Hypoxia suppresses AR expression and increases the CSC formation in RCC

We first examined hypoxia’s influence on the AR expression, and found hypoxia could decrease AR mRNA after 24 h in both RCC OSRC-2 and SW839 cells (Fig. 1A). Western-blot analysis also confirmed hypoxia could decrease AR protein expression after 36 h in both RCC OSRC-2 and SW839 cells (Fig. 1B and Additional file 1: Fig. S1A). Because hypoxia can suppress gene translation in some cases, we tested additional genes (EZH2 and SRC) to determine their protein expression level. The results showed that EZH2 decreased while SRC increased under hypoxia condition (Additional file 1: Fig. S1B). These results demonstrated that hypoxia-decreased AR may not be the result of general translation suppression.

Fig. 1

Hypoxia (1% oxygen) suppresses AR expression and increases the CSC formation in RCC cells. A SW839 and OSRC-2 cells were exposed to hypoxia for 0 h, 12 h, 24 h, and 36 h. Real-time PCR was used to detect AR expression. B SW839 and OSRC-2 cells were exposed to hypoxia for 0 h, 12 h, 24 h, and 36 h. Western-blots were used to detect AR expression. C SW839 and SW839 with sh-AR1 and sh-AR2 were exposed to hypoxia (H) for 2 days, sphere formation assays were performed to evaluate the CSCs number. After 14 days incubation, colonies in five random fields per each well were counted under a microscope. D SW839 cells were virally transduced with sh-AR and cells exposed to hypoxia (H) or normoxia (N) for 2 days. Total RNAs were analyzed for CSCs markers including CD24, CD133, PAX2, CD105, SOX2 by real-time PCR. E OSRC-2 and ACHN cells were virally transduced with overexpressed AR and exposed to hypoxia (H) for two days, sphere formation assays were performed to evaluate the CSCs number. After 14 days incubation, colonies in five random fields per each well were counted under a microscope. F OSRC-2 cells were virally transduced with oe-AR and cells were exposed to hypoxia (H) or normoxia (N) for 2 days. Total RNAs were analyzed for CSCs markers including CD24, CD133, PAX2, CD105 and SOX2 by real-time PCR.(AR: androgen receptor; GAPDH: Recombinant protein of human glyceraldehyde-3-phosphate dehydrogenase; pLKO: pLKO.1-puro plasmid; sh-AR1/2: The plasmid encoding AR-shRNA1/2; pWPI: Addgene plasmid 12,254,; oe-AR: The plasmid encoding AR sequence, H:hypoxia; N: normoxia. * P<0.05, ** P<0.01, ***P<0.001)

We then examined the consequence of hypoxia-suppressed AR signaling on the RCC progression, with a focus on CSC phenotype, as abundant evidence indicated that CSC might play important roles for tumor development and progression [12].

Using sphere formation assays to evaluate the CSC formation, we found knocking down AR (sh-AR) using two different sequences both increased the CSC formation (Fig. 1C) as well as the expression of CSCs biomarkers including CD24, CD133, PAX2, SOX2 and CD105 in RCC SW839 cells under hypoxia condition (Fig. 1D). In the following experiments we just used one sh-AR sequence. In contrast, overexpressing AR decreased the CSC formation (Fig. 1E) and expression of CSC biomarkers, including CD24, CD133, PAX2, SOX2 and CD105 in OSRC-2 cells under hypoxia (Fig. 1F). The efficiency of sh-AR and overexpression AR (oe-AR) were tested and shown in Additional file 1: Fig. S1C. We also confirmed the hypoxia condition by testing the protein of HIF-1α expression, the results demonstrated that hypoxia can decrease HIF-1α protein expression dramatically in OSRC-2 cells (Additional file 1: Fig. S1D).

Together, results from Fig. 1A–F and Additional file 1: Fig. S1A–C suggest that through suppression of AR expression, hypoxia may increase the CSC phenotype in RCC cells.

Hypoxia-suppressed AR expression may alter the expression of lncRNAs in RCC cells

To dissect the molecular mechanisms how hypoxia-suppressed AR can alter the RCC CSCs phenotype, we focused on the long noncoding RNAs (lncRNAs) as recent studies indicated that they might play important roles to promote tumor progression in a variety of settings. Using lncRNA microarray analysis, we subjected SW839 cells, a VHL-mutant RCC cell line, to hypoxia (1% oxygen level for 24 h) and carried out a comparative analysis of the expression of 40,000 lncRNAs and found the expression of nearly 7000 lncRNAs was increased, with a reduction of expression of nearly 13,000 lncRNAs in response to hypoxia with a subset of lncRNAs that showed more than two fold difference in expression (Fig. 2A and Additional file 1: Fig. S2A). We then focused on those lncRNAs with significant changes of expression in response to hypoxia (fold change > 0.6, and P value < 0.05), and chose 20 up-regulated lncRNAs and 20 down-regulated lncRNAs for further study (Fig. 2C and Additional file 1: Fig. S2B, C).

Fig. 2

Hypoxia-suppressed AR expression may alter the expression of lncRNAs in RCC cells. A SW839 cells, pre-cultured in the hypoxia condition (1% oxygen level for 24 h) and normoxia condition. Volcano plots of the differentially expressed lncRNAs: red spots represent fold change more than 0.6 and P valve less than 0.05. B Western-blot showed the efficiency of sh-AR in SW839 cells and overexpressed AR (oe-AR) in OSRC-2 cells. C The list of top 20 upregulated lncRNAs by hypoxia. D SW839 and OSRC-2 cells were virally transduced with sh-AR or oe-AR, and then cells were exposed to hypoxia (H) or normoxia (N) for 2 days. Total RNAs were analyzed by Q-PCR to show these 20 lncRNAs expression. The results showed that three lncRNAs (lnc-CTB-114C7.4, lnc-TCFL5-2(XLOC_013838)and LINC00312) were suppressed by AR. E SW839 cells were virally transduced with pLVTHM, sh-lnc-CTB-114C7.4, sh-lncTCFL5-2, sh-LINC00312 and cells were exposed to hypoxia (H) for 2 days. Sphere formation assay were performed to evaluate the CSCs numbers. After 14 days of incubation, colonies in five random fields per each well were counted under a microscope. F SW839 and OSRC-2 cells were virally transduced with sh-lncTCFL5-2 and then cells were exposed to hypoxia and normoxia for 2 days. Total RNAs were analyzed for CSCs markers CD24, CD133, PAX2, CD105, SOX2 by real-time PCR. G OSRC-2 cells were virally transduced with oe-lncTCFL5-2 and then cells exposed to hypoxia (H) for 2 days. Sphere formation assay were performed to evaluate the CSCs numbers. H SW839 and OSRC-2 cells were virally transduced with oe-lncTCFL5-2 and then cells exposed to hypoxia and normoxia for 2 days. Total RNA were analyzed for CSCs markers CD24, CD133, PAX2, CD105, SOX2 by real-time PCR. I ACHN and OSRC-2 cell were virally transduced with oe-AR or were co-transfected with oe-AR and oe-lncTCFL5-2, and then cells exposed to hypoxia for 2 days. Sphere formation assays were performed to evaluate the CSCs phenotype. (AR: androgen receptor; pLVTHM: Addgene plasmid 12,247; sh-lnc-CTB-114C7.4: The pLVTHM plasmid encoding lnc-CTB-114C7.4 shRNA, sh-lncTCFL5-2: The pLVTHM plasmid encoding lncTCFL5-2 shRNA, sh-LINC00312: The pLVTHM plasmid encoding LINC00312 shRNA; oe-AR: The plasmid encoding AR sequence; oe-lncTCFL5-2: The plasmid encoding lncTCFL5-2 sequence.)

Next, to link those 40 hypoxia-altered lncRNAs to the AR-modulated CSC phenotype under hypoxia, we altered AR expression (Fig. 2B) and examined its impact on these 40 lncRNAs, and results revealed that knocking down AR in SW839 cells (Fig. 2D) significantly altered the expression of some lncRNAs including lnc-CTB-114C7.4, lncTCFL5-2(XLOC_013838) and LINC00312 suggesting AR has an effect on the hypoxia-induced lncRNAs expression. Consistent with this finding, overexpression of AR in OSRC-2 cells suppressed the hypoxia-induction of these 3 lncRNAs (Additional file 1: Fig. S2D). Similarly, we found that AR regulates hypoxia-induced suppression of certain lncRNAs (Additional file 1: Fig. S2C).

Together, results from Fig. 2A-D and Additional file 1: Fig. S2A–D suggest that hypoxia can decrease AR expression, which may significantly alter expression of some special lncRNAs.

Hypoxia-decreased AR may result in de-repression of lncTCFL5-2, which contributes the CSC formation in RCC cells

To examine the role of AR-regulated lncRNAs in hypoxia-induced CSC phenotypes in RCC, we altered the expression of these 3 lncRNAs and found that knocking down lncTCFL5-2 with two different sh-sequence plasmids could both significantly suppress RCC sphere formation under the influence of hypoxia (Fig. 2E, Additional file 1: Fig. S2E). So we did the following experiments just using one sh-sequence plasmid.

Furthermore, knocking down lncTCFL5-2 also decreased expression of CSC related genes, including CD24, CD133, PAX2, SOX2 and CD105 in SW839 and OSRC-2 cells under hypoxia (Fig. 2F). In contrast, overexpressing lncTCFL5-2 increased CSC sphere formation (Fig. 2G) as well as expression of CSCs biomarkers, including CD24, CD133, PAX2, SOX2 and CD105 in SW839 and OSRC-2 under normoxia but not in a hypoxia condition (Fig. 2H).

Using interruption approaches, we also found exogenous overexpression of lncTCFL5-2 reversed/blocked AR-decreased CSC formation in both RCC OSRC-2 and ACHN cells under hypoxia (Fig. 2I). To further confirm AR effect on the CSC population through lncTCFL5-2, we also tested the relationship between AR and lncTCFL5-2 in Additional file 1: Fig. S2F, and found that the lncTCFL5-2 cannot change AR at either mRNA level or protein level.

Together, results from Fig. 2E–I and Additional file 1: Fig. S2E-F suggest that hypoxia-suppressed AR may result in de-repression of lncTCFL5-2, which may then increase the CSC population in RCC cells.

Mechanism dissection how AR suppresses lncTCFL5-2 expression

Based on the lncipedia data (https://lncipedia.org/db/gene/lnc-TCFL5-2), lncTCFL5-2 is encoded by the gene located in chromosome 20 (hg38) from 62,782,754–62,794,246 with a length of 463 bases. To dissect mechanisms how AR regulates lncTCFL5-2 expression, we first found knocking down AR increased the lncTCFL5-2 expression in SW839 cells (Fig. 3A), and overexpressing AR decreased the lncTCFL5-2 expression in OSRC-2 cells (Fig. 3B). Similar results were obtained when we replaced OSRC-2 cells with ACHN cells (Additional file 1: Fig. S3A). Importantly, treating with the anti-androgen enzalutamide to replace AR-shRNA, we also obtained similar results showing AR inactivation increased lncTCFL-5–2 expression (Fig. 3C). Consistent with this, there was increased expression of CSCs marker genes including CD24, CD133, PAX2, SOX2 and CD105 in OSRC-2 cells treated with Enzalutamide (Additional file 1: Fig. S3B).

Fig. 3

Mechanism dissection how AR suppresses lncTCFL5-2 expression. A SW839 cell were virally transduced with sh-AR1, sh-AR2 and cells were exposed to hypoxia or normoxia for 2 days. qPCR analysis of the expression of lncTCFL5-2. B OSRC-2 cells were virally transduced with oe-AR and cells exposed to hypoxia or normoxia for 2 days, qPCR analysis of the expression of lncTCFL5-2. C SW839 cells were treated with Enzalutamide (Enz) 10 µM and then cells exposed to hypoxia or normoxia for 2 days, qPCR analysis of the expression of lncTCFL5-2. D The interaction between AR and lncTCFL5-2 in both SW839 and OSRC-2 cells under normoxia and hypoxia condition were detected by RIP. E Bioinformatic analysis of potential AR binding sites in lncTCFL5-2 promoter. F Lysates of OSRC-2 cells were subjected to ChIP assay. ChIP products were amplified by PCR reaction. G, H SW838 and OSRC-2 cells were co-transfected with the indicated reporter constructs and Renilla luciferase plasmid. 48 h after transfection, reporter activity was then measured and plotted after normalizing against the Renilla luciferase activity (mean ± SD).. (pLKO: pLKO.1-puro plasmid, sh-AR1/2: The plasmid encoding AR-shRNA1/2, oe-AR: The plasmid encoding AR sequence, pWPI: Addgene plasmid 12,254, ARE:AR Element)

To further examine whether AR can regulate lncTCFL5-2 expression through transcriptional or post-transcriptional regulatory mechanisms, we tested whether AR could bind to the lncRNA. The RIP assay with anti-AR antibody together with IgG control followed by analysis of the associated RNA through reverse-transcription and quantitative PCR indicated that there was slight increase of AR association with the lncRNA compared to the negative control in both normoxia and hypoxia conditions (Fig. 3D). Thus, we concluded that AR unlikely interacted with the lncRNA in a functionally significant manner. On the other hand, through bioinformatics analysis (Ensembl and PROMO 3.0) of AR binding sites in the potential lncRNA promoter, we found 4 potential AREs within 2.5 kb upstream of the transcriptional start site (Fig. 3E). Chromatin immunoprecipitation (ChIP) assay with anti-AR antibody in OSRC-2 cells revealed that AR could bind to the ARE1 located on the 162nt to 154nt upstream of the transcription start site of lncTCFL5-2 in OSRC-2 cells (Fig. 3F).

Importantly, using luciferase reporter construct with 1 kb upstream sequence containing normal (wild type) or mutated ARE-1, we found overexpressing AR significantly decreased the reporter activity in OSRC-2 cells and sh-AR significantly increased the reporter activity in SW839 cells transfected with the wild type construct but not the mutant construct (Fig. 3G, H).

Together, results from Fig. 3A–H and Additional file 1: Fig. S3A-B demonstrated that AR could suppress lncTCFL5-2 expression at a transcriptional level via binding to the ARE-1 located in its 5’ promoter region.

Mechanism dissection how AR-suppressed lncTCFL5-2 may increase CSC phenotype via interacting with the YBX1 in RCC

The lncRNAs have been implicated in a variety of cellular functions broadly through specific mechanisms, such as signaling molecules, decoys, guides, and scaffolds to support multi-protein structures [13]. Among many possibilities suggesting that lncRNA-TCFL-2 can regulate CSC formation, we tested whether this lncRNA can interact with a protein known to be involved in CSC formation. Through a bioinformatic analysis of likely protein-RNA interactions (RBP map), we found that lncTCFL5-2 might be able to interact with YBX1, a protein reported to be a regulator of SOX2, a transcriptional factor critical for maintaining stem cell characteristics [14]. Results from RIP assays revealed that lncTCFL5-2 indeed interacted with YBX1 compared to AR that failed to interact with this lncRNA, consistent with the previous data (Fig. 4A, Fig. 3D). Furthermore the lncTCFL5-2 and YBX1 interaction could be demonstrated with the endogenous level in OSRC-2 cells, not through overexpression in either of them (Additional file 1: Fig. S3C, D).

Fig. 4

Mechanism dissection how AR-suppressed lncTCFL5-2 may increase CSCs the phenotype via interacting with the YBX1 in RCC. A 293 T cells were co-transfected with oe-AR and oe-lncTCFL5-2 or oe-YBX1 and oe-lncTCFL5-2, RIP detected the interaction between YBX1 and lncTCFL5-2. B 293 T cells were co-transfected with oe-YBX1 and four deletions mutants of lncTCFL5-2, RIP detected the interaction between YBX1 and the four mutants. C OSRC-2 cells were lentivirally transduced with sh-lncTCFL5-2 or oe-lncTCFL5-2, and then exposed to hypoxia or normoxia for 2 days. Western blot analysis of the expression of YBX1. D OSRC-2 cells were lentivirally transduced with pWPI or oe-lncTCFL5-2, then treated with 10 mg/ml cycloheximide (CHX). YBX1 protein levels were analyzed by western blot (lower right panel). The lower left panel is a graphic representation of YBX1 metabolic stability. MG132 (20 µM) was added in the OSRC-2 cells, Western blot analysis of the expression of YBX1. E SW839 and OSRC-2 cells were lentivirally transduced with sh-AR or oe-AR, respectively, and then cells were exposed to hypoxia or normoxia for 2 days. Western blot analysis of the expression of YBX1. F OSRC-2 cells were lentivirally transduced with oe-AR or oe-AR together with wildtype or mutant lncTCFL5-2. Sphere formation assays were performed to evaluate the CSC phenotype. After 14 days of incubation, colonies in five random fields per well were counted under a microscope. G OSRC-2 cells were lentivirally transduced with pLVTHM or wildtype or mutant lncTCFL5-2. Total RNAs were analyzed for CSC markers CD24, CD133, PAX2, CD105, and SOX2 by real-time PCR. H ACHN and OSRC-2 cells were lentivirally transduced with sh-lncTCFL5-2 or were co-transduced with sh-lncTCFL5-2 and oe-YBX1, and then cells exposed to hypoxia for 2 days. Sphere formation assay were performed to evaluate the CSC phenotype. After 14 days of incubation, colonies in five random fields per well were counted under a microscope. I SW839 cells were lentivirally transduced with sh-lncTCFL5-2 or co-transfected with sh-lncTCFL5-2 and oe-YBX1 and then cell exposed to hypoxia and normoxia for 2 days. Total RNAs were analyzed for CSC marker CD24, CD133, PAX2, CD105, SOX2 by real-time PCR (ab: Antibody, pLVTHM: Addgene plasmid 12,247; sh-lncTCFL5-2: The pLVTHM plasmid encoding lncTCFL5-2 shRNA, oe-lncTCFL5-2: The plasmid encoding lncTCFL5-2 sequence, oe-YBX1: The plasmid encoding YBX1 sequence; N: normoxia, H: hypoxia)

To further solidify the functional significance of lncTCFL5-2 with YBX1, we also mapped the YBX1 binding sequence in the lncRNA. There are 4 putative YBX1-binding areas with predicted binding motifs in the lncTCFL5-2. Through deletion analysis, we found that the 864-936nt region with 6 binding motifs (mutation-C) is critical for lncRNA-YBX1 interaction (Fig. 4B) while the remaining single binding motif in the 5’ and 3’ ends, as well as one single binding motif at 1150nt, appear not important for binding to YBX1.

Importantly, we found that YBX1 protein expression was also influenced by hypoxia and lncTCFL5-2 high expression (Fig. 4C). To directly test whether lncTCFL5-2 could regulate YBX1 protein stability, cycloheximide (CHX) was used to block the de novo protein synthesis, and the expression of YBX1 protein in OSRC-2 cells was tracked at various time points after CHX addition. We found the overexpression of lncTCFL5-2 prolonged YBX1 protein half-life (Fig. 4D). In addition, this increase of YBX1 stability is likely due to reduced proteasome-mediated degradation as inhibitor of proteasome, MG132, could increase the YBX1 level, more dramatically in cells with a knock down of lncTCFL5-2 (Fig. 4D lower left panel). It is likely that ubiquitin-mediated degradation plays a role in the protection by lncTCFL5-2 on YBX1 although the exact E3 ligase remains to be determined.

The AR/lncTCFL5-2 connection suggested that AR might regulate YBX1 expression and indeed we found overexpressing AR in ORSC-2 cells resulted in repressing the YBX1 expression in SW839 cells likely through repression of lncTCFL5-2 expression (Fig. 4E). Importantly, using an interruption approach, we found AR-suppressed CSCs formation could be reversed by overexpression of the wildtype lncTCFL5-2, but not by a mutant lncTCFL5-2 failed to bind to YBX1 in OSRC-2 cells (Fig. 4F), further solidifying the significance of lncTCFL5-2 and YBX1 interaction in regulating the CSCs phenotype. To confirm this interpreted result we tested the AR and lncTCFL5-2 levels by qPCR (Additional file 1: Fig. S3E).

The expression of CSC biomarkers in response to wildtype and mutant lncTCL5-2 also supported this conclusion (Fig. 4G). Furthermore, a reduction of lncTCFL5-2 resulted in reduced RCC sphere formation under hypoxia, and this reduction can be reversed by simultaneous exogenous expression of YBX1 (Fig. 4H). Similar results were also obtained when we replaced sphere formation assay with the measurement of the expression of CSC biomarkers (Fig.  4I).

Together, results from Fig.  4A–I and Additional file 1: Fig. S4C-E demonstrated that the AR/lncTCFL5-2/YBX1 signaling axis plays a critical role to regulate the CSC phenotype in RCC cells in response to hypoxia.

Mechanism dissection how AR/lncTCFL5-2/YBX1 signals may function via modulating SOX2 to alter the CSCs formation in RCC

To further dissect the mechanism of SOX2 regulation by YBX1, we focused on the potential transcriptional regulation of SOX2 by YBX1 as the latter could regulate SOX2 expression at both mRNA and protein levels in SW839 cells (Fig. 5A, B). Consistent with a previous report [14], bioinformatics analysis indicated that there are 10 putative YBX1 response elements (YBX1REs) located within 2.5 kb SOX2 promoter region (Fig. 5C). Chromatin immunoprecipitation (ChIP) assay with anti-YBX1 antibody revealed that YBX1 could bind to the YBX1RE located on the 1600–2000 nt region upstream of the transcription start site of SOX2 in OSRC-2 cells (Fig. 5D).

Fig. 5

Mechanism dissection how AR/lncTCFL5-2/YBX1 signals may function via modulating SOX2 to alter the CSCs formation and chemotherapy resistance in RCC cells. A, B SW839 cells were virally transduced with sh-YBX1 or oe-YBX1 and then exposed to hypoxia and normoxia for 2 days. SOX2 expression was evaluated by western blot (A) and qPCR (B). C Bioinformatic analysis of potential YBX1 binding sites in SOX2 promoter. D Lysates of OSRC-2 cells were subjected to ChIP assay. ChIP products were amplified by PCR reactions. E, F Co-transfection of SOX2 promoter constructs containing the wild type or mutant region into OSRC-2 cells and luciferase assay was applied to detect the luciferase activity. G OSRC-2 cells were lentivirally transduced with pLVTHM or oe-lncTCFL5-2 or co-transfected with oe-lncTCFL5-2 and sh-SOX2. Total RNAs were analyzed for CSCs markers CD24, CD133, SOX2, NANOG, and CD105, PAX2, by real-time PCR. H SW839 and OSRC-2 cells were lentivirally transduced with pLVTHM or oe-lncTCFL5-2 or co-transfected with oe-lncTCFL5-2 and sh-SOX2. Western blot analysis of the expressions of CD24, CD133, SOX2. I SW839 and OSRC-2 cells were lentivirally transduced with pLVTHM or oe-lncTCFL5-2 or co-transfected with oe-lncTCFL5-2 and sh-SOX2. Sphere formation assays were performed to evaluate the CSC phenotype. After 14 days of incubation, colonies in five random fields per each well were counted under a microscope. J MTT assay was used to determine viability of cells exposed to hypoxia comparing to normoxia in SW839 cells with sh-lncTCFL5-2 or without sh-lncTCFL5-2 with Sunitinib treatment. K SW839 and OSRC-2 cells were exposed to hypoxia and normoxia for 2 days, and then treated with THZ1. Total RNAs were analyzed for lncTCFL5-2 by real-time PCR. L SW839 cells were lentivirally transduced with or without sh-lncTCFL5-2, then exposed to hypoxia and normoxia, treated with DMSO or Sunitinib combined with CDK-7 inhibitor. MTT assay was used to detect the cell viability. M OSRC-2 cells were virally transduced with pWPI or wildtype or mutant lncTCFL5-2, then treated with different concentrations of THZ1 followed by MTT assay

We therefore constructed a luciferase reporter construct bearing 2.5 kb SOX2 promoter sequence as well as a mutant reporter construct with mutations in the YBX1REs (see details in Fig. 5E). As expected, results from the luciferase reporter assay revealed that expression of YBX1 significantly increased reporter luciferase activity in OSRC-2 cells transfected with wild type SOX2 promoter, but not in the cells with mutant SOX2 promoter (Fig. 5F).

To further demonstrate the functional significance of SOX2 expression in RCC CSCs formation, we examined the consequence of SOX2 expression in regulating CSC biomarker expression. The results revealed that the hypoxia-induced CSC phenotype is strongly correlated with the appearance of positive expression of CD24 and CD133 (Additional file 1: Fig. S3F). Significantly, knocking down SOX2 resulted in repression of CSCs biomarkers including mRNA levels of CD24, CD133, SOX2, NANOG, CD105 and PAX2, in OSRC-2 cells  (Fig. 5G), protein levels of CD133, CD24 and SOX2 in both OSRC-2 and SW839 cells (Fig. 5H), as well as sphere formation under hypoxia of SW839 and OSRC-2 cells (Fig. 5I).

Together, results from Fig.  5A–I suggest that the AR/lncTCFL5-2/YBX1 signaling axis regulates RCC CSC phenotypes via modulating the SOX2 expression.

AR/lncTCFL5-2/YBX1/SOX2-modulated CSC formation led to alter the chemotherapy resistance in RCC cells

To determine the biological consequence of the AR/lncTCFL5-2/YBX1/SOX2 signaling axis under hypoxia in the RCC progression, we focused on the chemotherapy resistance, since early studies indicated that CSCs formation is often negatively correlated with therapy resistance, such as resistance to Sunitinib, the first-line treatment of RCC [15, 16].

We first applied the MTT assay to examine the RCC cells viability in response to Sunitinib treatment. As shown in Fig. 5J, and consistent with the previous report [17], Sunitinib is less effective in suppressing RCC cell viability under hypoxia than under normoxia, likely as a result of increased stem cell formation under hypoxia. Indeed, knocking down lncTCFL5-2 can abolish this difference in the efficacy of Sunitinib between normoxia and hypoxia in RCC cells, consistent with the role of lncTCFL5-2 in enhancing CSCs formation and acting as a critical contributor for the drug sensitivity in RCC cells likely through regulation of CSC formation (Fig. 5J).

To explore the potential therapeutic application of suppressing lncTCFL5-2 in enhancing the current RCC targeted therapy, we examined the effect of a CDK7 inhibitor in this process. CDK7 inhibitors, such as THZ1, have been shown to repress CSCs formation as an inhibitor of transcriptional activation such as the activity of super enhancers [18, 19]. Results revealed that THZ1 could more suppress the expression the lncTCFL5-2 under hypoxia than under normoxia (Fig. 5K). Consistent with the function of lncTCFL5-2 in CSC formation, THZ1 could mimic the effect of knocked-down lncTCFL5-2, thus Sunitinib exhibited similar efficacy in cells both under normoxia and hypoxia while a simultaneous knock down of the lncRNA will further enhance the efficacy of Sunitinib (Fig. 5L). To further implicate the functional significance of lncTCFL5-2 and YBX1 interaction in regulating the sensitivity towards Sunitinib, we found that exogenous expression of the wildtype lncTCFL5-2 can partially reverse the reduction of cell viability in response to Sunitinib and THZ1, while mutant lncTCFL5-2 was even lower than the control, suggesting that THZ1 can suppress the lncTCFL5-2 expression likely through suppressing super enhancer [18] activity to resensitize RCC cells to Sunitinib under hypoxia in RCC cells (Fig. 5M).

Together, results from Fig. 5J–M suggest that lncTCFL5-2 and CSC formation may play critical roles to influence the efficacy of the RCC chemotherapy and THZ1 application may potentially result in significant enhancement of the current RCC targeted therapy by decreasing lncTCFL5-2 expression and CSC formation.

Clinical significance of AR/lncTCFL5-2/YBX1/SOX2 signaling axis

To link the above in vitro results in human RCC formation and progression, we examined the clinical significance of the AR/lncTCFL5-2/YBX1/SOX2 signaling axis in RCC patient samples. We first used IHC to detect AR, YBX1 and SOX2 protein level in 30 ccRCC patient samples, and the results indicated that AR expression is higher in normal tissues compared to tumor tissues and adjacent tumor tissues, while YBX1 and SOX2 expression is higher in tumor tissues and adjacent tumor tissues compared to normal tissues (Fig. 6A) which is consistent with TCGA database (Additional file 1: Fig. S4A). Consistent with the in vitro cell lines results, expression of lncTCFL5-2 is higher in RCC tissues than in the normal tissues as well as YBX1 and SOX2 (Fig. 6B), and there is a lower AR expression in RCC tissues compared to the normal tissues (Fig. 6C), consistent with a negative correlation trend between AR and lncTCFL5-2 and a positive correlation trend between SOX2 and YBX1 or lncTCFL5-2 (Fig. 6C, Additional file 1: Fig. S4B).

Fig. 6

Clinical data and xenograft model demonstrate AR/lncTCFL5-2/YBX1/SOX2/CSCs pathway. A IHC staining for AR, YBX1, and SOX2 of tumor (T) tissues, adjacent tumor (AT) tissues and normal (N) tissues in ccRCC samples. B Real-time RT-PCR assays for detecting lncTCFL5-2, YBX1 and SOX2 mRNA expressions in tumor and normal tissues. C Real-time RT-PCR assays for detecting AR mRNA expression in tumor tissues and normal tissues and the correlation analysis of AR and lncTCFL5-2. D 560 RCC patients from TCGA dataset were divided into two groups according to AR expression. Survival curves showed that patients who have a higher AR expression have a higher survival rate. E AR expression in M0 and M1 patients, in different grades and different stages. F AR expression in RCC sample with necrosis or no necrosis (left panel), the correlation of YBX1 and AR protein (middle panel). 560 RCC patients from TCGA dataset were divided into two groups according to CD24 expression. Survival curves showed that patients who have a higher CD24 expression have a lower survival rate (right panel). G-M Results from in vivo mouse studies. G IVIS imaging was used to determine the tumor sizes and metastases. H We treated mice with 40 mg/kg Sunitinib daily IP for two weeks, IVIS imaging was used to determine the tumor sizes and metastases. I After sacrificing the mice the tumor tissues were obtained as shown (N: normoxia, H: hypoxia). J Weights of the xenografts were shown in four groups. K Weights reduction of the xenografts were shown in two groups after Sunitinib treatment (weight reduction = mean weight treated without Sunitinib-weight treated with Sunitinib). L Representative bioluminescent images of different organ metastasis. (N: normoxia, H: hypoxia) M The IHC for AR, YBX1, SOX2 of xenografts in four groups (N: normoxia, H: hypoxia)

We also performed a similar analysis with the TCGA data set and consistent with a previous analysis [7], patients who had a higher AR expression have a higher survival rate (Fig. 6D). Patients who had metastases had a lower AR level on aggregate (Fig. 6E). With an increase of ccRCC stage and grade, there is a corresponding decrease of AR level (Fig. 6E). Furthermore, patients who had a necrosis in tumors, an indicator of hypoxia, have a lower AR expression (Fig. 6F left panel). In addition, there is a statistically significant negative correlation between AR and YBX1 in protein expression (Fig. 6F middle panel). And the CSC marker CD24 is also negatively correlated with patient overall survival (Fig. 6F right panel).

Together, these clinical results from Fig. 6A–F and Additional file 1: S4A-B are consistent role of AR/lncRNA TCFL5-2/YBX1/SOX2 signaling axis found through in vitro studies, and further demonstrated the clinical significance of this signaling axis for the patient prognosis likely as a result of regulating CSC formation.

The AR/lncTCFL5-2/YBX1/SOX2 signaling axis in RCC xenograft model

In order to test the validity of the in vitro data, we performed the in vivo orthotopic mouse tumor formation assay with the OSRC-2 cells expressing firefly luciferase. The cells were divided into 4 groups, vector control with or without hypoxia for 48 h, and sh-lncTCFL5-2 group with and without hypoxia for 48 h. These cells were inoculated into the left kidney capsule of nude mice and tumor sizes and metastases were evaluated. Once the tumor formation was detectable, we also compared the efficacy of Sunitinib treatment in the groups with and without lncTCFL5-2. The In Vivo Imaging Systems (IVIS) was used to monitor tumor growth. After 8 weeks, we sacrificed the mice and examined the tumor metastasis in lung, liver, spleen, and right kidney with the help of IVIS.

The results showed that RCC cells with a reduction of lncTCFL5-2 pre-cultured under hypoxia and normoxia resulted in significantly slower tumor growth than the control group (Fig. 6G). When mice were treated with 40 mg/kg Sunitinib for two weeks, the tumors were significantly smaller in the sh-lncTCFL5-2 group than the control group (Fig. 6H). After 8 weeks, we sacrificed the mice, and the tumor volumes and weight measurement also confirmed the conclusion that sh-lncTCFL5-2 suppresses tumorigenicity under both normoxia and hypoxia (Fig. 6I, J). In addition, treating mice with 40 mg/kg Sunitinib every day for 2 weeks, the tumor size reduction is more significant in cells with a reduction of lncTCFL5-2 group than the control group (Fig. 6I, K). The tumor metastasis was also evaluated via the IVIS, which revealed that less metastasis occurred in the sh-TCFL5-2 group under both normoxia and hypoxia (Fig. 6L). The IHC staining also indicated that AR expression decreased clearly under hypoxia, and both YBX1 and SOX2 expression decreased sharply in sh-lncTCFL5-2 group consistent with the in vitro finding (Fig. 6M).

Together, these data indicated that AR/lncTCFL5-2/YBX1/SOX2 signaling axis plays a critical role in regulating CSC formation in vivo. Repression of lncTCFL5-2 expression likely will promote Sunitinib efficacy for RCC in vivo (Fig. 7).

Fig. 7

Mechanisms and regulatory pathways of hypoxia increases RCC CSCs phenotype in RCC. Hypoxia may increase RCC CSCs phenotype via altering the AR/lncTCFL5-2/YBX1/SOX2 signaling axis