Abstract
We previously reported that BRCA1/2-mutated fallopian tube epithelium (FTE) collected during the luteal phase exhibits gene expression profiles more closely resembling that of high-grade serous carcinoma (HGSC) specimens than FTE collected during the follicular phase or from control patients. Since the luteal phase is characterised by high levels of progesterone, we determined whether the expression of progesterone receptor (PR) and PR-responsive genes was altered in FTE obtained from BRCA mutation carriers during the luteal phase of the menstrual cycle. RT-qPCR confirmed a decreased expression of PR mRNA in FTE during the luteal phase relative to follicular phase, in both BRCA1/2 mutation carriers and control patients. Immunohistochemistry using isoform-specific antibodies confirmed a low level of both PR-A and PR-B in HGSC and a lower level of staining in FTE samples obtained during the luteal phase compared with the follicular phase. No significant difference in PR-A or PR-B staining was found based on patient BRCA mutation status. Analysis of our previously reported gene expression profiles based upon known PR-A- and PR-B-specific target genes did not partition samples by BRCA mutation status, indicating that overall FTE PR response is not altered in BRCA mutation carriers. HGSC samples grouped separately from other samples, consistent with the observed loss of PR expression. These findings indicate no overall difference in PR signalling in FTE as a function of BRCA mutation status. Thus, the molecular similarity of BRCA1/2-mutated luteal phase FTE and HGSC likely results from an altered response to luteal phase factors other than progesterone.
Introduction
Epithelial ovarian cancers, which constitute the 5th leading cause of cancer-related death in North American women, are thought to arise through two basic pathways. While most histological subtypes likely arise from the ovarian surface epithelium (OSE), accumulating evidence provides strong support that high-grade serous adnexal cancer, commonly attributed to an ovarian origin, arises from the distal fallopian tube epithelium (FTE; Colgan et al. 2001, Piek et al. 2001, Finch et al. 2006, Medeiros et al. 2006, Crum et al. 2007, Kindelberger et al. 2007, Folkins et al. 2008, Shaw et al. 2009). We have previously demonstrated that non-malignant FTE collected during the luteal phase from breast cancer susceptibility gene 1/2 (BRCA1/2) mutation carriers exhibits gene expression profiles more closely resembling that of ovarian and fallopian tube high-grade serous carcinoma (HGSC) than FTE from control patients (Tone et al. 2008). This finding suggests that the FTE of BRCA mutation carriers (FTEb) responds differently to the luteal phase milieu, which may contribute to its predisposition to malignant transformation. Since elevated progesterone secretion is a hallmark of the luteal phase, the altered gene expression in these samples may reflect a differential response to progesterone.
Progesterone classically signals by binding to intracellular progesterone receptors (PR), leading to receptor dimerisation and binding to specific hormone response elements located in the promoter region of target genes to regulate transcription through interaction with co-regulatory proteins (Mulac-Jericevic & Conneely 2004, Shao et al. 2006, Gellersen et al. 2009). Progesterone can also lead to rapid activation of a number of genes through protein–protein interactions involving cytoplasmic PR with components of the Src/mitogen activated protein kinase, phosphoinositide 3-kinase/Akt and janus kinase/signal transducer and activator of transcription signalling pathways (Leonhardt et al. 2003, Boonyaratanakornkit et al. 2008, Gellersen et al. 2009). The canonical PR is expressed as two isoforms, PR-A and PR-B, which are encoded by a single gene but arise through alternative use of two promoters (Boonyaratanakornkit et al. 2008). PR-A and PR-B are co-expressed at roughly equivalent levels in most target tissues (Shao et al. 2006). PR-A is truncated at the amino-terminal domain and has been shown to exhibit differential functional properties compared with the full-length PR-B isoform, including regulation of distinct genes (Jacobsen et al. 2005) and differential effects on target tissues in knockout mice (Mulac-Jericevic & Conneely 2004, Shao et al. 2006). Interestingly, PR-A has been shown to act as a transdominant inhibitor of PR-B (Vegeto et al. 1993, Mote et al. 2002), and the relative expression of these two isoforms has been shown to vary in reproductive tissues as a result of development and hormonal status and carcinogenesis (Mulac-Jericevic & Conneely 2004). Phosphorylation of PR in the absence of ligand, through the activation of cytokine or growth factor signalling pathways such as epidermal growth factor, can activate PR-dependent gene expression. In breast epithelial cells, ligand-independent signalling is primarily through the PR-A isoform (Jacobsen et al. 2005). The genes regulated by ligand-independent signalling are largely distinct from those regulated by ligand-dependent PR activation (Jacobsen et al. 2005). Differential expression of PR isoforms in FTEb could thus underlie the altered gene expression profiles, particularly during the luteal phase when circulating progesterone levels are elevated.
The purpose of this study was to determine whether PR isoform expression is altered in FTEb relative to FTE from normal control patients (FTEn), particularly during the luteal phase. Real-time quantitative RT-PCR (RT-qPCR) and immunohistochemistry using isoform-specific antibodies were performed on patients' samples of FTE and HGSC. To our knowledge, this is the first comparative study of PR expression in non-malignant FTE in the follicular and luteal phases of the ovarian cycle in BRCA1/2 mutation carriers and control patients.
Materials and methods
Study samples
Gene expression data derived from laser capture microdissected (LCM) snap-frozen FTE and HGSC specimens as part of our original profiling study (Tone et al. 2008) were used to evaluate expression levels of PR and PR-responsive genes. Specific clinical characteristics associated with these samples were previously presented (Tone et al. 2008). Briefly, these samples included histologically normal fallopian tubes from 12 confirmed BRCA1/2 mutation carriers and 12 control patients undergoing salpingo-oophorectomy for reasons other than family history or adnexal malignancy, and 13 HGSC diagnosed as being of tubal or ovarian origin. Of the 12 BRCA patients, ten had confirmed mutations in BRCA1 and two had confirmed mutations in BRCA2. Non-malignant samples were stratified according to the phase of the menstrual cycle, with six samples collected during the luteal phase and six collected during the follicular phase in each group. RNA extracted from the same representative LCM samples used in our original study (histologically normal FTE from six BRCA1 mutation carriers and controls from both the luteal and follicular phases and six HGSC samples) was used to compare PR expression levels by RT-qPCR.
We constructed two tissue microarrays using formalin-fixed paraffin-embedded samples from the University Health Network Ovarian Tissue Bank, as described in our previous publication (Tone et al. 2008). The fallopian tube array included one 1.5 mm2 core from 11 luteal and 16 follicular phase non-malignant specimens from confirmed BRCA1/2 mutation carriers, and 11 luteal and 12 follicular phase specimens from normal control patients. This array included the FTEn and FTEb cases that were used for the original gene expression analysis, as well as additional samples from the same groups. The second array included two distinct 0.6 mm2 cores from 51 cases of HGSC naive to chemotherapy, all but two of which were independent from those subjected to gene expression profiling. Clinical features of all carcinoma specimens included in this second tissue array are shown in Supplementary Table 1, see section on supplementary data given at the end of this article. The use of all tissues in this study was approved by the University Health Network Research Ethics Board, and all patients provided informed consent.
RT-qPCR for PR mRNA
Real time RT-qPCR was performed as described previously (Tone et al. 2008) with the following modifications. Primer sequences used included PR: forward 5′-GAACAGCGGATGAA-AGAATCATC-3′ reverse 5′-AGGAACTCTTCTTGGCTAACTTGAAG-3′; ACTB: forward 5′-GCATTGTTACAGGAAGTCCCTTG-3′ and reverse 5′-CTATCACCTCCCCTGTGTG-GA-3′. The PR primer pair used recognises a specific sequence within the 1500 bp region immediately upstream of the poly(A) tail and therefore amplifies both PR-A- and PR-B-specific transcripts. All experiments included triplicate wells of each sample for both target and reference genes. The comparative CT method for relative quantitation was performed and normalised to ACTB expression. A one-way ANOVA followed by the Newman–Keuls multiple comparison test was used to compare groups. Statistical significance was determined as P<0.05.
PR isoform-specific immunohistochemistry
Immunohistochemistry was performed using standard procedures as described previously (Tone et al. 2008) with the following modifications. PR-A protein was detected using mouse monoclonal anti-PR antibody (PGR-312 (16), Novocastra Laboratories, Newcastle, UK; Mote et al. 2001) at a dilution of 1 in 200. PR-B was detected using anti-PR (B-form) antibody (PGR-B (SAN27), Novocastra Laboratories) at a dilution of 1 in 200. The ScanScope CS slide scanner (Aperio Technologies, Inc., Vista, CA, USA) was used to capture digitised images at 40× magnification. The images were then analysed core by core using the Positive Pixel Count Algorithm included with ImageScope software (Aperio Technologies, Inc., version 10.0). This algorithm counts the number of pixels within three user-defined intensity ranges (weak, positive and strong). Pixels that are stained but do not meet the positive colour specification are counted as negative-stained pixels; this is used to determine the overall percentage of positive-stained pixels. A pseudo-colour mark-up image is generated as an algorithm result to confirm the accurate measurement of pixels within each intensity range (examples shown in Supplementary Figure 1, see section on supplementary data given at the end of this article). To control for potential variability during immunohistochemistry, duplicate 5 μm sections of the fallopian tube array were included on each slide. Since the HGSC array consisted of duplicate distinct cores, the two values for each sample were averaged. Statistical analysis of mean percentage of strong staining by sample group was conducted using ANOVA followed by the Newman–Keuls multiple comparison test. The distribution of cases into quartiles was analysed using Pearson's χ2 test. Comparisons were considered statistically significant if P<0.05.
Gene expression analysis
Unsupervised hierarchical clustering was performed on our original gene expression data based upon corresponding probe sets of potential PR isoform-specific genes identified by Jacobsen et al. (2005). CEL files (Gene Expression Omnibus Series accession #GSE10971) were imported into GeneSpringGX (version 10.0, Agilent, Santa Clara, CA, USA). Normalisation was performed, first using the robust multi-array average (RMA) algorithm and then by the median measurement for each probe set across all samples. A two-way hierarchical cluster analysis with average linkage using a Pearson centered similarity distance metric was then performed.
Classification accuracy of subsets of PR isoform-specific genes was assessed using the supervised, shrunken centroid class prediction method (prediction analysis of microarrays (PAM); Tibshirani et al. 2002) found in the pamr (v1.44) package of R (v2.10.0, R Foundation for Statistical Computing, Vienna, Austria). Each gene set was tested against a predefined categorical binary variable relevant to a subset of patient expression data from Tone et al. (2008). Expression data for each set of patients under consideration were normalised in GeneSpringGX independently using the RMA algorithm as well as to the median expression value of each gene across all samples. All prediction analyses were performed with exported log-transformed values in R. In order to use all the probes under consideration in this prediction method, a fixed threshold of 0 for the shrinkage factor was used. Model building and prediction was done using a leave-one-out cross-validation, with the number of cross validation folds equal to the smallest number of samples in one of the classes under consideration.
Results
Decreased expression of PR mRNA in luteal phase FTE and HGSC specimens
Analysis of previously obtained gene expression profiles of microdissected non-malignant FTE and HGSC specimens (Tone et al. 2008) revealed a decreased expression of PR mRNA in FTE samples obtained during the luteal phase relative to follicular phase in both BRCA mutation carriers (P<0.001) and normal controls (P<0.01; Fig. 1A). No differences in PR mRNA expression were observed in samples obtained from mutation carriers versus normal controls during either phase of the ovarian cycle. A subset of four BRCA1-mutated luteal specimens, which we have previously highlighted because they grouped closely with HGSC based on global gene expression (denoted as FTEb(S); Tone et al. 2008), showed similar expression of PR compared with other samples obtained during the luteal phase. Finally, HGSC specimens exhibited decreased expression of PR, similar to that observed in the luteal phase FTE.
To confirm these findings, RT-qPCR was performed using RNA from the same representative LCM samples from each group used for RT-qPCR in our previous study. Similar to the Affymetrix profiling data, RT-qPCR revealed a decreased expression of PR mRNA in the luteal versus follicular samples in both mutation carriers (P<0.05) and controls (P<0.05). In addition, no differences in PR expression were observed between samples obtained from BRCA1/2 mutation carriers versus normal controls overall or within the same phase of the cycle (Fig. 1B). Importantly, PR mRNA was barely detectable in HGSC specimens, with expression levels >100-fold lower than those observed in FTEn follicular samples (P<0.01).
Decreased PR-A and PR-B immunostaining in luteal phase FTE and HGSC specimens
In light of previous studies showing selective loss of either PR-A or PR-B in ovarian (Akahira et al. 2000, 2002) and breast (Graham et al. 1995, Ariga et al. 2001, Mote et al. 2002, Jacobsen et al. 2005) cancers, immunohistochemistry was performed on tissue microarrays using PR-A- or PR-B-specific antibodies. Immunostaining for PR-A protein was predominately nuclear, with differences in staining observed due to stage of the ovarian cycle and not BRCA mutation status, consistent with our total PR mRNA results. A decreased percentage of FTE cells showing strong PR-A immunopositivity was observed in previously profiled luteal phase FTE relative to follicular phase samples in both normal controls and BRCA1/2 mutation carriers (P<0.001 for both). No differences in percentage of strong PR-A staining were observed in FTEn versus FTEb samples obtained within either the luteal or the follicular phase (Fig. 2A). Of particular note, expression of PR-A was similar in FTEb(S) and the remaining FTEb luteal phase cases not previously found to group with HGSC. These same differential expression patterns were also observed when PR-A immunostaining was evaluated in additional FTE cases not subjected to gene expression profiling in our previous study. In addition, PR-A protein was barely detectable in an independent set of 51 HGSC specimens (Fig. 2B).
The percentages of strong PR-A staining for all FTE cases were ranked by quartiles and compared between sample groups. A significant difference in distribution was observed between the luteal phase and follicular phase samples independent of BRCA status, with a greater proportion of luteal phase samples partitioning in the lower quartiles (Fig. 2C, P<0.001). While a slightly higher proportion of FTEb luteal versus FTEn luteal cases were within the lowest quartile, this distribution pattern was not statistically significant (P=0.372). Representative PR-A staining in follicular, luteal and HGSC specimens is shown in Fig. 2D–F respectively.
Similar results were obtained for PR-B immunohistochemistry. PR-B staining was decreased in non-malignant FTE during the luteal phase compared with that during the follicular phase in both normal controls and BRCA mutation carriers (P<0.001 for both), in samples obtained from both previously profiled (Fig. 3A) and additional cases (Fig. 3B). No differences in PR-B staining were observed in FTEn versus FTEb overall or within the same stage of the ovarian cycle, and PR-B was barely detectable in HGSC samples. A significant difference in quartile distribution of percentages of strong PR-B staining was observed between the luteal and follicular phase samples independent of BRCA status, with a greater proportion of luteal phase samples partitioning in the lower quartiles (Fig. 3C, P<0.001). The distribution of cases was strikingly similar in normal controls compared with mutation carriers. Representative staining for follicular, luteal and HGSC specimens is shown in Fig. 3D–F. Altogether, these data indicate that both PR-A and PR-B protein levels vary according to the stage of the ovarian cycle, and not BRCA mutation status. Similar results were obtained when images were analysed using an alternative image analysis program (Visiomorph software, Visiopharm, Hoersholm, Denmark; data not shown).
The analyses presented in this study largely reflect nuclear PR levels. Cytoplasmic staining was observed for PR-A and PR-B; however, the staining was less intense and present in fewer cells than nuclear staining. A difference due to the cycle stage was observed in cytoplasmic PR-A staining, with an absence of staining in the luteal phase samples regardless of BRCA mutation status. In contrast, cytoplasmic PR-B staining did not differ with either cycle stage or mutation status (Supplementary Figure 2, see section on supplementary data given at the end of this article).
Because PR-A has been found to affect signalling by PR-B, the relative levels of these two isoforms could be important to the resultant progesterone signalling. The ratio of the total percentage of cells positive for PR-A versus PR-B (PR-A:PR-B ratio) in the same selected area was calculated for each case and compared between FTE groups. No statistically significant differences in average PR-A:PR-B ratio were observed in the luteal phase compared with follicular phase samples, or normal controls versus BRCA mutation carriers (Fig. 4A). Interestingly, individual PR-A:PR-B values were tightly grouped in the follicular phase samples, whereas a greater range of ratios was observed in the luteal phase samples. This could potentially reflect different times of the luteal phase when samples were collected, as progesterone secretion is variable during this phase, reaching peak levels at 5 days post-ovulation. In contrast, very little progesterone is secreted throughout the duration of the follicular phase. Ranking of PR-A:PR-B ratio data by quartiles revealed an overall difference in distribution among the four FTE groups (P=0.006). A greater proportion of luteal phase compared with follicular phase samples partitioned in the lowest quartile (P=0.003; Fig. 4B). In contrast, there was no statistically significant difference in the distribution pattern due to BRCA mutation status overall (P=0.129) or in BRCA mutation carriers versus normal controls during the luteal phase (P=0.09). These data suggest a subtle shift in the PR-A:PR-B staining ratio during the luteal phase, potentially reflective of a slightly lower level of PR-A relative to PR-B during this phase.
Gene expression analysis of profiled FTE based on proposed PR-A and PR-B target genes
To determine whether the expression of PR-dependent genes separates FTE samples into meaningful subgroups, unsupervised clustering of all individual samples was performed using genes previously shown to be altered downstream of PR activation by Jacobsen et al. (2005). Genes with altered expression specifically due to PR-A or PR-B activation, both dependent and independent of ligand, were identified in oestrogen receptor (ER)-positive, PR-negative T47D breast cancer cell lines engineered to express PR-A or PR-B under the control of an inducible promoter (Jacobsen et al. 2005). Specifically, 37 distinct genes were found to be altered in PR-A+ (relative to PR null) T47D cells after 6 h, and 54 were found to be specifically altered in PR-B-expressing cells at the same time point (see Supplementary Tables 2 and 3, see section on supplementary data given at the end of this article for PR-A- and PR-B-dependent genes extracted from Jacobsen et al. that are included in the current analysis). Probe sets representing each of these genes were identified in the Affymetrix 2.0 Plus arrays (a total of 94 for PR-A-dependent genes and 118 for PR-B-dependent genes), and each list of probe sets was used separately to cluster our original profiling data.
In this study, unsupervised hierarchical clustering of FTE and HGSC based on probe sets corresponding to all (both ligand-dependent and -independent) PR-A-dependent genes resulted in two main cluster groups (I+II), each of which further divided into two subgroups (a+b; Fig. 5). A clear separation of non-malignant samples based on BRCA1/2 mutation status or cycle stage was not observed based on the expression of these genes. While group Ia consisted of mostly follicular samples and one luteal sample with the highest PR-A staining, group Ib consisted of a mix of luteal and follicular phase samples from both mutation carriers and normal controls. Notably, FTEb(S) samples all partitioned together (within group IIa) separate from the remaining FTEb luteal samples (group I). However, the presence of FTEn luteal samples in both of these cluster groups indicates that BRCA mutation status does not affect the expression of PR-A-responsive genes. Importantly, 12 of the 13 HGSC specimens partitioned separately from non-malignant FTE; since the HGSCs all express low levels of PR, this pattern is consistent with the PR dependence of this cassette of genes.
Similar results were obtained when unsupervised cluster analysis was performed using probe sets corresponding to genes specifically altered downstream of PR-B activation (Fig. 6). There were two main cluster groups, with no clear separation of non-malignant samples by BRCA mutation status or cycle stage. Group Ia consisted of a mix of FTEn and FTEb samples from both the follicular and luteal phases. FTEb(S) samples partitioned together within group Ib separate from the majority of HGSC in group II.
To extend our analysis of PR-dependent gene expression in FTE and HGSC, we tested the ability of these probe sets to predict a given sample category using a shrunken centroid classification methodology (Tibshirani et al. 2002; results shown in Table 1). A threshold of 0 for the shrinkage factor was used in order to include all probe sets in the prediction, and all results are based on a leave-one-out cross-validation analysis. Similar to the cluster analysis presented above, PR-A- and PR-B dependent genes were evaluated separately, and PR-A target genes were further subdivided into those found to be altered dependent (‘PR-A ligand-dependent’) and independent of the presence of ligand (‘PR-A ligand-independent’) in T47D cells. We chose to include the latter set of genes in our analysis (despite being altered in the absence of progesterone), since the ligand dependency of a particular gene may be influenced by cell or tissue type. As exposure to progesterone leads to downregulation of both PR isoforms (Lange et al. 2000), gene expression could also be altered by decreased levels of PR-A irrespective of the absence or presence of ligand. Furthermore, if ligand-independent genes were found to be predictive of BRCA mutation status, this could suggest a potential role for crosstalk between PR and cytokine signalling pathways that may be locally elevated following ovulation (Denner et al. 1990, Pierson-Mullany & Lange 2004). In contrast to PR-A, only two PR-B-specific genes were reported to be altered independent of progesterone in T47D cells (Jacobsen et al. 2005), so a separate analysis was not performed for PR-B. Consistent with unsupervised clustering, PR-dependent genes were predictive of whether a particular sample was non-malignant FTE or HGSC (as indicated by a high sample prediction accuracy for each list using dataset A, total n=37 samples). In contrast, none of the PR-dependent subsets (particularly probe sets corresponding to PR-A ligand-independent genes) were successful at predicting whether a particular non-malignant sample was obtained from a BRCA1/2 mutation carrier or control patient (as indicated by a low sample prediction accuracy for dataset B, total n=24 samples), indicating that the overall PR response is not altered due to BRCA mutation status. As expected, all lists were predictive of the stage of the ovarian cycle of a given FTE sample (dataset C, total n=24 samples), providing further support of the PR dependence of the genes identified by Jacobsen et al. (2005). Finally, certain subsets of PR-dependent genes (such as those activated by PR-A overall and PR-B) were relatively successful at predicting whether a particular luteal phase sample belonged to the FTEb(S) subgroup (dataset D, total n=12 samples).
Accuracy for predicting sample categories based on sets of PR-dependent genes using the shrunken centroid classification methodology of Tibshirani et al. with a threshold of 0 (all probes used for prediction)
Dataset | Predictiona | PR-A predictionb accuracy | PR-A LD predictionb accuracy | PR-A LI predictionb accuracy | PR-B predictionb accuracy |
---|---|---|---|---|---|
cA. Non-malignant FTE | Non-malignant | 24/24 (100%) | 24/24 (100%) | 24/24 (100%) | 24/24 (100%) |
versus HGSC | Cancer | 12/13 (92.3%) | 11/13 (84.6%) | 12/13 (92.3%) | 12/13 (92.3%) |
dB. FTEn | FTEn | 6/12 (50%) | 8/12 (66.7%) | 3/12 (25%) | 4/12 (33.3%) |
versus FTEb | FTEb | 7/12 (58.3%) | 8/12 (66.7%) | 4/12 (33.3%) | 5/12 (41.7%) |
dC. Luteal | Luteal | 9/12 (75%) | 9/12 (75%) | 9/12 (75%) | 10/12 (83.3%) |
versus follicular | Follicular | 11/12 (91.7%) | 11/12 (91.7%) | 10/12 (83.3%) | 11/12 (91.7%) |
eD. FTEb(S) | FTEb(S) | 3/4 (75%) | 3/4 (75%) | 2/4 (50%) | 3/4 (75%) |
versus other luteal | Other luteal | 7/8 (87.5%) | 6/8 (75%) | 7/8 (87.5%) | 8/8 (100%) |
LD, ligand dependent, LI, ligand independent.
Indicates the sample type attempting to be predicted based on the expression of a given set of PR-dependent genes.
Indicates the proportion of samples in a given category that are predicted correctly based on the expression of a given set of PR-dependent genes. A high prediction accuracy suggests that a given set of PR-dependent genes is successful at differentiating between the two sample types included in the analysis.
Expression data from all non-malignant FTE (n=24) and HGSC (n=13) samples included in prediction analysis.
Expression data from all non-malignant FTE (n=24) samples included in prediction analysis following independent normalisation.
Expression data from all non-malignant FTE obtained during the luteal phase (n=12) included in prediction analysis following independent normalisation.
Discussion
The data presented in this study demonstrate that PR expression in the FTE is altered as a function of the ovarian cycle. PR mRNA and protein were decreased during the luteal phase, and this represented a decreased expression of both PR-A and PR-B isoforms in both previously profiled and independent FTE samples. A major question addressed in this study was whether PR levels are affected by BRCA mutation status, and whether this could be PR isoform specific. Our results indicate that BRCA mutation status does not overtly impact PR-A and PR-B expression in the FTE. Unsupervised clustering and sample prediction analyses using genes reported as PR-A- and PR-B-specific targets further showed that BRCA-mutated samples did not exhibit an altered PR response.
The finding that PR mRNA is lowest during the luteal phase (in both carriers and controls) is consistent with previous studies of normal human FTE throughout the ovarian cycle, which have shown little or no PR expression during this phase as determined by semi-quantitative RT-PCR (Briton-Jones et al. 2005). Similar results have also been obtained for total PR protein, and the sharp decline in PR was found to coincide with the luteal elevation of serum progesterone concentration in these studies (Verhage et al. 1980, Pino et al. 1984, Helm et al. 1987). This is consistent with studies demonstrating the polyubiquitination and subsequent degradation of PR protein upon exposure to progesterone (Lange et al. 2000). Thus, the decline in PR expression in the luteal phase samples observed in this study likely represents downregulation of PR.
Our results further demonstrate a strong downregulation of PR expression in HGSC. This is consistent with several studies that have reported decreased expression of PR in ovarian and fallopian tube cancers. Lau et al. (1999) have demonstrated a marked reduction in PR mRNA in ovarian cancer cell lines compared with primary cultures of normal OSE using semi-quantitative RT-PCR, and others have found low PR protein expression in both BRCA1-associated and sporadic serous ovarian cancers by immunohistochemistry (Aghmesheh et al. 2005). However, recent studies have strongly suggested that the FTE is the source of HGSC (Colgan et al. 2001, Piek et al. 2001, Finch et al. 2006, Medeiros et al. 2006, Crum et al. 2007, Kindelberger et al. 2007, Folkins et al. 2008, Shaw et al. 2009), thus the OSE may not be the appropriate comparator. A single previous study has reported slightly decreased PR protein expression in fallopian tube carcinomas relative to areas of benign FTE in patients with or without a BRCA mutation (Cass et al. 2005). Our study extends these observations to both tubal and ovarian HGSC samples and to FTE samples obtained from women without adnexal cancer. Our studies further demonstrate a strong downregulation of PR expression in HGSC samples, such that PR expression is nearly absent in malignant cells. Interestingly, previous studies have demonstrated loss of heterozygosity (LOH) at the PR locus (11q22) in up to 50% of analysed ovarian carcinomas of varying histotypes (Gabra et al. 1995, Launonen et al. 1998), providing a potential explanation for PR loss in HGSC specimens. In one such study, 6/15 (40%) of informative serous carcinoma specimens exhibited LOH; furthermore, LOH at 11q22 was significantly associated with low tumour PR content in the full cohort of 38 ovarian carcinomas studied (Gabra et al. 1995). Whether LOH occurs in HGSC precursor lesions in the FTE remains to be determined.
Expression of PR has been reported to be a favourable prognostic indicator in ovarian cancer (Lee et al. 2005), but this finding may be misleading due to the inclusion of multiple histotypes in the analysis. Close examination of the reported data indicates that 64% of endometrioid carcinomas were considered positive for PR (>10% of cells stained), compared with only 25% of serous carcinomas of unspecified grade. Patients with endometrioid carcinomas typically present at an earlier stage and have a greatly enhanced survival (Gien et al. 2008), which could underlie the apparent impact of PR expression on prognosis in this study. The tissue microarray used in this study to assess PR expression in HGSC also contained other major histotypes of epithelial ovarian cancer (see Supplementary Table 1, see section on supplementary data given at the end of this article for the clinical characteristics and immunostaining results of individual carcinoma cases). Consistent with the data presented in the previous study, 11/14 (79%) and 12/12 (100%) of evaluable endometrioid cases, compared with 10/51 (20%) and 13/51 (25%) of HGSC cases, showed >10% positivity for PR-A and PR-B respectively (Supplementary Figure 3, see section on supplementary data given at the end of this article). Also consistent was our finding that none of the clear cell carcinomas and only one of six mucinous carcinomas examined showed positive PR-A or PR-B staining, suggesting a potential role for PR loss in these histotypes.
The relative levels of PR-A and PR-B influence the response of target cells to progesterone. For instance, breast cancer cells with predominant expression of PR-A have shown an exaggerated proliferative response to luteal phase levels of progesterone compared with cells with predominant expression of PR-B (Leo & Lin 2008). While PR-A and PR-B are expressed at roughly equivalent levels in normal human breast epithelium throughout the cycle, several studies have found a predominance of PR-A in a high proportion of invasive breast tumours (Graham et al. 1995, Ariga et al. 2001, Mote et al. 2002) and in ductal carcinomas in situ (Mote et al. 2002). A specific lack of expression of the PR-B isoform has also been observed in normal breast epithelium obtained from BRCA mutation carriers compared with control patients, resulting in predominant expression of PR-A in 40% of cases (Mote et al. 2004).
In this study, we found a decreased expression of both PR-A and PR-B in HGSC relative to its probable cell type of origin. Because expression of both isoforms in HGSC was barely detectable by immunohistochemistry, consistent with our RT-qPCR data, ratios of PR-A to PR-B were not calculated. Akahira et al. (2000) have demonstrated a lower expression of PR-A compared with PR-B for all epithelial ovarian cancer histotypes using both RT-PCR and immunohistochemistry; relative isoform levels did not vary with grade or stage of tumour. In a follow-up study, this group reported a progressive downregulation of PR-A, but not of PR-B, from normal OSE through benign, borderline and malignant serous ovarian tumours (Akahira et al. 2002). In normal OSE and serous adenomas, both isoforms were equally expressed, whereas PR-B predominated in borderline and malignant tumours. However, the grade of the serous carcinomas was not indicated, and while low-grade serous carcinomas are thought to be derived from borderline tumours (Singer et al. 2002, Shih & Kurman 2004, May et al. 2010), this is not thought to be the case for HGSC. In light of recent studies outlining the progression to HGSC from normal FTE (Crum et al. 2007, Lee et al. 2007, Folkins et al. 2008), studies of relative PR isoform expression in FTE exhibiting focal accumulation of p53 (p53 signatures), tubal intraepithelial carcinomas and their proposed intermediaries should be performed to determine whether differential isoform expression may play a role at any stage during HGSC development.
A study comparing steroid receptor expression in familial and sporadic serous carcinomas reported no statistical difference in either PR-A or PR-B expression (Aghmesheh et al. 2005). Initial studies investigating the effect of BRCA mutation status on PR expression in non-malignant cells did not distinguish by isoform, but rather focused on determining total PR protein levels. In these studies, no differential expression was observed in non-malignant FTE and OSE from mutation carriers and controls (Piek et al. 2001, 2003). PR protein was expressed abundantly in both ciliated and secretory FTE cells, and expression in morphologically normal and dysplastic areas from mutation carriers was similar (Piek et al. 2001). This is consistent with our findings of no differences in either PR-A or PR-B expression as a function of BRCA mutation status. In studies of BRCA-associated OSE, the only difference observed was an increased expression of PR protein in inclusion cyst epithelium compared with OSE, which the authors attributed to an increased exposure to ovarian stroma-derived hormones including oestrogen (Piek et al. 2003). Interestingly, PR expression was higher in inclusion cyst epithelium from mutation carriers compared with controls, potentially consistent with the repression of ER transactivation activity by wild-type BRCA1 (Zheng et al. 2001, Fan et al. 2002). BRCA1 has also been reported to function as a PR transcriptional repressor (Ma et al. 2006); thus we had expected to see a separation of FTEb and FTEn cases based on established PR-responsive genes. However, the response to progesterone was not affected by mutation status, as indicated by a lack of separation of FTEn and FTEb samples by unsupervised clustering based on PR-A- or PR-B-dependent genes. Furthermore, a low proportion of FTEn and FTEb samples were placed in the correct category when attempting to predict whether a particular FTE sample was obtained from a BRCA1/2 mutation carrier or a normal control using the same subsets of PR-dependent genes. If altered PR signalling is in fact an important contributor to the initiation of HGSC, we would expect to observe a clear separation of FTEn and FTEb samples by unsupervised clustering and a higher accuracy rate in sample prediction analysis given the greatly enhanced risk of HGSC in BRCA1/2 mutation carriers. In contrast to FTEn versus FTEb, PR-dependent genes could successfully differentiate malignant versus non-malignant samples, utilising both unsupervised clustering and sample prediction analysis. This is not unexpected, as HGSC samples showed a universal loss of PR expression compared with non-malignant FTE. It is important to note that the gene expression analysis presented in this study is based on target genes identified in breast cancer cells and does not reflect genes affected in a tissue-specific manner. A more definitive analysis therefore awaits the identification of PR target genes more specific to FTE cells.
Several studies suggest that HGSC arises from the secretory rather than the ciliated FTE (Talamo et al. 1982, Crum et al. 2007, Lee et al. 2007). One possibility, therefore, for the partitioning of luteal phase samples with HGSC may be an enrichment of secretory cells in the FTE during the luteal phase. However, studies indicate that the secretory cell type is not overrepresented during this phase. The relative proportions of secretory to ciliated cells in different segments of the human fallopian tube throughout the ovarian cycle were previously studied by Crow et al. (1994). They reported an increase in the proportion of ciliated cells (hence a decrease in the proportion of secretory cells) along the length of the tube, with the highest proportion of ciliated cells observed in the fimbriae (the location of the overwhelming majority of HGSC precursor lesions). Importantly, they reported only very minor differences in the percentages of secretory versus ciliated cells in the luteal versus follicular phase. For example, only 4% more secretory cells were observed in the fimbriated end in the luteal phase compared with the follicular phase. This slight difference would have a negligible impact on global gene expression. We found that luteal samples, in general, grouped separately from follicular phase samples by unsupervised clustering, which is consistent with the increased presence of progesterone during the luteal phase. Our finding of a high accuracy rate with PR ligand-independent gene expression in differentiating the luteal and follicular phase samples could reflect the changes in PR expression as well as indicate that signalling pathways resulting in PR phosphorylation differ with the ovarian cycle.
Our previous work provides evidence suggesting that a differential response by FTE from BRCA mutation carriers to the luteal phase milieu contributes to an increased propensity for malignant transformation. FTE from BRCA mutation carriers collected during the luteal phase molecularly resembled HGSC, rather than FTE collected during the follicular phase or from normal control patients. This was particularly apparent for four FTEb luteal samples, referred to as FTEb(S), which grouped with HGSC at the global gene expression level (Tone et al. 2008). In this study, we investigated the potential for altered PR expression and signalling to contribute to this differential gene expression pattern. Our results indicate that there is no overt difference in PR expression as a function of BRCA mutation status. Unsupervised cluster analysis further indicated that a differential response to progesterone based on BRCA mutation status also does not likely underlie the molecular similarity of FTEb(S) and HGSC samples, as FTEb(S) samples were found to group separately from the majority of HGSC according to the expression of PR-dependent genes. As we learn more about the specific role of PR-A and PR-B in the fallopian tube, a refined set of genes will likely emerge. It therefore remains possible that a differential clustering pattern could result with a tissue-specific gene list. However, based on the information that is presently available, our data indicate that differential response to factors associated with the luteal phase other than progesterone plays an important role in determining HGSC risk.
Supplementary data
This is linked to the online version of the paper at http://dx.doi.org/10.1530/ERC-10-0235.
Declaration of interest
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.
Funding
This work was supported by the Toronto Ovarian Cancer Research Network through funds raised by the Toronto Fashion Show, an Ontario Women's Health Council/Canadian Institute of Health Research (CIHR) Institute of Gender and Health Doctoral Research Award (to A A Tone), a Department of Defense Ovarian Cancer Research Program Idea Award (to P A Shaw) and an operating grant from the CIHR (MOP74726 to T J Brown).
Acknowledgements
We would like to thank the University Health Network Biobank for facilitating this work and Kelvin So for his assistance with immunohistochemical staining.
References
Aghmesheh M, Edwards L, Clarke CL, Byth K, Katzenellenbogen BS, Russell PJ, Friedlander M, Tucker KM & de Fazio A 2005 Expression of steroid hormone receptors in BRCA1-associated ovarian carcinomas. Gynecologic Oncology 97 16–25 doi:10.1016/j.ygyno.2004.12.030.
Akahira J, Inoue T, Suzuki T, Ito K, Konno R, Sato S, Moriya T, Okamura K, Yajima A & Sasano H 2000 Progesterone receptor isoforms A and B in human epithelial ovarian carcinoma: immunohistochemical and RT-PCR studies. British Journal of Cancer 83 1488–1494 doi:10.1054/bjoc.2000.1463.
Akahira J, Suzuki T, Ito K, Kaneko C, Darnel AD, Moriya T, Okamura K, Yaegashi N & Sasano H 2002 Differential expression of progesterone receptor isoforms A and B in the normal ovary, and in benign, borderline, and malignant ovarian tumors. Japanese Journal of Cancer Research 93 807–815 doi:10.1111/j.1349-7006.2002.tb01323.x.
Ariga N, Suzuki T, Moriya T, Kimura M, Inoue T, Ohuchi N & Sasano H 2001 Progesterone receptor A and B isoforms in the human breast and its disorders. Japanese Journal of Cancer Research 92 302–308 doi:10.1111/j.1349-7006.2001.tb01095.x.
Boonyaratanakornkit V, Bi Y, Rudd M & Edwards DP 2008 The role and mechanism of progesterone receptor activation of extra-nuclear signaling pathways in regulating gene transcription and cell cycle progression. Steroids 73 922–928 doi:10.1016/j.steroids.2008.01.010.
Briton-Jones C, Lok IH, Cheung CK, Po AL, Chiu TT & Haines C 2005 Ratio of mRNA expression of progesterone receptor isoforms AB is to B in human oviduct mucosal cells during the ovulatory cycle. Journal of Assisted Reproduction and Genetics 22 429–435 doi:10.1007/s10815-005-7203-5.
Cass I, Holschneider C, Datta N, Barbuto D, Walts AE & Karlan BY 2005 BRCA-mutation-associated fallopian tube carcinoma: a distinct clinical phenotype? Obstetrics and Gynecology 106 1327–1334 doi:10.1097/01.AOG.0000187892.78392.3f.
Colgan TJ, Murphy J, Cole DE, Narod S & Rosen B 2001 Occult carcinoma in prophylactic oophorectomy specimens: prevalence and association with BRCA germline mutation status. American Journal of Surgical Pathology 25 1283–1289 doi:10.1097/00000478-200110000-00009.
Crow J, Amso NN, Lewin J & Shaw RW 1994 Morphology and ultrastructure of fallopian tube epithelium at different stages of the menstrual cycle and menopause. Human Reproduction 9 2224–2233.
Crum CP, Drapkin R, Miron A, Ince TA, Muto M, Kindelberger DW & Lee Y 2007 The distal fallopian tube: a new model for pelvic serous carcinogenesis. Current Opinion in Obstetrics & Gynecology 19 3–9 doi:10.1097/GCO.0b013e328011a21f.
Denner LA, Weigel NL, Maxwell BL, Schrader WT & O'Malley BW 1990 Regulation of progesterone receptor-mediated transcription by phosphorylation. Science 250 1740–1743 doi:10.1126/science.2176746.
Fan S, Ma YX, Wang C, Yuan RQ, Meng Q, Wang JA, Erdos M, Goldberg ID, Webb P & Kushner PJ et al. 2002 p300 modulates the BRCA1 inhibition of estrogen receptor activity. Cancer Research 62 141–151.
Finch A, Shaw P, Rosen B, Murphy J, Narod SA & Colgan TJ 2006 Clinical and pathologic findings of prophylactic salpingo-oophorectomies in 159 BRCA1 and BRCA2 carriers. Gynecologic Oncology 100 58–64 doi:10.1016/j.ygyno.2005.06.065.
Folkins AK, Jarboe EA, Saleemuddin A, Lee Y, Callahan MJ, Drapkin R, Garber JE, Muto MG, Tworoger S & Crum CP 2008 A candidate precursor to pelvic serous cancer (p53 signature) and its prevalence in ovaries and fallopian tubes from women with BRCA mutations. Gynecologic Oncology 109 168–173 doi:10.1016/j.ygyno.2008.01.012.
Gabra H, Langdon SP, Watson JE, Hawkins RA, Cohen BB, Taylor L, Mackay J, Steel CM, Leonard RC & Smyth JF 1995 Loss of heterozygosity at 11q22 correlates with low progesterone receptor content in epithelial ovarian cancer. Clinical Cancer Research 1 945–953.
Gellersen B, Fernandes MS & Brosens JJ 2009 Non-genomic progesterone actions in female reproduction. Human Reproduction Update 15 119–138 doi:10.1093/humupd/dmn044.
Gien LT, Rosen B, Panzarella T, Chapman W & Dodge JE 2008 A comparison of outcomes between endometrioid and serous carcinomas of the ovary: are they different cancers? In 4th Canadian Conference on Ovarian Cancer Research Montreal, Quebec, Canada.
Graham JD, Yeates C, Balleine RL, Harvey SS, Milliken JS, Bilous AM & Clarke CL 1995 Characterization of progesterone receptor A and B expression in human breast cancer. Cancer Research 55 5063–5068.
Helm G, Batra S & Owman C 1987 Cytoplasmic and nuclear progesterone receptors in human fallopian tube and their relationship to plasma steroids during the menstrual cycle. International Journal of Fertility 32 162–166.
Jacobsen BM, Schittone SA, Richer JK & Horwitz KB 2005 Progesterone-independent effects of human progesterone receptors (PRs) in estrogen receptor-positive breast cancer: PR isoform-specific gene regulation and tumor biology. Molecular Endocrinology 19 574–587 doi:10.1210/me.2004-0287.
Kindelberger DW, Lee Y, Miron A, Hirsch MS, Feltmate C, Medeiros F, Callahan MJ, Garner EO, Gordon RW & Birch C et al. 2007 Intraepithelial carcinoma of the fimbria and pelvic serous carcinoma: evidence for a causal relationship. American Journal of Surgical Pathology 31 161–169 doi:10.1097/01.pas.0000213335.40358.47.
Lange CA, Shen T & Horwitz KB 2000 Phosphorylation of human progesterone receptors at serine-294 by mitogen-activated protein kinase signals their degradation by the 26S proteasome. PNAS 97 1032–1037 doi:10.1073/pnas.97.3.1032.
Lau KM, Mok SC & Ho SM 1999 Expression of human estrogen receptor-α and -β, progesterone receptor, and androgen receptor mRNA in normal and malignant ovarian epithelial cells. PNAS 96 5722–5727 doi:10.1073/pnas.96.10.5722.
Launonen V, Stenback F, Puistola U, Bloigu R, Huusko P, Kytola S, Kauppila A & Winqvist R 1998 Chromosome 11q22.3–q25 LOH in ovarian cancer: association with a more aggressive disease course and involved subregions. Gynecologic Oncology 71 299–304 doi:10.1006/gyno.1998.5186.
Lee P, Rosen DG, Zhu C, Silva EG & Liu J 2005 Expression of progesterone receptor is a favorable prognostic marker in ovarian cancer. Gynecologic Oncology 96 671–677 doi:10.1016/j.ygyno.2004.11.010.
Lee Y, Miron A, Drapkin R, Nucci MR, Medeiros F, Saleemuddin A, Garber J, Birch C, Mou H & Gordon RW et al. 2007 A candidate precursor to serous carcinoma that originates in the distal fallopian tube. Journal of Pathology 211 26–35 doi:10.1002/path.2091.
Leo JC & Lin VC 2008 The activities of progesterone receptor isoform A and B are differentially modulated by their ligands in a gene-selective manner. International Journal of Cancer 122 230–243 doi:10.1002/ijc.23081.
Leonhardt SA, Boonyaratanakornkit V & Edwards DP 2003 Progesterone receptor transcription and non-transcription signaling mechanisms. Steroids 68 761–770 doi:10.1016/S0039-128X(03)00129-6.
Ma Y, Katiyar P, Jones LP, Fan S, Zhang Y, Furth PA & Rosen EM 2006 The breast cancer susceptibility gene BRCA1 regulates progesterone receptor signaling in mammary epithelial cells. Molecular Endocrinology 20 14–34 doi:10.1210/me.2004-0488.
May T, Virtanen C, Sharma M, Milea A, Begley H, Rosen B, Murphy KJ, Brown TJ & Shaw PA 2010 Low malignant potential tumors with micropapillary features are molecularly similar to low-grade serous carcinoma of the ovary. Gynecologic Oncology 117 9–17 doi:10.1016/j.ygyno.2010.01.006.
Medeiros F, Muto MG, Lee Y, Elvin JA, Callahan MJ, Feltmate C, Garber JE, Cramer DW & Crum CP 2006 The tubal fimbria is a preferred site for early adenocarcinoma in women with familial ovarian cancer syndrome. American Journal of Surgical Pathology 30 230–236 doi:10.1097/01.pas.0000180854.28831.77.
Mote PA, Johnston JF, Manninen T, Tuohimaa P & Clarke CL 2001 Detection of progesterone receptor forms A and B by immunohistochemical analysis. Journal of Clinical Pathology 54 624–630 doi:10.1136/jcp.54.8.624.
Mote PA, Bartow S, Tran N & Clarke CL 2002 Loss of co-ordinate expression of progesterone receptors A and B is an early event in breast carcinogenesis. Breast Cancer Research and Treatment 72 163–172 doi:10.1023/A:1014820500738.
Mote PA, Leary JA, Avery KA, Sandelin K, Chenevix-Trench G, Kirk JA & Clarke CL 2004 Germ-line mutations in BRCA1 or BRCA2 in the normal breast are associated with altered expression of estrogen-responsive proteins and the predominance of progesterone receptor A. Genes, Chromosomes and Cancer 39 236–248 doi:10.1002/gcc.10321.
Mulac-Jericevic B & Conneely OM 2004 Reproductive tissue selective actions of progesterone receptors. Reproduction 128 139–146 doi:10.1530/rep.1.00189.
Piek JM, van Diest PJ, Zweemer RP, Jansen JW, Poort-Keesom RJ, Menko FH, Gille JJ, Jongsma AP, Pals G & Kenemans P et al. 2001 Dysplastic changes in prophylactically removed fallopian tubes of women predisposed to developing ovarian cancer. Journal of Pathology 195 451–456 doi:10.1002/path.1000.
Piek JM, Verheijen RH, Menko FH, Jongsma AP, Weegenaar J, Gille JJ, Pals G, Kenemans P & van Diest PJ 2003 Expression of differentiation and proliferation related proteins in epithelium of prophylactically removed ovaries from women with a hereditary female adnexal cancer predisposition. Histopathology 43 26–32 doi:10.1046/j.1365-2559.2003.01654.x.
Pierson-Mullany LK & Lange CA 2004 Phosphorylation of progesterone receptor serine 400 mediates ligand-independent transcriptional activity in response to activation of cyclin-dependent protein kinase 2. Molecular and Cellular Biology 24 10542–10557 doi:10.1128/MCB.24.24.10542-10557.2004.
Pino AM, Devoto L, Davila M & Soto E 1984 Changes during the menstrual cycle in cytosolic and nuclear concentrations of progestagen receptor in the human fallopian tube. Journal of Reproduction and Fertility 70 481–485 doi:10.1530/jrf.0.0700481.
Shao R, Weijdegard B, Ljungstrom K, Friberg A, Zhu C, Wang X, Zhu Y, Fernandez-Rodriguez J, Egecioglu E & Rung E et al. 2006 Nuclear progesterone receptor A and B isoforms in mouse fallopian tube and uterus: implications for expression, regulation, and cellular function. American Journal of Physiology. Endocrinology and Metabolism 291 E59–E72 doi:10.1152/ajpendo.00582.2005.
Shaw PA, Rouzbahman M, Pizer ES, Pintilie M & Begley H 2009 Candidate serous cancer precursors in fallopian tube epithelium of BRCA1/2 mutation carriers. Modern Pathology 22 1133–1138 doi:10.1038/modpathol.2009.89.
Shih IM & Kurman RJ 2004 Ovarian tumorigenesis: a proposed model based on morphological and molecular genetic analysis. American Journal of Pathology 164 1511–1518 doi:10.1016/S0002-9440(10)63708-X.
Singer G, Kurman RJ, Chang HW, Cho SK & Shih IM 2002 Diverse tumorigenic pathways in ovarian serous carcinoma. American Journal of Pathology 160 1223–1228 doi:10.1016/S0002-9440(10)62549-7.
Talamo TS, Bender BL, Ellis LD & Scioscia EA 1982 Adenocarcinoma of the fallopian tube. An ultrastructural study. Virchows Archive. A: Pathological Anatomy and Histology 397 363–368 doi:10.1007/BF00496577.
Tibshirani R, Hastie T, Narasimhan B & Chu G 2002 Diagnosis of multiple cancer types by shrunken centroids of gene expression. PNAS 99 6567–6572 doi:10.1073/pnas.082099299.
Tone AA, Begley H, Sharma M, Murphy J, Rosen B, Brown TJ & Shaw PA 2008 Gene expression profiles of luteal phase fallopian tube epithelium from BRCA mutation carriers resemble high-grade serous carcinoma. Clinical Cancer Research 14 4067–4078 doi:10.1158/1078-0432.CCR-07-4959.
Vegeto E, Shahbaz MM, Wen DX, Goldman ME, O'Malley BW & McDonnell DP 1993 Human progesterone receptor A form is a cell- and promoter-specific repressor of human progesterone receptor B function. Molecular Endocrinology 7 1244–1255 doi:10.1210/me.7.10.1244.
Verhage HG, Akbar M & Jaffe RC 1980 Cyclic changes in cytosol progesterone receptor of human fallopian tube. Journal of Clinical Endocrinology and Metabolism 51 776–780 doi:10.1210/jcem-51-4-776.
Zheng L, Annab LA, Afshari CA, Lee WH & Boyer TG 2001 BRCA1 mediates ligand-independent transcriptional repression of the estrogen receptor. PNAS 98 9587–9592 doi:10.1073/pnas.171174298.
(P A Shaw and T J Brown contributed equally to this work)