tissues around them, and the incision was closed with suture. For the sham operation to ovariectomy, after abdominal incision, the uterus, oviducts, and ovaries were exposed with the fat tissues and were put back, followed by closure with suture. At 6 weeks after operation, blood samples were collected after a 6-h fast. Glucose and insulin were measured by Glucose CII-Test Wako kit (Wako, Osaka, Japan) and a double-antibody enzyme immunoassay using a Glazyme Insulin EIA Kit (Sanyo Chemical Industries, Kyoto, Japan) with rat insulin as a standard, respectively. The homeostasis model assessment insulin resistance (HOMA-IR) values were calculated from the fasting concentrations of insulin and glucose using the following formula: fasting plasma insulin (μU/ml) × fasting plasma glucose (mg/dl)/405.

The experiments of castration and testosterone supplement were performed as described previously with minor modifications (28). Briefly, 11-week-old C57BL/6J male mice were castrated (CAST, n = 10) or sham-operated (SHAM, n = 9) under pentobarbital sodium anesthesia as mentioned above. At 6 weeks after operation, randomly selected castrated and sham-operated mice (n = 5 and n = 4, respectively) received one-shot subcutaneous injection of testosterone propionate (Wako) (0.5 mg/100 g body wt) every other day for 2 weeks, whereas other groups of castrated and sham-operated mice received the vehicle (polyethylene glycol) only. After a 6-h fast, mice were killed under anesthesia, and various tissues and blood were collected. This experimental protocol was approved by the Ethics Review Committee for Animal Experimentation of Osaka University.

Insulin sensitivity in mice was estimated as described previously, with minor modifications (29). Briefly, after a 6-h fast, ICR mice received a subcutaneous injection of saline containing epinephrine (1 μg/10 g body wt; Sigma) and propranolol (50 μg/10 g body wt; Sigma) to block the endogenous secretion of insulin. Subsequently, glucose (1 g/10 g body wt) and insulin (0.5 units/kg body wt) were injected intraperitoneally. Blood samples were obtained at 0 and 1 h after injections, and glucose concentrations were measured. Insulin sensitivity was estimated by percentage changes of plasma glucose concentrations.

3T3-L1 cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% FCS and differentiated with DMEM supplemented with 5 μg/ml insulin, 0.5 mmol/l 1-methyl-3-isobutyl-xanthin, and 1 μmol/l dexamethasone, 2 days after reaching confluence.

3T3-L1 cells on day 7 (7 days after differentiation-induction) were used for the experiments to examine the effects of hormones on the expression and secretion of adiponectin. Cells were washed twice with PBS and incubated with DMEM containing 1.5% BSA and indicated hormones or vehicle for 12 h. The medium was washed and replaced with a freshly prepared medium containing the same concentrations of hormones. After a 12-h incubation, the medium was collected for Western blotting to measure the secretion of adiponectin, and the cells were harvested for RNA and protein analyses.

Total RNA was isolated from mouse subcutaneous adipose tissues and 3T3-L1 cells using RNA STAT-60 (TEL-TEST) according to the manufacturer’s instructions. Isolated total RNA (10 μg) was separated by electrophoresis and transferred to nylon membranes. The membrane was hybridized with adiponectin or cyclophilin DNA probe labeled with [α-³²P]dCTP. The hybridized membrane was exposed to X-ray film, X-OMAT AR (Kodak). The band intensities were quantified by densitometry.

Plasma adiponectin concentration of mice, adiponectin secreted from 3T3-L1 cells into the media, and protein amounts in subcutaneous adipose tissues and in 3T3-L1 cells were analyzed by quantitative Western blotting. An equal aliquot of plasma (1 μl × 1/250) or medium sample (7 μl) was lysed in sample buffer (30) and subjected to SDS-PAGE. The subcutaneous adipose tissues of mice were homogenized and sonicated, or 3T3-L1 cells were sonicated in lysis buffer containing 10 mmol/l HEPES (pH 7.5), 142.5 mmol/l KCl, 5 mmol/l MgCl₂, 1 mmol/l EGTA, 0.2% NP-40, 1 mmol/l Na₃VO₄, 10 mmol/l NaF, 1 mmol/l Na₂P₂O₇, 1 mmol/l phenylmethylsulfonyl fluoride (PMSF), and 10 μg/ml aprotinin, followed by centrifugation. Equal amounts (2 μg) of the protein were subjected to SDS-PAGE. Western blotting was performed as described previously (31). The band intensities were quantified by densitometry.

3T3-L1 adipocytes (day 7) were plated onto six-well plates and were incubated with FCS-free DMEM for 12 h. For metabolic labeling, the cells were washed with PBS and incubated with methionine- and cysteine-free DMEM without serum for 30 min to deplete the intracellular pools. The depletion medium was removed, and the cells were incubated in 1 ml of methionine- and cysteine-free DMEM containing 100 μCi/ml of l-[³⁵S]methionine and l-[³⁵S]cysteine (Pro-mix l-[³⁵S] in vitro labeling mix; Amersham) for the specific times indicated (1, 2, and 4 h).

For the immunoprecipitation of radiolabeled adiponectin in medium and cell lysates, the metabolic labeling was performed for 2 h. Then, the labeling medium was replaced with 1 ml of complete DMEM (FCS-free) with testosterone (300 nmol/l) or vehicle (ethanol). The medium and cell lysates were collected at indicated times for immunoprecipitation assays.

For adiponectin immunoprecipitation, we used a procedure described previously (32), with modifications. Briefly, to prepare cell lysates, we washed the cells with PBS, resuspended them in 300 μl of disruption buffer (10 mmol/l Tris-HCl [pH 7.5], 150 mmol/l NaCl, 5 mmol/l EDTA, 10 mmol/l benzamidine, 1 mmol/l PMSF, 1% Nonidet P-40 [NP-40]), and lysed them with three repetitive freeze-thaw cycles. Cellular debris was removed by centrifugation. The cell lysates were adjusted to equal protein concentration with disruption buffer and subjected to immunoprecipitation. A total of 500 μl of cell lysates (100 μg of total protein) was mixed with an equal volume (500 μl) of disruption buffer lacking EDTA and NP-40 and immunoprecipitated with 5 μl of rabbit polyclonal antibody against murine adiponectin overnight at 4°C, followed by incubation with 40 μl of protein G beads (Amersham) for 2 h at 4°C. For medium, aliquots (450 μl) of the metabolically labeled culture medium were added by 1/10 volume of 10× immunoprecipitation buffer (IPB; 1× IPB = 10 mmol/l Tris-HCl [pH 7.5], 150 mmol/l NaCl, 1 mmol/l EDTA, 1% NP-40, 10 mmol/l benzamidine, 1 mmol/l PMSF) and immunoprecipitated as described above. Immunoprecipitates were washed three times with 1× IPB, followed by solubilization with sample buffer, and subjected to SDS-PAGE. After electrophoresis, radiolabeled proteins were autoradiographed. The band intensities were quantified by densitometry.

Data are expressed as mean ± SE. Statistical analyses, except for data shown in Fig. 7C, were performed with unpaired t tests. Statistical analysis of the data shown in Fig. 7C was performed with Pearson’s correlation coefficient test. P < 0.05 denoted the presence of a statistically significant difference.

First, to investigate the relationship to sex in adiponectin, we compared plasma concentrations of adiponectin between men and women. Figure 1 and Table 1 show the profiles of subjects from the Japanese Visceral Fat Syndrome study. There were no significant differences in plasma glucose and insulin concentrations between men and women. Plasma levels of TNF-α, which was demonstrated to reduce the expression and production of adiponectin (31), showed no sex difference either (Table 1). Figure 1A demonstrates that plasma adiponectin was 35% lower in men than in women. Furthermore, leptin plasma concentrations in men were lower than in women (Fig. 1A), as reported previously (33). Plasma adiponectin concentrations in premenopausal women were not significantly different from those in postmenopausal women (Fig. 1B). In contrast, plasma leptin concentrations in postmenopausal women were lower than in premenopausal women (Fig. 1B), as the increasing effect of estrogen on the leptin production was described previously (34). These results suggested that androgens might reduce plasma adiponectin concentrations in men.

For validating the above hypothesis, male mice were castrated, treated with or without testosterone, and adiponectin production was analyzed in these animals (Figs. 2 and 3). Figure 2A provides a schematic illustration of the experiment. From 6 weeks after the castration or sham operation, mice were treated with subcutaneous injections of testosterone or vehicle solution, every other day, for a total of eight injections. On the next day of the final injection, mice were killed. Figure 2B shows the weights of the body and various organs. The operation procedures and injections did not influence body weight or weight of subcutaneous white fat. Previous studies demonstrated that the weight of kidney was decreased after castration as a result of a fall in testosterone level and that such changes were reversed after testosterone supplement (35,36). In the present study, we observed a significant decrease in kidney and prostate weights by castration (28 and 77% of sham-operated mice, respectively) and found that testosterone treatment restored the weights of these organs to near-normal levels. These results validated the operation procedures and testosterone treatment in this study.

Figure 3A shows plasma adiponectin concentrations measured by Western blotting in these mice. Castration significantly increased the concentrations of plasma adiponectin (170% of sham-operated mice). Furthermore, testosterone injections reduced plasma adiponectin in castrated mice (46% of vehicle-treated castrated mice) and also in sham-operated mice.

We previously showed that adiponectin mRNA was expressed only in adipose tissues in human, monkey, and mouse (9,16–18). Figure 3B and C compare the protein and mRNA amounts of adiponectin in subcutaneous adipose tissues of these mice. Castration or testosterone treatment had no effect on adiponectin protein and mRNA levels. Considered together, the above data indicated that testosterone reduced plasma adiponectin concentrations without changing the mRNA and protein levels of adiponectin in the adipose tissue.

We further examined the effect of androgens on adiponectin production in in vitro 3T3-L1 adipocytes (Fig. 4). The amount of adiponectin secreted in media was markedly decreased by both testosterone and 5α-dihydrotestosterone (5α-DHT), a metabolite of testosterone that is not converted to estrogen (Fig. 4A). There were no apparent changes in adiponectin protein and mRNA levels in adipocytes treated with testosterone or 5α-DHT (Fig. 4B and C). These results obtained from the 3T3-L1 adipocytes were in agreement with the findings of in vivo studies demonstrating that testosterone reduced the production of adiponectin in plasma without changing the mRNA and protein levels of adiponectin in adipose cells.

To examine further the inhibitory effect of testosterone on adiponectin secretion from 3T3-L1 adipocytes, we performed the pulse-chase study. First, we established the procedure for adiponectin immunoprecipitation. 3T3-L1 adipocytes (day 7) were metabolically radiolabeled for 1, 2, or 4 h, and the cell lysates were collected and were immunoprecipitated with antiadiponectin antibody (Fig. 5A). The labeled adiponectin was increased in a time-dependent manner and not observed with nonimmune rabbit serum instead of adiponectin antibody for immunoprecipitation. Next, after a 2-h radiolabeling, the secretion into culture medium (Fig. 5B) and the amount in the cell lysates (Fig. 5C) of the newly synthesized radiolabeled adiponectin were chased in the presence and absence of testosterone. The secretion of labeled adiponectin was continuously inhibited with testosterone treatment from 1 to 12 h of chasing period (Fig. 5B). The testosterone-mediated inhibition was more significant in the acute phases such as 1 h (0.3-fold) and 2 h (0.5-fold). However, the labeled adiponectin was more retained in cells from 2 to 12 h with testosterone treatment. These results indicated that testosterone inhibited the secretion of adiponectin pool in 3T3-L1 adipocytes.

Next, we investigated the effects of sex hormones on plasma adiponectin concentration and insulin sensitivity in mice. Male ICR mice were castrated and female mice were ovariectomized, and plasma adiponectin concentrations were measured by Western blotting (Fig. 6). In sham-operated mice, plasma adiponectin concentrations were lower in male mice than in female mice, as was observed in human subjects (Fig. 1A). In male mice, castration resulted in increased plasma adiponectin levels, as observed in Fig. 3A. However, in female mice, there was no significant difference in adiponectin plasma concentrations between ovariectomized and sham-operated mice similarly to the female subjects before and after menopause, which indicated no effect of estrogen on the adiponectin production in plasma.

Fasting plasma concentrations of glucose and insulin were measured in these mice, and HOMA-IR was calculated. In male mice, castration reduced plasma glucose, insulin, and HOMA-IR to levels close to those of sham-operated female mice. These results indicated that androgens decreased insulin sensitivity, as described previously (37–39). The mirror image of adiponectin levels and HOMA-IR in Fig. 6 support the previously described concept in human, monkey, and rodents that adiponectin is an insulin-sensitizing hormone (9–11,21,22).

Using the mice described in Fig. 6, we also performed an insulin sensitivity test based on the steady-state plasma glucose concentration method, as described previously (Fig. 7). In this test, endogenous production of insulin is blocked by administration of epinephrine and propranolol, and glucose and insulin are injected to compare the insulin-mediated decrease in plasma glucose (Fig. 7A). In Fig. 7B, the ability of insulin to suppress plasma glucose concentration was much less in sham-operated male mice than in sham-operated female mice. In male mice, castration augmented insulin sensitivity to reduce plasma glucose concentration to levels close to those of sham-operated female mice. Ovariectomy did not affect insulin sensitivity in female mice. Figure 7C shows a plot of plasma adiponectin concentrations and the percentage change of the decrease in plasma glucose in the insulin sensitivity test, using data of all mice examined in Fig. 7B. There was a close correlation between plasma adiponectin concentrations and insulin sensitivity. Plasma leptin concentrations were similar in sham-operated and castrated male mice and, hence, showed no correlation to the increase of insulin sensitivity (data not shown). Taken together, these results suggest that testosterone-mediated suppression of adiponectin closely associates with less insulin sensitivity in male mice.

In the present study, we demonstrated that plasma concentrations of adiponectin, an adipocyte-derived plasma protein, were lower in men than in women but were not different between pre- and postmenopausal women. These data suggested that androgens act to reduce plasma adiponectin concentrations. In animal experiments, castration increased and testosterone supplement reduced plasma adiponectin concentrations in male mice. In cultured 3T3-L1 adipocytes, testosterone and 5α-DHT suppressed the secretion of adiponectin into the culture media. These results indicated that androgen decreased plasma adiponectin concentrations through its effect on adipocytes.

We reported previously the presence of low concentrations of plasma adiponectin in human subjects of obesity and type 2 diabetes (10,11,19). In a longitudinal study of rhesus monkeys that had free access to food and became obese and subsequently developed type 2 diabetes, the reduction of plasma adiponectin preceded the development of hyperinsulinemia and hyperglycemia (9). We also demonstrated, in monkey and human, that plasma adiponectin levels correlated significantly with insulin sensitivity in the whole body using the hyperinsulinemic-euglycemic glucose clamp technique (9,11). Recent studies demonstrated that adiponectin increased fatty acid oxidation (40), augmented insulin action in the muscle and liver, and further improved insulin resistance in lipoatrophic and genetically obese mice, both of which had hypoadiponectinemia (21,22). These observations indicate that adiponectin is an insulin-sensitizing hormone and that hypoadiponectinemia induces insulin resistance.

In addition, we showed that plasma adiponectin levels were associated with atherosclerosis: plasma adiponectin concentrations were low in patients with coronary artery disease (23). Adiponectin levels in plasma correlated inversely with the incidence of ischemic heart diseases in subjects with chronic renal failure (24). These results suggested that hypoadiponectinemia is also involved in the pathophysiology of atherosclerosis. In tissue-cultured cell experiments, adiponectin suppressed TNF-α-induced mRNA expression of several adhesion molecules, including vascular cell adhesion molecule-1, intracellular adhesion molecule-1, and E-selectin in vascular endothelial cells, and monocyte attachment to endothelial cells (23,25). Adiponectin also reduced lipid accumulation in human monocyte-derived macrophages presumably through its suppressive effect on the expression of macrophage scavenger receptor (26). It also inhibited the proliferation of vascular smooth muscle cells (Y. Arita, unpublished data). Moreover, we observed that in mice with hyperadiponectinemia induced by adiponectin-adenovirus, mechanical injury of vascular walls resulted in suppression of proliferation of vascular smooth muscle cells (M.M., unpublished data). These results demonstrate the antiatherogenic property of adiponectin and the involvement of hypoadiponectinemia in the pathophysiology of atherosclerosis. On the basis of these results, it is possible that testosterone-mediated reduction of plasma adiponectin may account, at least in part, for the higher incidence of atherosclerotic diseases in men.

Previous studies indicated that androgens reduced insulin sensitivity in humans and rodents (37–39). Female patients with polycystic ovary syndrome, which is characterized by hyperandrogenemia, also exhibit insulin resistance, although the exact mechanism remains undetermined. In the present study, castration-mediated increase in plasma adiponectin was accompanied by improvement of insulin sensitivity in male mice, estimated by HOMA-IR and insulin sensitivity test. Our results also showed a significant correlation between plasma adiponectin levels and the degree of insulin sensitivity when the data of all examined male and female mice were plotted. This was in accordance with our previous observation in humans and monkeys showing a significant correlation between plasma adiponectin levels and insulin sensitivity in the whole body (9,11). In the present study, plasma levels of leptin, which was shown to modify insulin sensitivity, were similar in male mice with or without castration. Taken together, testosterone-mediated suppression of plasma adiponectin may be responsible, at least in part, for the lower insulin sensitivity in males than in females.

In the present study, testosterone reduced adiponectin levels in plasma of mice and in media of 3T3-L1 adipocytes. These results were accounted for by the inhibitory effect of testosterone on the secretion of adiponectin from adipocytes, shown in the pulse-chase studies (Fig. 5). In Fig. 4B, total protein levels of adiponectin were not affected with testosterone treatment, although the newly synthesized adiponectin was more retained (Fig. 5C), suggesting that amount of adiponectin secreted into media was much smaller when compared with the intracellular stores. Although testosterone, classically, binds to its cognate nuclear receptor and enhances the transcription of the target genes by binding to the specific cis-element of the promoters (41), the adiponectin mRNA was not affected by the testosterone treatment in vivo and in vitro. Existence of another factor involved in the secretion machinery of adiponectin, whose activity is modulated by testosterone, may be suggested.

From preventive and therapeutic aspects, understanding the mechanisms involved in the regulation of adiponectin secretion may allow the design of a new class of agents to combat insulin resistance and atherosclerosis. In this regard, identification of the responsible factor for adiponectin secretion, whose transcription or activity is modulated by testosterone, should provide useful information for the therapeutic strategy.

In conclusion, the current study revealed that testosterone reduces plasma concentration of adiponectin, an adipose tissue-derived antidiabetic and antiatherogenic protein. Our finding may relate to the high risks of diabetes and atherosclerosis in men.

This work was supported in part by “Research for the Future” Program from the Japan Society for the Promotion of Science (JSPS-RFTF97L00801), Grants-in-Aid from the Ministry of Education, Science, Sports and Culture of Japan (12307022, 12557090, and 12671084), and a grant from Fuji Foundation for Protein Research.

We thank Dr. Shoji Tajima (Osaka University) for invaluable suggestions to the pulse-chase study. We also thank Sachiyo Tanaka and Yuko Matsukawa of our laboratory for helpful technical assistance.