Distinctive Reproductive Phenotypes in Peripubertal Girls at Risk for Polycystic Ovary Syndrome

Abstract

Context

Increased testosterone (T) levels are a cardinal feature of polycystic ovary syndrome (PCOS). Female relatives of affected women, including premenarchal daughters, have elevated T levels supporting a genetic susceptibility to this phenotype. Girls with obesity (OB-g) also have increased T levels throughout puberty, which may indicate risk for PCOS.

Objective

We tested the hypothesis that premenarchal daughters of women affected with PCOS (PCOS-d) have distinctive phenotypic features compared with OB-g.

Design, Setting, and Participants

Forty-eight PCOS-d, 30 OB-g, and 22 normal weight (NW-g) premenarchal girls were studied. Mothers of OB-g and NW-g had no evidence for PCOS.

Main Outcome Measures

Reproductive hormones were measured.

Results

Body mass index differed by design, was highest in OB-g, followed by PCOS-d (P > 0.001). PCOS-d and OB-g had similar increases in free T levels compared with NW-g (PCOS-d vs NW-g, P = 0.01; OB-g vs NW-g, P = 0.0001). Sex hormone binding globulin levels were lowest in OB-g and lower in PCOS-d than in NW-g (PCOS-d vs NW-g, P = 0.005; OB-g vs NW-g, P < 0.0001; PCOS-d vs OB-g, P < 0.0001). Anti-Müllerian hormone (AMH) levels in PCOS-d were significantly increased compared with OB-g, who tended to have lower AMH levels than NW-g (PCOS-d vs OB-g, P < 0.0001; PCOS-d vs NW-g, P = 0.10).

Conclusions

Despite similarly elevated free T levels, PCOS-d had increased AMH levels compared with OB-g. This finding suggests that OB-g lack alterations in ovarian folliculogenesis, a key reproductive feature of PCOS. Causal mechanisms may differ in PCOS-d or OB-g, or elevated T in OB-g may not be an early marker for PCOS.

We found similar degrees of hyperandrogenemia in premenarchal daughters of women with PCOS and girls with obesity but differing anti-Müllerian hormone levels, suggesting ovarian folliculogenesis differs in these groups.

Hyperandrogenemia is a cardinal reproductive phenotype of polycystic ovary syndrome (PCOS) and may play a causal role in disease pathogenesis (1). Approximately 40% of reproductive-age sisters of affected women have elevated total or bioavailable testosterone (T) levels (1). Furthermore, male (2) as well as nonreproductive-age female [i.e., postmenopausal and premenarchal (3, 4)] first-degree relatives have hyperandrogenemia, suggesting a genetic susceptibility to this phenotype. In animal models, including nonhuman primates (5–7), androgen exposure in utero (5, 8), neonatally (9), or peripubertally (7) can result in phenocopies of PCOS. Accordingly, we hypothesized that genetic variation resulting in hyperandrogenemia during key developmental windows programs the phenotypic features of PCOS (4, 10). Consistent with this hypothesis, we found that daughters of women affected with PCOS (PCOS-d) have evidence of increased global 5α-reductase activity in early childhood (10) and elevated T levels beginning in early puberty (4). In addition to these early differences in androgen production and metabolism, PCOS-d exhibit other reproductive and metabolic phenotypes, including elevations of anti-Müllerian hormone (AMH) levels over the course of childhood, as well as early metabolic and β-cell dysfunction by the peripubertal years (4, 11, 12).

Girls with obesity (OB-g) also have increased total and free T levels during the pubertal transition (13). Because of the proposed central role of androgens in the pathogenesis of PCOS (14), it has been hypothesized that hyperandrogenemic OB-g will develop PCOS after puberty (15, 16). Thus, elevated T levels may be an early biomarker for PCOS (16). However, no longitudinal studies have been performed to test this hypothesis.

An important distinguishing feature between PCOS-d and OB-g is their differing genetic risk for PCOS. All mothers of PCOS-d are genetically affected by definition. Accordingly, a substantial percentage of PCOS-d will inherit one or more PCOS susceptibility variants (17–19) from their affected mothers (20). In contrast, the prevalence of PCOS susceptibility loci in OB-g and their mothers is expected to reflect the background 7% to 15% population prevalence of PCOS (21). Therefore, the mechanisms for early hyperandrogenemia and the risk for future development of PCOS may differ in these distinct groups. We performed this study to test the hypothesis that premenarchal reproductive phenotypes in OB-g differ from those in PCOS-d.

Methods

Premenarchal PCOS-d (n = 48), OB-g (n = 30), and normal weight control girls (n = 22) aged 8 to 12 years with Tanner breast stages I to III were studied. OB-g had the additional inclusion criterion of body mass index (BMI) ≥97th percentile, and normal weight girls (NW-g) had a criterion of BMI <85th percentile; PCOS-d with any BMI were included. All girls were in good health and were not taking any medications known to alter reproductive hormone metabolism or glucose homeostasis for at least 1 month before the study. PCOS-d had a mother who fulfilled National Institutes of Health criteria for PCOS [hyperandrogenism and oligomenorrhea with the exclusion of other reproductive disorders (21)] as confirmed by us before the current study or by their personal physician. Mothers of OB-g and NW-g had regular menses every 27 to 35 days as well as no history of reproductive disorders and no signs or symptoms of androgen excess by validated questionnaire (1). Participants were recruited through outreach to women who previously participated in our studies of PCOS; control adult women were recruited through letters sent to eligible patients at our medical centers, as well as by advertisements in local media and online.

PCOS-d, OB-g, and NW-g were studied at the Ann & Robert H. Lurie Children’s Hospital, and additional PCOS-d were studied at the Milton S. Hershey Medical Center at Penn State College of Medicine. The institutional review boards of the Feinberg School of Medicine, Northwestern University, Ann & Robert H. Lurie Children’s Hospital, and Penn State College of Medicine approved this study. Written informed consent was obtained from a parent of each girl, and written assent was obtained from each girl aged 12 years or older prior to participation. Clinical and biochemical data from some of the study subjects have been previously reported (4, 22).

A physical examination, including Tanner breast staging determined by visualization and palpation, was performed by a single pediatric endocrinologist at the Ann & Robert H. Lurie Children’s Hospital and by a trained study coordinator at the Hershey Medical Center at Penn State. A fasting early morning blood sample was collected for measurement of AMH, T, sex hormone binding globulin (SHBG), ultra-sensitive estradiol, dehydroepiandrosterone sulfate (DHEAS), and androstenedione levels.

Assays

Androstenedione, ultra-sensitive estradiol, AMH, SHBG, and DHEAS levels were measured at the University of Virginia Ligand Core Laboratory (Charlottesville, VA). Androstenedione was measured by using the Siemens Diagnostics (Los Angeles, CA) RIA system [sensitivity, 0.1 ng/mL; intra-assay coefficient of variation (CV), 4.9%; interassay CV, 7.0%], ultra-sensitive estradiol was measured using the Siemens Diagnostics RIA system (sensitivity, 10 pg/mL; intra-assay CV, 6.3%; interassay CV, 8.1%), and AMH was measured using the Beckman Coulter (Brea, CA) two-site ELISA System (sensitivity, 0.16 ng/mL; intra-assay CV, 3.9%; interassay CV, 6.2%) (23). SHBG and DHEAS levels were measured by the Siemens Diagnostics chemiluminescence system (SHBG sensitivity, 2 nmol/L; intra-assay CV, 2.7%; interassay CV, 5.2%; and DHEAS sensitivity, 150 ng/mL; intra-assay CV, 5.4%; interassay CV, 6.5%). T levels were analyzed by liquid chromatography-tandem mass spectrometry (Brigham Research Assay Core, Boston, MA) (sensitivity, 2 ng/dL; intra-assay CV, 9.0% at 16 ng/dL; interassay CV, 15.8% at 12 ng/dL) (12). Free T was calculated as reported (24).

Statistical analysis

Data were log-transformed or square root transformed when necessary to achieve homogeneity of variance. Pearson correlation was performed to assess the association of the potential confounders of age and Tanner breast stage with all end points. For variables that significantly correlated with age and/or Tanner breast stage, analysis of covariance adjusting for these parameters was applied. Otherwise, unadjusted ANOVA was used to assess differences in end point variables between the groups. Tukey post hoc testing was used to determine which groups differed significantly. Categorical variables were compared by Fisher’s exact test. We performed a subgroup analysis in PCOS-d with obesity and OB-g of comparable BMIs to control for the independent effect of obesity on major end points. To isolate a cohort of PCOS-d and OB-g with comparable BMIs, PCOS-d with the lowest BMI and OB-g with the highest BMI were excluded in a stepwise fashion until we achieved a cohort with comparable BMIs (n = 27; limited to participants with BMI z scores of 1.7 to 2.6). For this analysis, Student t tests were applied. Normative ranges for end point variables were defined by Tanner breast stage‒specific thresholds of ±2 SD in the l-g. Statistical analyses were performed with SAS 9.4 (SAS Institute, Inc., Cary, NC). Data are reported as mean ± SD, with the α level set at 0.05.

Results

BMI differed between the groups by design, being highest in OB-g (Table 1). Both OB-g and PCOS-d were heavier than NW-g (Table 1). There were minimal differences in age between the groups (P = 0.05), but with a trend toward younger age in PCOS-d than in OB-g and NW-g (Table 1). Consistent with their younger age, the percentage of girls with Tanner stage I and II breast (P = 0.01) and pubic hair (P = 0.01) assessments was higher in PCOS-d than in OB-g and NW-g (Table 1). The groups also varied by race and ethnicity, with a higher prevalence of Hispanic and black girls in the OB-g group than in the NW-g and PCOS-d groups (P < 0.0001).

Table 1.

Baseline Clinical Characteristics and Reproductive Hormones

	PCOS-d (n = 48)	OB-g (n = 30)	NW-g (n = 22)	P	PCOS-d vs OB-g	PCOS-d vs NW-g	OB-g vs NW-g
Age, y	9.7 ± 1.3a (8.0–12.3)	10.3 ± 1.1 (8.3–12.3)	10.3 ± 1.1 (8.3–12.7)	0.05	0.09	0.14	0.99
BMI percentile	80 ± 23 (16–99)	99 ± 1 (97–99)	50 ± 19 (19–84)	NA	NA	NA	NA
BMI z score	1.1 ± 0.9 (−1.0 to 2.5)	2.4 ± 0.2 (1.9–2.9)	0.0 ± 0.5 (−0.9 to 1.0)	<0.0001	<0.0001	<0.0001	<0.0001
Tanner breast stage	44% Ib	33% I	32% I	0.01c	NA	NA	NA
	40% II	13% II	27% II
	16% III	54% III	41% III
Tanner pubic hair stage	67% I	33% I	32% I	0.01c	NA	NA	NA
	29% II	33% II	45% II
	2% III	33% III	18% III
	2% IV		5% IV
Total T,d ng/dL	8 ± 6 (1–30)	9 ± 6 (1–22)	9 ± 8 (2–33)	0.53	NA	NA	NA
SHBG,e nmol/L	54 ± 31 (9–180)	22 ± 10 (7–54)	71 ± 22 (38–121)	<0.0001	<0.0001	0.005	<0.0001
Free T,f ng/dL	0.13 ± 0.12 (0.01–0.58)	0.22 ± 0.15 (0.03–0.58)	0.10 ± 0.10 (0.02–0.40)	0.0002	0.19	0.01	0.0001
AMH,g ng/mL	3.28 ± 2.34 (0.13–13.00)	1.22 ± 0.72 (0.20–3.03)	2.08 ± 0.94 (0.79–4.52)	<0.0001	<0.0001	0.10	0.07
DHEAS,h ng/mL	493 ± 358 (91–1730)	635 ± 349 (179–1555)	549 ± 362 (145–1770)	0.34	NA	NA	NA
Androstenedione,i ng/mL	0.47 ± 0.37 (0.13–1.27)	0.57 ± 0.37 (0.11–1.77)	0.50 ± 0.41 (0.11–1.78)	0.52	NA	NA	NA
Estradiol,j pg/mL	13 ± 11 (4–59)	20 ± 14 (7–76)	14 ± 10 (5–43)	0.09	NA	NA	NA

P values listed from analysis of covariance correcting for age and Tanner breast stage unless noted otherwise. BMI percentiles reported for descriptive purposes only. Statistical assessment for differences between the groups performed only on BMI z score. To convert DHEAS from ng/mL to nmol/L, multiply by 2.714; androstenedione from ng/dL to nmol/L, multiply by 0.0349; and estradiol from pg/mL to pmol/L, multiply by 3.671.

Abbreviation: NA, not applicable.

^aMean ± SD (range).

^bPercentage of subjects included in each group is noted.

^cThe categorical variables Tanner breast and pubic hair stages were analyzed by Fisher’s exact test.

^dTotal T levels missing in three PCOS-d.

^eSHBG levels missing in two PCOS-d.

^fFree T levels missing in four PCOS-d.

^gAMH levels missing in nine PCOS-d, seven OB-g, and four NW-g.

^hDHEAS levels missing in six PCOS-d and two NW-g.

ⁱAndrostenedione levels missing in 28 PCOS-d.

^jEstradiol levels missing in 12 PCOS-d, six OB-g, and four NW-g.

Total T levels did not differ between the groups (P = 0.53) (Fig. 1). SHBG levels were lowest in OB-g compared with PCOS-d (P < 0.0001) and NW-g (P < 0.0001) and were also lower in PCOS-d than in NW-g (P = 0.005) (Fig. 1). Free T levels were similarly increased in PCOS-d (P = 0.01) and in OB-g (P = 0.0001) compared with NW-g (Fig. 1). In total, 39% of PCOS-d and 38% of OB-g had elevated free T levels. DHEAS (P = 0.34) and androstenedione levels (P = 0.52) (Table 1) did not differ between the groups. Estradiol levels were also similar between the groups (P = 0.09) (Table 1) and were strongly correlated with Tanner breast stage (r: −0.51; P < 0.0001). AMH levels were significantly higher in PCOS-d than in OB-g (P < 0.0001); there was a trend toward higher AMH levels in PCOS-d compared with NW-g (P = 0.10) (Fig. 1). Thirty-one percent of PCOS-d with elevated AMH levels also had elevated free T levels.

Figure 1.

Reproductive hormone levels in PCOS-d and OB-g. Although total T levels did not differ between PCOS-d, OB-g, and NW-g (upper left panel, ANCOVA P = 0.53), PCOS-d and OB-g did have similarly decreased SHBG levels compared with NW-g (upper right panel, ANCOVA P < 0.0001; NW-g vs OB-g, P < 0.0001; PCOS-d vs OB-g, P < 0.0001; PCOS-d vs NW-g, P = 0.005). Accordingly, PCOS-d and OB-g had significant increases in free T levels compared with NW-g (lower left panel, ANCOVA P = 0.002; NW-g vs OB-g, P = 0.0001; PCOS-d vs OB-g, P = 0.19; and PCOS-d vs NW-g, P = 0.01). AMH levels also differed between the groups, with PCOS-d having significantly increased AMH levels compared with OB-g and a trend toward increased AMH levels compared with NW-g, which did not reach statistical significance (lower right panel, ANCOVA P < 0.0001; NW-g vs OB-g, P = 0.07; PCOS-d vs OB-g, P < 0.0001; and PCOS-d vs NW-g, P = 0.10). Unadjusted means and SEM are pictured; statistical analyses were adjusted for differences in age and Tanner breast stage. ANCOVA, analysis of covariance.

In the subgroup analysis in PCOS-d with obesity and OB-g, BMI z score (P = 0.08), age (P = 0.10), and Tanner breast stage (P = 0.29) did not differ. AMH levels (P = 0.03) remained increased in PCOS-d. SHBG levels remained lower in OB-g (P = 0.03), whereas free T levels tended to be higher in OB-g (P = 0.09).

Discussion

We found that peripubertal PCOS-d and OB-g have distinct reproductive phenotypes. Although their elevations in free T were similar to those in NW-g, AMH levels were significantly higher in PCOS-d than in OB-g. It is possible that obesity contributed to the lower AMH levels in OB-g, as a negative correlation between BMI and AMH level has been reported in some (25, 26), but not all (27), studies in adult women, including those with PCOS (28). However, the significant increase in AMH levels in PCOS-d compared with OB-g persisted in subgroups of comparable BMIs. The lack of elevated AMH levels in OB-g suggests that they do not have alterations in ovarian folliculogenesis characteristic of PCOS (29).

Small preantral and antral ovarian follicles up to 6 mm in size are the primary source of circulating AMH in women and girls (30, 31). Polycystic ovaries have a twofold to threefold increase in small preantral follicles (32, 33), and increased circulating AMH levels are a key feature of the PCOS reproductive phenotype (21, 26, 28, 34). Even before menarche, circulating AMH levels correlate with the number of small and medium ovarian follicles (35). Previous studies in PCOS-d found increased AMH levels during infancy, childhood (11), and puberty (36), suggesting differences in ovarian folliculogenesis may begin early in development in these girls.

Recent studies have suggested that AMH may play a role in the development of PCOS. We found that ∼3% of women with PCOS have mutations in AMH that decrease the bioactivity of the encoded protein; this may contribute to PCOS by decreasing AMH-mediated suppression of testosterone production (37). It is also possible that AMH plays a role in the pathogenesis of PCOS through neuroendocrine mechanisms. The AMH receptor is expressed in the hypothalamus on GnRH neurons in mice and humans (38). AMH increases GnRH-dependent LH secretion (38). Prenatal administration of high doses of AMH in pregnant mice induced GnRH-mediated hyperandrogenism and aromatase blockade in the mothers, resulting in intrauterine androgen exposure and ultimately a PCOS reproductive phenotype in their female offspring (39).

There is considerable evidence that androgens play a central role in the development of PCOS. Androgen excess during critical developmental windows produces phenocopies of PCOS in animal models (5–7). In humans, androgens antagonize the ability of estradiol and progesterone to slow the GnRH pulses contributing to the disordered gonadotropin secretion characteristic of PCOS (14). Hyperandrogenemia is a consistent reproductive phenotype in male as well as female relatives of women with PCOS (1, 2, 4). In the daughters of affected women, T elevations begin before menarche, and there is evidence for global increases in 5α-reductase in childhood, which could enhance the local conversion of T to its more potent metabolite dihydrotestosterone (10). Therefore, it is biologically plausible that girls with elevated free T levels will develop PCOS (40). However, no prospective studies have tested this hypothesis.

Our findings suggest that there may be additional phenotypic features in girls at increased genetic risk for PCOS (i.e., PCOS-d) because they have inherited one or more PCOS susceptibility alleles from their affected mothers (21). However, a subset of PCOS-d will not have inherited maternal PCOS susceptibility alleles. Accordingly, the PCOS-d cohort represents a mix of genetically affected and unaffected individuals. Nevertheless, the increased mean AMH levels in PCOS-d suggests that at least some of these girls have the changes in folliculogenesis characteristic of PCOS, in addition to elevated free T levels. Longitudinal studies are needed to determine whether these girls develop PCOS.

Our study had several limitations. Because our primary objective was to compare early reproductive phenotypes in two distinct putative PCOS risk groups, premenarchal PCOS-d and OB-g, BMI differed between our study groups by design. Differences in AMH levels remained significant in our subgroup analysis of OB-g and PCOS-d with obesity and comparable BMIs; however, we lacked adequate power to determine whether reproductive phenotypes differed in normal weight PCOS-d and NW-g of comparable BMIs. However, previous investigators reported a distinct reproductive phenotype of elevated AMH levels during infancy and childhood (11) and elevated total T levels by late puberty in normal weight PCOS-d compared with control girls of similar weight (12). Further, we did not have adequate power to perform separate analyses by individual Tanner stage. Nevertheless, we did account for the independent impact of pubertal stage by including Tanner breast stage as a covariate in our analyses. In addition, our subjects were not stratified by race or ethnicity because there was no evidence for racial or ethnic differences in circulating reproductive hormone levels, including AMH (41, 42) and androgen levels (43).

In summary, we have shown that despite similar increases in free T levels, premenarchal PCOS-d and OB-g have distinct reproductive phenotypes. In contrast to OB-g, PCOS-d have elevated AMH levels, a biomarker for altered follicular development. The distinct phenotype in PCOS-d may reflect increased genetic risk for PCOS compared with OB-g. It remains possible that both groups are at increased risk for PCOS by different causal mechanisms.

Glossary

Abbreviations:

AMH

anti-Müllerian hormone

BMI

body mass index

CV

coefficient of variation

DHEAS

dehydroepiandrosterone sulfate

NW-g

normal weight girls

OB-g

girls with obesity

PCOS

polycystic ovary syndrome

PCOS-d

daughters of women affected with PCOS

SHBG

sex hormone binding globulin

T

testosterone