Candidate genes in polycystic ovary syndrome

Transcription

1 Human Reproduction Update, Vol.7, No.4 pp. 405±410, 2001 Genetics and Infertility II Candidate genes in polycystic ovary syndrome Stephen Franks 1, Neda Gharani and Mark McCarthy Imperial College School of Medicine, St Mary's and Hammersmith Hospitals, London, UK 1 To whom correspondence should be addressed at: Department of Reproductive Science and Medicine, Institute of Reproductive and Developmental Biology, Imperial College School of Medicine, Hammersmith Hospital, London W12 0NN. s.franks@ic.ac.uk The candidate gene approach has already paid some dividends in trying to understand the complex genetics of polycystic ovary syndrome (PCOS). In terms of steroidogenic abnormalities, CYP11aÐencoding P450 side chain cleavageðappears to be a major susceptibility locus. In relation to the well-described metabolic disturbances in PCOS, the insulin gene variable number tandem repeat (VNTR) appears to be a promising candidate, at least in populations studied in the UK. Finally, genes implicated in ovarian follicular development may have a role in the aetiology of PCOS, as demonstrated by recent identi cation of the follistatin gene as a potential disease locus. It seems unlikely that PCOS can be explained on the basis of a single gene disorder although, in a given family, one gene may have a predominant effect. An oligogenic model seems the most appropriate basis on which to understand the genetic origins of this very common disorder. The candidate gene approach has been useful to date, but it may prove important in the near future to perform an anonymous genome-wide scan to identify hitherto unheralded susceptibility loci. Key words: anovulatory/candidate genes/endocrinology/hirsutism/polycystic ovary syndrome TABLE OF CONTENTS Introduction Familial polycystic ovaries The biochemical phenotype of PCOS The candidate gene approach in PCOS Genes involved in the biosynthesis and metabolism of androgens Genes involved in the secretion and action of insulin Genes involved in folliculogenesis References polycystic ovariesðas de ned by pelvic ultrasonographyðis, however, wide and includes patients with hirsutism who have regular menstrual cycles. The results of ultrasound studies of `normal' populations suggest that polycystic ovaries are present in about 20% of women of reproductive age (reviewed by Franks, 1995). The cause(s) of polycystic ovaries and PCOS are not known for certain, but there is clear evidence for a primary abnormality of ovarian androgen secretion in the majority of cases. Introduction Polycystic ovary syndrome (PCOS) is a very common, heterogeneous endocrine disorder which accounts for about threequarters of all cases of anovulatory infertility and about 90% of the causes of hirsutism (Franks, 1995). The classic de nition includes the association of anovulatory menses (or oestrogenreplete amenorrhoea) with clinical and/or biochemical evidence of excess androgen secretion (Zawadzki and Dunaif, 1992). Using this de nition, the estimated prevalence of PCOS is in excess of 5% of the female population of reproductive age (Knochenhauer et al., 1998). The range of clinical presentation of women with Familial polycystic ovaries It has been recognized for many years that there is often an aggregation of cases of PCOS within families (for reviews see Simpson, 1992; Legro, 1995; Franks et al., 1997). Clinical genetic studies are hampered by the lack of consensus regarding de nition of polycystic ovary and PCOS and, particularly, by the fact that the disorder is only clinically expressed in women during their reproductive years. In the genetic studies based at this centre, we have de ned PCOS in the proband as the presence of polycystic ovary on ultrasound together with at least one symptom (of hyperandrogenism or anovulation, or both) (Carey Ó European Society of Human Reproduction and Embryology 405

2 S.Franks, N.Gharani and M.McCarthy et al., 1993; Franks et al., 1997). By contrast, in the north American studies, the group of Dunaif and Strauss have identi ed probands on the basis of oligo-amenorrhoea with elevated serum androgen concentrations (Legro et al., 1998; Urbanek et al., 1999). There is some controversy about the nature of a possible male phenotype. It has been observed that premature balding is more prevalent among male relatives of women with PCOS than in non-affected families or in the general population (Franks et al., 1997). However, not all investigators have been able to support this nding, and the de nition of the male phenotype remains contentious (Legro et al., 1998). Nevertheless, whether or not `affected' males are included in the analysis, most early studies indicated that, among simple Mendelian alternatives, autosomal dominant segregation provided the most likely mode of inheritance. Indeed, the results of our initial studies of 10 families with multiple cases of polycystic ovary were consistent with this suggestion (Carey et al., 1993). In our subsequent analysis of 23 affected families, the mode of inheritance was not clear-cut. An autosomal dominant mode of inheritance was not excluded, but we concluded that the prevalence of cases of polycystic ovary in families might best be explained on an oligogenic basis (Franks et al., 1997). An important observation in studies from our group and others (Legro et al., 1998) was that the heterogeneity in presentation between affected female members within the same family (e.g. menstrual pattern in hyperandrogenaemic women might be normal or abnormal). We therefore suggested that the observed heterogeneity of presentation of cases could be attributed to the interaction of a small number of genes with each other, and with environmental (mainly nutritional) factors. The biochemical phenotype of PCOS Although PCOS is typically heterogeneous in its clinical and biochemical features, there are certain biochemical characteristics that are common to all subjects with polycystic ovaries or, at least, features which occur commonly in major symptomatic subgroups. These clinical and biochemical characteristics provide a basis for investigation of the genetic origins of PCOS by indicating key `candidate' endocrine/metabolic pathways which are controlled by known genes. This is the rationale for using the candidate gene approach in this disorder. The most common biochemical abnormality in women with polycystic ovaries is hypersecretion of androgens (Franks, 1995). The ovary appears to be the predominant source of excess androgen production, although many studies have pointed to evidence for an additional adrenal abnormality. Nevertheless, the ovary is clearly the more important contributor to hyperandrogenaemia because suppression of LH in women with PCOS leads to a fall in androgen concentrations to levels that are indistinguishable from those in menopausal or oophorectomized women (Steingold et al., 1987). Investigations have shown that in human ovarian cell cultures, thecal cells from women with polycystic ovaries, regardless of presenting symptoms, produce some 20-fold more androstenedione in primary culture than do cells from women with normal ovaries (Gilling-Smith et al., 1994). Increased steroidogenic activity is, however, not con ned to androgen production. All stages of the steroidogenic pathwayðincluding progesterone productionðappear to be ampli ed in polycystic ovary theca. Importantly, these results have recently been con rmed using cells from polycystic ovary and normal theca that have undergone several passages in culture (Nelson et al., 1999). This suggests that this biochemical phenotype is an intrinsic feature of the polycystic ovary. Thus, it is unlikely that ovarian hyperandrogenism arises secondary to hypersecretion of LH in PCOS, particularly as this `typical' feature of PCOS occurs in little more than 50% of those with the classic syndrome and in the minority of those with hyperandrogenism and regular cycles (Adams et al., 1986; Franks, 1989). In recent years, there has been a great deal of interest in the metabolic associations of PCOS. The classic syndrome is characterized by a distinctive form of insulin resistance (reviewed by Dunaif, 1996). Women with PCOS have higher fasting and glucose-stimulated insulin concentrations and signi cantly reduced insulin sensitivity compared with weight-matched control subjects. The cause of this abnormality is unclear, but clinical and laboratory-based studies in PCOS have variously pointed to abnormalities of insulin receptor binding, or more plausibly, postreceptor signalling as well as to evidence for a primary abnormality of insulin secretion. Regarding the evidence for a primary pancreatic beta islet cell abnormality, it was demonstrated (Holte et al., 1995) that weight reduction in obese women with PCOS was accompanied by normalization of insulin sensitivity, but that rst-phase insulin secretion in response to an intravenous glucose challenge remained abnormal. These data also illustrate an important principle in understanding the aetiology of PCOS which is that, whatever the genetic basis for the syndrome, the phenotype can be in uenced by environmental (in this case nutritional) factors. A further phenotypic feature which needs to be considered in identifying potential candidate genes is the polycystic ovarian morphology itself. The polycystic ovary is characterized by the presence of an increased number not only of antral follicles but also of early growing and preantral follicles (Hughesden, 1982). Since these earlier stages of follicular development are thought to be largely independent of gonadotrophins, the implication is that local ovarian factors may have a role in genesis of the polycystic ovary. Many genes have been shown to have an in uence on early folliculogenesis, including those encoding members of the transforming growth factor b superfamily and growth factors signalling through tyrosine kinase-coupled receptors such as insulin-like growth factors I and II and transforming growth factor a. In a sense, any of these are candidate genes for PCOS but, as yet, there are insuf cient functional data to make a convincing target for a candidate gene approach. The candidate gene approach in PCOS Despite the obvious problems posed by the heterogeneous nature of PCOS, the descriptions above clearly indicate there are a number of well-characterized biochemical abnormalities which can provide a sound basis for adopting a candidate gene approach to the identi cation of susceptibility loci. Typically, this approach involves selection of the gene encoding the protein which is thought to function abnormally, identi cation of one (or, preferably, several) informative polymorphisms in, or very close to, the gene in question and the application of both association and linkage studies to determine whether there is any relationship 406

3 Candidate genes in PCOS between those variants and disease risk within populations or families. Association studies may involve the case-control approach, which addresses the question: does the variant allele occur more frequently in a series of women with PCOS than in an appropriate control population? In addition, family-based association methods may be used, such as the transmission disequilibrium test (TDT) (Spielman and Ewens, 1996) in which transmissions from parents to their affected offspring are the focus of analysis. TDT methods have the advantage of avoiding spurious positive associations which can be obtained in casecontrol studies when the two populations are not matched for ethnic background (so-called `population strati cation'). TDT also offers the prospect of assessing `parent-of-origin' effects wherein there is preferential transmission of disease alleles from either the mother or the father to affected offspring. This indeed may be the case in relation to the insulin gene in PCOS, as described below. Whilst TDT relies on one particular con guration of PCOS families, other family structures form the usual substrate for linkage analyses. Such analyses depend on the fact that polymorphic markers within, or closely-linked to, a diseasesusceptibility locus should show a tendency to segregate with the disease in families. A number of computer-assisted methods are available for linkage analysis. Traditionally, parametric logarithm of the odds ratio (LOD)-score based analytical methods have been used when there is clear evidence to support a particular mode of inheritance. Although, as we have seen, some of the available family data do support an autosomal dominant mode of inheritance to PCOS, there remains the concern that incorrect speci cation of the model could lead to reduced power to detect linkage. For this reason we have, for most of our linkage studies, used a non-parametric method of analysis (the GENEHUNTER programme) (Kruglyak et al., 1996) which requires no assumptions to be made about the mode of inheritance. Genes involved in the biosynthesis and metabolism of androgens Genes implicated in the pathway of androgen production and metabolism include those encoding the major endocrine regulator, LH, its receptor, and key P450 steroidogenic enzymes such as cholesterol side chain cleavage (P450scc) and 17-alpha hydroxylase/17-20 lyase (P450c17alpha). Also considered is CYP19, encoding P450 aromatase which is responsible for the conversion of androgen to oestrogen in granulosa cells. LH and its receptor A recent multicentre study investigating polymorphism in the LHb gene showed some interesting variations between populations, but failed to nd a clear causal link with PCOS (Tapanainen et al., 1999). We tested the hypothesis that an activating mutation in the LH receptor gene might be a cause of hyperandrogenism in PCOS, particularly in those subjects with normal serum LH and raised androgen concentrations. Using linkage analysis in families with multiple cases of PCOS, we identi ed ve families in whom polymorphic markers close to the LH gene appeared to segregate with affected status. There was no evidence for linkage in the remaining 18 families and overall, in the 23 families the NPL (non-parametric LOD) score did not reach signi cant levels (Figure 1), Nevertheless, in collaboration with Dr Deborah Segaloff and using DNA from the probands in the ve `affected' families, the relevant coding region of the LH receptor gene was sequenced. No mutations were found (Gharani et al., 1998). These negative data are in keeping with those in recently published studies (Urbanek et al., 1999) wherein a total of 37 potential candidate genes were examined in 150 families with PCOS. CYP11a, encoding P450 cholesterol side chain cleavage A polymorphic sequence [a pentanucleotide repeat ± (tttta) n ] in the 5 regulatory region of CYP11a was identi ed, and both casecontrol association studies and non-parametric linkage analysis were performed. Subjects were assigned to two groups according to genotype. The most common genotype, comprising four repeats occurred with a frequency of 0.59 and was designated 216. Subjects were subdivided according to whether this allele was present (216+) or absent (216±). We found that genotype was associated with serum testosterone concentrations, these being signi cantly higher in women with the 216± genotype (which consists of alleles of six repeats or longer) (Table I) (Gharani et al., 1997). On more detailed analysis, we found that this association held true only in those subjects with clinical evidence of hyperandrogenism. Supportive evidence for the association of PCOS with CYP11a comes from two recent European studies. In the rst (Pugeat et al., 1999), a relationship was reported between the (tttta) n polymorphism and androgen concentrations in 88 hirsute women. These authors found that the CYP11a genotype, together with endocrine markers, predicted the presence of polycystic ovary in hirsute women. In the second study (Diamanti-Kandarakis et al., 2000), it was shown that the CYP11a genotype was associated with both polycystic ovary and total testosterone concentrations in a case±control study. Further support for this notion was obtained from mutation screening of the CYP11a promoter in a 1.85 kb region 5 to the start site of translation. Direct sequencing of fragments (ampli ed by polymerase chain reaction) of DNA samples from affected and unaffected family members was carried out. Apart from the pentanucleotide repeat polymorphism at position ±466 from the start site of transcription (and a previously identi ed dinucleotide repeat polymorphism at position ±1314), no variation from the published sequence was found (Gharani et al., 1997). Structure± function studies of the promoter region using expression systems need to be performed to explore the putative functional role of this element. An alternative explanation is that this polymorphic marker is in linkage disequilibrium with the disease locus which may be located outside the promoter region itself. We examined the segregation of CYP11a in 20 families. With the aid of a number of polymorphic markers (D15S153, D15S125, CYP11a (ac) n, D15S169 and D15S211), in the region of CYP11a, we carried out non-parametric linkage analysis using the GENEHUNTER (multipoint linkage) programme (Gharani et al., 1997). Evidence was found for excess allele sharingði.e. linkageðat the CYP11a locus (NPL score 3.03, P = 0.003). In a parallel parametric analysis, assuming the autosomal dominant model for inheritance, we found evidence of genetic heterogeneity between families with ~60% of families showing linkage at the CYP11a locus. Thus, data from both association and linkage studies suggest that this is a major susceptibility locus for 407

4 S.Franks, N.Gharani and M.McCarthy Figure 1. Non-parametric LOD (NPL) plot for the chromosome 2 linkage analysis in the region of the LH receptor gene. NPL plot is a graphic representation of the non-parametric linkage results. The dashed line represents the mean of expected allele sharing distribution. There is no evidence for excess allele sharing across the entire map. cm = centimorgan. Table I. Comparison of CYP11a genotype distributions between different subgroups in the case±control data set Genotypes ± P-value(P c ) Analysis by quartile of serum testosterone Top quartile (0.03) 50th±75th percentile th±50th percentile 37 7 Bottom quartile 42 7 Analysis by clinical diagnosis: PCOS (0.12) apco/normal controls Quartiles of serum testosterone were obtained from the natural log distribution for the total data set; concentrations quoted are antilog values. Top quartile = women with serum testosterone >2.80 nmol/l; 50th±75th percentile = 2.10±2.80 nmol/l; `25th±50th percentile' = 1.50±2.10 nmol/l; Bottom quartile = <1.50 nmol/l for the entire data-set (affecteds and controls). Genotypes = observed number of individuals with either the 216+ or the 216± genotype for each subgroup. The `216+' genotype includes individuals with at least one 216 allele, and `216±' includes those with no 216 allele. PCOS group = women with polycystic ovaries and symptoms of anovulation and/or hirsutism. apco/normal controls = asymptomatic women with polycystic ovaries and control subjects with normal ovarian morphology. Chi-square contingency tables were used to compare genotype distributions between the different groups [both the uncorrected and corrected (P c ) P-values are given]. hyperandrogenism in PCOS. In one study (Urbanek et al., 1999), linkage analysis at this locus in 39 affected sibling-pairs also yielded a `nominally signi cant' result: the chi-square analysis showed a signi cant association between this locus and PCOS, but this was no longer signi cant after correction for multiple analyses. Although this remains to be con rmed in larger series, these ndings support the view that CYP11aÐor, conceivably a gene in linkage disequilibrium with (i.e. very close to) CYP11aÐ is causally related to PCOS. CYP17, encoding 17-a hydroxylase, 17/20-lyase Because of the reported abnormalities in regulation of 17- a hydroxylase/17±20 lyase in PCOS (Rosen eld et al., 1990), our initial studies focused on the role of CYP17 (coding for P450c17alpha) (Carey et al., 1994). Results of a preliminary case-control study suggested that a variant form of CYP17 was associated with PCOS, but that there was no relationship between genotype and serum testosterone concentrations. Subsequent, larger, case-control studies (from our own group as well as from other centres) have also been unable to con rm the putative association (Gharani et al., 1996; Franks et al., 1997; Urbanek et al., 1999). Furthermore, linkage analysis excluded CYP17 as a major susceptibility gene for PCOS within families (Carey et al., 1994). CYP19, encoding P450 aromatase Neither our linkage studies (Gharani et al., 1997) nor association studies (Urbanek et al., 1999) have revealed any evidence that variation at the CYP19 locus plays a signi cant part in the aetiology of PCOS. Genes involved in the secretion and action of insulin The insulin gene variable number tandem repeat (INS-VNTR) There is evidence that the insulin gene (INS) variable number tandem repeat (VNTR) is a major susceptibility locus for PCOS (Waterworth et al., 1997). The INS-VNTR lies in the 5 regulatory region of the gene; it has been shown to be involved in regulation of insulin gene expression and has been implicated in the aetiology of type 2 diabetes. We found that class III alleles in the VNTR were associated with anovulatory PCOS in two independent populations using two different methods of analysis (casecontrol studies and by the use of affected family based controls; AFBAC). With the aid of the GENEHUNTER linkage analysis programme, we also established that there was excess allele sharing at the INS-VNTR locus. The geometric mean of fasting serum insulin concentrations was signi cantly higher in families 408

5 Candidate genes in PCOS Figure 2. Association of class III alleles of the insulin gene VNTR in 69 parent-offspring `trios' with PCOS [i.e. patient with PCOS ( lled symbol) and her parents (open symbols)], using the transmission disequilibrium test (TDT). Alleles at this polymorphic region of the insulin gene VNTR have a biphasic distribution comprising shorter class I and longer class III alleles. Note preferential transmission of class III alleles [20 of 26 (77%) are class III alleles] from father to affected offspring compared with near equal transmission of class III and class I alleles (47% versus 53%) from the mother (this is a so-called `parent-of-origin' effect). in which linkage was demonstrated than in those families without evidence of linkage (Waterworth et al., 1997). This suggests a functional role for the VNTR variant in the expression of hyperinsulinaemia/insulin resistance in PCOS. These data have recently been extended (Ong et al., 1999). In contrast, others (Urbanek et al., 1999) found no evidence for excess allele sharing at this locus in their population, although it may be relevant that their diagnostic criteria for PCOS differed from those that we have used. We have also observed, using TDT analysis, a `parent of origin' effect in the transmission of alleles of the VNTR to affected subjects. Class III alleles were transmitted signi cantly more commonly from fathers than from mothers to affected daughter (Bennett et al., 1997) (Figure 2). Interestingly, this nding has been mirrored recently in an analysis of families with type 2 diabetes for which PCOS is a known risk factor (Huxtable et al., 2000). The insulin receptor gene Screening of the insulin receptor gene has been undertaken in two well-characterized populations of hyperinsulinaemic women with PCOS. In the rst study (Conway et al., 1994), the tyrosine kinase domain of the insulin receptor gene was examined in 22 patients, but no abnormalities were found. In a second study (Talbot et al., 1996), molecular scanning of the entire coding region of the gene was performed in 24 hyperinsulinaemic subjects with PCOS, and again no signi cant mutations were detected. Mutations of the insulin receptor gene are therefore unlikely to be a major cause of insulin resistance in PCOS. Evidence was found of an association of the insulin receptor gene locus with PCOS in a TDT analysis, but this effect proved to be non-signi cant after correction for multiple testing (Urbanek et al., 1999). Genes involved in folliculogenesis The follistatin gene As part of a panel of candidate genes related to gonadotrophin action, the follistatin locus on chromosome 5 was examined (Urbanek et al., 1999). Somewhat unexpectedly, the strongest evidence was found for linkage with PCOS of any of the 37 candidate genes chosen. In a affected sibling-pair analysis, 72% of sisters were concordant for the follistatin genotype, and this remained signi cant after correction for multiple testing (Urbanek et al., 1999). However, recent follow-up data from the same group suggest that this nding is no longer signi cant when further families are added to the database (Urbanek, 2001). This nding nevertheless remains intriguing, and the possibility arises that this and other genes implicated in folliculogenesis may have a causal role in this disorder which is, after all, characterized by disordered follicle development (Franks et al., 2000). References Adams, J., Polson, D.W. and Franks, S. (1986) Prevalence of polycystic ovaries in women with anovulation and idiopathic hirsutism. Br. Med. J., 293, 355±359. Bennett, S.T., Todd, J.A., Waterworth, D.M. et al. (1997) Association of insulin gene VNTR polymorphism with polycystic ovary syndrome (letter). Lancet, 349, 1771±1772. Carey, A.H., Chan, K.L., Short, F. et al. (1993) Evidence for a single gene effect in polycystic ovaries and male pattern baldness. Clin. Endocrinol., 38, 653±658. Carey, A.H., Waterworth, D., Patel, K. et al. (1994). Polycystic ovaries and premature male pattern baldness are associated with one allele of the steroid metabolism gene CYP17. Hum. Mol. Genet., 3, 1873±1876. Conway, G.S., Avey, C. and Rumsby, G. (1994) The tyrosine kinase domain of the insulin receptor gene is normal in women with hyperinsulinaemia and polycystic ovary syndrome. Hum. Reprod., 9, 1681±1683. Diamanti-Kandarakis, E., Bartzis, M.I., Bergiele, A.T. et al. (2000) Microsatellite polymorphism (tttta)(n) at ±528 base pairs of gene CYP11a alpha in uences hyperandrogenemia in patients with polycystic ovary syndrome. Fertil. Steril., 73, 735±741. Dunaif, A. (1996) Insulin resistance and the polycystic ovary syndrome: mechanism of action and implications for pathogenesis. Endocr. Rev., 18, 774±800. Franks, S. (1989) Polycystic ovary syndrome: a changing perspective. Clin. Endocrinol., 31, 87±120. Franks, S. (1995) Medical Progress Article: polycystic ovary syndrome. N. Engl. J. Med., 333, 853±861. Franks, S., Gharani, N., Waterworth, D. et al. (1997) The genetic basis of polycystic ovary syndrome. Hum. Reprod., 12, 2641±2648. Franks, S., Mason, D. and Willis, D. (2000) Follicular dynamics in the polycystic ovary syndrome. Mol. Cell. Endocrinol., 163, 49±

6 S.Franks, N.Gharani and M.McCarthy Gharani, N., Waterworth, D.M., Williamson, R. and Franks, S. (1996) 5 polymorphism of the CYP17 gene is not associated with serum testosterone levels in women with polycystic ovaries (letter). J. Clin. Endocrinol. Metab., 81, Gharani, N., Waterworth, D.M., Batty, S. et al. (1997) Association of the steroid synthesis gene CYP11a with polycystic ovary syndrome and hyperandrogenism. Hum. Mol. Genet., 6, 397±402. Gharani, N., Segaloff, D., Guo, S.-Y. et al. (1998) Molecular scanning of the LH receptor gene in hyperandrogenic women with polycystic ovary syndrome. Proceedings of the Endocrine Society, 80th Annual Meeting, New Orleans. Abstract P Gilling-Smith, C., Willis, D.S., Beard, R.W. and Franks, S. (1994) Hypersecretion of androstenedione by isolated theca cells from polycystic ovaries. J. Clin. Endocrinol. Metab., 79, 1158±1165. Holte, J., Bergh, T., Berne, C. et al. (1995) Restored insulin sensitivity but persistently increased early insulin secretion after weight loss in obese women with polycystic ovary syndrome. J. Clin. Endocrinol. Metab., 80, 2586±2593. Hughesden, P.E. (1982) Morphology and morphogenesis of the Stein- Leventhal ovary and of so-called hyperthecosis. Obstet. Gynecol. Surv., 37, 59±77. Huxtable, S.J., Saker, P.J., Haddad, L. et al. (2000) Analysis of parentoffspring trios provides evidence for linkage and association between the insulin gene and type 2 diabetes mediated exclusively through paternally transmitted Class III variable number tandem repeat alleles. Diabetes, 49, 126±130. Knochenhauer, E.S., Key, T.J., Kahsar-Miller, M. et al. (1998) Prevalence of the polycystic ovary syndrome in unselected black and white women of the south-eastern United States: a prospective study. J. Clin. Endocrinol. Metab., 83, 3078±3082. Kruglyak, L., Daly, M.J., Reeve-Daly, M.P. et al. (1996) Parametric and nonparametric linkage analysis: a uni ed multipoint approach. Am. J. Hum. Genet., 58, 1347±1363. Legro, R.S. (1995) The genetics of polycystic ovary syndrome. Am. J. Med., 98 (Suppl. 1A), 9S±16S. Legro, R.S., Driscoll, D., Strauss, J.F., III et al. (1998) Evidence for a genetic basis for hyperandrogenemia in polycystic ovary syndrome. Proc. Natl Acad. Sci. USA, 95, 14956± Nelson, V.L., Legro, R.S., Strauss, J.F., III and McAllister, J.M. (1999) Augmented androgen production is a stable phenotype in propagated theca cells from polycystic ovaries. Mol. Endocrinol., 13, 946±957. Ong, K.K., Phillips, D.I., Fall, C. et al. (1999) The insulin gene VNTR, type 2 diabetes and birth weight. Nature Genet., 21, 262±263. Pugeat, M., Cousin, P., Nicolas, M.H. et al. (1999) Pentanucleotide repeat polymorphism in the promoter region of the human CYP11a gene and polycystic ovaries in hirsute women. Proceedings of the Endocrine Society, 81st Annual Meeting, San Diego California. Abstract OR29-1. Rosen eld, R.L., Barnes, R.B., Cara, J.F. and Lucky, A.W. (1990) Dysregulation of cytochrome P450c 17 alpha as the cause of polycystic ovarian syndrome. Fertil. Steril., 53, 785±791. Simpson, J.L. (1992) Elucidating the genetics of polycystic ovary syndrome. In Dunaif, A., Givens, J.R., Haseltine, F.P. and Merriam, G.R. (eds). Polycystic Ovary Syndrome. Blackwell Scienti c Publications, Oxford, pp. 59±77. Spielman, R.S. and Ewens, W.J. (1996) The TDT and other family-based tests for linkage disequilibrium and association. Am. J. Hum. Genet., 59, 983± 989. Steingold, K., De Ziegler, D., Cedars, M. et al. (1987) Clinical and hormonal effects of chronic gonadotropin-releasing hormone agonist treatment in polycystic ovarian disease. J. Clin. Endocrinol. Metab., 65, 773±778. Talbot, J.A., Bicknell, E.J., Rajkhowa, M. et al. (1996) Molecular scanning of the insulin receptor gene in women with polycystic ovary syndrome. J. Clin. Endocrinol. Metab., 81, 1979±1983. Tapanainen, J.S., Koivunen, R., Fauser, B.C.J.M. et al. (1999) A new contributing factor to polycystic ovary syndrome: the genetic variant of luteinizing hormone. J. Clin. Endocrinol. Metab., 84, 1711±1715. Urbanek, M. (2001) In Dunaif, A.D., Chang, R.J. and Heindel, J. (eds) Proceedings: Polycystic Ovary Syndrome: Basic Biology and Clinical Intervention. National Institute of Environmental Health Sciences (NIH). Urbanek, M., Legro, R.S., Driscoll, D.A. et al. (1999). Thirty-seven candidate genes for polycystic ovary syndrome: strongest evidence for linkage is with follistatin. Proc. Natl Acad. Sci. USA, 96, 8573±8578. Waterworth, D.M., Bennett, S.T., Gharani, N. et al. (1997) Linkage and association of insulin gene VNTR regulatory polymorphism with polycystic ovary syndrome. Lancet, 349, 986±989. Zawadzki, J.K. and Dunaif, A. (1992) Diagnostic criteria for polycystic ovary syndrome: towards a rational approach. In Dunaif, A., Givens, J.R., Haseltine, F.P. and Merriam, G.R. (eds). Polycystic Ovary Syndrome. Blackwell Scienti c Publications, Oxford, pp. 377±384. Received on May 18, 2000; accepted on February 8,