Pathogenesis
Neuroendocrine disruption
Polycystic ovary syndrome is characterised by increased pulse frequency of gonadotrophin releasing hormone and reduced negative feedback from sex steroids at the level of the hypothalamus.4 12 Gonadotrophin releasing hormone is released from neurons in the hypothalamic infundibular nucleus in a pulsatile manner, resulting in increased secretion of luteinising hormone and follicle stimulating hormone. The pulse frequency of gonadotrophin releasing hormone is controlled by multiple upstream endocrine and neural factors, with a higher frequency favouring secretion of luteinising hormone and a lower frequency favouring secretion of follicle stimulating hormone. In women with polycystic ovary syndrome, raised levels of luteinising hormone cause excess production of ovarian thecal androgens, whereas relative deficiency of follicle stimulating hormone causes follicular arrest, polycystic ovarian morphology, and oligo-ovulation.4 The reduction in sex steroid feedback on release of gonadotrophin releasing hormone is thought to occur upstream of the hormone itself because gonadotrophin releasing hormone neurons do not have receptors for oestrogens or progesterone13 (figure 1). KNDy neurons have an important role in this regard (figure 1).
Kisspeptins are a family of peptides encoded by the KISS1 gene which act on the neuronal G protein coupled receptor KISS1R. KISS1 encodes prepro-kisspeptin, which is cleaved to produce the biologically active peptides KP54, KP14, KP13, and KP10.14 Two discrete neuronal populations exist: KNDy neurons in the infundibular nucleus function as the gonadotrophin releasing hormone pulse generator15 and mediate negative feedback from oestradiol,16 whereas a separate kisspeptin population located in the preoptic area mediates oestradiol positive feedback to produce the mid-cycle surge in luteinising hormone.16 17 Kisspeptin neurons express sex steroid receptors (progesterone and oestrogen receptors) required for negative feedback on gonadotrophin releasing hormone pulsatility.17 18 KISS1 is also expressed in adipose tissue where it is regulated independently of hypothalamic KISS1.19 Circulating levels of kisspeptin are higher in patients with polycystic ovary syndrome than in controls20 and although the origin of this excess is not entirely clear, a raised pulse frequency of kisspeptin in women with oligomenorrhoea and polycystic ovary syndrome suggests a hypothalamic source.21 Moreover, physiological coupling of kisspeptin and luteinising hormone pulsatility is lost in these women.21 The exact mechanisms for these effects are unclear, with inconsistent data from preclinical models on the existence and direction of dysregulated gonadotrophin releasing hormone pulsatility mediated by kisspeptin.22
Figure 1Pathophysiology and neuroendocrine disruption of the hypothalamo-pituitary-gonadal axis in polycystic ovary syndrome. (Left) Increased pulsatility of gonadotrophin releasing hormone (GnRH) causes increased secretion of luteinising hormone, consequent disrupted folliculogenesis, and increased production of ovarian androgens. Adrenal androgens are also increased, including 11-oxygenated androgens which are activated peripherally by renal 11β-hydroxysteroid dehydrogenase type 2 (HSD11B2) and aldo-keto reductase 1C3 (AKR1C3) in adipocytes. Steroid-5α-reductase (SRD5A) converts 11-ketotestosterone to 11-ketodihydrotestosterone. Excess levels of androgens stimulate deposition of abdominal adipose tissue which subsequently increases insulin resistance and hyperinsulinism. Hyperinsulinism stimulates AKR1C3 activity, increases androgen production from the ovaries (by its action as a co-gonadotrophin) and adrenal cortex, reduces production of hepatic sex hormone binding globulin, and inhibits progesterone mediated negative feedback onto GnRH neurons, worsening androgen excess in a vicious cycle. (Right) Kisspeptin, neurokinin B, and dynorphin A neurons (KNDy neurons) act in a paracrine and autocrine way to regulate release of kisspeptin onto GnRH neurons and consequent GnRH pulsatility. Neurokinin B binds to neurokinin 3 receptors (NK3R) to stimulate release of kisspeptin whereas dynorphin binds to kappa opioid receptors to inhibit kisspeptin release. γ-aminobutyric acid (GABA) and anti-müllerian hormone (AMH) bind to GABAA receptors (GABAAR) and AMH receptor type 2 (AMHR2), respectively, to stimulate GnRH pulsatility. Impaired negative feedback from oestradiol and progesterone is seen at the level of the hypothalamus. Neuroendocrine abnormalities in the control of these components are shown in red. OR=oestrogen receptor; PR=progesterone receptor
Neurokinin B and dynorphin are expressed by KNDy neurons and act in an autocrine and paracrine way to control release of kisspeptin (figure 1). Neurokinin B preferentially binds to the neurokinin 3 receptor (encoded by TACR3) to stimulate gonadotrophin releasing hormone pulsatility.4 23 Unlike KISS1 null mice, mice deficient in components of neurokinin B signalling can still generate surges in luteinising hormone and conceive, suggesting that compensatory pathways exist which contribute to the generation of kisspeptin and gonadotrophin releasing hormone pulses.17 24 25 This milder effect of neurokinin B blockade might avoid excessive reduction in gonadotrophin releasing hormone pulsatility, making it an attractive target for treatment.4 Dynorphin, which activates kappa opioid receptors on KNDy neurons to inhibit secretion of gonadotrophin releasing hormone,22 26 has been shown to mediate progesterone negative feedback on gonadotrophin releasing hormone neurons in sheep27 and humans.22 28
Neuronal activity of gonadotrophin releasing hormone is also regulated by other substances, including γ-aminobutyric acid (GABA) and anti-müllerian hormone, both of which stimulate gonadotrophin releasing hormone neurons directly. GABA exerts an excitatory effect on gonadotrophin releasing hormone neurons through GABAA receptors, and GABA levels in cerebrospinal fluid can be raised in patients with polycystic ovary syndrome.29 Anti-müllerian hormone is secreted by ovarian granulosa cells, where raised levels in women with polycystic ovary syndrome disrupt folliculogenesis and ovulation.30 Anti-müllerian hormone might also have neuroendocrine effects: 50% of gonadotrophin releasing hormone neurons in mice and humans express anti-müllerian hormone receptor type 2,31 with studies implicating anti-müllerian hormone in neuronal migration of gonadotrophin releasing hormone,32 gonadotrophin releasing hormone pulsatility, and secretion of luteinising hormone.30
Classical pathway of androgen synthesis
High levels of androgens is a primary defect in polycystic ovary syndrome. Cholesterol is converted to androgens by a cascade of enzymes common to all steroid producing organs, with tissue specific variations resulting in different steroid hormone profiles.33 In polycystic ovary syndrome, increased production of ovarian androgens by the classical pathway is driven by increased secretion of pituitary luteinising hormone, the action of insulin as a co-gonadotrophin, and increased thecal cell hypersensitivity to luteinising hormone.34–36 Figure 2 summarises the classical pathway of steroidogenesis. Through a sequence of reactions, cholesterol is converted to dehydroepiandrosterone, which is then converted to androstenedione by 3β-hydroxysteroid dehydrogenase type II and subsequently to testosterone by aldo-keto reductase type 1C3 (AKR1C3).35
Figure 2Classical pathway of androgen synthesis. Luteinising hormone stimulates the classical pathway of androgen synthesis in ovarian theca cells. Cholesterol is transported to the inner mitochrondrial membrane by steroidogenic acute regulatory protein (StAR). A cleavage system of the cytochrome P450 enzyme, CYP11A1, ferrodoxin, and ferrodoxin reductase converts cholesterol to pregnenolone. Expression of CYP11A1 is stimulated by activation of the luteinising hormone receptor. Pregnenolone is transported to smooth endoplasmic reticulum where it is converted to 17-hydroxypregnenolone and subsequently to dehydroepiandrosterone by the 17-hydroxylase and 17,20-lyase subunit of the CYP17A1 enzyme, respectively. Dehydroepiandrosterone is then converted to androstenedione or androstenediol and subsequently to testosterone by a combination of 3β-hydroxysteroid dehydrogenase type II (HSD3B2) and aldo-keto reductase type 1C3 (AKR1C3). 17β-hydroxysteroid dehydrogenase 1 (HSD17B1) also catalyses the conversion of dehydroepiandrosterone to androstenediol. HSD3B2 converts pregnenolone and 17-hydroxypregnenolone to progesterone and 17-hydroxyprogesterone, respectively, which are substrates for a back door alternative pathway of androgen synthesis. Androstenedione and testosterone diffuse into granulosa cells where they are converted to oestrogens by the action of aromatase (CYP19A1), under the control of follicle stimulating hormone receptor activation. Testosterone can be converted to dihydrotestosterone by steroid 5α-reductase (SRD5A) in peripheral tissues
Increased activity of ovarian 11β-hydroxysteroid dehydrogenase type 1 (11β-HSD1), which converts inactive cortisone to active cortisol, might also have a role in the pathogenesis of polycystic ovary syndrome.37 Overexpression of ovarian 11β-HSD1 in rats caused polycystic ovarian morphology, oestrous cycle, and reproductive hormone abnormalities.37 Although 11β-HSD1 is widely expressed, dysregulation seems to be tissue specific, because hepatic 11β-HSD1 activity is impaired and expression of 11β-HSD1 in subcutaneous adipose tissue is increased in patients with polycystic ovary syndrome.38 Raised circulating levels and ovarian expression of vascular endothelial growth factor also contribute to the hypervascular, hyperplastic appearance of the ovarian stroma and theca interna in polycystic ovary syndrome, and might contribute to increased ovarian androgen synthesis.39
Androgen synthesis in adrenal glands and peripheral tissues
Polycystic ovary syndrome was previously thought to be primarily a disease of excess production of androgens in the ovaries, but the adrenal glands and peripheral tissues are now considered important sources of androgens in patients with polycystic ovary syndrome. Increased concentrations of dehydroepiandrosterone sulphate, an almost exclusive product of the adrenal cortex,40 are apparent in 20-30% of patients with polycystic ovary syndrome.41 This finding seems to be the result of increased secretory activity of the adrenal cortex because no change in pituitary responsiveness to corticotrophin releasing hormone or reduction in the minimal stimulatory dose of adrenocorticotropic hormone required for adrenal hormone production is seen.42 Changes in steroidogenesis, such as increased enzymatic activity of the 17-hydroxylase subunit of the cytochrome P450 enzyme, CYP17A1, might account for this hyper-responsiveness.43
Other adrenal androgens are also secreted in excess, including 11β-hydroxyandrostenedione and 11β-hydroxytestosterone.5 44 The adrenal androgen 11β-hydroxyandrostenedione is abundant and was previously thought to have little physiological importance because of its weak androgenic activity. Recent studies, however, have shown that 11β-hydroxyandrostenedione can be metabolised to 11-ketotestosterone and 11-ketodihydrotestosterone, termed 11-oxygenated androgens, because of the presence of an oxygen atom on carbon 11.45 Both 11-ketotestosterone and 11-ketodihydrotestosterone bind to androgen receptors with similar affinity and potency to testosterone and dihydrotestosterone.46 47 Mass spectrometry analyses have shown that 11-oxygenated androgens are the dominant circulating androgens in women with polycystic ovary syndrome and correlate substantially with markers of metabolic risk.5 The synthesis of 11-oxygenated androgens is reliant on the peripheral activation of adrenal derived androgens (figure 3). 11β-hydroxysteroid dehydrogenase type 2 is an enzyme expressed by the kidney that converts 11β-hydroxyandrostenedione to 11-ketoandrostenedione, and 11β-hydroxytestosterone to 11-ketotestosterone.45 Adipose tissue also has enzymes responsible for potent androgen formation, however, and might represent the dominant source of circulating 11-oxygenated androgens.45 48
Expression of the androgen activating enzyme, AKR1C3, in subcutaneous adipose tissue is increased in women with polycystic ovary syndrome compared with matched controls.6 49 Thus concentrations of androgens in adipose tissue are increased in women with polycystic ovary syndrome, accompanied by inhibition of lipolysis and increased de novo lipogenesis.6 These observations suggest that inhibition of AKR1C3 might be an attractive therapeutic target in patients with polycystic ovary syndrome.
Figure 3Pathway for 11-oxygenated androgen synthesis, which begins in the adrenal cortex. Androstenedione and testosterone are produced by the classical pathway (figure 2). Dehydroepiandrosterone is diverted to downstream androgens or sulphonated to dehydroepiandrosterone sulphate by the sulphotransferase, SULT2A1. Androstenedione and testosterone are hydroxylated by 11β-hydroxylase (CYP11B) to produce abundant 11β-hydroxyandrostenedione (11OHA4) and smaller amounts of 11β-hydroxytestosterone (11OHT). Renal 11β-hydroxysteroid dehydrogenase type 2 (HSD11B2) converts 11OHT to 11-ketotestosterone (11KT) and 11OHA4 to 11-ketoandrostenedione (11KA4). In adipose tissue, 11KA4 is metabolised to 11KT and 11-ketodihydrotestosterone (11DHKT) by aldo-keto reductase type 1C3 (AKR1C3) and steroid-5α-reductase (SRD5A), respectively. 11OHA4 is metabolised to 11OHT and 11β-hydroxydihydrotestosterone (11OHDHT) by 17β-hydroxysteroid dehydrogenase 2 (HSD17B2) and SRD5A, respectively. 11KT and 11KDHT are potent agonists of the androgen receptor whereas 11OHT and 11OHDHT have milder potency. StAR=steroidogenic acute regulatory protein; HSD3B2=3β-hydroxysteroid dehydrogenase type II; CYP11A1, CYP17A1, CYP11B1=cytochrome P450 enzymes
Hyperinsulinism
Insulin resistance, and the consequent hyperinsulinism, have an important role in driving androgen synthesis in many endocrine tissues. Insulin acts as a co-gonadotrophin in the ovaries,36 impairs progesterone mediated inhibition of the gonadotrophin releasing hormone pulse generator,50 and facilitates synthesis of androgens in the adrenal glands by increasing adrenocorticotropic hormone stimulated steroidogenesis.51 AKR1C3 expression and activity in adipocytes is increased by insulin, contributing to increased synthesis of androgens in adipocytes in polycystic ovary syndrome.52 Insulin also inhibits sex hormone binding globulin, facilitating hyperandrogenism by increasing the percentage of free biologically active androgens.53 Excess production of androgens then stimulates hyperinsulinism, leading to a vicious cycle between androgen and insulin excess.7 54 Several studies have also implicated hyperandrogenism in the accumulation of abdominal and visceral adipose tissue in polycystic ovary syndrome55 56; this hyperandrogenism further drives insulin resistance and consequent production of androgens (figure 1).
In common with hyperandrogenism, insulin resistance is not a universal feature of polycystic ovary syndrome, although a systematic review of hyperinsulinaemic-euglycaemic clamp studies of 1224 women with polycystic ovary syndrome and 741 controls showed that insulin sensitivity was lower in women with polycystic ovary syndrome than in controls (mean effect size −27%, 99% confidence interval −21 to −33).57 Studies exploring steroid metabolomics in patients with polycystic ovary syndrome might give more information. One such cross sectional study (n=488) combining machine learning with mass spectrometry multisteroid profiling has identified three distinct groups of patients based on the predominant source of androgens.58 These subgroups have distinct steroid metabolomes and risk of metabolic complications: a gonadal derived classical androgen excess group, an adrenal derived androgen excess group (comprising 11-oxygenated androgens), and a group with comparably mild androgen excess.58 The adrenal derived androgen group had the highest rates of hirsutism, insulin resistance, and type 2 diabetes. These insights challenge our understanding of polycystic ovary syndrome as one entity and might prompt a reconsideration of the classification of the disease based on the metabolomic signature.
Changes in adipocyte structure and function
Changes in white adipose tissue morphology and function is seen in women with polycystic ovary syndrome, including enlarged adipocytes, reduced lipoprotein lipase activity,59 and increased secretion of proinflammatory cytokines.60 The function of brown adipose tissue might also be disrupted because women with polycystic ovary syndrome showed reduced postprandial thermogenesis compared with controls matched for body mass index.61 This defect could be driven by androgen excess, because prenatally androgenised sheep have reduced postprandial thermogenesis in adulthood,62 accompanied by reduced adipose expression of thermogenic uncoupling proteins and sympathetic activity. Adolescent prenatally androgenised sheep also showed reduced hepatic expression and circulating levels of fibroblast growth factor 21,63 a hormone that regulates adipocyte function, insulin sensitivity, and energy balance. Targeting expression of fibroblast growth factor 21 during an appropriate period in development might be a therapeutic option.
Gut microbiota and bile acid metabolism
Recent studies have implicated changes in the gut microbiome in the pathogenesis of polycystic ovary syndrome. Women with polycystic ovary syndrome have higher intestinal levels of Bacteroides vulgatus and lower levels of glycodeoxycholic acid and tauroursodeoxycholic acid.64 Oral gavage of wild-type mice with faecal microbiota from individuals with polycystic ovary syndrome or pure B vulgatus caused insulin resistance, changes in bile acid metabolism, reduced secretion of interleukin 22, and disrupted oestrous cycle and ovarian morphology.64 Administration of interleukin 22 or glycodeoxycholic acid to mice treated with B vulgatus improved insulin sensitivity, testosterone levels, and oestrous cycles. Hence modifying the gut microbiota or bile acid metabolism, increasing levels of interleukin 22, or a combination of these actions, might be therapeutically valuable in polycystic ovary syndrome.64
Insights from genome-wide association studies
Genome-wide association studies have identified numerous susceptibility loci for polycystic ovary syndrome, including 11 in Han Chinese populations,65 66 eight in European populations,67 68 and eight in a Korean population.69 Robust candidate susceptibility loci are near genes belonging to metabolic (insulin receptor (INSR), insulin gene-variable number of tandem repeats (INS-VNTR), and DENN domain containing protein 1A (DENND1A))70 and neuroendocrine (follicle stimulating hormone receptor, luteinising hormone receptor, and thyroid adenoma associated (THADA)) pathways.70 Meta-analyses of genome-wide association studies have shown that the genetic architecture of polycystic ovary syndrome is consistent across different diagnostic criteria and ethnic groups.71 72 These observations indicate a shared ancestry for polycystic ovary syndrome and reinforce the importance of neuroendocrine and metabolic pathways in the pathogenesis of the disease.
Developmental programming
Genetic loci identified by genome-wide association studies currently account for only 10% of the known heritability (about 70%) of polycystic ovary syndrome,73 74 suggesting other influences on the pathogenesis of the disease. Emerging evidence indicates that polycystic ovary syndrome might have its origins in utero, and thus could be subject to developmental programming and epigenetic modifications. Prenatal exposure to androgens in several preclinical models caused a permanent polycystic ovary syndrome-like phenotype postnatally.75–77 A programming effect might also persist transgenerationally, because pregnant mice treated with dihydrotestosterone produced female offspring with polycystic ovary syndrome-like phenotypes from the first to the third generations of offspring.78 Cautious interpretation is needed, however, because these models might not accurately reflect the human phenotype. Anti-müllerian hormone might also be involved in in utero programming: levels of anti-müllerian hormone increased significantly in pregnant women with polycystic ovary syndrome (P<0.001), and use of this hormone caused gonadotrophin releasing hormone neuronal hyperactivity and androgen excess in pregnant mice.79 Epigenetic mechanisms might also be involved in mediating susceptibility to polycystic ovary syndrome, with differential methylation patterns and microRNA expression detected in adipose tissue and ovarian tissue of patients with polycystic ovary syndrome compared with controls.80