Sexual Differentiation, Pregnancy, Calorie Restriction, and Aging Affect the Adipocyte-Specific Secretory Protein Adiponectin

The effects of puberty, pregnancy, lactation, ovariectomy (OVX), and estrogen, prolactin, and bromocriptine implants were studied in FVB mice (Charles River Laboratories) housed in groups of two to five in filter-top cages in a controlled environment (22–25°C, 40–50% humidity, and 12-h light-dark cycle with lights on from 0600–1800). The mice had free access to water and Teklad Rodent Diet no. 8604 (Harlan Teklad) at the Association for Assessment of Laboratory Animal Care-accredited facility of the Albert Einstein College of Medicine. Avertin 1.2% (Sigma-Aldrich) was used for general anesthesia (0.2 ml/10 g body wt) to perform bilateral OVX and introduce implants. Silastic capsules of 15 mm length, 1.47 mm internal diameter, and 1.96 mm outer diameter were packed with either 17β-estradiol (E2) and cholesterol (at a 1:9 ratio) or cholesterol (Steraloids). Model 1007D osmotic pumps (Alzet) delivered prolactin (Dr. A.F. Parlow, National Institute of Diabetes and Digestive and Kidney Diseases) and 2-bromo-α-ergocryptine (Sigma). The animal research review committee at the Albert Einstein College of Medicine approved the procedures.

The effects of neonatal and adult castration (CX) were studied in Long-Evans rats (Charles River Laboratories) housed individually in polypropylene cages covered with polyester filter bonnets in an accredited animal facility at Johns Hopkins University. The animals were maintained on a constant 16/8-h light-dark cycle with lights on at 0600 h Eastern standard time. Food and sterile tap water were available ad libitum. Female rats were bred in the laboratory, and gestational day 1 was based on the presence of a sperm plug. Pregnancy was monitored, and the day of birth constituted postnatal day (PND) 1. At PNDs 2–4, all pups were removed from the dams and held on a heating pad (with towels layered on top, to avoid overheating) to maintain body temperature. During this time, male pups assigned to have their gonads removed were anesthetized by hypothermia and castrated bilaterally, and incisions were closed with adhesive. Control males received sham operations. The procedures were approved by the Johns Hopkins animal care and use committee.

The effects of adult CX in the context of glucocorticoid treatment were studied in Swiss Webster mice (Charles River Laboratories) electronically tagged (Biomedic Data System) in plastic cages (one animal per cage) under standard laboratory conditions with a 12-h light-dark cycle at a constant temperature of 20°C and at 48% humidity. The mice had free access to water and standard Agway RMH 3000 rodent diet (Agway). Orchidectomy and prednisolone pellet implants were performed at the same time. Slow-release pellets (Innovative Research of America) of placebo or 2.1 mg · kg⁻¹ · day⁻¹ prednisolone were implanted, as previously described (13). Acrp30 was measured after 28 days. The Division of Laboratory and Animal Medicine at the University of Arkansas for Medical Sciences approved the protocols.

The effects of CR and aging were studied in C57BL/6JxBALB mice (F1) maintained in a pathogen-free barrier at the Jackson Laboratories (Bar Harbor, ME). Starting at ∼8 weeks of age, food intake (NIH-31 rodent diet) was measured weekly, and adjustments were made to ensure that the calorie-restricted group was allotted 60% of the ad libitum baseline intake. Mice were free of known pathogens (as detailed in quarterly reports from The Jackson Laboratory). The Jackson Laboratory animal care and use committee approved the protocol for these studies.

Blood samples were obtained from the tail vain or the orbital sinus by heparinized capillary tubes (Fisher) between 0900 and 1100 and stored at −20°C until measurement of Acrp30.

The chloramine T method (14) with one mCi of 125I (New England Biolabs) was used to radiolabel full-length murine recombinant Acrp30 (6) for clearance studies and rabbit IgGs used for Western blots. Unincorporated label was removed by gel filtration on a PD-10 column (Amersham Pharmacia Biotech). The radiolabeled product used in the clearance studies was detected as a single band on SDS-PAGE at 30 kDa.

3T3-L1 murine fibroblasts were propagated and differentiated as previously described (15). Dulbecco’s modified Eagle’s medium was purchased from Cellgro. E2 and ICI 182,780 were purchased from Steraloids and ICI Pharmaceuticals, respectively, whereas all other chemicals were purchased from Fisher.

3T3-L1 cells were washed twice with cold PBS and lysed in TNET buffer (1% Triton X-100, 150 mmol/l NaCl, 5 mmol/l EDTA, and 50 mmol/l Tris, pH 8.0) containing 60 mmol/l octylglucoside (Roche) in the presence of protease inhibitors. Samples were centrifuged at 15,000g for 15 min, and the supernatant was transferred to a fresh tube. Protein determinations were made on the supernatants using the bicinchoninic acid assay (Pierce Chemical, Rockford, IL), and equal amounts of protein were used to measure Acrp30. Monoclonal antibody-15 AER611 (Lab Vision) was used for Western blot analysis of the estrogen receptor (ER).

Total RNA from tissue and 3T3-L1 adipocytes was isolated with Trizol (Life Technologies). For Northern blot analysis, 20 μg of total RNA was used, as previously described (16). Blots were prehybridized in Ultrahyb (Ambion) for 2 h at 42°C. Denatured 32P-labeled DNA probes were added (2 ×10 cpm/ml), and blots were hybridized overnight (42°C). The filters were washed in 2 × sodium chloride-sodium citrate (SSC)/0.1% SDS and 0.1 × SSC/0.1% SDS before autoradiography.

Fat pads were dissected after sacrifice, immersed in cold extraction buffer (20 mmol/l Tris-HCl, pH 7.5, 5 mmol/l EDTA, 10 mmol/l KCl, and 1 mmol/l phenylmethylsulfonyl fluoride), and sonicated for 20 s. The resulting homogenate was centrifuged at 4°C to separate lipids, lysate, and insoluble material. The lysate infranatant was removed and mixed with hot SDS (95°C) to a final concentration of 2% wt/vol. After 5 min, the lysates were cooled to room temperature and assayed for protein concentration by bicinchoninic acid (Pierce Protein Assay Kit; Pierce Chemical) and assayed for Acrp30 content by quantitative Western blot analysis.

Acrp30 in plasma (3 μl) or tissue (50 μg protein) was measured by Western blot analysis. After SDS-PAGE (12%), proteins were transferred to BA83 nitrocellulose (Schleicher & Schuell). The blots were exposed to 125I-labeled rabbit antibody against murine Acrp30 and analyzed with a PhosphorImager (Molecular Dynamics, Sunnyvale, CA), using ImageQuant 1.2 software. Each gel contained four standards of purified mouse Acrp30 at four different concentrations to ensure linearity and reproducibility of the signal. Intra- and interassay variations for Acrp30 were <10%, and the limit of detection was ∼5 ng of Acrp30 (17).

The results are shown as the means ± SE. Statistical analysis was performed by one- or two-way ANOVA, unless otherwise indicated. Significance was accepted at P < 0.05.

At 1 week of age, circulating Acrp30 values in male and female FVB mice were 3.0 ± 0.8 μg/ml. By 6 weeks, Acrp30 reached 10.2 ± 2.0 and 30.1 ± 4.2 μg/ml in males and females, respectively (P < 0.001) (Fig. 1A). The clearance of intravenous [125I]Acrp30 did not differ between adult males and females (Fig. 1B). A number of studies have linked elevated Acrp30 levels with improved insulin sensitivity (6). Elevated serum Acrp30 levels in females are consistent with the improved insulin sensitivity of female FVB mice compared with their age-matched male counterparts, as judged by decreased glucose and insulin levels (Figs. 2A and B), as well as an improved insulin tolerance test (Fig. 2C). We measured both Acrp30 mRNA and protein levels in various fat pads isolated from male and female mice. Surprisingly, Northern blot analysis did not reveal significant sex differences in Acrp30 mRNA of gonadal white and interscapular brown adipose tissue (Fig. 3A). In contrast, Western blot analyses showed that Acrp30 protein levels within the fat depots are significantly lower in males than females, suggesting sexually dimorphic protein production and secretion in adipocytes (Fig. 3B).

Elevated neonatal androgen exposure is associated with sex-specific imprinting. Adult Acrp30 levels were therefore measured in Long-Evans rats that had undergone CX 2–4 days after birth. Neonatal CX allowed plasma Acrp30 in adult males to rise above controls (P < 0.001) and reach female values. CX in adult rats had no affect (Fig. 4A). Acrp30 levels in Swiss Webster and FVB mice were also unaffected by adult CX (Fig. 4B and data not shown). However, Acrp30 was significantly upregulated (P < 0.001) in an osteoporosis model that combined CX with the synthetic glucocorticoid analog prednisolone (13,18) (Fig. 4B). Secondary hypogonadism and increased glucocorticoid activity are also found under other physiological conditions, such as chronic food restriction. Accordingly, Acrp30 levels were higher in both young and old male F1 (C57BLxBALB) mice under chronic calorie-restriction than in animals with ad libidum access to food (28.2 + 2.0 and 30.0 + 2.0 μg/ml vs. 17.0 + 2.0 and 18.0 + 2.0 μg/ml) (Fig. 4C). In agreement with observations in Fig. 3, changes in serum Acrp30 are brought about posttranscriptionally, since mRNA levels in gonadal white adipose tissue remain unchanged (Fig. 4D).

To determine whether ovarian maturation during puberty plays a role in the sex-specific expression of Acrp30 in females, the ovaries were surgically removed in FVB mice at 1 week of age. OVX before puberty (1 week) did not prevent Acrp30 from reaching normal female levels (Fig. 5A). In fact, OVX in adults allowed plasma Acrp30 to increase 10 days after surgery above sham-OVX controls (42.1 ± 3.2 vs. 30.5 ± 4.2 μg/ml, respectively, in OVX and sham-OVX; P < 0.05). The effect of estrogen withdrawal associated with OVX was reflected in the uterine weights (18 ± 4 and 98 ± 10 mg, respectively, in OVX and sham-OVX; P < 0.001). Mice that received E2 implants and OVX were examined to determine whether estrogen might suppress Acrp30. In agreement with a suppressive effect of estrogen, Acrp30 levels decreased after 10 days of continuous exposure to supraphysiological levels of E2 (250–350 pg/ml) (Fig. 5A). Once again, uterine weights reflected the bioactivity of the E2 implants (92.4 ± 8.3 and 21.9 ± 4.0 mg, respectively, in E2- and vehicle-implanted mice).

This suggests that estrogen exerts a negative impact on Acrp30 levels in vivo. Because estrogen levels drop dramatically after ovarian cessation, we predicted that levels of Acrp30 should increase in old anovulatory mice. To test this hypothesis, we performed a cross-sectional study, looking at age-related differences in serum Acrp30 levels in young (4–5 months) and old (24–30 months) F1(C57BLxBALB) mice (Fig. 5B). Acrp30 levels are ∼80% higher in old mice. We cannot, of course, exclude other age-related influences on Acrp30 expression. However, if the increase in Acrp30 under food restriction is a consequence of secondary hypogonadism, we would expect to see a rise of Acrp30 in response to food restriction in young mice (reduced estrogen in response to reduced caloric intake) but not in old anovulatory mice (low estrogen under all conditions), and this is indeed the case (Fig. 5B). Similar to the observations in Figs. 3 and 4, changes in serum Acrp30 are brought about posttranscriptionally, since mRNA levels in gonadal white adipose tissue remain unchanged (Fig. 5C). This observation emphasizes that depending on the physiological state, tissue mRNA and serum protein levels do not always correlate. The suppression of Acrp30 in vivo may, however, be an indirect effect of elevated E2 caused by a number of downstream effects, such as the estrogen-mediated suppression of pituitary leuteinizing hormone and follicle-stimulating hormone. Hence, 3T3-L1 adipocytes were used to determine whether the effects of estrogen in vivo were direct or indirect. FCS (10%) used to maintain adipocyte cultures is already rich in gestational estrogens, but the addition of increasing amounts of E2 suppressed Acrp30 levels (Fig. 6A; left panel), and the effect was blocked by the potent ER antagonist ICI 182,780 (19). Secretion of the protein into the extracellular medium was blocked as well (Fig. 6A; right panel). The specificity of estrogen action is supported by the presence of ER in differentiated 3T3-L1 adipocytes detected with a monoclonal antibody against murine ERα (Fig. 6B). A similar 66-kDa band was also detected in uterine extracts without immunoprecipitation. Interestingly, treatment of 3T3-L1 adipocytes with E2 suppressed Acrp30 mRNA levels, suggesting that in this isolated in vitro system, mRNA and protein go hand in hand (Fig. 6C). This effect reflects direct action of the ER on the Acrp30 gene, since similar effects can be observed in the presence of the protein biosynthesis inhibitor cycloheximide. Pregnancy is associated with a remarkable increase of various estrogens (20), and consistent with the E2 effects described herein, Acrp30 in mice decreased in the second and third week of gestation (P < 0.05) (Fig. 7A). Food intake increases by nearly 100% late in gestation, but the suppression of plasma Acrp30 late in gestation was evident even after 1 week of 50% CR (P < 0.05) (Fig. 7B). Despite the postpartum drop in estrogen, Acrp30 levels remained suppressed during lactation (P < 0.05) (Fig. 7A).

The effect of elevated lactogens seen during lactation and late gestation was assessed with osmotic minipumps releasing 50–100 μg of prolactin per day. Consistent with a suppressive effect of prolactin, circulating Acrp30 levels decreased after 72 h of elevated prolactin exposure (P < 0.05) (Fig. 8A). In contrast, Acrp30 levels increased with bromocriptine implants, presumably under dopaminergic inhibition of prolactin synthesis and secretion (P < 0.05) (Fig. 8A).

The correlation between low body weight, low prolactin, elevated insulin sensitivity, and high Acrp30 levels seen with CR is also evident in the Snell dwarf mouse (Fig. 8B). This model is homozygous for a mutation in pituitary transcription factor-1 (Pit-1), which plays a key role in the pituitary’s expression of prolactin, growth hormone, and thyroid-stimulating hormone (21).

We observed that circulating Acrp30 levels in mice increased abruptly by the third week of postnatal life but were relatively stable thereafter. In contrast to Acrp30, circulating levels of another adipose-specific marker, leptin, are relatively constant from 1 to 6 weeks of age and only begin to rise by the seventh week (22). What triggers the rise of Acrp30 early in puberty is not known. Our results, however, indicate that it does not involve the maturation of a gonadal feedback loop because surgical excision of the testes or ovaries before puberty did not interfere with the rise of Acrp30 in adults. Marked differences between the peripubertal secretions of male and female gonads support the likelihood that the pubertal Acrp30 surge is mediated by nongonadal mechanisms, and the similar timing in the male and female Acrp30 surge suggests a common mechanism. Previous studies suggest that the peripubertal surges of growth hormone, IFG-I, prolactin, and luteinizing hormone are also triggered by nongonadal mechanisms (23–25).

Adult female mice have higher Acrp30 levels than males, in agreement with the sexual dimorphism previously described in humans (10,11,26). A similar sexual dimorphism has also been described for leptin, even after accounting for differences in adipose tissue mass (27). Like humans and mice, the Long-Evans rat exhibits a sex-specific difference in plasma Acrp30 as well as leptin. We demonstrate for the first time that neonatal CX clearly feminized both Acrp30 and leptin levels in the adult male. It is unlikely that the lack of androgens during the neonatal period directly affects Acrp30 levels. Much rather, the lack of androgens during this critical postnatal period exerts its effect through an imprinting mechanism presumably at the level of the brain or adipose tissue. In either mouse or rat, adult CX did not have a significant effect on Acrp30. Hence, the testes may only have an irreversible effect in the sexual differentiation of adipose tissue during the critical neonatal period, as previously shown in the hypothalamus, pituitary, and liver (28–30). We observed that male fat depots had lower Acrp30 levels, even though Acrp30 mRNA levels were indistinguishable between males and females, indicative of a sexual dimorphism with respect to posttranscriptional regulation. We report similar discrepancies between mRNA and serum protein levels under two additional conditions in vivo. This emphasizes the need to measure serum protein levels as the final readout and highlights the dangers of drawing any conclusions based on tissue mRNA levels alone. However, in many instances, mRNA and protein levels do agree with each other, and we cannot draw the conclusion from these results that serum Acrp30 levels are exclusively governed by posttranscriptional mechanisms.

Mature ovaries have a mild but significant influence on adult Acrp30 levels. OVX in young cycling mice induced plasma Acrp30 (10 days after surgery), and E2 implants reversed the effect. Increased adiposity is generally associated with OVX. However, the increase of Acrp30 in this case was measured before weight gain. The ability to block the effect of E2 in 3T3-L1 adipocytes with ICI 182,780, an ER antagonist, suggests that E2 acts directly on adipose tissue. However, whereas 3T3-L1 adipocyte mRNA and protein levels correlate, such correlations were not observed in the in vivo setting. The suppression of Acrp30 we observed during pregnancy and the elevation of Acrp30 in anovulatory mice aged 24–30 months are consistent with opposite extremes of estrogen action. The parallel increase of adipose mass, aromatase activity, and nonovarian estrogens may also play a key role in the suppression of plasma Acrp30 in obese individuals (31). The correlation during puberty between increasing estrogen and Acrp30 is therefore distinct from their inverse relationship in adulthood. However, more rigorous testing is needed to determine the physiological significance of estrogen action on Acrp30.

Do plasma Acrp30 levels thus far reflect the insulin-sensitizing effects of recombinant Acrp30 and all of the correlative data that illustrate the relationship between insulin resistance and low plasma Acrp30? Changes in insulin action during puberty have been poorly described in rodent models. We showed that in young adult mice, the fasting glucose, insulin, and insulin tolerance tests indicate that females are more insulin sensitive than their male littermates. Early studies in adult rats showed that 95% pancreatectomy resulted in a higher incidence of diabetes in males than females even after CX or OVX (32). The development of diabetes after pancreatectomy did not differ between female rats that were masculinized by neonatal exposure to elevated testosterone and male rats (33). A sexual dimorphism has also been described for non-insulin-dependent diabetes in the Wistar fatty rat on a high-sucrose diet (34).

Both humans and rodents develop insulin resistance during pregnancy (35,36). Here, we show for the first time that Acrp30 levels are suppressed from mid-gestation until weaning. Of all the endocrine changes that peak around pregnancy, placental lactogens and prolactin best fit the time course for the suppression of Acrp30. Placental prolactin-like molecules appear in the maternal circulation around mid-gestation, whereas pituitary prolactin levels increase progressively during gestation and remain elevated until suckling ceases (37). Isolated hyperprolactinemia is associated with insulin resistance, which is sometimes treated with various dopamine receptor agonists like bromocriptine (38,39). Indeed, the expected effects of prolactin and bromocriptine implants were observed on plasma Acrp30 in adult mice. Once again, the parallel rise of prolactin and Acrp30 in plasma seen during puberty should be distinguished from their apparently inverse relationship in adulthood.

It is well known that chronic CR (not malnutrition) increases insulin sensitivity in adult rodents compared with ad libitum feeding (40). Based on the consistent correlation between plasma Acrp30 and insulin sensitivity, it was not surprising to see that CR is associated with elevated plasma Acrp30 in young male and female mice. CR in rodents has also been used to understand the process of aging since the life-extending effects of food restriction were first in described rats (41). One theory is that CR delays the onset of changes associated with senescence, and we tested this by measuring Acrp30 levels in young and old ad libitum-fed and CR mice. The progression of biological senescence in common strains of laboratory mice is well advanced by 23–30 months, yet ad libitum-fed males maintained the same levels of Acrp30 as seen at 3–5 months. Moreover, CR led to a similar induction of Acrp30 in young and old males. On the other hand, old female mice had higher Acrp30 levels than young mice, and CR failed to stimulate Acrp30. The induction of Acrp30 we observed with the combination of CX and prednisolone treatment in Swiss Webster mice may be related to the mild stress response and secondary hypogonadism triggered by CR (42,43).

Because of a mutation in Pit-1, the Snell dwarf mouse (−/−) is a genetic model of anterior pituitary hormone deficiency of prolactin, thyroid-stimulating hormone, and growth hormone (42,44,45). We observed a significant elevation of plasma Acrp30 in the Snell dwarf mouse that is consistent with its marked elevation in insulin sensitivity (46,47). Elevated insulin sensitivity is considered to play a key role in life extension across phyla (48).

The evidence for a role of Acrp30 in insulin sensitivity was recently underlined by the development of mice carrying a chromosomal deletion of Acrp30. Thus far, two independent reports describe that in oral glucose tolerance tests and insulin tolerance tests, the Acrp30 (+/−) and Acrp30 (−/−) mice have decreased insulin sensitivity (49,50). Future studies may compare the insulin sensitivity between male and female Acrp30 (−/−) mice to determine the significance of Acrp30 in sex-specific differences or CR.

How is it possible that a female steroid hormone like estrogen could suppress Acrp30 if basal Acrp30 values are clearly greater in females than males? This paradox suggests that the inhibitory effect of elevated estrogen exposure in females cannot eliminate the sex-specific “set point.” We showed that early postnatal CX is the only manipulation that could alter the sexual dimorphism, suggesting that neonatal testicular secretions play a major role in establishing a sex-specific set point for Acrp30. Therefore, we propose that the set point, which determines basal Acrp30 levels in males and females, is established by the neonatal testosterone surge. Estrogen in adults only modulates Acrp30 values.

This work was supported by a Hormones/Membrane Interactions Training Grant from the National Institutes of Health (T32 DK 07513-15; to T.P.C.), the University of California at Davis, a National Research Service Award (F32 DK61228; to T.P.C.), an American Diabetes Medical Scientist Training Grant (to A.H.B.), an American Diabetes Association Research Grant Award (to M.E.P.), the Adler Foundation (M.E.P. and J.C.J.-C.), a National Institutes of Health Medical Scientist Training Grant (T32-GM07288; to P.I.), a VA Research Enhancement Award Program Grant and VA Merit Review Grant from the Office of Research and Development, Department of Veterans Affairs, and National Institutes of Health grants R01-AR46191 (to R.S.W.) and 1R01-DK55758 (to P.E.S.).

We are thankful to the members of Dr. Scherer’s laboratory for their critical comments on the manuscript, Drs. K. Flurkey and D. Harrison for kindly providing serum samples from Snell dwarf mice (The Jackson Laboratory), Dr. B. Frenkel for initiating the glucocorticoid studies (USC Institute for Genetic Medicine), Dr. A.F. Parlow for prolactin (National Hormone and Peptide Program, NIDDK) and Drs. P. Wise (University of California at Davis) and A. Etgen (Albert Einstein College of Medicine) for their helpful suggestions.