Prenatal Androgen Excess Negatively Impacts Body Fat Distribution in a Nonhuman Primate Model of Polycystic Ovary Syndrome (PCOS)

Abstract

Introduction

Prenatally androgenized (PA) female rhesus monkeys share metabolic abnormalities in common with PCOS women. Early gestation exposure (E) results in insulin resistance, impaired pancreatic beta-cell function and type 2 diabetes, while late gestation exposure (L) results in supranormal insulin sensitivity that declines with increasing body mass index (BMI).

Objective

To determine whether PA females have altered body fat distribution.

Design

Five EPA, 5 LPA, and 5 control adult female monkeys underwent somatometrics, dual x-ray absorptiometry (DXA) and abdominal computed tomography (CT). Five control and 5 EPA females underwent an intravenous glucose tolerance test to assess the relationship between body composition and glucoregulation.

Results

There were no differences in age, weight, BMI, or somatometrics. LPA females had ∼20% greater DXA-determined total fat and percent body fat, as well as total and percent abdominal fat than EPA or control females (p≤0.05). LPA females also had ∼40% more CT-determined non-visceral abdominal fat than EPA or control females (p≤0.05). The volume of visceral fat was similar among the 3 groups. EPA (R²=0.94, p≤0.01) and LPA (R²=0.53, p=0.16) females had a positive relationship between visceral fat and BMI, although not significant for LPA females. Conversely, control females had a positive relationship between non-visceral fat and BMI (R²=0.98, p≤0.001). There was a positive relationship between basal insulin and total body (R²=0.95, p≤0.007), total abdominal (R²=0.81, p≤0.04), and visceral (R²=0.82, p≤0.03) fat quantities in EPA, but not control females.

Conclusions

Prenatal androgenization in female rhesus monkeys induces adiposity-dependent visceral fat accumulation, and late gestation androgenization causes increased total body and non-visceral fat mass. Early gestation androgenization induces visceral fat-dependent hyperinsulinemia. The relationship between the timing of prenatal androgen exposure and body composition phenotypes in this nonhuman primate model for PCOS may provide insight into the heterogeneity of metabolic defects found in PCOS women.

INTRODUCTION

Prenatally androgenized (PA) female rhesus monkeys, a nonhuman primate model of polycystic ovary syndrome (PCOS), exhibit metabolic abnormalities similar to those found in PCOS women (1-3). Female monkeys exposed to testosterone excess in early gestation (early-treated) exhibit insulin resistance, impaired pancreatic beta-cell function and type 2 diabetes, while those exposed in late gestation (late-treated) develop supranormal insulin sensitivity that declines with increasing body mass index (BMI) (3-5).

Women with PCOS have increased upper body obesity and visceral fat, independent of total body mass (6-9), similar to findings in PA female rhesus monkeys exposed to testosterone excess during early gestation (10). Previous studies in rhesus monkeys indicate that increased abdominal adiposity may be a potential mechanism for the observed metabolic abnormalities in early-treated PA females (3, 10). Whether late-treated females harbor similar alterations in body composition is unknown. Moreover, the effect of altered body composition on glucoregulation in PA females has not been elucidated.

The purposes of this study were to examine body composition using dual x-ray absorptiometry (DXA) and total abdominal computed tomography (CT) and to assess the relationship of body fat distribution with glucoregulatory dynamics.

MATERIALS AND METHODS

Animals

Fifteen adult female rhesus monkeys (Macaca mulatta) were used in this study and were maintained at the National Primate Research Center, University of Wisconsin, Madison (WNPRC) according to standard protocol (11). Animals were fed Purina monkey chow (product no. 5038, Ralston Purina, St. Louis, MO) with occasional supplementation of fresh fruits. This formulation of monkey chow provides 70% of calories as carbohydrate, 13% as fat, and 17% as protein. The animal care and use committee of the Graduate School of University of Wisconsin (Madison, WI) approved all experiments and animal protocols. Recommendations of the Guide for the Care and Use of Laboratory Animals and the Animal Welfare Act with its subsequent amendments were followed.

PA female rhesus monkeys were developed as previously reported (12). Ten PA females were produced by subcutaneously injecting pregnant rhesus monkeys with 10 mg of testosterone propionate (TP) for 15-35 consecutive days. TP was initiated on either day 40 (early-treated; n=5) or day 100-118 (late-treated; n=5) of gestation (gestation=165 days). The control group was comprised of five unexposed females of similar age, weight and body mass index (BMI; 13) to both PA female monkey groups (Table 1). Intervals between menstrual cycles tended (p<0.09) to be greater in early-treated PA compared to control females (∼80% decrease in cycle frequency in early-treated PA females). Baseline (early follicular phase or anovulatory period) serum androgen levels were comparable between female groups (Table 1). As found in this study, adult androgen excess is not always demonstrable in PA monkeys (14, 15) if repeated sampling (16) or ovarian endocrine challenges (5) are not performed.

Table 1.

Baseline characteristics of control, early-treated and late-treated PA female rhesus monkeys

Parameter	Control females (n=5)	Prenatally androgenized females
		Early treated (n=5)	Late treated (n=5)
Age (years)	18.4±0.9	21.2±0.2	19.0±1.0
Body weight (kg)	8.3±0.3	8.6±0.6	8.6±0.3
Body mass index (kg/m2)	36.3±1.6	37.6±1.3	36.1±0.8
Menstrual cycle  duration (days)	32.2±2.2	59.0±21.4a	42.7±3.7
Baseline serum androgen  parameters:
Testosterone (pg/ml)	360.0±99.5	360.0±101.5	240.0±75.7
Androstenedione (pg/ml)	251.7±43.9	303.4±43.8	239.8±56.6

^ap<0.09 versus controls

Measures of Body Composition

Somatometrics

Animals were anesthetized with an intramuscular (im) injection of ketamine hydrochloride (7 mg/kg) and xylazine (0.6 mg/kg). Xylazine was partially reversed with yohimbine (0.06 mg/kg iv). Somatometric measurements were performed immediately before DXA scans. Measurements of body weight, body length, body and limb circumference, and skinfold thickness were performed by a single operator (STB). All assessments were performed with the animal in a supine position with the exception of abdominal and chest circumferences, which were accomplished with the animal in a right lateral recumbent position. Skinfold thickness was determined using a Lange caliper, as previously reported (10). BMI was calculated as body weight (kg) divided by the square of the crown rump length (m²) (17).

Dual x-ray absorptiometry

Dual x-ray absorptiometry (DXA) was performed on anesthetized animals immediately after somatometric measurements. DXA has a precision (CV%) of <4% for total body and regional composition in rhesus monkeys (18). Animals were placed in a supine position for scanning with DXA scanner model DPX-L (GE/Lunar Corp., Madison WI). 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 (18). Each scan lasted ∼15 minutes. The abdominal region of interest was defined by area between the thoracic (T) 12-lumbar (L) 1 interspace and the L6-L7 interspace. To compare CT and DXA adipose volumes, DXA-derived adipose mass was converted to volume by dividing by the average density of triglyceride (0.9 kg/L).

Computed Tomography

Computed Tomography (CT) was performed within 7.7 ± 6.0 days of the DXA scan. Animals were transported to the University of Wisconsin Veterinary School and were anesthetized with one of three anesthesia protocols (protocol 1: ketamine hydrochloride 10mg/kg, atropine 0.4-1mg, and xylazine 5mg im, followed by reversal with yohimbine 1mg im; protocol 2: ketamine 100mg, xylazine 2-5mg, and atropine 0.16-1mg im followed by reversal with atipamezole 2.5mg im; protocol 3: ketamine 10mg/kg and xylazine 5-10 mg im followed by reversal with yohimbine 0.5-0.8mg iv). Two animals in protocol 3 required an additional dose of ketamine (20mg) or acepromazine (2mg) administered intravenously (iv) to maintain sedation. The abdomen was scanned using a HiLight Advantage scanner (GE, Brookfield, WI) at 120 kV and 200 mA with a scan time of 2 seconds and slice thickness of 0.5cm. The images were converted to compact disc from optical disk format tape and were stored until analysis. The abdominal region was analyzed using SliceOMatic (version 4.3 rev0h, Tomovision, Montreal, Canada). The analyses were performed by a single operator (CMB), blinded to animal identity. A computer-generated region of interest containing 45 pixels of adipose tissue alone was placed in 3 regions (anterior subcutaneous, anterior intraperitoneal and left perinephric gutter) of the L4-5 slice. The mean and variance of the Hounsfield units (HUs) in these three regions were recorded. The standard deviations and means from these three regions were averaged together respectively to yield an average HU mean and standard deviation. To estimate the lower limit of adipose tissue density, 20 standard deviations were added to the mean HU. Using this method, the average upper limit of adipose tissue density was -46 ± 1.5 HU (mean ± SEM) and was similar in the three groups (early-treated PA: -45 ± 2.6, late-treated PA: -50 ± 1.5, C: -44 ± 3.2). The HU value calculated for the upper limit of adipose tissue by this method visually corresponded to slightly less than the nadir of the distribution overlap between nonadipose tissue HU and adipose tissue HU. The HU value selected for the upper limit of adipose tissue density was set at -200. Each abdominal slice between T12-L1 and the top of the sacrum was analyzed for total and visceral fat. First, visceral fat area was assessed by “painting” the abdominal region inside the peritoneum with a mouse-driven cursor, while excluding colon contents and adipose tissue posterior to the longitudinal axis of the kidneys and great vessels. The adipose regions posterior to the kidneys and great vessels and outside the peritoneal lining was then painted and the surface area was recorded as the total adipose tissue. The surface areas of each slice were multiplied by the slice thickness (0.5 cm) to yield a volume of adipose tissue. Non-visceral (or subcutaneous) adipose was calculated as the difference between the total and visceral compartments. Between-group differences were assessed based on the adipose volume between the T12-L1 interspace and the top of the sacrum. The adipose volume from T12-L1 through L6-7 assess by CT was used to correlate abdominal adipose volume with the same abdominal region assessed by DXA. Eighteen CT images were re-analyzed by the same operator (CMB) for both total and visceral adipose area to provide an estimate of within-operator variability. CT measurements of total and visceral fat were highly reproducible (R²=0.99, p≤0.0001 and R²=0.99, p≤0.0001, respectively). Abdominal adipose volume measurements by CT were highly correlated with DXA measurements (y=0.7+156.4, R²=0.92, p≤0.0001).

Frequently-sampled intravenous glucose tolerance test

Early PA and C females underwent a single, three-hour, frequently-sampled intravenous glucose tolerance test (FSIGT) starting at 0700 h to 0900 h, according to the tolbutamide-modified minimal model of Bergman (Eisner, et al. 2000). FSIGTs were performed within 24.1 ± 5.2 days from the DXA scan and 25.7 ± 5.1 days from the CT scan. Animals had not been anesthetized for other procedures within the previous 4 weeks. Animals were fasted overnight and were anesthetized with ketamine hydrochloride (15mg/kg body weight, im) and diazepam (1-1.25 mg/kg body weight, im). Additional ketamine (5-10mg/kg, im) and acepromazine (2mg, iv) were given as needed to maintain sedation. A central iv catheter was placed for blood sampling and administration of 50% dextrose (300mg/kg body weight) at 0 min and tolbutamide (5mg/kg body weight, Orinase Diagnostic, Upjohn, Kalamazoo, MI) at 20 min. Since DXA and CT scans on the late-treated PAs were performed up to 491 days from glucose and insulin determinations, these glucose and insulin values were excluded from the analyses.

Assay procedures

Plasma insulin concentrations were measured by RIA as previously reported (Linco Research, Inc, St. Louis, MO) (3). Plasma glucose concentrations were measured by the glucose oxidase method (Yellow Springs Instruments, Yellow Springs, OH). Inter- and intra-assay coefficients of variation for insulin were 8.1% and 4.8%, respectively. Inter-and intra-assay coefficients of variation for glucose were 4.0% and 2.9%, respectively.

Statistical methods

Summary measures derived from FSIGT

The tolbutamide-modified minimal model (version 3.0, R.N. Bergman) was used to generate estimates of insulin sensitivity (S_I) and glucose effectiveness (S_G). S_I reflects the ability of insulin to promote glucose uptake and to inhibit hepatic glucose production. S_G reflects insulin-independent glucose uptake and suppression of hepatic glucose production. Further measures derived from the FSIGT included basal glucose (Gb; mean of the four prechallenge glucose values, -15, -10, -5, and -2 min), basal insulin (Ib; mean of the four prechallenge insulin values), acute insulin response to glucose (AIRg; average increase in insulin above basal at 2-4 minutes, post challenge), and disposition index (DI; beta-cell compensation index; product of S_I and AIRg).

Statistical Analyses

Baseline characteristics in Table 1 were analyzed by ANOVA (Systat v.5.2.1, Evanston, IL) except for menstrual cycle duration. The latter analysis was performed using Kruskal Wallis because of lack of normality in the data (19). All body composition variables were compared using analysis of covariance (ANCOVA; Systat Version 5.2.1) with type of female (early-treated PA, late-treated PA, control) as a factor and BMI as a covariate after homogeneity of slopes was confirmed. There was an interaction between BMI and nonvisceral fat by CT scan (p<0.04) excluding ANCOVA for the latter parameter. Post-hoc analyses, using Fisher’s LSD tests, were performed when the overall ANCOVA p-value was ≤ 0.05. Plots of residuals showed a homogeneous variance across the range of BMI. Multiple linear regression analysis was used to assess the relationships between body composition variables and FSIGT variables. The correlation between CT and DXA measurements of abdominal fat was assessed by linear regression. Intra-observer variability was examined using linear regression. A p-value ≤ 0.05 was considered significant. Data are presented as mean ± SEM.

RESULTS

Somatometrics

There were no significant between-group differences in abdominal circumference or skinfold thickness measurements (data not shown).

DXA

Total and percent body fat mass, as well as abdominal and percent abdominal fat mass, as determined by DXA, were greater in late-treated PA compared to early-treated PA and control females (p≤0.05; Table 2). Total and abdominal lean mass were similar for all three groups, although late-treated PA females had a lower proportion of total body and abdominal lean mass than early-treated PA and control females (p≤0.05; Table 2).

TABLE 2.

Body composition of early prenatally androgenized, late prenatally androgenized and control females

Body Composition Parameter	EPA	LPA	C	p value
	(n=5)	(n=5)	(n=5)
DXA-derived
Total body fat (g)a	2023.75±152.43	2647.50±150.53b, c	1844.77±149.58	0.008
% Total body fat	24.46±1.83	30.30±1.81b, c	21.86±1.80	0.019
Total body lean mass (g)	6402.52±302.56	6184.57±298.80	6602.91±296.90	0.621
% Total body lean mass	75.73±1.82	70.10±1.80b, c	78.36±1.79	0.022
Total abdominal fat (g)	983.15±89.27	1319.08±88.16b, c	940.57±87.60	0.022
% Total abdominal fat	36.28±2.42	44.68±2.39	33.66±2.37	0.019
Total abdominal lean mass(g)	1764.78±118.71	1668.42±117.23b, c	1811.79±116.49	0.684
% Total abdominal lean mass	63.74±2.46	55.80±2.43b, c	66.34±2.42	0.027
CT-derived
Total abdominal fat (cm3)	1042.96±85.81	1405.62±84.74b, c	1014.80±84.20	0.013
Visceral fat (cm3)	679.46±73.88	786.94±72.96	650.29±72.50	0.407
% Visceral fat	65.08±4.67	55.29±4.61	64.62±4.58	0.282
Non-visceral fat (cm3)	363.51±62.64	618.68±61.86b, c	364.51±61.47	0.021
% Non-visceral fat	34.92±4.67	44.71±4.61	35.38±4.58	0.282

EPA: Early-treated PA

LPA: Late-treated PA

C: Control

^aValues are the mean ± SEM

^bp≤0.01 vs. C

^cp≤0.05 vs. EPA

CT

CT-determined volumes of visceral fat were similar for all three female groups (Table 2). Late-treated PA females exhibited a larger volume of non-visceral (subcutaneous) abdominal fat than early-treated PA or control females (p≤0.05; Table 2).

Association between CT-determined body composition and BMI

Early-treated PA females preferentially deposit visceral fat at BMIs above the mean for normal females (Figure 1A). A similar, parallel relationship is exhibited in late-treated PA females, though not statistically significant (p=0.16). If, however, the late-treated PA female with excessive visceral fat (929cm³) relative to her BMI (∼35kg/m²; Figure 1A) is removed from the analysis, the late-treated PA females exhibit (p<0.008) an almost identical positive relationship (r²=0.98; y=89×-2515) between visceral fat and BMI as found in early-treated PA female monkeys. In contrast, control females do not significantly accumulate visceral fat with increasing BMI, but instead accumulate non-visceral fat with increasing BMI (Figure 1B), unlike both PA female groups.

FIGURE 1.

(A) Visceral fat is positively correlated with BMI in early-treated PA (dotted line, y=75x-2091, R²=0.94, p≤0.01), but not late-treated PA (dashed line, y=74x-1893, R²=0.53, p=0.16) or C (solid line, y=6x+437, R²=0.02, p=0.84) females. If, however, the data from a late-treated PA female monkey with relatively excessive visceral fat accumulation (929cm³) relative to BMI (∼35kg/m²) is removed, then visceral fat is positively correlated with BMI in late-treated PA females (y=89x-2515, R²=0.98, p≤0.008). (B) Non-visceral fat is positively correlated with BMI in C (solid line, y=45x-1258, R²=0.98, p≤0.001), but not early-treated PA (dotted line, y=-29x+1452, R²=0.26, p=0.38) or late-treated PA (dashed line, y=15x+76, R²=0.10, p=0.60) females.

C: Control

EPA: Early-treated PA

LPA: Late-treated PA

Correlation between measures derived from basal glucose and insulin and DXA- and CT-determined estimates of body composition

Basal serum insulin levels in early-treated PA females were positively correlated with total body (Figure 2A), total abdominal (Figure 2B) and visceral (Figure 2C) fat quantities, while similar relationships between Ib and body composition in control females did not occur. The associations between basal insulin and body fat parameters among early-treated PA females were reexamined using a rank-based method (Spearman Rank-Order Correlation) that is robust to outliers. In this analysis, the strength of these associations persisted (basal insulin vs. total body fat, r_s=0.90, p=0.04; vs. total abdominal fat, r_s=1.00, p<0.0001; vs. visceral fat, r_s=0.80, p=0.10). There were no statistically significant relationships between S_G, S_I, DI, Gb, or AIRg and body composition variables in the three female groups (data not shown).

FIGURE 2.

In early-treated PA (dotted line), but not C females (sold line), basal insulin (Ib) is positively correlated with total body (A: y=0.09x-148.91, R²=0.95, p≤0.007), total abdominal (B: y=0.15x-120.53, R²=0.81, p≤0.04), and visceral fat quantities (C: y=0.161x-66.86, R²=0.82, p≤0.03).

C: Control

EPA: Early-treated PA

^aTo convert to picomolar concentrations, multiply by 6.945.

Validation of a single-slice determination of abdominal fat compartments

The amount and proportion of visceral fat volumes in the single L4-5 abdominal slice were highly correlated with the total visceral fat volume of the entire abdomen (amount: y=0.03×+1.88, R²=0.86, p≤0.001, proportion: y=1.13×-7.85, R²=0.96, p≤0.001). The product of percent visceral fat in the L4-5 slice and DXA-derived total abdominal fat also was highly correlated with the CT-derived total abdominal visceral fat volume (y=1.13×-65.98, R²=0.91, p≤0.001). Analysis of covariance utilizing this combined CT- and DXA-derived measure yielded results similar to that using the direct measurement of total abdominal visceral fat by CT (p≤0.05). Similarly, the product of the percent non-visceral fat in the L4-5 slice and the DXA-derived total abdominal fat were highly correlated with the CT-derived total abdominal nonvisceral fat (p≤0.05).

Discussion

This study confirms and extends our previous body composition study of PA female rhesus monkeys (10) showing preferential accumulation of abdominal visceral fat in early-treated PA female monkeys, further demonstrating that preferential visceral fat accumulation depends upon increasing BMI. The lack of difference in absolute fat accumulation (regardless of BMI) between early-treated PA females and controls in our current study versus our previous one (10) can be explained by the narrower BMI range of the animals in the present study (35.5 to 42.0 kg/m²; average BMI: 37.6 kg/m²) compared to the earlier study (28.7 to 49.7 kg/m²; average BMI: 39.3 kg/m²) (10). Given preferential accumulation of visceral fat with increasing BMI in early-treated PA females, we might not have noted an increase in visceral fat in the early-treated PA group because their BMI range was lower than previously reported (10). Further, in support of prenatal androgen exposure reprogramming BMI-dependent visceral fat accumulation, LPA females demonstrate a similar positive relationship to that shown by early-treated PA females. In this latter regard, while data on sex differences in visceral fat accumulation in rhesus monkeys are lacking, sex differences in visceral fat accumulation in humans would be consistent with the notion of prenatal androgen programming (20). Men demonstrate a positive relationship between visceral fat and BMI throughout the adult male BMI range. In contrast, women only do so when BMI reaches overweight or obese categories (20). Since maximal BMI values in our control females reach just over 40 kg/m² (Figure 1A), well short of the obese range for female rhesus monkeys in our colony (21), the lack of association between visceral fat and BMI in control females in this study would be consistent with the absence of obesity.

Early-treated PA females also exhibit a significant correlation between Ib and total body fat, total abdominal fat, and visceral fat, agreeing with data from PCOS women showing an increase in visceral fat with fasting serum insulin levels (8). Whether inter-individual differences among early- and late-treated PA females may reflect differences in fetal exposure or sensitivity to gestational androgen excess (5), genotypic differences in androgen receptor (22), or androgen biosynthetic enzyme expression (23) remains to be determined.

Body composition analyses by DXA and CT indicate that late-treated PA female rhesus monkeys have increased total body fat, percent total body fat, total abdominal fat, and non-visceral (subcutaneous) abdominal fat compared to early-treated and control monkeys. In other words, visceral and nonvisceral abdominal fat accumulations are different metabolic consequences of early and late-gestation exposure to androgen excess, respectively. Both gestational treatments induce preferential accumulation of abdominal adiposity. They just vary in the intra-abdominal site at which the accumulation occurs. Interestingly, early-gestation testosterone programming results in additional adiposity-dependent glucoregulatory dysfunction, indicated by correlations between Ib and total body fat, total abdominal fat, and visceral fat. In addition, early androgenization results in impaired pancreatic beta cell function (3) and increased incidence of type 2 diabetes (4). We do not yet have enough data to conclude if late gestation testosterone excess affects glucoregulation, but we have previously reported a negative relationship between S_I and BMI in this female group (3). Such differences in body fat distribution, based upon timing of gestational testosterone excess, extend our understanding of PCOS phenotypic variation from differential exposure to prenatal androgen excess (5) and could additionally contribute to the heterogeneity in metabolic consequences of PCOS.

The association of accumulated upper body fat with metabolic complications (including insulin resistance) is widely documented in humans (reviewed in 24), and is an important clinical characteristic of women with PCOS (reviewed in 25). The increased insulin resistance in PCOS women can be accounted for by increased abdominal fat (26). Even though insulin resistance in PCOS women can be independent of obesity, obesity and PCOS act synergistically to adversely affect glucose homeostasis (27). In these regards, our current findings that early gestation androgen excess programs both hyperinsulinemia from adiposity-dependent insulin resistance and preferential accumulation of visceral adiposity in the same animals, (5, 10, 28), strongly implicate fetal programming of metabolic dysfunction in our nonhuman primate model for PCOS and suggest a fetal origin for this syndrome.