Fetal Programming of Polycystic Ovary Syndrome by Androgen Excess: Evidence from Experimental, Clinical, and Genetic Association Studies

Context: Polycystic ovary syndrome (PCOS) is a common endocrine disorder of premenopausal women, characterized by hyperandrogenism, polycystic ovaries, and chronic anovulation along with insulin resistance and abdominal obesity as frequent metabolic traits. Although PCOS manifests clinically during adolescence, emerging data suggest that the natural history of PCOS may originate in intrauterine life.

Evidence Acquisition: Evidence from experimental, clinical, and genetic research supporting the hypothesis for the fetal origins of PCOS has been analyzed.

Evidence Synthesis: Female primates, exposed in utero to androgen excess, exhibit the phenotypic features of PCOS during adult life. Clinical observations also support a potential fetal origin of PCOS. Women with fetal androgen excess disorders, including congenital 21-hydroxylase deficiency and congenital adrenal virilizing tumors, develop features characteristic of PCOS during adulthood despite the normalization of androgen excess after birth. The potential mechanisms of fetal androgen excess leading to a PCOS phenotype in humans are not clearly understood. However, maternal and/or fetal hyperandrogenism can provide a plausible mechanism for fetal programing of PCOS, and this, in part, may be genetically determined. Thus, genetic association studies have indicated that common polymorphic variants of genes determining androgen activity or genes that influence the availability of androgens to target tissues are associated with PCOS and increased androgen levels. These genomic variants may provide the genetic link to prenatal androgenization in human PCOS.

Conclusion: Prenatal androgenization of the female fetus induced by genetic and environmental factors, or the interaction of both, may program differentiating target tissues toward the development of PCOS phenotype in adult life.

POLYCYSTIC OVARY SYNDROME (PCOS) is a common endocrine and metabolic disorder affecting 5–10% of women in their reproductive life (1–4). The endocrine manifestations of PCOS include excess androgen production of ovarian and/or adrenal origin and arrested follicular development leading to chronic oligo- or anovulation (5). An important feature of PCOS, resulting from aberrant folliculogenesis, is accumulation of small follicular cysts and increased ovarian stromal thickness yielding a characteristic morphology on ultrasound (5). Often women with PCOS have increased LH secretion and decreased SHBG levels in serum, which in turn contribute to hyperandrogenemia and increased availability of free androgens, respectively (5). Insulin resistance with compensatory hyperinsulinemia, which further exacerbates the hyperandrogenism and ovulatory dysfunction, is a frequent metabolic trait, which in association with other features of the metabolic syndrome imposes an increased risk for type 2 diabetes and cardiovascular disease (6–8).

The spectrum of the above endocrine and metabolic abnormalities may vary among affected women creating heterogeneous clinical and biochemical phenotypes, and this has implications for precise diagnosis of the syndrome. Recently a revised consensus diagnosis of PCOS was agreed by an International Consensus Working Group in The Netherlands (9). According to this consensus, PCOS can be diagnosed after exclusion of other conditions that cause hyperandrogenism and if at least two of the following three criteria are present: oligoovulation or anovulation (manifesting as oligomenorrhea or amenorrhea), elevated serum androgen levels or clinical manifestations of androgen excess (hirsutism, acne, or androgenic alopecia), and polycystic ovaries as defined by ultrasound examination (9).

The pathogenesis of the syndrome is largely unknown, but there is increasing evidence for a strong genetic basis (10, 11). PCOS usually becomes clinically manifest during adolescence along with maturation of the hypothalamic-pituitary-gonadal axis (12). However, recent evidence suggests that the natural history of PCOS may originate in very early development, even in intrauterine life. Indeed, experimental animal research and clinical observations have led to the developmental origin hypothesis of PCOS (13). According to this hypothesis, fetal exposure to androgen excess can induce changes in differentiating tissues leading to PCOS phenotype in adult life (13, 14).

We review the growing evidence from experimental, clinical, and genetic association studies on the role of prenatal androgen excess in the developmental origin of PCOS.

Evidence for the Developmental Origin of PCOS

Experimental animal studies

Experimental evidence strongly supports the hypothesis that the intrauterine environment may influence the phenotypic expression of PCOS.

Experiments in rats 40 yr ago showed that the pattern of gonadotropin-hormone release is programmed at the hypothalamus by the concentration of androgens during early development (15). In addition, female rats exposed to androgen excess may develop anovulatory sterility and polycystic ovaries (16).

Studies in sheep showed that exposure to androgens in utero reduces the sensitivity of GnRH neural network to steroid-negative feedback, leading to increased LH secretion and aberrant ovarian follicular development in the female offspring (17, 18). This was postulated to be caused by androgen-induced reprogramming of neural development such that GnRH neurons are desensitized to steroid feedback. A recent study showed that prenatal androgenization of transgenic female mice alters the drive of the secretion of γ-aminobutyric acid to GnRH neurons in adult mice, implicating that the γ-aminobutyric acid secretion system may be reprogrammed by androgens in utero (19). In addition, the intrafollicular availability of activin appeared to be compromised in prenatally androgenized ewes as evidenced by increased expression of follistatin along with the reduced expression of activin-βΒ mRNA in the majority of follicles in the androgenized animals, compared with controls (20). The activin-follistatin-inhibin system appears to play an important role in normal folliculogenesis, and its dysregulation is implicated in the impaired follicular development in PCOS. Activin promotes granulosa cell proliferation, enhances FSH receptor expression, decreases LH-induced androgen production, increases pituitary FSH secretion, and generally promotes follicular growth (21). Follistatin, a binding protein for activin, and inhibin antagonize most of the actions of activin. Therefore, a decrease in the concentration or functional activity of activin as well as an increase in follistatin might contribute to the defective follicular development characteristic of PCOS. Indeed, clinical studies have reported increased circulating follistatin and decreased activin concentrations in women with PCOS, compared with controls (22, 23). Furthermore, prenatal androgen exposure of sheep not only leads to multifollicular ovarian development but also results in intrauterine growth retardation and postnatal catch-up growth (24).

A more appropriate model for understanding the reproductive outcome of female prenatal androgen excess is the nonhuman primate. Studies have shown that female rhesus monkeys exposed in utero to levels of testosterone equivalent to those found in fetal males develop clinical and biochemical features in adult life resembling those observed in women with PCOS. More precisely, prenatally androgenized female rhesus monkeys exhibit all three main features required for the consensus diagnosis of PCOS (14). Indeed, they exhibit ovarian hyperandrogenism as indicated by higher basal and human chorionic gonadotropin (hCG)-stimulated serum testosterone levels (25). They also manifest enhanced basal and ACTH-stimulated adrenal androgen levels causing adrenal hyperandrogenism comparable with that found in PCOS women (26). In addition, prenatally androgenized females demonstrate ovulatory dysfunction resulting in about 40–50% fewer menstrual cycles than normal females (27). This menstrual dysfunction is greatest in those females with high body mass index, suggesting that hyperinsulinemia may also contribute to anovulation. Finally, a high proportion of prenatally androgenized female monkeys have enlarged polyfollicular ovaries resembling the morphology of polycystic ovaries seen in PCOS women (27). Collectively, the above reproductive-endocrine defects qualify prenatally androgenized female rhesus monkeys for a PCOS diagnosis (14).

It is of interest that differential timing of fetal androgen excess may be important for the expression of PCOS traits beyond the diagnostic criteria, in a fashion similar to the heterogeneity in phenotypes expressed in PCOS women. Thus, LH hypersecretion is found only in female monkeys that are exposed to excess androgens in early gestation (14, 28).

In addition to its impact on the reproductive endocrine axis, prenatal androgenization of female rhesus monkeys may also induce metabolic defects that characterize PCOS, including aberrant adipose tissue distribution and defects in insulin secretion and action. Thus, female rhesus monkeys exposed to androgen excess in utero manifest preferential accumulation of visceral fat in adulthood, which is independent of obesity (29).

Furthermore, prenatal androgen excess may perturb the insulin-glucose homeostasis, depending on the timing of androgen exposure during gestation. Early-treated females exhibit impaired insulin secretion, whereas late-treated females show decreased insulin sensitivity with increasing adiposity (30). On the other hand, adult male monkeys exposed to excess androgens in utero may also develop defects in insulin secretion and action in a similar way to female counterparts (31).

Collectively, these findings support the view that prenatal androgen excess can induce metabolic defects similar to those expressed in PCOS women and that the timing of fetal exposure is important in determining phenotypic presentation.

Clinical Observations

The developmental origin hypothesis of PCOS emerged after astute clinical observations in women with fetal androgen excess disorders and is substantiated by the experimental studies on the prenatally androgenized primate model, discussed above. Examples of prenatally androgenized humans are women with congenital adrenal hyperplasia from 21-hydroxylase deficiency and women with congenital adrenal virilizing tumors. Despite the normalization of androgen excess with treatment or removal of the tumor after birth, these women manifest PCOS traits including anovulatory cycles, ovarian hyperandrogenism, LH hypersecretion, polycystic-appearing ovaries, and insulin resistance in adult life (32, 33). Further experiments of nature associated with prenatal androgenization are women with rare loss-of-function mutations of the P450 aromatase gene or the gene of SHBG. Such patients were reported to develop features of hyperandrogenism and polycystic ovaries (34, 35).

The possible role of PCOS itself as a cause for prenatal androgen excess was evaluated in a recent study (36). It was reported that pregnant PCOS women have higher concentrations of androgens than normal pregnant women. The origin of the androgen excess during pregnancy in PCOS women is uncertain, but it could be due to increased androgen production in the maternal theca cells or the placenta, stimulated by hCG. Normally maternal androgens or fetal adrenal androgens are rapidly converted to estrogens by the activity of the placental enzyme aromatase. However, when the enzyme activity is inhibited, the availability of androgens could be increased. Insulin has been shown to inhibit aromatase activity in human cytotrophoblasts and stimulate 3β-hydroxysteroid dehydrogenase activity (37–39). Therefore, in pregnant PCOS women, hyperinsulinemia could potentially contribute to excess androgen exposure in their female offspring (33).

In addition to increased androgen levels during pregnancy, PCOS women may also deliver small-for-gestational-age newborns at a higher prevalence than control mothers (36). It is postulated that the prenatal exposure to androgens in offspring of PCOS mothers may provide the stimulus for both the low birth weight and development of the PCOS phenotype later in life. In this regard, recent studies in girls have shown that low birth weight is related to the development of premature pubarche followed by functional hyperandrogenism, insulin resistance with hyperinsulinism, and dyslipidemia during adolescence (40, 41). It has been suggested that these manifestations may have a common early origin (41).

Apart from the associations cited above, epidemiological studies have also shown that low birth weight, reflecting alterations in intrauterine nutritional milieu, is associated with an increased risk of developing features of the metabolic syndrome including central obesity, insulin resistance, type 2 diabetes, and hypertension (42). These epidemiological associations, replicated by studies in animals, have led to the fetal origin hypothesis of adult disease (42, 43). This is thought to be the consequence of programing whereby a stimulus or insult at a crucial, sensitive period of early life may induce long-term changes in physiology and metabolism (44). It appears that the phenotypic expression in adulthood is dependent on the nature of stimulus during intrauterine life. In the case of exposure to androgen excess, this may lead to adult PCOS phenotype.

Genetic Association Studies

The clinical observations and experimental animal research cited above suggest a common prenatal etiology for the multiple manifestations of PCOS that may be programed in utero or early postnatal life by androgen excess. The potential origin of the excess androgens during intrauterine life to account for fetal programming of PCOS in humans is not known. Theoretically this could be from maternal sources, placental production, or endogenous fetal origin. As discussed above, it is unlikely that an excess of maternal androgens can be implicated because the human fetus is normally protected from maternal androgens by increased placental aromatase activity and the high levels of SHBG that characterize a normal pregnancy. However, this buffering effect may be overcome if placenta aromatase activity is inhibited or the production of SHBG is reduced by genetic or other factors, i.e. by hyperinsulinemia in the case of pregnant PCOS women (36).

On the other hand, although it is thought that the human fetal ovary is quiescent in terms of sex steroid production, it is possible that genetic or other factors may influence the steroidogenic activity in utero in response to hCG or during infancy by the burst of gonadotropin secretion resulting in a hyperandrogenic ovary (45, 46). Maternal hyperinsulinemia may induce excessive placental hCG production leading to fetal ovarian hyperplasia and hyperandrogenism (45). A hyperandrogenic adrenal cortex may also contribute to ovarian androgen production because fetal ovaries are able to convert steroid precursors to functional ovarian androgens (47).

Ovarian hyperandrogenism is a key diagnostic feature and is a heritable PCOS trait (48). Indeed, increased functional activity of cytochrome P450 steroidogenic enzymes, 3β-hydroxysteroid dehydrogenase enzyme, and intracellular kinase proteins important for ovarian theca cell steroidogenesis constitute the molecular phenotype of PCOS theca cells (49, 50). Furthermore, recent studies using DNA microarrays in cultured theca cells from PCOS women identified the genes encoding aldehyde dehydrogenase-6 and retinol dehydrogenase-2 as candidate genes for PCOS (46). These factors play a role in all-trans-retinoic acid synthesis and the transcription factor GATA6 that, in turn, increase the expression of 17α-hydroxylase, a functional characteristic of PCOS theca cells (49).

These observations prompted the consideration of genes encoding steroidogenic enzymes as potential candidate genes for the etiology of ovarian hyperandrogenism in PCOS. Almost all initial candidate genes regulating theca cell steroidogenesis, including CYP17 and CYP11α encoding for the P450c17a and P450scc enzymes, respectively, have failed to show association or linkage to PCOS phenotype (51, 52). Thus, apart from the genes regulating transcription factors involved in theca cell steroidogenesis, genetic variants of steroidogenic enzymes have not been linked to PCOS.

On the other hand, human theca cells have been shown to express androgen receptors that may act as transcriptional factors in molecular pathways involved in cellular differentiation and function (53). In this regard, ovarian hyperandrogenism might relate to genetic variation in androgen receptor sensitivity. Indeed, shorter androgen receptor (AR) gene CAG repeat number is linked to increased receptor sensitivity in vitro (54) and is associated with variation in androgen activity in vivo in both females and males (55, 56). In support of this, a recent study (57) showed that genetically higher AR sensitivity, as indicated by shorter AR (CAG)n repeat alleles, is a risk factor for the development of ovarian hyperandrogenism among Spanish adolescent girls. This study demonstrated that clinical hyperandrogenism and androgen levels were increased in those women with shorter AR (CAG)n repeat alleles, indicating that higher AR sensitivity is associated, through a positive feedback mechanism, with increased ovarian androgen production (57). The same study also showed that both low birth weight and shorter CAG repeats are independently related to ovarian hyperandrogenism and hyperinsulinemia, and their effects appear to be additive (57).

Another study by the same authors showed that both insulin and free androgen levels were independently associated with central fat mass in adolescents with a history of low birth weight and precocious pubarche, indicating that both hyperinsulinemia and increased androgen activity may promote an android distribution of body fat (58).

The role of hyperandrogenism in the pathogenesis of the metabolic trait of central adiposity in PCOS, however, remains controversial. On the one hand, androgens appear to enhance catecholamine-induced lipolysis and testosterone administration to decrease visceral fat mass in men (59). Conversely, lowering androgens in PCOS women by pituitary desensitization to GnRH increases visceral adiposity (60), similar to the increase in adipose mass in elderly men with declining androgen production (61) or middle-aged men with low free testosterone levels (62). These data suggest that hyperandrogenism per se may not be directly involved in central fat deposition, but rather hyperinsulinemia or some other factor may influence body fat distribution. On the other hand, in females there is an association between visceral fat accumulation and hyperandrogenicity, despite the documented effects of testosterone on lipid mobilization and the expected decrease in visceral fat depots (63). Indeed, long-term testosterone administration to young nonobese female-to-male transsexuals increases the amount of visceral fat (64). In addition, an increase in weight in this hyperandrogenic state leads to a preferential storage of fat in the visceral depot. These data support the view that in women, a hyperandrogenic milieu or an increase in the androgen to estrogen ratio may predispose to an android body fat distribution and that the resultant hyperinsulinemia may aggravate this phenomenon.

In this regard, recent evidence suggests that in PCOS women, catecholamine-induced lipolysis is increased in visceral fat cells in a similar way to men but decreased in sc fat cells, unlike normal women (65). The latter effect may promote the development of obesity, whereas the visceral effect may lead to elevated portal fatty acid levels, leading to liver dysfunction, insulin resistance, and hyperinsulinemia, which may further induce central obesity creating a vicious circle (65).

Given the fact that regional fat distribution is regarded as a secondary sex characteristic, this indicates a potential role of sex steroid hormones in determining the site of fat deposition. It is possible that this sexually dimorphic trait is entrained in intrauterine life as was suggested by the prenatally androgenized primate model (29).

Because the AR gene is involved in prenatal sexual differentiation (66), it is plausible that a relatively pronounced androgenic effect during fetal development may influence the differentiation of sexually dimorphic tissues including adipose tissue distribution (66). Although the association of genomic variants of AR with PCOS has been confirmed in some case-control studies, other population studies have reported no linkage or association of PCOS with AR locus (51, 67). These discrepant findings may be due to the reported wide population differences in AR CAG repeat distribution and could relate to ethnic differences in body habitus and PCOS phenotypes (68).

Additional ways that might contribute to excess androgen exposure of the female fetus, regardless of the source of androgens, could be a reduced binding capacity of androgens by SHBG or reduced aromatization of androgens leading to increased tissue androgen availability or a high androgen to estrogen ratio, as was mentioned earlier.

With regard to SHBG, serum levels are often low in women with PCOS and individuals at risk for diabetes and cardiovascular disease (69, 70). Although this is thought to be the result of hyperinsulinemia and hyperandrogenemia (71, 72), there is also evidence that SHBG production may be genetically determined (35). In this regard, a (TAAAA)n pentanucleotide repeat polymorphism at the promoter of the human SHBG gene has been described and reported to influence its transcriptional activity in vitro (73).

We have recently shown in a case-control study (74) an association between the (TAAAA)n polymorphism of the SHBG gene and PCOS among Greek women. In particular, women with PCOS were more frequently carriers of longer (TAAAA)n alleles, compared with controls. Furthermore, carriers of the longer allele genotypes had lower SHBG levels and higher free androgen index than those with shorter alleles. Similar findings were reported among French women with hirsutism (75) and in a more recent study in young men (76).

These studies suggest that there may be a genetic contribution to decreased SHBG levels in women with PCOS. Those individuals with genetically determined low SHBG levels may be exposed to higher free androgen levels throughout life but, more importantly, during fetal life when programming of the differentiating target tissues takes place.

In a similar way to SHBG variation, genetic variation in the aromatase (CYP19) gene may also contribute to increased prenatal androgenization in humans (77). Although earlier linkage and association studies failed to find an association between CYP19 and PCOS, probably due to low statistical power (51), a recent association study (78) showed a strong link between genetic variations in CYP19 and androgen excess in females. In particular, the study demonstrated that a common genomic variant [an intronic single-nucleotide polymorphism (SNP) close to exon 8 of CYP19, the SNP50] was associated with early adrenarche, pubertal ovarian hyperandrogenism, and variation in PCOS symptom score in both girls and young women in two different ethnic populations.

Along the same lines, a recent study also indicated that, in premenopausal women, a short microsatelite (TTTA)n repeat allele in the fourth intron of the CYP19 gene, is associated with elevated androgens, perturbed regulation of the hypothalamic-pituitary-adrenal axis, and abdominal obesity (79).

In summary, genetic variation in androgen activity as well as androgen availability to target tissues as indicated by shorter AR (CAG)n repeat alleles, longer SHBG (TAAAA)n repeats, and polymorphic variants of the aromatase gene may contribute to prenatal androgenization and thus provide the genetic link to the fetal origin of PCOS (13). In turn, prenatal androgen exposure could be a developmental factor implicated in the etiology of PCOS.

Plausible Biological Mechanisms

The potential mechanisms of fetal programming of the PCOS phenotype by androgen excess are not well understood. Theoretically, however, three plausible biological mechanisms could be implicated in this phenomenon.

The first and more likely one involves altered target tissue differentiation. In this context, exposure of the female fetus to androgens at a time when target organ systems, in particular those regulating reproduction and metabolism, are differentiating may influence their ontogenic development and phenotypic expression.

In the reproductive context, an important key trait of PCOS is the hyperandrogenic phenotype of the ovarian theca cells, as discussed above. It is likely that the structural and functional phenotype of PCOS theca cells is programmed during differentiation by the altered intrauterine sex-steroid milieu. This notion is further supported by observations on the phenotype of the aromatase knockout female mice, in which the androgen-to-estrogen balance is altered in favor of excess androgens. The ovaries of these mice exhibit an increased interstitium with the presence of theca cells morphologically resembling Leydig cells (80). Thus, the altered androgen-to-estrogen balance appears to influence the differentiation of the ovarian theca cells toward a male-type phenotype. The molecular mechanisms, however, underlying this effect are not known. Another reproductive trait of the PCOS phenotype is LH hypersecretion, probably due to altered hormonal-negative feedback on the hypothalamic-pituitary axis. The latter may also be induced prenatally during neuroendocrine development by androgen excess. This effect, at least in the nonhuman primate model, appears to be dependent on the timing of the fetal exposure to androgens relative to neuroendocrine development because LH hypersecretion is found only in early-treated females (14).

On the other hand, a metabolic trait of PCOS that may also be programed in utero at a time of tissue differentiation is visceral fat deposition associated with insulin resistance and hyperinsulinism (58). Because body fat distribution in humans is sexually dimorphic, the central-android accumulation of adipose tissue in PCOS may, in part, reflect a masculinized pattern of fat distribution.

A second plausible mechanism of fetal programming by androgen excess could be epigenetic change in gene expression. Although this mechanism has been implicated in the fetal programming of the metabolic syndrome by maternal nutrition (81), there is no evidence that imprinting of genes can be influenced by prenatal exposure to androgens.

Finally, a third mechanism that has also been implicated in the developmental origin of metabolic syndrome due to fetal undernutrition involves adaptive homeostatic process conferring a survival advantage (70). However, whether this mechanism also applies to PCOS remains a speculation. In this context, shorter AR (CAG) repeat alleles, longer SHBG (TAAAA)n alleles, or the SNP50 of the aromatase gene could represent thrifty genotypes because increased androgenic activity and central fat accumulation may confer a survival advantage during periods of undernutrition. However, in our modern affluent society, these thrifty genes may be disadvantageous (82).

Conclusion and Future Perspectives

Clinical observations and experimental research have convincingly demonstrated that exposure of the female fetus to androgen excess during intrauterine life programs the various organ systems in such a way that they manifest the characteristics of the PCOS phenotype in adult life. Regarding PCOS in humans, the potential origin of increased androgenic activity during fetal life is not clear. However, recent association studies have indicated that a number of genes may be implicated. These are genes determining androgen activity or genes involved in the availability of androgens to target tissues. It is postulated that common genetic variants of these genes may contribute to excess androgenization during fetal life and thus provide a plausible genetic link to the developmental origin of PCOS. Other genetic factors (i.e. related to insulin resistance) and environmental exposure (i.e. dietary factors) or the interaction between the two may also contribute to prenatal androgenization and expression of the PCOS phenotype (Fig. 1).

Fig. 1.

Prenatal androgen excess and the development of PCOS phenotype.

The evidence, at least from the primate model, clearly implicates hyperandrogenism during critical periods of fetal development in the pathogenesis of the adult PCOS phenotype. The hypothesis formulated above, that genetically determined hyperandrogenism during intrauterine life may program the human fetus for the development of PCOS in adulthood, opens up new prospects in the understanding and treatment of PCOS in humans. Further research is needed to test this hypothesis and gain insight on the genetic influence of the hormonal environment during prenatal life and how this programs the differentiation of fetal target tissues. This may pave the way for possible intervention at this critical period of prenatal life to facilitate a favorable outcome during adult life.

Disclosure Statement Summary: N.X. and A.T. have nothing to declare.

Abbreviations:

 
• AR,

  Androgen receptor;

 
• hCG,

  human chorionic gonadotropin;

 
• PCOS,

  polycystic ovary syndrome;

 
• SNP,

  single-nucleotide polymorphism.