Hyperandrogenaemia in adolescent girls: origins of abnormal gonadotropin‐releasing hormone secretion

Polycystic ovarian syndrome is a common disorder characterised by ovulatory dysfunction and hyperandrogenaemia. Its origins begin peripubertally, as adolescent hyperandrogenaemia commonly leads to adult hyperandrogenaemia and decreased fertility. Hyperandrogenaemia reduces the inhibition of gonadotropin-releasing hormone pulse frequency by progesterone, causing rapid luteinising hormone pulse secretion and further increasing ovarian androgen production. Obese girls are at risk for hyperandrogenaemia and develop increased luteinising hormone pulse frequency with elevated mean luteinising hormone by late puberty. Many girls with hyperandrogenaemia do not exhibit normal luteinising hormone pulse sensitivity to progesterone inhibition. Thus, hyperandrogenaemia may adversely affect luteinising hormone pulse regulation during pubertal maturation, leading to persistent hyperandrogenaemia. Polycystic ovarian syndrome (PCOS) is the leading cause of infertility, affecting approximately 6–8% of women of reproductive age, and is associated with obesity in 60% of affected women in the USA, with insulin resistance and hyperinsulinaemia in 50–70%, diabetes mellitus in 4–10% and markers of cardiovascular disease risk in 33%.1–3 Adolescents with PCOS have a 30–60% prevalence of metabolic syndrome, which is four to five times higher than that in age- and body mass index-matched girls.4 Hallmarks of PCOS are clinical and/or biochemical evidence of hyperandrogenaemia (HA) (i.e. excessive acne, hirsutism or elevated free testosterone levels) and evidence of anovulatory cycles (i.e. oligo/amenorrhoea). Different definitions of PCOS currently contribute to the heterogeneity of the clinical and pathophysiological features within the diagnosis. The National Institutes of Health criteria (1990) require hyperandrogenism and ovulatory dysfunction, but not polycystic ovarian morphology. The Rotterdam criteria (2003) identify women having two of three features: androgen excess, ovulatory dysfunction and polycystic morphology on ovarian ultrasound. This definition is more inclusive, as HA is not a required feature of PCOS. The inclusion of ovarian morphology in the definition of PCOS has been less validated and may not be useful in adolescent girls. Enlarged ovarian volume occurs in 50% of asymptomatic adolescent postmenarcheal girls, probably a developmental stage of maximal ovarian size and a normal variant in most girls.5 Therefore, the ability to discern abnormalities of morphological change in pubertal girls via ovarian ultrasound may be limited. The origins of PCOS are possibly pre- or peripubertal, as clinical manifestations of the disorder frequently develop shortly after menarche. Both adolescents and adult women with PCOS have disruptions in the regulation of the gonadotropin-releasing hormone (GnRH) pulse generator, characterised by rapid luteinising hormone (LH) (and hence GnRH) pulsatility,6,7 with impaired inhibition by progesterone.8,9 LH pulse sensitivity to slowing by progesterone can be restored in women with PCOS by antiandrogen therapy, suggesting that HA is responsible for the impaired feedback.10 HA during adolescence is recognised as a forerunner to adult PCOS, as it is associated with higher androgen levels in adulthood and lower fertility rates.11 Obese adolescents are at increased risk of HA12 and later development of PCOS.13 In utero factors have been suggested to be an important contributor to insulin resistance and HA. Animal models support this idea, as intrauterine androgen exposure in primates14 and sheep15,16 causes changes in LH secretion, LH pulse sensitivity to progesterone feedback and insulin metabolism, especially when the animals are overfed postnatally.16 Low birthweight seems to be a factor in the development of PCOS in some girls in Spain and Italy.17,18 However, studies in Finland, Amsterdam and the UK have not supported this relationship in their populations.13,19,20 In premenopausal women, approximately one-half of circulating testosterone is derived from the peripheral conversion of androstenedione, particularly in adipose tissue,21 with the remainder being derived from ovarian and adrenal sources. Obesity alone increases androgen production, as adipose tissue, particularly from the abdominal region, contains enzymes of steroidogenesis.22 LH is a major physiological stimulus for ovarian androgen production from theca cells. In addition, hyperinsulinaemia related to obesity can contribute to HA through several mechanisms. Insulin can act as a co-gonadotropin with LH on ovarian theca cells to increase androgen production,23,24 can increase adrenal responsiveness to adrenocorticotropin hormone for further androgen production,25 can potentiate human chorionic gonadotropin-mediated ovarian follicle arrest,26 and can cause pituitary hyperresponsiveness to GnRH for increased LH secretion in in vitro studies.27 Women with PCOS often have elevated LH and insulin levels and are obese, leading to compounded mechanisms for increased androgen production. HA is associated with excess weight during puberty,12,28,29 with a prevalence of 60–94% in our cohort of obese girls (body mass index for age ≥95%),12 and can be ameliorated by weight loss.28,29 Circulating androgen levels normally increase slightly during puberty, with early morning elevations of serum testosterone seen in prethelarchal girls.30 In normal weight pubertal girls, the levels of total testosterone increase and sex hormone binding globulin decrease, leading to higher circulating free testosterone levels.12,28 Obese girls have even higher levels of testosterone, with diminished sex hormone binding globulin throughout all stages of puberty, leading to marked elevations of free testosterone (Figure 1).12 Dehydroepiandrosterone sulphate, an adrenal androgen, is modestly increased in obese compared with normal weight girls in early puberty, but the differences disappear by the end of puberty.12 Early morning hormone levels in nonobese (open squares) and obese (filled squares) peripubertal girls by breast Tanner stage. Data are presented as the mean ± SEM. Differences were assessed with Wilcoxon rank sum test before Bonferroni correction: *P < 0.05, **P ≤ 0.01, ***P ≤ 0.001, ***P ≤ 0.0001. Conversion from conventional to SI units: total testosterone (T) × 3.47 (nmol/l). SHBG, sex hormone binding globulin. (McCartney et al.,12 adapted with permission from The Endocrine Society.) The GnRH pulse frequency designates the preferential production of LH (via high-frequency pulses) versus follicle-stimulating hormone (FSH) (via low-frequency pulses) in normal adult women.31 The pulse frequency is regulated by progesterone in the presence of estradiol,8 such that increased progesterone production by the corpus luteum slows LH pulse frequency to favour FSH production, which aids in follicular development for the next menstrual cycle (Figure 2). Women with PCOS have abnormally rapid LH pulses with reduced response to progesterone feedback,8 contributing to elevations in serum LH:FSH ratios. As LH stimulates theca cell steroid production and FSH regulates follicular conversion of androgen to estrogen, this relative increase in LH leads to increased ovarian androgen synthesis with limited aromatisation to estradiol. The resultant increase in serum testosterone contributes to the maintenance of increased LH pulse frequency,10 creating a vicious loop, leading to the production of more ovarian androgens. Schematic diagram of hormonal patterns during an ovulatory menstrual cycle. E2, estradiol; FSH, follicle-stimulating hormone; GnRH, gonadotropin-releasing hormone; LH, luteinising hormone; P, progesterone. (Marshall and Eagleson,36 with permission from Elsevier.) Clues to the aetiological factors related to the instigating events leading to PCOS can be obtained from adolescents. In girls, LH is independently and positively correlated with free testosterone when adjusting for age, pubertal stage, body mass index, dehydroepiandrosterone sulphate and insulin.28 This might suggest that pre-existing neuroendocrine abnormalities affecting LH production during puberty lead to HA. Alternatively, HA might impair the inhibition of the GnRH pulse generator, leading to high-frequency pulses that favour LH production. Therefore, further study of normal and abnormal development of GnRH pulse generator regulation during puberty may help to discern more precise aetiological mechanisms for HA in PCOS. The maturation of the GnRH pulse generator undergoes a typical developmental progression during childhood. During infancy in girls, the GnRH pulse generator is active, producing FSH to high adult levels by about 2 months of age and LH to early pubertal levels by about 4 months of age, which both then decrease slowly over the first year to prepubertal levels by 4 years of age.32 As early as 5–6 years of age (prethelarchal), small pulses of LH (and, by inference, GnRH) can be seen intermittently throughout the day, with pulses of increased amplitude entrained to sleep,30,33,34 which are quite sensitive to progesterone inhibition.9 These augmented night-time LH pulses are associated with overnight increases in sex steroid production (primarily progesterone and testosterone), which diminish during the following daytime hours.30,35 As puberty progresses, LH pulses gain increasing amplitude and frequency, initially throughout the night-time hours and then during the day.34,35 By the end of puberty, daytime LH pulses have a higher frequency than night-time pulses with increased amplitude during sleep (Figure 3).34,35 Late evening and overnight luteinising hormone (LH) characteristics in nonobese (NO, open squares) and obese (OB, filled circles) peripubertal girls by breast Tanner stage. Data are presented as the mean ± SEM. The numbers of subjects are given below the Tanner stage labels. The last column shows obese Tanner 5 girls only with polycystic ovarian syndrome (PCOS) (filled triangles) and without PCOS (open triangles). As a result of blood volume constraints, time point 19.00–23.00 (subject awake) was used as a surrogate for the daytime hormone levels. (McCartney et al.,35 with permission from The Endocrine Society.) Obese girls, however, have altered maturation of the GnRH pulse generator (Figure 3). Initially, their LH pulses have lower amplitude and frequency overnight compared with normal weight girls, consistent with lower mean LH concentrations.35 By mid-puberty, however, they surpass normal weight girls in LH pulse frequency during both the day and night, although their pulse amplitude and mean LH concentrations remain lower. By the end of puberty, obese girls have significantly higher LH pulse frequency which diminishes only slightly with sleep. Although they still have lower LH pulse amplitudes, their mean LH concentrations have become elevated as in adults with PCOS, probably reflecting the increased LH pulse frequency. Obese girls with clinical evidence of PCOS have even less night-time diminution of LH pulse frequency, greater pulse amplitude and significantly higher mean LH concentrations than obese girls of similar pubertal stage without clinical evidence of PCOS. Perhaps this designates a continuum of developmental LH pulse abnormalities during puberty, which leads to adolescent PCOS symptoms in more strongly affected individuals. During early puberty, the GnRH pulse generator is exquisitely sensitive to inhibition by progesterone. LH pulses are essentially abolished by administered progesterone during early pubertal stages.9 As puberty progresses, normal girls exhibit reduced sensitivity to progesterone inhibition, similar to normal adult women (Figure 4). In adolescents with HA, however, approximately 60% do not appropriately suppress LH pulsatility after progesterone administration, similar to adult women with PCOS (Figure 4).9 Interestingly, in girls with HA, the reduction in LH pulse frequency after progesterone decreases with increasing fasting insulin levels, suggesting that hyperinsulinaemia may further contribute to impaired GnRH regulation.9 Percentage change in luteinising hormone (LH) pulse frequency during 11 hours following 7 days of oral estradiol (E2) and progesterone (P) plotted as a function of mean plasma P on day 7 in controls (A) and adolescent girls with hyperandrogenaemia (HA) (B). Filled circles represent girls with breast Tanner stages 1–2; open circles represent girls with Tanner stages 3–5. The shaded background areas represent the range of response to a similar protocol in adult control women (A) and women with polycystic ovarian syndrome (PCOS) (B). The outlined area in (B) represents the range of responses in control adolescent girls. Conversion from conventional to SI units: P × 3.18 (pmol/l). (Blank et al.,9 with permission from The Endocrine Society.) Progressively diminishing sensitivity to inhibition by progesterone may help to determine the developmental patterns of the GnRH pulse generator during puberty. Progesterone levels increase overnight during early puberty in normal girls at the same time as the development of detectable sleep-associated LH pulses.35 On the following day, progesterone again falls to lower levels. In subsequent stages of puberty, daytime progesterone levels diminish to a lesser degree with a continued slight increase overnight.35 The overnight rise of progesterone may contribute to the subsequent suppression of daytime LH pulses during early puberty through exquisitely sensitive feedback inhibition. Free testosterone has a similar overnight rise in plasma,30 whereas estradiol does not; however, levels of both hormones increase during daytime as puberty progresses (Figure 5).35 Late evening and overnight sex steroid concentrations in nonobese (NO, open squares) and obese (OB, filled circles) peripubertal girls by breast Tanner stage. Data are presented as the mean ± SEM. The numbers of subjects are given below the Tanner stage labels. The last column shows obese Tanner 5 girls only with polycystic ovarian syndrome (PCOS) (filled triangles) and without PCOS (open triangles). Conversion from conventional to SI units: progesterone × 3.18 (pmol/l); free testosterone (pmol/l); estradiol × 3.671 (pmol/l). (McCartney et al.,35 with permission from The Endocrine Society.) Given that androgens decrease the feedback inhibition of LH pulsatility by progesterone in adult women with PCOS,10 we hypothesise that increasing androgen levels throughout puberty gradually impair the inhibitory effects of the overnight rise in progesterone. This results in progressive daytime increases in LH pulsatility, as is seen in normal weight girls as they progress through puberty (Figure 3). In obese girls with HA, increased androgen levels may further impair the normal patterns of progesterone inhibition during puberty, leading to earlier development of elevated daytime LH pulsatility.7 Over time, this change in the maturational pattern of LH pulses could enhance ovarian androgen production, especially in later puberty, leading to a symptomatic PCOS phenotype. Further studies are needed to elucidate the precise mechanisms for the development of abnormal GnRH regulation in obese girls. The source of overnight increases in progesterone is unclear, as ovarian hormone production is thought to be minimal in early puberty. Alternatively, the adrenal gland and not the ovary may be responsible for this diurnal variation in progesterone levels, similar to diurnal cortisol variability. The role of androgens in decreasing progesterone sensitivity during pubertal maturation and the impact of progesterone inhibition also remain to be demonstrated. Studies to determine whether antiandrogen administration to obese girls could restore sensitivity to progesterone and normalise the development of LH pulsatility are required to establish this concept. There are no relevant or significant financial, personal, political, intellectual or religious conflicts of interest to disclose. CM Burt Solorzano is the primary author of this article, writing all drafts up to the final manuscript. JC Marshall contributed extensive manuscript intellectual, organisational and editorial mentorship. CR McCartney, SK Blank and KL Knudsen contributed intellectual and editorial advice to the preparation of the final manuscript. This is a review article and does not include previously unpublished human study data. All information relayed in this article has been collected from previously published studies, which have therefore undergone the ethical scrutiny of each corresponding journal. Studies previously reported from our research group have all received Institutional Review Board approval from the University of Virginia and our General Clinical Research Center. Our studies were supported by the National Institutes of Health U54 Grant HD 28934–16 from the Eunice Kennedy Shriver Institute for Child Health and Human Development. No funding was received or given for publication, writing or editorial assistance. We are grateful to L. Lockhart, MPH for research coordination; M. DeBoer, MD, W. Clarke, MD and S. Cluett, CPNP for subject recruitment for our studies; the GCRC staff and nurses; and the Ligand Core Laboratory of the Center for Research in Reproduction.