Active Estrogen Receptor-Alpha Signaling in Ovarian Cancer Models and Clinical Specimens

Author Manuscript Published OnlineFirst on January 10, 2017; DOI: 10.1158/1078-0432.CCR-16-1501 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Active estrogen receptor-alpha signaling in ovarian cancer models and clinical specimens

Courtney L. Andersen1,2,3,12, Matthew J. Sikora1,3, Michelle M. Boisen3,4, Tianzhou Ma5, Alec Christie3,

George Tseng5, Yongseok Park5, Soumya Luthra6, Uma Chandran6, Paul Haluska7, Gina M. Mantia-

Smaldone8, Kunle Odunsi9, Karen McLean10, Adrian V. Lee1,3, Esther Elishaev11, Robert P. Edwards4, Steffi

Oesterreich1,2,3

Affiliations: 1Department of Pharmacology & Chemical Biology, University of Pittsburgh; 2Molecular

Pharmacology Training Program, University of Pittsburgh; 3Women’s Cancer Research Center, University of

Pittsburgh Cancer Institute; 4Department of Obstetrics, Gynecology, & Reproductive Sciences, Magee-

Womens Hospital of UPMC; 5Department of Biostatistics, University of Pittsburgh; 6Department of

Biomedical Informatics, University of Pittsburgh; 7Oncology, Merck Research Laboratories; 8Department of

Surgical Oncology, Fox Chase Cancer Center; 9Department of Gynecologic Oncology, Roswell Park Cancer

Institute; 10Division of Gynecologic Oncology, University of Michigan; 11Department of Pathology, Magee-

Womens Hospital of UPMC; 12Present address: IMED Oncology, AstraZeneca

Corresponding Author: Steffi Oesterreich, PhD.: B411 Womens Cancer Research Center, University of

Pittsburgh Cancer Institute, Magee-Womens Research Institute, 204 Craft Ave, Pittsburgh, PA 15213. Tel:

(412)641-8555, Fax: (412)641-2458, [email protected]

Keywords: ovarian cancer, high-grade serous cancer, estrogen receptor-alpha, endocrine therapy, tumor explants, ultra-low attachment, fulvestrant, predictive biomarkers

Microarray data (GEO): GSE81612

COI: Dr. Andersen is now an employee of AstraZeneca Pharmaceuticals.

Abbreviations: HGSOC, high-grade serous ovarian cancer; ERα, estrogen receptor-alpha; PDX, patient-

derived xenograft; 4OHT, 4-hydroxytamoxifen; AI, aromatase inhibitor

Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2017 American Association for Cancer Research. Author Manuscript Published OnlineFirst on January 10, 2017; DOI: 10.1158/1078-0432.CCR-16-1501 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Translational Relevance

High-grade serous ovarian cancer (HGSOC) is a malignancy with extremely poor prognosis and limited therapeutic options. Targeting estrogen receptor-alpha has shown promise in laboratory models and in clinical trials but identification of the appropriate patient subset has remained elusive. We characterized endocrine response in cell line and patient-derived HGSOC models to identify features associated with estrogen-responsive HGSOC. In these studies, we observed that a subset of HGSOC models require estrogen for growth and survival. Further, we identified genes (e.g. the ER-alpha target IGBP3) which were associated with clinical endocrine response. We also determined that fulvestrant may be more effective than tamoxifen at blocking cell proliferation in HGSOC. Our data may enable the identification of patients with ovarian cancer who would benefit from endocrine therapy.

Abstract

Purpose: High-grade serous ovarian cancer (HGSOC) is an aggressive disease with few available targeted

therapies. Despite high expression of estrogen receptor-alpha in ~80% of HGSOC and some small but

promising clinical trials of endocrine therapy, estrogen receptor-alpha has been understudied as a target in

this disease. We sought to identify hormone-responsive, estrogen receptor-alpha-dependent HGSOC.

Experimental Design: We characterized endocrine response in HGSOC cells across culture conditions (2-

D, 3-D, forced suspension) and in patient-derived xenograft (PDX) explants, assessing proliferation and gene expression. Estrogen-regulated transcriptome data were overlapped with public datasets to develop a comprehensive panel of estrogen receptor-alpha target genes. Expression of this panel and estrogen receptor-alpha H-score were assessed in HGSOC samples from patients who received endocrine therapy.

Time on endocrine therapy was used as a surrogate for clinical response.

Results: Proliferation is estrogen receptor-alpha-regulated in HGSOC cells in vitro and in vivo, and is partly dependent on 3-D context. Transcriptomic studies identified genes shared by cell lines and PDX explants as estrogen receptor-alpha targets. The selective estrogen receptor-alpha down-regulator (SERD)

3 fulvestrant is more effective than tamoxifen in blocking estrogen receptor-alpha action. Estrogen receptor- alpha H-score is predictive of efficacy of endocrine therapy, and this prediction is further improved by inclusion of target gene expression, particularly IGFBP3.

Conclusions: Laboratory models corroborate intertumor heterogeneity of endocrine response in HGSOC but identify features associated with functional ERα and endocrine responsiveness. Assessing ERα function

(e.g. IGFBP3 expression) in conjunction with H-score may help select patients who would benefit from endocrine therapy. Preclinical data suggest that SERDs might be more effective than tamoxifen.

1 Introduction

2 High-grade serous ovarian cancer (HGSOC) is an aggressive and often lethal disease with limited

3 options for therapy. HGSOC typically responds to surgical debulking and platinum-based chemotherapy as

4 first-line treatment but the majority of patients relapse and ultimately succumb to the disease (1). Identifying

5 targeted, individualized treatment strategies for ovarian cancer will be essential for improving patient

6 survival.

7 One promising but understudied therapeutic target for HGSOC is estrogen receptor-alpha (ERα).

8 ERα is expressed in ~80% of HGSOC (2–4), and estrogen exposure (e.g. oral contraceptive use, hormone

9 replacement therapy) affects the risk of ovarian cancer (4–6). Preclinical studies have shown that estrogen

10 can promote proliferation and migration of HGSOC cell lines and mouse models and in part these effects

11 are blocked by antiestrogens (7–12).

12 Several clinical trials have evaluated endocrine therapy in ovarian cancer. Trials were small (n=14-

13 105 patients), patients were heavily pre-treated with chemotherapy, and ERα status was infrequently used

14 as an inclusion criterion (4). Nevertheless, in each trial a subset of patients benefited from tamoxifen (~20%

15 of patients (13–20)), aromatase inhibitors (~17% (21–24)), or fulvestrant (~40% (25)). Though consistent

16 inclusion of ERα-status may improve response rates, superior biomarkers for ERα function and endocrine

17 responsiveness are needed.

18 We sought to identify HGSOC likely to be ERα-dependent and endocrine responsive. Therefore, we

19 comprehensively characterized estrogen and antiestrogen response with regards to growth, survival, and

20 gene expression in HGSOC cell lines and patient-derived xenograft (PDX) explants. Based on those data,

21 we built an assay for endocrine response and profiled tumors from patients with ovarian cancer who

22 received endocrine therapy to identify genes associated with clinical response. Here we show that ERα H-

23 score with expression of other biomarkers (e.g. IGFBP3) can identify patients with HGSOC who benefit from

24 endocrine therapy.

25 Materials and Methods

26 Antibodies and reagents

27 Chemicals: estradiol (E2) (Sigma-Aldrich, St. Louis, MO); 4-hydroxytamoxifen (4OHT) (Sigma-

28 Aldrich); ICI 182,780 (fulvestrant) (Tocris Bioscience, Bristol, UK); staurosporine (STS) (Tocris Bioscience);

29 and Z-VAD (Tocris Bioscience). E2, 4OHT, and fulvestrant were solubilized in 200-proof ethanol prior to

30 use. STS and Z-VAD were solubilized in sterile DMSO.

31 Antibodies: ERα 6F11 clone (Leica Biosystems, Buffalo Grove, IL), ER (SP1 Clone, Biocare

32 Medical, Concord, CA), BrdU (Bu20a clone, Cell Signaling Technology, Danvers, MA), Ki67 (M1B clone,

33 Dako, Carpinteria, CA), Tubulin (Sigma-Aldrich), beta-actin (Sigma-Aldrich).

35 Cell lines and culture conditions

36 PEO1, PEO4, OVSAHO, and MCF-7 cells were maintained in DMEM (Invitrogen) + 10% fetal bovine

37 serum (FBS) (Gibco). OVCA432 cells were maintained in RPMI-1640 + 10% FBS. Cell line identity was

38 verified by short tandem repeat (STR) profiling. PEO1 and PEO4 cells were derived from the same patient

39 (26), PEO1 after her first recurrence and PEO4 after the tumor became platinum-resistant.

40 Hormone deprivation was performed as described previously (27) using IMEM + 10% charcoal-stripped

41 serum (CSS) for PEO1, PEO4, and OVCA432 or 5% CSS for MCF-7 and OVSAHO. Unless otherwise

42 specified, hormones were used at the following concentrations: 1 nM E2, 1 µM fulvestrant, and 1 µM 4OHT.

43 For standard (2-D) proliferation assays, growth was analyzed using the FluoReporter dsDNA

44 quantitation kit (Molecular Probes, Eugene, OR) as previously described (27). Cells (2000-4000 / well) were

45 seeded in 96-well plates (Fisher). After cells adhered (16-24 hrs), drug was added directly to the wells. For

46 ultra-low attachment (ULA) assays, cells (5000-10000 / well) were treated at the time of seeding in ULA

47 (Corning). Viability was measured using the CellTiter-Glo assay and apoptosis with the CaspaseGlo-3/7

48 assay (Promega, Madison, WI). For 3-D assays, dishes were coated with phenol red-free matrigel (BD

49 Biosciences, Erembodegem, Belgium). Cells were seeded on top of the matrigel in media + 2% matrigel.

51 Cell line gene expression analyses

52 Hormone-deprived cells were treated in biological quadruplicate with vehicle, E2, 4OHT +/- E2, or

53 fulvestrant +/- E2 as previously described (27). RNA was isolated using the Illustra RNAspin Mini kit (GE

54 Healthcare, Buckinghamshire, UK). Gene expression was measured on Affymetrix U133A 2.0 arrays. Data

55 were RMA normalized using the affy() package in R (command “rma()”, http://www.bioconductor.org/).

56 Differentially expressed genes were identified with limma. When genes were represented by multiple

57 probes, the probe with greatest variation (largest dynamic range) across samples was chosen for

58 downstream analysis. Heat maps were generated using the Multiple Experiment Viewer (MeV,

59 http://www.tm4.org/mev.html). E2-regulated genes were considered “blocked” by fulvestrant or 4OHT if E2

60 produced significant (p<0.001) changes in expression compared to vehicle but fulvestrant+E2 or 4OHT+E2

61 did not. To determine overlap with E2-regulated genes in breast cancer, MCF-7 data was obtained from the

62 GEMS database. Significantly E2-regulated genes were defined as those with q<0.05. Because the

63 treatment for our microarray studies was 3 hours, we used only the “early” (3-4 hour E2 treatment) GEMS

64 data set.

65 cDNA conversion and qRT-PCR were performed using iScript and Universal SYBR RT Supermix

66 (Bio-Rad, Hercules, CA). Primer sequences are in Table 1.

68 Immunoblotting

69 Cells were lysed in RIPA buffer + Halt Phosphatase/Protease Inhibitor Cocktail (Pierce). Protein (20

70 µg) was run on a 10% SDS polyacrylamide gel and transferred to a PVDF membrane. Membranes were

71 incubated with primary antibody overnight at 4°C. Blots were imaged on the Olympus LiCor system.

72 Antibody dilutions were as follows: ER 6F11, 1:500; Tubulin, 1:10000.

74 Xenograft studies

75 All animal studies were approved by the University of Pittsburgh Institutional Animal Care and Use

76 Committee. C.B.17/IcrHsd-PrkdcscidLystbg-J (SCID/Beige, Harlan) mice were used for all studies. For PEO4

77 xenografts, mice underwent ovariectomy followed by sub-cutaneous pellet implantation (placebo or 0.03 mg

78 E2, Innovative Research of America). Each group had n=5 mice. Two weeks after surgery, 106 PEO4 cells

79 in 1:1 RPMI + matrigel were injected intraperitoneally (IP). Mice were monitored for 11 weeks after injection

80 and then sacrificed. Tumors were harvested immediately after euthanasia. Tissue was either flash frozen in

81 liquid nitrogen or fixed in 10% NBF. Patient-derived xenografts (PDXs) were provided by Dr. Paul Haluska

82 and processed as previously described (28). Tissue was collected at necropsy in the same manner as the

83 PEO4 xenografts.

85 Explants

86 PDXs were passaged <5 times prior to use. PDXs were cultured ex vivo using an established

87 protocol for primary tumors (29). Briefly, fresh PDX tissue harvested and dissected into ~1 mm3 fragments.

88 Fragments were cultured on VetSPON sponges (Henry Schein, Dublin, OH) partially submerged in media

89 (IMEM + 5% FBS + 10 μg/mL insulin + 10 μg/mL hydrocortisone) + vehicle, fulvestrant, or 4OHT. Three

90 tissue fragments were placed on each sponge. After three days, explants were pulsed with 30 μg/mL BrdU

91 (Invitrogen, Carlsbad, CA) for 4-6 hours. Tissue was either fixed in 10% neutral-buffered formalin (NBF) for

92 immunohistochemistry (IHC) or snap-frozen in liquid nitrogen. Snap-frozen tissue was processed using the

93 RNEasy Mini Kit (Qiagen, Hilden, Germany). qRT-PCR was performed as described above. IHC details are

94 provided below. For PH045, PH053, and PH070, two sponges per treatment group (six explant pieces total)

95 were used for each assay type (i.e. two sponges for RNA collection and two for fixation/IHC). For PH242,

96 only one sponge (three fragments) was used per treatment group although several time points (day 1-3)

97 were assessed. Each tissue fragment was treated as a unique biological replicate.

99 Immunohistochemistry

100 Antigen retrieval was performed in citrate buffer (pH 6) for 30 min using a boiling water bath.

101 Sections were then blocked in 5% BSA in PBS + 0.5% Tween-20 for 1 hour at room temperature. Sections

102 were incubated in primary antibody overnight at 4°C. Staining was visualized with DAB. Slides were

103 counterstained with hematoxylin. BrdU staining was quantified by determining %Brdu+ cells (# BrdU+

104 cells/total cells*100%) for a given field of view. For each treatment group, 10 fields of view were counted,

105 spanning multiple explant pieces.

106 Immunohistochemistry of clinical samples was performed by the research histology core at UPMC.

107 Antigen retrieval was performed in citrate buffer (pH 6) at 120°C. Staining was detected using Envision Dual

108 Link+ HRP Polymer and DAB (Dako). Hematoxylin was used for counterstaining. Antibody dilutions were as

109 follows: ER 6F11, 1:50; BrdU, 1:200; Ki67, 1:300, ER SP-1, pre-dilute.

110

111 Clinical samples

112 Paraffin-embedded tumor samples from the University of Pittsburgh Medical Center (UPMC), Fox

113 Chase Cancer Center (FCCC), Roswell Park Cancer Institute (RPCI), and the University of Michigan (Mich)

114 were centrally reviewed by a pathologist (E.E.) to confirm >50% tumor and >50% viable cells. RNA was

115 isolated using the AllPrep FFPE kit (Qiagen). Expression of genes on the EndoRx panel (see methods,

116 Supplementary File 1) was measured on the NanoString nCounter as previously described (30). Data were

117 normalized to internal controls and the geometric mean of four housekeeping genes. All work was approved

118 by local Institutional Review Boards.

119

120 Design of the EndoRx assay

121 To develop a comprehensive assay for estrogen response, we overlapped our microarray results

122 with publically available preclinical studies of E2 response in breast, bone, ovarian, and endometrial cancer,

123 and with genes differentially expressed between ERα-positive and ERα-negative breast and ovarian tumors

124 and genes specific to “hormonally responsive” endometrial tumors from The Cancer Genome Atlas (TCGA)

125 (31–33). Ad hoc additions were made including mediators of ERα signaling (e.g. NCOAs), genes with

126 known associations to endocrine resistance, immune response, or tumor-stromal interactions, and genes

127 that correlated with response in clinical trials of endocrine therapy (34,35).The final assay comprised 350

128 genes (Supplementary File 1).

129

130 Statistical methods

131 Significance was determined at p=0.05 unless otherwise specified. Unpaired, two-tailed t-tests were

132 used to compare two groups. For three or more groups, one-way ANOVA and Tukey’s post-hoc test were

133 used. Growth curves in ULA were fit with simple logarithmic regression (log(y)=log(y0)+k*x) and k was

134 compared between groups using sum-of-squares F-test. Chi-square test was used to compute significance

135 of overlap in E2-regulated genes between cell lines. On figures, asterisks indicate significance: ****,

136 p<0.0001; ***, p<0.001; **, p<0.01; *, p<0.05.

137 Clinical specimens were dichotomized by time on endocrine therapy: “long” (>120 days, n=43) and

138 “short” (<120 days. n=25). Differentially expressed genes were identified with edgeR; significance was

139 determined using a likelihood ratio test and Benjamini-Hochberg correction for multiple comparisons. To

140 construct a prediction model for classifying patients by time on endocrine therapy, support vector machines

141 (SVM) with linear kernel was used. SVM is a supervised machine learning algorithm used to solve

142 classification problems. It is a generalization of the maximal margin classifier. Given a separation of the

143 hyperplane and when data are separable, the maximal margin is defined as the minimum distance of the

144 objects to the hyperplane. In addition, we applied an SVM-RFE (Recursive Feature Extraction) to return a

145 ranking of the features of our classification problem by training an SVM with a linear kernel and removing

146 the feature with the smallest ranking criterion. Model accuracy was assessed through leave-one-out cross-

147 validation. Log-rank test was used to determine significance level of the survival curves. ANOVA was used

148 to test differences of cohort means from major clinical variables. Differences in pre- and post-endocrine

149 therapy CA-125 levels were determined by paired t-test.

150

151 Results

152 Endocrine response in HGSOC cell lines

153 To determine if estrogen regulates growth, we evaluated ERα expression and response to E2,

154 fulvestrant, and 4OHT in four ERα+ HGSOC cell lines. PEO1, PEO4, and OVCA432 cells expressed high

155 ERα but expression in OVSAHO cells was lower (Figure 1A); all lines were ERβ-negative (data not shown).

156 In 2-D assays, E2 stimulated proliferation of PEO4 and PEO1 cells in a dose-dependent manner, which was

157 abrogated by fulvestrant and 4OHT. In contrast, E2 had no effect on growth of OVCA432 and OVSAHO

158 cells (Figure 1B).

159 Since ERα status alone did not predict E2 response in HGSOC cells, we examined markers of

160 ERα function. Previous efforts to identify ERα target genes in HGSOC are limited (7,9). To create a

161 comprehensive picture of the ERα transcriptome, we performed whole-genome microarrays in PEO4 and

162 PEO1 cells after treatment with E2 +/- fulvestrant or 4OHT. E2 regulated the expression of 221 and 291

163 genes in PEO1 and PEO4, respectively (Figure 1C, Supplementary File 2). Notably, fulvestrant was more

164 effective than 4OHT at blocking E2 effects; 4OHT mitigated E2-mediated expression of 70% of genes in

165 PEO4 cells and 89% in PEO1 cells, whereas fulvestrant mitigated 96% and 99.5%.

166 There was significant overlap among E2-regulated genes in HGSOC lines (n=175/515, p<0.001)

167 versus MCF-7 cells (36) (Figure 1D), including canonical ERα targets GREB1, CCNG2, and MYC. We

168 validated E2-regulation (and blockade by fulvestrant and 4OHT) of GREB1, CCNG2, and MYC by qRT-PCR

169 in PEO1 and PEO4 cells (Figure 1E). This suggests ERα targets in HGSOC cells comprise a largely

170 classical E2 response.

171 We then evaluated endocrine response of HGSOC cells lines in models mimicking in vivo tumor

172 growth, first via 3-D growth in a matrigel. Similar to growth in 2-D, E2 increased spheroid formation in PEO4

173 cells (Figure 2A) but not in OVCA432 cells. To model ascites, a common clinical manifestation in late-stage

174 HGSOC, we grew cells in forced suspension using ultra-low attachment (ULA) plastics. Cells in ULA grow in

175 aggregates (Supplementary Figure 1A). E2 significantly increased PEO1 and PEO4 cell number; OVSAHO

176 and OVCA432 cells gained E2-responsiveness, as E2 now increased cell number (Figure 2B). ULA

177 increased ERα mRNA and protein levels versus 2-D conditions (Fig 2C-D), which may mediate the novel E2

178 response. The effect of E2 on PEO1 appeared to be more through a decrease of cell death than increased

179 proliferation. Survival in ULA/forced suspension typically requires resistance to anoikis (apoptosis due to

180 detachment) but we observed no effect of E2 on caspase-3/7 activity (Supplementary Figure 1B),

181 suggesting E2 did not inhibit anoikis but may mediate other survival mechanisms. We then sought to

182 directly assess estrogen response in vivo. Because PEO4 displayed hormone-dependence across all three

183 culture methods, we chose this cell line for our xenograft experiments. E2 treatment increased tumor

184 burden versus placebo (Figure 2E) and induced GREB1 and MYC expression, consistent with our in vitro

185 results. Taken together, these data demonstrate that some HGSOC cells are E2-responsive but that

186 response may be dependent on three-dimensional context.

187

188 Antiestrogen response in HGSOC explants

189 To examine antiestrogen response in models more closely mimicking clinical HGSOC, we utilized

190 four PDXs that recapitulate classic ovarian cancer phenotypes. They were established from HGSOC

191 tumors of advanced stage (III/IV) and grade (III/IV). Our models also capture the varying ERα expression of

192 HGSOC - encompassing high (PH045), medium (PH053, PH242), and low expression (PH070) (Figure 3A);

193 none expressed ERβ (data not shown). However, HGSOC PDXs often take months to form detectable

194 tumors (28,37), limiting the feasibility of large-scale in vivo studies. We therefore assessed proliferation

195 (BrdU incorporation) and gene expression in PDX explants (29) following endocrine treatment (Fig. 3B).

196 Treatment of PH045 explants with 4OHT significantly decreased proliferation (Figure 3C-D)

197 (p=0.014, median change of -50%); 4OHT also decreased GREB1 and PGR expression and induced the

198 E2-repressed gene CCNG2 (Figure 3E). 4OHT did not affect MYC expression as was observed in cell lines

199 (data not shown). Fulvestrant produced similar but more pronounced effects; the median change in

200 proliferation was -50% (p=0.005) but the maximal decrease was greater (-95% vs. -58%). Fulvestrant

201 decreased ERα protein (mean change -55%), consistent with its mechanism of action, and produced

202 stronger effects on gene expression versus 4OHT. Similar observations were made in PH053 explants

203 (Figure 3F-H); fulvestrant significantly decreased proliferation (p=0.0049) while effects of 4OHT on

204 proliferation and gene expression were weaker. In contrast, fulvestrant did not affect PH070 and PH242

205 explants (Supplementary Figure 2). Lack of fulvestrant response was expected for PH070 (low/absent ERα)

206 but PH242 had ERα levels similar to PH053. The fulvestrant resistance in PH242 thus supports clinical

207 observations that ERα itself is limited as a biomarker of endocrine response in HGSOC (25,34,38,39).

208 Therefore, though some HGSOC PDXs are endocrine responsive, additional biomarkers are necessary to

209 differentiate response versus resistance.

210

211 Genes associated with clinical endocrine response

212 To identify novel biomarkers of endocrine response, tumor specimens were procured from 70

213 patients with ovarian cancer who received tamoxifen and/or an AI at four medical centers (Table 2). Median

214 age at diagnosis was 63 and the majority (85%) of patients presented at late stages (III/IV). No patient or

215 tumor characteristics differed across centers except age; RPCI patients were significantly older (p=0.0013).

216 There were also no significant differences in overall survival (time from diagnosis to death) or duration of

217 endocrine therapy across cohorts (Supplementary Figure 3).

218 Endocrine therapy was given after chemotherapy, often in the setting of recurrent disease (Figure

219 4A). While data on disease progression after endocrine therapy were limited, we have single values post-

220 treatment CA-125 measurements for 45/70 patients and these were significantly higher than pre-treatment

221 levels (p=0.031, Supplementary File 3). However, in light of the heterogeneity in our patient population

222 (patients were placed on endocrine therapy at various points in their disease course and treated at different

223 institutions) and to also capture patients who achieved disease stability, we chose to use time on endocrine

224 therapy as a surrogate for clinical response.

225 ERα status (positive vs. negative) was not associated with response (data not shown), so we

226 examined whether ERα levels as a continuous variable or biomarkers of ERα function were predictive of

227 time on endocrine therapy in our cohort. ERα H-scores ranged from 0-270 (median=60, Figure 4B). Patients

228 with an H-score >60 stayed on endocrine therapy longer, suggesting a higher likelihood of response

229 (p=0.002, Figure 4C). In parallel, we evaluated expression of ERα target genes (EndoRx panel,

230 Supplementary Figure 4A). This gene set (n=350) includes ERα targets identified using public data on E2

231 treatment in hormone-responsive cancers (n=207) and genes with known roles in antiestrogen resistance

232 and ERα function (n=77) or the tumor microenvironment (n=66). The ERα target subset was validated in

233 silico using public data from breast tumors (METABRIC (40) and Van’t Veer (41)), where it distinguished

234 ERα-positive and -negative disease (Supplementary Figure 5A), and in vitro using PEO4 and MCF-7 cells

235 (Supplementary Figure 5B).

236 We compared expression of the EndoRx panel between patients with a long (>4 months) or short

237 (<4 months) duration of endocrine therapy. We specifically chose 4 months for a cut-off to allow adequate

238 time to achieve disease stability or response with oral hormonal agents, which are often slower to affect

239 tumor growth than cytotoxic therapies. Among the top 11 genes significantly associated >4 months

240 endocrine therapy were ESR1 (ERα, p=0.0005) and ERα targets IGFBP3 (p=1,5x10-4), PGR (p=0.008), and

241 MYC (p=0.0055). However, only IGFBP3 was significant after correction for multiple comparisons (q=0.026)

242 (Figure 4D). IGFBP3 was previously described as E2-repressed in ovarian cancer (42) and we observed

243 similar ERα-regulation in our hormone-responsive models (Figure 4E). We did not see ERα-regulation of

244 IGFBP3 expression in OVCA432 cells (Supplementary Figure 5). Using log-rank comparisons, the p-value

245 for the association with IGFBP3 and endocrine therapy was better than that of H-score (p=0.00002 vs

246 p=0.002, Figure 4F). We then designated tumors as ERαhigh or ERαlow (H-score >60 or<60, respectively)

247 and IGFBP3high or IGFBP3low (above or below third-quartile expression). The majority (30/33) of ERαhigh

248 tumors were IGFBP3low (i.e. with ERα actively suppressing IGFBP3) and patients with these tumors

249 remained on endocrine therapy longer than the rest of the cohort. Strikingly, patients with ERαlow/IGFBP3high

250 tumors (i.e. no functional ERα) benefited less from endocrine therapy than their ERαlow/IGFBP3low

251 counterparts (Figure 4G, p=0.023). Patients with ERαlow/IGFBP3low tumors had outcomes comparable to

252 patients with ERαhigh/IGFBP3low tumors, suggesting some ERαlow tumors retain active ERα signaling.

253 In light of these data, we asked if combining H-score with an aggregation of genes would provide

254 stronger predictive power of time on endocrine therapy than a single gene. We used an SVM algorithm to

255 identify features (genes and/or H-score) associated with duration of endocrine therapy. The top 30 features

256 were significantly (p<2x10-16) associated with endocrine response (Figure 4H), including H-score and

257 IGFBP3 (Table 3). The majority of features were known ERα targets, suggesting that assessing ERα

258 function might be a stronger predictor of time on endocrine treatment than ERα alone.

259

260 Discussion 261 Clinical trials of endocrine therapy suggest a subset of patients with ovarian cancer benefit from

262 endocrine therapy. However, unlike for breast cancer, tumor ERα status (positive vs. negative) is not

263 sufficient to predict response in HGSOC. Further, linking ERα immunoscore with endocrine response in

264 ovarian cancer has produced mixed results (13,34,38). Implementing biomarkers that complement ERα will

265 be critical to identifying appropriate patient populations for endocrine therapy.

266 Three studies have previously tried to identify biomarkers of endocrine response (34,35,43), all

267 utilizing a small panel of IHC markers. We pursued a comprehensive profile of potential biomarkers by

268 designing and evaluating the EndoRx panel. Our analysis indicated lower IGFBP3 was significantly

269 associated with prolonged duration of endocrine therapy (Figure 4), corroborating previous reports that

270 IGFBP3 expression correlates with response to therapy (43,44). IGFBP3 expression was more strongly

271 associated with time on therapy than H-score (Figure 4). Moreover, low IGFBP3 identified a subgroup of

272 ERαlow patients who received longer endocrine therapy. Given that IGFBP3 is ERα-suppressed (Figure 4)

273 (35), our results suggest a direct output of ERα function better designates endocrine responsiveness than

274 ERα itself. Further supporting this is the SVM analysis; while H-score was among the top features, it ranked

275 29th, indicating 28 genes, including ERα targets, carried stronger associations with endocrine response.

276 Excepting IGFBP3, we did not find strong associations between previously reported biomarkers of

277 endocrine response and outcomes in our cohort (34,35,43). This could be attributable to different

278 methodology (gene expression vs. IHC) or cohort size (ours is the largest to date). Though vimentin was

279 shown to be associated with fulvestrant response (34), difference in therapy may account for this

280 discrepancy; independent classes of endocrine therapy may require different predictive markers.

281 Our EndoRx panel was designed based on studies in models that recapitulate the varying endocrine

282 response seen in clinical HGSOC. Consistent with previous reports that E2 promotes ovarian cancer cell

283 growth (7,8,11), PEO1 and PEO4 cells were endocrine responsive in 2-D, 3-D, and ULA cultures. While

284 PEO1 and PEO4 cells were isolated from the same patient (9), they exhibit different E2-response

285 phenotypes. We believe that this is attributable to differences in cell line biology s the lines were isolated at

286 different points in the patient’s disease course. The PEO1 cells were isolated from her first recurrence

287 whereas the PEO4 were isolated from a later recurrence after the tumor became platinum-refractory.

288 In contrast to the E2-responsive PEO1 and PEO4 cells, OVCA432 and OVSAHO were E2-

289 independent in 2-D and 3-D but became E2-responsive in ULA. None of our models expressed ERβ,

290 suggesting these effects are ERα-mediated and that ERα has unique roles throughout HGSOC

291 progression. Elevated ERα protein in ULA supports the association between ERα levels and clinical

292 endocrine response. Moreover, this alludes to a functional link between ERα protein levels and

293 transcriptional activity. It is, however, also possible that a different E2-binding receptor is mediating these

294 effects. Assessing the potential of other receptors as well as the tie between ERα and ERα protein levels is

295 an important direction for future investigation as it may provide insight into the efficacy of fulvestrant vs.

296 4OHT. This also emphasizes the necessity of translational models such as PDXs and explants to fully

297 understand ERα action in HGSOC.

298 We provide the first evidence of endocrine response in patient-derived HGSOC models. Fulvestrant

299 produced stronger effects on explant proliferation and gene expression than 4OHT, suggesting modality of

300 endocrine therapy will be an important consideration in HGSOC treatment. Selective ERα modulators (e.g.

301 4OHT) exhibit partial agonism in certain tissues and cancers (27) whereas selective ER degraders (e.g.

302 fulvestrant) are pure antagonists. Potential tamoxifen agonism in HGSOC has not been explored but

303 tamoxifen exposure was reported to promote fallopian tube and ovarian lesions (45,46). Further

304 comparisons of fulvestrant and 4OHT with other antiestrogens will be necessary to understand any

305 differential class effects in HGSOC.

306 Our explant studies also suggest heterogeneous endocrine response across regions of HGSOC

307 tumors: 4OHT and fulvestrant response varied in terms of both proliferation and gene expression between

308 explants from the same PDX (Figure 3). It is possible that interactions between different regions would

309 facilitate response of the bulk tumor. However, strategies for combination therapy should also be

310 considered. Two such possibilities are MAPK and Src, which are known to crosstalk with ERα and drive

311 endocrine resistance in ovarian cancer (47,48). Co-targeting PGR is also promising given its interaction with

312 ERα (49) and recent reports demonstrating PGR agonists induce senescence in ovarian cancer cells (50).

313 We focused our study on HGSOC as it is the most common clinical subtype of ovarian cancer.

314 However, there is likely potential for endocrine therapy in endometrioid ovarian cancer as well, which is also

315 frequently ERα-positive. Evaluating the role of ERα and IGFBP3 in this subtype is an important area for

316 future investigation.

317 Our clinical analysis is somewhat limited by its retrospective nature. Modality of endocrine therapy

318 and number of previous therapies vary across patients and inconsistent post-treatment data necessitate the

319 use of surrogates for clinical responsiveness. Prospective studies with post-treatment specimen collection,

320 standardized timing of endocrine therapy, and sufficient power to compare different endocrine agents will be

321 necessary to solidify the utility of any biomarkers. Additionally, the mechanistic role of these markers (e.g.

322 IGFBP3) in ovarian cancer should be followed up in preclinical studies.

323 In summary, ERα modulates growth, survival, and gene expression in a subset of HGSOC and

324 targeting ERα can be effective clinically. Inhibiting ERα with fulvestrant and 4OHT modulates expression of

325 MYC, PGR, and IGFBP3 in HGSOC models. Moreover, expression of these genes reflects clinical

326 endocrine response. Our findings may enable selection of HGSOC patients who would benefit from

327 endocrine therapy.

328

329 Acknowledgements

330 We thank Dr. Marc Becker for guidance on establishing PDX models and Dr. Brian Szender for

331 assistance procuring RPCI specimens.

332 This work was supported by the National Institutes of Health (F31CA186736 to CLA, T32GM008424

333 to CLA, K99CA193734 to MJS, and R01CA184502 to PH), the Magee-Womens Research Foundation (SO),

334 the University of Pittsburgh Cancer Institute (UPCI), and the ARCS Foundation (CLA). Additional support

335 was provided by the Department of Defense (CDMRP W81XWH-13-1-0205 to SO, Ovarian Cancer

336 Academy Early Career Investigator Award W81XWH-15-0194 to KM), the RPCI-UPCI SPORE

337 (5P50CA159981-03 to RPE, KO), and the Mayo Clinic SPORE (2P50CA136393 to PH). This project used

338 the UPCI Cancer Biomarkers Core Facility, which is supported by NIH P30CA047904.

References 1. Vaughan S, Coward JI, Bast RC, Berchuck A, Berek JS, Brenton JD, et al. Rethinking ovarian cancer: recommendations for improving clinical outcomes. Nat Rev Cancer 2011;11. 2. Sieh W, Köbel M, Longacre TA, Bowtell DD, deFazio A, Goodman MT, et al. Hormone-receptor expression and ovarian cancer survival: an Ovarian Tumor Tissue Analysis consortium study. Lancet Oncol 2013;14:853-62. 3. Lau KM, Mok SC, Ho SM. Expression of human estrogen receptor-alpha and -beta, progesterone receptor, and androgen receptor mRNA in normal and malignant ovarian epithelial cells. Proc Natl Acad Sci 1999;96:5722–7. 4. Modugno F, Laskey RA, Smith AL, Andersen CL, Haluska P, Oesterreich S. Hormone response in ovarian cancer: time to reconsider as a clinical target? Endocr Relat Cancer 2012;19:R255-79. 5. Group C, Cancer O, Beral V, Gaitskell K, Hermon C, Moser K, et al. Menopausal hormone use and ovarian cancer risk: individual participant meta-analysis of 52 epidemiological studies. Lancet 2015;6736:1835–42. 6. Narod SA, Risch H, Moslehi R, Dorum A, Neuhausen S, Olsson H, et al. Oral contraceptives and risk of hereidtary ovarian cancer. N Engl J Med 1998;389:424–8. 7. Spillman MA, Manning NG, Dye WW, Sartorius CA, Post MD, Harrell JC, et al. Tissue-specific pathways for estrogen regulation of ovarian cancer growth and metastasis. Cancer Res 2010;70:8927–36. 8. Langdon SP, Hawkes MM, Lawrie SS, Crew AJ, Miller WR, Smyth JF. Oestrogen receptor expression and effects of oestrogen and tamoxifen on the growth of human ovarian carcinoma cell lines. Br J Cancer. 1990;1:213–6. 9. Langdon SP, Hirst GL, Miller EP, Hawkins RA., Tesdale AL, Smyth JF, et al. The regulation of growth and protein expression by estrogen in vitro: A study of 8 human ovarian carcinoma cell lines. J Steroid Biochem Mol Biol. 1994;50:131–5. 10. Moyle-Heyrman G, Schipma MJ, Dean M, Davis DA, Burdette JE. Genome-wide transcriptional regulation of estrogen receptor targets in fallopian tube cells and the role of selective estrogen receptor modulators. J Ovarian Res 2016;9:5. 11. Laviolette LA, Hodgkinson KM, Minhas N, Perez-Iratxeta C, Vanderhyden BC. 17beta-estradiol upregulates GREB1 and accelerates tumor progression in vivo. Int J Cancer. 2014;135:1072-84. 12. Laviolette LA, Garson K, Macdonald EA, Senterman MK, Courville K, Crane CA, et al. 17Beta- Estradiol Accelerates Tumor Onset and Decreases Survival in a Transgenic Mouse Model of Ovarian Cancer. Endocrinology 2010;151:929–38. 13. Hatch KD, Beecham JB, Blessing JA, Creasman WT. Responsiveness of patients with advanced ovarian carcinoma to tamoxifen. A Gynecologic Oncology Group study of second-line therapy in 105 patients. Cancer 1991;68:269–71. 14. Hurteau JA, Brady MF, Darcy KM, McGuire WP, Edmonds P, Pearl ML, et al. Randomized phase III trial of tamoxifen versus thalidomide in women with biochemical-recurrent-only epithelial ovarian, fallopian tube or primary peritoneal carcinoma after a complete response to first-line platinum/taxane chemotherapy with an evaluation . Gynecol Oncol 2010 ;119:444–50. 15. Schwartz PE, Keating G, MacLusky N, Naftolin F, Eisenfeld A. Tamoxifen therapy for advanced ovarian cancer. Obstet Gynecol 1982;59:583–8. 16. Hofstra LS, Mourits MJ, de Vries EG, Mulder NH, Willemse PH. Combined treatment with goserelin and tamoxifen in patients with advanced chemotherapy resistant ovarian cancer. Anticancer Res 1999;19:3627–30. 17. Marth C, Sorheim N, Kaern J, Trope C. Tamoxifen in the Treatment of Recurrent Ovarian Carcinoma. Int J Gynecol Cancer 1997;7:256–61.

18. Tropé C, Marth C, Kaern J. Tamoxifen in the treatment of recurrent ovarian carcinoma. Eur J Cancer 2000;36 Suppl 4:S59-61. 19. Weiner SA, Alberts DS, Surwit EA, Davis J, Grosso D. Tamoxifen therapy in recurrent epithelial ovarian carcinoma. Gynecol Oncol 1987;27:208–13. 20. Tevaarwerk AJ, Wang M, Zhao F, Fetting JH, Cella D, Wagner LI, et al. Phase III Comparison of Tamoxifen Versus Tamoxifen Plus Ovarian Function Suppression in Premenopausal Women With Node-Negative, Hormone Receptor-Positive Breast Cancer (E-3193, INT-0142): A Trial of the Eastern Cooperative Oncology Group. J Clin Oncol 2014;10:3948-58. 21. Kavanagh JJ, Hu W, Fu S, Deavers M, Moore C, Coleman RL, et al. Anti-tumor activity of letrozole in patients with recurrent advanced low malignant potential or low-grade serous ovarian tumors. ASCO Meet Abstr 2007;25:5582. 22. Papadimitriou CA, Markaki S, Siapkaras J, Vlachos G, Efstathiou E, Grimani I, et al. Hormonal therapy with letrozole for relapsed epithelial ovarian cancer. Long-term results of a phase II study. Oncology 2004;66:112–7. 23. Ramirez PT, Schmeler KM, Milam MR, Slomovitz BM, Smith JA, Kavanagh JJ, et al. Efficacy of letrozole in the treatment of recurrent platinum- and taxane-resistant high-grade cancer of the ovary or peritoneum. Gynecol Oncol 2008;110:56–9. 24. Smyth JF, Gourley C, Walker G, MacKean MJ, Stevenson A, Williams ARW, et al. Antiestrogen therapy is active in selected ovarian cancer cases: the use of letrozole in estrogen receptor-positive patients. Clin Cancer Res 2007;13:3617–22. 25. Argenta PA, Thomas SG, Judson PL, Downs LS, Geller MA, Carson LF, et al. A phase II study of fulvestrant in the treatment of multiply-recurrent epithelial ovarian cancer. Gynecol Oncol 2009;113:205–9. 26. Langdon SP, Lawrie SS, Hay FG, Hawkes MM, Mcdonald A, Hayward IP, et al. Characterization and Properties of Nine Human Ovarian Adenocarcinoma Cell Lines Characterization and Properties of Nine Human Ovarian Adenocarcinoma Cell Lines. Cancer Research 1988;6166–72. 27. Sikora MJ, Cooper KL, Bahreini A, Luthra S, Wang G, Chandran UR, et al. Invasive lobular carcinoma cell lines are characterized by unique estrogen-mediated gene expression patterns and altered tamoxifen response. Cancer Res 2014;74:1463-74. 28. Weroha SJ, Becker MA, Enderica-Gonzalez S, Harrington SC, Oberg AL, Maurer MJ, et al. Tumorgrafts as in vivo surrogates for women with ovarian cancer. Clin Cancer Res 2014;20:1288–97. 29. Centenera MM, Gillis JL, Hanson AR, Jindal S, Taylor RA, Risbridger GP, et al. Evidence for efficacy of new Hsp90 inhibitors revealed by ex vivo culture of human prostate tumors. Clin Cancer Res 2012;18:3562–70. 30. Nayak SR, Harrington E, Boone D, Hartmaier R, Chen J, Pathiraja TN, et al. A Role for Histone H2B Variants in Endocrine-Resistant Breast Cancer. Horm Cancer 2015;6:214–24. 31. The Cancer Genome Atlas, Cancer T, Atlas G. Integrated genomic analyses of ovarian carcinoma. Nature 2011;474:609–15. 32. Kandoth C, Schultz N, Cherniack AD, Akbani R, Liu Y, Shen H, et al. Integrated genomic characterization of endometrial carcinoma. Nature 2013;497:67–73. 33. Comprehensive molecular portraits of human breast tumours. Nature 2012;490:61–70 34. Argenta PA, Um I, Kay C, Harrison D, Faratian D, Sueblinvong T, et al. Predicting response to the anti-estrogen fulvestrant in recurrent ovarian cancer. Gynecol Oncol 2013;131:368-73. 35. Walker G, MacLeod K, Williams ARW, Cameron D a, Smyth JF, Langdon SP. Insulin-like growth factor binding proteins IGFBP3, IGFBP4, and IGFBP5 predict endocrine responsiveness in patients with ovarian cancer. Clin Cancer Res 2007;13:1438–44.

36. Ochsner SA, Steffen DL, Hilsenbeck SG, Chen ES, Watkins C, McKenna NJ. GEMS (Gene Expression MetaSignatures), a Web resource for querying meta-analysis of expression microarray datasets: 17beta-estradiol in MCF-7 cells. Cancer Res 2009;69:23–6. 37. Ricci F, Bizzaro F, Cesca M, Guffanti F, Ganzinelli M, Decio A, et al. Patient-derived ovarian tumor xenografts recapitulate human clinicopathology and genetic alterations. Cancer Res 2014;74:2980- 90. 38. Bowman A, Gabra H, Langdon SP, Lessells A, Stewart M, Young A, et al. CA125 Response Is Associated with Estrogen Receptor Expression in a Phase II Trial of Letrozole in Ovarian Cancer : Identification of an Endocrine-sensitive Subgroup CA125 Response Is Associated with Estrogen Receptor Expression in a Phase II Trial of Let. 2002;2233–9. 39. Stasenko M, Plegue M, Sciallis AP, McLean K. Clinical Response to Antiestrogen Therapy in Platinum-Resistant Ovarian Cancer Patients and the Role of Tumor Estrogen Receptor Expression Status. Int J Gynecol Cancer 2014;0:1–7. 40. Curtis C, Shah SP, Chin S-F, Turashvili G, Rueda OM, Dunning MJ, et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012;486:346– 52. 41. van ’t Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AAM, Mao M, et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002;415:530–6. 42. O’Donnell AJM, Macleod KG, Burns DJ, Smyth JF, Langdon SP. Estrogen receptor-alpha mediates gene expression changes and growth response in ovarian cancer cells exposed to estrogen. Endocr Relat Cancer 2005;12:851–66. 43. Walker G, MacLeod K, Williams ARW, Cameron DA, Smyth JF, Langdon SP. Estrogen-regulated gene expression predicts response to endocrine therapy in patients with ovarian cancer. Gynecol Oncol 2007;106:461–8. 44. Luo L-L, Zhao L, Wang Y-X, Tian X-P, Xi M, Shen J-X, et al. Insulin-like growth factor binding protein- 3 is a new predictor of radiosensitivity on esophageal squamous cell carcinoma. Sci Rep 2015;5:17336. 45. Brown AP, Morrissey RL, Crowell JA, Levine BS. Thirteen-week oral toxicity study of difluoromethylornithine in combination with tamoxifen citrate in female dogs. Cancer Chemother Pharmacol 1999;43:479–88. 46. Sonnendecker HE, Cooper K, Kalian KN. Primary fallopian tube adenocarcinoma in situ associated with adjuvant tamoxifen therapy for breast carcinoma. Gynecol Oncol. 1994;52:402–7. 47. Simpkins FA, Hevia-Paez P, Sun J, Ullmer W, Gilbert C, da Silva T, et al. Src inhibition with saracatinib reverses fulvestrant resistance in ER-positive ovarian cancer models in vitro and in vivo. Clin Cancer Res 2012;18:5911-23. 48. Hou JY, Rodriguez-Gabin A, Samaweera L, Hazan R, Goldberg GL, Horwitz SB, et al. Exploiting MEK inhibitor-mediated activation of ERα for therapeutic intervention in ER-positive ovarian carcinoma. PLoS One 2013;8:e54103. 49. Mohammed H, Russell IA, Stark R, Rueda OM, Hickey TE, Tarulli GA., et al. Progesterone receptor modulates ERα action in breast cancer. Nature 2015;523:313–7. 50. Diep CH, Charles NJ, Gilks CB, Kalloger SE, Argenta PA, Lange CA. Progesterone receptors induce FOXO1-dependent senescence in ovarian cancer cells. Cell Cycle 2013;12:1433–49.

Tables

Gene Fwd Rev RPLP0 TAAACCCTGCGTGGCAATC TTGTCTGCTCCCACAATGAAA GREB1 GGTTCTTGCCAGATGACAATGG CTTGGGTTGAGTGGTCAGTTTC ESR1 GAGTATGATCCTACCAGACCCTTC CCTGATCATGGAGGGTCAAATC IGFBP3 CACAGATACCCAGAACTTCTCC CAGGTGATTCAGTGTGTCTTCC CCNG2 GTTTGGATCGTTTCAAGGCG CCTCTCCACAACTCATATCTTCAC PGR TCGCCTTAGAAAGTGCTGTC GCTTGGCTTTCATTTGGAACG MYC GCTGCTTAGACGCTGGATTT GAGTCGTAGTCGAGGTCATAGT Table 1: Primer sequences for qPCR

UPMC (N=14) RPCI (N=19) FCCC (N=8) Mich (N=26) Age at dx 60.3 ± 10.5 72.4 ± 7.3 56.1 ± 7.7 58.9 ± 15.9 Primary/Recur samples 14/0 17/2 5/3 26/0 Grade Low 2 0 1 3 High 12 19 7 23 Stage Early (1-2) 1 0 3 2 Late (3-4) 13 19 5 24 Histology Serous 10 17 5 15 Clear Cell 0 1 0 1 Endometrioid 1 0 3 1 Other 3 1 0 9 Pre-rx CA-125* 340.2±544.0 (5) 392.4±646.0 (1) 463.3±785.0 (4) 105.0±143.4 (0) (# censored) Endo rx Tamoxifen 11 18 7 14 AI 2 0 1 6 Both 2 1 0 6 Days on endo rx 351 ± 358 419 ± 661 168 ± 115 260 ± 310 Min 38 22 31 30 Max 908 2850 396 1470 Survival after endo rx 882 ± 564 717 ± 880 738 ± 653 834 ± 774 (days) Overall survival (days) 2346 ±1302 2019 ± 1372 2129±1104 1880 ± 1228 Table 2: Clinical features of the patient cohort. *Pre- and post-endocrine therapy CA-125 was available for 45 patients.

FeatureName Expression long vs short rx IGFBP3 Higher in short rx RASGRP1 Lower in short rx KIAA1467 Lower in short rx RNF144A Higher in short rx BCAR3 Lower in short rx CDCA7 Lower in short rx PHLDA1 Lower in short rx SDHA Lower in short rx MUC1 Higher in short rx PTEN Lower in short rx PDGFRL Higher in short rx STAT3 Lower in short rx PIK3CA Higher in short rx CD8 Lower in short rx MYC Lower in short rx SMTNL2 Higher in short rx MYBL1 Higher in short rx SLC25A29 Higher in short rx MTUS1 Higher in short rx ID2 Higher in short rx DAAM1 Higher in short rx INPP4B Higher in short rx TNF Lower in short rx DNAJB4 Lower in short rx ARHGAP26 Lower in short rx SLC30A1 Higher in short rx RDX Lower in short rx YPEL2 Higher in short rx Hscore.nona Lower in short rx GATA3 Higher in short rx Table 3: Top 30 features associated with endocrine response from SVM analysis

Figure Legends

Figure 1: Endocrine response in HGSOC cell lines grown in 2-D. A. Expression of ERα mRNA (ESR1) and protein in HGSOC cell lines. The hormone-responsive breast cancer cell line MCF-7 was included as a positive control. B. Effect of E2, fulvestrant, and 4OHT on growth of HGSOC cells after six days.

Hormone-deprived cells were treated with increasing doses of E2, fulvestrant, or 4OHT. Fulvestrant and

4OHT were added in the presence of 100 pM E2. Data are shown as fold change (FC) vs. vehicle. Points represent the mean of six biological replicates; error bars show standard deviation. Graphs are representative of >2 independent experiments. Lines represent best fit non-linear regressions. Curves could not be fit for OVCA432 nor OVSAHO. C. Heat maps depicting gene expression changes (log2FC vs. vehicle) after 3-hour treatment with E2 +/- 4OHT or fulvestrant in PEO4 and PEO1 cells. Genes shown are significantly regulated by E2 compared to vehicle (p<0.001). Heat maps were generated using MeV. D.

Overlap of E2-regulated genes in PEO1 and PEO4 cells versus MCF7 breast cancer cells (GEMS early data). E. Mean log2(FC) (treatment vs. vehicle) of ERα target genes in HGSOC cells. Gene expression was measured by qRT-PCR after 8-hour treatment with Vhc, E2, E2 + fulvestrant, or E2 + 4OHT. Error bars show standard deviation of three biological replicates.

Fig. 2: Endocrine response in 3-D, ULA, and cell line xenografts. A. PEO4 and OVCA432 cells were plated in matrigel, treated as indicated, and allowed to grow for 10 days. Images are representative of two biological replicates. B. Hormone-deprived cells were plated in ULA +/- E2. Data are presented as blank- corrected luminescence (mean of six replicates +/- standard deviation). Graphs are representative of three experiments. C-D. Cells were plated in 2-D or ULA plates for 24 hours. ESR1 mRNA levels were measured by qRT-PCR (C) and ERα protein levels measured by immunoblot (D). Numbers below the band indicate ERα/tubulin ratio. E. PEO4 cells were injected IP into mice after ovariectomy (OVX) plus placebo or E2 pellet supplementation. Tumor burden was measured after 11 weeks and calculated as (tumor weight

/ total body weight)*100%. Each point represents an individual mouse. F. Gene expression in xenografts was measured by qRT-PCR. Each point represents an individual mouse. Differences were not statistically significant.

Fig. 3: ERα expression and endocrine response in HGSOC PDX explants. A. PDX tissue was collected

when mice became moribund (8-16 weeks after engraftment). ERα (ESR1) mRNA was measured by qRT-

PCR. Protein was assessed by Western blot and IHC. H-scores were calculated by a pathologist (E.E.) by

reviewing the slide shown). B. Workflow for explant studies. For each model, 3-6 explants were used for

each experiment (i.e. 3 for proliferation and 3 for qRT-PCR). Each fragment was treated as a unique

biological replicate. C-D. Effect of fulvestrant and 4OHT treatment on ERα levels and BrdU incorporation in

PH045 explants. E. Effect of fulvestrant and 4OHT on gene expression of PH045 explants. F-H. Effects of

fulvestrant and 4OHT treatment on PH053 explant proliferation (F-G) and gene expression (H). For graphs

in D and G, each point represents a separate field of view and bars show the mean. For E and H, each dot

represents an individual explant piece. Bars show the mean.

Fig. 4: Identification of ERα targets associated with clinical response to endocrine therapy. A.

Representative sample timeline of a patient in our cohort. After diagnosis (dx), treatment starts with

debulking surgery and chemotherapy. Endocrine therapy is typically given after multiple rounds of

chemotherapy if patients have a “biochemical” recurrence, determined by rising serum CA-125 levels.

Endocrine therapy is continued until the disease progresses by CA-125 or other evidence of disease (e.g.

imaging). B. Representative stains of ERα in our patient cohort (SP1 clone antibody) and distribution of H-

scores. C. Patients with a higher H-score have better response to endocrine therapy. D. IGFBP3

expression in patients with short vs. long duration of endocrine therapy. E. ERα-mediated repression of

IGFBP3 in preclinical HGSOC models. E2 represses IGFBP3 expression in PEO1 and PEO4 cells and this

is reversed by fulvestrant. Fulvestrant and 4OHT increase IGFBP3 expression in PH053 and PH045

explants. F. Kaplan-Meier analysis of association between IGFBP3 expression and time on endocrine

therapy. G. Separation of ERαhigh and ERαlow groups by IGFBP3 expression. H. Separation of long and short treatment groups based on top 30 features identified by SVM.

Active estrogen receptor-alpha signaling in ovarian cancer models and clinical specimens

Courtney L Andersen, Matthew J Sikora, Michelle M Boisen, et al.

Clin Cancer Res Published OnlineFirst January 10, 2017.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-16-1501

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2017/01/10/1078-0432.CCR-16-1501.DC1

Author Author manuscripts have been peer reviewed and accepted for publication but have not yet been Manuscript edited.

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2017/01/10/1078-0432.CCR-16-1501. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.