Funding
This work was supported by academic grants to Paolo Moghetti from the University of Verona.
Conflicts of interest/competing interests
The author has nothing to disclose.
Introduction
Polycystic ovary syndrome is a highly prevalent and heterogeneous condition in women of reproductive age, primarily characterized by hyperandrogenism and reproductive abnormalities. However, there are other frequent abnormalities in these women. In particular, many of these subjects show insulin resistance and other metabolic alterations, which may play relevant roles in terms of pathogenesis and general health.
According to the conclusions of a joint consensus workshop of the European Society of Human Reproduction and Embryology (ESHRE) and the American Society for Reproductive Medicine (ASRM), held in Rotterdam in 2003, diagnosis is currently based on the presence of at least two of three clinical features, i.e., clinical and/or biochemical hyperandrogenism, chronic oligoanovulation and micropolycystic ovarian morphology (PCOm), after exclusion of secondary causes . Although different opinions still exist, and some authors have suggested that the presence of hyperandrogenism should always be required for diagnosing PCOS , these criteria have been widely adopted. However, an important issue is that they contributed to increasing the clinical heterogeneity of the syndrome. In fact, according to the different possible combinations of these diagnostic features in each subject with PCOS, adoption of the Rotterdam criteria has introduced different clinical phenotypes of the syndrome. These phenotypes were subsequently named Complete or A (characterized by hyperandrogenism, oligoanovulation and PCOm), Classic or B (hyperandrogenism and oligoanovulation, without PCOm), Ovulatory or C (hyperandrogenism and PCOm), and Normoandrogenic or D (oligoanovulation and PCOm). It remains unclear whether these phenotypes really reflect the same overall condition and the same long-term risks. For these reasons, a subsequent NIH workshop, while confirming the use of the Rotterdam criteria for diagnosing PCOS, has recommended that diagnosis should be accompanied by the indication of the specific clinical phenotype . Interestingly, there is evidence that this distinction may be important also in terms of individual metabolic risk .
In examining the metabolic alterations of PCOS and the association between these features and the clinical phenotypes of PCOS, it should be kept in mind that the key diagnostic aspects of the syndrome are often assessed by inaccurate methods. This is especially true as regards hyperandrogenism. Clinical hyperandrogenism is generally measured by subjective scores, which are operator dependent and are also affected by constitutional and ethnic factors . Even more important, assessment of biochemical hyperandrogenism is often biased by use of inaccurate and imprecise methods, as gold standard assays are generally not available in clinical laboratories. When using these methods there is substantial overestimation of hormone concentrations, as compared with figures obtained by reference methodology, i.e., liquid chromatography-tandem mass spectrometry and equilibrium dialysis . Moreover, the imprecision of most available methods is a critical issue from both a clinical and pathophysiological perspective, as many women may be misclassified in terms of biochemical hyperandrogenism. In our experience, up to 30% of subjects with PCOS may be wrongly classified by these methods as hyperandrogenemic or normoandrogenemic, with assignment of many PCOS women to an inexact clinical phenotype .
Major problems also exist in the measurement of insulin resistance, as the gold standard glucose clamp methodology is not applicable in clinical practice. For this reason, surrogate measures are generally used. However, as discussed in detail below, they lack sensitivity, showing a low negative predictive power .
Difficulties and pitfalls in the measurement of in vivo insulin action
Unfortunately, direct in vivo assessment of insulin sensitivity is a difficult and complex task. The gold standard procedure is the hyperinsulinemic euglycemic clamp, a technique proposed by DeFronzo et al. . Briefly, this methodology is based on a prolonged iv insulin infusion at a primed-continuous rate, combined with a concurrent infusion of variable amounts of glucose, as needed by each individual to maintain plasma glucose at the basal level. This requires a real-time, frequent assessment of glucose levels in arterialized blood. In this condition, provided that insulin infusion rate is sufficient to inhibit endogenous glucose production, the rate of glucose infusion, measured when the steady state in insulin action has been reached, equals insulin induced glucose utilization and is therefore a direct measure of whole body insulin action. However, this procedure is time consuming and requires personnel with specific expertise, which limits its adoption in research and makes it difficult to use in clinical practice. The frequently sampled intra-venous glucose tolerance test, with computer modeled analysis of data, is also considered a reliable method. Nevertheless, this technique is also complex, time-consuming and expensive, with the need for multiple assays, which limits its use. As an alternative, a number of surrogate indexes of insulin sensitivity are often used, mainly based on glucose and insulin levels measured at fasting or after oral glucose, such as the HOMA index or the Matsuda index. However, while these indexes may be useful in epidemiological studies and in analyzing the relationships between insulin sensitivity and other features, they are inadequate for distinguishing between insulin-resistant and insulin-sensitive individuals . In fact, they show a good positive predictive value but a low negative predictive value and have therefore low sensitivity in recognizing insulin-resistant subjects. In other terms, they can be used to confirm but not to exclude insulin resistance in individual subjects.
The leading reason limiting the reliability of surrogate indexes of insulin action is the fact that insulin levels, which mainly guide these indexes, are affected by several factors, not only by insulin sensitivity. Besides blood glucose, which is a key element in the regulation of insulin secretion and is taken into account by all these algorithms, beta-cell function is another obvious determinant of insulin levels. Moreover, additional factors that play a major role in determining plasma insulin concentrations are insulin clearance and body fat excess.
A study carried out in the multicenter EGIR cohort, a large cohort of nondiabetic subjects of both sexes with different BMI, who had previously been submitted to an hyperinsulinemic euglycemic clamp, investigated the extent of overlap between hyperinsulinemia and impaired glucose utilization during the clamp, and the predictors of insulin levels in multivariable analysis . Overlap of direct and indirect evidence of insulin resistance was only 60%, with 43% of subjects being hyperinsulinemic with normal M-clamp values, and 38% insulin resistant with normal insulin levels. In multivariable analysis, including glucose utilization during the clamp, fasting blood glucose, BMI and insulin clearance among the independent variables, the variance of fasting insulin explained by the model was about 30%. Interestingly, insulin sensitivity contributed to a relatively small percentage of insulin variance in these subjects, whereas the independent contribution of BMI and insulin clearance was greater.
The role of body fat excess in this phenomenon is complex and of particular interest. Obesity has a clear, strong relationship with insulin resistance, which is also evident in women with PCOS. Total body fat and, to a greater extent, truncal fat are inversely associated with in vivo insulin action, as measured by the hyperinsulinemic euglycemic clamp . Nevertheless, obesity affects insulin levels independently of its effects on insulin sensitivity. Comparison of the relationships between the M-clamp value and fasting plasma insulin in obese vs normal-weight subjects included in the EGIR cohort showed that obese individuals had significantly higher insulin levels for any degree of insulin sensitivity . Moreover, consistent findings were recently provided by another study in which plasma insulin kinetics, before and after glucose ingestion, were compared in obese vs lean subjects matched on multiorgan insulin sensitivity, as assessed in terms of stimulation of glucose uptake and inhibition of lipolysis and glucose production by a two-stage hyperinsulinemic euglycemic clamp combined with glucose and palmitate tracer infusion and FDG-PET . In this study, plasma insulin levels at fasting and after oral glucose load were higher in obese than in lean individuals. Furthermore, when comparing insulin response to glucose ingestion in obese subjects at baseline and after a low-calorie diet, in a subgroup of subjects who did not show changes in insulin sensitivity despite significant weight loss, insulin levels were significantly lower after BMI reduction .
Another important factor that can affect plasma insulin concentrations is insulin clearance. Plasma insulin levels depend on the balance between hormone secretion rate from the beta-cells and metabolic clearance rate from the plasma compartment. In particular, insulin degradation occurs following internalization of the insulin-receptor complex in target cells, under the action of the insulin-degrading enzyme and other lysosomal enzymes . Interestingly, some data indicate that this phenomenon may be of specific importance in women with PCOS. In fact, marked reduction of insulin clearance, assessed by the fasting plasma C-peptide over insulin ratio, was reported in these patients, as compared with both lean and obese controls . This phenomenon was associated with impaired insulin degradation in cultured lymphocytes from PCOS women, suggesting an intrinsic defect. Moreover, in the same study, testosterone revealed a biphasic effect on insulin binding and insulin degradation in cultured cells from healthy subjects. Indeed, hormone binding and degradation were impaired when cells were exposed to moderately high concentrations of testosterone, corresponding to those that can be observed in women with PCOS, whereas they were unaffected at higher testosterone concentrations, corresponding to those typical of men. These findings call to mind the sexual dimorphism observed in the metabolic effects of androgens. Use of plasma C-peptide over insulin ratio is a rough index of insulin clearance . However, consistent results have been recently reported in another study, in which metabolic clearance of insulin was calculated as the insulin infusion rate divided by the serum insulin concentrations during the hyperinsulinemic euglycemic clamp, which is a direct measure of this phenomenon. In this study insulin clearance was markedly reduced in women with PCOS, as compared with findings in healthy controls ( Fig. 1 ). Interestingly, in multivariable analysis, insulin clearance was independently predicted, in these patients, by body fat, the estimated insulin secretion rate and serum androgens, all with inverse relationships. It is noteworthy that insulin clearance was not associated with insulin sensitivity in this analysis. In contrast, insulin secretion was independently predicted by age, insulin clearance and insulin sensitivity, with inverse relationships, and by body fat, with a direct relationship. These findings suggest that, in women with PCOS, obesity can contribute to hyperinsulinemia by increasing insulin secretion and reducing insulin clearance, independent of its effects on insulin resistance, whereas age and insulin sensitivity regulate insulin secretion, and serum androgens specifically regulate insulin degradation. There appears also to be a mutual modulation between insulin secretion and insulin clearance. Hence, in women with PCOS, androgens could contribute to determining increased insulin levels by reducing insulin clearance, besides their potential effects on insulin action. It is noteworthy that hyperinsulinemia entirely due to reduced insulin clearance has recently been reported in subjects with classical congenital adrenal hyperplasia, although in this study the interpretation of findings is complicated by ongoing therapy .
The unhealthy connection between insulin and PCOS
A number of data support a linkage between insulin resistance and PCOS, although the mechanisms underlying this relationship and the respective roles of the impaired insulin action and the associated hyperinsulinemia still remain unclear . The evidence of a potential relationship between insulin resistance and hyperandrogenism firstly emerged in the seventies, when the development of radioreceptor assays allowed researchers to study the interaction between insulin and its receptor on the cell surface in some rare syndromes characterized by severe insulin resistance, with frequent glucose tolerance alterations and acanthosis nigricans. Interestingly, affected women also had severe PCOS-like symptoms, such as amenorrhea and marked hyperandrogenism . It is noteworthy that these features appeared to be independent of the specific mechanisms underlying insulin resistance, as similar clinical characteristics were found in subjects with congenital insulin receptor deficiency and in those with acquired autoimmunity against the insulin receptor. Moreover, in the autoimmune syndrome, insulin binding and clinical features improved in parallel after in vivo removal of autoantibodies by plasma exchange, confirming the key role of insulin signaling in these alterations .
These findings stimulated researchers to assess insulin levels and insulin action in subjects with PCOS, and it was found that many of these women had hyperinsulinemia and insulin resistance, independent of obesity . Early research by Dunaif and her coworkers suggested potential PCOS-specific mechanisms of insulin resistance. In particular, increased serine phosphorylation of the insulin receptor and its substrate IRS-1 was reported in cultured fibroblasts and skeletal muscle, associated with altered metabolic insulin signaling . Nevertheless, these molecular alterations were not universal in examined women, suggesting that different mechanisms could indeed account for the impaired insulin action in women with PCOS. Further studies by the same group reported that mitogenic insulin signaling was stimulated in cultured cells from these subjects . These results suggested that pathways which remain unaltered in PCOS could be responsible for increased insulin action, determined by the insulin excess accompanying insulin resistance, on the routes specifically involved in the origin of PCOS, such as androgen production in the ovary . However, these intriguing hypotheses were not fully confirmed, as subsequent studies that assessed the potential molecular mechanisms responsible for an altered insulin signaling in tissues from insulin-resistant women with PCOS gave inconsistent results. Indeed, some studies did not identify defects in the proximal part of the insulin signaling cascade . Likewise, the alternative hypothesis of alterations in the AMPK pathway, supported by data obtained in skeletal muscle of lean, insulin-resistant subjects with PCOS, was not confirmed by findings in obese, insulin-resistant individuals . Thus, the putative PCOS-specific defects in insulin signal transduction are still to be identified and there is possible heterogeneity in the mechanisms responsible for insulin resistance in these women.
Yet, studies carried out with the reference glucose clamp procedure have confirmed that many subjects with PCOS have impaired insulin action . This phenomenon is extremely frequent in women with body fat excess, but can also be found to a considerable degree, around 60%, of normal-weight individuals .
Obesity is a common finding in these subjects, although some studies have suggested that the association may be boosted by considerable referral bias and contributes to generating insulin resistance in women with PCOS. Interestingly, the effect of body fat excess on insulin sensitivity seems to be greater in these women than in non-PCOS subjects. In fact, it was observed that in women with PCOS glucose utilization during the clamp is reduced to a greater extent with the increase in BMI, as compared to the relationship found in controls . It was hypothesized that this may be due to a relative increase of intra-abdominal fat deposition, driven by androgen excess . However, while disease-specific alterations in body fat distribution were not consistently found in these women , changes in morphology and function of subcutaneous adipocytes have been reported, which could be important issues in this phenomenon . Another interesting finding is that although subcutaneous adipocytes are enlarged and appear to be resistant to the lipolytic action of catecholamines , visceral adipocytes show an increased lipolytic response to adrenergic stimulation in women with PCOS , due to divergent changes in beta(2)-adrenergic receptors, the regulatory and catalytic components of protein kinase A, and hormone-sensitive lipase.
Multivariable analysis of data in a large sample of women with PCOS has suggested that the association of hyperandrogenism with total and visceral fat could be indirect, mediated by the relationships that both these different aspects have with insulin resistance . Nevertheless, the topic is very complex, and this piece of the PCOS puzzle is really difficult to put in the right place. Available data indicate several bidirectional links between body fat excess and PCOS . In this regard, it has been proposed that a PCOS secondary to obesity may exist, which should be considered as a distinct disorder .
Bidirectional Mendelian randomization analysis, a method which assesses whether common genetic variations in a potential underlying trait are linked to an outcome of interest, has recently been used as a tool to investigate the presence of a potential causal association between obesity and PCOS, in two different large cohorts of subjects. Interestingly, in both these studies, single nucleotide polymorphisms associated with obesity predicted increased risk of PCOS, whereas the opposite was not true . While these findings strongly support a causal role of body fat excess in the origin of PCOS, it should be kept in mind that currently known genetic probes for PCOS still have intrinsic limitations. Therefore, caution is appropriate in excluding, based on these results, a primary role of PCOS in fat accumulation.
Indeed, there is in vitro evidence that androgens can directly affect insulin sensitivity in cultured subcutaneous adipocytes . Moreover, in vivo data obtained in subcutaneous adipose tissue microdialysate have recently shown increased expression of the androgen-activating enzyme aldo-ketoreductase type 1 C3 (AKR1C3) in women with PCOS , as compared with BMI-matched healthy controls, with evidence of increased peripheral synthesis of testosterone in this tissue. Interestingly, in this study, both in vivo and in vitro findings have consistently shown that androgen excess can activate lipogenesis and reduce lipolysis and fatty acid oxidation in subcutaneous adipocytes, changes that promote fat mass accumulation and may favor the development of insulin resistance. Furthermore, it was also reported that AKR1C3 expression and activity were increased by insulin in cultured adipocytes , showing a potential intra-adipose mechanism of reciprocal potentiation between androgen excess and insulin resistance occurring in PCOS.
Another potential mechanism linking hyperandrogenism, body fat accumulation and insulin resistance is inflammation. Increased serum inflammatory markers have been found in obesity and in several other insulin-resistant states, including PCOS . To make the interpretation of these relationships even more complex, adipose tissue plays a significant role in the association between inflammatory cytokines and insulin resistance . Nevertheless, in multivariable analysis both obesity and insulin resistance reveal independent relationships with C-reactive protein levels in women with PCOS . Moreover, although this inflammatory marker shows only indirect relationships with serum androgens, pentraxin-3, which is another acute phase protein that also plays a role in oocyte development, showed a negative association with serum testosterone, independent of body fat and insulin resistance, indicating a potential role of androgens in these phenomena .
In accordance with this hypothesis, some studies that have investigated the effects of short-term experimental hyperandrogenism in women have shown an increase, at fasting and after oral glucose, of the expression in mononuclear cells of activated NF-kB, a molecule playing a central role in metabolic inflammation . Consistently, plasma TNFalpha levels were higher after androgen administration in these volunteers. These authors have also recently reported that saturated fat ingestion triggers an exaggerated inflammatory response in women with PCOS, independent of body fat excess and correlated with androgen response to stimulation . As a whole, these findings support the hypothesis that hyperandrogenism can modulate inflammation, directly and/or by amplifying the effects of nutrients on this phenomenon, resulting in complex interactions with insulin resistance.
A primary role of androgen excess in the relationship between insulin and PCOS is supported by the results of some intervention studies, designed to investigate the effect on insulin action of either androgen administration, in nonhyperandrogenic women, or of androgen action attenuation, in hyperandrogenic women. Two studies have measured insulin sensitivity by multistep euglycemic clamp, before and after the administration of supraphysiological amounts of androgens, in both female-to-male transsexuals (treated for 4 months) and healthy volunteers (treated for 10–12 days) . These studies showed that whole-body insulin action is rapidly impaired after androgen administration. The effect on insulin resistance of the attenuation of hyperandrogenism is more controversial. However, in a sample of hyperandrogenic women, mostly with PCOS, we have found that glucose utilization during the euglycemic clamp studies improved after 3–4 months of treatment with different drugs capable of reducing androgen levels or androgen action, although insulin resistance of these subjects was not completely reversed by treatment, even when it was maintained for 1 year .
Evidence of a potential primary role of androgens in this relationship derives also from studies in animal models, with results that suggest the potential contribution of different mechanisms. Testosterone-treated ovariectomized female rats, with moderate serum androgen excess, develops marked insulin resistance and hyperinsulinemia. In these animals, skeletal muscle showed fiber shift toward a fast, less insulin sensitive phenotype, with a concurrent reduction in capillary density . Animal models of cell-specific androgen receptor knockout showed that in mice with DHT-induced PCOS the development of metabolic traits requires androgen action in the brain . Moreover, insulin resistance induced by DHT administration in female mice was abolished by selective androgen receptor knockout in beta-cells .
Animal models have also supported the hypothesis of a developmental origin of PCOS, in which androgens and AMH can both play a key role by altering the firing rate of GnRH neurons, inducing oversecretion of LH and ovarian androgens both in the mother and in the offspring, with the appearance of PCOS traits in the adult female progeny . In this regard, an interesting point is that in women with PCOS serum androgens and AMH levels are higher than in healthy controls also during pregnancy . These phenomena may be relevant in the neuroendocrine mechanisms potentially triggering the origin of PCOS and associated metabolic alterations.
As a whole these data support the presence of multiple bidirectional links between insulin resistance and PCOS, playing a key role in the pathophysiology of the syndrome.
Insulin resistance or hyperinsulinemia: Which is the player in the pathophysiology of PCOS?
Besides the still unclear specific molecular mechanisms involved, another important question that remains unsolved is whether it is the impaired insulin action or the adaptive insulin excess that plays the main role in the pathophysiology of PCOS. Indeed, separate in vivo analysis of the respective responsibility of these two factors is extremely difficult as they are strictly associated.
In vitro studies showed that insulin increases androgen secretion in ovarian stroma and theca cells . Interestingly this effect of insulin was strikingly higher in cultured theca cells from women with PCOS than in cells from healthy controls . These findings suggest that insulin is a physiologic regulator of ovarian steroidogenesis and that there are intrinsic abnormalities in ovarian cells of PCOS subjects that enhance this effect of insulin.
Findings obtained in an animal model are consistent with the hypothesis of a primary role of hyperinsulinemia in the origin of PCOS. Increased insulin levels can be found in female mice with diet-induced obesity and are associated with several PCOS traits, such as LH hypersecretion, infertility, and abnormally high serum testosterone. Interestingly, in this model, selective knockout of the insulin receptor in theca-interstitial cells of the ovary strikingly improved the reproductive and endocrine features, whereas it had no effect in lean normoinsulinemic control animals . These results indicate that direct effects of insulin on the ovary play a key role in determining these PCOS-like alterations.
Some short-term in vivo studies in humans also support the hypothesis that insulin may play a significant role in the regulation of steroidogenesis, at least in women with PCOS. A comparison of serum levels of several adrenal hormones and their precursors in hyperandrogenic women studied both at baseline insulin levels and during hyperinsulinemia, obtained by a 2-h hyperinsulinemic euglycemic clamp, showed no difference between these conditions. However, after i.v. ACTH injection the response of 17-hydroxypregnenolone and 17-hydroxyprogesterone, which are 17a-hydroxycorticosteroid intermediates in the glucocorticoid and androgen pathways of adrenal steroidogenesis, was significantly higher at high insulin than at baseline insulin levels . In a subsequent study this protocol was modified, prolonging the duration of the comparison to 8 h, including 4 h of ACTH infusion in the second half of studies, in order to better investigate the effect of insulin on basal and ACTH-stimulated adrenal steroidogenesis in these subjects. In this study, absolute adrenal hormone secretion was quantified by a gas-liquid chromatographic assay of C19 and C21 steroid metabolites in urine collected after the first 4 h of insulin or saline infusion, and then after further 4 h of concurrent ACTH infusion . As in the previous study, response to ACTH of serum 17a-hydroxycorticosteroid intermediates increased during hyperinsulinemia. Moreover, although no differences in serum cortisol and androgens were found between the protocols, either before or after ACTH infusion, urinary excretion of ACTH-stimulated C19 and C21 steroid metabolites was significantly greater during hyperinsulinemia than at basal insulin levels ( Fig. 2 ).
