The Pathophysiology of Polycystic Ovary Syndrome

Tasoula Tsilchorozidou; Caroline Overton; Gerard S. Conway


Clin Endocrinol. 2004;60(1) 

In This Article


Despite being one of the most common endocrinopathies, a comprehensive explanation of pathophysiology is still lacking. The heterogeneity of PCOS may well reflect multiple pathophysiological mechanisms, but the definition of each contributing mechanism has been slow to emerge. Traditionally, it has been useful to consider the polycystic ovary syndrome as the result of a 'vicious cycle', which can be initiated at any one of many entry points. Altered function at any point in the cycle leads to the same result: ovarian androgen excess and anovulation.

Several theories have been proposed to explain the pathogenesis of PCOS:

  • A unique defect in insulin action and secretion that leads to hyperinsulinaemia and insulin resistance.

  • A primary neuroendocrine defect leading to an exaggerated LH pulse frequency and amplitude.

  • A defect of androgen synthesis that results in enhanced ovarian androgen production.

  • An alteration in cortisol metabolism resulting in enhanced adrenal androgen production.

It must be accepted, however, that each of these are artificial stating points to our understanding of the metabolic-ovarian-pituitary circuitry being closely interrelated.

The first recognition of an association between glucose intolerance and hyperandrogenism was made by Achard & Thiers (1921) and was called 'the diabetes of bearded women'. The association between increased insulin resistance and PCOS is now well recognized. Insulin resistance is defined as a reduced glucose response to a given amount of insulin. There are several mechanisms contributing to the state of insulin resistance: peripheral target tissue resistance, decreased hepatic clearance, or increased pancreatic sensitivity. Studies with the euglycaemic clamp technique indicate that insulin resistance is a common feature of the syndrome, and both obese and nonobese women with the syndrome are more insulin-resistant and hyperinsulinaemic than age- and weight-matched normal women (Chang et al., 1983; Dunaif et al., 1989, 1992). However, obese PCOS women had significantly decreased insulin sensitivity compared with nonobese PCOS women (Dunaif, 1995). For example, Morales et al. (1996) demonstrated reduced insulin sensitivity in lean PCOS compared with lean controls, a further decrease in obese controls and a twofold further reduction in obese PCOS, suggesting that obesity is additive to insulin resistance related to PCOS. Consistent with the degree of insulin resistance, the manifestation of compensatory hyperinsulinaemia in lean PCOS women was incipient, being evident only in response to meals. Collectively, these observations indicate that insulin resistance is a common finding in women with PCOS independent of obesity and that insulin resistance in obese PCOS is composed of dual contributions, one unique to PCOS and the other obesity-specific.

Not all studies, however, have shown insulin resistance in lean PCOS subjects. Reports using an intravenous glucose tolerance test (Herbert et al., 1990), a continuous glucose infusion model (Dale et al., 1992) or a hyperinsulinaemic euglycaemic clamp (Ovesen et al., 1993) found normal insulin action in normal weight PCOS patients. There are several possible reasons for these discrepancies which an any case common in PCOS research papers. The composition of study groups varies depending on local diagnostic criteria; there is no standard definition and assessment of insulin resistance; ethic variations in in central adiposity, insulin sensitivity and the prevalence of a positive family history of type 2 diabetes are rarely taken into account; many studies with a small numbers of subjects may contain type 2 statistical errors.

In addition to decreased insulin sensitivity, pancreatic β-cell secretory dysfunction has been reported in PCOS (Ehrmann et al., 1995a; Dunaif & Finegood, 1996a). The β-cell defect - increased secretion of insulin under basal conditions and decreased secretion after meals - results in insufficient insulin secretion to compensate for the degree of insulin resistance. The decreased postprandial secretory responses in these patients resembles the β-cell dysfunction of type 2 DM and are much more pronounced in PCOS women who have a first-degree relative with type 2 DM, suggesting an increased risk for developing glucose intolerance. Weight loss results in significantly improved insulin resistance, but the β-cell defect remains (Holte et al., 1995), suggesting that it may be the primary abnormality in PCOS. However, Rodin et al. (1998) have showed that polycystic ovaries unlike type 2 DM, were not associated with a defect in the secretion of insulin. Finally, a reduction in the insulin clearance rate due to decreased hepatic insulin extraction has been reported to be partially responsible for the elevations in insulin concentration by some investigators (Mahabeer et al., 1989; O'Meara et al., 1993).

Several groups have focus on mechanisms of insulin signalling in order to define the pathogenesis of insulin resistance in PCOS. Insulin action is mediated through a protein tyrosine kinase receptor. Tyrosine autophosphorylation increases the insulin receptor's tyrosine kinase activity, whereas serine phosphorylation inhibits it. The tyrosine-phosphorylated insulin receptor phosphorylates intracellular substrates, such as IRS-1 and IRS-2 (insulin receptor substrate-1 and -2) initiating signal transduction and the plieotropic actions of insulin. A potential mechanism for insulin resistance in at least 50% of PCOS women appears to be related to excessive serine phosphorylation of insulin receptor. A factor extrinsic to the insulin receptor, presumably a serine/threonine kinase, causes serine phosphorylation of the insulin receptor, leading to inhibition of signalling (Dunaif et al., 1995; Dunaif, 1997). It is believed that the defect in insulin action is limited to glucose metabolism, whereas other biologic actions of insulin - including those involved in steroidogenesis - are not impaired. Interestingly, serine phosphorylation of IRS-1 appears to be the mechanism of TNF-α-mediated insulin resistance of obesity (Rosen & Spiegelman, 1999). Serine phosphorylation also appears to modulate the activity of the key regulatory enzyme of androgen biosynthesis, P450c17, present in both the adrenal and ovarian steroidogenic tissue. Thus, serine phosphorylation has been shown to increase enzyme activity and androgen synthesis (Zhang et al., 1995). It is therefore possible that a single defect - serine phosphorylation - produces both the insulin resistance and the hyperandrogenism in a subgroup of PCOS women (Fig. 1).

Pathways leading to androgen excess in PCOS.

To further evaluate the postbinding defect in insulin action in PCOS, Book & Dunaif (1999) examined the metabolic and mitogenic actions of insulin and IGF-I in cultured skin fibroblasts from PCOS and control women. They concluded that: (a) there is a selective defect in insulin action in PCOS fibroblasts that affects metabolic, but not mitogenic signalling pathways, (b) that there is a similar defect in IGF-I action, suggesting that insulin and IGF-I stimulate glycogen synthesis by the same postreceptor pathways, and (c) that insulin receptor substrate-1-associated phosphatidylinositol 3-kinase activation by insulin and IGF-I is similar to the control value, suggesting that the metabolic signalling defect is in another pathway or downstream of this signalling step in PCOS fibroblasts.

The hypothesis for the existence of a postreceptor defect in insulin action in PCOS is also consistent with reports of molecular studies found no structural abnormality in the insulin receptor (Conway et al., 1994; Sorbara et al., 1994; Talbot et al., 1996).

The Association Between Hyperinsulinaemia and Hyperandrogenism. Several studies have demonstrated a positive correlation between fasting insulin levels and androgen levels (Burghen et al., 1980; Lobo et al., 1983). Furthermore, the severity of hyperinsulinaemia correlates with the degree of clinical expression of the syndrome (Conway et al., 1990; Robinson et al., 1993). Whether hyperandrogenism results from the hyperinsulinaemia of insulin resistance, or vice versa, has been debated since this correlation was demonstrated. Most of the evidence supports hyperinsulinaemia as the primary factor, especially the experiments in which decreasing the hyperandrogenaemia by bilateral oophorectomy (Nagamani et al., 1986) or the administration of a GnRH-agonist (Geffner et al., 1986; Dunaif et al., 1990) has not demonstrated changes in the hyperinsulinaemic state in PCOS. Diamanti-Kandarakis et al. (1995) have also reported that antiandrogen therapy did not alter insulin sensitivity in PCOS. It is possible, however, that androgens may contribute to some extent to the associated insulin resistance of PCOS, as some investigators have found that insulin resistance was partially reversed during androgen suppression (Elkind-Hirsch et al., 1993) or antiandrogen treatment (Moghetti et al., 1996a). In summary, these findings indicate that endogenous androgens do not play a central pathophysiologic role in sustaining insulin resistance in women with PCOS and that disordered insulin action precedes the increase in androgens.

It is generally accepted that hyperinsulinaemia augments androgen production in PCOS (Fig. 1). Insulin may act:

  • directly, as a cogonadotrophin augmenting LH activity through stimulation of ovarian receptors of insulin and insulin-like growth factors;

  • indirectly, by enhancing the amplitude of serum LH pulses.

Despite evidence that insulin promotes ovarian androgen production in PCOS, the exact mechanism on cellular level remains unclear. Initially, insulin cross-reaction with the IGF-I receptor - similar in structure to insulin receptor - on the ovarian thecal cells (Bergh et al., 1993) was proposed as a possible mechanism of insulin-mediated hyperandrogenism. In view of the known actions of IGF-I in augmenting the thecal androgen response to LH (Cara et al., 1990), activation of IGF-I receptors by insulin would lead to increased androgen production in thecal cells. However, insulin has been shown to bind the IGF-I receptor with an affinity of 50-500 times lower than that of IGF-I. The cross-over effect of insulin with the type-I IGF receptor therefore is an important consideration at high insulin concentrations (Jacobs, 1990). The existence of hybrid and atypical insulin/IGF-I receptors - which consist of a combination of α- and β-subunits of both receptors - have also been described (Siddle et al., 1994). It is believed that these receptors can bind insulin and IGF-I with similar affinity. Also, it has been proposed that insulin has specific actions on steroidogenesis acting through its own receptor (Poretsky et al., 1985), a pathway supported by in vitro studies of both granulosa (Willis & Franks, 1995; Willis et al., 1996) and thecal cells (Barbieri et al., 1984; Nestler et al., 1998a). Interestingly, Willis & Franks (1995), using anti-insulin receptor and antitype-I IGF receptor antibodies, not only demonstrated that insulin effects on human granulosa cell steroidogenesis in vitro must be mediated via its own receptor, but also excluded both the insulin/type-I IGF hybrid receptor or the type-I IGF receptor as possible insulin action-mediated receptors.

Preliminary studies suggest that insulin enhances the amplitude of LH pulses but not their frequency in obese women with PCOS (Nestler & Jakubowicz, 1996; Nestler, 1997). This is consistent with previous report by Yen et al. (1993) that there is a general concordance of the diurnal pattern of LH levels and that of insulin levels in these women. Studies in which insulin levels have been suppressed by insulin-lowering agents suggest that insulin might also contribute to changes in ovarian androgen secretion through effects at the pituitary level. An alternative possibility is that the reduced secretion of LH with these therapies could be secondary to high progesterone levels following ovulation. Finally, insulin receptors have been identified in human pituitary tissue (Unger et al., 1991) and insulin found to stimulate gonadotropin release in vitro, at least in rats (Adashi et al., 1981).

Nestler et al. (1996) proposed insulin-mediated increase of ovarian cytochrome P450c17a activity, as an additional mechanism of insulin action in a subgroup of obese PCOS women. In this study, a decrease in serum insulin concentrations with metformin, followed by a reduction of ovarian P450c17a activity, as demonstrated by a substantial reduction in the response of serum 17α-hydroprogesterone to the administration of leuprolide. An analogous report by Moghetti et al. (1996b) that hyperinsulinaemia may stimulate cytochrome P450c17a activity in another steroidogenic tissue of women with PCOS - the andrenal gland - further supports this hypothesis.

There are two other important actions of insulin which contribute to hyperandrogenism in PCOS: the inhibition of hepatic synthesis of serum sex hormone-binding globulin (SHBG), which allows more free androgen and oestrogen to be bioavailable, and the inhibition of hepatic production of IGFBP-1, which allows an increase in circulating levels of IGF-I and greater local activity (Bach, 1999; Le Roith et al., 1992). This is now known to be the mechanism for the frequently observed inverse correlation between peripheral insulin and SHBG levels. This relationship is so strong that SHBG concentrations are good markers for hyperinsulinaemic insulin resistance, and reduced SHBG concentration is a predictor for the development of type 2 DM (Lindstedt et al., 1991; Nestler, 1993).

The clinical implication of these findings is that amelioration of hyperandrogenism in women with PCOS may be achieved by interventions that improve insulin sensitivity and reduce circulating insulin levels. Indeed, weight loss in women with PCOS improves the endocrine and ovarian dysfunction (Kiddy et al., 1992; Guzick et al., 1994), while the pharmacological approach with agents that either decrease insulin secretion, like diazoxide (Nestler et al., 1989) or somatostatin (Prelevic et al., 1990), or that improve insulin sensitivity, like metformin (Velazquez et al., 1994, 1997; Nestler & Jakubowicz, 1996; Ehrmann et al., 1997a), or troglitazone (Dunaif et al., 1996; Ehrmann et al., 1997b) have demonstrated conclusively that a reduction in serum insulin levels is associated with a reduction of ovarian androgen secretion in PCOS. The improvement of serum androgen levels with different drug classes, with different mechanism of action, suggests an effect mediated by reduction in circulating insulin levels, although a direct ovarian effect especially of the insulin-sensitizing agents can not be excluded. These changes were independent of changes in body weight, although for metformin there still exists some controversy.

The insulin-sensitizing agents, metformin and troglitazone, not only reduce circulating insulin concentrations, but also reverse the metabolic and endocrine anomalies (decreased androgens, increased SHBG, decreased PAI-1 consistent with improved fibrinolytic capacity and decreased LH), and more recently, restoring menstrual abnormalities and improving the reproductive outcome in anovulatory PCOS women (Nestler et al., 1998b; Mitwally et al., 1999; Pirwany et al., 1999; Moghetti et al., 2000; Vandermolen et al., 2001). We are currently at a stage where the few randomized controlled studies that exist include small numbers of subjects and many reports use inferior retrospective/prospective comparisons to evaluate the efficacy of metformin. Additional questions that emerge are: what predicts a good response to metformin - the effect can vary greatly; how safe is metformin in pregnancy and lactation; could metformin reduce the higher prevalence of miscarriage or gestational diabetes in PCOS? Preliminary, uncontrolled studies have shown that continuing metformin throughout pregnancy in women with PCOS appears to safely reduce first-trimester spontaneous abortion (Glueck et al., 2001) as well as the development of gestational diabetes (Glueck et al., 2002). In a subsequent retrospective study, Jakubowicz et al. (2002) also reported reduced incidence of first-trimester pregnancy loss after the administration of metformin in women with PCOS. While these studies infer a degree of safety of metformin throughout pregnancy, as no adverse fetal outcomes have been noted among women treated with metformin, it must be noted that formal safety data are lacking. It should also be noted that lifestyle measures are probably more effective than metformin in addressing hyperinsulinaemia if we extrapolate from the experience in preventing type 2 DM (Rubin et al., 2002; Diabetes Prevention Programme).

If insulin resistance and hyperinsulinaemia have an important pathogenetic role in PCOS, why are not all patients with hyperinsulinaemia also hyperandrogenic; like many women with type 2 DM? Furthermore, how is it that ovaries appear to be insulin-responsive in an insulin-resistant state? Indeed, does insulin activate a signalling system separate from glucose transport to stimulate steroidogenesis? Conn et al. (2000) have shown that although 82% of women with type 2 DM had polycystic ovaries in ultrasound, only 52% had clinical evidence of hyperandrogenism and/or menstrual disturbance, suggesting that hyperinsulinaemia alone is not sufficient for the expression of the syndrome. In another study in a more genetically homogenous group of Asian women, Rodin et al. (1998) reported that the effects of type 2 DM and polycystic ovaries on insulin-sensitivity were independent, suggesting that these changes in insulin sensitivity involve different mechanisms. It is possible that the insulin resistance and the reproductive disturbances reflect separate genetic defects and that insulin resistance unmasks the syndrome in genetically susceptible women. In general, studies strongly suggest that only women with the endocrine syndrome of hyperandrogenism and chronic anovulation appear to be insulin-resistant (Dunaif et al., 1987; Robinson et al., 1993; Sampson et al., 1996). Ovulatory women with hyperandrogenism or polycystic ovaries morphology in ultrasound are not insulin-resistant.

Glucose Intolerance and Diabetes. Insulin resistance and β-cell dysfunction are both known to precede the development of glucose intolerance and type 2 DM (Reaven, 1988). Thus, PCOS women would be predicted to be at an increased risk for type 2 DM. Dunaif et al. (1987) have originally reported that up to 40% of obese PCOS women had impaired glucose tolerance or frank type 2 DM, when the WHO criteria are used. These prevalence rates are substantially higher than those found in a major population-based studies in women of similar age (10·3% by WHO criteria, Harris et al., 1987). In a recent, large, prospective study of 254 affected women, Legro et al. (1999) have documented that women with PCOS are at significantly increased risk for glucose intolerance (31·1% IGT) and type 2 DM (7·5% undiagnosed diabetes) compared to concurrently studied age-, weight- and ethnicity-matched controls of the reproductive age (7·8% IGT; 1% undiagnosed diabetes, WHO criteria). Furthermore, they found that nonobese PCOS women may also have glucose intolerance (10·3% IGT; 1·5% diabetes). Not surprising, women with PCOS are more likely to develop gestational diabetes (Lanzone et al., 1995). Finally, consistent with these findings, a study in postmenopausal women with a history of PCOS found a 15% prevalence of type 2 DM (Dahlgren et al., 1992).

Cardiovascular Disease. Insulin resistance is considered to be a risk factor for coronary heart disease as it is associated with impaired glucose tolerance and type 2 DM (Reaven, 1988), hypertension (Ferrannini et al., 1987), abdominal obesity and adverse lipid profiles (Orchard et al., 1983), all features of the so-called 'metabolic syndrome X'.

Retrospective studies suggest that there is an association between PCOS and cardiovascular disease. Wild et al. (1990) studied 102 pre- and postmenopausal women undergoing cardiac catheterization for the investigation of chest pain. Arterial lesions were seen in 52 women and these women were more likely to report hirsutism, diabetes mellitus, hypertension and previous coronary artery disease. Birdsall et al. (1997) carried out a pelvic ultrasound scan for polycystic ovary morphology on 143 women undergoing cardiac catheterization. They reported no significant difference in the prevalence of polycystic ovaries in women with coronary artery lesions and normal arteries. However, women with PCOS had more affected segments. These studies suggest that there may be some association between PCOS and coronary artery disease risk. However, the studies are retrospective, contain a mixed group of pre- and postmenopausal women, are not controlled for obesity and do not have rigorous criteria for the PCOS. Guzick et al. (1996) studied atherosclerosis at a different site. They measured carotid artery intimal thickness by ultrasound in women with PCOS. The carotid artery intima-media thickness was significantly increased in women with PCOS but there was no significant difference in the number of women with atherosclerotic plaques.

Haemodynamic changes have also been reported in women with PCOS. In two consecutive studies, Prelevic et al. (1995, 1996) reported lower cardiac flow velocity, higher resting forearm flow during reactive hyperaemia and lower incremental forearm flow in PCOS than in age-matched control women. In a subsequent study, Lees et al. (1998) reported a constrictor response to transdermal glyceryl trinitrate - a potent vasodilator which acts through the endothelial nitric oxide system - on uterine artery Doppler velocimetry in women with PCOS. Furthermore, Lakhani et al. (2000) reported a paradoxical constrictor response to 5% carbon dioxide, acting as a cerebrovasodilator, in the internal carotid artery in women with PCOS compared with women with normal ovaries. More recently, Paradisi et al. (2001) showed a positive correlation between abnormal endothelial function and testosterone levels in hyperandrogenic insulin-resistant women with PCOS, an association which was stronger than that of insulin sensitivity. All of these findings probably represent an abnormality in endothelial function in women with PCOS and are indicative of widespread changes in cardiovascular function in these women.

Finally, despite the strong association between PCOS with cardiovascular risk factors, a long-term follow-up study of 786 women diagnosed with PCOS (Pierpoint et al., 1998) did not show an excess mortality from circulatory disease. In follow-up, Wild et al. (2000) reported no excess of coronary heart disease mortality or morbidity among middle-aged women with a history of PCOS, despite increased prevalence of several cardiovascular risk factors. However, mortality and morbidity from diabetes and risk of nonfatal cerebrovascular disease were higher among women with PCOS. The reason for the discrepancy between prevalence of cardiovascular risk factors and expected prevalence of cardiovascular disease is unknown. It is interesting to note the prevalence of a positive family history of type 2 DM among women with PCOS compared with controls in this study. Also, it has been suggested that protective mechanisms may be operative in PCOS such as prolonged exposure to unopposed oestrogen or elevated levels of vascular endothelial growth factor (VEGF). Indeed, increased serum VGEF concentrations have been reported in women with polycystic ovaries and PCOS recently (Agrawal et al., 1998; Tulandi et al., 2000).

Cardiovascular Risk Factors. Although a positive relationship between insulin and blood pressure has been demonstrated in many populations (Bonora et al., 1987; Zavaroni et al., 1989; Pollare et al., 1990) it is possible that this association does not exist in PCOS: women with PCOS do not appear to be hypertensive compared to control subjects matched for body composition, even if they have significant insulin resistance (Zimmerman et al., 1992; Sampson et al., 1996). Total plasma renin levels, however, were found to be higher in normotensive women with PCOS than in healthy women, independent of the degree of insulin resistance (Hacihanefioglu et al., 2000). Additionally, in a retrospective long-term study of postmenopausal women with a history of PCOS, Dahlgren et al. (1992) have found a significant higher rate of hypertension (39%) compared to controls (11%). This may represent a bias of the study design, an effect of obesity or a longer-term impact of blood pressure not apparent in a younger healthier population.

Women with PCOS would be expected to be at high risk for dyslipidaemia due to elevated androgen levels, body fat distribution and hyperinsulinaemic insulin resistance. A number of studies have shown that women with PCOS exhibit an abnormal lipoprotein profile characterized by raised concentrations of plasma triglycerides, marginally elevated low-density lipoprotein (LDL) cholesterol, and reduced high-density lipoprotein (HDL) cholesterol. Recently, two studies have shown that women with PCOS have an atherogenic lipoprotein profile characterized not only by the above mentioned abnormalities but also by raised concentrations and proportions of atherogenic small, dense LDL-III relative to body mass index (BMI)-matched controls (Dejager et al., 2001; Pirwany et al., 2001). Furthermore, an increased hepatic lipase activity has been documented. These metabolic disturbances appear to be related more closely to adiposity/insulin metabolism than to circulating androgen levels. Indeed, Wild et al. (1990) found that suppressing androgen levels does not alter lipid profile in PCOS. In contrast, Diamanti-Kandarakis et al. (1998) have demonstrated for the first time that treatment with a pure androgen receptor blocker, flutamide, improves the lipid profile and that this effect may be due to direct inhibition of androgenic actions.

PCOS women have been also found to have increased circulating levels of plasminogen activator inhibitor, PAI-1 (Sampson et al., 1996). Elevated PAI-1 activity levels are linked both to the insulin resistance state (Dawson & Henney, 1992) and to increased risk of thrombotic vascular events (Hamsten et al., 1987), being an independent risk factor for atherosclerosis. In PCOS, these levels decreased with improvement in insulin sensitivity mediated by weight loss (Andersen et al., 1995) or insulin-sensitizing agents (Ehrmann et al., 1997b; Velazquez et al., 1997).

Lately, Diamanti-Kandarakis et al. (2001) have demonstrated elevated endothelin-1 (ET-1) levels in women with PCOS, independently of the presence of obesity. It is suggested that high ET-1, a potent vasoconstrictor peptide, may represent an early sign of abnormal vascular reactivity, which precedes that detectable by dynamic tests of endothelial dysfunction. Furthermore, a positive correlation between plasma ET-1 levels and testosterone levels was reported, which is in accordance with the findings by Paradisi et al. (2001) referred to above. Interestingly, 6 months of metformin therapy reduced ET-1 concentrations in these women, suggesting that increase insulin sensitivity may offer benefit by protecting and/or restoring the endothelial barrier.

LH hypersecretion - both basally and in response to GnRH administration - is a characteristic hallmark of PCOS (Yen et al., 1970; Barnes et al., 1989). This phenomenon has been considered to be the primary abnormality in classic PCOS and thus, the cause of androgen excess.

It is believed that the elevated LH levels are partly due to an increased sensitivity of the pituitary to GnRH stimulation, manifested by an increase in LH pulse amplitude and frequency, but mainly amplitude (Venturoli et al., 1988; Hayes et al., 1998). The gonadotrophin pattern (high LH and low to normal FSH) can also be due to increased pulse frequency of GnRH secretion (Waldstreicher et al., 1988), attributed to a reduction in hypothalamic opioid inhibition because of the chronic absence of progesterone (Berga & Yen, 1989; Cheung et al., 1997). It is likely that this increased activity is taking place at both hypothalamic and pituitary sites. This is consistent with an abnormal diurnal pattern of LH secretion that has been reported both in adolescent girls with PCOS (Apter et al., 1994) and in adults with PCOS (Yen et al., 1993), with the highest LH values occurring in late afternoon rather than at night. Finally, there is evidence that bioactive LH is elevated in many patients with PCOS, in whom immunoactive LH is normal (Fauser et al., 1992; Imse et al., 1992).

Marshall et al. (1991; Marshall & Eagleson, 1999) have focused on pulsatile patterns of LH secretion as an indicator of altered hypothalamic secretion of GnRH. It is believed that although a pulsatile GnRH stimulus is required to maintain gonadotropin synthesis and secretion, it is the frequency and amplitude of GnRH pulses that determine gonadotrophin subunit gene expression and secretion of pituitary LH and FSH. Thus, in ovulatory cycles, an increase in GnRH frequency during follicular phase favours LH synthesis prior to the LH surge, while following ovulation, luteal steroids slow GnRH pulses to favour FSH synthesis. In PCOS, LH/GnRH pulses are persistently rapid and favour LH synthesis, hyperandrogenaemia and impaired follicular maturation. Christman et al. (1991) have shown that the administration of progesterone in anovulatory women with PCOS can slow GnRH pulse secretion, favour FSH secretion and induce follicular maturation. Furthermore, in a more recent study, the same group of authors showed evidence suggesting an insensitivity of the GnRH pulse generator to suppression by oestradiol and progesterone in PCOS women (Pastor et al., 1998). Taken together, these data suggest that the increased plasma LH and GnRH/LH pulsatile secretion in PCOS are not simply a consequence of the low levels of progesterone due to anovulation, but reflect an underlying insensitivity of the hypothalamic GnRH pulse generator to oestrogen/progesterone inhibition. Finally, it was suggested that such insensitivity during pubertal maturation could be a potential mechanism for the perimenarchal abnormalities seen in hyperandrogenaemic adolescents who appear to exhibit early manifestations of PCOS (Marshall & Eagleson, 1999). The mechanisms underlying the reduced hypothalamic sensitivity, however, remain unclear. The potential roles of hyperinsulinaemia and hyperandrogenaemia often present in women with PCOS, in modifying ovarian steroid regulation of GnRH pulse generator have to be clarified, although an intrinsic abnormality could not also be excluded.

Ovarian-Pituitary Feedback of LH Secretion. Classic studies in primates have demonstrated that pulsatile secretion of GnRH is an important prerequisite for normal pituitary function (Mais et al., 1986) and that the regulation of gonadotrophin levels is controlled by ovarian steroid feedback on the anterior pituitary cells. Thus, low levels of oestrogen inhibit LH and FSH at the pituitary level; FSH more than LH (Gharib et al., 1990). High levels of oestrogen exhibit a positive stimulatory feedback with LH, inducing the LH surge at midcycle, whereas high steady levels of oestrogen lead to sustained elevated LH secretion (Keye & Jaffe, 1975). In addition, low levels of progesterone acting at the level of the pituitary gland enhance the LH response to GnRH and are responsible for the FSH surge at midcycle. As a consequence, the question of whether the gonadotropin abnormalities seen in PCOS might be secondary to an oestrogenic effect on the pituitary or even more to an effect of androgens, independent of their aromatization to oestrogens, at the level of the hypothalamus and/or the pituitary, has been considered. Although the data are inconsistent, they mainly suggest that if a primary abnormality of serum sex steroids concentrations has a stimulatory effect on LH secretion in PCOS, this effect must be minimal (Ehrmann et al., 1995b). In addition, raising serum androgen concentrations in normal women (Spinder et al., 1989) or women with the PCOS (Dunaif, 1986) does not stimulate the secretion of luteinizing hormone.

An additional consideration reported by Barnes et al. (1989), is that the pattern of gonadotrophin responses to GnRH agonist administration in PCOS women appears to be 'masculinized'. Recent data support the hypothesis that perinatal exposure of the neuroendocrine axis to excess levels of androgen may bring about such masculinization by programming the neuroendocrine system to secrete excessive LH at puberty, thus resulting in ovarian hyperandrogenism (Barnes et al., 1994). This hypothesis has been further supported in primate studies by Dumesic et al. (1997) who have shown that prenatal exposure of female rhesus monkeys to testosterone propionate increases serum LH levels in adulthood.

Although many hypotheses have been proposed for the aetiology of pituitary hypersecretion of LH, none of these fully explains the underlying neuroendocrine abnormality that leads to exaggerated LH pulse frequency. On the other hand, it is also clear that an elevated LH concentration is not necessary for ovarian dysregulation (Ehrmann et al., 1992). Indeed, hypersecretion of LH occurs only in approximately one-third of women with PCOS (Balen et al., 1995), particularly in nonobese patients. Furthermore, Shoham et al. (1992) have characterized a subgroup of hypogonadotropic patients with ultrasound finding of polycystic ovaries, in which the ovarian response to ovulation induction with pulsatile GHRH, was similar to patients with PCOS. In a subsequent study, Schachter et al. (1996) not only found much higher serum LH concentrations following ovulation induction with GnRH in this group of patients compared to hypogonadotropic patients with ultrasonographically normal ovaries, but also found that this difference preceded any observed changes in oestradiol levels. This suggests that the primary lesion in PCOS is in the ovary, with pituitary hypersecretion of LH secondary to disturbed ovarian feedback signalling.

A number of authors have proposed an alternative model of PCOS as a form of gonadotropin-dependent ovarian hyperandrogenism in which the central abnormality is an elevated intraovarian androgen concentration (Fig. 2) (Barnes et al., 1989; Rosenfield et al., 1990). They presented data suggesting that women with PCOS have increased formation of 17α-hydroxyprogesterone and androstenedione in response to LH (Barnes et al., 1989) because of abnormal enzymatic regulation ('dysregulation') of steroidogenesis (Rosenfield et al., 1990, 1994). In a subsequent in vitro study using a primary monolayer culture system of human thecal cells, Gilling-Smith et al. (1994) have demonstrated a significant increase in both basal- and LH-stimulated androstenedione production per cell in theca from polycystic ovaries compared to the response in normal ovaries, although the magnitude of response to LH was similar. Importantly, the addition of LH in vitro did not significantly alter the androstenedione/progesterone ratio, suggesting that these observations cannot be explained solely by the exposure of thecal cells to high concentrations of LH in vivo.

Ovarian steroid biosynthesis. Although each cell type of the ovary possess the complete enzymatic complement required for steroid hormone synthesis, the predominant hormones formed differ among cell types. In the ovarian follicle, the Δ5-pathway is preferred for the formation of androgens and oestrogens, because theca cells of human ovary metabolize 17-OH Pregnenolone more efficiently than 17-OHP. The main pathway of steroid synthesis in human corpus luteum is the Δ4-pathway, which involves conversion of pregnenolone to progesterone. The names of each enzyme are shown by each reaction. (1) P450scc, mitochondrial cholesterol side-chain cleavage enzyme, mediates 20α hydroxylation, 22 hydroxylation and scission of the C20-22 bond. (2 and 3) P450c17 in the endoplasmic reticulum mediates both 17α-hydroxylation and scission of the C17,20 bond. (4) 3βHSD, a non-P450 enzyme bound to the endoplasmic reticulum, mediates both 3β-hydroxysteroid dehydrogenase and Δ5-Δ4 isomerase activities. The type II isoform is expressed both in gonads and adrenal glands. (5) 17βHSD, a non-P450 enzyme of the endoplasmic reticulum, converts estrone to oestradiol, androstenedione to testosterone, and DHEA to androstenediol, and vice versa. Nine human 17βHSD isoforms have been cloned and characterized to date. (6) P450arom in the endoplasmic reticulum mediates aromatization of the A ring of the steroid nucleus. The factors that determine which steroid is secreted by each cell type include the levels of gonodotrophin and gonodotrophin receptors, the expression of steroidogenic enzymes, and the availability of LDL cholesterol.

More recent studies from the group of Strauss and MacAllister have also shown that increased androgen production is a stable steroidogenic phenotype of PCOS theca cells propagated in long-term culture, strongly supporting the hypothesis that the hyperandrogenaemia associated with PCOS results from an intrinsic abnormality of ovarian theca cell steroidogenesis (Nelson et al., 1999, 2001). The theory that PCOS results from a primary abnormality of androgen biosynthesis further supported by Ehrmann et al. (1995b) who concluded that the dysregulation of the steroid biosynthesis and metabolism, prominently but not exclusively, involves P450c17 enzyme activities. This enzyme performs both 17-hydroxylation and 17,20-lyase functions in both the ovarian and adrenal steroidogenic tissue. Thus, this dysregulation may be apparent as ovarian dysfunction alone, as adrenal dysfunction alone, or as both together. Ovarian stimulation testing has suggested that ovarian hyperandrogenism is a result of dysregulation of the androgen-producing steroidogenic enzymes. On the other hand, ACTH stimulation testing is consistent with dysregulation of adrenal steroidogenic enzymes in about two-thirds of hyperandrogenic women (Ehrmann et al., 1992). Although adrenal androgen concentrations may be elevated in women with PCOS, the weight of evidence suggests that in most cases, the ovary is the main contributor to excess androgen secretion. Furthermore, genetic screening of anovulatory women with polycystic ovaries has not detected sequence variations in the promoter or coding regions of the P450c17 gene (Techatraisak et al., 1997).

In an attempt to identify the biochemical basis for the increased testosterone production in PCOS theca cells, Nelson et al. (2001) examined 17β-hydroxysteroid dehydrogenase (17βHSD) isoform expression in long-term cultures of theca and granulosa cells isolated from normal and PCOS ovaries. 17βHSD-specific isoforms catalyse the final step in the conversion of androstenedione to testosterone. They reported P450c17 and 3βHSD enzyme activities were increased by more than 500% and 1000%, respectively, in PCOS theca cells compared with controls, whereas 17βHSD enzyme activity was unaffected. They concluded that the increased synthesis of testosterone precursors is probable the primary factor driving enhanced testosterone secretion in PCOS. Interestingly, although a number of authors have proposed that C17,20 lyase activity is disproportionately increased in PCOS, this new data suggests that both 17α-hydroxylase and C17,20-lyase activities are coordinately increased in PCOS theca cells.

Role of Gonadotrophins and Growth Factors. Other evidence for a primary defect at the level of the ovary comes from the classic polycystic ovary morphology. The presence of many follicles with a high androgen to oestrogen ratio was first thought to represent a high rate of follicular atresia. Subsequently, the granulosa cells were shown to be viable and able to respond to FSH production (Mason et al., 1994; Almahbobi et al., 1996). The functional picture that emerges of arrested granulosa cells and very active theca cells is consistent with a blockage of FSH response, probably through various growth factors. As a consequence the follicles are unable to successfully change their microenvironment from androgen dominance to oestrogen dominance, the change that is essential for continued follicular growth and development.

A number of peptides modulate gonadotrophin-dependent ovarian folliculogenesis and steroidogenesis acting by autocrine, paracrine and endocrine mechanisms. Thus, LH stimulation of androgen synthesis appears to be augmented by factors such as inhibin and downregulated by factors such as activin, epidermal growth factor (EGF), transforming growth factor-α (TGF-α) and transforming growth factor-β1 (TGF-β1). It has long been known that inhibin augments LH-stimulated androstenedione production in cultured human theca cells (Hsueh et al., 1987), whereas activin has the opposite effect (Hillier et al., 1991). As inhibin also has the ability to selectively inhibit FSH secretion, it was suggested that increased production of inhibin B from the multiple small viable follicles could also be a factor in suppressing FSH secretion (Lockwood et al., 1998). Gonadotrophin-induced ovarian function is also known to be modulated by growth factors. IGFs, their receptors, binding proteins and binding protein proteases are important in normal ovarian follicle development. Overall, it is believed that IGFs stimulate ovarian cellular mitosis, inhibit apoptosis and increase steroidogenesis - mainly by enhancing P450scc enzyme activity and synergizing with LH to upregulate P450c17 production. It is interesting that the pattern of expression of the components of the IGF system in the follicular fluid from polycystic ovaries was found to be the same as in atretic follicles, finding consistent with limitation of IGFs activity in PCOS (Cataldo & Giudice, 1992; San Roman & Magoffin, 1992). GH itself may also be involved to the genesis and maintenance of hyperandrogenic chronic anovulation and polycystic ovarian morphology, at least in the subgroup of lean PCOS women in whom it is hypersecreted (Morales et al., 1996). GH actions on growth factors and their binding proteins are known to be similar to those of insulin (Conway et al., 1990).

Leptin. Leptin, the product from the obesity gene (ob gene), correlates positively with BMI but also has variation during the menstrual cycle; leptin levels peak in the luteal phase of the cycle, correlating with maximum progesterone (Hardie et al., 1997). These changes suggest a direct physiological role for leptin in regulating ovarian function. Disruption of such an effect could play a role in menstrual irregularities generally observed in both obese and undernourished women and may offer a pathophysiological mechanism in women with PCOS (Jacobs & Conway, 1999). Most studies, however, report that leptin levels in women with PCOS do not differ significantly from normal controls, regardless of their bodyweight (Rouru et al., 1997). In addition there is no evidence for mutations of leptin or leptin receptor genes in women with polycystic ovary syndrome (Oksanen et al., 2000).

A Primate Model for the Aetiology of PCOS. A prenatal aetiology for the multiple features manifest in women with PCOS has been proposed by studies of adult female rhesus monkeys exposed prenatally to androgens, a nonhuman primate model for the aetiology and mechanism of PCOS. In rhesus monkeys, multiple fetal organ systems are reprogrammed by experimentally induced androgen excess, leading to an adult phenotype similar to that found in women with PCOS: LH hypersecretion (Dumesic et al., 1997) and impaired insulin secretion and action (Eisner et al., 2000) are accompanied by hyperandrogenism and anovulation (Abbott et al., 1998). Hyperandrogenism seems to be the core functional disorder in both women with PCOS and adult female rhesus monkeys, androgenized prenatally (Abbott et al., 2002). Obesity and hyperinsulinaemia exacerbate this core hyperandrogenic disorder into anovulation. Recently, Eisner et al. (2002) have shown that prenatal androgen excess in female rhesus monkeys causes perturbations in ovarian and adrenal steroidogenesis during adulthood, which may both contribute to hyperandrogenism. Although it is unlikely that a fetal origin of PCOS in humans is based on exogenous (environmental or maternal) hyperandrogenism, androgen excess originating from the fetal ovary or adrenal cortex, both of which are steroidogenically active during the second trimester of prenatal life could well explain many of the manifestations that present as PCOS.

Although observations in animals strongly support the hypothesis that abnormalities of PCOS are initiated in utero, this has been examined only recently in human beings. Cresswell et al. (1997), in a sample of 235 middle-aged women whose size at birth was recorded in detail, related the prevalence of polycystic ovaries and the plasma concentrations of gonadotropin hormones and androgens to their body size at birth, and the length of gestation. They concluded that the two common forms of PCOS have different origins in intrauterine life: obese, hirsute women with polycystic ovaries have higher than normal ovarian secretion of androgens that are associated with high birth weight and maternal obesity, while the thin women with polycystic ovaries have altered hypothalamic control of LH release resulting from prolonged gestation. A strong association between polycystic ovaries and large birth weight has also been described recently by Michelmore et al. (2001) in a population-based study. In this study there were no differences in birth weight between women with polycystic ovaries alone and those with features of PCOS. However, the nature of the association between high birth weight and polycystic ovaries/PCOS is uncertain and contrasts with recent reports of low birth weight in girls with precocious pubarche and hyperinsulinaemic hyperandrogenism (Ibanez et al., 1998, 1999).

An increased adrenal androgen production found in 25% of PCOS women (Ehrmann et al., 1992; Turner et al., 1992), possibly as a result of a genetic trait or secondary to ovarian hormonal secretion (Moran & Azziz, 2001). A more recently proposed mechanism is an alteration in cortisol metabolism. The principal pathways of cortisol metabolism include irreversible inactivation by 5α-reductase (5α-R) and 5β-reductase (5β-R) in liver, and reversible interconversion with cortisone by 11βHSD in liver and adipose tissue. According to this theory, increased peripheral cortisol metabolism either by increased 5α-R activity and thus increased inactivation of cortisol (Stewart et al., 1990; Chin et al., 2000) or impaired 11βHSD activity and thus impaired regeneration of cortisol (Rodin et al., 1994) results in compensatory increase of ACTH secretion via a decrease in the negative feedback signal, maintaining normal serum cortisol levels at the expense of adrenal androgen excess (Fig. 3). In support of this hypothesis, urinary metabolites of cortisol were found to be abnormal in women with PCOS.

Increased peripheral cortisol metabolism as a proposed mechanism for the development of PCOS.

Stewart et al. (1990) first documented increased total cortisol metabolite excretion in the urine of PCOS women using gas chromatography and mass spectrometry, compared to controls. In particular, they found that the ratio of 5α to 5β cortisol metabolites were higher in PCOS subjects than in controls, indicating enhanced 5α-R activity. This steroidogenic enzyme is responsible for both 5α-reduction of testosterone to 5α-dihydrotestosterone in skin and cortisol to 5α-dihydrocortisol in liver. Therefore, it was suggested that increased activity of 5α-R mediates both hirsutism and enhanced hepatic cortisol metabolism. Previous in vitro studies in genital skin fibroblasts have shown that 5α-R activity is upregulated by androgens (Mowszowicz et al., 1983), an effect that might be mediated by IGF-I (Horton et al., 1993). Finally, a downregulation of this enzyme by high prolactin levels has been supported as a protective mechanism against clinical presentation of unwanted hair growth in hirsute women (Seppala & Hirvonen, 1975). Luppa et al. (1995) have documented changes in the excretion of urinary androgen and gestagen metabolites in response to GnRH agonist stimulation, suggesting a functional alteration of the pituitary-ovarian axis.

Rodin et al. (1994) found evidence for dysregulation of 11βHSD enzyme activity in PCOS women. In particular, they found that the ratio of 11-hydroxy to 11-oxo cortisol metabolites were lower in PCOS women than in controls, indicating impaired activity of 11βHSD1. Two isoenzymes of 11βHSD catalyse the interconversion of active cortisol to inactive cortisone and vice versa. 11βHSD1 is an oxoreductase expressed mainly in human liver and adipose tissue, responsible for cortisone to cortisol conversion. Impaired activity of this enzyme is compatible with increased metabolic clearance of cortisol, resetting of ACTH secretion, and consequent generalized adrenocortical over activity. Because liver and adipose tissue are also targets of insulin action, they speculated that the dysregulation of this enzyme might be the result of hyperinsulinaemia. Alternatively, they suggested that 11βHSD dysregulation could be due to hyperandrogenaemia itself.

The mechanism of altered 5α-R and/or 11βHSD1 activity in women with PCOS is still uncertain. Although, more than half of women with PCOS may be overweight, and obesity may cause abnormalities of cortisol metabolism, this mechanism cannot fully account for abnormalities of 5α-R and 11βHSD1 activities in PCOS. Stewart et al. (1990) found increased 5α-R activity in PCOS subjects, compared with controls of similar weights. Similarly, the altered 11βHSD1 activity in PCOS reported by Rodin et al. (1994) was also confirmed in lean PCOS subjects. Recently, Walker et al. (2000) have also excluded the increased production of endogenous inhibitors of 11βHSD1, measured in urine, as a mechanism of abnormal cortisol metabolism in PCOS. Another proposed mechanism is that high oestrogen levels in PCOS, especially in the form of oestrone, could downregulate 11βHSD1 activity in liver. However, recent evidence suggests that oestrogen does not have a potent effect on 11βHSD1 activity in humans (Andrew et al., 1998; Finken et al., 1999). Finally, the association between the activities of these enzymes with insulin resistance and hyperinsulinaemia in women with PCOS needs to be clarified and might explain the altered cortisol metabolism in these women.

Genetics of PCOS. Recent family studies suggest that hyperandrogenaemia per se is the major reproductive endocrine phenotype in premenopausal first-degree relatives of PCOS probands (Legro et al., 1998; Kahsar-Miller et al., 2001). In a prospective study of first-degree female relatives of women with PCOS, of the 46% of sisters thus affected, approximately one-half have PCOS and one-half have hyperandrogenaemia with regular menstrual cycles. The distribution of testosterone levels is bimodal, suggesting a monogenic trait controlled by two alleles at an autosomal locus (Urbanek et al., 2000a). Insulin resistance also demonstrates familial aggregation consistent with a genetic trait. A twin study of PCOS women showed high correlation within twin pairs for both circulating androgen and fasting insulin levels (Jahanfar et al., 1997). Interestingly, Legro et al. (2002) have also found an association between insulin resistance and hyperandrogenaemia rather than menstrual irregularity in the sisters of women with PCOS. It was suggested that phenotyping sisters of woman with PCOS on the basis of circulating testosterone levels may also help to identify those at risk for insulin resistance and, accordingly, for the long-term health consequences of PCOS. Thus, insulin resistance and hyperandrogenaemia may reflect variable expression of a single gene or the interaction of two or more genes. Further to that, Waterworth et al. (1997) have found an association between PCOS and allelic variation at the INS VNTR (variable number of tandem repeats) locus of the insulin gene, implying that PCOS is due, in part, to an inherited alteration in insulin production.

These observations encouraged a study of the gene CYP17, which codes for the enzyme P450c17a, as a candidate genetic locus in the aetiology of PCOS. A linkage analysis study excluded the possibility that gene CYP17 is the primary genetic locus responsible for PCOS (Carey et al., 1994). In the same study, however, a common polymorphism (substitution of C for T at -34 bp from the initiation site of translocation) was discovered and was proposed to affect the phenotype, conferring increased susceptibility to PCOS. However, extending the initial study to a larger case-control study, the frequency of CYP17 promoter polymorphism in women with PCOS did not differ significantly from the frequency observed in the reference population (Gharani et al., 1996).

Using a candidate gene approach, Franks et al. (1997), reported evidence for the involvement of two key genes in the aetiology of PCOS. From the results of both linkage and association studies, they suggested that the steroid synthesis gene CYP11a which codes for P450 side chain cleavage, and the insulin variable number tandem repeats regulatory polymorphism are important factors in the genetic basis of PCOS and may explain the heterogeneity of the syndrome. PCOS appears to represent a quantitative trait in which a relatively small number of key genes contribute, in conjunction with environmental (particularly nutritional) factors, to the observed clinical and biochemical heterogeneity (Franks et al., 2001).

Because of its conceivable role of FSH and its receptor in the pathophysiology of PCOS, PCOS patients were screened for the Finnish FSH receptor gene mutation C556T (Conway et al., 1999; Al-Hendy et al., 2000). All the tested samples showed homozygous normal allele. Urbanek et al. (1999) tested 37 candidate genes for linkage and association with PCOS or hyperandrogenaemia in data from 150 families. The strongest evidence for linkage was with the follistatin gene, for which affected sisters showed increased identity with descent (72%). The linkage results for CYP11A were also nominally significant, but were no longer significant after correction. Further studies showed that the follistatin gene is unlikely to contribute to the aetiology of PCOS (Urbanek et al., 2000b). An IGF-I or insulin receptor genes are currently under investigation.