Long Noncoding RNA LBCS Inhibits Self-Renewal and Chemoresistance of Bladder Cancer Stem Cells through Epigenetic Silencing of SOX2

With a high rate of recurrence and tumor heterogeneity, bladder cancer is one of the most common and lethal malignancies worldwide (1, 2). Chemotherapy is a significant component of current first-line treatment for bladder cancer. It reduces tumor masses in most patients initially; however, a majority of patients progressively become unresponsive after multiple treatment cycles, and ultimately suffer tumor relapse (3, 4). Recent studies demonstrate that the main cause of chemoresistance is existence of cancer stem cells (CSCs), which have a survival advantage in response to chemotherapy (4, 5). Therefore, CSCs are responsible for the initiation, propagation, metastasis, chemoresistance, and relapse of cancers (3, 5, 6). Previous studies have verified the existence of bladder cancer stem cells (BCSCs; refs. 7, 8), which can be distinguished by various biomarkers and particular characteristics, such as CD44 (7), high aldehyde dehydrogenase (ALDH) activity (9), and sphere formation (10). Given the critical role of CSCs in tumorigenesis, intervention to block CSCs' self-renewal could be a potential strategy for tumor treatment. However, the biological characters and molecular mechanisms of chemotherapy-induced BCSCs remain largely unknown.

Long noncoding RNAs (lncRNA) are transcripts longer than 200 nucleotides without protein-coding capacity (11). Accumulating evidence indicates that lncRNAs are involved in diverse physiologic and pathologic progresses, including embryonic development, organ formation, and tumorigenesis (12, 13). Aberrant expression of lncRNAs has been observed in many types of cancers and may play important roles in regulating proliferation, chemoresistance, and metastasis of cancer cells (14, 15). Recent studies show that lncRNAs are key regulators of self-renewal in CSCs of liver, renal, and colon cancer, such as lncTCF7 (16), lncARSR (17), and RBM5-AS1 (18). However, whether and how lncRNAs regulate the self-renewal and chemoresistance of BCSCs remains largely unknown.

In the present study, we discovered a novel lncRNA termed Low expressed in Bladder Cancer Stem cells (lnc-LBCS), which is significantly downregulated in BCSCs and inhibits self-renewal and chemoresistance of BCSCs in vitro and in vivo. Interestingly, lnc-LBCS guides heterogeneous nuclear ribonucleoprotein K (hnRNPK)–enhancer of zeste homolog 2 (EZH2) complex to suppress SOX2 expression by mediating H3K27me3 of SOX2 promoter, leading to inhibition of BCSCs’ self-renewal and chemoresistance.

The human bladder cancer cell lines (UM-UC-3, 5637, HT-1376, and J82) and human embryonic kidney (HEK-293T) cells were purchased from the American Type Culture Collection from 2013 to 2015. UM-UC-3, HT-1376, J82, and HEK-293T cells were cultured in DMEM (Gibco), whereas 5637 cells were cultured in RPMI 1640 (Gibco). All medium was supplemented with 10% FBS (Shanghai ExCell Biology) and 1% penicillin/streptomycin (Gibco). Cells were cultured in a humidified atmosphere of 5% CO₂ at 37°C. These cells were characterized using short tandem repeat markers and were confirmed to be mycoplasma-free (last tested in 2018).

A total of 120 pair snap-frozen fresh bladder cancer tissues and normal adjacent tissues (NAT), termed Cohort 1, were obtained by surgery with the written consent of patients who underwent surgery at Sun Yat-sen Memorial Hospital, Sun Yat-sen University. The workflow of patients was selected in Cohort 1 (Supplementary Fig. S17). Three tissue microarrays containing 86 bladder cancer specimens and 20 normal tissues, termed Cohort 2, were purchased from US Biomax (catalogue numbers BL244, BC12011b, and T124b). All the samples were pathologically confirmed as transitional cell carcinoma of the bladder by two pathologists. Ethical consent was approved by Sun Yat-sen University's Committees for Ethical Review of Research involving Human Subjects. The study was conducted in accordance with the International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS). The characteristics and clinicopathologic features of the patients are listed in Supplementary Table S1.

All animal studies were conducted with the approval of the Sun Yat-sen University Institutional Animal Care and Use Committee and were performed in accordance with established guidelines. Male BALB/c nude mice (4–5 weeks old) were purchased from the Experimental Animal Center of Sun Yat-sen University and housed in SPF barrier facilities. The tumor-initiating capacity assay and in vivo chemotherapy assay were detailed in the Supplementary Materials and Methods section. The size of the tumor was measured every 3 days. The mice were euthanized, and tumors were dissected surgically at the end of experiment. The tumor specimens were fixed in 4% paraformaldehyde.

The LncRNA+mRNA Human Gene Expression Microarray (CapitalBio) was used to investigate the differentially expressed lncRNAs in UM-UC-3 4th spheres, spheres readhered for 10 hours, and 1 day, and adherent UM-UC-3 cells. All primary data in the microarray analysis have been uploaded to the Gene Expression Omnibus with the accession number GSE107857.

Patients’ clinical profiles in The Cancer Genome Atlas (TCGA) bladder cancer cohort are available at https://cancergenome.nih.gov/ (19). The expression of the TCGA lnc-LBCS cohort comprising 209 patients was obtained from TANRIC (ref. 20; http://ibl.mdanderson.org/tanric/_design/basic/query.html), but patients with no available clinical data were excluded from the analysis. So a total of 185 patients were used for further analysis. The characteristics and clinicopathologic features of the patients are listed in Supplementary Table S1. For the overall survival (OS) and disease-free survival (DFS) analysis, the Kaplan–Meier survival analysis of SOX2 in 402 cases in a TCGA cohort was obtained from GEPIA (ref. 21; http://gepia.cancer-pku.cn/index.html).

Sphere culture was conducted as previously described (22, 23) and as detailed in the Supplementary Materials and Methods section. Cells were assayed for ALDH activity using the Aldefluor Kit (Stem Cell Technologies) according to the manufacturer's instructions.

The chemosensitivity assay and apoptosis analysis were performed as described previously (23–25) and as detailed in the Supplementary Materials and Methods section. Caspase-3/7 activity was measured using a Caspase-Glo 3/7 Assay kit (Promega) as previously described (23, 24). The terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) assay was conducted using the In Situ Cell Death Detection Kit (Roche), following the manufacturer's instructions and as described previously (25).

Lnc-LBCS expression was also examined using ISH in formalin-fixed, paraffin-embedded (FFPE) samples, as previously described (26) and as detailed in the Supplementary Materials and Methods section. The IHC analyses and score calculation were conducted as described previously (23, 24). Anti-SOX2 antibodies (1:500) were used to detect the expression of SOX2 in bladder cancer tissues. The expression of lnc-LBCS and SOX2 in bladder specimens was quantified by using the histochemical score (H-score) as described previously (23, 24). The staining intensity was graded as follows: 0 (no staining), 1 (weak staining, light yellow for IHC, light blue for ISH), 2 (moderate staining, brown for IHC, moderate blue for ISH), and 3 (strong staining, brown red for IHC, strong blue for ISH). The intensity of staining was multiplied by the percentage of positive cells (0%–100%), and the H-score (0–300) of each tissue was obtained for statistical analysis. The samples were classed as low (score < 50) or high (score ≥ 50) lnc-LBCS expression. The samples were classed as low (score < 200) or high (score ≥ 200) SOX2 expression. Anti-Ki67 antibodies (1:1,000) were used to detect the expression of Ki67 in mouse tumors. The score of ISH and IHC in the FFPE samples was blindly quantified by two pathologists, and the average H-score (0–300) of each tissue was obtained for statistical analysis.

RNA isolation and qRT-PCR were performed as previously described (25, 27). Relative expression was calculated using the 2^−ΔΔCt method (Ct, cycle threshold). All specific primers are listed in Supplementary Table S10. Western blotting was performed as previously described (25, 27). The information of antibody was detail in the Supplementary Materials and Methods section. The full images of all Western blots are shown in Supplementary Figs. S18 and S19.

Lnc-LBCS full-length sense, antisense, and serial deletion sequences were prepared via in vitro transcription using a TranscriptAid T7 High Yield Transcription Kit. The RNA pulldown assay was performed using a Magnetic RNA-Protein Pull-down kit according to the manufacturer's instructions. The samples were separated using electrophoresis, and lnc-LBCS–specific bands were identified using mass spectrometry and retrieved from a human proteome library. The RIP was performed as described previously (25) and as detailed in the Supplementary Materials and Methods section.

The chromatin isolation by RNA purification (ChIRP) was conducted using a Magna ChIRP RNA Interactome kit (Millipore) according to the manufacturer's instructions and as described previously (25). Chromatin immunoprecipitation (ChIP) was conducted using an EZ-Magna ChIP A/G kit (Millipore) according to the manufacturer's instructions and as previously reported (23–25). The method of ChIRP and ChIP was detail in the Supplementary Materials and Methods section.

Quantitative data were presented as the mean ± the SD of three independent experiments. Differences between two groups were analyzed with the unpaired/paired Student t test (two-tailed tests), and one-way ANOVA followed by Dunnett multiple comparisons tests was performed when more than two groups were compared. Data of clinical analysis were shown as median with the interquartile range. The Mann–Whitney U test was used for independent samples when the population could not be assumed to be normally distributed. Pearson χ² test was used to analyze the clinical variables. Spearman correlation analysis was performed to determine the correlation between two variables. Cumulative survival time was calculated using the Kaplan–Meier method and analyzed by the log-rank test. The best point cutoff value was used to define lnc-LBCS expression level (low vs. high) for all survival analyses in this study. A multivariate Cox proportional hazards model was used to estimate the adjusted HRs and 95% confidence intervals, and to identify independent prognostic factors. All statistical analyses in this study were performed using SPSS 19.0 software. A P value < 0.05 was considered significant.

Previous studies have demonstrated that CSCs, such as breast and bladder CSCs, are selectively enriched after chemotherapy via enhanced survival (4, 28). We took advantage of this finding to determine if we could enrich BCSCs by consecutively passaging bladder cancer cells in NOD/SCID mice treated with chemotherapy. We obtained the cells termed UM-UC-3 4th from four chemotherapy passages in vivo (Supplementary Fig. S1A). To study whether UM-UC-3 4th cells were resistant to chemotherapy, we performed MTT assays, flow cytometry, and caspase-3/7 activity assays in UM-UC-3 4th and parental UM-UC-3 cells. Notably, UM-UC-3 4th cells displayed higher chemotherapy resistance and a higher IC₅₀ value, and a lower apoptosis rate and caspase-3/7 activity than the parental UM-UC-3 cells when treated with gemcitabine or cisplatin (Supplementary Fig. S1B–S1D). This suggested that UM-UC-3 4th cells were chemoresistant bladder cancer cells.

To evaluate whether UM-UC-3 4th cells had a stronger self-renewal capacity, we performed a sphere formation assay in vitro and a tumor initiation assay in vivo. UM-UC-3 4th cells formed more and larger spheres than UM-UC-3 cells (Supplementary Fig. S1E and S1F). A serial sphere formation assay showed that UM-UC-3 4th cells had a stronger long-term self-renewal over four generations of sphere formation compared with UM-UC-3 cells (Supplementary Fig. S1G). Tumor initiation assay showed that UM-UC-3 4th cells had a stronger tumorigenic capacity and contained remarkably higher ratios of BCSCs than parental UM-UC-3 cells (Supplementary Fig. S1H–S1J). Furthermore, UM-UC-3 4th-formed spheres contained more ALDH⁺ cells, a functional marker of BCSCs, and increased expression of OCT4, SOX2, and NANOG, which are key stem cells markers (Supplementary Fig. S1K and S1L). Therefore, we used the UM-UC-3 4th spheres as BCSCs for further study.

To identify lncRNAs involved in BCSCs, we conducted a transcriptome microarray analysis of UM-UC-3 4th spheres, spheres readhered for 10 hours, 1 day, and adherent UM-UC-3 cells. As shown in Fig. 1A, 614 lncRNAs decreased or increased in UM-UC-3 4th spheres compared with readhered spheres, including CASC2 (29), XIST (30), and OVAAL (31), which are associated with human cancer progression. We then further examined the expression of 5 most decreased lncRNAs and 5 most increased lncRNAs in a series of BCSCs models, including UM-UC-3 4th spheres differentiation, four other bladder cancer spheres, UM-UC-3, and 5637 ALDH⁺ cells. The expression levels of three lncRNAs were closely correlated with a series of BCSCs’ models (Fig. 1B–D). Moreover, sphere formation assay was performed to identify which lncRNA plays a key role in self-renewal of BCSCs using lentivirus-mediated shRNAs (Supplementary Fig. S2A). Finally, we identified a novel lncRNA (ENSG00000228412), termed lnc-LBCS, that was significantly downregulated in BCSCs and affected sphere formation (Fig. 1E).

Lnc-LBCS is located on human chromosome 6p22.3 and is modestly conserved among mammals (Supplementary Fig. S2B). In our assessment, the full-length lnc-LBCS transcript was 731 nt in the bladder cancer cell, which was examined using 5′ and 3′ rapid amplification of cDNA ends (RACE) and verified by Northern blotting (Supplementary Fig. S2C and S2D). In addition, lnc-LBCS displayed no protein coding potential (Supplementary Fig. S2E and S2F). Furthermore, cellular fractionation assays and RNA FISH showed that lnc-LBCS was mainly localized in the nuclei of bladder cancer cells (Supplementary Fig. S2G and S2H).

To investigate whether lnc-LBCS was involved in clinical bladder cancer progression, we detected and analyzed lnc-LBCS expression in three independent large-scale cohorts of bladder cancer specimens. Remarkably, lnc-LBCS was downregulated in bladder cancer tissues compared with normal tissues, and in high-grade compared with lower-grade tumors examined in a 120-case Cohort 1 (Fig. 1F and G). Moreover, statistical analysis revealed that lnc-LBCS expression was negatively correlated with pathologic grade and stage (P = 0.002 and 0.001, respectively, Supplementary Table S2). This result was further confirmed by analyses of a 185-case cohort in TCGA database and by ISH of an 86-case Cohort 2 (Fig. 1H–K; Supplementary Tables S3 and S4, and Supplementary Fig. S2I). Notably, lnc-LBCS displayed higher expression in neoadjuvant chemotherapy responders than nonresponders, suggesting that lnc-LBCS might be associated with chemoresistance (Fig. 1L). Furthermore, the Kaplan–Meier survival analysis showed that patients with low lnc-LBCS–expressing bladder cancers had significantly shorter OS and DFS in Cohort 1 and the TCGA Cohort (Fig. 1M–O). Multivariate analyses revealed that high lnc-LBCS expression was independent prognostic factor for OS and DFS in bladder cancer patients (Supplementary Tables S5 and S6). These data demonstrate that lnc-LBCS associates with cell differentiation of bladder cancer and may as a marker of good prognosis in bladder cancer.

Interestingly, further analyses of TCGA databases showed that lnc-LBCS expression was also significantly downregulated in other types of human cancer, such as lung, breast, stomach, and thyroid cancer, and head and neck squamous cell carcinoma (Supplementary Fig. S3A–S3F). High expression of lnc-LBCS correlated with good prognosis in many human cancers, such as lung cancer, head and neck squamous cell carcinoma, sarcoma, rectum adenocarcinoma, and uterine corpus endometrial carcinoma (Supplementary Fig. S3G–S3O), further supporting the tumor-suppressor role of lnc-LBCS in cancers.

To investigated the role of lnc-LBCS in BCSCs, we overexpressed lnc-LBCS in 5637 and BCSCs’ UM-UC-3 4th cells, whereas silenced lnc-LBCS in 5637 and non-BCSCs’ UM-UC-3 cells via lentivirus infection. The overexpression or knockdown efficiency was confirmed using qRT-PCR (Fig. 2A and B). Lnc-LBCS overexpression markedly decreased the number and size of spheres formation in UM-UC-3 4th and 5637 cells (Fig. 2C; Supplementary Fig. S4A). Serial sphere formation assays showed that lnc-LBCS overexpression inhibited long-term self-renewal over four generations of sphere formation (Supplementary Fig. S4B). Conversely, lnc-LBCS depletion significantly enhanced sphere formation in UM-UC-3 and 5637 cells (Fig. 2D; Supplementary Fig. S4C and S4D). Furthermore, flow cytometry analysis revealed that the population of ALDH⁺ cells was significantly decreased by lnc-LBCS overexpression, whereas it was increased by lnc-LBCS depletion (Fig. 2E and F; Supplementary Fig. S4E and S4F). The in vivo tumor initiation assay showed that lnc-LBCS–transduced cells had remarkably weaker tumorigenic capacity and lower ratios of BCSCs than control cells (Fig. 2G–I; Supplementary Fig. S4G; Supplementary Fig. S5A–S5C and S5G), suggesting the repressive function of lnc-LBCS in tumor initiation. Besides, lnc-LBCS–transduced cells showed reductive tumor propagation compared with control cells (Fig. 2G–I). In contrast, lnc-LBCS depletion contributed to much stronger tumor initiation and propagation, as well as higher tumorigenic cell frequency (Fig. 2J–L; Supplementary Fig. S4H; Supplementary Fig. S5D–S5F and S5H). To validate the results of lnc-LBCS knockdown, we also established lnc-LBCS KO cells using a CRISPR/Cas9 approach (Supplementary Fig. S6A and S6B). Lnc-LBCS KO cells displayed enhanced self-renewal capacity and tumor initiation of BCSCs in vitro and in vivo (Supplementary Fig. S6C–S6H), which agreed with the results of lnc-LBCS knockdown. Collectively, our results demonstrate that lnc-LBCS inhibits the self-renewal, tumor initiation, and propagation of BCSCs.

Emerging evidence shows that CSCs are the main cause of chemotherapy resistance (5, 6). Gemcitabine and cisplatin are the first-line chemotherapeutic drug for bladder cancer. Therefore, we explored whether lnc-LBCS plays an important role in chemoresistance of BCSCs using the MTT assay, flow cytometry, and caspase-3/7 activity. Interestingly, lnc-LBCS overexpression markedly reduced chemotherapy resistance and the IC₅₀ value, and increased the apoptosis rate and caspase-3/7 activity in UM-UC-3 4th and 5637 cells treated with gemcitabine or cisplatin, whereas the opposite outcome was observed after lnc-LBCS depletion (Fig. 3A–D; Supplementary Fig. S7A–S7C). Similar observations were obtained in the lnc-LBCS KO cells (Supplementary Fig. S9A–S9D). The ability of lnc-LBCS to regulate chemoresistance in BCSCs was further examined using an in vivo tumor model. Interestingly, after treated with GC chemotherapy, the relative tumor growth, the size, and weight in the lnc-LBCS overexpression group were significantly decreased compared with the control group (Fig. 3E–G; Supplementary Fig. S8A–S8C). Moreover, the tumors derived from the lnc-LBCS overexpression group exhibited a higher proportion of TUNEL-positive cells and lower Ki67 expression compared with the control group when treated with GC chemotherapy (Fig. 3H; Supplementary Fig. S7D and S7E; Supplementary Fig. S8G and S8I). Conversely, lnc-LBCS depletion significantly enhanced tumor growth and reduced apoptosis when treated with GC chemotherapy (Fig. 3I–L; Supplementary Fig. S7D and S7E; Supplementary Fig. S8D–S8F, S8H, and S8I). Consistent with the lnc-LBCS knockdown findings above, lnc-LBCS KO promoted the chemoresistance of BCSCs in vivo (Supplementary Fig. S9E–S9J). Overall, these data strongly indicate that lnc-LBCS suppresses chemoresistance of BCSCs in vitro and in vivo.

To further identify the molecular mechanism and binding partners of lnc-LBCS in BCSCs, we performed RNA pulldown with biotin-labeled lnc-LBCS. Two overtly differential bands appeared by silver staining and were identified as hnRNPK and EZH2 by mass spectrometry (Fig. 4A; Supplementary Fig. S10A and S10B). We confirmed the special interaction of lnc-LBCS, hnRNPK, and EZH2 using Western blotting (Fig. 4B). We also verified this result using RIP assay and found that lnc-LBCS were enriched in both hnRNPK and EZH2 precipitates (Fig. 4C). We next constructed a series of lnc-LBCS truncations to map its binding fragment with the hnRNPK and EZH2. We found that the 240–480 nt fragment of lnc-LBCS was sufficient to bind hnRNPK (Fig. 4D). Considering that hnRNPK binds to pyrimidine-rich sequences (32, 33), we identified the accurate binding site of hnRNPK in lnc-LBCS (256–280 nt) via RNA pulldown assays (Fig. 4E). The 240–731 nt fragment of lnc-LBCS was sufficient to bind EZH2, but not 1–480 nt or 480–731 nt, suggesting that the binding site was around 480 nt (Fig. 4D). Considering that EZH2 binds to stem-loop structure of RNA (34), we deduced that EZH2 bound to the 468–523 nt fragment of lnc-LBCS which formed a stem-loop structure (Fig. 4F).

To determine whether hnRNPK and EZH2 interact with lnc-LBCS independently or as a multiprotein complex, we performed co-IP assay followed by Western blotting (Fig. 4G). Interestingly, hnRNPK and EZH2 interactions were observed significantly in lysates incubated with the lnc-LBCS RNA, but not the mock group. However, these interactions were abolished by treatment of lysates with ribonuclease (RNase; Fig. 4G), suggesting that lnc-LBCS RNA was essential to the formation of hnRNPK–EZH2 complex. Taken together, these results indicate that lnc-LBCS directly binds to hnRNPK and EZH2, and serves as a scaffold to induce the formation of this complex.

To explore the target genes of lnc-LBCS in BCSCs, we investigated self-renewal and chemoresistance pathway genes in lnc-LBCS–silenced cells. Interestingly, SOX2 was the most significant upregulated gene in lnc-LBCS–silenced cells (Fig. 5A). Furthermore, the mRNA and protein levels of SOX2 were increased in lnc-LBCS–silenced and lnc-LBCS-KO cells, whereas they were decreased in lnc-LBCS–overexpressing cells (Fig. 5B and C; Supplementary Fig. S11A). To further verify the correlation between lnc-LBCS and SOX2, we detected the expression of SOX2 in two independent cohorts of bladder cancer specimens by qRT-PCR and IHC. Notably, the mRNA and the protein levels of SOX2 correlated negatively with the lnc-LBCS level (Fig. 5D–F, P < 0.001, P < 0.001, respectively). In summary, these data indicated that lnc-LBCS inhibited SOX2 expression.

To provide direct evidence that lnc-LBCS associated with the promoter region of SOX2, we performed ChIRP experiment and detected the enrichment of specific regulatory regions by qPCR. We observed obvious enrichment of the –810 to –927 nt region of the SOX2 promoter using lnc-LBCS probes, but no enrichment of GAPDH or other regions of the SOX2 promoter, as compared with LacZ probes. TERC and its downstream gene WNT-1 served as positive controls for RNA and DNA enrichment (Fig. 5G and H; Supplementary Fig. S11B–S11E). To further identify more precisely the direct binding sites between lnc-LBCS and the SOX2 promoter, we analyzed potential triplex-forming oligos (TFO) and corresponding triplex target sites (TTS) using the Longtarget software (35). Fluorescence resonance energy transfer (FRET) was performed using in vitro synthesized predicted TFOs of lnc-LBCS and TTSs of the SOX2 promoter. Upon excitation at 460 nm, the emission at 580 nm increased, whereas the signal at 520 nm decreased in the lnc-LBCS (246–266 nt)/SOX2 TTS group compared with that of the control RNA/SOX2 TTS (Fig. 5I). However, lnc-LBCS (mut)/SOX2 TTS and lnc-LBCS (246–266 nt)/SOX2 TTS (mut) did not influence the signals at 520 and 580 nm (Fig. 5I; Supplementary Fig. S11G). Taken together, these data indicate that lnc-LBCS 246–266 nt fragment directly forms triplexes with the promoter sequences of SOX2.

Through analysis of the histone modification profile in the UCSC genome browser, we found that the lnc-LBCS–binding region in the SOX2 promoter had a high H3K27me3 level, suggesting that lnc-LBCS might recruit hnRNPK–EZH2 complex to mediate H3K27me3 on SOX2 promoter (Supplementary Fig. S11F). To address this, we performed a ChIP assay and qPCR. Lnc-LBCS knockdown resulted in decreased location of hnRNPK, EZH2, and H3K27me3 but increased location of RNA polymerase-II on the promoter regions of SOX2, but not on the negative control (Fig. 5J; Supplementary Fig. S11H). Furthermore, we found that lnc-LBCS knockdown enhanced the luciferase activity in the wild-type–binding region of SOX2, but not in the mutated region (Fig. 5L). Conversely, lnc-LBCS overexpression augmented the location of hnRNPK, EZH2, and H3K27me3, but reduced RNA polymerase-II binding to the SOX2 promoter (Fig. 5K; Supplementary Fig. S11H). Similarly, lnc-LBCS overexpression attenuated the luciferase activity, but not when including the mutated region (Fig. 5L). Furthermore, knockdown of hnRNPK or EZH2 rescued the depression effect of lnc-LBCS on chemoresistance and SOX2 in bladder cancer cells (Fig. 5M and N; Supplementary Fig. S12). Overall, these data indicate that lnc-LBCS inhibits SOX2 transcription by directly guiding hnRNPK–EZH2 complex to mediate H3K27me3 of the SOX2 promoter.

Accumulating evidence shows that SOX2 plays a key role in tumor initiation and self-renewal of CSCs in several cancers (36); however, its effect on self-renewal and chemoresistance of BCSCs remains unclear. We stably upregulated SOX2 in control vector and lnc-LBCS–overexpressing cells, which was verified by qRT-PCR and Western blotting (Supplementary Fig. S13A and S13B). SOX2 upregulation significantly promoted BCSCs’ sphere formation and increased the population of ALDH⁺ cells in vitro, and enhanced tumor initiation in vivo. Moreover, SOX2 overexpression also rescued the inhibitory effects of lnc-LBCS on the self-renewal of BCSCs in vitro and in vivo (Fig. 6A–D; Supplementary Fig. S13C–S13E). Notably, SOX2 overexpression markedly enhanced the chemoresistance of BCSCs to gemcitabine and cisplatin in vitro and in vivo. In addition, SOX2 overexpression also rescued the repressive roles of lnc-LBCS on the chemoresistance of BCSCs (Fig. 6E–I; Supplementary Fig. S14A–S14G). In summary, these findings indicate that SOX2 is required for self-renewal and chemoresistance of BCSCs and rescues the suppressive effect of lnc-LBCS.

To further verify the roles of lnc-LBCS in BCSCs in an SOX2-dependent manner, we overexpressed lnc-LBCS and truncated lnc-LBCS without the SOX2 promoter binding site (lnc-LBCS Sdel) in UM-UC-3 4th and 5637 cells, respectively. Interestingly, the lnc-LBCS Sdel group did not inhibit sphere formation or reduce the population of ALDH⁺ cells compared with the control vector group (Supplementary Fig. S15A–S15C). Similarly, compared with the control vector group, overexpression of lnc-LBCS Sdel did not affect the chemotherapy resistance, IC₅₀ value, apoptosis rate, and caspase-3/7 activity in UM-UC-3 4th and 5637 cells treated with gemcitabine or cisplatin (Supplementary Fig. S15D–S15H). Overall, the SOX2-binding site of lnc-LBCS is essential for lnc-LBCS to exert the inhibitory effect on SOX2 expression and BCSCs’ functions.

Finally, we investigated the clinical significance of SOX2 expression in bladder cancer tumorigenesis and progression in three independent large-scale cohorts. Interestingly, high expression levels of SOX2 were found in bladder cancer samples and high-grade samples, indicating the potential role of SOX2 in bladder cancer initiation and stemness (Fig. 6J and K). Moreover, clinical correlation analysis revealed that SOX2 expression was strongly correlated with pathologic grade and stage (P = 0.001 and P < 0.001, respectively; Supplementary Table S7). This finding was further confirmed by IHC of an 86-case Cohort 2 and TCGA cohort bladder cancer samples (Fig. 6L and M; Supplementary Table S8; Supplementary Fig. S16A). Notably, SOX2 displayed higher expression in neoadjuvant chemotherapy nonresponders than responders, further conforming that SOX2 was associated with chemoresistance (Supplementary Fig. S16B). Furthermore, the Kaplan–Meier survival analysis found that patients with high SOX2 expression had shorter OS and DFS in Cohort 1 and in the TCGA cohort (Fig. 6N–Q). Besides its role in stemness maintenance and development regulation, we revealed the critical role of SOX2 in bladder cancer initiation, progress, and prognosis prediction.

Collectively, the absence of lnc-LBCS in BCSCs attenuates the recruitment and location of hnRNPK–EZH2 complex on the SOX2 promoter, which increases SOX2 expression, contributing to the survival advantage in response to chemotherapy and subsequent tumor relapse. In non-BCSCs, lnc-LBCS could guide the hnRNPK–EZH2 complex to suppress SOX2 expression, leading to good response to chemotherapy (Fig. 6R).

LncRNAs are reported to have important epigenetic regulatory roles in diverse biological cellular processes including tumorigenesis and cancer chemoresistance. In this study, we first reported that a novel lncRNA-LBCS is significantly downregulated in BCSCs and cancer tissue, and correlates with tumor grade, chemotherapy response, and prognosis. Moreover, lnc-LBCS markedly suppressed the self-renewal and chemoresistance of BCSCs by guiding the hnRNPK–EZH2 complex to inhibit SOX2 expression, thereby contributing to attenuate bladder cancer initiation and chemoresistance. These findings indicate that lnc-LBCS acts as a tumor suppressor in bladder tumorigenesis and progression, and could be considered as a potential prognostic indicator and therapy target for bladder cancer.

Recent studies show that lncRNAs, including lncTCF7 (16), lncARSR (17), RBM5-AS1 (18), and ICR (37), participate in CSCs' self-renewal of liver, renal, and colon cancer. However, the BCSCs-associated lncRNAs remain largely unknown. Here, through transcriptome microarray analysis and verification in several BCSCs models, we identified a novel lncRNA-LBCS that was significantly downregulated in BCSCs, and its expression gradually increased during spheres readherence. Importantly, lnc-LBCS expression was negatively correlated with bladder cancer grade and stage in three independent cohorts, whereas it was positively correlated with neoadjuvant chemotherapy response and good prognosis, suggesting that lnc-LBCS could regulate cell differentiation and chemosensitivity. Furthermore, analyses of TCGA databases showed that lnc-LBCS was also downregulated and correlated with prognosis in various types of human cancers, suggesting a common tumor-suppressor role of lnc-LBCS.

LncRNAs can guide and recruit histone protein modification enzymes or transcription factors to specific genomic loci, leading to inactivation or activation of genes (38). HnRNPK is an essential RNA- and DNA-binding protein that plays a critical role in several cancers (24, 39). Our previous study found that hnRNPK regulates diverse functions in bladder cancer by directly mediating transcription of target genes; however, the detailed mechanism remained unclear (24). EZH2, a histone methyl transferase subunit of the polycomb repressor complex 2 (PRC2), is an oncogene that is central to tumor proliferation, metastasis, and self-renewal (40, 41). In the present study, we found that lnc-LBCS physically interacts with hnRNPK and EZH2, and lnc-LBCS serves as a scaffold to induce the formation of hnRNPK–EZH2 complex. Furthermore, lnc-LBCS recruited this complex to the SOX2 promoter and suppressed SOX2 expression via mediating H3K27me3. Although hnRNPK and EZH2 are overexpressed and exert oncogenic function in several cancers, recent study indicates that some lncRNAs could recruit hnRNPK or EZH2 to exert a tumor-suppressor role. For example, lincRNA-p21 recruits hnRNPK to inhibit p53-pathway genes (42), and MEG3 recruits EZH2 to inhibit metastasis of breast cancer cells by repressing TGFβ pathway genes (43). These findings suggested that recruitment and guidance by lncRNAs might decide target gene regulation and the function of hnRNPK and EZH2 in cancer. The low expression of lnc-LBCS in BCSCs would result in attenuated recruitment and location of hnRNPK–EZH2 complex on the SOX2 promoter, contributing to SOX2 upregulation in BCSCs and bladder cancer progression.

SOX2 is a master regulator that maintains stemness in embryonic stem cells (44) as well as self-renewal of CSCs in several malignancies (36, 45). In addition, SOX2 also plays a critical role in drug resistance to paclitaxel (45) and gemcitabine (46) in several cancers. However, the function and epigenetic regulatory mechanism of SOX2 in BCSCs remain to be determined. In this study, we demonstrated that SOX2 is required for the self-renewal and chemoresistance to gemcitabine and cisplatin of BCSCs and tumor propagation by acting as a key oncogene. Intriguingly, a previous study revealed that the expression of SOX2 during the course of carcinogenesis correlates with a loss of the repressive H3K27me3 chromatin mark at the SOX2 promoter (36), but the underlying mechanism was unclear. Here, we found that SOX2 is markedly repressed by lnc-LBCS, which recruits hnRNPK–EZH2 complex to SOX2 promoter leading to H3K27me3. While we were submitting this article, we noted a similar finding by Ooki and colleagues recently reported that YAP1 and COX2 coordinately enhanced SOX2 to promote chemoresistance of BCSCs (47). Similarly, Zhu and colleagues reported that SOX2 is a marker for stem-like tumor cells in bladder cancer (48). Consistent with our finding, these two studies indicate that SOX2 plays a major role in BCSCs. Collectively, our finding reveals a novel epigenetic regulation mechanism of SOX2 by lnc-LBCS in BCSCs, suggesting that lnc-LBCS may represent a new target for intervention in SOX2-dependent BCSCs.

In conclusion, it is our novel discovery that lnc-LBCS can inhibit BCSCs’ self-renewal and chemoresistance by repression of SOX2 via guiding hnRNPK–EZH2 complex to the SOX2 promoter and inducing H3K27me3. Therefore, our findings provide insight into lnc-LBCS might be used in diagnosis and prognosis for bladder cancer, as well as in the development of novel therapeutic drugs against chemoresistant BCSCs.

No potential conflicts of interest were disclosed.

Conception and design: X. Chen, J. Huang, T. Lin

Development of methodology: X. Chen, P. Gu, B. Wang, G. Zhong, Z. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Chen, R. Xie, P. Gu, M. Huang, W. Dong, W. Xie

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Han, W. Dong, G. Zhong

Writing, review, and/or revision of the manuscript: X. Chen, J. Huang, T. Lin

Study supervision: W. He, J. Huang, T. Lin

This study was supported by the National Natural Science Foundation of China (grant nos. 81825016, 81702523, 81772719, 81772728, 81572514, and 81472384), National Natural Science Foundation of Guangdong (grant nos. 2016A030313321, 2016A030313244, and 2015A030311011), Science and Technology Program of Guangzhou (grant nos. 201804010041, 201604020156, and 201604020177), the Science and Technology Planning Project of Guangdong Province (grant no. 2017B020227007), Guangdong Special Support Program (2017TX04R246), the Fundamental Research Funds for the Central Universities (for X. Chen, 18ykpy18), Project Supported by Guangdong Province Higher Vocational Colleges & Schools Pearl River Scholar Funded Scheme (for T. Lin), Yat-Sen Scholarship for Young Scientist (for X. Chen), Sun Yat-sen Initiative Program for Scientific Research (for X. Chen, YXQH201708), Cultivation of Major Projects and Emerging, Interdisciplinary Fund, Sun Yat-Sen University (grant no. 16ykjc18), and National Clinical Key Specialty Construction Project for Department of Urology and Department of Oncology. This study was also supported by grant KLB09001 from the Key Laboratory of Malignant Tumor Gene Regulation and Target Therapy of Guangdong Higher Education Institutes, Sun-Yat-Sen University, and grant [2013]163 from Key Laboratory of Malignant Tumor Molecular Mechanism and Translational Medicine of Guangzhou Bureau of Science and Information Technology.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.