Active FOXO1 Is a Key Determinant of Isoform-Specific Progesterone Receptor Transactivation and Senescence Programming

This article is featured in Highlights of This Issue, p. 125

Approximately 1.3% of women will be diagnosed with ovarian cancer in their lifetime (1). Increased ovarian cancer risk is associated with progesterone deficiencies during infertility or with increasing age, and may occur as the result of genetic loss of heterozygosity at the progesterone receptor (PR) gene locus (2, 3). Conversely, elevated progesterone levels experienced during pregnancy transiently and reversibly increase breast cancer risk (4), but significantly reduce ovarian cancer risk in Asian, European, and North American populations (5). Consistent with these findings, use of hormonal contraceptives (that include a progestin) is associated with increased risk of breast cancer (6–9) but reduced risk of ovarian cancer (9–11). During ovulation, the release of a mature follicle into the fallopian tube space requires shedding of ovarian epithelial cell layers followed by local proliferation that is analogous to wound repair. High levels of progesterone experienced during pregnancy or during the luteal phase of the normal menstrual cycle prevent follicle development. Thus, progesterone is believed to exert its protective effect on ovarian cancer in part by reducing the number of times ovulation occurs (i.e., via fewer cycles of ovulation-associated “damage and repair”).

Similar to PR expression in luminal (estrogen receptor (ER)-positive) breast tumors, PR expression in ovarian tumors is a favorable prognostic marker associated with longer progression-free survival (12–20). Roughly 35% of ovarian tumors (all subtypes) express PR; PR expression is highest in the endometrioid (67%) and serous (35%) subtypes (21, 22). Full-length PR-B and N-terminally truncated PR-A (-164 amino acids) isoforms are encoded by the same gene and mRNAs. PR isoforms are ligand-activated transcription factors with distinct transcriptional activities. Following ligand binding, PRs dimerize (A:A, B:B, and A:B) and are retained in the nucleus where they repress or activate transcription of PR-target genes, either directly through binding to progesterone response elements (PRE) on chromatin, or indirectly via tethering interactions with other transcription factors (e.g., AP1, SP1, STATs, FOXO; refs. 23–26).

Studies of isoform-specific knockout mice determined that PR-B is required for mammary gland development, while PR-A is required for uterine development and reproductive actions (27–30). While PR-A and PR-B share structural and sequence identity downstream of the BUS (B-upstream segment), they are unique transcriptional regulators of distinct gene sets (31). Little is known about how PR isoform–specific transcription is regulated in PR-expressing tissues and tumors; total PR rather than PR isoform expression is measured clinically. Progesterone and progestins, acting through PR-B, are proliferative in the breast (32). However, in the endometrium of the uterus, paracrine signals secreted from PR-A–containing stromal cells antagonize estrogen-induced epithelial hyperplasia (29).

Our previous study (21) demonstrated that ligand-activated PR-B induces cellular senescence via induction of known cellular senescence mediators, including p21 and p15 via a FOXO1-dependent mechanism in ovarian cancer cells. FOXO1 interacts with steroid hormone receptors (SR), including the androgen receptor (AR; refs. 33, 34), ERα (35), and both PR isoforms, PR-A and PR-B (36, 37). PR-B and FOXO1 were corecruited to the same PRE-containing region of the p21 upstream promoter. Both proteins were required to activate p21 expression; stable knockdown of FOXO1 or inhibition of FOXO1 activity using the small-molecule inhibitor AS1842856 blunted progestin-induced p21 expression in PR-B–expressing cells and blocked PR-dependent cellular senescence.

PR isoform–specific actions have not been studied in ovarian cancer models. In vitro studies primarily performed in breast or uterine cancer models have demonstrated PR-A transrepression of PR-B, as well as other SRs, including AR and ERα (38). Notably, PR-A expression is markedly reduced relative to PR-B in ovarian tumors (16, 39, 40). To study PR isoform–specific gene regulation and biologic consequences, we engineered ovarian cancer (ES-2) cells to stably express either empty vector (control), PR-A-only, or PR-B-only. Our studies indicate that PR-B is the dominant driver of cellular senescence in ovarian cancer cells and reveal a novel mechanism of regulation of hormone sensitivity via PR isoform–specific target gene expression; the presence of activated FOXO1 confers potent PR-B–like transcriptional activity to PR-A. Remarkably, active (dephosphorylated) FOXO1 is required for phosphorylation of PR-A on Ser294 and transactivation of PR-B at PR-B target genes. A clear understanding of PR isoform–specific actions, interactions, and required coregulators may reveal novel ways to pharmacologically select for PR-driven inhibitory over proliferative actions in hormone-driven cancers.

The human PR-B gene was previously cloned into the pEGFP-N3 vector (Clontech Laboratories, Inc.), which also served as the empty vector (EV) control vector (21, 41). GFP-tagged EV control, PR-A, and PR-B (with the isoform A start site mutated to Ala) stable clonal cell lines were generated using the parental ES-2 cell line as a model system. Stable cell lines were generated by transfecting cells with 2 μg of their respective plasmids using FuGene HD transfection reagent (Roche, #04709691001) according to manufacturer's instructions. Twenty-four hours posttransfection, cells were selected and maintained with McCoy 5A modified medium supplemented with 10% charcoal-stripped FBS (i.e., DCC; Hyclone, #SH30068.03), 100 units/mL penicillin, 100 μg/mL streptomycin, and 0.5 mg/mL of G418 sulfate (Corning, #61-234-RG). FACS with a FACSAria II cell sorter (BD Biosciences) was used to purify GFP+ EV, PR-A-, and PR-B–containing cells by removing any low and non-GFP–expressing cells. Clones were then established and cultured from the FACS-purified population.

Stable FOXO1 expression cells were generated by infecting ES-2 PR-A–expressing cells (clone #1, #5) with the retroviral pBabe puro L vector (which also served as the EV control) containing the constitutively active FOXO1 (FOXO1-AAA; ref. 42). The pBabe puroL HA FKHR AAA plasmid was a gift from William Sellers (Addgene #9025). Cells were selected in and maintained as described earlier with 1 μg/mL of puromycin (21).

Ovarian cancer tissues were provided by the University of Minnesota Biological Materials Procurement Network (BioNet). All deidentified tissue samples received in this study were obtained with written informed consent in accordance with the University of Minnesota Institutional Review Board under exemption status.

Dissection, plating, and treatments of ovarian cancer tissue were performed as described previously (43–45), with a few exceptions. After surgical excision and pathologic examination, fresh ovarian cancerous tissue was placed in 10 mL of McCoy 5A medium supplemented with 5% DCC for transport back to the laboratory. Tissue was dissected into 1-mm³ pieces and cultured in duplicate wells containing presoaked gelatin sponges (Ethicon, Inc., #1969) in 12-well plates containing 1.5 mL explant media [McCoy 5A medium supplemented with 10% DCC, hydrocortisone (0.01 mg/mL), and insulin (0.01 mg/mL)] within 1 hour of receiving tissue from BioNet. Tissue cultures were placed in a 37°C incubator with 5% CO₂ for 24 hours. Afterwards, media were gently aspirated from each well. Explant media (1.5 mL) containing 10 nmol/L R5020 or equal volume vehicle (ethanol) were added to corresponding treatment wells and placed in a 37°C incubator with 5% CO₂. To equilibriate the media were gently aspirated every 1 hour and replenished with fresh explant media containing 10 nmol/L R5020 or equal volume vehicle (ethanol) for a total of three times. Plates were returned to a 37°C incubator with 5% CO₂ for 48 hours. Afterwards, tissues were gently removed from the sponges with sterile forceps and processed for RNA or protein isolation as described below.

Cells were treated with the following reagents (when applicable): R5020 (Perkin Elmer, #NLP004005MG) and AS1842856 (EMD Millipore, #344355). For experiments with AS1842856, cells were pretreated for 18 to 24 hours prior to the addition of R5020 in combination treatment studies.

Cells were cotransfected overnight using FuGene HD transfection reagent (Roche) according to manufacturer's instructions with either 0.5 μg of a PRE-containing (ref. 31; 2X-PRE) firefly luciferase reporter construct, 1 μg of a p21 promoter-containing firefly luciferase reporter construct (21), or 4 μg (for p21 promoter-luciferase assay) or 0.5 μg (for RT-qPCR analysis of endogenous genes in PR-B cells) of GFP-tagged N3-PRA vector. The constitutively active pRL-TK-Renilla luciferase construct (Promega, #E2241; 10 ng) was cotransfected as a transfection control. Luciferase assays were performed as previously described (21) using the dual luciferase reporter assay (Promega, #E1910).

In experiments using HeLa cells, 10 ng of GFP-tagged EV or PR-A vector, 10 ng pcDNA3 empty vector (pc-EV) or pcDNA3 Flag FKHR (ref. 46; pc-FOXO1 WT), 0.5 μg of a PRE-containing (ref. 31; 2X-PRE) firefly luciferase reporter construct, and 10 ng of a constitutively active pRL-TK-Renilla luciferase construct (Promega, #E2241) were transiently cotransfected. pcDNA3 Flag FKHR was a gift from Kunliang Guan (Addgene plasmid # 13507).

One microgram of pcDNA3 EV, pcDNA3 Flag FKHR WT, or Flag FKHR AAA (46) were transiently transfected in PR-B+ (clone #1) cells. pcDNA3 Flag FKHR WT and FKHR AAA were a gift from Kunliang Guan (Addgene plasmid # 13507, 13508).

Clonal ES-2 cells stably expressing GFP-tagged EV (clone #3), PR-A (clone #7), and PR-B (clone #1) and breast cancer cells T47D Y, YA, and YB originally described by Sartorius and colleagues (47) were hormonally starved in modified improved MEM (IMEM; Gibco, catalog #A10488) plus 5% DCC for 24 hours. Afterwards, cells were treated with R5020 (10 nmol/L) or vehicle control for 24 hours prior to RNA extraction using the RNeasy kit (Qiagen, #74104). DNase I–treated (QIAgen, #79254) triplicate RNA samples were prepared for expression analysis using the Illumina HT-12v4 bead chip platform according to the manufacturer's protocols. Data were analyzed within R software (48) using the Bioconductor (49) package, lumi (50) where raw intensities were log₂ transformed and quantile normalized. Differentially expressed genes were analyzed using the limma package (51), where empirical Bayes was used to better estimate the variance of the genes. Gene expression data presented contain log₂-normalized intensities and biologic comparisons presented (e.g., R5020/vehicle) contain log₂ fold change with the Benjamini and Hochberg (BH) adjusted P value (52). Heatmaps were generated by unsupervised hierarchical clustering of genes via the “aheatmap” function in the NMF R package (53). Clustering was performed using Pearson distance and average linkage. Rows were scaled to have mean zero and SD equal to one. All gene expression data are available in the NCBI Gene Expression Omnibus (GEO) database (accession number: GSE69296): http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=chkzqyeojjwnrql&acc=GSE69296.

Total RNA was extracted from cells in triplicate wells using TriPure Isolation Reagent (Roche, #11667165001) and isopropanol precipitation. RNA (1.0 μg) was reverse transcribed to cDNA according to manufacturer's instructions using the qScript cDNA SuperMix (Quanta Biosciences, #95048-100). qPCR was performed using Light Cycler FastStart DNA Master SYBR Green I (Roche, #12239264001) on a Light Cycler 480 II Real-Time PCR System (Roche). qPCR cycling conditions with primer sequences (available upon request) and concentrations were performed as described previously (21).

Western blots were performed as previously described (21). Protein lysates from the ex vivo culture of primary human ovarian tumors were assayed with 4%–20% gradient SDS-PAGE precasted gels. Western blots were probed using the following primary antibodies: PR-A/B (H-190) and p21 (C-19) were purchased from Santa Cruz Biotechnology; FOXO1 (#2880), phospho-AKT (Ser473; #9271), and AKT (#9272) were purchased from Cell Signaling Technology; and Actin from Sigma-Aldrich (#A4700). Phospho-Ser190 PR anti-sera was purchased from NeoMarkers (#MS-1331-P1). Custom made phospho-Ser294 PR antibodies (clone 8508) were commissioned from Thermo Fisher Scientific designed to recognize the following phospho-specific peptide sequence: C-PMAPGR(pS)PLATTV-amide. HRP-conjugated goat anti-rabbit and goat anti-mouse (Bio-Rad, #170-6515 and #170-6516) secondary antibodies were used to detect their respective primary antibodies, and immunoreactive proteins were visualized on Kodak X-OMAT LS film (Carestream Health, #864-6770) following ECL detection with Super Signal West Pico Maximum Sensitivity Substrate (Pierce, #34087). All Western blotting experiments were performed at a minimum in triplicate, and representative experiments are shown in each respective figure.

Cells were treated in 5% DCC with either 10 nmol/L R5020 or equal volume vehicle (ethanol) for 96 hours. Cells were collected, fixed, and stained as described previously (21). Propidium iodide staining was detected using a LSRII (BD Biosciences, #H4760). Cells were gated for cell-cycle phases using BD Biosciences FACSDiva 8.0 software. Cell-cycle profiles were analyzed using FlowJo vX software (Tree Star Inc.).

Cells were treated for 1 hour with either R5020 (10 nmol/L) or equal volume vehicle (ethanol) in 5% DCC and cell samples were fixed, harvested, and lysed according to optimized manufacturer's instructions using the ChIP-IT Express Magnetic Chromatin Immunoprecipitation Kit (Active Motif, #53008). Samples were homogenized using a Bioruptor sonicator (Diagenode, Inc.). ChIP reactions were incubated overnight on an end-to-end rotator using 100 μL of isolated chromatin and either 2 μg of PR-A/B antibody (Ab-8), FKHR (H-128) antibody (Santa Cruz Biotechnology, #sc-11350), or 0.4 μg of normal mouse or rabbit IgG (Santa Cruz Biotechnology, #sc-2025 # sc-2027). Samples were washed, eluted, reverse cross-linked, and treated with Proteinase K according to manufacturer's instructions (Active Motif). DNA was analyzed by RT-qPCR as described above.

Cells were plated in 96-well plates in IMEM plus 5% DCC. Twenty-four hours later, cells were treated with either R5020 (10 nmol/L) or equal volume vehicle (ethanol) in IMEM plus 5% DCC for 96 hours. Three hours prior to the 96-hour timepoint, 10X 5-bromo-2′-deoxyuridine (BrdUrd) antibody from the BrdU Cell Proliferation Assay Kit (Cell Signaling, #6813) was prepared in IMEM and then added to each well for a final concentration of 1×. Plates were returned to a 37°C incubator with 5% CO₂ for 3 hours. Cells were fixed, DNA was denatured, and detection of BrdUrd incorporation into cellular DNA during cell proliferation was conducted according to manufacturer's protocol.

Cells were continuously treated for 96 hours in IMEM plus 5% DCC. Cells were washed, fixed, and stained for senescence associated-β-galactosidase activity (SAβGal) activity according to manufacturer's instructions using the SAβGal (Cell Signaling Technology, #9860). Staining of cell nuclei was achieved with DAPI containing ProLong Gold antifade reagent (Invitrogen, #P-36931). All bright-field and fluorescent cell images described herein were acquired with a Leica DM4000B microscope (Leica Microsystems, Inc.) and captured using the Leica Application Suite software (version 4.2). Percentage of positive SAβGal cells (blue staining with enlarged nuclei) was determined from quantifying three fields at 100× magnification using ImageJ software. Values were normalized to total nuclei present per field from DAPI staining.

All reported values represent the mean ± SD. Data shown are representative of the indicated replicates of each experiment. Statistical analyses were performed using a Student t test where significance was determined with 95% confidence (*P ≤ 0.05, **P ≤ 0.01).

Cell line models derived from human tumors frequently lose SR expression when propagated in monolayer tissue cultures. We detected low to negligible levels of PR mRNA and protein in a panel of well-characterized ovarian cancer cell lines including the clear cell carcinoma cell line, ES-2 (21). PR expression in these models was only modestly induced by estrogen. Thus, to study the impact of progesterone/PR signaling on ovarian cancer cell biology without the added complexity of estrogen addition (PR is an ER target gene), we generated ES-2 cell models stably expressing either PR-A or PR-B. We selected ES-2 cells due to their inherent aggressive nature and rapid growth rate (54), low endogenous PR mRNA levels (21), and molecular and genomic characteristics that are similar to primary tumors of high-grade serous ovarian carcinomas (the most common cancer of the ovary; ref. 55).

We initially created pooled populations of ES-2 cells stably transfected with GFP-tagged PR-A, GFP-tagged PR-B (with the PR-A start site mutated), or GFP-only EV control (Fig. 1A). GFP-tagged PRs are fully functional relative to untagged PRs in COS and HeLa cell models (41, 56). To verify that GFP-tagged PR isoforms behave as ligand-activated transcription factors in progestin-treated ES-2 cells, PR transcriptional activity in pooled populations of ES-2 cells expressing EV control, PR-A, and PR-B was evaluated using PRE-luciferase reporter assays. Similar to results obtained in breast cancer cell models (57), PR-B was more transcriptionally active (∼5-fold) relative to PR-A expressed in ES-2 ovarian cancer cells as measured using this minimal (2X-PRE-driven) promoter; the synthetic progestin R5020 (10 nmol/L; for 18 hours) significantly increased luciferase expression in ES-2 cells expressing PR-A (1.8-fold) and PR-B (9.5-fold) relative to vehicle or EV (GFP-only) controls (Fig. 1A). Both isoforms underwent a slight up-shift in gel mobility upon R5020 treatment (Fig. 1A inset), indicating direct PR phosphorylation (further addressed below). Endogenous PR-A (HEF1) and PR-B (BIRC3) target genes were also selectively upregulated in an isoform-specific manner in pooled ES-2 cell populations relative to vehicle and EV controls (Fig. 1B). Interestingly, basal BIRC3 mRNA levels were higher in cell pools expressing PR-A only relative to EV control and untreated PR-B–expressing cells (Fig. 1B), suggestive of potential ligand-independent actions of PR-A in this setting (further addressed below).

PR expression in pooled populations likely varies between individual cells. To directly compare PR isoform–specific actions in ES-2 ovarian cancer cells to well characterized breast cancer models, multiple clonal PR+ ES-2 cell lines were isolated and expanded after selection for PR protein expression levels comparable with PR expressed in T47D breast cancer cell lines stably expressing either PR-A only (YA), PR-B only (YB), or constitutively expressing both endogenous isoforms in the absence of estrogen (CO; ref. 47; Fig. 1C). One GFP-only EV control clone (EV #3), four GFP-PR-A–expressing clones (PR-A #1, #5, #4, #7), and two GFP-PR-B–expressing clones (PR-B #1, #3) were selected for further study. PR phosphorylation status was examined in PR-containing clonal cell lines following R5020 (10 nmol/L for 1 hour) treatment. Similar to pooled populations (Fig. 1A inset), R5020 induced a subtle upshift in PR gel mobility in studies with clonal cell lines as measured by Western blotting (Fig. 1D). PR Ser190 is a constitutive phosphorylation site similarly present in each clone. PR Ser294 is a proline-directed (i.e., primarily phosphorylated by CDKs and/or MAPKs; refs. 58–60) phosphorylation site robustly phosphorylated in PR-B–expressing cells in the presence of R5020 (a nonspecific protein just below PR-B was detected in all lanes). Notably, PR-A–expressing clones exhibited varied ligand-dependent phosphorylation of Ser294, with robust levels detected in PR-A–expressing clone #7 cells relative to clone #4 cells. Phosphorylation of Ser294 occurs on PR-B, but not PR-A, in breast cancer cell models (57, 61). Together, these data confirm that GFP-tagged PR-A and PR-B stably expressed in ES-2 cells can bind ligand, undergo phosphorylation at well-characterized sites, and selectively regulate isoform-specific endogenous target genes.

To assign isoform-specific transcriptional actions to PR-A and PR-B stably expressed in ovarian cancer cells, we performed genome-wide transcriptional profiling of representative clonal ES-2 cell lines expressing either EV (clone #3), PR-A (clone #7), and PR-B (clone #1) following treatment without or with R5020 (10 nmol/L for 24 hours) and created representative heatmaps (Fig. 2A) and Venn diagrams (Fig. 2B and C) illustrating isoform-specific gene regulation (>2-fold up or downregulated genes are shown). Cells expressing either PR-A–only or PR-B–only regulate distinct gene clusters relative to the same parental cells expressing EV (GFP-only). Simply the expression of PRs in the absence of exogenously added ligand dramatically altered the transcriptome. Strikingly, in the absence of ligand, PR-A altered the expression of more genes (100 genes downregulated and 61 genes upregulated) relative to PR-B (31 genes downregulated and 17 genes upregulated; Fig. 2B). Ligand-independent actions of PR-A (as measured by gene array) were also reported using isoform-specific and inducible breast cancer models (62, 63). In contrast to our findings with PR-A–expressing cells, in the presence of ligand, PR-B (132 genes upregulated and 69 genes downregulated) was the more active transcriptional regulator relative to PR-A (53 genes downregulated and 38 genes upregulated). These data suggest that PR-B is the dominant hormone-sensitive PR isoform in ovarian cancer cells, while PR-A may mediate significant ligand-independent actions, but functions predominantly (in the absence or presence of ligand) via repressive actions (further addressed below).

Regulation of isoform-specific PR target genes in ES-2 cells suggests important functional differences between PR isoforms (Fig. 2B and C). To understand what specific cellular pathways may be isoform specifically regulated, we performed Gene Ontology (GO) analysis for gene terms significantly regulated by either PR-A or PR-B (Supplementary Fig. S1). In the presence of progestin, gene terms significantly upregulated by PR-A (38 genes, 27 of which were also upregulated by PR-B) were enriched in cell adhesion, regulation of cell-substrate adhesion, extracellular matrix organization, and growth factor binding. Gene terms significantly downregulated by PR-B in response to progestin (69 genes, 30 of which were also down-regulated by PR-A) were also associated with regulation of the extracellular matrix, but significantly included genes associated with regulation of the basement membrane. Gene terms associated with pathways regulating the cellular response to oxygen levels were significantly upregulated by ligand-bound PR-B. Other PR-isoform–specific regulated gene terms significantly associated with known pathways as defined by GO are listed in Supplementary Fig. S1.

In contrast to its protective role in reproductive tissues, progesterone is a potent mitogen in the mammary gland (64). PR-B, but not PR-A, is required for ductal side-branching and alveologenesis that occurs during mammary gland development. Breast cancer cell models expressing PR-B, but not PR-A, grow in soft agar in response to progestin treatment (65–67). Indeed, both overlapping and isoform-specific PR target gene up- and downregulation has been extensively reported in breast cancer models (31, 62). To illustrate potentially distinct transcriptional actions of PR isoforms in ovarian relative to breast cancer models, we repeated whole genome profiling studies in T47D cells stably expressing either PR-A–only (YA) or PR-B–only (YB), using the naturally occurring PR-null variant parental T47D cell line (Y) as a control. As above for ES-2 cells, T47D cells were treated with R5020 (10 nmol/L) for 24 hours, and global gene expression profiles were measured using whole genome gene expression bead (Illumina) arrays. As in our ovarian cancer cell models, transcriptional differences between PR-null T47D cells and parental T47D cells stably expressing either PR-A or PR-B were readily observed (Fig. 2D). However, in sharp contrast to hormone-naïve ovarian cancer models, in the absence of progestin, breast cancer cells expressing PR-B upregulated more genes (194 genes) relative to cells expressing PR-A (111 genes; Fig. 2E). In the absence of ligand, expression of PR-A (168 genes) or PR-B (157 genes) resulted in downregulation of a similar number of genes from distinct but largely overlapping subsets (103 genes in common; Fig. 2E). In the presence of progestin, PR-A and PR-B both activated and repressed numerous genes from distinct but largely overlapping subsets (Fig. 2F). Thus, in contrast to PR expression in ovarian cancer models (where there is modest overlap between PR-A- and PR-B–regulated gene sets, and PR-B appears to be the dominant hormone–regulated isoform) PR expression in breast cancer models exhibits moderate overlap between PR-A- and PR-B–regulated gene sets and both isoforms are highly responsive to hormone when these models are assayed using similar conditions (i.e., compare Fig. 2 to Supplementary Fig. S2). We detected no differences in transcriptional activity between GFP-tagged PRs and untagged PRs in multiple cell types (data not shown).

Subsequently, differential regulation of PR isoform-specific target genes identified using the Illumina platform (Fig. 2A–C) was validated by RT-qPCR in multiple ES-2 clones (Fig. 3). Progestin-induced expression of selected PR-A and PR-B target genes was evaluated at 24 and 96 hours (R5020; 10 nmol/L). PR-A target genes, CRISPLD1 and WISP1, known to be involved in the regulation of the extracellular matrix, were robustly induced following R5020 treatment of PR-A–expressing clones (#4 and #7) relative to modest or insignificant regulation in PR-B–expressing clones (#1 and #3). CRISPLD1 belongs to a super-family of proteins that have extracellular endocrine or paracrine functions and are also involved in the regulation of extracellular matrix and in cell–cell adhesion during fertilization (68). WISP1 binds to proteoglycans, decorin, and biglycan, present in the extracellular matrix of connective tissues (69). PR-B target genes BIRC3 (a mediator of cancer cell prosurvival and a well-characterized PR-B target gene in breast (66) and endometrial (70) cell models) and GZMA [a novel PR-B target gene identified herein and a mediator of cytotoxic T-cell responses significantly elevated in the serum of women with ovarian cancer (71)] were robustly induced in PR-B–expressing clones (#1 and #3) relative to weak or insignificant induction in PR-A–expressing clones (#4 and #7) or controls (Fig. 3B). Basal expression of BIRC3 noted in PR-A+ ES-2 pools (Fig. 1B) validated in clonal PR-A–expressing clones (Fig. 3B; #4 and #7 at 24 hours) and increased over time in untreated cells (compare vehicle-treated 24 to 96 hours for each clone) but was primarily hormone-regulated in PR-B+ clones. Similarly, PDLIM1 [a novel PR target gene identified herein that encodes a cytoskeleton-associated protein linked to breast cancer progression as a mediator of cell migration and invasion (72)] was similarly validated with robust induction at both time points in PR-A as well as PR-B–expressing clones; highest total levels of hormone-induced PDLIM1 expression were noted in PR-B–expressing clones.

To more fully understand the basis of PR isoform–specific gene regulation in ovarian cancer cells, we considered the largely repressive actions of (primarily unliganded) PR-A relative to the predominantly hormone-sensing actions of (liganded) PR-B (compare Fig. 4A, left to right Venn diagram). Only four genes were significantly repressed by PR-A and activated by liganded-PR-B: KYNU, FOXO1, p21, and ZDHHC9 (Fig. 4A). KYNU is involved in the biosynthesis of NAD cofactors from tryptophan catabolism through the kynurenine pathway. The expression of KYNU is significantly downregulated in several cancer types, such as in invasive ductal breast carcinomas (73), pediatric acute myeloid leukemias harboring IDH mutations (74), and highly aggressive osteocarcinoma cell lines (75). ZDHHC9 is a palmitoyltransferase specific to HRAS and NRAS (76) and reported to be widely overexpressed in human cancers (77). Notably, we previously identified both FOXO1 and p21 as required factors for progestin-mediated cellular senescence in PR-B+ ovarian cancer cells (21). Our microarray experiments revealed repression of both p21 and FOXO1 mRNA expression in untreated or treated PR-A–expressing (clone #7) ES-2 cells relative to EV (clone #3) controls (Fig. 4B). In contrast to PR-A+ cells, R5020 treatment (24 hours) significantly upregulated p21 and FOXO1 mRNA in PR-B+ cells (clone #1). These results (i.e., robust PR-B–selective upregulation of FOXO1, p21, and the FOXO1 target gene, p15) were validated by RT-qPCR in multiple clones of ES-2 cells stably expressing either PR-B–only or PR-A–only and treated with R5020 for 24 and 96 hours (Fig. 4C–E). FOXO1, p21, and p15 mRNA levels were either slightly upregulated or unchanged in PR-A+ cells relative to their pronounced induction in PR-B+ cells (24 hours). These genes were also insensitive to progestin in EV+ (clone #3) controls (Fig. 4C–E). Western blot analysis confirmed these results (Fig. 4F), demonstrating robust upregulation of FOXO1 and p21 protein with 96-hour R5020 treatment in PR-B+ clones (#1 and #3). Interestingly, R5020 treatment weakly downregulated FOXO1 protein levels in PR-A+ clones relative to PR-B+ clones (Fig. 4F), while FOXO1 and p21 levels remained unaltered in EV+ control cells following 96-hour R5020 treatment (data not shown). Taken together, these data indicate that while progestin-treated PR-A+ cells are capable of modestly inducing FOXO1 and p21 at the mRNA level, changes in protein expression do not persistently manifest (i.e., by as late as 96 hours). In the presence of progestin, PR-B is a stronger transcriptional activator of known senescence mediators relative to PR-A.

To demonstrate the repressive action of PR-A on the p21 promoter, we performed chromatin immunoprecipitation (ChIP) assays to determine whether liganded PR-A binds to PRE-containing regions of the endogenous p21 promoter. Indeed, in the presence of progestin, PR-A was significantly recruited (9- and 21-fold in clones #4 and #7, respectively) to a PRE-containing region downstream of the p21 transcriptional start site previously identified by ChIP-Seq studies conducted in PR+ breast cancer models (78) and validated in our previous study (21). While some basal level of PR was detected in vehicle-treated EV (PR-low) controls, ligand-dependent PR recruitment was minimal and comparable with background (IgG control) levels (Fig. 4G). Previously (21), we detected very low PR mRNA levels in ES-2 cells relative to T47D breast cancer cells. While PR protein expression is undetectable in ES-2 (EV) cells as measured by Western blotting and PRE-driven luciferase reporter assays, low levels of PR protein may occupy the p21 promoter as detected by highly sensitive ChIP assays. Additional data (Fig. 4G) representing the average fold recruitment (R5020/Vehicle) of three separate experiments illustrates weak PR recruitment in EV control (clone #3) cells relative to PR-A–expressing clones. We previously reported (21) similar ligand-dependent recruitment (5-fold relative to controls) of PR-B to the same PRE-containing region of the p21 promoter.

Taken together, our data suggest that PR-B–containing transcriptional complexes activate p21 transcription, while PR-A–containing complexes are repressive. Changes in gene expression are linked to epigenetic histone tail modifications, such as methylation and acetylation. Histone H3 Lys4 dimethylation (H3K4me2) is an epigenetic modification associated with transcriptional activation. To measure the level of H3K4me2 (i.e., as a surrogate marker of p21 gene activation) at the same PRE/PR-containing p21 promoter region, EV control (clone #3), PR-A–expressing (clone #7), and PR-B–expressing (clone #1) cells were treated with vehicle control or R5020 (10 nmol/L for 1 hour) and the degree of histone H3 Lys4 dimethylation was measured by ChIP assay (Fig. 4H). H3K4me2 levels were significantly elevated (6-fold) only in progestin-treated ES-2 cells stably expressing PR-B (clone #1) relative to ES-2 cells stably expressing PR-A (clone #7) and EV (clone #3) control. These results suggest that PR-A is recruited to the p21 promoter, but is incapable of mediating robust hormone-dependent p21 gene activation.

To account for the differences in PR isoform-specific regulation of senescence mediators (Fig. 4), we evaluated the senescence phenotype in PR-A- and PR-B–expressing ES-2 cells. Our previous study demonstrated that in the presence of progestin, PR-B promoted cellular senescence, a cell fate wherein cells remain viable and long-lived, but are nonproliferative (21). A common marker of cellular senescence is the accumulation of endogenous lysosomal β-galactosidase. We estimated the percentage of senescent ES-2 cells via measurement of the activity of SAβGal at pH 6 (ref. 79; i.e., blue stained cells with enlarged nuclei). Representative images showing SAβGal and DAPI staining performed in clonal ES-2 cell lines expressing EV (clone #3), PR-A–only (clone #7), or PR-B–only (clone #1) and treated (96 hours) with either R5020 or vehicle control are depicted (Fig. 5A) along with quantitation of SAβGal-positive cells in five clonal cell lines (Fig. 5B). As expected, R5020 (96 hours) significantly induced senescence as measured by SAβGal in PR-B–expressing ES-2 cells (clones #1 and #3 exhibited 51% and 59% positive cells, respectively) relative to vehicle controls. In contrast, PR-A–expressing ES-2 cells (clones #4 and #7) exhibited significantly less SAβGal (21% and 22%, respectively) relative to PR-B–expressing ES-2 cells. Minimal SAβGal levels were observed in EV #3 (2%) and vehicle-treated cells (Fig. 5A and B).

Cellular senescence is often associated with a decrease in cell proliferation that is accompanied with cell-cycle exit (i.e., cells accumulate in G₁). Thus, we examined BrdUrd incorporation to detect altered proliferation following progestin treatment of PR-expressing ovarian cancer cells. After 96-hour R5020 treatment, BrdUrd incorporation was significantly inhibited in PR-B–expressing cell clones, indicating that R5020 attenuated DNA synthesis and halted proliferation. R5020 treatment did not significantly alter BrdUrd incorporation in EV (clone #3) and PR-A–expressing (clone #4) cells, and only weakly inhibited PR-A+ clone #7 cells relative to that observed in PR-B–expressing clones (Fig. 5C). Finally, cell-cycle (FACS) analysis of PR-B–expressing cells, as measured by propidium iodide staining, confirmed a significant increase in the percentage of cells in the G₀–G₁ phase of the cell cycle accompanied by a decrease in the percentage of cells in S-phase following 96 hours of R5020 exposure relative to vehicle controls (Fig. 5D). Similar to the above findings using BrdUrd, progestin treatment elicited minimal or insignificant effects on the cell-cycle distribution of ES-2 cells expressing either EV (clone #3) or PR-A (clones #4 and #7; weak but significant effects were again observed in PR-A clone #7). Taken together, our data suggest that in the presence of progestin, PR-B is a stronger promoter of cellular senescence relative to PR-A in ovarian cancer cells.

Similar to cell-cycle studies performed in breast cancer models (65, 80), relative to PR-B, PR-A often fails to elicit robust induction of cell-cycle mediators (p21) and their associated biologic responses (i.e., biphasic regulation of cell proliferation in breast models), but may weakly mimic PR-B. Recall that progestin treatment weakly induced FOXO1 mRNA expression (Fig. 4C) but weakly downregulated FOXO1 protein levels (Fig. 4F) in FOXO1-expressing PR-A+ clones (#4 and #7) relative to PR-B+ clones (at 96 hours). We thus evaluated whether FOXO1 activity was required for the weak but significant induction of cellular senescence in PR-A–expressing clone #7 using the selective FOXO1 small-molecule inhibitor, AS1842856 (AS; refs. 21, 81; Supplementary Fig. S3). R5020 treatment of PR-A+ (clone #7) ES-2 cells for 96 hours modestly induced FOXO1 protein and mRNA expression relative to vehicle controls (Supplementary Fig. S3A and S3B). Treatment with AS alone did not affect basal PR-A or FOXO1 protein and mRNA expression, but AS blocked R5020-induced FOXO1 protein and mRNA expression (Supplementary Fig. S3A and S3B). FOXO1 gene expression is known to be autoregulated (i.e., FOXO1 contributes to regulation of its own gene expression; ref. 82). Similar results were observed for R5020-induced p21 and p15 mRNA expression (Supplementary Fig. S3C and S3D). Consistent with these results, progestin-induced senescence, as measured by SAβGal activity, was significantly reduced upon inhibition of FOXO1 by AS (100 nmol/L) in combination with R5020 for 96 hours relative to R5020-only treated controls (Supplementary Fig. S3E). Together, these data implicate FOXO1 as a required mediator of senescence-associated gene expression in PR-A+ ES-2 cells treated with progestin.

Progesterone, via PR-dependent rapid signaling actions, transiently activates numerous protein kinases, including the PI3K/AKT pathway in diverse cell types (83, 84). Because AKT signaling is a key input to FOXO1 inactivation (i.e., via direct phosphorylation), we examined the level of AKT activity in our PR+ ES-2 ovarian cancer cell clones. ES-2 cells were plated in media devoid of growth factors (see Materials and Methods) and treated with a time course (0–6 hours) of R5020 (10 nmol/L). Notably, basal AKT activity was undetectable in multiple clones of untreated PR-A+ or PR-B+ ES-2 cells. However, R5020 treatment of either PR-A- or PR-B–expressing ES-2 cells transiently activated AKT as measured by Ser473 phosphorylation at 30 minutes with peak levels observed at approximately 1 hour; AKT activity subsided by 4 hours and returned to near basal levels by 6-hour R5020 treatment (Fig. 6A). Similar results were observed in additional distinct clones of PR-A+ and PR-B+ cells (data not shown).

We next speculated that overexpression of FOXO1 in PR-A–expressing cells would confer PR-B–like behavior and thereby “rescue” hormone sensitivity and senescence-associated gene regulation similar to that induced in PR-B+ cells. PR-A–expressing ES-2 clones were screened for low to undetectable FOXO1 protein levels by Western blot analysis (Fig. 6B). These clones were fully capable of inducing robust expression of the PR-A target gene WISP1, but failed to upregulate PR-B–selective genes including FOXO1 following 96-hour R5020 treatment (data not shown). The PI3K/AKT pathway is constitutively activated in a majority of ovarian cancers (85). Phosphorylation of FOXO1 at Thr24, Ser256, and Ser319 by AKT or other kinases negatively regulates its nuclear localization and transcriptional activity (46, 86). We therefore utilized the constitutively active form of FOXO1 (FOXO1-AAA) that cannot be phosphorylated by AKT (42). FOXO1-AAA or EV control were stably expressed by retroviral transduction of PR-A+/FOXO1-low (clones #1 and #5) ES-2 cells. Expression of total FOXO1 protein was confirmed by Western blotting (Fig. 6B).

Notably, R5020 only weakly increased p21 mRNA expression after 96 hours in PR-A+ (clone #1) cells expressing EV (FOXO1-null) control. However, forced FOXO1-AAA expression in these cells significantly further increased p21 mRNA induction in the presence of R5020 relative to vehicle controls at both 24 and 96 hours (Fig. 6C). These results essentially repeated in an additional ES-2 cell line coexpressing PR-A (clone #5) and FOXO1-AAA (Fig. 6C). Similar results were observed in HeLa cells transiently expressing PR-A and FOXO1 as measured via 2X-PRE-luciferase reporter assays and RT-qPCR analysis of p21 expression (Supplementary Fig. S4). ChIP assays revealed that upon R5020 treatment (10 nmol/L; 1 hour), PR-A was again recruited to the p21 promoter in PR-A+ EV-expressing (clone #1) cells, and significantly further recruited upon exogenous FOXO1-AAA expression (Fig. 6D). FOXO1-AAA recruitment to this site also occurred in the absence of ligand but was significantly increased in response to R5020 treatment (Fig. 6D). Consistent with these results, overexpression of FOXO1-AAA in PR-A+ cells (clones #1 and #5) conferred increased SAβGal activity upon R5020 treatment relative to vehicle and EV controls (Fig. 6E). The level of SAβGal activity detected in FOXO1-AAA–overexpressed PR-A+ ES-2 cells was similar to that of PR-B+ cells when treated with R5020 (Fig. 5A and B). Induction of SAβGal in R5020-treated PR-A+ cells expressing FOXO1-AAA (but not EV) was accompanied by cellular senescence-associated changes in cell-cycle distribution (i.e., increased percentage of cells in G₀–G₁; data not shown).

These data suggest that expression of activated FOXO1-AAA in PR-A+ cells confers PR-B–like behavior (i.e., hormone-responsive induction of p21 and senescence). To get at the mechanism for this effect, we assayed ligand-dependent and independent PR phosphorylation as a surrogate marker of heightened progesterone responsiveness (87). Notably, phosphorylation of PR-B Ser294 mediates changes in PR-B promoter selection and hormone sensitivity, primarily assayed in breast cancer models (57, 88–90). We included both R5020 (a PR agonist) and the selective PR antagonist, CDB-4124 (CDB). Surprisingly, while PR-Ser190 phosphorylation remained unaltered across all conditions, addition of either R5020 or CDB similarly induced robust PR-Ser294 phosphorylation in PR-A+ (clone #1) ES-2 cells expressing FOXO1-AAA relative to EV-expressing and vehicle-treated controls (Fig. 6F). Similar results were observed in a separate PR-A+ ES-2 clone (#5; data not shown). These data suggest that overexpression of activated FOXO1-AAA enhances PR-A Ser294 phosphorylation and transcriptional activity in a manner similar to PR-B activity on PR-B–selective target genes (e.g., p21). Surprisingly, overexpression of FOXO1-AAA in PR-A+ (clones #1 and #5) cells also significantly increased mRNA expression of the PR-B–selective genes, GZMA and IGFBP1 (Fig. 6G). Together, these data suggest that FOXO1 is a key mediator of enhanced hormone sensitivity on selected PR target genes (p21, GZMA, IGFBP1). Ligand-bound PR-A is fully capable of undergoing phosphorylation at Ser294 and regulating PR-B–only target genes when activated FOXO1 is present.

PR-A is capable of inhibiting the transcriptional activity of PR-B, as well as other SRs, including glucocorticoid, mineralocorticoid, androgen, and estrogen receptors (38). Transrepression of PR-B by PR-A has been primarily measured using reporter-gene assays (91, 92) rather than on endogenous genes and requires PR-A SUMOylation (93). Our data suggest that PR-A is primarily repressive on the p21 promoter, but high levels of activated FOXO1 reversed this effect and enhanced both PR-A Ser294 phosphorylation and hormone sensitivity (i.e., of PR-A on PR-B target genes; Fig. 6). To address PR-B target gene expression when PR-A and PR-B are coexpressed, we transiently transfected PR-A into PR-B–expressing ES-2 cells. PR-B+ ES-2 cells (clone #1) were transiently transfected with either GFP-tagged EV (N3-EV, control) or GFP-tagged PR-A (N3-PRA) and a p21-promoter–driven luciferase reporter (ref. 21; Fig. 7A). As previously reported (21) R5020 (10 nmol/L for 24 hours) significantly induced p21 promoter activity in PR-B–only expressing cells containing the EV control (∼2-fold) as measured using p21 luciferase assays. Surprisingly, transient expression of PR-A further increased ligand-induced p21 promoter activity (5-fold) in PR-B+ ES-2 cells. These results were confirmed at the level of transcription of the endogenous p21 promoter; addition of PR-A to PR-B+ ES-2 cells (clone #1) significantly increased p21 mRNA expression (i.e., from 3-fold in EV-expressing cells to 6-fold upon addition of PR-A; Fig. 7B). Furthermore, in the presence of progestin, the PR-B–selective genes, GZMA and IGFBP1, were significantly increased upon expression of PR-A in PR-B+ ES-2 cells (clone #1; Fig. 7C). Together, our data suggest that when FOXO1 is present (a PR-B target gene), PR-A is equally hormone responsive (i.e., relative to PR-B) and capable of robust transactivation of selected PR-B target genes (p21, GZMA, IGFBP1).

To determine cooperative transcriptional activity in cells expressing both PR-A and PR-B was FOXO1-dependent, we utilized the selective FOXO1 small-molecule inhibitor, AS1842856 (AS). Empty vector (N3-EV, control) or PR-A (N3-PRA) was transfected into PR-B+ ES-2 cells followed by treatment with either vehicle control, AS (100 nmol/L), R5020 (10 nmol/L) or the combination of AS and R50 for 24 hours. Again, the PR-B target genes, GZMA and IGFBP1, were significantly increased upon expression of PR-A in PR-B+ cells following R5020 treatment (Fig. 7D). Treatment with AS alone did not alter basal levels of GZMA or IGFBP1, but AS significantly blocked R5020-induced GZMA and IGFBP1 expression when either EV or PR-A was expressed in PR-B+ cells (Fig. 7D).

The loss of progestin responsiveness when FOXO1 activity was blocked in PR+ cells suggested that PR Ser294 phosphorylation may also be downregulated. We therefore measured PR Ser294 phosphorylation following exposure (1 hour) to either vehicle control, AS (100 nmol/L), R5020 (10 nmol/L), or the combination of AS and R5020 in PR-A+ and PR-B+ cells (Fig. 7E). As in Fig. 1D, PR Ser294 phosphorylation was detected in PR-A+ (clone #7) and PR-B+ (clone #1) cells following R5020 treatment, and this site was not regulated in vehicle or AS-treated controls. Notably, AS blocked R5020-induced Ser294 phosphorylation in PR-A- and PR-B-expressing cells. Similar results were observed in additional PR+ clonal cell lines (data not shown). Conversely, overexpression of FOXO1 increased Ser294 phosphorylation (Supplementary Fig. S5). Flag-tagged FOXO1 wild-type (WT), Flag-tagged FOXO1-AAA, or EV control plasmids were transfected into PR-B–expressing cells (clone #1) and PR Ser294 phosphorylation was detected by Western blotting. Forced expression of either WT or AAA-mutant FOXO1 increased PR-B Ser294 phosphorylation (∼2-fold) in response to R5020 and relative to EV controls. FOXO1-AAA expression also weakly increased Ser294 phosphorylation relative to similarly transfected WT FOXO1. Overall, these data implicate FOXO1 as a key cofactor required for maximal hormone sensitivity that mediates increased PR Ser294 phosphorylation and robust transactivation of selected PR-B target genes (89, 90, 94).

To validate our novel in vitro findings in ovarian cancer cell line models defined herein, we assessed the impact of PR signaling in primary human ovarian tumors using an innovative ex vivo culture system (43–45). Notably, tumor tissues cultured using this three-dimensional system retain features of the original tumor, including SR expression. As illustrated in Fig. 8A, ex vivo cultures of fresh primary ovarian tissue obtained from cytoreductive surgery were established using dissected tissues mounted on presoaked gelatin sponges equilibrated in culture media containing either vehicle control or R5020 (10 nmol/L, 48 hr). A total of seven deidentified ovarian carcinoma tissues were obtained from the University of Minnesota's tissue procurement facility (BioNet) and cultured as described in Materials and Methods (Fig. 8B). Tumors included in this study expressed the PR-B isoform only or both PR-A and PR-B (Fig. 8B); we did not observe any tumors expressing PR-A only. A high-grade serous subtype tumor (OVC-1) retained expression of the PR-B isoform (two film exposures depicting PR isoform expression in tumor lysates and T47D CO whole-cell lysate as a positive control are shown; Fig. 8C). In the presence of progestin, FOXO1 and p21 mRNA expression were significantly induced relative to vehicle controls (Fig. 8C). A separate PR-B+ primary ovarian tumor (OVC-6) was similarly responsive to progestin treatment (48 hours); p21 mRNA but not FOXO1 mRNA expression was significantly induced upon R5020 treatment (Fig. 8D). Similarly, FOXO1 mRNA but not p21 mRNA expression was significantly induced with R5020 treatment in a PR-A+/PR-B+ primary ovarian tumor (OVC-7; Fig. 8E). Additional hormone-responsive tumors are indicated (Fig. 8B). Notably, only three of seven primary ovarian tumors were responsive to progestin treatment, as measured by upregulation of either FOXO1 and/or p21. As noted above, high AKT activity in ovarian tumors may account for loss of progesterone sensitivity via inactivation of FOXO1. We therefore screened duplicate samples of primary ovarian tumors in our study with an antibody against phospho-Ser473 (active) AKT. As predicted, tumors with high AKT activity (OVC-3, OVC-4, OVC-5, OVC-6) either did not respond or significantly downregulated FOXO1 or p21 mRNA levels, whereas tumors with low basal AKT activity (OVC-1, OVC-7) remained sensitive to progestin (Fig. 8F). These data confirm that endogenously expressed PR-B induces p21 and FOXO1 expression in unmodified human primary tumors exposed to progestin. However, significant tumor heterogeneity in the response of ovarian tumors to progestins may relate in part to PR isoform expression as well as FOXO1 expression and activity (i.e., as influenced by AKT and other kinases that inactivate FOXO1 via phosphorylation events; see Discussion).

Herein, our findings are consistent with a model that underscores the importance of activated (i.e., dephosphorylated) FOXO1 as a mediator of PR hormone responsiveness (Fig. 9). In the absence of FOXO1 expression, PR-A primarily represses cellular senescence mediators, p21 and p15, and remains relatively insensitive to added progestin. In sharp contrast, PR-B–expressing cells induce abundant FOXO1, p21, and p15 expression in response to hormone. FOXO1 and PR-B further cooperate to promote a robust cellular senescent phenotype (21). When both isoforms are present, active FOXO1 confers increased hormone responsiveness to PR-A at PR-B target genes, effectively “switching” the genetic programming to induce robust cellular senescence. Overall, we conclude that limiting expression of FOXO1 or AKT-driven FOXO1 phosphorylation and inactivation are important inputs to predicting progesterone/PR-driven responses in ovarian cancers.

Herein, we modeled PR isoform expression in ovarian cancer cells to understand mechanisms of progesterone/PR-driven gene expression and cell biology. To date, two studies with large patient cohorts have examined the relative distribution of total PR levels within ovarian carcinomas. We first reported that in a cohort of 504 ovarian tumors, 35% were PR+, with the highest PR expression in endometrioid (67%) and serous (35%; low grade serous, 64%) subtypes (21). In accordance with our study, the international Ovarian Tumor Tissue Analysis (OTTA) consortium examined the association of ER and PR expression and survival in 2,933 invasive epithelial ovarian tumors with similar results (22). Furthermore, the OTTA study confirmed the prognostic significance of PR expression in ovarian tumors; high PR expression (≥50% tumor cell nuclei staining) in high-grade serous ovarian carcinomas was associated with a significant improvement in survival.

PR isoforms are expressed from a single gene via alternate usage of at least two internal translational start sites (thus potentially creating full-length PR-B, truncated PR-A, or further truncated and transcriptionally inactive PR-C isoforms). Although numerous mRNAs capable of encoding both PR-A and PR-B have been characterized (95), isoform-specific expression is also regulated via two distinct upstream (i.e., distal and proximal) promoter regions (96). At least three studies (16, 39, 40) have reported differential expression of PR isoforms in ovarian tumors. Akahira and colleagues (16) investigated PR-A and PR-B by immunohistochemistry of 107 patient tumors. They observed the dominant expression of PR-B across all of the histologic subtypes, with PR-B frequently expressed in the serous subtype; PR-A was weakly expressed in mucinous followed by serous (16). In a follow-up study, Akahira and colleagues evaluated PR isoform expression via immunohistochemical staining in normal (n = 8), benign serous cystadenomas (n = 10), borderline serous adenocarcinomas with low malignant potential (n = 8), and malignant serous adenocarcinomas (n = 24; ref. 39). They reported that PR-A expression was decreased or absent in malignant tissues relative to normal, whereas PR-B expression was present across all tissue cases in that study. The authors concluded that the absence of PR-A expression in malignant ovarian tissues is associated with the development of ovarian tumors. Finally, Lenhard and colleagues (40) examined the expression of PR isoforms in 155 ovarian patient cases, and reported PR-B expression in serous and endometrioid subtypes, while PR-A was present in serous followed by modest expression in endometrioid and clear cell (40). Ultimately, these studies observed dominant expression of PR-B in ovarian tumors across all subtypes with reduced or absent expression of PR-A. The mechanism for loss of PR-A relative to PR-B in malignant ovarian tumors is unknown. However, in endometrial cancer cells, enhanced DNA methylation was detected at the PR gene locus when PTEN expression was deleted (97). Approximately 7% of ovarian tumors evaluated in the 2011 TCGA analysis (98) harbored PTEN deletions, while 18% and 9% contained activating mutations of either PI3K or AKT, respectively (98). Similar to these findings in ovarian cancer, methylation of the PR-A but not PR-B promoter upstream of the PR gene occurs in advanced breast cancers and is significantly associated with tamoxifen resistance (99).

To our knowledge, we are the first to examine the biologic consequences of PR isoform-specific gene regulation in ovarian cancer models, as well as to directly compare PR isoform-selective transcriptomes between ovarian and breast cancer models cultured under similar conditions (Figs. 2 and 3, Supplementary Fig. S2). Remarkably, in ovarian cancer models, we found that PR-A predominantly up- or downregulates genes independently of ligand, while PR-B is the dominant hormone-responsive receptor; we were also surprised to discover relatively modest overlap between PR-A and PR-B target genes. In contrast, we observed much target gene overlap between PR-A and PR-B in similarly designed breast cancer models, where relative to PR-A, PR-B predominantly upregulated target genes in the absence of progestin. Ligand-independent actions of PR-B have been well characterized in breast cancer models (67, 90, 100). Surprisingly, there was also relatively little overlap between ovarian and breast cancer models with regard to hormone-regulated PR-target genes, with nine genes regulated by PR-A and 16 genes regulated by PR-B in both breast and ovarian cancer models (Supplementary Fig. S2). However, while PR-A- and PR-B–regulated genes were distinct in ovarian relative to breast cancer models, both isoforms regulated similar pathways involved in cell growth, proliferation, cell death, and survival (as predicted using Ingenuity Pathway analysis).

The paradoxical effects of progesterone observed in ovarian relative to breast cancer models may be attributed to differential cofactor availability (ref. 101; discussed below) as well as differential cross-talk between PR and growth factor–mediated signaling pathways (i.e., protein kinases). PR-B, but not PR-A, functions outside the nucleus to rapidly activate protein kinases (c-Src, MAPK, AKT) in part via ligand-induced interaction between a poly-Pro-rich region unique to PR-B and c-Src kinase (wherein c-Src is activated by SH3-domain ligation with PR; refs. 102–104). In breast cancer models, rapid progesterone-induced MAPK activation (i.e., downstream of c-Src) or CDK2 robustly phosphorylates PR-B (but not PR-A) at Ser294 (65, 89). In this context, PR-B Ser294 phosphorylation is required for regulation of selected target genes in either the absence (e.g., IRS-2, STC1) or presence (e.g., MSX2, RGS2, MAP1A, PDK4) of hormone (90). Thus, as expected, we observed robust ligand-induced phosphorylation of PR-B Ser294 in intact ES-2 cells (Fig. 1D). However, in stark contrast to breast cancer models [i.e., where PR-A is not well phosphorylated on Ser294 (57, 61)], we were surprised to detect phosphorylation of PR-A Ser294 in intact ES-2 cells (Fig. 1D). In breast cancer models, PR-B Ser294 phosphorylation confers hypersensitivity to progestin, increases the rate of ubiquitinylation of PR-B (an activation step for several SRs; ref. 105), delays SUMOylation on K388 (57), promotes ligand-independent transcriptional activity (106), and profoundly alters promoter selectivity (90). Ligand- or growth factor–induced PR-A Ser294 phosphorylation is generally undetectable (57, 61). Consistent with these findings, we demonstrated that PR-A K388 SUMOylation (a transcriptionally repressive modification) occurs very rapidly (15 minutes) and is sustained relative to PR-B in response to progestin or progestin plus growth factor (EGF) treatment (57). Surprisingly, herein, we observed increased PR-A Ser294 phosphorylation when FOXO1-AAA was expressed in ES-2 ovarian cancer models. Similar results were observed in PR-B+ cells (Supplementary Fig. S5) and the FOXO1 inhibitor blocked Ser294 phosphorylation of both isoforms (Fig. 7E). These data suggest that activated FOXO1 may serve to corecruit protein kinases capable of phosphorylating Ser294 (MAPKs, CDKs) to PR-containing transcriptional complexes. Taken together, our data suggest that increased PR-A Ser294 phosphorylation in the presence of activated FOXO1 may account for its ability to transactivate PR-B–specific target genes when these isoforms are coexpressed in PR-B+ cells (Fig. 7). PR-A and PR-B cross-talk in the context of multiple posttranslational modifications, including SUMOylation, is a topic of further study.

Herein, our data reveal a mechanism whereby PR isoforms may maintain either distinct or overlapping transcriptional gene programs via context-dependent upregulation of key PR cofactors. PR-B but not PR-A induces FOXO1, the same pathway cofactor it requires to induce robust p21-dependent (i.e., via PR-B/FOXO1 complexes) cellular senescence in ovarian cancer cells (21). In contrast, the relative in-availability of FOXO1 severely limits hormone sensitivity in ovarian cancer cells expressing PR-A. In PR-A+ ES-2 cells, ligand-independent repression of FOXO1 mRNA may account for concomitant minimal regulation of p21 and p15 mRNAs and weak induction of cellular senescence when FOXO1 levels are also relatively low. Surprisingly, exogenous expression of active FOXO1 (FOXO1-AAA) restored hormone sensitivity in PR-A+ cells and conferred PR-B–like transcriptional activity at selected PR-B target genes (p21, GZMA, IGFBP1). In this context (PR-A plus activated FOXO1-AAA), progestin induced a robust cellular senescence phenotype similar to that observed in PR-B+ cells. Ultimately, our study underscores the critical importance of FOXO1 as a required cofactor for PR-dependent actions in the ovary.

Steroid hormone action in ovarian cancer is grossly understudied relative to breast or prostate cancers. There is an urgent need to focus on the early events related to the contribution of hormones in the context of altered signaling events (loss of p53 or PTEN, activation of AKT signaling) that may predispose women to increased risk of ovarian cancer. Deregulation of FOXO1 is associated with tumorigenesis and cancer progression across multiple cancer types. A recent PAthway Representation and Analysis by Direct Inference on Graphical Models (PARADIGM) analysis of mRNA expression and copy number data revealed 80 significant pathways altered in primary ovarian serous cystadenocarcinomas; the FoxO family signaling pathway was identified as significantly altered in this subtype (107). FOXO1 expression and activity is downregulated through alterations in upstream regulators, posttranslational deregulation, or by genetic mutations (108). Notably, AKT negatively regulates FOXO1 through phosphorylation and prevents FOXO1 nuclear accumulation, thus impairing target gene regulation (108). In other studies, AKT-dependent phosphorylation of FOXO1 was identified as both a mechanism of FOXO1 nuclear export and protein turnover/downregulation (46, 86). Activating mutations of PI3Ks or inactivating mutations (i.e., functional loss) of PTEN are common early events in ovarian cancer. For example, activated AKT (i.e., downstream of these events) may impair PR-induced senescence signaling by nuclear exclusion of FOXO1 (Fig. 8E). The early loss or inactivation of FOXO1 may render PR “incompetent” at genes required for the induction of cellular senescence, leading to the loss of protective “sensing” by progesterone in ovarian tumors, and perhaps also in early PR+ breast tumors. Notably, analysis of The Cancer Genome Atlas data revealed significantly decreased FOXO1 mRNA expression in breast tumors relative to normal breast tissue (P < 2.2 × 10⁻¹⁶). Thus, we propose that FOXO1 acts as a “molecular switch” that confers highly hormone-sensitive cell growth inhibition/senescence to PR+ epithelial cells, but may allow progesterone to stimulate proliferation in a highly context-dependent manner. Upon loss of FOXO1, PR interaction with other available cofactors may mediate profound changes in gene expression and associated cell biology (i.e., from senescence induction to activation of proliferative/prosurvival programs). For example, STAT3 and STAT5 appear to be primary cofactors that direct PR-B–selective gene expression and increased breast cancer cell proliferation and survival (109–111). Furthermore, in proliferating (S-phase) breast cancer models, CK2 confers ligand-independent PR-B target gene selection via phosphorylation of Ser81 (not found on PR-A) and recruitment of MKP3/DUSP6 (111). These examples illustrate how early genetic events (i.e., mutations that alter kinase signaling) have profound consequences on SR-mediated (i.e., epigenetic) regulation of gene programs and associated changes in cell and cancer biology.

Our data highlight the requirement of FOXO1 for PR transcriptional regulation and specificity. A previous study demonstrated that constitutively active FOXO1 enhanced the ligand-dependent transcriptional activity of PR-A in PRE-luciferase assays (36). One caveat of luciferase assays is that they primarily measure SR transcriptional activity when bound to minimal promoter elements in artificial contexts, limiting interpretations relevant to promoter selectivity during transcriptional regulation of endogenous gene promoters (i.e., in chromatin). Thus, while PR-A may appear to be “functional” in PRE-luciferase assays, it may repress selected endogenous genes in the context of chromatin. Herein, we validated that FOXO1 enhanced the ligand-dependent transcriptional activity of PR-A as measured via the levels of endogenous p21, GZMA, and IGFBP1 mRNAs (ES-2 cells; and in HeLa cells for p21) and by readout of PRE-luciferase (HeLa cells). As noted above, regulation of FOXO1 is complex. Notably, progestins transiently activated AKT in our ES-2 cells (Fig. 6A). In this context, FOXO1 protein turnover (in response to R5020) is likely more visible in PR-A–expressing cells relative to PR-B–expressing cells (Fig. 4F), as progestins activate AKT in cells expressing either isoform (Fig. 6A), but only PR-B (but not PR-A) induces robust FOXO1 mRNA and protein expression (Fig. 4). When we stably expressed the constitutively active FOXO1-AAA in PR-A+ ES-2 lacking detectable FOXO1, the sensitivity of PR-A to hormone was greatly enhanced and this formerly repressive isoform transactivated the p21 promoter (Fig. 6). We conclude that constitutively active FOXO1 modulates PR-A hormone sensitivity and influences the specificity of gene regulation, thereby mimicking the actions of PR-B.

In light of our findings herein, PR-B dominance in ovarian tumors is intriguing. Early loss or inactivation of FOXO1 (i.e., via functional loss of PTEN and/or activation of PI3K/AKT signaling) is predicted to block induction of the senescence pathway, effectively “lifting” progesterone-mediated protection in PR-B+ ovarian cancer cells. It is tempting to speculate that in the absence of active FOXO1, PR-B acts as a driver of a proliferative/prosurvival transcriptional program similar to that observed in breast cancer models (67, 90, 100, 111). In support of this concept, flow cytometry experiments revealed a large spike in the percentage of PR-B+ ES-2 cells in S-phase upon FOXO1 knockdown relative to shRNA controls (21). The context-dependent actions of PR-B in ovarian cancer models with chronically elevated AKT (i.e., inactivated FOXO1) signaling are a topic of further investigation.

While total PR levels are routinely measured in breast and endometrial cancers for clinical management and disease treatment, very few studies have examined the levels of PR isoforms in ovarian cancers. We collected a total of seven primary ovarian tumors and detected a dominance of PR-B–only in four tumors, while three tumors coexpressed PR-B and PR-A. In the ex vivo culture system, three high-grade serous ovarian tumors recapitulated our in vitro findings of PR-B–dependent induction of FOXO1 and/or p21 upon progestin treatment. As discussed above, aside from considerable tumor heterogeneity, several factors may account for lack of sensitivity to progestin in the other ovarian tumors; unresponsive tumors may harbor mutations that render FOXO1 inactive and unable to direct PR-dependent regulation of FOXO1 and p21 (i.e., required for senescence induction).

In summary, our data reveal distinct tissue-specific actions of PR-A and PR-B in ovarian relative to extensively studied breast cancer models. Our observations have important clinical implications with regard to the development of isoform-specific and/or tissue-selective progestins for endocrine therapies. The ability of PR-A to inhibit hyperplastic proliferation in the uterus, together with reduced proliferative activity when PR-A is activated in ovarian and mammary gland tissue strongly indicates that targeted activation of PR-A with isoform-specific agonists will likely have a protective effect against uterine, breast, and ovarian carcinogenesis. Conversely, it may be desirable to inhibit the actions of PR-B, particularly when FOXO1 is inactivated in the context of heightened kinase signaling that is a hallmark of hormone-driven cancers. The quantification of PR isoforms and their key cofactors and target genes (i.e., FOXO1, p21) rather than total PR levels will be essential to improving the efficacy of disease management by isoform-specific selective PR modulators (SPRM).

No potential conflicts of interest were disclosed.

Conception and design: C.H. Diep, C.A. Lange

Development of methodology: C.H. Diep, C.A. Lange

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.H. Diep, C.A. Lange

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.H. Diep, T.P. Knutson, C.A. Lange

Writing, review, and/or revision of the manuscript: C.H. Diep, T.P. Knutson, C.A. Lange

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.H. Diep, C.A. Lange

Study supervision: C.A. Lange

The authors thank members of the University of Minnesota Masonic Cancer Center's Flow Cytometry Core Facility, the University Imaging Centers, the Minnesota Supercomputing Institute, and the University of Minnesota Genomics Center for their assistance in data acquisition, processing, and analysis. CDB-4124 (Proellex) was a kind gift from Ronald Wiehle (Repros Therapeutics).

This study was supported by NIH grant R01 CA159712 and R01 CA159712-S1 (to C.A. Lange), Cancer Biology Training Grant NIH T32 CA009138 (to C.H. Diep), National Center for Advancing Translational Sciences of the NIH Award UL1TR000114 (to C.H. Diep), and a University of Minnesota Doctoral Dissertation Fellowship (to C.H. Diep).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.