PGR +331 A/G and Increased Risk of Epithelial Ovarian Cancer

Estrogens and progesterone are the principal steroid hormones of reproduction in women, and it is established that both childbearing and use of oral contraceptives lower the risk of developing ovarian cancer. In reviewing the literature on the effects of these hormones, we suggested that the protective effects of parity and oral contraceptive use were at least in part due to increased progesterone or progestin exposures (1). Since that time, additional studies have supported this hypothesis (2). The effects of progesterone are mediated through the progesterone receptor, which exists in two morphologically and functionally distinct protein isoforms, hPR-A and hPR-B. Mice with loss of PR-A and exclusive production of PR-B have increased uterine epithelial cell proliferation in response to estrogen alone and to estrogen plus progesterone (3). The two progesterone receptor isoforms are produced from the same gene by separate promoters and by initiation of translation at distinct signals (3). Biochemical assays of the progesterone receptor gene (PGR) show that the A allele of the promoter region polymorphic variant, +331 A/G, increases transcription of the hPR-B receptor isoform compared with GG carriers (4). We therefore examined the association between PGR +331 A/G genotype and risk of ovarian cancer in a case-control study in Connecticut.

A population-based case-control study was carried out for the study of ovarian cancer. This study was approved by the Connecticut Department of Public Health, the Yale Human Investigation Committee, and all 31 other Connecticut hospital institutional review boards. Potentially eligible case subjects were English-speaking women resident in Connecticut, diagnosed at 35 to 79 years of age, between September 1, 1998 and February 28, 2003, with new primary borderline or invasive epithelial ovarian tumors. Case subjects were identified by use of a rapid reporting system, in which staff visited pathology departments and hospital tumor registrars of all 32 Connecticut hospitals at 2- to 4-week intervals. Records were also obtained for Connecticut residents receiving care in major cancer referral centers in adjacent states, and the Connecticut Tumor Registry was queried to identify any missed cases from the hospitals information. Pathology reports were obtained for all potentially eligible cases and reviewed by us for eligibility and histologic classification of tumor type. For cases with diagnoses of borderline or mucinous tumors, we procured tumor slides and systematically reviewed them according to standardized criteria. Following physician consent, letters of introduction were sent to cases briefly explaining the study, the voluntary nature of participation, and noting which doctor gave approval for contact. This was followed up by telephone calls from the interviewer to schedule the in-person interview. At interview, after providing informed consent, subjects were asked the risk factor questionnaire, and then four buccal mucosa brushings (one upper and one lower on each side of the mouth) were taken.

A representative sample of the general population of the study area was used for control subjects. Controls under age 65 years were identified by list-based random digit dialing methods. To improve participation, sampled telephone numbers found to have addresses in reverse telephone directories were mailed study introduction letters before initial random digit dialing contact. Controls 65 years of age and over initially were randomly chosen from rosters of the residents of Connecticut obtained from the Health Care Financing Administration and later were identified by the same random digit dialing methods as were used for the younger controls. All case subjects had telephone numbers found within 1+ residential blocks, which validated the sampling frame of the random digit dialing controls. Control subjects were ascertained over an eligibility period from April 15, 1999 to September 17, 2003 in three age strata (35-49, 50-64, and 65-79 years) to match the age distribution of cases. Potential control subjects received letters of introduction briefly explaining the study and the voluntary nature of participation. This was followed up by telephone calls from the interviewer to determine eligibility and, if eligible, to schedule the in-person interview. Any potential control who had had cancer of the ovary or who had had bilateral oophorectomy (or oophorectomy but uncertain whether it was total) was excluded from the study. Interview and buccal swab procedures identical to those for the cases were used for the controls.

In total, 720 eligible cases were identified and 497 (69%) enrolled in the study. Reasons for nonparticipation included death before study contact (12%), subject refusal (15%) or too ill (1%), or lost to follow-up (3%). Of the 497 cases, 490 (99%) provided buccal cell samples. Nine-hundred two eligible controls were interviewed and 551 (61%) were enrolled. Thirty-three percent of eligible controls refused participation, 1% were too ill, and 5% were lost to follow-up. Of the 551 interviewed, 534 (97%) gave buccal cell samples.

After interview, buccal cell samples were briefly kept, refrigerated, and either extracted shortly thereafter or stored long-term at −80°C. Fresh and thawed samples had DNA extracted by standard phenol/chloroform methods. Genotyping was accomplished by PCR and dot blot methods. The forward and reverse primers used for the PGR +331 A/G polymorphism were 5′GCACTACTGGGATCTGAGATC3′ and 5′ACGCAGAGTACTCACAAGTCC3′, respectively. Each 25 μL PCR was cycled at 97°C for 2 minutes, 40 cycles of 96°C for 30 seconds, 55°C for 30 seconds, 72°C for 40 seconds, and then finally, 72°C for 5 minutes. The allele-specific oligonucleotide probes for the +331 A/G variant were allele 1 (A), 5′-AGATAAAAGAGCCGC-3′, and allele 2 (G), 5′-AGATAAAGGAGCCGC-3′. Autoradiographs of the same blot probed for different alleles were generated to distinguish homozygous from heterozygous alleles. Subjects with no spot intensities for either allele were rerun. Three cases and one control still did not yield results and were omitted from statistical analyses. For quality control purposes, a 5% random sample of all subjects was rerun and found blindly to be completely concordant with the original results. Spots were read blindly with respect to case-control status, and both case and control samples were included in each 96-well plate.

Analysis of allele status was carried out using unconditional multivariate logistic regression methods. The GLIM computer program (5) was used for calculations. Regression models included adjustment for the three categories of the age matching, continuous terms for age within each of the three age categories, years of education, race (African-American versus other), menopausal status, number of full-term pregnancies, years of oral contraceptive use, average months of lactation per pregnancy, history of tubal ligation, and history of hysterectomy (at least 5 years in the past). Two cases and no controls were homozygous AA carriers; thus this genotype was considered together with heterozygotes. The three genotype groups were consistent with Hardy-Weinberg equilibrium among controls (P = 0.32). All statistical tests were two sided and considered to reach statistical significance at the 0.05 level.

Characteristics of the study subjects are shown in Table 1. Within each of the categories of age, the age matching was very good. As has been seen in virtually all studies of ovarian cancer, cases tended to have fewer full-term pregnancies and to use oral contraceptives less frequently than controls.

Table 2 gives the numbers of subjects and risk of ovarian cancer according to genotype for all women and separately for premenopausal and postmenopausal women. The increased risk seen with carriage of the A allele among all women is due to an effect only in postmenopausal women. Most of the participants in the present study were White, for whom carriage of the A allele had an odds ratio (OR) of 1.59 [95% confidence interval (95% CI), 1.02-2.48] among all and 2.05 (95% CI, 1.16-3.63) among the postmenopausal. An increased risk was also observed for African-American women, but numbers of subjects were too small (3 case, 1 control carrier) to provide adequate OR estimates. Among controls, there were no associations between allele carriage and any of the reproductive factors in Table 1 (data not shown).

Finally, Table 3 shows the risks of ovarian cancer associated with carriage of the A allele according to tumor histologic type. Again, no associations were seen among premenopausal women, but for postmenopausal women, increased risk was found for each of the histologic groups. Women with mucinous ovarian tumors had the highest risks of all of the histologic types. The same high increased risks in postmenopausal women were found when the mucinous tumors were subclassified as endocervical/Mullerian (OR, 4.67) and intestinal/enteric (OR, 4.67).

Germ-line genetic variation in PGR has been under study for about a decade. Initial reports suggested that the PROGINS allele was associated with risk of ovarian cancer; however, subsequent studies have not been confirmatory (reviewed in ref. 6). The functional significance of the PROGINS variant is not yet well defined.

In the present study, we observed statistically significantly increased risk of ovarian cancer for carriage of the PGR +331 A/G A allele among all women and especially among postmenopausal women and a similar picture within all tumor histologic groups. A virtually identical pattern of increased risk with the A allele among all women and particularly postmenopausal ones has been observed for breast cancer in a nested case-control sample from the Nurses' Health Study (7). The breast cancer analysis involved 990 cases and 1,364 matched controls and showed statistically significant ORs of 1.33 to 1.41 for carriage of the A allele. A second nested sample of 197 cases of endometrial cancer and 397 matched controls from the Nurses' Health Study also showed increased risk of that disease (OR, 1.90; 95% CI, 1.10-3.29) associated with carriage of the A allele (4). On the other hand, an ovarian cancer case-control study that looked at the +331 A/G variant did not show associations with risk (8). That study involved Caucasian subjects from both North Carolina and Australia and found ORs below unity for the A allele for all histologic types. Given the lack of a priori reason for focusing on one specific type and thus the multiple comparisons issue, none of the associations reached statistical significance. A second ovarian cancer study, in eastern Massachusetts and New Hampshire, also overall did not show increased risk for the +331 A allele, although some risk elevation was observed for women with invasive serous or undifferentiated tumors (OR, 1.3; ref. 9). Nevertheless, the two studies do not really confirm the present results.

It has been suggested that the enhancement of PR-B transcription by PGR +331 A increases epithelial cell proliferation in response to estrogen and estrogen + progesterone stimulation (3) and that such constitutive up-regulation could thus increase cancer risk of sex hormone-dependent organs. A second functional effect of the +331 A/G variant has also been identified. Among premenopausal women in Sweden, carriers of the A allele were found to have on average >60% higher serum prolactin levels than GG homozygotes (10). Progesterone receptors are absent from the pituitary; thus, the influence of progesterone on prolactin release occurs in the brain, through progesterone receptors, involving interactions with endorphins, serotonin, or dopamine (10). The effect of prolactin on increasing risk of ovarian cancer development is unknown, although it has been shown that elevated serum prolactin levels strongly predict subjects with ovarian cancer compared with controls (11). Prolactin also markedly inhibits apoptosis of ovarian carcinoma cells (12). Thus, possibly, the overall effect of the PGR +331 A allele is actually not to modulate steroid hormone-related induction of new cancers but rather to inhibit apoptosis of existing but undetected tumors. This is consistent with the rather general increased risk by carriage of the A allele seen for three cancer sites and for all ovarian cancer histologic groups and with the lack of association with the known ovarian cancer reproductive risk factors among controls.

This case-control study was large and had sufficient power to detect ORs of the magnitude seen. Examination of germ-line PGR variants was a specific aim of the study, and the observed associations were in the directions expected to be found. With response fractions of 69% for cases and 61% for controls, it is possible that our subject sample could differ somewhat from a completely representative one. However, our case-control differences in parity and oral contraceptive use are identical to those in the literature and suggest that our subject sample reflects the population base. In addition, inclusion of all of the adjustment variables made little change to the magnitudes of the ORs, suggesting that little confounding was present and thus that, even if our sample were not truly representative on these factors, the results would still be close to the true associations.

In summary, we found that carriage of the A allele of the PGR +331 A/G polymorphism was associated with increased risk of ovarian cancer, particularly among postmenopausal women. In light of the two other ovarian cancer studies that did not observe this association but two of breast and endometrial cancer that did, further examination of this association may be warranted.

We thank the 32 Connecticut hospitals, including the Stamford Health System, for participation in this study.