Genetics of Polycystic Ovary Syndrome

Polycystic ovary syndrome (PCOS) is a complex disease, one whose clinical manifestations and even definition are an ongoing controversy among researchers in the field. Yet the initial forays into its genetic roots have suggested a simple solution: most of the family studies have shown a dominant form of inheritance, and positive findings have been reported with candidate genes in both association and linkage studies. However, closer study of the families and more extensive genetic analysis is likely to reveal a genetic complexity to the syndrome that matches its clinical diversity. Despite these words of caution, genetic investigation of women with PCOS and their families may provide major insight into this common endocrine abnormality as well as into many of the metabolic sequelae that have been reported, including diabetes, gynecologic cancers, and heart disease.

The purpose of this chapter is to review the literature on the ethnicity and genetics of PCOS, discuss some preliminary results from the molecular genetic investigations into PCOS, and discuss our model and preliminary results of research.

There is no consensus as to the clinical definition of PCOS. Unfortunately, to date, there is no consistent clinical marker or phenotype that is unique to PCOS distinguishing it from other forms of hyperandrogenism. As a reference point, we will use the criteria based on the 1990 NIH-NICHD conference on PCOS: hyperandrogenism and/or hyperandrogenemia, oligo-ovulation, and exclusion of other known disorders, such as congenital adrenal hyperplasia, hyperprolactinemia, or Cushing's syndrome.¹ This may be summarized as unexplained hyperandrogenic chronic anovulation. We have further narrowed this definition to the documentation of hyperandrogenemia (testosterone or unbound testosterone more than two standard deviations above a control group of cycling reproductive-age women), combined with six or fewer episodes of vaginal bleeding a year.

Although many other groups have relied on the basic concept of unexplained hyperandrogenic chronic anovulation, many have accepted the presence of hirsutism or acne as evidence of hyperandrogenism, instead of serum androgen assays. Others have chosen different androgen assays (dehydroepiandrosterone sulfate [DHEA-S] or androstenedione) or have relied on dynamic testing to expose abnormalities in steroidogenic enzymes. Others have relied on indirect evidence of hyperandrogenism, such as gonadotropin abnormalities (elevated levels of luteinizing hormone [LH] or elevated ratio of LH to follicle-stimulating hormone [FSH]). Each group has used different criteria for excluding phenocopies of the syndrome or has consciously included them in the studies. Others have downplayed the clinical criteria of hyperandrogenic chronic anovulation and instead have focused on the ovarian morphology, more recently as determined by ultrasound examination, to determine affected status.

These definitions must be carefully examined in any study of PCOS, especially the genetics of PCOS, because there is no universal concept of what the phenotype is and therefore no universal concept of who is affected and who is unaffected.

The prevalence of PCOS among the population and within families varies according to the diagnostic criteria used. Each paper must be interpreted within that framework (i.e., the limitations of the definition used), and findings based on one set of criteria may not be a priori transferred to a set of families or a population defined on the basis of other criteria. Most of the criteria used for diagnosing the syndrome are continuous traits such as the extent of hirsutism, circulating androgen levels, extent of menstrual irregularity, and even the morphology of the ovary. Some definitions are based on comparison to a “normal” population, but just as there are many definitions of PCOS, there are also many definitions of normal used for the control women. Some criteria have a more arbitrary cutoff for assigning abnormal status. For instance, what is the absolute number of subcapsular small follicles that are required to have polycystic ovaries? This is viewed as a nominal trait, but most likely it represents a continuous trait. What absolute value on a Ferriman-Galway hirsutism assessment qualifies as hirsute—is it a score of 6 or 8 or 10? Even the designation of menstrual irregularity, which we would consider the sine qua non of the syndrome (or at least of its underlying metabolic sequelae) is subject to the same scrutiny. Does a woman with eight menses a year not qualify as a candidate for a diagnosis of PCOS? How about one with 36 episodes?

It is within the context of the definition that these genetic epidemiology studies must be interpreted. The best evidence for the population distribution of one of the frequently used criteria for PCOS is that of polycystic ovaries. Polycystic ovaries are defined on ultrasound according to the frequently cited criteria of Adams and associates,² which include the presence of 8 or more peripheral follicular cysts of 10 mm or less with increased central ovarian stroma. When Polson and colleagues³ examined a large group of volunteers from the general population in England, they found that 22% of 257 women had polycystic ovaries by ultrasound examination; however, one third of these had regular menstrual cycles. A similar prevalence of polycystic ovary morphology detected by ultrasound was found in a New Zealand population.⁴ Other studies have confirmed that about 25% of normal, cycling women have polycystic ovaries on ultrasound examination.⁵ When a subset of women with polycystic ovaries were evaluated endocrinologically, fewer than half had an abnormally elevated testosterone level.⁶ However, as the criteria for diagnosis of the endocrine syndrome were expanded to include symptoms (irregular menses and/or hirsutism) or a biochemical abnormality (elevated testosterone and/or LH level), eventually 92% of these women with polycystic ovaries had another abnormality. Of course, it is tempting to postulate what percentage of women with normal ovaries on ultrasound would also have one of these abnormalities. Polycystic ovaries do not necessarily indicate the presence of the endocrine syndrome.

The prevalence of menstrual disorders among reproductive-age US women was reported to be 53 per 1,000 women, according to National Health Interview Survey data.⁷ This was the most common gynecologic condition reported and accounted for more than 50% of all gynecologic complaints. These included women up to age 50 (many of whom may be menopausal or have experienced amelioration of their menstrual disorders with age) and excluded women less than 18 (PCOS is thought to present with perimenarchal menstrual disorders), so it may tend to underestimate the prevalence of menstrual disorders among what we refer to as reproductive-age women. In our experience, up to 80% of women who are recruited on the basis of oligomenorrhea (6 or fewer menses per year) have PCOS. It is more difficult to determine the range of normal testosterone values in the female population; this is very much dependent of the type of assay used as well as the source of the normal controls.

Thus, it is difficult from these diverse data to predict what percentage of the female population may have the full-blown endocrine syndrome of PCOS. Extrapolating from the menstrual disorder prevalence data and the ovarian morphology data, we have estimated that 5% to 10% of the female population may be affected.

Ethnic Studies

Although we lack the population-based knowledge to define the prevalence of many of the criteria of PCOS, there have been multiple case reports or series suggesting that it exists in most major ethnic groups, although the phenotype varies according to ethnicity. Aono and coworkers⁸ identified a group of 11 Japanese women with polycystic ovaries identified on laparoscopy or laparotomy who had a significantly elevated mean testosterone level and LH/FSH ratio compared to ethnic controls. Patients with polycystic ovaries had an exaggerated response of LH secretion to both a gonadotropin-releasing hormone infusion and a conjugated estrogen infusion to the same degree as had previously been reported in PCOS patients from US and European groups.

Carmina and associates⁹ studied a cohort of 75 patients with hyperandrogenic chronic anovulation composed of 25 Japanese, 25 Italian, and 25 Hispanic Americans compared to ethnic controls. Participants were characterized on the basis of history, physical examination, ultrasonic appearance of the ovaries, levels of gonadotropins and sex steroids, and insulin sensitivity. Women from Japan were less obese and were not hirsute compared to the other ethnic groups. All groups had similar testosterone and LH levels and a similar incidence of polycystic ovaries on ultrasound. Adrenal androgens were elevated in comparable numbers of patients and to a similar degree. Insulin resistance, measured by a dissociation constant of an insulin tolerance test, was significantly elevated but similar in all groups. These data suggest that ethnicity may play a significant role in the phenotype of the syndrome.

Familial Studies

Previous studies of families of PCOS women have found a high incidence of affected relatives. Almost all have suggested a dominant mode of inheritance. On face value, this would suggest a simple genetic disease, one readily amenable to linkage analysis given a mendelian form of inheritance. However, these studies have used various criteria for identifying probands with the syndrome, as well as incompletely or superficially characterizing other family members. Also, no study has been free of the ascertainment bias cited above, which tends to study families with multiple affected females. Many of the studies suffer also from an information bias in attempting to determine reproductive characteristics in persons who are not of reproductive age. For instance, how do you assign affected status to a postmenopausal (or surgically castrated) mother of a proband with PCOS—on her recall of her menstrual history or hyperandrogenic status? What do you do with a sister who is premenarchal? And, of course, all these studies beg the question as to the male phenotype, if this even exists. Table 1 summarizes the suggested mode of inheritance. Table 2 gives the percentage of affected sisters and mothers. The proposed male phenotype in shown in Table 3.

TABLE 1. Summary of Diagnostic Criteria for the Proband in Familial Studies of PCOS and Proposed Mode on Inheritance

PCOS, polycystic ovary syndrome; CAH, congenital adrenal hyperplasia; SHBG, sex hormone binding globulin.

TABLE 2. Summary of Female Relative Affected by Trait in Families of Probands with PCOS

PCOS, polycystic ovary syndrome.

TABLE 3. Proposed Male Phenotypes in Family Studies of PCOS

PCOS, polycystic ovary syndrome.

Cooper and Clayton,¹⁰ in the first larger study of familial PCOS, attempted to identify and characterize other family members of affected white women. The affected patients were all identified to have “Stein-Leventhal syndrome” with both clinical and biochemical abnormalities implied, although the exact manifestations were not identified. All probands, however, had ovaries diagnosed as polycystic, either on the basis of a wedge resection or culdoscopy. Only first-degree female relatives of the identified probands were studied, and they were compared to a control group. A history of oligomenorrhea was more common in mothers and sisters of PCOS patients than controls. Although male relatives were not specifically studied, a questionnaire revealed that male relatives were noted to have increased “pilosity.” This was one of the first published inferences that males could also be affected. The proposed mechanism of inheritance was autosomal dominant with decreased penetrance.

Givens and associates from the University of Tennessee in Memphis^11,12,13 have reported on multiple kindreds showing affected members in several generations. Ethnicity is not specifically stated for each family, but the largest pedigree studied was identified on the basis of a “Black” female. Diagnostic criteria were hirsutism and oligomenorrhea with enlarged ovaries. Some pedigree members were studied in considerable detail. Controls were not used as in other studies. These studies were the first to reveal some of the severe metabolic sequelae that may accompany the syndrome (e.g., diabetes mellitus, insulin resistance, lipid abnormalities, hypertension, and arteriosclerosis). The study of these pedigrees also underscored the variability of phenotype in PCOS, even within the same kindred.

Some males were also studied more closely for the first time. In one kindred, there were several males with oligospermia and one with Klinefelter's syndrome (47,XXY).¹⁴ Elevated LH levels were discovered in some males. The researchers concluded that there is abnormal gonadotropin secretion and testicular function in some male kindred members.

When the researchers classified female kindred members on the basis of hirsutism and oligomenorrhea, they found a high percentage of females affected through both maternal and paternal transmission, although the paternal transmission appeared stronger. This would suggest inheritance in an X-linked dominant manner, although in later publications a probable dominant mode of inheritance was emphasized.

Ferriman and Purdie¹⁵ reported on a larger group of 700 hirsute patients with or without oligomenorrhea. The affected group was classified on the basis of hirsutism and enlarged ovaries (documented by an outmoded air-contrast technique known as gynecography). A significantly higher prevalence of hirsutism, oligomenorrhea, and infertility was noted among first-degree relatives of hirsute women than nonhirsute or control women. Also noted in this study, on the basis of a questionnaire, was an increased incidence of baldness among male relatives of a subgroup of hirsute female patients. Patients and affected family members were not systematically characterized either clinically or endocrinologically. The authors concluded that the mode of inheritance was a “modified dominant form(s) of inheritance.”

More recent studies from Europe have focused on polycystic ovaries identified by ultrasound to characterize PCOS. Hague and colleagues¹⁶ used high-resolution ultrasonography to identify polycystic ovaries in women presenting to a reproductive endocrinology clinic complaining of menstrual disturbances, hyperandrogenic phenomena, obesity, and infertility. First-degree female relatives were then subjected to ovarian ultrasound examination. Males were not studied, and ethnicity was not stated. The ultrasonic appearance of the ovaries was considered to be a more sensitive diagnostic marker than either symptoms or biochemical markers. Segregation ratios were in excess of those expected in an autosomal dominant mendelian inheritance.

Lunde and co-workers¹⁷ studied a group of 132 Norwegian women who had been identified on the basis of an ovarian wedge resection compared to a control group. Criteria for inclusion as a proband included “multicystic ovaries” and two or more of the following symptoms: menstrual irregularities, hirsutism, infertility, and obesity. Findings were consistent with the earlier study by Ferriman and Purdie. Female first-degree relatives of PCOS patients had a significantly higher percentage of polycystic ovary-related symptoms (hirsutism, menstrual irregularity, and infertility) compared to controls; male first-degree relatives of PCOS patients were more likely to have early baldness or excessive hairiness compared to controls. No clear mode of inheritance was ascertained, although the authors considered the findings to be consistent with an autosomal dominant mode of inheritance for a large number of the families.

A report by Carey and colleagues⁶ in which affected probands and family members were more fully characterized suggested a single gene with an autosomal dominant pattern of transmission as the cause of polycystic ovaries. Probands were identified on the basis of ovarian morphology on ultrasound as per the criteria reviewed above. Probands and family members, including some males, underwent a more extensive evaluation, consisting of history; measurement of physical indices and hirsutism; measurement of serum androgens and other steroids, including 17-OH progesterone, gonadotropins, and prolactin; assessment of insulin resistance by an oral glucose tolerance test in obese patients; and ultrasonic visualization of polycystic ovaries in women. Fourteen families were identified, although information was available on only 10 families (of varying ethnicities) to perform classic segregation analysis. Affected status was assigned in first-degree relatives on the basis of ultrasound examination consistent with polycystic ovaries, and in the extended family in some cases on the basis of a positive history suggestive of PCOS. First-degree female relatives were found to have a 51% chance of being affected. Premature balding was found to be an accurate phenotype for male carriers. If male pattern baldness is accepted as the male phenotype, the segregation is consistent with autosomal dominant inheritance.

Continuing this encouraging trend toward more intensive phenotyping of families of PCOS probands, Norman and associates¹⁸ reported that polycystic ovaries and male pattern baldness are common in first-degree female and male relatives, respectively. This group also reported that many family members were affected by hyperinsulinemia and hypertriglyceridemia compared to control groups. This may in fact have been related to the fact that one of the diagnostic criteria for PCOS in the probands was decreased levels of sex hormone binding globulin. Sex hormone binding globulin is thought to be inversely regulated by circulating insulin levels,¹⁹ so this factor may have selected for hyperinsulinemia in the families. Also, this study examined only a small number of families (five).

Twin Studies

Until recently, there was a relative dearth of twin studies of PCOS. Case reports have identified affected sets of female twins.^20,21 A larger twin study from Australia by Jahanfar and colleagues²² reported on twins, both mono- and dizygotic, who were studied with ultrasound as well as clinical and biochemical parameters. The ethnicity of the twins and controls was not discussed. From a starting population of 500 female-female twins who were contacted to participate, eventually only 34 pairs were analyzed. There was also an unusually high incidence of polycystic ovaries on ultrasound, with 50% of the study population affected. This study noted a high degree of discordance among the twins for polycystic ovaries on ultrasound. The study suggested that PCOS may have a more complex inheritance pattern than autosomal dominant, perhaps X-linked or polygenic. It also suggested that environmental factors may play a significant role. There also appeared to be a significant genetic component to the fasting insulin level, further supporting insulin resistance as a potential familial characteristic.

Summary

The uncertain phenotypic criteria make assignment of affected status difficult, and different authors have used different criteria. Many of the studies rely mainly on historical criteria to do this. PCOS remains a diagnosis of exclusion, and many studies have failed to systematically exclude potential phenocopies. Few of the studies have fully characterized the endocrinologic and metabolic sequelae of multiple pedigrees. The male phenotype remains uncertain and incompletely studied, although the common thread seems to be some disorder of androgen metabolism. Despite these shortcomings, the study of familial aggregates has consistently suggested that the mode of inheritance appears to be dominant. This fact, in and of itself, would tend to exclude many of the other rare etiologies of hyperandrogenism, such as steroidogenic enzyme deficiencies, which are autosomal recessive.

Genetic investigations of PCOS may be divided into chromosomal/human leukocyte antigen (HLA) studies, direct sequencing of candidate gene regions, association studies, and linkage studies. Karyotypes were the first genetic tools used in the study of PCOS. There have been isolated case reports or small series reporting polyploidies²³ and aneuploidies, specifically X chromosomal aneuploidies. These include XX/XXX and XX/XO mosaics.^24,25 Larger cytogenetic series of PCOS patients, however, have found normal karyotypes. Stenchever and colleagues²⁶ reported normal karyotypes in 41 patients; Knorr and associates²⁷ reported the same in 16 patients. HLA association studies of familial groupings of PCOS have shown conflicting results. Mandel and coworkers²⁸ studied four families with two affected siblings and found no linkage to the HLA types studied. Hague and associates²⁹ reported an association with DRW6 in 75 patients with polycystic ovaries compared to 110 control women. However, when 16 families with PCOS were studied, no linkage with an HLA was noted. In a similar but smaller study, Ober and colleagues³⁰ reported an association with DQA1*0501 among 19 women with PCOS compared to 46 controls.

Multiple genetic causes of adult-onset hyperandrogenism and chronic anovulation have been identified. The prevalence of many of these mutations among hyperandrogenic women is still being established, although they tend to be rare compared to the relative frequency of PCOS. Mutations in steroidogenic enzymes such as in the 21-hydroxylase gene and now the 3β-hydroxysteroid reductase gene have been identified.³¹ Mutations in the insulin receptor have also been identified in women with frank diabetes mellitus and significant insulin resistance with accompanying hyperandrogenism to the point of virilization.³² However, these have not yet been documented in PCOS. Complete sequencing of all 22 exons of the insulin receptor in 2 women with PCOS has not yielded any mutations,³³ nor has molecular scanning of the insulin receptor gene in 24 women with PCOS.³⁴ Despite these initial results, positive association and linkage have been reported with an insulin gene variable number of tandem repeats (VNTR) locus (see below).

Association Studies

Association studies are not concerned with familial inheritance of specific traits. They also offer the advantage of studying diseases whose mode of inheritance is uncertain and whose presentation is poorly defined or subject to late onset or variable penetrance. They represent a form of case-control study where a comparison is made between the frequency of a given allele in unrelated affected persons compared to unrelated unaffected persons in a given population. An allele is considered to be significantly associated with the trait if it is more often found in the affected group than the unaffected group. Several case-control studies have found a positive association between PCOS and alleles of candidate genes.

There are many potential areas of criticism for these types of studies. Perhaps the most significant is the fact that the allele frequency may vary significantly, depending on the ethnicity of the population studied. Thus, comparisons between mixed populations may be spurious due to the different ethnic makeup of the two groups. Choice of a control group and the means used to exclude affected persons are another source of bias in these studies.

The positive associations reported here, one with a neurotransmitter receptor, two with steroidogenic enzymes, and one with the insulin gene, form a diverse but plausible group of candidate genes (Table 4). However, these studies are all relatively small and thus suffer from a potential type I error. This error has already been demonstrated in the association reported with CYP17.³⁵ This finding was not replicated by another published study.³⁶ When the investigators expanded the size of their case-control set, the significant association disappeared.³⁷ The ethnicity of the cases and controls in three of the positive associations are not specifically mentioned. In the same paper, families of probands used in the linkage studies discussed below are from a variety of ethnic backgrounds, including white, Iranian, Asian, and Afro-Caribbean. The association study by Legro and associates,³⁸ although confined to one ethnic group, phenotyped cases and controls only on the basis of menstrual history and serum testosterone values.

TABLE 4. Positive Associations with Candidate Alleles Reported in PCOS

PCOS, polycystic ovary syndrome.
* This association was no longer found by the authors to be significant on testing a larger sample size.

In the same sense, the negative associations between PCOS and such genes as the dopamine D2 receptor,³⁹ the androgen receptor,⁴⁰ the insulin receptor,⁴¹ and glycogen synthetase⁴² may be type II errors, and larger sample sizes may still reveal significant associations. Given the bias against negative results, it may be inferred that many more negative associations have never been discovered and not reported. This is also an important point in terms of the above questions of type I errors. Most of the associations found have been tentative, as suggested by the broad confidence intervals of the odds ratios reported. Each association test may be considered independent of another, and statistical correction is necessary to control for this. Lander and Schork⁴³ recommended that in each n loci with K alleles tested, the probability value should be adjusted by an n(k - 1) factor. This would require a p value of 0.00001 if one were testing 100 markers with 6 alleles each. This sort of multiple testing also lends itself to post hoc division of phenotypes to better fit tentative genetic association.

Association studies such as the ones discussed above are useful as preliminary tests for an association between candidate genes and the disease phenotype. All of these criteria are applicable to the study of PCOS. Only the affected patients (and, where relevant, unaffected controls) need be studied. This would eliminate some of the difficulty in classifying premenarchal and postmenopausal women as well as men. However, association studies may be weakened by the heterogeneity, both genetic and nongenetic, of the syndrome.

Spielman and Ewens⁴⁴ suggested testing positive associations with a transmission disequilibrium test. This involves looking at the parents of the affected proband and at the inheritance of the putative disease allele. The parent who is heterozygous for this allele should more often transmit the disease allele to the affected offspring than the other allele(s).⁴⁴ Tests such as these allow for appropriate internal control, lessening the problems of an inappropriate or ethnically diverse control group. A variation of this was done with the insulin VNTR for one of the alleles. It was not initially significant, when the transmission disequilibrium test was done for all heterozygous parents. However, when these data were divided on the basis of parent transmitting, the test was positive for paternal transmission (20 transmissions, 6 nontransmissions; 77% transmission rate, p < 0.0006).⁴⁵

Linkage Analysis

Linkage analyses in PCOS have been mainly performed by the group in London who reported the significant association findings noted above^46,47 (Table 5). Their work has been characterized by simultaneous reporting of both linkage and association studies from separate populations. Given the uncertain mode of inheritance, these investigators have used both parametric and nonparametric analyses in their families. The linkage findings with the insulin VNTR remained significant even when men assigned affected status on the basis of premature balding were excluded from the analysis. Larger linkage studies with either larger kindreds or greater numbers will provide greater power to determine the significance of these initial findings.

TABLE 5. Studies Reporting Significant Linkage in PCOS Families

PCOS, polycystic ovary syndrome.

Available studies suggest there is a strong familial component to PCOS. Certainly one of the largest criticisms of these family studies has been the researchers' failure to investigate fully all kindred members with the same systematic screen for the metabolic and reproductive abnormalities in question. There is no substitute for direct biometric or biochemical proof of the phenotype. Large family clusterings of PCOS offer the best opportunity for identifying unique strains of PCOS. Familial clusterings of PCOS may represent a homogeneous etiology of the syndrome, despite significant phenotypic heterogeneity within a given pedigree. Linkage analysis can be performed between polymorphic markers spaced at regular genetic intervals, and these familial traits may identify critical regions for further investigation. Choosing candidate genes for beginning linkage analysis may prove as unproductive at discerning the true etiology of the syndrome as our clinical investigations to date. These intensive studies are now ongoing by a number of groups, and the preliminary results hold great promise for the future.

Supported in part by NIH grants KO8 HDO1118 to Dr. Legro and GCRC grant M01 RR 10732 to the Pennsylvania State University College of Medicine.