FOXR2 Is an Epigenetically Regulated Pan-Cancer Oncogene That Activates ETS Transcriptional Circuits

Abstract

Forkhead box R2 (FOXR2) is a forkhead transcription factor located on the X chromosome whose expression is normally restricted to the testis. In this study, we performed a pan-cancer analysis of FOXR2 activation across more than 10,000 adult and pediatric cancer samples and found FOXR2 to be aberrantly upregulated in 70% of all cancer types and 8% of all individual tumors. The majority of tumors (78%) aberrantly expressed FOXR2 through a previously undescribed epigenetic mechanism that involves hypomethylation of a novel promoter, which was functionally validated as necessary for FOXR2 expression and proliferation in FOXR2-expressing cancer cells. FOXR2 promoted tumor growth across multiple cancer lineages and co-opted ETS family transcription circuits across cancers. Taken together, this study identifies FOXR2 as a potent and ubiquitous oncogene that is epigenetically activated across the majority of human cancers. The identification of hijacking of ETS transcription circuits by FOXR2 extends the mechanisms known to active ETS transcription factors and highlights how transcription factor families cooperate to enhance tumorigenesis.

Significance: This work identifies a novel promoter that drives aberrant FOXR2 expression and delineates FOXR2 as a pan-cancer oncogene that specifically activates ETS transcriptional circuits across human cancers. See related commentary by Liu and Northcott, p. 2977.

Figure 1.

FOXR2 is aberrantly expressed across adult and pediatric cancers through multiple mechanisms. A,FOXR2 mRNA expression log₂(normalized count + 1) generated by RSEM of all TCGA samples across cancer subtypes (n = 8,391). Mean expression for each cancer subtype is indicated by a black horizontal line. Cancer subtypes labeled blue indicate mean FOXR2 expression > 0.1. Cancer subtypes: melanoma (n = 443); EC, endometrial cancer (n = 230); NSGCT, nonseminomatous germ cell tumor (n = 86); CSCC, cervical squamous cell carcinoma (n = 248);NSCLC (n = 994); HNSCC, head and neck squamous cell carcinoma (n = 515); RNCCC, renal nonclear cell carcinoma (n = 348); sarcoma (n = 253); BUC, bladder urothelial carcinoma (n = 407); HCC, hepatocellular carcinoma (n = 363); CA, colorectal adenocarcinoma (n = 365); DG, diffuse glioma (n = 513); IBC, invasive breast carcinoma (n = 1,082); RCCC, renal clear cell carcinoma (n = 510); PA, pancreatic adenocarcinoma (n = 177); CeA, cervical adenocarcinoma (n = 46); GBM, glioblastoma (n = 160); AC, adrenocortical carcinoma (n = 78); PrA, prostate adenocarcinoma (n = 460). Black dotted line, log₂(normalized count + 1) > 1. B,FOXR2 mRNA expression FPKM scores across pediatric cancer subtypes within the INFORM data set (n = 1,279). Mean expression for each cancer subtype is indicated by black horizontal line. Cancer subtypes labeled blue indicate mean FOXR2 expression > 1.0. Cancer subtypes: neuroblastoma (n = 162); HGG, high-grade glioma (n = 209); GCT non-CNS, germ cell tumor, noncentral nervous system (n = 21), GCT CNS, germ cell tumor, central nervous system (n = 9), osteosarcoma (n = 137), medulloblastoma (n = 66); ependymoma (n = 92); RMS, rhabdomyosarcoma (n = 181); Ewing sarcoma (n = 147); B-ALL, B-lymphoblastic leukemia (n = 64); ACC, adrenocortical carcinoma (n = 8); AML, acute myeloid leukemia (n = 34); ATRT, atypical teratoid rhabdoid tumor (n = 21); HCC, hepatocellular carcinoma (n = 13); hepatoblastoma (n = 14); T-ALL, T-lymphoblastic leukemia (n = 16); ETMR, embryonal tumor with multilayered rosettes (n = 13); NHL, non-Hodgkin lymphoma (n = 28); Wilms tumor (n = 44). Black dotted line, FPKM > 0. C, Normal postnatal expression of FOXR2 from the GTEx data set. Mean and individual TPM values are displayed. ***, P < 0.0001 (Tukey multiple comparisons test). Lineages include the following: brain (cortex; n = 255), breast (n = 459), colon (n = 779), lung (n = 578), skin (n = 1,304), testis (n = 361). D,FOXR2 isoforms identified in normal testis, which harbor additional exons to the single exon annotated in RefSeq (Exon +1). The position of each additional exon is shown, with the corresponding annotation used throughout this manuscript to refer to each of them: Exon −1, Exon −2, Exon −3, Exon −4, Exon −5, Exon −6, and Exon −7. Aligned reads from a non–FOXR2-expressing glioma are shown as a negative control. E, The proportion of FOXR2-expressing tumors identified to harbor SVs across 152 tumors encompassing all adult and pediatric lineages. Also shown are the different lineages represented in SV and non-SV FOXR2-expressing tumors. F, Fraction and percentage of tumors within each brain location are shown that harbor SVs or aberrant promoters to activate FOXR2. SVs are enriched in cortical tumors whereas midline gliomas tend to activate FOXR2 through the use of aberrant promoters. ****, P < 0.0001 as determined by Fisher exact test for mutual exclusivity.

Figure 2.

FOXR2 is necessary in FOXR2-expressing cancers. A, Top three tracks: Expression and structure of FOXR2 isoforms in FOXR2-expressing DIPG-IV diffuse midline glioma (DMG) cells (green), non–FOXR2-expressing DIPG-XIII DMG cells (blue), and normal testis (black), as determined by RNA-seq. Bottom four tracks: H3K4me3 promoter and H3K27ac enhancer binding in the same FOXR2-expressing DIPG-IV DMG and non–FOXR2-expressing DIPG-XIII DMG cells. Bars below the H3K4me3 and H3K27ac ChIP-seq tracks indicate the position of statistically enriched peaks (FDR < 0.01). The position of Exon-3 is depicted with a gray bar across the different tracks.B, Firefly to Renilla luciferase relative light unit (RLU) ratio in FOXR2-expressing NCI-H838 NSCLC cells transduced with reporter constructs that harbor the Exon −3 promoter sequence relative to vector control. *, P < 0.05, as determined by a two-tailed unpaired t test comparing Exon −3 promoter sequence to control at 7 days following selection across three independent replicate experiments. C, Firefly to Renilla luciferase RLU ratio in FOXR2-expressing A375 melanoma (SKCM) cells transduced with reporter constructs that harbor the Exon −3 promoter sequence relative to vector control. *, P < 0.05, as determined by two-tailed unpaired t test comparing Exon −3 promoter sequence to control at 7 days following selection across three independent replicate experiments. D, Schematic showing the location of CRISPR interference guides (red, labeled sgRNA_1^KRAB, sgRNA_2^KRAB, sgRNA_3^KRAB, and sgRNA_4^KRAB) in relation to the FOXR2 Exon −3 promoter peak. Four CRISPR-Cas9 KO guides targeting the FOXR2 Exon +1 are also shown. E–G, Representative Western immunoblots showing FOXR2 protein levels following transduction with CRISPRi lentivirus with guides targeting different regions of the FOXR2 Exon −3 promoter peak, or nontargeting vector control using lysates derived from the following FOXR2-expressing cancer cell lines DIPG-IV DMGE (E), NCI-H838 NSCLC (F), and A375 melanoma (SKCM; G). H, Gene effect (y-axis) following FOXR2 ablation in genome-scaled CRISPR-Cas9 loss-of-function screens across 948 cancer cell lines with low vs. high levels of FOXR2 expression (x-axis). Values depict mean ± SEM. ***, P < 0.0001, as determined by two-tailed unpaired t test. I, CRISPR interference of FOXR2 using sgRNA_1-4^KRAB in FOXR2-expressing DIPG-IV DMG cell line attenuates proliferation. Values indicate mean ± SEM across three replicate experiments. *, P < 0.05, as determined by two-tailed unpaired t tests comparing each guide relative to nontargeting control at 120 hours. J, Normalized confluence (relative to day 0) of FOXR2-expressing NCI-H838 NSCLC cells following transduction with CRISPRi vectors and guides targeting the Exon −3 FOXR2 promoter. Values indicate mean ± SEM across three replicate experiments. **, P < 0.01; ***, P < 0.001 as determined by two-tailed unpaired t tests at 120 hours. K, Normalized confluence (relative to day 0) of FOXR2-expressing A375 melanoma (SKCM) cells following transduction with CRISPRi vectors and guides targeting the Exon −3 FOXR2 promoter. Values indicate mean ± SEM across three replicate experiments. *, P < 0.05 as determined by two-tailed unpaired t tests at 120 hours. ns, not significant.

Figure 3.

FOXR2 is aberrantly expressed through epigenetic mechanisms. A, Heatmap of CCLE cell lines depicting z-scores (relative to mean of methylation for each CpG site) for all CpG sites within the FOXR2 TAD on the X chromosome as a function of FOXR2 expression. FOXR2-expressing cell lines contain a 2-Mb region that is significantly hypomethylated relative to other regions of the X chromosome. FOXR2-expressing cell lines are delineated by FOXR2 expression (right). B, Quantification of FOXR2 TAD methylation score for each CCLE cancer cell line (P = 1.965e−15, as determined by Welch t test). Each cell line is represented by an individual dot, and FOXR2 expression (low vs. high) is indicated on the x-axis. C, Correlation between methylation at CpG loci and FOXR2 expression (x-axis) and the probability of these correlations (y-axis). Blue, CpG sites within the FOXR2 TAD; orange, the entire X chromosome. D, CTCF binding of the hypomethylated region shown in A that encompasses FOXR2 in FOXR2-expressing DIPG-IV diffuse midline glioma (DMG; bottom) and non–FOXR2-expressing BT245 cells (top). Three independent replicates are shown for each cell line. Two regions are highlighted: within the FOXR2 TAD (chrX: 55,320,00–55,540,000; left) and outside of the FOXR2 TAD (chrX: 64,730,000–64,830,000; right). E, Quantification of the CTCF pileup specifically in the indicated region encompassing the FOXR2 TAD in three replicates of non–FOXR2-expressing cell lines (BT245) and three replicates of FOXR2-expressing DMG cells (DIPG-IV; P = 0.0424, unpaired t test). F, Quantification of total CTCF peaks across the genome in three replicates of non–FOXR2-expressing BT245 DMG cells and three replicates of FOXR2-expressing DIPG-IV DMG cells (P = 0.1078, unpaired t test). ns, not significant.

Figure 4.

FOXR2 is sufficient in vitro and in vivo to promote proliferation and tumor formation. A, Normalized confluence of H9 hNSCs expressing FOXR2 relative to control. Values indicate mean ± SEM across three replicate experiments. **, P < 0.01, as determined by a two-tailed unpaired t test on day 7. B, Normalized spheroid size (relative to day 0) of mNSCs transduced with lentiviral vectors to induce expression of FOXR2 or vector control. Values indicate mean ± SEM across three replicate experiments. *, P < 0.05, as determined by a two-tailed unpaired t test on day 7. C, Survival curves of control (n = 13) and FOXR2-expressing (n = 10) IUE pediatric high-grade glioma (PdgfraD842V + DNp53) mouse models. ***, P < 0.0001, as determined by the log-rank Mantel–Cox test. D, Representative whole-brain images of control and FOXR2-expressing IUE pediatric high-grade glioma (PdgfraD842V + DNp53) mouse models. GFP expression indicates the cortical location of IUE-targeted pediatric high-grade glioma models. Scale bar, 1 mm. E, Representative hematoxylin and eosin (H&E) staining of control and FOXR2-expressing IUE cortical gliomas. Scale bar, 100 μm. F, Percentage of Ki67-positive cells in tumors collected from control or FOXR2-expressing IUE cortical gliomas. Values indicate mean ± SEM from tumors obtained from six mice per condition. ***, P < 0.0001, as determined by a two-tailed unpaired t test. G, Survival curves of control (n = 11) and FOXR2-expressing (n = 10) IUE brainstem DMG (PdgfraD842V + DNp53 + H3.3K27M) mouse models. ***, P < 0.0001 as determined by log-rank Mantel–Cox test. H, Representative images of control and FOXR2-expressing IUE brainstem DMG models labeled with GFP (green) and counterstained with Hoechst (blue). GFP expression indicates the brainstem location of IUE-targeted DMG models. Hoechst stains nuclei as a marker of all cells. I, Representative hematoxylin and eosin staining of control and FOXR2-expressing IUE brainstem DMG models. Scale bar, 100 μm. J, Percentage of Ki67-positive cells in tumors collected from control or FOXR2-expressing IUE brainstem DMG models. Values indicate mean ± SEM from tumors obtained from six mice per condition. ***, P < 0.0001, as determined by a two-tailed unpaired t test.

Figure 5.

FOXR2 is enriched for ETS motifs. A, Heatmaps of HA ChIP-seq peaks overlapping with H3K27ac and H3K4me3 peaks in H9 hNSCs expressing HA-FOXR2. Each row is centered on a HA-FOXR2 peak. The regions are rank ordered by the HA-FOXR2 signal. Intensity indicates average binding intensity at that site. B, Top 10 most significantly enriched known motifs from HA ChIP-seq in hNSCs expressing HA-FOXR2. Transcription factor, motif sequence, P value, and percentage of target sequences are shown. C, Top 11 TFs with significant binding overlap to HA-FOXR2 from Cistrome GIGGLE analysis. D, Percent of total peaks found in the 28 ETS family TFs compared with 28 randomly selected TFs across two replicate experiments. *, P < 0.05, as determined by a two-tailed unpaired t test.

Figure 6.

FOXR2 specifically regulates ETS transcriptional circuits. A, GSEA of C3-regulatory target gene sets in FOXR2-expressing H9 hNSCs. The top five most significant (q value < 0.05) FOX, ETS, and MYC/MAX gene sets are plotted by −log₁₀(q value). B, GSEA of C3-regulatory target gene sets in FOXR2-expressing tracheobronchial epithelial cells (AALE). The top five most significant (q value < 0.05) FOX, ETS, and MYC/MAX gene sets are plotted by −log₁₀(q value). C, GSEA of C3-regulatory target gene sets in tumor models generated by IUE, comparing FOXR2-expressing glioma tumors to non–FOXR2-expressing controls. The top 10 most significant (q value < 0.05) gene sets are plotted by normalized enrichment score (NES).D, Heatmap depicting relative expression of the FOXR2 target gene set, 15 ETS transcription factor gene sets, and 8 MYC/MAX gene sets within the C3 database in FOXR2-expressing versus non–FOXR2-expressing samples across different human cancers. Gradient depicts NES for each lineage, with red showing high enrichment, blue indicating negative enrichment, and white indicating no significant enrichment of each gene set. Human tumors include TCGA melanoma (SKCM) human tumors (total, n = 472; FOXR2, n = 48), TCGA-LUSC human tumors (total, n = 551; FOXR2, n = 15), osteosarcomas human tumors (total, n = 150; FOXR2, n = 5), human diffuse midline gliomas (DMGs; total, n = 67; FOXR2, n = 5), and human cortical gliomas (total, n = 32; FOXR2 = 6). All colored squares are significantly enriched or depleted with a q value < 0.05, with the exception of osteosarcoma, for which a q-value threshold of < 0.25 was applied. White squares are not significant based on this FDR threshold. E, GSEA of C3 regulatory target gene sets in FOXR2-expressing models (red, right), combining both H9 hNSCs and tracheobronchial epithelial cells (AALE). GSEA was generated by comparing hNSCs and AALE cells transduced with FOXR2 relative to hNSCs and AALE cells transduced with a vector control (n = 3 replicates per condition). The top 10 most significant gene sets are plotted by NES with q value = 4.66E–8. GSEA of C3-regulatory target gene sets in FOXR2 CRISPRi suppression models (blue, left), combining both DMG DIPG-IV and melanoma (SKCM) A375. Primary FOXR2-expressing cancer cell lines with FOXR2 suppression transduced with CRISPRi targeting the Exon -3 promoter were compared with primary FOXR2-expressing cancer cell lines transduced with a CRISPRi control vector (n = 3 replicates per condition). The top 10 most significant gene sets are plotted by their NES with q value = 9.56E–9.

Figure 7.

ETS TFs are necessary for FOXR2-mediated proliferation. A, Normalized confluence of H9 hNSCs expressing FOXR2 + vector control, FOXR2 + ETS1^sg2_KRAB, HcRed + vector control, and HcRed + ETS1^sg2_KRAB. Values indicate mean ± SEM across three replicate experiments. *, P < 0.05, as determined by a two-tailed unpaired t test on day 5. B, Normalized confluence of H9 hNSCs expressing FOXR2 + vector control, FOXR2 + ETV3^KRAB, HcRed + vector control, and HcRed + ETV3^KRAB. Values indicate mean ± SEM across three replicate experiments. *, P < 0.05, as determined by a two-tailed unpaired t test on day 5. C, Representative Western immunoblot depicting ETS1 protein levels in H9 hNSCs expressing FOXR2 + vector control, FOXR2 + ETS1^sg2_KRAB, HcRed + vector control, and HcRed + ETS1^sg2_KRAB. D, Representative Western immunoblot depicting ETV3 protein levels in H9 hNSCs expressing FOXR2 + vector control, FOXR2 + ETV3^KRAB, HcRed + vector control, and HcRed + ETV3^KRAB. E, Schematic depicting FOXR2 protein domains. The MYC-binding domain is indicated in red and forkhead DNA-binding domain is indicated in green. F, Growth of neurospheres (measured as spheroid size) of mNSCs expressing FOXR2 WT relative to mNSCs expressing ΔMYC, ΔForkhead, and control, and normalized to day 0 values. Values indicate mean ± SEM across three replicate experiments. **, P < 0.01, as determined by two-tailed unpaired t tests. G, Representative Western immunoblot depicting HA-tagged FOXR2 protein levels in mNSCs expressing FOXR2 WT, ΔMYC, and ΔForkhead. H, Quantification of HA protein expression in Western immunoblots performed using lysates derived from mNSC-transduced FOXR2 WT, ΔMYC, and ΔForkhead. Values from three independent experiments and their mean ± SEM are shown. ns, not significant as determined by a two-tailed unpaired t test with P > 0.05. I, Percentage of GFP-positive cells that are Ki67-positive cells in control, FOXR2 WT, ΔMYC, and ΔForkhead expressing IUE tumors. Values represent Ki67% positivity in four independent tumors per condition. Black horizontal line indicates the mean across all four tumors, with error bars representing SEM. **, P < 0.01; ***, P < 0.001 as determined by two-tailed unpaired t tests.