High-Confidence Differentially Expressed (DE) lncRNAs in ESCC.

To identify a set of lncRNAs dysregulated in ESCC, RNA-seq performed from paired cancerous and noncancerous tissues of 13 ESCC patients in the Yonsei Severance Hospital (YSH) cohort (Dataset S1) and was subjected to transcriptome assembly, and lncRNA annotations using our computational pipeline (27) (SI Appendix, Fig. S1). In total, 6,411 lncRNAs from 4,842 annotated loci and 1,924 from 1,657 unannotated loci were identified (Dataset S2). Of the total of 8,335 lncRNAs, 465 (305 up-regulated and 160 down-regulated) were significantly dysregulated in ESCC, exhibiting greater than 2-fold differences in expression, with a false discovery rate (FDR) ≤ 0.01 (Fig. 1A and Dataset S3). Then, to identify an ethnically independent set of DE lncRNAs in ESCC, 113 DE lncRNAs commonly dysregulated in all 3 ESCC cohorts (YSH, Genotype-Tissue Expression [GTEx] + The Cancer Genome Atlas [TCGA], and Chinese) were selected (Fig. 1A). Of the 113 confidence DE lncRNAs, 20 were newly annotated, 32 were up-regulated, and 81 were down-regulated in ESCC (SI Appendix, Fig. S2 A–C). A majority of the confidence DE lncRNA genes were either located in intergenic regions or were antisense to other genes (Fig. 1B); their genomic and clinical features, such as subcellular localization of the lncRNA (Fig. 1C), associations with enhancers (Fig. 1D), and DNA methylation (Fig. 1E), were systematically examined (Dataset S4 A–D). As previously reported, many overlapped with enhancers (Fig. 1D) and seemed to be associated with epigenetic markers of other genes (Fig. 1E).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Characterization of a common set of DE lncRNAs in ESCC. (A) Venn diagram of the DE lncRNAs detected in 3 ESCC cohorts (YSH, Chinese, and GTEx+TCGA). The pie charts (B–F) show the number of DE lncRNAs categorized according to their genomic location (B), subcellular localization (C), enhancer overlap (D), association with DNA methylation (E), and prognostic power (F). (G) The 4 ESCC subclasses based on the 6 prognostic marker lncRNAs in F. The top section presents cohort information, clinical history, pathological features, and survival information from the YSH and TCGA ESCC patients. The various categories are represented as different colors, as shown in the legend on the right. OS is overall survival. The expression patterns of the 6 prognosis-related lncRNAs in the RNA-seq datasets are shown with a colored heatmap in the bottom section (red indicates the top 33% highly expressed lncRNAs associated with a HHR; blue indicates the top 33% highly expressed lncRNAs associated with a LHR).

Six lncRNAs Define 4 ESCC Subclasses with Different Clinical Outcomes.

To find clinically relevant lncRNAs that are associated with survival outcomes of patients, Kaplan–Meier survival analysis for all 113 DE lncRNAs were performed with TCGA ESCC datasets comprising 95 patient samples (Fig. 1F and Dataset S4E). Six DE lncRNAs were significantly associated with survival rates, 2 (HERES and RP11-1L12.3) of which were associated with a high hazard ratio (HHR; P ≤ 0.05) and 4 (RP11-114H23.1, RP11-114H23.2, CTD-2319I12.1, and LINC00330) of which were associated with a low hazard ratio (LHR; P ≤ 0.05). To delineate how the expression of the 6 lncRNAs stratifies ESCC patients, we clustered samples from the TCGA and YSH cohorts including additional 10 RNA-seq samples where the clinical values were available based on the binary expression patterns (high and low) of the 6 lncRNAs, revealing 4 distinct classes of patients: class L1∼L4 (Fig. 1G and Dataset S5). Noticeably, class L1, in which only the HHR markers are highly expressed, showed a worse survival rate than class L3 (P < 0.05) and the other classes (P = 0.01; Fisher’s exact test). Class L3 tended to display a greater overall survival rate than other classes (P < 0.05; Fisher’s exact test). Importantly, class L1 appeared to be significantly associated with smoking (P < 0.05; Fisher’s exact test), compared to other classes. Taken together, these results indicate that the 6 lncRNAs represent prognostic signature genes that can stratify ESCC patients based on clinical outcomes.

HERES Encodes Alternative Splicing Isoforms, Up-Regulated in ESCCs.

Since HERES, 1 of the HHR markers, was greatly up-regulated in ESCCs compared to paired adjacent noncancerous samples (SI Appendix, Fig. S3) and most strongly associated with poor vital status (Fig. 1G), we investigated whether HERES might be an ESCC-driving lncRNA. The lncRNA gene encodes 2 isoforms with the same transcription start sites (TSSs) (28) and cleavage and polyadenylation sites (CPSs) (29) (Fig. 2A). Isoform #1, HERES.1, is 2,160 nt and contains 2 exons, whereas isoform #2, HERES.2, is an intron-retained, single-exonic transcript that is 6675 nt. Both isoforms were confirmed to lack coding potential (Fig. 2A). Analysis of ESCC RNA-seq datasets (SI Appendix, Fig. S4A) and qRT-PCR in an ESCC cell line (KYSE-30) (Fig. 2 A and B) showed that HERES.1 is the major isoform. Only a short region in HERES exon 1 displays sequence conservation with a region in the mouse genome; the intergenic HERES locus between the GLS and NAB1 genes on chromosome 2 shows synteny conservation in mouse (SI Appendix, Fig. S4B).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

HERES is a highly expressed lncRNA in ESCC. (A) The HERES genomic locus with CAGE-seq and 3P-seq signals. qRT-PCR primer sets were designed to recognize exonic (2) and intronic regions (3). The coding potentials calculated by CPC and CPAT are indicated on the right. (B) qRT-PCR results using the 5 primer sets in KYSE-30 cells. (C) The HERES expression level was measured in 5 ESCC cell lines and a normal esophageal cell line (Het-1A). (B and C) Error bars represent the mean ± SD from 3 independent experiments. (D) The box plots show the HERES expression levels in normal, noncancerous, and cancerous tissues from the YSH cohort (paired), the Chinese cohort (paired), and the TCGA cohort. (E) HERES expression levels measured by qRT-PCR in additional frozen tissue samples including YSH ESCC (n = 66) and adjacent samples (noncancer; n = 66) (Left) and in normal mucosa tissues (n = 21) from reflux symptom patients (Right). (F) Survival analyses of YSH and TCGA patients from whom the ESCC samples were obtained based on the HERES expression level. **P ≤ 0.01, ***P ≤ 0.001.

Elevated HERES expression was then validated in other ESCC cell lines. Compared to that in a normal esophageal epithelial cell line (Het-1A), HERES expression appeared to be up-regulated greater than 10-fold in all tested ESCC cell lines (Fig. 2C), as observed in ESCC samples (Fig. 2D). HERES was significantly up-regulated not only in ESCC, but also in esophageal adenocarcinoma (ESAD) and other squamous carcinomas (head and neck, and lung squamous cell carcinoma; HNSC and LUSC), but not in lung adenocarcinoma (LUAD) (SI Appendix, Fig. S4C). These expression changes were further confirmed in 66 ESCC samples from the YSH cohort using qRT-PCR, which revealed an elevated level of HERES in cancers compared to adjacent noncancerous samples (Fig. 2 E, Left). HERES expression in the adjacent noncancerous samples was higher than that in normal mucosa from the normal population (Fig. 2 E, Right). As observed in the YSH cohort (Fig. 2 F, Left and Dataset S6), HERES levels were significantly correlated with stage-free survival rates in the TCGA ESCC cohort (Fig. 2 F, Right), and multivariate analysis with clinical information revealed that the HERES level was strongly associated with tumor grade (P = 0.004; Fisher’s exact test) but not with other clinicopathological factors (SI Appendix, Fig. S4D).

HERES Promotes Cell Proliferation, Migration, Invasion, and Colony Formation.

To investigate whether HERES is involved in cancer development and progression, the effects of HERES knockdown on cell proliferation, migration, invasion, and colony formation were explored with siControl- and siHERES-treated cells. Introduction of an siRNA targeting HERES (siHERES_1 or siHERES_2) to KYSE-30 and HCE-7 cells significantly reduced HERES expression (SI Appendix, Fig. S5). The proliferation indices (optical density [O.D.] values) were significantly reduced in both siHERES_1 and siHERES_2-treated cells compared to siControl-treated cells (Fig. 3A), and the introduction of a HERES pcDNA expression construct partly rescued the proliferation activity (Fig. 3A), indicating that HERES can regulate cell proliferation. Migration and invasion assays showed that both cell migration and invasion were greatly reduced in siHERES-treated cells compared to siControl-treated cells (Fig. 3 B and C). In addition, HERES knockdown also reduced colony formation measured at 14 d after siRNA transfection (Fig. 3D). A role for HERES in malignant ESCC progression was confirmed by the reduction of N-cadherin and vimentin levels in siHERES-treated cells and by the rescue of these levels by the introduction of the HERES pcDNA construct to the cells (Fig. 3E and SI Appendix, Fig. S6). These results suggest that HERES can promote cancer progression and metastasis.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

HERES modulates cell proliferation, migration, invasion, and colony formation. (A) Cell viability was measured using an MTS assay in KYSE-30 and HCE-7 cells transfected with siControl (NC), siHERES (si_1 and si_2), or siHERES followed by pcDNA-HERES (si_1+pcDNA, si_2+pcDNA). Growth curves were compared between siHERES- and siControl-transfected cells, and between pcDNA-HERES+siHERES- and siHERES-transfected cells. Wound healing assays (B), invasion assays (C), and colony formation assays (D) were performed in KYSE-30 and HCE-7 cells after HERES knockdown (all assay images at 4× magnification). The bar graphs represent the frequency of wound closure (B) and the number of invading cells (C) and colonies formed (D). Data represent the mean ± SD from 3 independent experiments (A–D). (E) Expression of EMT markers in KYSE-30 and HCE-7 cells transfected with siControl or the indicated combinations of siHERES and pcDNA-HERES as determined by immunoblot. *P ≤ 0.05, **P ≤ 0.01.

HERES Affects Wnt Signaling Pathway-Related Genes.

To study the means by which HERES promotes cancer development and progression, the changes in expression of ∼730 cancer-related pathway genes were analyzed in siHERES-treated and siControl-treated KYSE-30 cells using the NanoString nCounter PanCancer Pathways Panel (Fig. 4A and Dataset S7). Seventy-seven cancer-related genes (34 for up-regulation and 43 for down-regulation) were dysregulated greater than 2-fold in siHERES-treated cells compared to siControl cells (Fig. 4A); the expression changes of the 2 most up-regulated genes (CACNA2D3 and SFRP2) and the 2 most down-regulated genes (BMP7 and GRIN1) in this group were confirmed by qRT-PCR (SI Appendix, Fig. S7 A–D). Noticeably, among the genes dysregulated by HERES reduction, 14 belong to the Wnt signaling pathway, and half of the 10 most up-regulated genes (CACNA2D3, SFRP2, CACNA1E, CXXC4, and SFRP4) are involved in the Wnt signaling pathway (Dataset S7). CACNA2D3, which encodes a subunit of the calcium channel protein complex, was previously shown to be induced in ESCC (30) and other cancers (31, 32) via epigenetic mechanisms, and its down-regulation led to inactivation of Wnt/Ca²⁺ signaling pathway (32). SFRP2 encodes a member of the SFRP family that modulates the Wnt signaling pathway; SFRP2 hypermethylation is known to enhance cell invasiveness in both cancers and noncancerous diseases (33, 34). The enrichment of canonical and noncanonical Wnt signaling pathway-related genes among the genes that respond to HERES depletion, together with results from previous studies, suggest that HERES may regulate cancer development via control of Wnt signaling pathways.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

HERES epigenetically regulates genes involved in canonical and noncanonical Wnt signaling pathways. (A) Changes in the expression of cancer-related genes in response to siHERES treatment compared to siControl are shown. The colored circles indicate genes that are up-regulated (red) or down-regulated (blue) under HERES-depleted conditions. Changes in the expression of the highlighted genes were experimentally confirmed by qRT-PCR. (B) HERES expression in the nuclear and cytoplasmic fractions of KYSE-30 cells as determined by qRT-PCR. (C) Log-scaled fold-changes of expression (x axis) and DNA methylation (y axis) of each gene in the HERES-high versus HERES-low sample groups from the TCGA dataset. The red dots indicate DE genes in ESCC. The highlighted genes are those for which there is anti-correlation between expression and DNA methylation. DEG: DE gene. N.S.: not significant. (D) The CACNA2D3 genomic locus with CpG island tracks, DNA methylation (beta value), and RNA expression (read count) in the HERES-high and HERES-low groups. (E) CACNA2D3 and LDOC1 DNA methylation patterns (methylation [M] and unmethylation [UM]) were measured by MS-PCR in KYSE-30 cells transfected with siControl or siHERES. (F) Immunoblots of histone modification markers in siControl- or siHERES-transfected KYSE-30 cells. (G) ChIP-qPCR analysis of the H3K27me3 level of CACNA2D3 in siControl- or siHERES-transfected KYSE-30 cells. Data represent the mean ± SD from 3 independent experiments. (H) Immunoblots of the products of the 3 HERES target genes (CACNA2D3, SFRP2, and CXXC4), components (β-catenin and NLK) of the Wnt/β-catenin and Wnt/Ca²⁺ signaling pathways, and downstream proteins associated with cell proliferation (CMYC and CCND1), invasion (MMP7), and EMT (SNAIL).

HERES Regulates Wnt Signaling Pathways at the Epigenetic Level.

As HERES appeared to be enriched in the nucleus rather than the cytoplasm (Fig. 4B and SI Appendix, Fig. S2C) and nucleus-localized lncRNAs are often reported to be epigenetic regulators, a potential epigenetic role for HERES was investigated by analyzing publicly available array-based DNA methylation and RNA-seq data from TCGA ESCC samples. Based on their HERES expression level, the ESCC samples were first divided into 2 subgroups, HERES-high and HERES-low, and the changes in expression and DNA methylation of the protein-coding genes were then compared between the subgroups (Fig. 4C and Dataset S8). Of the genes, CACNA2D3 and LDOC1 (Fig. 4D and SI Appendix, Fig. S7E) were down-regulated and hypermethylated in the HERES-high group, whereas EPSTI1, SLC15A3, and BST2 were up-regulated and hypomethylated in the HERES-high subgroup. The expression and DNA methylation changes were confirmed in HERES-depleted KYSE-30 cells by qRT-PCR and methylation-specific (MS) PCR. Only 2 down-regulated genes (CACNA2D3 and LDOC1) were confirmed to have both expression and DNA methylation changes (Fig. 4E and SI Appendix, Fig. S7 F–L). On the other hand, although SFRP2 and CXXC4 did not display DNA methylation changes in the analysis of the TCGA ESCC samples, the expression and DNA methylation signals of the Wnt signaling-related genes were changed similarly to those of CACNA2D3 and LDOC1 in the siHERES-treated cells compared to the siControl-treated cells (SI Appendix, Fig. S7B and S8 A–C).

Because DNA methylation is often associated with histone modifications (35, 36), global changes in histone modification markers in response to HERES knockdown were examined, revealing a marked decrease in H3K27me3 levels (Fig. 4F). We then investigated where the H3K27me3 signal was depleted in the genomic regions of 3 Wnt signaling pathway genes in siHERES-treated cells using chromatin immunoprecipitation (ChIP)-qPCR analysis. Significantly reduced H3K27me3 signals were observed at specific sites in the genes (recognized by primer 5 for CACNA2D3, primers 3, 4, and 5 for SFRP2, and primers 9 and 10 for CXXC4) in siHERES-depleted cells (Fig. 4G and SI Appendix, Fig. S8 D and E).

Previous studies reported that CACNA2D3 down-regulation inhibited the noncanonical Wnt/Ca²⁺ signaling pathway by decreasing the intracellular calcium level and NLK expression (32) and that SFRP2 and CXXC4 play roles as negative regulators of the canonical Wnt signaling pathway (34, 37). We thus examined changes in the expression of 2 Wnt signaling-related factors, NLK and β-catenin, in siHERES-treated cells (Fig. 4H and SI Appendix, Fig. S9). As expected, HERES reduction increased the NLK level and decreased β-catenin in KYSE-30 and HCE-7 cells. In addition, changes in the expression of Wnt downstream targets were also confirmed (Fig. 4H and SI Appendix, Fig. S9). In contrast, introducing the pcDNA-HERES construct to cells reverted the expression levels of NLK, β-catenin, and Wnt downstream targets (Fig. 4H and SI Appendix, Fig. S9). Taken together, these results suggest that HERES down-regulation in cancers perturbs and promotes canonical and noncanonical Wnt signaling pathways via epigenetic regulation, resulting in the inhibition of cancer progression.

Inhibition of HERES Arrests the Cell Cycle at the G1/S Checkpoint and Induces Apoptosis.

Because 2 of the downstream targets of HERES, CCND1 and CACNA2D3, are known to regulate the cell cycle and apoptosis (30, 32), the effect of the loss of HERES on the cell cycle and apoptotic processes was examined. Cell counting showed that siHERES-treated cells were arrested at G0/G1 (Fig. 5A). Flow cytometry showed that siHERES-treated cell populations exhibited significantly increased levels of apoptosis compared to siControl-treated cells (Fig. 5B). An induction of apoptotic factors, such as cleavage of poly (ADP ribose) polymerase (PARP), cleaved caspase-9, and Bax, and a reduction of the anti-apoptotic factor, Bcl-2, were confirmed in siHERES-treated cells. However, the rescue of HERES expression reverted the expression of these factors to levels in control cells (Fig. 5C).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

The effect of HERES on the cell cycle and apoptosis. Cell cycle (A) and apoptosis (B) assays were performed on siRNA-transfected KYSE-30 and HCE-7 cells. (A) Cell cycle analysis of siRNA-transfected KYSE-30 and HCE-7 cells by flow cytometry. The bar graph shows the percentage of cells in sub-G0, G1, S, and G2 phases in siRNA-transfected KYSE-30 and HCE-7 cell populations. (B) Apoptosis was measured by flow cytometry using PI/Annexin V staining. The bar graph represents the percentage of apoptotic cells in each population. Data represent the mean ± SD from 3 independent experiments. (C) Apoptosis markers were assessed by immunoblot in KYSE-30 and HCE-7 cells transfected with siControl or siHERES and/or pcDNA-HERES. *P ≤ 0.05, **P ≤ 0.01.

HERES Interacts with EZH2 to Regulate CACNA2D3.

We then asked how HERES regulates the expression of target genes at the epigenetic level. To address this question, binding sites for possible epigenetic modulators that can drive the histone methylation of target genes were first examined using publicly available ChIP-seq datasets from the ENCODE project. We found that all 3 HERES target genes contained enhancer of EZH2 binding sites in their promoter regions (SI Appendix, Fig. S10). Because EZH2, a subunit of the PRC2, has a well-known role in histone methylation to generate H3K27me3 and is known to interact with nuclear lncRNAs (38), we suspected that EZH2 would be a binding partner of HERES. To examine the molecular relationship between HERES and EZH2, EZH2 RNA and protein levels were quantified in siControl- and siHERES-treated KYSE-30 cells, showing that HERES reduction decreased the EZH2 protein level but not the RNA level (Fig. 6A and SI Appendix, Fig. S11A). Subsequently, RNA immunoprecipitation (RIP) (Fig. 6B) and EZH2 IP (Fig. 6C) assays showed the interaction of HERES and CACNA2D3 with EZH2. To validate a direct interaction between HERES and PRC2-EZH2, we then searched for PRC2-EZH2 binding motifs in the HERES sequence. Because the PRC2 complex including EZH2 is known to be recruited by G-rich motifs (39), we scanned for G-rich regions in HERES transcripts, leading to the identification of 6 regions including 2 potential g-quadruple structure motifs (SI Appendix, Fig. S11B). EZH2-IP and qRT-PCR showed that a single region with 4 GGW repeats (index 1) was significantly enriched in EZH2-IP (SI Appendix, Fig. S11C). To further investigate if the HERES GGW repeat sequence (index 1) is necessary for an interaction with EZH2, we constructed a plasmid vector that harbors a HERES sequence that lacks the GGW repeat region (HERES-Mut) (Fig. 6D). A RIP assay confirmed that EZH2 failed to interact with HERES-Mut in KYSE-30 cells (Fig. 6E). RNA fluorescence in situ hybridization (FISH) of HERES and fluorescein isothiocyanate (FITC) staining of CACNA2D3 validated that HERES was principally localized to the nucleus and that the GGW sequence (index 1) is necessary for the interaction with EZH2 to down-regulate CACNA2D3 (Fig. 6F). Cells transfected with HERES-Mut exhibited significantly increased CACNA2D3 at both the RNA and protein level, whereas HERES overexpression (pcDNA-HERES) reduced the CACNA2D3 level (Fig. 6 G and H).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

HERES regulates CACNA2D3 via interaction with EZH2. (A) Immunoblots of EZH2 and DNMT1 in KYSE-30 cells transfected with either siControl or siHERES. (B) RIP assays were performed with anti-EZH2 in KYSE-30 cell lysates. The quantity of HERES in the cell lysates (input) and the immunoprecipitates was measured by qRT-PCR. (C) IP assays were performed with anti-EZH2 in KYSE-30 cell lysates. The quantity of CACNA2D3 in the cell lysates (input) and the immunoprecipitates was measured by immunoblot. (D) Representation of the WT and mutated (HERES-Mut) HERES sequences used for IP with anti-EZH2. HERES-Mut contains a deletion of the G-rich sequence (index 1) presented in SI Appendix, Fig. S11B. (E) RIP assays were performed with anti-EZH2 in lysates of KYSE-30 cells transfected with either pcDNA-HERES or HERES-Mut (Upper). The bar graph shows the relative amount of HERES after anti-EZH2 IP using lysates of cells transfected with either pcDNA-HERES or HERES-Mut (Lower). (F) RNA FISH (40× magnification) to visualize HERES (red) and FITC staining of CACNA2D3 (green) in KYSE-30 cells transfected with pcDNA (Upper), pcDNA-HERES (Middle), or HERES-Mut (Lower). (Scale bar, 20 µm.) Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue). CACNA2D3 RNA (G) and protein (H) levels were measured in KYSE-30 cells transfected with pcDNA, pcDNA-HERES, or HERES-Mut by qRT-PCR and immunoblot, respectively. Data represent the mean ± SD from 3 independent experiments (B, E, and G). **P ≤ 0.01.

HERES as a Candidate Therapeutic Target.

To investigate whether HERES controls tumor growth in vivo, we carried out xenograft assays with siControl- and siHERES-treated cancer cell lines (Fig. 7). Both the volume and weight of tumors derived from HERES-depleted samples were significantly reduced compared to tumors derived from control cells 4 wk after the injection (Fig. 7 A–C). We further examined changes in the expression of HERES and its target genes, finding that the reduction of HERES was maintained for 4 wk after siHERES injection (SI Appendix, Fig. S12), whereas the levels of HERES targets were significantly increased in tumor samples derived from HERES-depleted cells compared to control cells (Fig. 7D). We also confirmed that global H3K27me3 and EZH2 levels were decreased in HERES-depleted tumor samples (Fig. 7D), suggesting that HERES is a promising candidate therapeutic target that controls tumor growth through the regulation of canonical and noncanonical Wnt signaling pathways in vivo (Fig. 7E).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 7.

Expression of HERES regulates tumorigenicity in xenograft models. KYSE-30 cells transfected with either siControl or siHERES were injected into nude mice (6 mice for each group). The resulting xenograft tumor volumes (A and B) and weights (C) are shown. (A) Tumor growth curves showing that tumors in the siHERES group grew markedly slower than those in the siControl group. (B) Images of tumor volumes from the xenograft models. (C) Tumor weights in the siHERES and siControl groups 4 wk after cell injection. Data represent the mean ± SD. (D) Immunoblot analysis (#1 and #2) of levels of key components (CACNA2D3, SFRP2, and CXXC4) from the canonical and noncanonical Wnt signaling pathways and of H3K27me3 and EZH2 in the siHERES and siControl xenograft models. (E) A graphic illustration of HERES-regulated canonical and noncanonical Wnt signaling pathways in ESCC. *P ≤ 0.05.