Impaired Preadipocyte Differentiation Into Adipocytes in Subcutaneous Abdominal Adipose of PCOS-Like Female Rhesus Monkeys

Abstract

Metabolic characteristics of polycystic ovary syndrome women and polycystic ovary syndrome-like, prenatally androgenized (PA) female monkeys worsen with age, with altered adipogenesis of sc abdominal adipose potentially contributing to age-related adverse effects on metabolism. This study examines whether adipocyte morphology and gene expression in sc abdominal adipose differ between late reproductive-aged PA female rhesus monkeys compared with age-matched controls (C). Subcutaneous abdominal adipose of both groups was obtained for histological imaging and mRNA determination of zinc finger protein 423 (Zfp423) as a marker of adipose stem cell commitment to preadipocytes, and CCAAT/enhancer binding protein (C/EBP)α/peroxisome proliferator-activated receptor (PPAR)δ as well as C/EBPα/PPARγ as respective markers of early- and late-stage differentiation of preadipocytes to adipocytes. In all females combined, serum testosterone (T) levels positively correlated with fasting serum levels of total free fatty acid (r² = 0.73, P < .002). PA females had a greater population of small adipocytes vs C (P < .001) in the presence of increased Zfp423 (P < .025 vs C females) and decreased C/EBPα (P < .003, vs C females) mRNA expression. Moreover, Zfp423 mRNA expression positively correlated with circulating total free fatty acid levels during iv glucose tolerance testing (P < .004, r² = 0.66), whereas C/EBPα mRNA expression negatively correlated with serum T levels (P < .02, r² = 0.43). Gene expression of PPARδ and PPARγ were comparable between groups (P = .723 and P = .18, respectively). Early-to-mid gestational T excess in female rhesus monkeys impairs adult preadipocyte differentiation to adipocytes in sc abdominal adipose and may constrain the ability of this adipose depot to safely store fat with age.

Increased adiposity enhances the severity of polycystic ovary syndrome (PCOS) (1, 2), beginning in adolescence (3, 4) and increasing with age. Women with PCOS develop greater abdominal adiposity than normal women of comparable body mass index (BMI), which positively correlates with insulin resistance, altered adipose function, and glucose dysregulation (5). Moreover, sc abdominal adipocyte morphology appears altered in PCOS women (6), potentially linking altered adipocyte morphology with a perturbed capacity of sc adipose to safely store fat (7, 8). This phenomenon is important, because metabolic dysfunction in humans likely results from ectopic lipid accumulation in nonadipose cells, causing lipotoxicity (9). Thus, when energy intake exceeds the capacity of normal sc adipose to safely store fat, excess free fatty acids (FFAs) become deposited in abnormal locations, such as muscle and liver, where they induce oxidative/endoplasmic reticulum stress tightly linked with insulin resistance and inflammation (6–10).

Altered adipose function also occurs in prenatally androgenized (PA) adult female monkeys exposed during early-to-mid gestation to male levels of testosterone (T) (11). Similar to PCOS women, PA monkeys express ovarian hyperandrogenism, ovulatory dysfunction and polycystic ovaries (12–14), and progressively manifest metabolic dysfunction with age (15), thereby increasing abdominal adiposity in combination with hyperlipidemia, insulin resistance, and type 2 diabetes (5, 11, 16).

These findings suggest that gestational T excess in female nonhuman primates may alter adipogenesis at any of its 3 stages of development: 1) adipose stem cell (ASC) commitment to preadipocytes via bone morphogenetic protein 4 (BMP4)-associated zinc finger protein 423 (Zfp423) activation; 2) early-stage preadipocyte differentiation via CCAAT/enhancer binding protein (C/EBP)β and C/EBPδ activation and peroxisome proliferator-activated receptor (PPAR)δ and C/EBPα induction; and/or 3) late-stage preadipocyte differentiation to mature adipocytes through further activation of PPARγ, C/EBPα, and other transcriptional factors (17–22). In support of this, exposure of cultured human ASCs (hASCs) from sc abdominal adipose of normal women to androgen impairs their commitment to preadipocytes and subsequent early-stage preadipocyte differentiation through androgen receptor action (23). To determine whether gestational T excess similarly affects sc abdominal adipogenesis in vivo, the present study examines whether changes in sc abdominal adipose morphology and gene expression occur in relation to metabolic and reproduction dysfunction in PA female monkeys.

Materials and Methods

Ethics statement

The Institutional Animal Care and Use Committee of the Graduate School of the University of Wisconsin-Madison approved all procedures used in the study, and the care and housing of the monkeys was in accordance with the recommendations of the Guide for the Care and Use of Laboratory Animals and Animal Welfare Act with its subsequent amendments.

Animals

The 12 adult female rhesus monkeys (Macaca mulatta) used in this study were maintained at Wisconsin National Primate Research Center (WNPRC), according to standard protocol (24, 25). They were fed Purina monkey chow (product number 5038; Ralston Purina) with occasional supplementation of fresh fruits. This formulation of monkey chow provides 70% of calories as carbohydrate, 13% as fat, and 17% as protein. Age and BMI (body weight [kg]/crown-rump length [m²]) (26) of both female groups were comparable (Table 1). Six of the female monkeys were exposed to gestational T excess by sc injection of their dams with 10-mg T propionate starting on gestational days 40–44 for 15–35 consecutive days (early-to-mid gestation, PA). Only 6 of the 14 previously described PA monkeys were studied (27), because they exhibited high T levels and intermittent menstrual cycles (16) compared with controls (C) (Table 1) and provided appropriate quality tissue at necropsy. The 6 C female monkeys in this study were not exposed to exogenous T excess in utero and were selected from monkeys not otherwise manipulated during gestation by other investigators or colony management at the WNPRC.

Table 1.

Selected Parameters of Study Monkeys

Parameter	C Monkeys	PA Monkeys
Age/somatometric
Age (y)	24.0 ± 0.9	25.8 ± 0.7
BMI (kg/m2)	41.9 ± 2.2	38.1 ± 2.9
% Abdominal fat	48.6 ± 3.8	33.3 ± 3.1a
Subcutaneous abdominal fat (cm3)	816 ± 362	346 ± 84
Visceral fat (cm3)	1066 ± 98	630 ± 99b
Ratio sc abdominal/visceral fat	0.7 ± 0.3	0.7 ± 0.2
Metabolic physiology
Basal total FFAs (mM/L)	522 ± 84	1052 ± 380c
Area-under-the curve total FFAs (mM/L per 180 min)	75.7 ± 0.9	104.1 ± 15.2
Basal insulin (μU/mL)	52.7 ± 12.7	19.3 ± 36.7
Basal glucose (mg/dL)	57.9 ± 1.4	73.5 ± 10.2
Insulin sensitivity (SI)	1.4 ± 1.0	3.4 ± 2.7
Glycosylated hemoglobin (%)	8.0 ± 0.4	9.6 ± 1.6
Reproductive physiology
Basal T (ng/mL)	0.2 ± 0.1	0.4 ± 0.1d
% Intermittent/absent menstrual cycles	0	67e
% Polycystic ovaries	0	83f
Ratio of mRNA expression
Zfp423/C/EBPα	0.12 ± 0.01	0.35 ± 0.02g
Zfp423/PPARλ	0.40 ± 0.03	0.41 ± 0.03
Zfp423/PPARδ	0.57 ± 0.07	0.62 ± 0.02
Positive correlation with basal T levels (ng/mL)
Basal insulin (μU/mL)	r2 = 0.65, P < .029 (All female monkeys combined)
Glycosylated hemoglobin (%)	r2 = 0.62, P < .035 (All female monkeys combined)
Total FFAs (μM/L)	r2 = 0.73, P < .002 (All female monkeys combined)

Data are mean ± SEM.

^aP < .021 vs C.

^bP < .029 vs C.

^cP < .059 vs C.

^dP < .020 vs C.

^eP < .062 vs C.

^fP < .016 vs C.

^gP < .001 vs C.

In some of the female monkeys in this study, somatometric measures, basal serum T levels, menstrual cycle duration, and ovarian morphology were previously reported but are included here to provide appropriate context for analyses of adipose tissue (12, 13, 16, 27–31). Blood samples providing serum for hormone analyses were obtained from animals accustomed to using a tabletop restraint without anesthesia (29).

Hormone assays

Serum T levels were determined by enzyme-immunoassay, plasma insulin concentrations were measured by RIA (Linco Research, Inc), and plasma glucose concentrations were measured by the glucose oxidase method (Yellow Springs Instruments) in the WNPRC/Institute of Clinical Translational Research Hormone Assay Services Laboratory, as previously reported (30, 32). T measurements were performed after diethyl ether extraction of serum and solvent fraction separation by celite chromatography. Intra- and interassay coefficient of variation (CVs) for quality control (QC) preparation values were, respectively, T, 3.5% and 14.0%; insulin, 8.1 and 4.8%; and glucose, 4.0 and 2.9%. Because T values have a wide range of variability, we adjusted them within the range of values observed in the study samples (QC preparation value [n = 57]: 0.18 ± 0.04 ng/mL, mean ± SD), with intra- and interassay CVs for QC values of 4.9% and 21.1%, respectively. Glycosylated hemoglobin was assayed by the Glyco-Teck affinity column method, as previously established for rhesus monkeys at WNPRC (33, 34).

Ovarian morphology

During laparoscopic assessment of ovarian dimensions (35, 36) while the animals were sedated with ketamine HCl (10 mg/kg, im), photographic images were taken of transilluminated ovaries (37) at their largest diameter during the early follicular phase (menstrual cycle d 1–5) or an anovulatory interval. Ovarian images with more than 10, approximately 1–3 mm in diameter follicles were scored as polyfollicular (Table 1), a criterion modified from the prevailing ultrasonographic determination of polycystic ovaries in women (38) before the Rotterdam consensus (39), contemporary with ovarian observation.

Frequently sampled iv glucose tolerance test (FSIGT)

Monkeys underwent a single, 3-hour, FSIGT starting at 7–9 am after an overnight fast, according to the tolbutamide-modified minimal model of Bergman and at approximately 4 weeks before dual-x-ray absorptiometry (DXA) and computed tomography scans, as previously reported (31). All FSIGTs were performed during the early follicular phase of a menstrual cycle or an anovulatory period and provided the glucose, insulin, and total FFA data for parameters presented in Table 1.

Abdominal fat determination

Crown-rump length measurements for BMI determination were obtained from each animal while anesthetized for DXA scan.

For DXA, DXA scanner model DPX-L (GE/Lunar Corp) had a precision (CV%) of less than 4% for total body and regional composition in rhesus monkeys. Animals were placed in a supine position for scanning, and images were acquired and analyzed with pediatric software (versions 1.5e and 4.0a, respectively) to determine total body and abdominal lean and fat mass, as reported previously (11). The abdominal region of interest was defined by area between the thoracic 12-lumbar (L) interspace and the L6-L7 interspace.

Computed tomography was performed within 7–8 days of the DXA scan, as previously reported (11), using a HiLight Advantage scanner (GE) at 120 kV and 200 mA with a scan time of 2 seconds and slice thickness of 0.5 cm. The abdominal region was analyzed using SliceOMatic (version 4.3 rev0h; Tomovision).

Preparation and morphological assessment of sc abdominal fat

Subcutaneous fat was collected after monkeys were killed. Monkeys were put through timed injection with ketamine and barbiturate before necropsy. During the necropsy protocol, abdominal fat was obtained from sc depots adjacent to the umbilicus. Subcutaneous fat was placed into sterile cryovials, which were immediately placed into liquid nitrogen and transported to a −80°C freezer soon afterwards.

For cell sizing studies, approximately 0.5-g frozen sc abdominal fat was embedded in optimal cutting temperature medium, cut into 10-μm sections, stained with hematoxylin and eosin following standard protocols, and digitally imaged using Aperio ScanScope. Six sections were cut from each sample, and a minimum of 100 cells per section was analyzed. Adipocyte area (AA) and circumference were determined from the digital images using Aperio ImageScope software, which is internally calibrated (50 000 pixels per inch, 0.5 μm per pixel).

mRNA expression

Total cellular RNA from frozen tissue samples was isolated using an RNeasy kit (QIAGEN) according to manufacturer's protocol. mRNA expression of Zfp423, PPARδ, C/EBPβ, C/EBPα, and PPARγ was determined by real-time quantitative RT-PCR (qRT-PCR) under standard conditions. All primers were provided by A&B Applied Biosystems. Human primer for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used to detect GAPDH mRNA as an internal C housekeeping gene (Life Technologies Corp). The relative expression of target genes was measured using the comparative critical threshold (C_t) method. The results were expressed as mRNA expression relative to GAPDH using the formula 2^−δCt(δC_t is the difference between the mean C_t values of genes of interest and the mean C_t value GAPDH). The results were expressed as fold changes (PA vs C). As a negative C, first strand DNA synthesis was performed in the absence of mRNA followed by qRT-PCR.

Statistical analyses

All results were expressed as mean ± SEM. Adipocyte size was transformed using the square root transformation for display purposes. The empirical cumulative distribution function was calculated and plotted for each group (treated/untreated). The empirical cumulative distribution function was constructed by finding the probability that the adipocyte size was less than or equal to the given value. The treated and untreated distributions were compared using the two-sample Kolmogorov-Smirnov test. For qRT-PCR studies, each sample was subjected to at least 3 independent experiments performed in triplicates. Statistical significance was determined as P < .05. Regression analysis was used to examine the associations between gene expression and area under the curve (AUC) for total FFA or circulating T levels. Similar analysis was performed to determine the association between basal total FFA and circulating T levels.

Results

Monkey baseline parameters

Although the percentages of abdominal and visceral fat content were lower in PA than C monkeys, both female groups had a similar approximately 0.7 ratio of sc abdominal to visceral fat (Table 1). Glucoregulatory parameters were similar between female groups, except for a trend for an increase in basal (fasting) circulating concentrations of total FFA in PA monkeys.

Reproductive parameters distinguished the 2 female groups, with increases in circulating T levels, and incidences of oligo-ovulation and polycystic ovaries in PA monkeys (Table 1). In all females combined, circulating T levels positively correlated with overnight basal circulating levels of total FFA, insulin, and glycosylated hemoglobin (Table 1; also, see figure 3 below), with PA monkeys contributing both high T concentrations and increased total FFA, insulin, and glycosylated hemoglobin levels.

Adipocyte morphology

A similar number of cells was analyzed between C and PA females. No significant differences in adipocyte morphology, mean AA (C = 6052 ± 827 μm² vs PA = 6224 ± 1085 μm²), mean adipocyte circumference (C = 331 ± 45 μm vs PA = 332 ± 75 μm), or ratio of AA/adipocyte circumference (C = 15.6 ± 3.8 vs PA = 16.0 ± 3.8) were observed between female groups (Figure 1, A–C). The cell size distribution analysis, however, showed a significant shift (P < .001) toward smaller sc adipocyte diameter in the PA compared with C females (Figure 1D).

Figure 1.

Quantification of adipocyte cell size in PA vs C adult female rhesus monkeys. Representative images of sc abdominal adipocyte samples stained with hematoxylin and eosin (H&E) in C (A) and PA monkeys (B). C, Graphical representation of the sc AA (mean ± SEM) in PA (n = 6) vs C (n = 6) monkeys. D, Subcutaneous adipocyte size distributions shows a significant shift toward smaller sc adipocyte diameter in PA vs C monkeys; *, P < .001 vs C.

Gene expression analysis

mRNA expression of Zfp423 in sc adipose of PA monkeys was increased (2.77 ± 0.17-fold change vs C; P < .02) (Figure 2A), whereas C/EBPα mRNA expression was reciprocally decreased (0.26 ± 0.05-fold change vs C; P < .002) (Figure 2C). No female-type differences were detected in sc adipose of PPARδ (1.15 ± 0.04-fold change vs C; P = .723) (Figure 2B) or PPARγ (2.05 ± 0.50-fold change vs C; P = .18) mRNA expression (Figure 2D). Furthermore, in a subset of PA and C females combined, sc adipose Zfp423 mRNA expression (relative to GAPDH) positively correlated with AUC total FFA levels during an FSIGT (P < .004, r²=0.66) (Figure 3A). Subcutaneous adipose C/EBPα mRNA expression (relative to GAPDH) negatively correlated with circulating T levels (P < .02, r² = 0.43, all animals combined) (Figure 3B). Subcutaneous abdominal adipose Zfp423, C/EBPα, PPARδ, and PPARγ mRNA expression was not associated with adipocyte morphology, glucoregulatory parameters, BMI, or age.

Figure 2.

mRNA expression of adipogenic genes in PA vs C adult female rhesus monkeys. Subcutaneous abdominal adipose tissue collected from 6 PA and 6 C was immediately placed into liquid nitrogen after animals were killed. mRNA extraction and qRT-PCR were performed as indicated in Materials and Methods. Zfp423 (A), PPARδ (B), CEBPα (C), and PPARγ (D) mRNA levels were quantified by qRT-PCR relative to the housekeeping gene GAPDH. Results represent fold changes in PA (n = 6) vs C (n = 6) monkeys; *, P < .020; **, P < .001 vs C.

Figure 3.

Correlation between selected parameters of all monkeys combined. Regression analysis was performed in PA (black) vs C (no fill) monkeys to determine (A) Zfp423 mRNA expression vs AUC total FFA, C (n = 3), PA (n = 6); (B) C/EBPα mRNA expression vs circulating T, C (n = 5), PA (n = 6); and (C) basal total FFA vs circulating T, C (n = 5), PA (n = 5). mRNA expression of specific genes relative to the housekeeping gene GAPDH was determined by qRT-PCR; circulating T and total FFA were determined by RIA and radiochemical assay, respectively. FFAs were measured throughout the FSIGT, and the basal measurements presented represent the first measurements taken during the test.

Discussion

Altered metabolic function accompanied by increased abdominal adiposity is a salient characteristic of PCOS women (14, 40) and PCOS-like, PA female monkeys alike (14, 15), implicating possible abnormalities of adipogenesis from ASC commitment to preadipocytes to subsequent differentiation of preadipocytes to adipocytes (Figure 4). Using several adipogenic gene markers, the present study shows in sc abdominal adipose of aging female rhesus monkeys that gestational T excess impairs preadipocyte differentiation to mature adipocytes while promoting earlier ASC commitment to preadipocytes, as evidenced by a reciprocal C/EBPα decrease and Zfp423 increase in mRNA expression, respectively. These findings support, in part, our previous observation in normal women that androgens reduce preadipocyte differentiation to mature adipocytes, perhaps limiting adipocyte numbers and fat storage in sc abdominal adipose (23).

Figure 4.

Schematic representation of adipogenesis. ASC commitment to preadipocytes is governed by Zfp423 mRNA expression. Activation of C/EBPβ and PPARδ, and C/EBPα and PPARγ regulates early and late stage preadipocyte differentiation to mature adipocytes, respectively.

In support of this, PA monkeys had a greater population of small sc adipocytes compared with C, suggesting a developmental programming role for T in constraining adipocyte size, as previously shown in female sheep exposed to excess T during gestation (41). Such a phenomenon in PA monkeys may also diminish sc abdominal storage of lipid and, therefore, explain why visceral fat, as an alternative fat depot, increases in amount with BMI in PA, but not C, female monkeys (11). Moreover, an increase in the number of small sc adipocytes is linked with human metabolic dysfunction (42), as is an alternative increase in sc adipocyte size in some PCOS women (6), implying that a change in either direction away from the optimal size of sc adipocytes might have metabolic consequences.

A strength of our study is the determination of adipocyte size using fixed histological sections of adipose tissue instead of using freshly isolated adipocytes obtained by collagenase digestion, which could lyse larger adipocytes leading to inaccurate measurements of the size population of adipocytes (43).

Along with androgen-induced impairment of preadipocyte differentiation as a function of constrained sc abdominal adipose storage capacity, circulating T levels negatively correlated with sc abdominal adipose C/EBPα gene expression while simultaneously positively correlating with total FFA levels, in all monkeys combined, agreeing with PCOS women showing several defects in lipid metabolism (44–46). Moreover, increased FFA release during FSIGT positively correlated with elevated sc abdominal adipose Zfp423 mRNA expression, presumably to enhance ASC commitment to preadipocytes from impaired onward differentiation of preadipocytes to adipocytes.

PA monkeys have reduced numbers of mature sc adipocytes (this study), along with a defect in the ability to increase sc lipid storage in proportion to BMI (11). They also exhibit dyslipidemia, an abdominal fat volume that positively correlates with basal insulin levels (11, 16), and a subtle impairment in insulin sensitivity in some (11), but not all (16), studies that increases the incidence of type 2 diabetes compared with age- and BMI-similar C (15). Our present findings may therefore have identified a molecular component(s) underpinning how prenatal androgen excess constrains the ability of the sc abdominal adipose depot to safely store fat, causing lipotoxicity from ectopic lipid accumulation in nonadipose cells in PCOS-like monkeys.

Our present data also show similarities and differences to those of other reports of T-regulated adipogenic genes. Although Zfp423 gene expression is up-regulated in PA female monkeys, cultured hASCs from women show T inhibition of Zfp423-BMP4-mediated gene expression (23). This discrepancy may be attributed to the differences in species (human vs monkey), study design (in vitro vs in vivo), supraphysiological vs pharmacological concentrations, and/or stimulation techniques (in vitro T and BMP4 stimulation vs in vivo gestational T excess).

Similar to decreased C/EBPα mRNA expression in sc abdominal adipose of PA females, exposure of cultured hASCs and mouse preadipocytes to T also inhibits C/EBPα mRNA expression (23, 47). In contrast, however, to the inhibitory effect of T on PPARγ gene expression in hASCs (23), PPARγ mRNA levels were not affected in sc abdominal adipose of PA monkeys. These unexpected results challenge contemporary understanding that C/EBPα and PPARγ gene expression engage concurrently during late-stage preadipocyte differentiation to mature adipocytes. One possible explanation of the present monkey data might be that PA treatment epigenetically reprograms the androgen receptor-chromodomain helicase DNA-binding complex (48), leading to preferential inactivation of C/EBPα rather than PPARγ (49). Developmental programming of PA monkeys could also select for specific microRNAs critical for adipogenesis such as microRNA-31, which primarily down-regulates C/EBPα expression but only secondarily associates with PPARγ expression (50, 51). Whether developmental programming in PA monkeys affects the earliest stage of preadipocyte differentiation to adipocytes is unclear, because C/EBPβ, a critical gene involved during this process, was undetectable.

An important limitation of our study is its inability to distinguish the long-term effect of prenatal androgen exposure from the short-term action of high circulating androgen levels on sc abdominal adipose function, particularly because androgen inhibits sc abdominal lipoprotein lipase activity in women (52). As a result, this confounding action of circulating androgen levels on sc abdominal adipose function, apart from the earlier developmental programming effect of prenatal androgen exposure, may also vary with age, because the T to estradiol ratio in the circulation of late-reproductive-aged women increases across the perimenopausal transition (53). Moreover, visceral adipose volume increases in hyperandrogenic anovulatory women after pituitary desensitization to GnRH, whereas sc abdominal adipose does not (54), making extrapolation of androgen action on sc abdominal to visceral adipose unwarranted. Equally relevant, the negative impact of androgen on adipogenesis and lipid synthesis, although similar in direction by sex type, differs in magnitude between men and women (52), so that additional studies are required to understand the developmental consequences of prenatal androgen exposure in men on preadipocyte differentiation in sc abdominal adipose.

The collective data suggest that early-to-mid gestation exposure to T excess in adult female rhesus monkeys impairs preadipocyte differentiation and reduces adipocyte size in sc abdominal adipose. Dysfunctional adipocyte maturation may thus constrain the capacity of this adipose depot to safely store fat, promoting insulin resistance through lipotoxicity.