Characterization of a novel estrogen- and progesterone-responsive endometrial cancer cell line: HCI-EC-23

Most endometrial cancers express the hormone receptor estrogen receptor alpha (ER) and are driven by excess estrogen signaling. However, evaluation of the estrogen response in endometrial cancer cells has been limited by the availability of hormonally responsive in vitro models, with one cell line, Ishikawa, being used in most studies. Here, we describe a novel, adherent endometrioid endometrial cancer (EEC) cell line model, HCI-EC-23. We show that HCI-EC-23 retains ER expression and that ER functionally responds to estrogen induction over a range of passages. We also demonstrate that this cell line retains paradoxical activation of ER by tamoxifen, which is also observed in Ishikawa and is consistent with clinical data. The mutational landscape shows that HCI-EC-23 is mutated at many of the commonly altered genes in EEC, has relatively few copy-number alterations, and is microsatellite instable high (MSI-high). In vitro proliferation of HCI-EC-23 is strongly reduced upon combination estrogen and progesterone treatment. HCI-EC-23 exhibits strong estrogen dependence for tumor growth in vivo and tumor size is reduced by combination estrogen and progesterone treatment. Molecular characterization of estrogen induction in HCI-EC-23 revealed hundreds of estrogen-responsive genes that significantly overlapped with those regulated in Ishikawa. Analysis of ER genome binding identified similar patterns in HCI-EC-23 and Ishikawa, although ER exhibited more bound sites in Ishikawa. This study demonstrates that HCI-EC-23 is an estrogen- and progesterone-responsive cell line model that can be used to study the hormonal aspects of endometrial cancer.


Introduction
Cancer of the uterine corpus is the most common gynecologic cancer, 4th most common overall cancer type, and 6th most common cause of cancer death for women in the US, with an expected 65,950 new cases and 12,550 deaths in 2022 [1]. Uterine cancer remains one of the few cancers with increasing incidence and mortality, in part due to its strong association with the obesity epidemic [2,3]. The vast majority (>90%) of cancers of the uterine corpus are cancers of the endometrium, the inner epithelial lining of the uterus. Endometrial cancers are broadly separated into two histopathological subtypes [4]. Type I tumors make up ~85% of endometrial tumors and have low-grade (grade I or II) endometrioid histology. Type I tumors express high levels of estrogen receptor alpha (ER) and are believed to be hormonally driven with excess estrogen signaling being a significant risk factor. In contrast, type II tumors are thought to be hormone-independent and include high-grade (grade III) endometrioid tumors along with tumors of other histologic subtypes, including serous, clear cell, carcinosarcoma, and mixed histology. The broad histopathological classification of endometrial cancer also associates with molecular classifications identified by The Cancer Genome Atlas, where type I tumors fall largely into the copy-number low, microsatellite instable (MSI), and POLE mutated etiologies, while type II tumors are enriched for the copy-number high etiology and are enriched for TP53 mutations [5][6][7].
There have been many cell lines established from type I tumors in order to model disease biology, such as Ishikawa [8,9], AN3CA [10], HEC-1-A [11], RL95-2 [12], MFE-280 [13], MFE-296 [13], JHUAS-1 [14], and JHUEM-2 [14]. However, the majority of these lines have limited ability to assess the estrogen driven aspects of type I endometrial cancer biology due to loss of ER expression and response to estrogens. This has left the Ishikawa cell line as the prevailing model for studying the hormonal aspects of endometrial cancer in vitro [15][16][17][18][19][20][21][22]. Early studies made use of another hormone receptor positive cell line, ECC-1, which has been recently identified to be contaminated by the Ishikawa line at some point, rendering its utility limited [23,24]. The lack of additional cell lines to study the >85% of endometrial cancers that are hormonally-driven has limited the generalizability of previous studies. This is especially true when considering the example of hormone receptor positive breast cancer cell lines, where there are variable responses to hormones, therapeutic responses, and resistance mechanisms to anti-hormone therapy [25][26][27][28][29]. Many of these differences can be attributed to the unique genetic background of each of the cell lines [29]. Therefore, we are limited to studying the estrogen-driven aspects of type I endometrial cancer through the lens of a single genetic background, that of the MSI-high Ishikawa cell line [24]. The increasing use of patient-derived xenograft (PDX) models in endometrial cancer has increased the availability of hormone-responsive model systems [30][31][32][33]. Although these PDX models are highly informative, cell lines remain the predominant model system where large numbers of cells are needed, such as drug and genetic screens, or for genetic manipulation to model clinically-relevant mutations.
We addressed the lack of hormonally-responsive endometrial cancer cell lines by isolating and characterizing endometrial cancer cell lines from PDX models, which led to the development of a cell line model, HCI-EC-23, that retains estrogen-responsiveness in vitro and in vivo. We demonstrate that this cell line is epithelial in nature and grows well as a monolayer in vitro. We also describe the genetic background of HCI-EC-23 by showing that it is mutated at many of the top mutated genes in endometrioid endometrial carcinoma and that it is MSI-high. Critically, the HCI-EC-23 cell line retains ER expression and estrogen-responsiveness in vitro and in vivo as assayed by western blot, luciferase reporter assays, growth assays, and RNA-seq. We also show that the HCI-EC-23 is similar but distinct from Ishikawa in its response to 17b-estradiol (E2) on a transcriptional level and where ER binds to the genome. We show that ER in both HCI-EC-23 and Ishikawa is paradoxically activated by tamoxifen, while HCI-EC-23 shows an anti-proliferative response to progesterone (P4). This study details the characterization of a novel endometrial cancer cell line that retains hormone-responsiveness and will advance the field's understanding of the hormonal underpinnings of endometrial cancer by providing an additional yet novel model system.
In order to generate a new EEC cell line that retains ER expression and hormonal responsiveness, we leveraged a set of PDX models [22]. PDXs were dissociated and then grown on collagen-coated plates in a complex media (described in Methods). A major concern when initially growing these dissociated xenografts is the presence and overgrowth of murine stromal cells that out-compete the human tumor cells. To assess the level of murine cell contamination in these initial outgrowths, we used species-specific PCR amplicon length (ssPAL) [34,35], to determine the fraction of cells that are either human or mouse. ssPAL analysis revealed that two of our dissociated xenografts were almost exclusively murine stromal cells; EC-PDX-32 was >99% murine and EC-PDX-33 was 100% murine (no human DNA detected) ( Figure S1a). These two xenografts had an overwhelmingly fibroblast-like appearance and eventually stopped growing in culture with no outgrowth of human clones. In contrast, the EC-PDX-23 dissociation led to an outgrowth that was predominantly human (>99%) as detected by ssPAL ( Figure S1a). The resulting line, which we have named HCI-EC-23, was subsequently adapted to standard tissue culture-treated dishes and standard RPMI-1640 media supplemented with 10% FBS. Additional testing by IDEXX BioAnalytics (Columbia, MO, USA) detected no murine content and found the line to be mycoplasma negative ( Figure S1a). Short tandem repeat (STR) cell line authentication was performed and yielded a novel signature indicating that HCI-EC-23 is not contaminated with cell lines known in either the IDEXX or DSMZ cell line databases (data not shown).

HCI-EC-23 is an endometrial cancer cell line that retains estrogen signaling
The primary tumor leading to HCI-EC-23 was obtained from a 66-year-old woman who was diagnosed with a grade 2 stage 1A endometrioid endometrial carcinoma ( Table 1). Additionally, this primary tumor was microsatellite instability-high (MSI-high) and was positive for expression of ER (Table 1, Figure 1b). The PDX model from which HCI-EC-23 was cloned (EC-PDX-23), retained grade 2 histology, MSI-high, and ER positivity (Table 1, Figure 1b). MSI testing of the HCI-EC-23 cell line confirmed an MSI-high status ( Table 1). The HCI-EC-23 cell line grows as a monolayer exhibiting characteristic epithelial cobblestone-like morphology, similar to the Ishikawa cell line, although the HCI-EC-23 cell line appears to exhibit an increased degree of cytoplasmic projections and cell-cell contacts (Figure 1c). Epithelial cell adhesion molecule (EPCAM) positivity, as assessed by immunofluorescent staining, confirms that HCI-EC-23 is an epithelial cell line similar to Ishikawa (Figure 1d).
Unlike many of the EEC cell lines described earlier, HCI-EC-23 retains ER expression as detected by western blot (Figure 1a) and qRT-PCR (Supplemental Figure S1c). ER is expressed in HCI-EC-23 at approximately half the amount detected in Ishikawa at both the RNA and protein levels (Supplemental Figure  S1c and Figure 1a, respectively). ER is also upregulated at the RNA level in HCI-EC-23 upon E2 treatment, while it is repressed in Ishikawa in response to E2 (Supplemental Figure S1c).
We next determined if HCI-EC-23 was responsive to estrogen, similar to the known hormonally responsive cell line Ishikawa. Using a luciferase reporter under the control of tandem estrogen response elements (ERE), ER's preferred binding site, we found that HCI-EC-23 is indeed responsive to treatment with 10 nM E2 in a manner and magnitude similar to Ishikawa (Figure 1e). HCI-EC-23 retains its E2 responsiveness at least through passage 35 (Figure 1f), suggesting it durably remains hormonally responsive in culture.
It is known that the selective ER modulator tamoxifen, which is used to block ER activity for breast cancer treatment, is paradoxically an ER agonist in the normal endometrium and endometrial cancer cells, including the Ishikawa cell line [16,17,36,37]. We found that treatment with 1 µM of the active metabolite of tamoxifen, 4hydroxy-tamoxifen (4OHT), induces ER activity in both HCI-EC-23 and Ishikawa, demonstrating that these cell lines retain the paradoxical activity of tamoxifen observed in the endometrium (Figure 1e).

Mutation spectra of HCI-EC-23 is consistent with MSI-high endometrioid endometrial carcinoma
To define the mutational background of the HCI-EC-23 cell line, whole-exome sequencing was performed. We employed a stringent filtering strategy in order to call variants in this cell line as we had no normal sample to remove germline variation; similar strategies have been utilized by the catalog of somatic mutations in cancer (COSMIC) to assess mutations in cell lines [38]. This filtering strategy reduced the number of variants from 29,958 to 3,445, which are presumed to be somatic (Supplemental File S1). The filtered, non-silent mutations were largely missense mutations (n = 1,651) followed by frameshift mutations (n = 435) (Supplemental Figure S2a). The mutational profile was largely C >T and most closely matched the mutation signature SBS6, which is associated with defective DNA mismatch repair (cosine-similarity 0.873, Supplemental Figure S2b). Next, we compared the mutated genes in HCI-EC-23 to the top twenty most frequently mutated genes in the TCGA EEC UCEC cohort. We found that HCI-EC-23 contained mutations in 11 of the 20 most frequently mutated genes (Figure 2a). Importantly, HCI-EC-23 is mutated at PTEN, PIK3CA, ARID1A, CTCF, and ATM, but remains wild-type for PIK3R1, CTNNB1, KRAS, and TP53. Consistent with our goal of establishing a hormonally responsive EEC cell line, the genes that encode progesterone receptor (PR, PGR) and ER (ESR1) are also wild-type in HCI-EC-23. Copy number alteration (CNA) inference using whole-exome sequencing depth and variant allele fraction (VAF) showed that HCI-EC-23 is relatively copy number neutral aside from the relatively common [5] amplification of chromosome 1q and a few focal alterations (Figure 2b). The mutational landscape, CNA profile, absence of POLE exonuclease domain mutation, and MSI-high status demonstrate that HCI-EC-23 is an MSI hypermutated cell line according to the molecular classification identified in the TCGA analysis [5] of endometrial cancer.

HCI-EC-23 exhibits hormone-responsive growth characteristics
Knowing that HCI-EC-23 expresses ER, we next examined the in vitro growth characteristics of HCI-EC-23 in the presence of E2 alone and in combination with the growth-suppressive hormone P4. Neither HCI-EC-23 or Ishikawa showed increased growth rates upon treatment with 10 nM E2 (Figure 3a). The absence of a statistically significant E2-driven effect on in vitro growth rate for Ishikawa has been previously reported and may be related to differing effects between in vitro and in vivo conditions [21,22]. Upon concurrent E2 and P4 treatment (10 nM and 100 nM, respectively), we observed growth reduction in HCI-EC-23, but not in Ishikawa cells (Figure 3b). The P4-mediated growth reduction for HCI-EC-23 is consistent with the growth-suppressive effect of progesterone in the endometrium. The lack of inhibitory effect in Ishikawa may be related to five-fold lower levels of PGR expression compared to HCI-EC-23 (Figure 4e).
We next assessed the growth of HCI-EC-23 in vivo. HCI-EC-23 cells were implanted subcutaneously in immunocompromised mice that were ovariectomized to prevent endogenous estrogen production. Mice were also implanted with pellets containing E2, E2 + P4, or vehicle. Individual tumors grew similarly within treatment groups, suggesting that HCI-EC-23 can robustly grow as a xenografted cell line (Supplemental Figure S3). HCI-EC-23 tumors in mice treated with E2 grew substantially faster than tumors in vehicle treated mice; tumors in vehicle treated mice exhibited a very slow growth rate that was 10-fold lower than tumors in E2 treated mice ( Figure 3c). These results demonstrate a strong estrogen-dependence for growth in vivo, similar growth characteristics have been observed for Ishikawa [22] which far outpaces the effects observed in vitro. Tumors in mice treated with E2 + P4 grew similarly to tumors in E2 treated mice at first, but tumor volume began to diverge towards the end of the experiment, where E2 treated tumors were significantly larger than E2 + P4 treated tumors ( Figure 3c). This suggests that progesterone impedes in vivo growth at later time points, similar to its impact on in vitro growth. Combined, the in vitro and in vivo growth characteristics of HCI-EC-23 demonstrate increased growth upon E2 treatment, especially in vivo, while P4 reduced E2 driven growth, consistent with the action of these hormones in the endometrium.

HCI-EC-23 exhibits a global estrogen response similar to Ishikawa
To assess the transcriptional response to estrogen in HCI-EC-23, we performed RNA-seq following E2 induction. We identified 591 differentially expressed genes following E2 treatment, 389 upregulated and 202 downregulated (Figure 4a). Gene set enrichment analysis (GSEA) demonstrated that both early (NES = 1.58, FDR q-value = 0.011) and late (NES = 1.45, FDR q-value = 0.031) hallmark estrogen responses are upregulated in HCI-EC-23 cells treated with E2, further supporting the estrogen-responsive nature of the cell line (Figure 4b).
We compared the E2 response in HCI-EC-23 and Ishikawa cells. Analysis of RNA-seq data identified four distinct classes of differentially expressed genes in the two cell lines upon E2 induction: 1) shared E2responsive genes (n = 176), 2) genes regulated in opposite directions (n = 21), 3) genes that are specific to HCI-EC-23 (n = 394), and 4) genes that are specific to Ishikawa (n = 1570) (Figure 4c, Supplemental Table S1). Although the Ishikawa cell line had far more E2-regulated genes than the HCI-EC-23 cell line, we observed statistically significant overlaps between genes upregulated (n = 114, hypergeometric p-value = 1.10E-110) and downregulated (n = 62, hypergeometric p-value = 5.72E-65) in both cell lines, suggesting that they share a core estrogen response (Figure 4d). PGR, a known ER target gene, was significantly upregulated in both cell lines (Figure 4e). The two cell lines exhibited a majority of genes whose differential expression was cell line specific. For example, SOX17, a transcription factor essential for endometrial gland formation [39][40][41], was upregulated upon E2 induction in HCI-EC-23, while Ishikawa exhibited lower SOX17 expression that was not affected by E2 treatment (Figure 4e). Conversely, MMP17, a matrix metallopeptidase implicated in extracellular matrix degradation, was upregulated in the Ishikawa cell line, consistent with previous work [21,42], but was very lowly expressed in HCI-EC-23 (Figure 4e). Together, these data suggest that HCI-EC-23 and Ishikawa share a core estrogen response along with a larger set of cell line specific estrogen-regulated genes.
We also broadly compared the gene expression profile of the E2 treated HCI-EC-23 and Ishikawa cell lines to that of patient samples from TCGA Uterine Corpus Endometrial Carcinoma (TCGA-UCEC) cohort [5]. By principal component analysis (PCA), we found that the TCGA-UCEC cohort can be broadly clustered by histology with endometrioid and serous tumors clustering separately (Supplemental Figure S4b). The HCI-EC-23 cell line clusters with the endometrioid tumors, while the Ishikawa line clusters near the boundary between endometrioid and serous tumors, suggesting that HCI-EC-23 may retain a more differentiated phenotype than Ishikawa (Supplemental Figure S4b). These data are consistent with higher expression of PGR and SOX17 being associated with more well-differentiated tumors [43][44][45][46][47].
Knowing that HCI-EC-23 has a transcriptional response to estrogen similar to that observed in Ishikawa, we next asked if genomic binding of ER is similar upon E2 induction in the two lines. To assess the genomic binding of ER in these lines, we performed ChIP-seq after E2 induction. We found that ER was bound at 1,542 sites in the HCI-EC-23 cell line and 12,361 sites in the Ishikawa cell line (Figure 5a). We hypothesize that the larger number of E2 regulated genes in Ishikawa compared to HCI-EC-23 (Figure 4c) is due to this increased genomic binding of ER in Ishikawa compared to HCI-EC-23. The majority (n = 896, 58.1%) of HCI-EC-23 ER ChIP-seq peaks were shared with Ishikawa (Figure 5a-b). Similar overlap frequencies were found when HCI-EC-23 ER peaks were compared to previously published Ishikawa ER peak sets GSE32465 [48] (57.8%) and GSE129805 [22] (46.0%). ER binding within the 3' UTR of PGR was unchanged between HCI-EC-23 and Ishikawa, consistent with PGR being a shared E2-regulated gene (Figure 5c). In contrast, SOX17 (E2-regulated in HCI-EC-23) and MMP17 (E2-regulated in Ishikawa) had unique ER binding at their respective loci that are consistent with their cell line specific expression patterns (Figure 5c). The region of the MMP17 locus that was uniquely bound by ER in Ishikawa cells has been previously shown to be directly involved in the E2-regulation of MMP17 expression [42,49]. To assess the importance of these ER bound sites on E2-regulated gene expression, we analyzed the proximity of E2-regulated genes to ER bound sites in both cell lines and found that E2-regulated genes are significantly closer (Wilcoxon signed rank test, p < 2.2E-16) than genes not regulated by E2 to ER bound sites in either HCI-EC-23 or Ishikawa (Supplemental Figure S5a). The close proximity of E2-regulated genes to ER bound sites compared to genes not regulated by E2 suggests functional importance for many ER bound sites in regulating E2-responsive genes.
As expected, motif analysis identified ESR1 (ER) as the dominant motif in shared, HCI-EC-23-specific, and Ishikawa-specific ChIP-seq peaks (Supplemental Figure S5b). Among shared ER ChIP-seq sites, PITX2 and ZN770/MAF/ZSC22/SMAD3 motifs are also significantly enriched (Supplemental Figure S5b). HCI-EC-23 specific ER ChIP-seq sites were enriched for ATF3 and ZN770/ZSC22/ZFX/CRX motifs, while Ishikawa specific ER ChIP-seq sites were enriched for AP-1 family members and PITX2/ZN264 motifs (Supplemental Figure  S5b). These differences in motif enrichment suggest that different transcription factors could contribute to differential ER genomic binding or subsequent effects on transcription. Looking more closely at ERE sequences within the ER ChIP-seq sites, we found that 61.7% of shared sites contained an ERE sequence as compared to 43.0% of HCI-EC-23 specific sites or 25.6% of Ishikawa specific sites (Supplemental Figure S5c). The frequency of EREs in these binding sites suggests that ER binding in HCI-EC-23 is strongest at sites with frequent EREs as compared to Ishikawa, where more sites are bound that do not harbor EREs. In addition to differences in ERE frequency, differences in chromatin states between HCI-EC-23 and Ishikawa may explain cell line specific ER binding. To interrogate chromatin states at ER bound sites, we assessed H3K27ac ChIP-seq signal in DMSO treated HCI-EC-23 and Ishikawa cells as a baseline marker of active regulatory regions. We found that H3K27ac signal was relatively high in both cell lines at shared ER bound sites (Figure 5d, Supplemental Figure S5d). Conversely, cell line specific ER bound sites are associated with reduced H3K27ac signal in the other line, suggesting that baseline chromatin state is important for cell line specific ER binding (Figure 5d, Supplemental Figure S5d). We also observed differences in the distribution of ER bound sites relative to transcription start sites between HCI-EC-23 and Ishikawa cell lines (Supplemental Figure S5e). ER is bound more at distal intergenic regions in HCI-EC-23 while in Ishikawa ER binding is expanded at promoters and 5'UTRs. Overall, these gene regulation studies indicate that HCI-EC-23 cells exhibit a robust transcriptional response to estrogen that is similar to Ishikawa.

Discussion
Although most endometrial cancers are hormonally driven and estrogen-responsive, only one currently available endometrial cancer cell line has retained this estrogen-responsiveness, the Ishikawa cell line. The lack of additional cell line models to examine the estrogen response in endometrial cancer has limited our understanding of the mechanisms of pathogenesis and the study of therapeutic interventions. In this study, we described a novel estrogen-responsive cell line, HCI-EC-23, that will benefit the field by providing an additional cell line model to study the hormonally driven aspects of endometrial cancer. Importantly, this model has characteristics that are distinct from the Ishikawa cell line, including sensitivity to progesterone treatment. The sensitivity of HCI-EC-23 to progesterone will be particularly important for translational studies, since synthetic progesterone (progestins) are the only approved pharmaceutical treatment for primary endometrial cancer. We provided our strategy to develop cell lines from PDX tumors along with methods to rapidly determine murine contamination levels by using ssPAL. This strategy led to the isolation of HCI-EC-23, from a PDX model derived from a grade 2, MSI-high, and ER-positive endometrioid endometrial tumor. Critically, HCI-EC-23 retained ER expression as shown by western blot and qRT-PCR and ER was functional as assayed by ERE-driven luciferase. HCI-EC-23 also displays paradoxical activation of ER by tamoxifen, an effect known in the endometrium and Ishikawa cell line [16,17,36,37]. HCI-EC-23 showed a strong P4-induced proliferation arrest in vitro. HCI-EC-23 xenografts showed strong E2-dependence in vivo with a reduction in tumor volume with co-administration of P4. These results demonstrate that HCI-EC-23 is a translationally relevant estrogen-responsive endometrial cancer cell line.
Whole-exome sequencing and MSI typing provided evidence that HCI-EC-23 harbors mutations to many of the genes frequently mutated in endometrioid endometrial cancer. We also showed that HCI-EC-23 has relatively few copy-number alterations and is MSI-high, placing it into the MSI-high molecular subgroup identified by TCGA. Ishikawa is also MSI-high but harbors mutations distinct from HCI-EC-23. Among the top 20 mutated genes in endometrioid endometrial carcinoma, Ishikawa is mutant for PIK3R1, FAT4, KMT2C, PCLO, TP53, and LRP1B [50]. In contrast, HCI-EC-23 is mutant for PIK3CA, KMT2B, TAF1, and NSD1. Both lines are mutant for PTEN, ARID1A, KMT2D, CTCF, ZFHX3, FAT1, and ATM, although the specific alterations to these genes are different between the two lines [50]. This demonstrates that HCI-EC-23 and Ishikawa have similar but distinct genetic backgrounds. For instance, both cell lines harbor two common alterations to the PI3K pathway, sharing altered PTEN but differ in having either mutant PIK3CA or PIK3R1. Used together, the similar but distinct genetic backgrounds between the two lines are more reflective of the mutational heterogeneity of patient tumors when studying the hormonal aspects of endometrial cancer than either line alone.
Molecular characterization of the estrogen response using RNA-seq and ER ChIP-seq in HCI-EC-23 showed many parallels to Ishikawa. A large core set of genes were up-or down-regulated upon E2 induction in both HCI-EC-23 and Ishikawa. This coordinated regulation in both cell lines may be attributable to the significant overlap of ER binding sites in the two lines. However, there remain estrogen-regulated genes that are unique to each cell line. This difference may be due to the different genetic backgrounds and chromatin states of the two lines. Alternatively, the relatively larger number of ER binding sites in Ishikawa may confer regulation of a proportionally larger set of genes, which is likely related to the higher expression of ER in Ishikawa compared to HCI-EC-23. Differences in the degree of cellular differentiation in the two lines could also alter the set of estrogenregulated genes, as suggested by RNA-seq clustering that indicates HCI-EC-23 is more similar to endometrioid TCGA tumors while Ishikawa clusters among a subset of endometrioid tumors closer to serous tumors. HCI-EC-23 expresses higher levels of both progesterone receptor and SOX17 than Ishikawa, both of which are transcription factors essential for endometrial gland development and function [39-41, 51, 52]. In the context of endometrial cancer, both progesterone receptor and SOX17 are tumor suppressors and exhibit reduced expression that is associated with less differentiated, high-grade tumors [43][44][45][46][47]. Progesterone receptor and SOX17 have also been shown to cooperatively regulate target genes as part of a transcriptional network [53][54][55]. Together, these data suggest that HCI-EC-23 cells may have a more differentiated state than Ishikawa.
As HCI-EC-23 is an adherent cell line, it is more easily amenable to screening efforts and genetic manipulation than PDX models due to the large numbers of cells and conditions required. Although the power of PDX models to recapitulate tumor heterogeneity is clear, the hormonal aspects of these models are often less clear and are largely only assessed by hormone-receptor positivity [30][31][32][33]. We believe that HCI-EC-23 is uniquely situated to vastly improve the quality of research regarding the molecular aspects of hormonally driven endometrial cancer by providing another well-characterized cell line model for investigators to utilize.

Endometrial cancer patient-derived xenograft (PDX) models
PDX models were generated from human tissue acquired with patient informed consent and experiments performed in accordance with University of Utah Institutional Review Board approved guidelines and regulations [22]. All animal procedures were conducted following approved University of Utah Institutional Animal Care and Use Committee guidelines and regulations. Animals were housed in the Comparative Medicine Center at the University of Utah under standard conditions. Tumor specimens from patients with endometrial cancer were collected in the operating room by the Biorepository and Molecular Pathology Shared Resource. Tumor specimens were transported in ice-cold RPMI-1640 base medium to the lab. Each tumor specimen was mechanically divided into multiple pieces using sterile disposable scalpels. A minimum of three female 6-to 8week old nude mice (nu/nu, Jackson Laboratory) were used for tumor implantation. Orthotopic implantation of ~4 mm 3 tumor fragments into the lumen of the mouse uterus was performed as described previously [22]. Nude mice were not given supplemental estrogen. Tumor size was measured twice per week. Mice were sacrificed and necropsies performed once tumor size reached 2 cm in diameter.

Immunohistochemical staining of tumor specimens
Immunohistochemical staining of FFPE slides was performed by ARUP Laboratories using an automated immunostainer. Briefly, 5 μm tissue sections were obtained using a standard microtome and de-paraffinized with the EZ Prep solution. The slides were then treated with Cell Conditioning 1 (CC1), pH 8.5 for 68 min at 95°C. The primary antibody (#ab80922, Abcam) was applied for 1 hr at a dilution of 1:100 at 35°C. Following removal of the primary antibody, the secondary antibody (#B8895, Sigma-Aldrich) was applied for 1 hr at a dilution of 1:100 at 37°C. Slides were then exposed to the IView DAB Map detection kit (Ventana) and counterstained with hematoxylin for 8 min. After dehydration in graded alcohol, coverslips were put on and slides were read by boardcertified pathologists. Hematoxylin and eosin (H&E) staining was performed on 5 μm tissue sections.

Tumor dissociation
Patient

ssPAL (species-specific PCR amplicon length)
ssPAL was performed as described previously [34,35]. Primer sets used were reported in Schneeberger et al. 2016 [35]. Primer set 5 (Forward: TCATTGGCTTAAAATGTGT, Reverse: FAM-TTTATTTTAAGGGGTTGTAATG) amplifies a region of Ribonuclease P, yielding a 272 nt amplicon in humans and a 278 nt amplicon in mice. Primer set 43 (Forward: CTATTCCTATAGCACAAAGG, Reverse: FAM-GATGGTGTACACCCATCATG) amplifies a region downstream of RC3H2, yielding a 211 nt amplicon in humans and a 206 nt amplicon in mice. PCR was performed using 2X Phusion HF Mastermix (NEB M0531L) in 10 µL reactions with 20 ng of template DNA. Cycling conditions were 98°C 5 minutes; 30 cycles of 98°C 30 seconds, 52°C 30 seconds, 72°C 30 seconds; 72°C 10 minutes. Samples were diluted 1:20 and then mixed 1:1 by DNA source allowing for analysis of multiple assays for one specimen at a time. Samples were then sent to GeneWiz (South Plainfield, NJ, USA) for fragment analysis by capillary electrophoresis. Fragment traces were then analyzed using the Peak Scanner module on ThermoFisher Cloud. Peaks were called human in the following ranges (primer set 5 271-273 nt; primer set 43 210-212 nt) and mouse (primer set 5 278-279 nt; primer set 43 205-207 nt). Percent human or mouse DNA was calculated by dividing the area of the human or mouse peak by the sum of both peaks.

MSI testing
Microsatellite Instability (MSI) testing was assessed at the quasimonomorphic mononucleotide runs

ERE (estrogen response element) luciferase reporter assay
HCI-EC-23 and Ishikawa cells were grown in hormone-deprived media for 3 days prior to plating at 15,000 cells per well in a 96-well plate in hormone-deprived media. One day after plating, cells were transfected, according to manufacturer's protocol, the following day with dual ERE-driven firefly luciferase and constitutive renilla luciferase constructs (Qiagen, CCS-005L) using FuGENE HD transfection reagent (Promega, E2312). One day post-transfection, media was changed and cells were treated with either DMSO, 10 nM E2, or indicated concentrations of 4OHT in hormone-deprived media. Twenty-four hours post-treatment, luciferase activity was measured using the Dual-Luciferase Reporter assay system (Promega, E1960) on a Promega Glo-Max luminometer (Promega, 9301-010). All experiments were performed in technical triplicate with three biological replicates. Firefly luciferase was normalized to renilla luciferase and then fold change determined to respective DMSO control. Statistical significance was determined by Student's T-test for comparison of treatments to control DMSO.

Cell proliferation assays
Ishikawa and HCI-EC-023 cells were initially cultured in regular media and transferred to hormonedeprived media 48 hours prior to plating for proliferation experiments. For proliferation assays, 96-well plates were seeded with 10,000 cells per well and incubated overnight at 37°C. The following day, control DMSO, 10 nM estrogen, or the combination of 10 nM E2 and 100 nM P4 were added to wells, after which the plates were loaded onto the IncuCyte ZOOM live cell imaging system (Sartorius, 4459). Cell proliferation was monitored via time-lapse image acquisition (2 images per well) every 4 hours for up to 96 hours using the IncuCyte(R) ZOOM software (Sartorius, vers. 2016B, 2018A). The initial acquisition was used as 0-hour timepoint. Doubling times were calculated by first normalizing confluency measurements to cell density at 0-hour timepoint. The log2 values of each normalized confluency measurement were then calculated, and the first 48 hours of each growth curve were fitted to a linear regression model. Doubling times were determined using 1/slope of each best fit line, and statistical significance was assessed by Student's T-test using GraphPad Prism. Cell proliferation assays were performed as two biologic replicates assayed in technical triplicate.

qRT-PCR (quantitative reverse-transcription polymerase chain reaction)
HCI-EC-23 and Ishikawa cells were grown in hormone-deprived media for five days. Total RNA was harvested following an 8 hour DMSO or 10 nM E2 induction with buffer RLT plus (Qiagen, 1053393) supplemented with 1% beta-mercaptoethanol (Sigma Aldrich, M3148-100ML). Total RNA was purified using a Quick-RNA Miniprep kit (Zymo Research, R1055). Twenty-five nanograms of RNA was used as starting material per sample for qRT-PCR using the Power SYBR Green RNA-to-Ct 1-Step kit (Thermo Scientific, 4389986) on a Bio-Rad CFX96 thermocycler. Expression measurements were calculated using the ΔΔCt method using CTCF as a control. Primers used were for ESR1 (Forward: GGGAAGTATGGCTATGGAATCTG, Reverse: TGGCTGGACACATATAGTCGTT) and CTCF (Forward: ACCTGTTCCTGTGACTGTACC, Reverse: ATGGGTTCACTTTCCGCAAGG). Experiments were performed in biologic triplicate and statistical significance was determined by Student's T-test to respective DMSO control.

Whole-exome sequencing
Whole-exome sequencing was performed on the HCI-EC-23 cell line using the Agilent SureSelect XT Human All Exon v7 capture kit. The library was then sequenced on an Illumina NovaSeq 6000 yielding 150 bp paired-end reads. The Whole Exome Sequencing -BWA + GATK 4.0 pipeline (https://cgc.sbgenomics.com/public/apps/admin/sbg-public-data/whole-exome-sequencing-bwa-gatk-4-0/41) was then utilized on SevenBridges Genomics to align reads to hg38 and generate a variant call file (VCF). The SevenBridges pipeline yielded 162,787,374 mapped reads leading to an overall 96.01% of targeted bases having at least 30X coverage as determined by Picard (https://broadinstitute.github.io/picard/) CollectHsMetrics. VCF2MAF (https://github.com/mskcc/vcf2maf) was used to convert the VCF file to a mutation annotation file (MAF) while also annotating variant effects using the Ensembl Variant Effect Predictor [57] (VEP) v102. This MAF file contained 29,958 variants and was filtered to remove presumed germline mutations using the following strategy. All variants tagged as a "common variant" (defined as an allele frequency in at least one gnomAD [58] subpopulation >0.04%) were removed. The remaining variants were filtered to include those without existing variant identifiers, present in COSMIC [38], while removing those with identifiers in dbSNP [59] with any reported allele frequency. This resulted in a filtered MAF file containing only 3,445 mutations which are presumed to be somatic. The MAFTools [60] R package was used to generate mutational frequency, mutation-type, and mutational signature plots. Mutational signature plots were made using the mutational profiles outlined in Alexandrov et. al. 2020 [61]. MAFTools also contained a MAF file of the TCGA uterine corpus endometrial cancer (UCEC) cohort [5,62] that was filtered for endometrioid histology and used to generate the barplot comparing mutational frequency in TCGA to HCI-EC-23 mutations. CNVkit [63] was used to assess copy number alterations (CNA) from the whole-exome sequencing. Because of the lack of paired-normal DNA, CNA analysis with CNVkit was performed against a flat reference of the hg38 genome. The scatter function in CNVkit was used to generate a plot of computed CNA in the HCI-EC-23 cell line along with a plot of VAF using the VCF from the SevenBridges pipeline (described above).

RNA-seq
HCI-EC-23 or Ishikawa cell lines were cultured in hormone-deprived media for five days and then induced with 10 nM E2 or DMSO (vehicle) for eight hours. RNA was harvested by washing cells with PBS and lysing cells with buffer RLT plus (Qiagen, 1053393) supplemented with 1% beta-mercaptoethanol (Sigma-Aldrich, M3148). RNA was then purified using the Zymo Quick-RNA Miniprep kit (Zymo Research, R1055). RNA-seq libraries were generated using the KAPA Stranded mRNA HyperPrep kit (Roche, KK8580/08098115702) using 500 ng of RNA. RNA-seq libraries were pooled and sequenced on either the Illumina HiSeq 2500 or NovaSeq 6000. Two replicates of each cell line were sequenced on each sequencing platform. Sequencing reads were aligned to the hg38 build of the human genome using HISAT2 [64]. SAMtools [65] was used to convert SAM files to BAM files. Genes were defined by the University of California Santa Cruz (UCSC) Known Genes and reads mapped to known genes were assigned with featureCounts [66]. Within each cell line, read counts were normalized and analyzed for differential expression via the DESeq2 [67] package in R [68] while controlling for the sequencing platform by using the design formula (= ~Sequencer + Treatment), where sequencer reflects the sequencing platform and treatment is 10 nM E2 or DMSO. Heatmaps of gene expression z-score by gene were made using the pheatmap R package.

The Cancer Genome Atlas UCEC RNA-sequencing analysis
The Cancer Genome Atlas (TCGA) RNA-seq dataset for Uterine Corpus Endometrial Carcinoma (TCGA, PanCancer Atlas) was downloaded from Xena browser (https://gdc-hub.s3.us-east-1.amazonaws.com/download/TCGA-UCEC.htseq_counts.tsv.gz). Histological data for the TCGA samples was obtained from cBioPortal [69] (https://www.cbioportal.org/study/summary?id=ucec_tcga_pan_can_atlas_2018). Counts were converted from log(2+1) to raw counts. Counts for E2 treated Ishikawa and HCI-EC-23 samples were determined as described in the RNA-seq section. These raw counts were then combined with the TCGA samples and analyzed using DESeq2 [67]. Counts were corrected for their source using the design formula (=~Group). Group=Gertz vs TCGA. Batch effects of the different groups were removed from the variance stabilized counts using the limma [70] package function removeBatchEffect(). Normalized counts were then used to create a PCA plot.

ChIP-seq
HCI-EC-23 and Ishikawa cells were grown in hormone-deprived media for 5 days prior to collection. For ER ChIPs, cells were induced with 10 nM E2 for 1 hour. For H3K27ac ChIPs, cells were induced with DMSO for 8 hours. Cells were then fixed using 1% formaldehyde (Sigma Aldrich, 252549) for 10 minutes at room temperature to cross-link DNA and proteins. Cross-linking was stopped by addition of glycine (Fisher Scientific, BP381) to a final concentration of 125 mM. Cells were then washed with cold PBS and harvested via scraping in Farnham lysis buffer supplemented with protease and phosphatase inhibitors (Thermo Scientific, A32959). Farnham lysis buffer consisted of 5 mM PIPES pH 8.02, 85 mM KCl, and 0.5% NP-40. Chromatin immunoprecipitation was performed as previously described [71] with 5 µg of an Anti-ERα (Millipore Sigma, 06-935) or Anti-H3K27ac (Active Motif, 39133) antibody. ChIP-seq libraries were sequenced on an Illumina NovaSeq 6000. Sequencing reads were aligned to the hg38 build of the human genome using Bowtie [72] using parameters: -m 1 -t -best -q -S -I 32 -e 80 -n 2. MACS2 [73] was used to call peaks with a P-value cutoff of 1x10 -10 and mfold parameter between 15 and 100. Input control libraries from either cell line were used as controls for each ChIP-seq experiment. All ChIP-seq experiments were performed in biologic duplicates. Overlaps between peaks were determined using a 1-bp minimum overlap using BEDTools [74]. Heatmaps and profile plots of ChIP signal were made using deepTools [75]. Motif discovery was performed on 500-bp regions surrounding the summit of identified peaks. Motifs between 6 and 30 bp in size were identified by MEME-ChIP, part of the MEME suite [76], with a motif distribution of zero to one occurrence per sequence. Percentage of peaks containing EREs was identified using AME, part of the MEME suite [76], with the ESR1_HUMAN.H11MO.0.A position weight matrix from HOCOMOCOv11 [77]. Genomic annotation of binding sites was performed using CEAS [78].
Previously published Ishikawa ER ChIP-seq peak calls (narrowPeak or broadPeak files) were downloaded from GEO datasets GSE32465 [48] (sample GSM803422) and GSE129805 [22] (samples GSM3722307 and GSM3722311). Overlaps between individual replicates were determined as described above and peaks were remapped from hg19 to hg38 using UCSC Genome Browser's LiftOver tool. These data were then used to identify overlapping Ishikawa ER peaks with HCI-EC-23 ER peaks with BEDTools.

Immunofluorescence
HCI-EC-23 and Ishikawa cells were grown in regular media in 8-well chamber slides (Thermo Scientific, 177402PK). Cells were washed with 1x PBS and then fixed using 4% formaldehyde (Sigma Aldrich, 252549) in 1x PBS for 10 minutes at room temperature. After fixation, cells were washed 3x with ice-cold 1x PBS-T (1x PBS, 0.1% Tween-20). Cells were then permeabilized with Triton X-100 (Sigma Aldrich, T8787) diluted in 1x PBS (1x PBS, 0.1% Triton X-100) for 10 minutes at room temperature. Cells were then washed 3x for 5 minutes with 1x PBS-T. Cells were then blocked using a BSA blocking buffer (1x PBS-T, 1% BSA, 22.5mg/mL glycine) for 30 minutes at room temperature. Blocking buffer was then removed and cells were incubated with an APCconjugated anti-human EpCAM (BioLegend, 369809) antibody diluted 1:20 in dilution buffer (1x PBS-T, 1% BSA) for 1 hour at room temperature in the dark. Cells were then washed 3x 5 minutes in 1x PBS-T in the dark. Cells were then counterstained with DAPI using 1 µg/mL DAPI (Thermo Scientific, D1306) for 1 minute at room temperature. Cells were then washed 3x with 1x PBS-T. Coverslips were then mounted to the slides using ProLong Gold antifade mounting media (Thermo Scientific, P10144) and sealed with clear nail polish. Images were taken with a 60x objective on a Nikon Ti-E microscope with Andor Zyla CMOS camera using the APC and DAPI channels.

Phase-contrast microscopy
HCI-EC-23 and Ishikawa cells were grown in regular media until roughly 80% confluent. Photomicrographs of cells were taken using a 20x objective on a Life Technologies EVOS FL microscope.

In vivo xenograft model
All animal experiments were conducted following approved University of Utah Institutional Animal Care and Use Committee guidelines and regulations and are reported in accordance with ARRIVE guidelines. A total of 14 mice (6-9 weeks old; 17-27 g) were used in the in vivo xenograft growth study. Mice were housed in a temperature and humidity-controlled room with 12 hour light-dark cycles. Mice had ad libitum access to standard rodent food and water for the course of the study. Ovariectomized NRG (NOD.Cg-Rag1 tm1Mom Il2rg tm1Wjl ISzJ, Jackson Laboratory) mice were implanted with beeswax pellets containing 0.57 mg E2 or combination of 0.57 mg E2 and 10 mg P4 under the skin on the back, between the shoulder blades. Beeswax hormone pellets were made as described previously [79]. Beeswax pellets have been shown to release hormones with a linear increase for the first 7 days post-implantation followed by sustained release with increased serum levels detected for at least 10 weeks [80]. Mice either received E2 pellets (n = 5), E2 + P4 pellets (n = 5), or no pellets (control, n = 5). Mice were then injected with 1x10 6 HCI-EC-23 cells into the flank. 32 days after cell injection, mice were sacrificed and tumors excised and weighed. Tumors were measured using calipers starting 10 days after injection and tumor volume was calculated using the following equation: Volume = (Length x Width 2 )/2.

Data access
Raw and processed sequencing data for RNA-seq and ChIP-seq experiments generated in this study are available under the NCBI Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/) accession numbers: GSE210123 (RNA-seq) and GSE210122 (ChIP-seq). In consultation with the NCI Office of Data Sharing, only the germline-filtered MAF file describing presumed somatic mutations is available as Supplemental File S1. The STR cell line authentication profile for HCI-EC-23 can be provided at the time of cell line distribution or upon request.