Association of aromatase (CYP 19) gene variation with features of hyperandrogenism in two populations of young women

C. J. Petry1, K. K. Ong1, K. F. Michelmore2, S. Artigas3, D. L. Wingate1, A. H. Balen4, F. de Zegher5, L. Ibáñez3 and D. B. Dunger1,6

1 Department of Paediatrics, University of Cambridge, Cambridge CB2 2QQ, 2 Division of Public Health and Primary Health Care, University of Oxford, Oxford OX3 7LF, UK, 3 Endocrine Unit, Hospital Sant Joan de Déu, University of Barcelona, 08950 Barcelona, Spain, 4 Reproductive Medicine Unit, The General Infirmary, Leeds LS2 9NS, UK and 5 Department of Paediatrics, University of Leuven, 3000 Leuven, Belgium

6 To whom correspondence should be addressed at: Department of Paediatrics, Addenbrooke's Hospital, Hills Road, Cambridge CB2 2QQ, UK. Email: dbd25{at}cam.ac.uk


    Abstract
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
BACKGROUND: Aromatase catalyses the conversion of androgens to estrogens and thus variation in the aromatase gene could contribute to female syndromes of androgen excess, such as precocious pubarche (PP) and polycystic ovarian syndrome (PCOS). METHODS: Two groups, one case–control containing girls from Barcelona, Spain with PP (n=186) or healthy controls (n=71), and the other a population study of young women from Oxford, UK, who volunteered for a study of normal women's health (n=109), were genotyped at four aromatase gene haplotype-tag single nucleotide polymorphisms (SNP). Clinical features and hormone concentrations relevant to hyperandrogenism were compared across haplotypes or genotypes. RESULTS: Distributions of aromatase haplotypes (P<0.0001) and aromatase SNP_50 genotype (P=0.001) were significantly different between PP girls and Spanish controls. The AGGG haplotype was associated with an odds ratio (95% confidence interval) of 0.5 (0.3–0.9) (P=0.005) for the presence of PP compared to GAGG. In 84 post-pubertal PP girls, aromatase haplotype was associated with functional ovarian hyperandrogenism (P<0.05), independently of insulin sensitivity. In the Oxford population, SNP_50 was associated with variation in PCOS symptom score (P=0.008) and circulating testosterone concentrations (P=0.02). CONCLUSIONS: This study suggests that common variation at the aromatase gene (and not just rare loss-of-function mutations) is associated with androgen excess in girls and young women.

Key words: genetic association/insulin resistance/polycystic ovarian syndrome/premature pubarche/testosterone


    Introduction
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
Polycystic ovarian syndrome (PCOS) is characterized by hyperandrogenism and chronic anovulation, commonly leading to infertility. In some populations, women who develop clinical signs of ovarian androgen excess post-menarche present earlier with exaggerated adrenarche or precocious pubarche (PP), i.e. the appearance of pubic hair before the age of 8 years (Ibáñez et al., 1993Go, 2000Go). As with PCOS, girls with PP become insulin resistant and dyslipidaemic, particularly if PP was preceded by low birthweight (Ibáñez et al., 1998Go), suggesting that PP may be an antecedent of PCOS. Sisters of women with PCOS have increased circulating testosterone concentrations (Legro et al., 1998Go) and PP has been associated with genetic variation at the androgen receptor (Ibáñez et al., 2003Go), suggesting that hyperandrogenism in both PP and PCOS may be partly genetically determined.

Aromatase (EC 1.14.14.1) is a member of the cytochrome P450 family of enzymes (subfamily 19) which catalyses the conversion of C19 androgens to aromatic C18 estrogens. It is induced by FSH and is present in a number of different tissues including adrenals, muscle, placenta, skin, adipose and nervous tissue. Reduced aromatase activity may lead to the development of PCOS, since PCOS has been observed in patients with aromatase deficiency caused by rare loss-of-function mutations (Harada et al., 1992Go; Ito et al., 1993Go; Belgorosky et al., 2003Go) and antral follicles taken from PCOS women exhibited no aromatase activity (Takayama et al., 1996Go).

The human aromatase gene is ~130 kb long. Its 10 exons (the final nine of which are coding) are located within 30 kb of each other (Bulun et al., 2003Go), and the 93 kb 5'-flanking region is thought to have a regulatory role. In this study we explored the association between common variation in aromatase and hyperandrogenism in two separate populations of girls and young women, using a haplotype-tag single nucleotide polymorphism (htSNP) genotyping approach.


    Materials and methods
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
Barcelona case–control study
All girls with PP were recruited from among all cases that presented with a PP-like phenotype over a 5 year period. The policy at the Hospital Sant Joan de Déu (Barcelona, Spain) is to follow such patients auxologically until after completion of puberty, and to perform a hormonal assessment sometime pre-pubertally, in mid-puberty and after menarche. Girls were considered eligible for the study if they had PP due to premature adrenarche, as suggested by elevated plasma dehydroepiandrosterone sulphate (DHEAS) and/or androstenedione concentrations at clinical diagnosis (Ibáñez et al., 1993Go). Enrolment exclusion factors included the presence of acanthosis nigricans, thyroid dysfunction, Cushing's syndrome, hyperprolactinaemia, a family or personal history of diabetes mellitus, late-onset congenital adrenal hyperplasia (Ibáñez et al., 2000Go) or the taking of drugs such as oral contraceptives or those known to affect carbohydrate or lipid metabolism. A total of 186 girls with PP fulfilled these criteria and had DNA available for study.

Eighty-four of these subjects were subsequently also studied post-menarche in the follicular phase (days 3–8) of the menstrual cycle, with assessment of biochemical and clinical features of ovarian hyperandrogenism, including hirsutism score, the severity of any acne lesions, serum testosterone, androstenedione, DHEAS and sex hormone-binding globulin (SHBG). They also underwent a 1.75 g/kg body weight (maximum 75 g) 120 min oral glucose tolerance test (Ibáñez et al., 1998Go). Serum 17-hydroxyprogesterone (OHP) levels in response to leuprolide acetate, a GnRH agonist (Procrin, 500 µg sc; Abbott, Spain) were measured, and biochemical ovarian hyperandrogenism was defined as a peak 17-OHP >160 ng/dl (Ibáñez et al., 1994Go).

Healthy controls (n=71) were either short, normal Barcelona-Spanish girls (heights between the 10th and 25th percentiles, i.e. within the familial range and well above what is considered pathological) attending a clinic for concerns about growth (none were subsequently found to have a pathological cause) or were subjects undergoing minor non-endocrine-related surgical procedures such as those on the ear, nose or throat. None of the controls had a history of PP or PCOS or showed any symptoms or signs of androgen excess. The study protocol was approved by the Institutional Review Board of Barcelona University Hospital of Sant Joan de Déu. Informed consent was obtained from parents and/or study subjects, including permission for the collection and genotyping of DNA samples. Birthweight and gestational age data were obtained from hospital records or from the girls' paediatricians, and transformed into SD scores (Ibáñez et al., 1998Go).

Oxford population ptudy
A total of 224 young women volunteers who considered themselves normal were recruited by invitation from general practice surgeries and by advertisement (requesting healthy volunteers to enrol for a study of women's health) in Oxford universities (Michelmore et al., 1999Go, 2001Go). All volunteers were aged between 18 and 25 (mean 21.5) years and 97% of them were Caucasian in ethnic origin (the remainder being Indian or Afro-Caribbean). In these young women the prevalence of polycystic ovaries on transabdominal ultrasound was higher than that reported in other population-based studies (Polson et al., 1988Go; Tayob et al., 1990Go), but no direct evidence of selection bias was detected (Michelmore et al., 2001Go). A total of 115 women who were recruited in the original survey were excluded from this study as they were using oral contraceptives (the only exclusion criterion). Blood samples for hormonal analyses were collected in the fasting state during the early follicular phase of the menstrual cycle. Ovaries were scanned by ultrasound for polycystic appearance. A symptom score for PCOS was calculated on the basis of number of features from: menstrual irregularity, acne, hirsutism, testosterone >3 nmol/l, and LH > 10 IU/l (Michelmore et al., 1999Go, 2001Go). Around one third of the volunteers were shown to have polycystic ovaries, and of these 75–80% had symptoms of PCOS (Michelmore et al., 1999Go, 2001Go). Ethical approval for the study was granted by the Central Oxford Research Ethics Committee and informed consent was obtained from all subjects.

Assays
For the Barcelona cases and controls, serum hormone concentrations [insulin intra- and inter-assay coefficients of variation (CV) were 4.7 and 7.2% respectively, testosterone 9.2 and 10.4%, androstenedione 6 and 11.9%, DHEAS 5.3 and 3.9%, SHBG 3.0% and 4.4%] were measured by radioimmunoassay or enzyme-linked immunosorbent assay (ELISA) kits as previously described (Ibáñez et al., 1993Go, 1998Go, 2000Go). Glucose was measured using a glucose oxidase method. For the Oxford population study, plasma hormone concentrations (including testosterone inter-assay CV <10%, estradiol inter-assay CV < 3%, androstenedione inter-assay CV <7%, 17-hydroxyprogesterone inter-assay CV <4%, SHBG inter-assay CV <5%, LH inter-assay CV <2%, FSH inter-assay CV <3%, insulin CV: intra-assay 8.3% and inter-assay 12.2%, and c-peptide intra-assay CV 3.4% and inter-assay CV 10.0%) were measured by radioimmunoassay or ELISA as previously described (Michelmore et al., 1999Go, 2001Go).

Aromatase genotyping
Linkage disequilibrium (LD) maps of aromatase single nucleotide polymorphisms (SNP) have been constructed to define htSNP that allow high predictability of haplotypes (Haiman et al., 2003Go). Four blocks of strong LD were found, the largest of which covered a 50 kb region containing the entire aromatase coding region and part of the 3'-untranslated region. The haplotypes in this region could be inferred from eight htSNP genotypes (Haiman et al., 2003Go). Close inspection, however, revealed that the increased diversity of aromatase in African-Americans alone accounts for four of the htSNP genotypes necessary to define this region (Haiman et al., 2003Go). In races such as Caucasians and Latinas, haplotypes in this LD-block could be inferred from just four htSNP genotypes, so we adopted this strategy for constructing aromatase coding-region haplotypes in our populations. For convenience, SNP genotypes are reported according to the nomenclature of Haiman et al. (2003)Go, and genotyping was performed on genomic DNA that had been extracted from leukocytes (Ibáñez et al., 1993Go; Michelmore et al., 2001Go). Replicate genotyping on the same DNA gave <1% inconsistent calls throughout.

SNP44 genotyping
SNP 44 (Haiman et al., 2003Go) (hCV8234947, rs12907866) was genotyped using PCR amplification and detection of SNP alleles by restriction fragment length polymorphism (RFLP) analyses. DNA (10 ng) was amplified (20 µl total volume) along with 1 x reaction buffer, 200 µmol/l each dNTP (Promega Ltd, UK), 1 mmol/l magnesium, 12 pmol each primer (forward: 5'-GCATTGCAGGTAGAGAGAATGAGAAAC-3'; reverse: 5'-CCTGCTCCATGTTCTCAAGACAGATATA-3'), 10% (v/v) glycerol and 0.5 IU Biotaq DNA Polymerase (Bioline, UK). The mix was heated to 94°C for 5 min, followed by 20 cycles of 94°C (45 s), 59°C (45 s, dropping 0.5°C per cycle) and 72°C (45 s). After this the samples underwent 15 cycles of 94°C (45 s), 49°C (45 s) and 72°C (45 s), prior to a final incubation at 72°C for 10 min. Ten microlitres of the resulting PCR product was incubated at 65°C for 16 h along with 5 IU of TaqI (New England Biolabs, UK). Separating the resulting products by agarose gel electrophoresis produced a 425 bp band for the ‘A’ allele and 258 and 167 bp bands for the ‘G’ allele.

SNP50 genotyping
SNP 50 (Haiman et al., 2003Go) (rs2414096) was genotyped using mismatch primer PCR–RFLP (Hao et al., 2002Go). DNA (10 ng) was amplified (20 µl total volume) along with 1 x reaction buffer, 200 µmol/l each dNTP, 1.5 mmol/l magnesium, 12 pmol of each primer (forward: 5'-TTGTTACCCTCAAAAAAGCGACC-3'; reverse: 5'-CTCAAACTCCAATCTAGAGGTTCAAAG-3'), 5% (v/v) dimethylsulphoxide and 0.5 IU Biotaq DNA Polymerase. The mix was heated to 94°C for 5 min, followed by 20 cycles of 94°C (45 s), 55°C (45 s, dropping 0.5°C per cycle) and 72°C (20 s). After this the samples underwent 15 cycles of 94°C (45 s), 45°C (45 s) and 72°C (20 s), prior to a final incubation at 72°C for 10 min. Of the resulting PCR product, 10 µl was incubated at 60°C for 16 h along with 2.5 IU of BsiEI (New England Biolabs). This produced a 124 bp band for the ‘A’ allele and 102 and 22 bp bands for the ‘G’ allele.

SNP60 genotyping
SNP 60 (Siegelmann-Danieli and Buetow, 1999Go; Haiman et al., 2003Go) (int 7_14A, IVS7+26, G-A) was genotyped using tetra-primer amplification refractory mutation system (ARMS) PCR (Ye et al., 2001Go). DNA (30 ng) was amplified (10 µl total volume) along with 1 x reaction buffer, 200 µmol/l each dNTP, 2.5 mmol/l magnesium, 10 pmol each inner primer (forward: 5'-CACAGTCATAACATATGTGGCATTGA-3'; reverse: 5'-GAGGTACTGACCTGAACTAACTGTAAGTC-3') and 1pmol each outer primer (forward: 5'-TGGGATTACAGAACTGCTAGAGAAAGTA-3'; reverse: 5'-TAGACAATATTTTGACTTTTTTTCCAGC-3'), 10% (v/v) glycerol and 0.5 IU Biotaq DNA Polymerase. The mix was heated to 94°C for 5 min, followed by 20 cycles of 94°C (45 s), 57°C (45 s, dropping 0.5°C per cycle) and 72°C (45 s). After this the samples underwent 15 cycles of 94°C (45 s), 47°C (45 s) and 72°C (45 s), prior to a final incubation at 72°C for 10 min. This produced a 194 bp band for ‘A’ allele and a 135 bp band for the ‘G’ allele (with a full PCR product band at 274 bp).

SNP64 genotyping
SNP 64 (Haiman et al., 2003Go) (rs4646) was also genotyped by ARMS PCR. The cycling conditions were similar to those for SNP 60, except that the annealing temperatures were 7°C higher throughout. The reaction mix was the same as for SNP 60 with the exception of the inner primer sequences (forward: 5'-GGGTTGTCACCAAGCTAGGTGCTACTT-3prime; reverse: 5prime;-CTGGTGTGAACAGGAGCAGATGGCC-3prime) and outer primer sequences (forward: 5prime;-ACCCCAAGAAACTCAGACAGGTGTCTG-3prime; reverse: 5prime;-TTGTTAATGAAGGCCTATCCTTCTCAAAGCACA-3'). This produced a 223 bp band for ‘G’ allele and a 176 bp band for the ‘T’ allele (with a full PCR product band at 346 bp).

Statistical analyses
All genotype frequencies were assessed for consistency with Hardy–Weinberg equilibrium (the Barcelona samples being assessed separately for cases and controls due to the possibility of natural selection occurring for these loci in the cases). Statistical analyses were performed using SPSS for windows (version 10.0.7) (SPSS Inc., USA). In the Barcelona study there was 80% statistical power to be able to detect a difference of 0.4 SD of a quantitative variable at the quoted SNP 50 allele frequencies. For the Barcelona SNP genotypes there was 80% statistical power to detect a difference between minor allele frequencies of 0.5 and 0.3. In the Oxford population study there was 80% statistical power to be able to detect a difference of 0.6 SD of a quantitative variable at the quoted SNP 50 allele frequencies. Aromatase coding-region haplotypes were reconstructed from the four SNP genotypes using Phase (version II) software (Stephens et al., 2001Go). For both studies, haplotypes were reconstructed using data from the complete set of samples (i.e. in Barcelona both the cases and controls combined and in the Oxford population the full group including those taking oral contraceptives). Haplotypes were expressed in terms of aromatase alleles in the order SNP 44, 50, 60 then 64. Hormone concentrations and quantitative phenotypic data were compared across aromatase genotypes and haplotypes by one-way ANOVA. Where appropriate, adjustment was made for insulin sensitivity [calculated as homeostatic model assessment (HOMA) (Levy et al., 1998Go) from fasting plasma insulin and glucose concentrations], age (at presentation or assessment), puberty stage (when analysing puberty-related hormone concentrations), body mass index (BMI) or birthweight. In the Barcelona study, frequencies of aromatase haplotypes were compared using the {chi}2 goodness-of-fit test. Frequencies of individual haplotypes were compared (presence against absence in cases and controls) by the {chi}2-test.


    Results
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
Barcelona case–control study
Table I shows the baseline phenotypic and biochemical characteristics of the PP girls and controls. Table IITable II shows the frequencies of their aromatase genotypes and haplotypes (along with those from the Oxford population of young women). The distributions of haplotypes (P<0.0001) and SNP 50 genotypes (P=0.001) were significantly different between PP girls and controls. The odds ratios with 95% confidence intervals (CI) relating to the risk of having PP (with reference to the GAGG haplotype, using an additive model) were: AAGG 1.5 (0.7–3.0) (P=0.07), AGGG 0.5 (0.3–0.9) (P=0.005), AGGT 0.6 (0.3–1.0) (P=0.02) and AGAT 0.7 (0.3–1.5) (P=0.1). The AAGG haplotype was proportionally more common in PP girls than in controls (P=0.007) whereas the AGGG haplotype was less common (P=0.02).


View this table:
[in this window]
[in a new window]
 
Table I. Phenotypic and biochemical characteristics of precocious pubarche (PP) girls and controls from Barcelona

 

View this table:
[in this window]
[in a new window]
 
Table IIa. Frequencies of genotypes in aromatase single nucleotide polymorphisms (SNP) expressed in number (%) of subjects in precocious pubarche (PP) girls and controls from Barcelona and young women from Oxford

 

View this table:
[in this window]
[in a new window]
 
Table IIb. Frequencies of aromatase haplotypes expressed as number (%) of chromosomes in precocious pubarche (PP) girls and controls from Barcelona and young women from Oxford

 
Frequencies of all the aromatase SNP genotypes in these samples were generally consistent with Hardy–Weinberg equilibrium. However, SNP 50 genotypes from girls with PP were not in Hardy–Weinberg equilibrium (P<0.005), unlike genotypes from the Spanish controls which were. Variation in this SNP genotype was not associated with variation in estradiol, insulin sensitivity or birthweight, but SNP 50 genotype was associated with plasma testosterone concentrations, after adjustment for puberty stage (P=0.03; Table III) and this persisted after adjustment for birthweight SDS (P=0.03). Adjusting for puberty stage also revealed significant associations with DHEAS concentrations (P=0.018) and insulin sensitivity (P=0.039; Table III).


View this table:
[in this window]
[in a new window]
 
Table III. Associations between single nucleotide polymorphism (SNP) Arom 50 genotypes and serum concentrations of testosterone and DHEAS and insulin sensitivities in a mixed cohort of girls with premature pubarche (PP) and healthy controls from Barcelona

 
In 84 post-pubertal PP subjects for whom data were available, aromatase haplotype was related to functional ovarian hyperandrogenism (P=0.049), after adjustment had been made for insulin sensitivity (with reduced insulin sensitivity itself being related to functional ovarian hyperandrogenism, P=0.027).

None of the other genotypes and haplotypes were associated with clinical or biochemical features of hyperandrogenism in this group.

Oxford population study
All aromatase SNP genotypes in this cross-sectional study were consistent with Hardy–Weinberg equilibrium. The SNP 50 genotype was associated with PCOS symptom scores (P=0.008 with dominant model; Table IV), but the aromatase coding-region haplotype was not associated with any clinical or biochemical feature of hyperandrogenism (all P>0.05; SNP genotype and haplotype prevalences are shown in Table IITable II). This association with SNP 50 was independent of BMI, insulin sensitivity and ultrasound diagnosis of polycystic ovaries, all of which were independently associated with the PCOS symptom score (all P<0.03). The SNP 50 genotype was also associated with variation in circulating testosterone concentrations (P=0.02 with dominant model; Table IV). There was no detectable association between this genotype and other features of PCOS or with insulin sensitivity (P=0.7), even after adjusting for BMI (P=0.6). Variation in estradiol quartile was also associated with the SNP 50 genotype (P=0.014 with additive model; Table IV) but it was not related to the estradiol to testosterone ratio (P=0.4).


View this table:
[in this window]
[in a new window]
 
Table IV. Associations between aromatase single nucleotide polymorphism (SNP) genotypes and polycystic ovarian syndrome (PCOS) scores, serum concentrations of testosterone and estradiol quartiles in young women from Oxford

 
There were no associations between variation in these phenotypic variables and variation in aromatase SNP 44, 60 or 64 (Table IV).


    Discussion
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
This study demonstrates that common genetic variation in the coding region of aromatase is associated with risk for pre-pubertal hyperandrogenism, and with PCOS risk in both girls and young women. More specifically in our groups an intronic SNP close to exon 3 in aromatase (SNP 50) was associated with PP, post-menarchal ovarian hyperandrogenism and variation in PCOS symptom score. These associations could be explained by the additional association with circulating testosterone concentrations in the two separate groups.

The aetiology of PP, ovarian hyperandrogenism and PCOS has been linked to insulin resistance and obesity (Dunaif, 1997Go). Our findings were independent of insulin sensitivity and BMI. Genotype distributions between PP girls and healthy controls were different, with the SNP 50 ‘A’ allele (which was more prevalent in PP girls) being associated with increased testosterone concentrations in both the Barcelona PP case–control study and the Oxford population study (where it was also associated with increased PCOS symptom scores). Aromatase coding-region haplotypes in both our Barcelona controls and Oxford young women showed similar prevalences to those of Haiman et al. (2003)Go. Variation in them was also shown to be associated with alterations in risk for PP, itself a risk factor for PCOS (Ibáñez et al., 1993Go, 2000Go). In particular, comparing haplotypes differing in the SNP 50 genotype, one variant was associated with a 2-fold decreased risk for having PP. In 84 post-pubertal subjects, reduced insulin sensitivity and aromatase coding region haplotypes were independently associated with functional ovarian hyperandrogenism risk. The lack of confirmation of the haplotype effects in the Oxford population study may well have been due to a lack of sufficient statistical power, with the four SNP genotyped being resolved into nine different haplotypes in that group of only 109 young women.

The significant findings in this study appear to relate to the SNP 50 genotypes. The question therefore arises as to whether this SNPS is causing these associations directly or whether it is just in LD with another causal SNPS. Clearly there are many SNPS in this aromatase block that are in strong LD (Haiman et al., 2003Go), hence the relatively small number of SNP that need to be typed to be used as haplotype tags. Indeed one SNP, only 667 base pairs away from SNP 50 [SNP 51 in Haiman et al. (2003)Go] in exon 3 of the gene, is in strong LD with SNP 50 (Haiman et al., 2003Go) and has already been shown to be associated with breast cancer risk in women (Siegelmann-Danieli and Buetow, 1999Go) and with circulating estradiol concentrations (Somner et al., 2004Go). However, in those studies the issue of causality has not been addressed either. The DNA sequence around SNP 50, when the ‘G’ allele is present, is a consensus branchpoint sequence for pre-RNA processing, albeit further from the 3' splice site than branchpoints are usually found (Buvoli et al., 1997Go). When the ‘A’ allele is present this branchpoint sequence disappears, so if it were physiologically active there may be an alteration in the efficiency of aromatase pre-RNA processing and ultimately aromatase expression. This suggestion remains speculative at present, however.

While our samples sizes may be relatively small for genetic association studies, the detailed phenotype assessments in both populations would increase the power to identify true positive associations. The consistent associations in two very separate populations further increases the validity of the findings and reduces the likelihood of population stratification bias, although family-based transmission studies would be needed to confirm this. The Barcelona control girls were not recruited from a completely unselected population; however, they had no evidence of endocrine or any other pathology. Similarly, while it is possible that the Oxford volunteers may have over-selected themselves for the presence of PCOS features, no selection bias was detected (Michelmore et al., 2001Go), and as in the Barcelona control girls, the genotype and haplotype frequencies were similar to those observed in other populations (Haiman et al., 2003Go). Our associations were largely with hyperandrogenism, and future case–control studies would be needed to confirm the relevance of these findings to actual PCOS risk. Such studies would need to be much larger, in view of the widely heterogeneous pathogenesis of PCOS.

Previously described rare loss-of-function mutations in aromatase have demonstrated that variation in this estrogen biosynthesis pathway can indeed lead to PCOS in females (Harada et al., 1992Go; Ito et al., 1993Go; Conte et al., 1994Go; Morishima et al., 1995Go; Mullis et al., 1997Go; Belgorosky et al., 2003Go). So far the aromatase mutations described have been in exons 5 (Belgorosky et al., 2003Go), 9 (Belgorosky et al., 2003Go; Morishima et al., 1995Go; Mullis et al., 1997Go) and 10 (Ito et al., 1993Go; Conte et al., 1994Go), and in the boundary between exons and introns 3 (Mullis et al., 1997Go) and 6 (Harada et al., 1992Go). Our findings which relate to SNP 50, which is also in the coding-region, suggest that more subtle, common variations in aromatase are also associated with variation in risk for PCOS. Overall PCOS risk may involve oligogenic contributions (Franks et al., 2001Go) [as well as environmental components (Crosignani and Nicolosi, 2001Go)] including that made by aromatase, in addition to other genes such as those of the androgen receptor (Ibáñez et al., 2003Go), insulin receptor (Tucci et al., 2001Go), cholesterol side-chain cleavage enzyme (Gharani et al., 1997Go), follistatin (Urbanek et al., 1999Go) and the insulin VNTR (Michelmore et al., 2001Go). Alternatively, as PCOS is such a heterogeneous disorder, different genetic variants could associate with PCOS risk in different populations (Sanders et al., 2002Go).

In summary, this is the first association study that has found a link between genetic variation in the aromatase gene and androgen excess in females. This was confirmed in two independent populations, one a population survey and the other a case–control study. The fact that these two study groups, which gave very similar associations with aromatase genotypes, were established by different selection criteria gives these findings extra credence. Previously both a linkage (Gharani et al., 1997Go) and an association (Urbanek et al., 1999Go) study failed to find a relationship between aromatase and PCOS, but these studies were performed without the power associated with the use of htSNPS and therefore some of the haplotypic variation in the gene may have been missed.


    Acknowledgements
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
The authors are extremely grateful to all the girls who agreed to participate in this study. This study was funded by the Wellcome Trust.


    References
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
Belgorosky A, Pepe C, Marino R, Guercio G, Saraco N, Vaiani E and Rivarola MA (2003) Hypothalamic–pituitary–ovarian axis during infancy, early and late prepuberty in an aromatase-deficient girl who is a compound heterozygote for two new point mutations of the CYP19 gene. J Clin Endocrinol Metab 88, 5127–5131.[Abstract/Free Full Text]

Bulun SE, Sebastian S, Takayama K, Suzuki T, Sasano H and Shozu M (2003) The human CYP19 (aromatase P450) gene: update on physiologic roles and genomic organization of promoters. J Steroid Biochem Mol Biol 86, 219–224.[CrossRef][ISI][Medline]

Buvoli M, Mayer SA and Patton JG (1997) Functional crosstalk between exon enhancers, polypyrimidine tracts and branchpoint sequences. EMBO J 16, 7174–7183.[Abstract/Free Full Text]

Conte FA, Grumbach MM, Ito Y, Fisher CR and Simpson ER (1994) A syndrome of female pseudohermaphrodism, hypergonadotropic hypogonadism, and multicystic ovaries associated with missense mutations in the gene encoding aromatase (P450arom). J Clin Endocrinol Metab 78, 1287–1292.[Abstract]

Crosignani PG and Nicolosi AE (2001) Polycystic ovarian disease: heritability and heterogeneity. Hum Reprod Update 7, 3–7.[Abstract/Free Full Text]

Dunaif A (1997) Insulin resistance and the polycystic ovary syndrome: mechanism and implications for pathogenesis. Endocr Rev 18, 774–800.[Abstract/Free Full Text]

Franks S, Gharani N and McCarthy M (2001) Candidate genes in polycystic ovary syndrome. Hum Reprod Update 7, 405–410.[Abstract/Free Full Text]

Gharani N, Waterworth DM, Batty S, White D, Gilling-Smith C, Conway GS, McCarthy M, Franks S and Williamson R (1997) Association of the steroid synthesis gene CYP11a with polycystic ovary syndrome and hyperandrogenism. Hum Mol Genet 6, 397–402.[Abstract/Free Full Text]

Haiman CA, Stram DO, Pike MC, Kolonel LN, Burtt NP, Altshuler D, Hirschhorn J and Henderson BE (2003) A comprehensive haplotype analysis of CYP19 and breast cancer risk: the Multiethnic Cohort. Hum Mol Genet 12, 2679–2692.[Abstract/Free Full Text]

Hao K, Niu T, Sangokoya C, Li J and Xu X (2002) SNPkit: an efficient approach to systematic evaluation of candidate single nucleotide polymorphisms in public databases. Biotechniques 33, 822–828.[ISI][Medline]

Harada N, Ogawa H, Shozu M and Yamada K (1992) Genetic studies to characterize the origin of the mutation in placental aromatase deficiency. Am J Hum Genet 51, 666–672.[ISI][Medline]

Ibáñez L, Potau N, Virdis R, Zampolli M, Terzi C, Gussinye M, Carrascosa A and Vicens-Calvet E (1993) Postpubertal outcome in girls diagnosed of premature pubarche during childhood: increased frequency of functional ovarian hyperandrogenism. J Clin Endocrinol Metab 76, 1599–1603.[Abstract]

Ibáñez L, Potau N, Zampolli M, Prat N, Gussinye M, Saenger P, Vicens-Calvet E and Carrascosa A (1994) Source localization of androgen excess in adolescent girls. J Clin Endocrinol Metab 79, 1778–1784.[Abstract]

Ibáñez L, de Zegher F, Francois I and Potau N (1998) Precocious pubarche, hyperinsulinism and ovarian hyperandrogenism in girls: relation to reduced fetal growth. J Clin Endocrinol Metab 83, 3558–3562.[Abstract/Free Full Text]

Ibáñez L, Dimartino-Nardi J, Potau N and Saenger P (2000) Premature adrenarche-normal variant or forerunner of adult disease? Endocr Rev 21, 671–696.[Abstract/Free Full Text]

Ibáñez L, Ong KK, Mongan N, Jaaskelainen J, Marcos MV, Hughes IA, De Zegher F and Dunger DB (2003) Androgen receptor gene CAG repeat polymorphism in the development of ovarian hyperandrogenism. J Clin Endocrinol Metab 88, 3333–3338.[Abstract/Free Full Text]

Ito Y, Fisher CR, Conte FA, Grumbach MM and Simpson ER (1993) Molecular basis of aromatase deficiency in an adult female with sexual infantilism and polycystic ovaries. Proc Natl Acad Sci USA 90, 11673–11677.[Abstract/Free Full Text]

Legro RS, Driscoll D, Strauss JF, 3rd, Fox J and Dunaif A (1998) Evidence for a genetic basis for hyperandrogenemia in polycystic ovary syndrome. Proc Natl Acad Sci USA 95, 14956–14960.[Abstract/Free Full Text]

Levy JC, Matthews DR and Hermans MP (1998) Correct homeostasis model assessment (HOMA) evaluation uses the computer program. Diabetes Care 21, 2191–2192.[Free Full Text]

Michelmore KF, Balen AH, Dunger DB and Vessey MP (1999) Polycystic ovaries and associated clinical and biochemical features in young women. Clin Endocrinol 51, 779–786.[CrossRef][ISI][Medline]

Michelmore K, Ong K, Mason S, Bennett S, Perry L, Vessey M, Balen A and Dunger D (2001) Clinical features in women with polycystic ovaries: relationships to insulin sensitivity, insulin gene VNTR and birth weight. Clin Endocrinol 55, 439–446.[CrossRef][ISI][Medline]

Morishima A, Grumbach MM, Simpson ER, Fisher C and Qin K (1995) Aromatase deficiency in male and female siblings caused by a novel mutation and the physiological role of estrogens. J Clin Endocrinol Metab 80, 3689–3698.[Abstract]

Mullis PE, Yoshimura N, Kuhlmann B, Lippuner K, Jaeger P and Harada H (1997) Aromatase deficiency in a female who is compound heterozygote for two new point mutations in the P450arom gene: impact of estrogens on hypergonadotropic hypogonadism, multicystic ovaries, and bone densitometry in childhood. J Clin Endocrinol Metab 82, 1739–1745.[Abstract/Free Full Text]

Polson DW, Adams J, Wadsworth J and Franks S (1988) Polycystic ovaries—a common finding in normal women. Lancet 1, 870–872.[CrossRef][ISI][Medline]

Sanders EB, Aston CE, Ferrell RE and Witchel SF (2002) Inter- and intrafamilial variability in premature pubarche and polycystic ovary syndrome. Fertil Steril 78, 473–478.[CrossRef][ISI][Medline]

Siegelmann-Danieli N and Buetow KH (1999) Constitutional genetic variation at the human aromatase gene (CYP19) and breast cancer risk. Br J Cancer 79, 456–463.[CrossRef][ISI][Medline]

Somner J, McLellan S, Cheung J, Mak YT, Frost ML, Knapp KM, Wierzbicki AS, Wheeler M, Fogelman I, Ralston SH et al. (2004) Polymorphisms in the P450 c17 (17-hydroxylase/17,20-lyase) and P450 c19 (aromatase) genes: association with serum sex steroid concentrations and bone mineral density in postmenopausal women. J Clin Endocrinol Metab 89, 344–351.[Abstract/Free Full Text]

Stephens M, Smith NJ and Donnelly P (2001) A new statistical method for haplotype reconstruction from population data. Am J Hum Genet 68, 978–989.[CrossRef][ISI][Medline]

Takayama K, Fukaya T, Sasano H, Funayama Y, Suzuki T, Takaya R, Wada Y and Yajima A (1996) Immunohistochemical study of steroidogenesis and cell proliferation in polycystic ovarian syndrome. Hum Reprod 11, 1387–1392.[Abstract/Free Full Text]

Tayob Y, Robinson G and Adams J (1990) Ultrasound appearance of the ovaries during the pill-free interval. Br J Fam Plann 16, 94–96.

Tucci S, Futterweit W, Concepcion ES, Greenberg DA, Villanueva R, Davies TF and Tomer Y (2001) Evidence for association of polycystic ovary syndrome in caucasian women with a marker at the insulin receptor gene locus. J Clin Endocrinol Metab 86, 446–449.[Abstract/Free Full Text]

Urbanek M, Legro RS, Driscoll DA, Azziz R, Ehrmann DA, Norman RJ, Strauss JF, 3rd, Spielman RS and Dunaif A (1999) Thirty-seven candidate genes for polycystic ovary syndrome: strongest evidence for linkage is with follistatin. Proc Natl Acad Sci USA 96, 8573–8578.[Abstract/Free Full Text]

Ye S, Dhillon S, Ke X, Collins AR and Day IN (2001) An efficient procedure for genotyping single nucleotide polymorphisms. Nucleic Acids Res 29, E88.[CrossRef][Medline]

Submitted on December 22, 2004; resubmitted on January 24, 2005; accepted on March 4, 2005.