Introduction

Epithelial ovarian cancer (EOC) is the most lethal gynecologic malignancy, and high-grade serous cancer (HGSC) is the most prevalent EOC subtype1,2. The majority of HGSC cases are diagnosed at late clinical stages. Once diagnosed, EOC and HGSC treatment consists of primary debulking surgery and platinum/taxane combination chemotherapy, typically leading to a robust clinical response. Unfortunately, most patients diagnosed in late stage ultimately relapse with chemoresistant disease3. Although there has been significant recent progress in ovarian cancer treatment4,5,6, there remains an urgent need for improved therapeutic approaches, particularly in the recurrent disease setting.

Cancer-testis antigens (CTAs), also known as cancer-germline genes, show low expression in normal somatic tissues but are expressed in germ cells of the adult testis and fetal ovary, and in placenta7,8. CTAs can show highly elevated expression in cancer, which appears most often to result from epigenetic alterations, particularly DNA hypomethylation9,10,11. Some CTAs are immunogenic (hence the name), in part because their normal expression is restricted to immune privileged sites. The immunogenicity of specific CTAs has led to the development of immunotherapies to target them in cancer, using vaccines and adoptive cell therapies8,9. Importantly, specific CTAs directly promote oncogenic phenotypes, suggesting they are not just cancer passengers12,13,14. This opens up new opportunities for therapeutic targeting of CTAs unrelated to immunotherapy, which is a crucial development, as only a limited number of CTAs are likely to be immunogenic.

A sizable number of CTAs, including the most frequently studied members of this superfamily, are located on the X-chromosome (CT-X genes). However, most CTAs were recently shown to be encoded on autosomes13,15. Amongst these, POTEs are the only multigene family described to date, POTEs consist of 14 primate-specific genes distributed on seven chromosomes, and are divided into three phylogenetic groups16,17,18. The POTE family originated from an ancestral ankyrin repeat domain 26 (ANKRD26) gene17. POTEs contain a conserved 3′UTR LINE-1 element, which promoted POTE dispersal in the primate genome, and several Chr. 2 POTEs contain a C-terminal in-frame fusion with Actin resulting from transposition16,19 (Table 1). Structurally, POTE proteins contain a N-terminal cysteine-rich region, central ankyrin repeats, and C-terminal spectrin-like α-helices, suggesting participation in protein-protein interactions and association with cell membranes19,20.

Table 1 Human POTE Gene Family.
Full size table

An important early study of POTE expression in cancer showed differential POTE expression in cancer tissues, including ovarian cancer. However, the analysis of ovarian cancer was limited to an endpoint RT-PCR study of five ovarian cancer samples of unknown classification21. A limitation to early studies of POTEs was that the high homology of POTEs made it difficult to resolve expression of individual POTEs. However, in recent years, the field has experienced the advent of RNA-sequencing (RNA-seq), which can readily resolve individual POTEs, as well as great progress by consortia-based projects for depositing extensive RNA-seq data from normal human tissues, human tumors, and human cancer cell lines2,22,23,24. These data allow the opportunity to measure POTE expression in different contexts, including ovarian cancer. Here we report several new and extensive analyses of POTE expression, including in normal tissues, ovarian cancer tumors and cell lines, normal control cells, and an initial study in pan-cancer tissues and cell lines.

Results

POTE expression in normal human tissues

We first analyzed expression of 13/14 members of the POTE gene family (data was not available for POTEB3) (Table 1), using GTEx RNAseq data22, primarily to determine if POTEs show a testis-specific or testis-enriched expression characteristic of CTAs7. Notably, Groups 1 & 2 POTEs, which are more closely related to the ancestral ANKRD26 gene17, displayed testis-specific expression (Supplementary Fig. S1), despite the fact that ANKRD26 was widely expressed in normal tissues (data not shown). In contrast to Group 1 & 2 POTEs, Group 3, and particularly the POTE-actin genes, showed widespread normal tissue expression (Supplementary Fig. S1). The only exception was POTEH, a Group 3 POTE that showed significant expression only in testis and prostate. We conclude that Groups 1 & 2 POTEs (A, B, B2, C, D) have normal tissue expression consistent with CTAs, while Group 3 POTEs (E, F, G, H, I, J, KP, M) do not (Table 1). Widespread expression of POTE-actin genes suggests a function in normal tissues.

POTE expression in EOC

We measured Pan-POTE expression by RT-qPCR in EOC and bulk normal ovary (NO) tissues. Supplementary Table S1 lists the characteristics of the EOC samples. Pan-POTE was significantly overexpressed in EOC compared to NO, with approximately one-third of cases showing >10-fold increased expression (Fig. 1a). Pan-POTE expression significantly associated with increased clinical stage and pathological grade (Fig. 1b,c). We separated EOC into HGSC (serous histology, grade 2/3) and other EOC. While Pan-POTE was elevated in both groups compared to NO, HGSC showed significantly higher expression (Fig. 1d). Individual histological subgroups did not contain sufficient samples to make meaningful comparisons (Supplementary Fig. S2). Next, to assess individual POTE gene expression in EOC, we used Affymetrix microarrays to examine EOC (n = 40) and NO (n = 3). In agreement with Pan-POTE data, sub-sets of POTEs showed elevated expression in EOC (Supplementary Fig. S3). However, this methodology was limited by extensive POTE gene overlap.

Figure 1
figure 1

Pan-POTE expression in NO and EOC tissues. (a) NO and EOC. (b) NO and EOC separated by stage. (c) NO and EOC separated by grade. (d) NO and EOC separated into HGSC (serous histology, grade 2/3) and other EOC. Graphs show median values, and two-tailed Mann-Whitney tests with significant differences, after performing Bonferroni correction, are shown. Samples with no detectable Pan-POTE expression were plotted at log10 (−7) for clarity.

Full size image

POTE expression in HGSC

HGSC frequently originates from precursor lesions in the fallopian tube epithelia (FTE), and the TCGA ovarian cancer project specifically focused on HGSC2,25,26. To focus our studies of POTEs on HGSC, and to examine individual POTEs using RNA-seq, we used Toil analyses23. As a control for HGSC, we combined normal tissue GTEx data from both ovary and fallopian tube (FT), as utilization of FT alone was not feasible due to limited sample size (n = 5), and because unseparated FT is only an approximation of FTE. This analysis revealed significant overexpression of 10/13 POTEs in HGSC (Fig. 2a–c). Amongst Groups 1 & 2 POTEs, A, B2, and C showed significant upregulation, along with generally low or absent expression in control tissues (Fig. 2a). All Group 3 POTE-actin genes showed altered expression in HGSC, with all but one (POTEJ) being upregulated (Fig. 2b). We noted that POTE-actin expression was significantly upregulated in HGSC despite expression in the control tissues. Other Group 3 POTEs (POTEs G/H/M) were also highly upregulated in HGSC, but showed lower expression in control tissues than POTE-actin genes (Fig. 2c). Comparison of the expression of all POTEs revealed that POTEs C, E, F, and I show highest overall expression in HGSC (Fig. 2d).

Figure 2
figure 2

POTE expression in fallopian tube (FT) + ovary and HGSC. (a–c) Comparisons of POTE expression in normal controls (n = 93) vs. HGSC (n = 419). (a) Groups 1 & 2 POTEs. (b) Group 3 POTE-actin genes. (c) Other Group 3 POTEs (i.e. POTEs G/H/M). (d) Comparison of POTE gene expression in HGSC. Box and whiskers plot, with medians, 10–90%iles, and ranges indicated. Two-tailed Mann-Whitney tests with significant differences are shown.

Full size image

We used unsupervised hierarchical clustering to compare POTE expression in TCGA HGSC data. POTEs generally clustered into three expression sub-groups: i) Groups 1 & 2, ii) Group 3 POTE-actin genes, and iii) POTEs G/H/M (Fig. 3a). We also identified different tumor clusters characterized by specific POTE expression patterns (Fig. 3a, right labels), and the most prominent clusters were characterized by high expression of POTEC and/or POTE-actin genes. We conducted Spearman rank correlation testing of POTE expression, which confirmed that the three aforementioned POTE subgroups show correlated expression (Fig. 3b). In agreement with earlier data, the two POTEs that did not correlate within their respective subgroups (POTEs D and J) either showed very low expression in HGSC or were downregulated in HGSC compared to normal controls (Fig. 2a,b).

Figure 3
figure 3

POTE expression in HGSC. (a) Expression heatmap of POTEs in TCGA HGSC data. Toil log2 normalized read counts shown, and coloring indicates row min to row max (see key). Samples showing enrichment for specific POTE expression patterns are labelled at right. (b) Spearman rank correlation matrix heatmap of POTE gene expression in TCGA HGSC. In both panels, POTE font color indicates POTE group: Groups 1 & 2 (blue), Group 3 POTE-actin (red), POTE G/H/M (black).

Full size image

HGSC patients often develop recurrent chemoresistant disease3. We compared POTE expression in patient-matched primary and recurrent HGSC using two independent RNA-seq data sets27,28. Data from Patch et al., showed altered expression of several POTEs, and identified POTEF, I, and M with significant upregulation in recurrent HGSC, both in individual patients and overall (Fig. 4a,c). In addition, POTEs C and E were upregulated in several patients. Data from Kreuzinger revealed a similar pattern of altered POTE expression, with increased expression of POTEs C, F, I, and M in recurrent HGSC. Only POTEC was significantly upregulated over the entire patient population (Fig. 4b,d). Upregulated POTEs included at least one member of each previously identified POTE expression subgroup (i.e. Groups 1 & 2, POTE-actin genes, and POTE G, H, M).

Figure 4
figure 4

POTE expression in patient-matched primary and recurrent HGSC. (a,b) Expression heatmaps showing log2 fold changes for recurrent/primary HGSC. (c,d) POTE expression averages. (a,c) Data from Patch et al.27 (n = 12 pairs). (b,d) Data from Kreuzinger et al.28 (n = 66 pairs). Font color indicates POTE group: Groups 1 & 2 (blue), Group 3 POTE-actin (red), other Group 3 (black). (c) Data from27 (n = 12 pairs). (d) Data from28 (n = 66 pairs). Bars plot means + SEM; two-tailed student’s t-test with significant differences are shown. In panels (a,b) POTE font color indicates POTE group: Groups 1 and 2 (blue), Group 3 POTE-actin (red), POTE G/H/M (black).

Full size image

POTE expression and overall survival (OS) in EOC and HGSC

We tested the association of Pan-POTE expression with OS in EOC. Consistent with the observed increase of Pan-POTE with stage, grade, and HGSC (Fig. 1b–d), Pan-POTE associated with reduced OS in a univariate analysis, but not in a multivariate analysis (Fig. 5a; data not shown). We next tested the association of individual POTEs with OS using HGSC TCGA data, and observed that POTEE associated with reduced OS, using either two or three expression sub-groups (Fig. 5b,c). Consistently, POTEE was upregulated in HGSC compared to normal controls, showed heterogeneous expression in HGSC, and select patients showed increased POTEE expression at recurrence (Figs 2b, 3a, 4a,b). Other POTEs were not associated with HGSC OS (data not shown).

Figure 5
figure 5

POTE expression and overall survival (OS) in EOC and HGSC. (a) Pan-POTE expression and OS in EOC (n = 114). (b,c) POTEE expression and OS in TCGA HGSC (n = 417), using either two (b) or three (c) expression subgroups.

Full size image

POTE expression in ovarian cancer cell lines

Cancer cell lines are valuable tools for functional studies29. We measured Pan-POTE in a panel of cell lines relevant to EOC and HGSC, including cancer cell lines, and normal and immortalized ovarian surface epithelia (OSE) and FTE cells (Supplementary Table S2). Consistent with primary tumor data, Pan-POTE expression was significantly increased in ovarian cancer cells compared to control cells (Fig. 6a). Next, we examined the pattern of expression of individual POTEs in a large panel of ovarian cancer cell lines, using data from the cancer cell line encyclopedia (CCLE)24. POTE expression in CCLE ovarian lines segregated into the three POTE sub-groups described above (Fig. 6b,c). A large proportion of cell lines had elevated expression of POTE-actin genes (Fig. 6b).

Figure 6
figure 6

POTE expression in ovarian cancer and control cell lines. (a) Pan-POTE expression in control cells (ovarian surface epithelia, OSE; fallopian tube epithelia; FTE) and ovarian cancer cell lines. See Supplementary Table S1 for list of cell lines utiized. Box and whiskers plot, with medians, 25–75%iles, and ranges indicated. Two-tailed Mann-Whitney test result shown. (b) POTE expression RNA-seq read counts in CCLE ovarian cancer cell lines (n = 50). Cell line names are shown, and samples showing enrichment for specific POTE expression patterns are labelled at right. (c) Spearman rank correlation matrix heatmap of POTE gene expression in CCLE ovarian cancer cell lines. In panels (b,c) POTE font color indicates POTE group: Groups 1 & 2 (blue), Group 3 POTE-actin (red), POTE G/H/M (black).

Full size image

POTE expression in pan-cancer TCGA and CCLE data

We utilized in silico resources to conduct an initial examination of POTE expression in pan-cancer24,30. Pan-cancer TCGA data showed similar POTE expression sub-groups and sample clusters as observed in HGSC (Fig. 7a,b). However, although the data were overall similar to HGSC, there were distinctions, including elevated POTEJ expression in a sub-set of tumors (Fig. 7a). In pan-cancer CCLE data, again similar POTE expression patterns were apparent, including sample clusters with increased expression of POTEC, POTE-actin genes, and Group 3 POTEs (Fig. 8a). Moreover, the three previously identified POTE expression sub-groups (Group 1 & 2, POTE-actin genes, and POTE G/H/M) perfectly segregated in pan-cancer CCLE data (Fig. 8a,b).

Figure 7
figure 7

POTE expression in TCGA pan-cancer tissues (n = 9345), determined using RNA-seq data from the UCSC Xena browser Toil. (a) Unsupervised hierarchical clustering of individual POTEs and pan-cancer cases. log2 normalized read counts are shown. Samples showing enrichment for specific POTE expression patterns are labelled at right. (b) Spearman rank correlation matrix heatmap of POTE gene expression in TCGA pan-cancer data. POTE font color indicates POTE group: Groups 1 & 2 (blue), Group 3 POTE-actin (red), POTE G/H/M (black).

Full size image
Figure 8
figure 8

POTE expression in Cancer Cell Line Encyclopedia (CCLE) pan-cancer data. (a) POTE expression read counts in CCLE pan-cancer cell lines (n = 1076). Samples showing enrichment for specific POTE expression patterns are labelled at right. (b) Spearman rank correlation matrix heatmap of POTE gene expression in CCLE pan-cancer cell lines. POTE font color indicates POTE group: Groups 1 & 2 (blue), Group 3 POTE-actin (red), POTE G/H/M (black).

Full size image

Discussion

Pan-POTE expression is frequent in EOC and correlates with increased stage and grade, HGSC, and reduced OS. Although these data are valuable, it is important to determine individual POTE gene expression in the context of normal tissues and cancer. Due to extensive sequence homology this previously was difficult, requiring PCR cloning and Sanger sequencing21. To overcome this limitation, we utilized microarrays and, more extensively, RNA-seq. Microarray studies indicated that POTE sub-groups have increased expression in EOC compared to NO. Due to the availability of extensive RNA-seq data for HGSC2, and given our observation of Pan-POTE overexpression in this EOC subtype, we focused subsequent studies on HGSC. We used TCGA HGSC data, and GTEx normal FT and ovary as the control, and our analyses revealed that most individual POTEs (10/13 genes) are overexpressed in HGSC. Importantly, GTEx data revealed that Groups 1 & 2, but not Group 3, POTEs show a testis-specific expression pattern characteristic of CTAs. We conclude that Groups 1 & 2 POTEs are CTAs that can be overexpressed in HGSC (3/5 genes), with POTEC showing the most robust overexpression. In contrast, Group 3 POTEs are not CTAs but are more commonly overexpressed in HGSC (7/8 genes), with POTEJ the lone exception. A caveat to our analysis is that GTEx used bulk tissues, not specifically isolated epithelial cells22. Because FTE secretory cells are the progenitor cell for HGSC, future studies should determine POTE expression in this cell type, as well as in HGSC precursor lesions in the distal FT25,26,31. Additionally, a recent study suggests that evaluation of testis-specific expression in the context of CTA gene classification benefits from the use of isolated testicular germ cells15.

POTEs showed patterns of correlated gene expression, and the three sub-groups were: i) Groups 1 & 2 POTEs, ii) Group 3 POTE-actin genes, and iii) other Group 3 POTEs (i.e. POTEs G/H/M). These data suggest transcriptional co-regulation with sub-groups and divergence between groups. As CTA genes are regulated by epigenetic mechanisms9, it becomes relevant to determine whether epigenetics states, and/or specific transcription factors, explain the observed POTE expression sub-groups.

In addition to ovarian cancer, we conducted an initial examination of POTE expression in pan-cancer data sets from TCGA and CCLE. The data showed relative similarity of POTE expression patterns in pan-cancer. For example, sample sub-groups showed high enrichment of POTE-actin genes, POTEC, and Group 3 POTEs. Additionally, the three HGSC expression sub-groups were also apparent in pan-cancer data. Moving forward, it now becomes relevant to determine whether specific tumor types or lineages are enriched for specific patterns of POTE expression.

In contrast to POTE gene expression, POTE protein expression data in large cancer data sets is currently unavailable. In addition, commercial POTE antibodies recognize all or most POTEs, restricting their utility (data not shown). Supporting the relevance of our mRNA expression data, our prior studies of CTAs in EOC, including CTCFL (BORIS), CT45, and PRAME, revealed significant correlations between mRNA and protein expression32,33,34. Nevertheless, an important goal is to measure POTE protein expression levels in EOC and HGSC and to determine the relationship of protein expression to clinicopathology. Of note, a recent proteomic study reported increased POTEE expression in breast cancer35. It is intriguing that we observed that POTEE was the only POTE gene associated with reduced OS in HGSC, given the fact that HGSC has high genomic similarity to basal breast cancer36.

POTE protein expression was previously detected in human testis and spermatids, where it was associated with apoptosis37,38. Moreover, studies of cancer cells provide tentative support of a role for POTEs in apoptosis39,40. In addition, POTE-actin proteins appear likely to play a role in cytoskeletal function given their structure. Future work on POTE function in ovarian and other cancers might thus focus on apoptosis and cytoskeletal functions as starting points for investigation.

For functional cancer studies, cell lines are an invaluable tool. In this context, we observed that EOC/HGSC cell lines have significantly elevated POTE expression compared to normal OSE and FTE controls. In particular, POTE expression in the CCLE cell lines provides useful insight into model choice to study of POTE function in ovarian and other cancers.

The fact that several POTEs are not CTAs, combined with the high conservation of POTE proteins, could make immunological approaches to target POTEs difficult, despite the fact that POTE epitopes are capable of generating human CTL responses41. Regardless of the limitations in immunological targeting of POTEs, frequent POTE overexpression in EOC, HGSC, and other cancers, along with limited or absent expression in most normal tissues, supports POTEs as potential therapeutic targets. An important next step will be to determine whether (and which) POTEs have oncogenic function. Such data will provide insight into the potential of POTE-targeted approaches for cancer treatment.

Methods

POTE expression in human adult normal tissues

We determined the expression of individual POTEs in human adult normal tissues using GTEx22. We obtained GTEx RNAseq data using GeneCards (http://www.genecards.org/).

Pan-POTE expression in human EOC and normal ovary (NO) tissues

We obtained fresh-frozen human EOC and bulk normal ovary (NO; obtained from patients without malignancy). All samples were collected using IRB-approved protocols at Roswell Park Comprehensive Cancer Center (RPCCC)42. All experiments using human samples were approved by the Institutional Review Board of the RPCCC and the Institutional Review Board of the University of Nebraska Medical Center (UNMC), and all methods were performed in accordance with relevant guidelines and regulations. Informed consent was obtained from all subjects and all subjects were over the age of 18. We processed tissues as described43. We extracted RNA using TRIzol (Invitrogen) and synthesized cDNA using iScript cDNA Synthesis Kit (BioRad). We performed qPCR using the BioRad CFX Connect system with SYBR green master mix (Qiagen), and primers from IDT. We amplified Pan-POTE (i.e. all POTE genes) as described19. We also determined POTE expression in EOC (n = 40) and NO (n = 3) using Affymetrix HG 1.0ST arrays, performed by the University at Buffalo Center of Excellence in Bioinformatics and Life Sciences (UBCOE). We normalized microarray probe cell intensity data (.cel) using the Affymetrix Expression Console (version 1.3.0.187) software running the Robust Multi-chip Averaging (RMA) background correction and quantile normalization using a linear scale.

POTE expression in fallopian tube (FT), ovary, and HGSC tissues

We obtained Toil GTEx data for FT and ovary, and Toil TCGA data for HGSC and pan-cancer. All data correspond to RNA-seq normalized read counts. We obtained data from the UCSC Xena Browser (https://xenabrowser.net)23.

POTE expression in primary and recurrent HGSC

We obtained POTE RNA-seq data from patient-matched primary and recurrent HGSC using the European Genome-phenome Archive (EGA) https://ega-archive.org/. We analyzed EGAD00001000877 (n = 12 pairs) and EGAD00010001403 (n = 66 pairs)27,28.

POTE expression and overall survival (OS) in EOC and HGSC

For EOC, we defined overall survival (OS) as the time between the date of diagnosis and death, and censored patients who were alive at the time of analysis at the date of last follow up. We split EOC patients into Pan-POTE expression tertiles and compared OS using Kaplan-Meier analysis and Logrank test. For HGSC, we analyzed individual POTE expression vs. HGSC survival using the UCSC Xena Browser (https://xenabrowser.net).

POTE expression in ovarian cancer, OSE, and FTE cells

We measured Pan-POTE expression as described above19. We obtained OVCAR3, A2780, and OVCAR429 from ATCC and cultured as described43. We obtained and cultured Kuramochi, OVSAHO, SNU119, COV318, COV362, OVCAR4, and SV40 large T-antigen immortalized normal human OSE (IOSE-SV) cells as described44. We obtained SKOV3 from ATCC and cultured in McCoy’s media with standard supplementation. We obtained primary human OSE from ScienCell and cultured according to manufacturers’ instructions. We obtained CAOV3 and OVCAR5 from Dr. Anirban Mitra and cultured as described45. We obtained OVCAR8 cells from the NCI and cultured in DMEM, using standard supplementation. We obtained EFO-21 from the MD Anderson Cancer Center (MDACC) Cell Line Core and cultured in RPMI 1640 and 20% FBS with standard supplementation. We obtained FU-OV1 from MDACC Cell Line Core and cultured in DMEM/F12 with standard supplementation. We obtained and cultured FT190, FT237, FT282, and FT282-CCNE1 as described31,46,47. We generated a clonal FT282 cell line, FT282-c11, and FT282-c11-FOXM1c cells as described in Supplementary Methods. We obtained CCLE RNA-seq data (normalized read counts, release date: May 2, 2018), generated and funded by Broad Cancer Dependency Map (https://depmap.org/broad/), using the Broad CCLE Portal (https://portals.broadinstitute.org/ccle/data). We analyzed data for both ovarian cancer cell lines in CCLE (n = 50) and pan-cancer cell lines (n = 1076).

Statistical analyses

We used descriptive statistics as described in the individual figure legends to compare group differences. We used Spearman rank order tests to measure expression correlations. We assigned p < 0.05 as the cutoff for statistical significance. We used GraphPad Prism to conduct statistical analyses. Statistical analyses relevant to survival are described above.