Fetal androgen excess provides a developmental origin for polycystic ovary syndrome

Our 14-year study of a nonhuman primate model for polycystic ovary syndrome (PCOS) has provided substantial, experimentally derived evidence for fetal origins of PCOS that leads to lifelong reproductive and metabolic consequences [1–4]. Fetal androgen excess programming of PCOS-like pathophysiology is well established across the mammalian order, including rodents (mice and rats [5,6]), ungulates (sheep [7]) and higher primates (rhesus monkeys [8]), with such prenatally androgenized females displaying an increased prevalence of at least two diagnostic criteria for PCOS in women.

Our focus on the developmental origins of PCOS begins with the search for its elusive etiology, since without such knowledge physicians can only ameliorate adult signs and symptoms of PCOS, without establishing a pre-emptive intervention that could lead to an effective cure (comprehensively discussed in a recent issue of Expert Review of Obstetrics and Gynecology[9]). We consider here the case for experimentally induced or naturally occurring fetal androgen excess origins of PCOS, focusing mainly on higher primates as they most closely resemble humans in terms of reproduction, metabolic physiology and aging, and exhibit PCOS-like traits spontaneously [10], as well as following experimentally induced fetal androgen excess [2].

Polycystic ovary syndrome in women

Polycystic ovary syndrome is a complex, heterogeneous health problem with reproductive and metabolic consequences, afflicting approximately 10% of women in their reproductive years [11]. It is the most prevalent female endocrinopathy and increases the lifetime risk of Type 2 diabetes, cardiovascular disease and endometrial cancer [12–14]. The current diagnostic criteria, established by the 2003 Rotterdam consensus conference [15], are indicative of the heterogeneity of PCOS and the continuing refinement of its clinical definition [16]. A PCOS diagnosis requires the presence of two out of three characteristic signs and symptoms:

• • Criterion 1: clinical and/or biochemical hyperandrogenism

• • Criterion 2: intermittent or absent menstrual cycles

• • Criterion 3: polycystic ovaries

This is excluding other related or similar-appearing conditions that include classical and nonclassical congenital adrenal hyperplasia, Cushing’s syndrome, androgen secreting tumors, hyperprolactinemia and hyperthyroidism.

Such diagnostic criteria generate four distinct PCOS phenotypes [17]:

• • Phenotype 1: severe PCOS (all three criteria)

• • Phenotype 2: hyperandrogenism and chronic anovulation (criteria 1 and 2)

• • Phenotype 3: ovulatory PCOS (criteria 1 and 3)

• • Phenotype 4: mild PCOS (criteria 2 and 3)

Despite such phenotypic diversity, approximately 75% of PCOS women manifest the two most severe phenotypes (phenotypes 1 and 2), with the metabolic syndrome particularly prevalent among PCOS women with such pronounced hyperandrogenic phenotypes [4]. On the other hand, the recent Androgen Excess Society refinement of the Rotterdam criteria [16] concludes that PCOS is principally a hyperandrogenic disorder, thus, removing the mild PCOS phenotype (phenotype 4) from diagnostic consideration. Our findings from the nonhuman primate model also suggest that limitation of PCOS to a hyperandrogenic disorder is highly appropriate, since we and others have repeatedly shown the intimate involvement of androgen excess in the developmental origins and persistence of PCOS in female mammals [2,18].

Dysfunction of reproductive neuroendocrinology accompanies PCOS in approximately 90% of cases [19]. Increased frequency of episodic release of luteinizing hormone (LH) from pituitary gonadotropes in such PCOS women most likely reflects increased hypothalamic gonadotropin-releasing hormone pulsatile release from reduced steroid hormone negative feedback on LH secretion mediated through androgen receptor action [20]. While LH hypersecretion may enhance ovarian hyperandrogenism in PCOS women, the PCOS ovary is, nevertheless, intrinsically hyperandrogenic [21,22].

Polycystic ovary syndrome is highly familial, with recent twin [23] and genetic [24] studies implicating a single genetic component underlying its etiology. All but one potential gene candidate, however, have failed confirmation of an association with PCOS. The currently viable gene candidate is the D19S884 allele 8 that maps to chromosome 19p13.2 within intron 55 of the fibrillin-3 gene, a member of the TGF-β superfamily encoding for extracellular matrix proteins [24]. These proteins play critical roles in ovarian follicle [25] and testicular [26] development and androgen exposure increases their expression [27,28]. Fetal androgen excess could, thus, potentially enhance TGF-β-regulated extracellular matrix protein production and, in doing so, sufficiently disrupt ovarian differentiation, causing a polycystic phenotype.

In this regard, since TGF-β, SMAD3 (one of its intracellular signaling proteins) and other TGF-β family members (including antimullerian hormone) regulate the expression of the key androgen biosynthetic enzyme, CYP17 [29,30], there is the added possibility that altered TGF-β action could perturb ovarian or adrenal androgen biosynthesis. Such interactions between TGF-β superfamily members and androgen biosynthesis are particularly noteworthy since serum and follicular fluid levels of AMH are consistently elevated in women with PCOS [31]. Given such TGF-β-mediated regulation of androgen biosynthesis, allele 8’s association with PCOS could reflect either its primary role in inducing a hyperandrogenic polycystic ovary or its secondary role in responding to a hyperandrogenic intrauterine environment. Regardless, both scenarios provide a pathophysiological basis for the developmental origins of PCOS that can have genetic or epigenetic origins. The latter would be consistent with our reliable ability to employ fetal androgen excess to generate phenocopies of PCOS in animal models [3,8,18].

Association of fetal androgen excess with PCOS in women

While little is known concerning the intrauterine endocrine environment during pregnancy in PCOS women, there is an increased likelihood of complications, such as gestational diabetes and preeclampsia [32], that compromise nutrition of the fetus and alter development of the child [9]. Convincing circumstantial evidence suggests fetal androgen excess programming of PCOS in women. During the second trimester of pregnancy in primates, when the fetal ovary transiently expresses the key androgen biosynthetic enzyme, CYP17 [33] and produces androgens [34], approximately 50% of umbilical vein blood levels of unbound testosterone in human fetal females from otherwise ‘normal’ pregnancies are elevated into the fetal male range [35]. Since the PCOS ovary is hyperandrogenic from enhanced activity of multiple steroidogenic enzymes, including CYP17, and their regulatory elements [21,22], its excessive capacity to produce androgens may well be expressed during mid-gestation. Thus during the second trimester of fetal development, a hyperandrogenic primate ovary could thereby alter the development of multiple target tissues expressing androgen receptors, including the ovary, hypothalamus, pancreas, liver and kidney.

Further circumstantial evidence for fetal androgen excess programming of PCOS arises from human female fetuses exposed to gestational androgen excess from fetal adrenal deficiency of 21-hydroxylase or other enzymes involving cortisol biosynthesis (classical, early-onset congenital adrenal hyperplasia [CAH]). There is an increased incidence (25–50%) of PCOS in women with early-onset CAH compared with the approximately 10% incidence of PCOS found in women without CAH [36]. In addition, a female fetus with an androgen-secreting tumor removed shortly after birth developed PCOS in adulthood [37], suggesting that fetal androgen excess per se is crucial for the expression of adult PCOS-like traits, regardless of its source.

Lack of genital virilization in PCOS individuals compared with women with early-onset CAH agrees with an ovarian, rather than an adrenal, source of fetal androgen excess in PCOS, beginning in the second trimester of fetal development. This is because the fetal primate adrenal cortex produces androgens during the first trimester (when the immature reproductive tract and external genitalia express androgen receptors), but adrenocorticotropic hormone-driven adrenal androgenic output is normally minimized by cortisol negative feedback to safeguard female sexual development [38]. By contrast, the fetal ovary produces androgens during the second trimester when the female reproductive tract and external genitalia lose androgen receptor expression, except for the clitoris [39]. Such a scenario corresponds with the findings of PCOS-like traits in female monkeys with normal external genitalia or virilized genitalia following exposure to androgen excess during late gestation or during the transition between first and second trimesters, respectively [2].

To date, however, there is no convincing evidence that gestational maternal hyperandrogenism in PCOS pregnancies [40] contributes to fetal androgen excess per se. As in normal pregnancies, increased maternal androgen levels during pregnancy in PCOS women usually occur without fetal virilization. The human (and higher primate) placenta has an extensive capacity to inactivate and aromatize androgens that exceeds greatly that found in non-primate mammals [41]. Therefore, unless placenta function is compromised substantially in PCOS women (as it is in women with aromatase deficiency [42] or undernutrition [43]), or unless pregnant PCOS women have profoundly diminished sex hormone-binding globulin (thereby elevating circulating free testosterone levels [41]), female fetuses are protected from hyperandrogenic excursions in the circulation of PCOS mothers.

Pregnancy following experimentally induced gestational androgen excess & in PCOS women

Fetal androgen excess in female monkeys and other female mammals has been conventionally considered as engaging masculinization and/or defeminization processes. PCOS-like phenotypes are induced in monkeys by injecting dams with testosterone propionate 10–15 mg for 15–41 days starting on either gestation days 40–60 (early gestation exposure) or days 100–155 (late gestation exposure) that elevate circulating testosterone levels in fetal females to those found in the low-normal range of fetal males [3]. It is now clear that such in utero endocrine environments disrupt development and function of multiple female organ systems [2].

Fetal androgen excess in nonprimate female mammals, in addition to inducing PCOS-like traits in adulthood, also causes fetal growth restriction [44,45]. Such fetal impairment implicates placental insufficiency leading to low birth weight, with subsequent postnatal catch-up growth [45]. In primates, fetal growth restriction is not found in females exposed to intrauterine androgen excess [2]. The same holds true in most PCOS pregnancies [32,43,46,47], except for two PCOS populations of Spanish descent [9,48]. Testosterone-mediated androgen excess in fetal female primates does not elevate circulating estrogen levels in the fetus or dam, probably owing to the rapid bioinactivation of estrogens by the primate liver and placenta [3].

Metabolic mechanisms of fetal developmental programming may also exist. For example, prenatally testosterone-treated male monkeys exhibit insulin resistance and diminished insulin response to glucose in adulthood [49], despite normal male levels of circulating testosterone during fetal life. Therefore, the combination of steroid and metabolic abnormalities in utero might perturb development of several fetal organ systems and increase the risk of developing reproductive and metabolic diseases in later life regardless of the sex of the fetus. In this regard, it is interesting to note that the most profound phenotypes in male close relatives of women with PCOS manifest metabolic dysfunction [50,51]. Consistent with this notion of metabolic programming, monkey pregnancies experimentally exposed to androgen excess result in mild-to-moderate glucose intolerance in exposed dams [52]. It is thus not surprising that androgen-exposed fetuses demonstrate increased biparietal diameter. Such increased fetal size, however, normalizes at term, but is associated with increased infant bodyweight by 2 months of age [53]. Impaired maternal glucoregulation is also found in PCOS pregnancies, and PCOS mothers are at increased risk for gestational diabetes [31]. Interestingly, CAH mothers are also at an increased risk for gestational diabetes [54]. Nevertheless, macrosomia is not prevalent in PCOS pregnancies and Boomsma and colleagues have speculated that such lack of macrosomia may be due to placental insufficiency (from increased preeclampsia and pregnancy-induced hypertension) counteracting increased glucose load [32].

Fetal androgen excess in female monkeys alters their developmental trajectory towards a PCOS phenotype

While we know virtually nothing about PCOS phenotypes during fetal life and early childhood in humans, prenatally androgenized female monkeys exposed to androgen excess during early gestation provide some insight. By late gestation and in the first month of infancy, LH excess is a programmed trait [3,55], probably reflecting fetal androgen excess entrainment of the hypothalamic pulse generator to diminished steroid hormone negative feedback [56], as found in fetal male monkeys [57].

After birth, hyperandrogenism is evident in prenatally androgenized infant female monkeys that may reflect prolonged ovarian exposure to LH excess [3]. The infant adrenal gland may contribute to this postnatal endogenous androgen excess, since adrenarche occurs around the time of birth in rhesus monkeys coincident with regression of the adrenal fetal zone [58], and adult prenatally androgenized female monkeys exhibit adrenal hyperandrogenism [59]. A relative hypersecretion of insulin may contribute to the approximately 10% increase in prenatally androgenized bodyweight by 2 months of age [54] that could progress into increased abdominal adiposity in adulthood [60].

While there are no data concerning prepubertal endocrine status or metabolically relevant parameters in juvenile, prepubertal female prenatally androgenized monkeys, these androgen-exposed female monkeys exhibit enhanced male-typical behavior [61] analogous to altered behavior in prepubertal CAH [62] and PCOS [63] girls. By adolescence, though, we can see the PCOS phenotype beginning to emerge. Menarche, an event regulated by both hypothalamic neuroendocrine function and body fat mass [64], is delayed or normal in prenatally androgenized female monkeys [65,66], delayed or normal in girls with CAH [67], and delayed or advanced in PCOS adolescents [9,48]. Following menarche, menstrual cycle disturbances ensue. Prenatally androgenized female monkeys initially exhibit longer periods of amenorrhea followed by a greater incidence of luteal phase defects [63], while adolescent PCOS girls exhibit persisting intermittent or absent menstrual cycles [68], increased overweight and metabolic syndrome [69], and hyperandrogenism with accompanying LH excess [70,71]. Reproductive and metabolic defects thus appear together yet again in young monkeys or children destined to become PCOS in adulthood.

Adult PCOS-like phenotypes in prenatally androgenized female monkeys

The PCOS-like phenotypes in adult monkeys are comprehensive and include both reproductive and metabolic dysfunction. The PCOS-like traits expressed, however, depend on the timing of gestational androgen excess. Early gestation-exposed prenatally androgenized female monkeys contribute approximately 70% of cases of polyfollicular ovaries [18] and exhibit basal hyperandrogenism as well as human chorionic gonadotropin (hCG)-stimulated ovarian hyperandrogenism [2,18], intermittent or absent ovulatory menstrual cycles [2,8], basal LH excess and gonadotropin-releasing hormone stimulated pituitary LH hyperresponsiveness [2], diminished oocyte developmental competence [72], defects in insulin action and secretion [2,73], increased incidence of Type 2 diabetes [18], increased abdominal adiposity [61,74] and hyperlipidemia [75]. Late gestation-exposed prenatally androgenized female monkeys, on the other hand, exhibit hCG-stimulated ovarian hyperandrogenism [2], intermittent or absent ovulatory menstrual cycles [2], diminished oocyte developmental consequence [72], exaggerated adiposity-related insulin resistance [73] and increased total body fat [74]. Exposure to androgen excess after the majority of neuroendocrine, pancreatic and ovarian differentiation has occurred in the monkey [18], may explain the absence of defects in LH, insulin and basal testosterone release in these late gestation-exposed female monkeys.

As in women with PCOS, defects in insulin action, ovulation and steroidogenesis in prenatally androgenized monkeys all improve on treatment with pioglitazone, an insulin sensitizer (early and late gestation-exposed females combined [4,75]). Reducing circulating insulin levels and insulin resistance are key to normalizing reproductive function impaired by fetal androgen excess in prenatally androgenized monkeys and in women with PCOS. Deficient insulin signaling in multiple organ systems is, therefore, a fundamental common defect. It is interesting to speculate whether this is induced by fetal programming from a relatively hyperglycemic pregnancy, a hyperandrogenic fetus or a combination of both.

Compensatory hyperinsulinemia from adiposity-related insulin resistance may also play a mechanistic role in diminishing oocyte quality in prenatally androgenized female monkeys. Early, but not late, gestation-exposed monkeys undergoing recombinant human follicle-stimulating hormone (rhFSH)/hCG combination therapy for IVF show relative hyperinsulinemia and LH excess at oocyte retrieval [72], suggesting that the timing of fetal androgen excess may influence susceptibility of the oocyte to androgen programming in utero by altering follicle differentiation through subtle metabolic and/or neuroendocrine dysfunction. Equally important, low intrafollicular estradiol and androstenedione responses to rhFSH in all prenatally androgenized monkeys are associated with profound impairment of blastocyst development in early gestation-exposed monkeys, but only a subtle impairment of blastocyst development in late gestation-exposed monkeys [18]. The relatively hypoandrogenic responses of prenatally androgenized monkeys to rhFSH, however, resemble those observed in IVF patients with diminished ovarian reserve [76], rather than similarly treated PCOS women [77]. Consistent with the notion of oocyte susceptibility to the consequences of androgen excess programming, obese PCOS patients experience low oocyte fertilization and failure of embryos to implant in their own uterus or those of their surrogates [78], implicating impaired oocyte development. Furthermore, the transcriptome of mature oocytes retrieved from PCOS women following ovarian stimulation for IVF fails to undergo typical transcription silencing prior to fertilization [79]. Together, these findings raise the concern that fetal androgen programming may have transgenerational consequences for female offspring in addition to any genetically heritable traits.

Translation of a developmental origins hypothesis for PCOS into human therapies

The prenatally androgenized female monkey model for PCOS makes a compelling argument for fetal androgen excess resetting the female reproductive trajectory, while a combination of hyperglycemic pregnancy and later-onset metabolic abnormalities influence the severity of the adult phenotype. In support of such a ‘two-hit hypothesis’, the ‘first hit’ represents fetal androgen excess alteration of neuroendocrine regulation of ovarian function reducing hypothalamic sensitivity to steroid negative feedback. The resulting LH hypersecretion may aid in perpetuating ovarian hyperandrogenism. The ‘second hit’, compensatory hyperinsulinemia from in utero hyperglycemia or adult adiposity-dependent insulin resistance may further complicate reproductive dysfunction since postnatal weight gain in prenatally androgenized monkeys [8] amplifies reproductive dysfunction, and such dysfunction can be readily ameliorated by insulin sensitizer-induced reductions in circulating insulin levels [75].

That such developmental origins for PCOS may be influenced strongly by the postnatal environment, yet go virtually unrecognized until adulthood, raises profound healthcare concerns regarding the effect of postnatal obesity on an individual’s susceptibility to a plethora of diseases. With a progressive epidemic of obesity likely to induce metabolic abnormalities that amplify fetal androgen excess programming, the need for nonhuman primate models to understand the developmental origins of PCOS-like reproductive and metabolic dysfunction is crucial to the improvement of clinical strategies for diagnostics that target abnormalities in maternal–fetal and postnatal environments and reduce the risk of long-term adult disease.

Financial & competing interests disclosure

This work was supported by US NIH grants R01 RR013635, P50 HD044405, U01 HD044650 and P51 RR000167 (WNPRC base operating grant), and was partly conducted at a facility (WNPRC) constructed with support from Research Facilities Improvement Program grant numbers RR015459-01 and RR020141-01. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.