hnRNPK is responsible for the malignant phenotypes of cancer cells
We investigated the role of hnRNPK in malignant phenotypes including metastatic potential and proliferation in HeLa cells. To investigate the role of hnRNPK in cancer malignancy, two independent HNRNPK-specific siRNAs were designed and introduced into HeLa cells. For overexpression of hnRNPK, we used a pcDNA/Flag-hnRNPK vector constructed previously (8). Individual and mixture of HNRNPK siRNAs efficiently decreased hnRNPK expression (Fig. 1a). Conversely, introduction of Flag-hnRNPK resulted in a significant increase of hnRNPK in a dose-dependent manner (Fig. 1b). Using the siRNAs, we tested the effect of knockdown of hnRNPK on the invasiveness of HeLa cells. The invasive ability was reduced by approximately 65% (siRNA #1), 52% (siRNA #2), and 54% (mixed siRNA), respectively, compared to the control siRNA (Fig. 1c). Conversely, overexpression of hnRNPK enhanced the invasive ability by approximately 1.8–2.7-fold compared to the blank Flag vector (Fig. 1d), indicating that hnRNPK is closely associated with the invasiveness of cancer cells.
Two other distinctive features of malignancy, the proliferation rate and clonogenicity, were also examined. We observed that knockdown of hnRNPK using two individual siRNAs or a mixture resulted in a decrease in the proliferation rate (Fig. 1e). On the other hand, hnRNPK-overexpressing cells showed higher proliferation rate than the blank vector control cells (Fig. 1f). Next, we examined the colony forming ability following knockdown or overexpression of hnRNP. Knockdown of hnRNPK abrogated the colony forming ability to less than 50% of the control (Fig. 1g). In contrast, the number of colonies was dose-dependently increased following hnRNPK overexpression (Fig. 1h). These results showed that increased hnRNPK resulted in higher proliferative potential. Collectively, our findings demonstrate that hnRNPK is responsible for the malignant characteristics including high invasiveness and rapid proliferation.
LINC00263 is identified as a novel hnRNPK-regulated lincRNA
We hypothesized that long intergenic non-coding RNAs (lincRNAs) are involved in hnRNPK-mediated cancer malignancy. To identify hnRNPK-regulated lincRNAs, we performed RNA sequencing using hnRNPK-silenced HeLa cells. The various plots representing RNA sequencing data (scatter plot, volcano plot, and volume plot) and gene ontology (GO) analysis are shown in Supplementary Fig. 1a and 1c. Based on the data analysis and processing, five lincRNAs were identified to be significantly regulated by hnRNPK (Fig. 2a and 2b). Of the five lincRNAs, two lincRNAs, LINC00618 and LINC02246 were upregulated following knockdown of hnRNPK. In contrast, the expression levels of LINC01137, LINC00263, and LINC00162 were lower in hnRNPK-silenced cells than in control cells. Since LINC00263 showed the most significant effect on the metastatic potential (data not shown) of the cells, we chose to investigate its role in the control of cancer malignancy through hnRNPK.
To verify the RNA sequencing data, we performed transient knockdown of hnRNPK and observed substantial decrease in hnRNPK expression with two individual siRNAs (Fig. 2c). Further, knockdown of hnRNPK also reduced the level of LINC00263 significantly (Fig. 2d). We next sought to determine whether LINC00263 affects hnRNPK expression. Following transient transfection of cells with two independent LINC00263-specific siRNAs, the levels of HNRNPK mRNA and LINC00263 were determined by RT-qPCR analysis. Although both LINC00263-targeting siRNAs caused a substantial decrease in LINC00263 level, they did not affect the levels of hnRNPK protein and mRNA (Fig. 2e and 2f, respectively). In addition, a specific siRNA targeting the 3'UTR of the HNRNPK mRNA was designed and introduced into HeLa cells with Flag-hnRNPK overexpression vector. Knockdown of hnRNPK by 3'UTR-targeting siRNA efficiently decreased the expression of hnRNPK without significant change in the ectopic hnRNPK (Flag-hnRNPK) (Fig. 2g). In accordance with previous results, introduction of the 3'UTR-targeting siRNA resulted in decreased expression of LINC00263. However, the level of LINC00263 was restored to control level following ectopic expression of hnRNPK (Fig. 2h).
From above results, we identified LINC00263 as a novel hnRNPK-regulated lincRNA. Therefore, we investigated detailed molecular mechanism by which hnRNPK regulates the expression of LINC00263. Since five hnRNPK motives are predicted in the sequence of LINC00263 using a bioinformatic tool for RBP binding site prediction (http://rbpmap.technion.ac.il/) (Supplementary Fig. 2a and 2b), the direct interaction between hnRNPK and LINC00263 was examined through RNP-IP experiment. LINC00263 was found to be highly enriched approximately 20-fold in endogenous hnRNPK IP material compare to control IgG (Fig. 2i). In addition, RNP-IP using full-length Flag-hnRNPK and its various deletion mutants (ΔKH1, ΔKH1/2, ΔKH2, and ΔKH3) revealed that interaction of hnRNPK with LINC00263 was dependent on its K homology 1 (KH1) and KH2 domains (Fig. 2j). These results suggested that hnRNPK might regulate LINC00263 at post-transcriptional level. Therefore, we examined whether hnRNPK influences the stability of LINC00263 (Fig. 2k). Knockdown of hnRNPK induced more rapid decrease in LINC00263 compare to control. The approximate half-life of LINC00263 in control and hnRNPK-silenced cells was calculated as 13.5 h and 7.6 h, respectively. However, the level of GAPDH mRNA was barely affected by knockdown of hnRNPK. Collectively, we demonstrate that hnRNPK stabilized LINC00263 through direct interaction and LINC00263 is a novel target of hnRNPK.
LINC00263 promotes malignant phenotypes including invasiveness, proliferation, and clonogenicity
To investigate whether LINC00263 is responsible for hnRNPK-mediated invasiveness, we performed transient knockdown or overexpression of LINC00263 and examined the invasive ability of the cells using Transwell invasion assay. Knockdown of LINC00263 in HeLa cells using two independent siRNAs resulted in approximately 50% decrease in the number of invading cells (Fig. 3a). Conversely, overexpression of LINC00263 potentiated the invasive ability of HeLa cells in a dose-dependent manner (Fig. 3b). The level of LINC00263 in the overexpressing cells was verified by RT-qPCR analysis (Supplementary Fig. 3a). Next, the proliferation rate and clonogenicity were assessed in LINC00263-silenced and -overexpressing HeLa cells. We observed decreased proliferation rate in LINC00263-silenced cells by two individual siRNAs (Fig. 3c). In contrast, the proliferation rate tended to increase in a dose-dependent manner in LINC00263-overexpressing cells (Fig. 3d). We also examined the effect of LINC00263 knockdown on colony forming ability of the cells. We found that approximately 40% decrease in the number of colonies in two individual siRNA-transfected cells (Fig. 3e), while the colony forming ability was increased following overexpression of LINC00263 (Fig. 3f). These results indicate that LINC00263 is associated with the oncogenic function of hnRNPK.
miR-147a is involved in the regulation of cancer malignancy by hnRNPK/ LINC00263
To determine the molecular mechanism through which LINC00263 positively regulates malignant properties, we first examined the subcellular localization of LINC00263 using cellular fractionation assay. For verification of the appropriate cellular fractions, the levels of a-tubulin (cytosolic marker) and lamin B (nuclear marker) were analyzed in each fraction by Western blot analysis (Fig. 4a). The level of LINC00263 in each fraction was determined by RT-qPCR analysis. The levels of 18S, GAPDH, and ACTB mRNA were assessed for reference (Fig. 4b). The cellular fractionation assay revealed that LINC00263 was mainly localized in the cytosolic fraction. Additionally, we performed Argonaute 2 immunoprecipitation (Ago2-IP) assay to examine whether LINC00263 was associated with the function of miRNAs (Fig. 4c). LINC00263 was more enriched in Ago2-IP compared to control IgG-IP, indicating that LINC00263 is involved in the regulatory pathway of miRNAs (results of three independent experiments are shown in Supplementary Fig. 5a).
From the above results, we concluded that LINC00263 is mainly located in the cytosol and co-immunoprecipitated with Ago2 antibody, which suggests that LINC00263 may function as a competitive endogenous RNA (ceRNA) for miRNA. Therefore, we hypothesized that LINC00263 might be closely involved in the oncogenic function of hnRNPK by sponging tumor-suppressor miRNAs. To identify LINC00263-associated miRNAs, seven ASOs consisting of complementary sequences of LINC00263 were designed to perform ASO pull-down experiments. Four ASOs for LacZ were used for control IP (Fig. 4d and Supplementary Fig. 4a). To test the efficiency of the ASO pull-down, the levels of LINC00263 and ACTB mRNA in ASO pull-down materials were determined by RT-qPCR analysis. Whereas ACTB mRNA was not enriched, LINC00263 was selectively enriched in the pull-down materials using the corresponding ASOs as compared to LacZ ASO (Fig. 4e). To screen LINC00263-bound miRNAs, small RNA sequencing was performed using the RNA isolated from the ASO pull-down. Analysis of sequencing data revealed that 24 miRNAs showed higher enrichment in LINC00263 ASO pull-down material than in LacZ ASO pull-down, suggesting that they bound to LINC00263 directly or indirectly (Fig. 4f and Supplementary Fig. 4b). Next we predicted the potential miRNA binding sites within LINC00263 sequence using a miRNA target discovery tool RNA22 (https://cm.jefferson.edu/rna22). This bioinformatic tool revealed that LINC00263 possessed MREs for only four miRNAs (miR-147a, miR-492, miR-601, and miR-1268a) out of the 24 miRNAs found by the ASO pull-down analyses (Supplementary Fig. 4c). Since miR-147a showed the most significant folding energy, we chose to further investigate whether miR-147a was responsible for the oncogenic function of hnRNPK/LINC00263.
To confirm the interaction between miR-147a and LINC00263, we examined the enrichment of LINC00263 in Ago2-IP. Precursor miR-147a (pre-miR-147a) was used for overexpression and the level of miR-147a in transfected cells was found to be markedly increased (Supplementary Fig. 3b). Ago2-IP assay indicated that overexpression of miR-147a resulted in an increase in LINC00263 in Ago2 IP materials, indicating that miR-147a guided the interaction of LINC00263 with Ago2 to form miRNA-induced silencing complex (miRISC) (Fig. 4g). In contrast, inhibition of miR-147a using anti-miR-147a (Supplementary Fig. 3c) decreased the level of LINC00263 in Ago2-IP (Fig. 4h). These results indicate that LINC00263 is associated with miR-147a-guided RISC. Therefore, we analyzed whether miR-147a regulated LINC00263 expression. Decreased level of LINC00263 was observed in miR-147a-overexpressing cells compared to that in the control (Fig. 4i); conversely, LINC00263 was highly expressed following miR-147a knockdown using anti-miR-147a (Fig. 4j). In general, the reduction of a ceRNA leads to an increase in the ceRNA-bound miRNAs, thus the repressing function of miRNA is potentiated in ceRNA-silenced cells. Accordingly, we assessed the level of miR-147a in hnRNPK- and LINC00263-silenced cells (Fig. 4k). Knockdown of both hnRNPK and LINC00263 resulted in approximately 2-fold increase of miR-147a, indicating that LINC00263 acts as a ceRNA for miR-147a. Two MREs of miR-147a in LINC00263 were predicted by bioinformatic tool (Supplementary Fig. 7a and 7b). Consequently, we constructed luciferase reporter vectors containing wild-type or mutant sequence of miR-147a MREs. In both reporter vectors, overexpression of miR-147a suppressed the expression of luciferase in wild-type reporter vector but not in the mutant (Fig. 4l). The results of Ago2-IP and luciferase reporter assay suggest that miR-147a directly binds to LINC00263.
Next, we tested whether miR-147a influences malignant phenotypes. Invasiveness was reduced to less than 30% of the control following overexpression of miR-147a; conversely, inhibition of miR-147a using anti-miR-147a resulted in approximately 4-fold increased invasive ability (Fig. 4m and 4n, respectively). In addition to invasiveness, proliferation rate and colony forming ability were also regulated by miR-147a. Under conditions of high miR-147a levels, the proliferation rate and clonogenicity were diminished (Fig. 4o and 4q, respectively). On the other hand, decrease in miR-147a level resulted in higher proliferative and clonogenic abilities compared to those of the control (Fig. 4p and 4r, respectively). Collectively, we concluded that LINC00263 controls malignant properties by functioning as a ceRNA of the tumor-suppressor, miR-147a.
CAPN2 is a target of hnRNPK/ LINC00263 /miR-147a axis
To search for target genes responsible for the oncogenic function of hnRNPK/LINC00263/miR-147a, we conducted RNA sequencing using total RNA isolated from hnRNPK- and LINC00263-silenced HeLa cells (Supplementary Fig. 1a and 1b) and tried to identify common target genes. TargetScan (http://www.targetscan.org) was used to predict miR-147a target genes. As shown in Fig. 5a, eight genes (CAPN2, CCND1, CDKN1A, CSDC2, L1CAM, PAQR4, PARP12, and TRIM47) were identified as common target genes that are simultaneously regulated by hnRNPK, LINC00263, and miR-147a. RT-qPCR analysis indicated that knockdown of either hnRNPK or LINC00263 significantly decreased the level of CAPN2 mRNA, suggesting that it may be a putative target of hnRNPK/LINC00263/miR-147a (Supplementary Fig. 6). To check the possibility that hnRNPK regulates CAPN2 by directly binding to CAPN2 mRNA, RNP-IP experiment was performed. Whereas LINC00263 was highly enriched in hnRNPK IP material, CAPN2 mRNA was barely bound to hnRNPK (Fig. 5b), indicating that hnRNPK regulates the expression of CAPN2 in an indirect way. Subsequently, we examined the role of hnRNPK/LINC00263/miR-147a in regulating CAPN2 expression. Overexpression of miR-147a resulted in decreased CAPN2 protein expression without any change in hnRNPK. Consistent with the results of the Western blot and RT-qPCR analyses, overexpression of miR-147a decreased the level of CAPN2 mRNA (Fig. 5c). Conversely, inhibition of miR-147a using anti-miR-147a resulted in increased CAPN2 protein and mRNA expression (Fig. 5d). Overexpression and inhibition efficiencies of the pre-miR-147a and anti-miR-147a, respectively, were confirmed by determining the level of miR-147a by RT-qPCR analysis (Supplementary Fig. 3b and 3c, respectively). To determine direct binding between CAPN2 mRNA and miR-147a, the level of CAPN2 mRNA in Ago2-IP material was assessed. Ago2-IP revealed that miR-147a increased the enrichment of CAPN2 mRNA in miRISC (Fig. 5e); conversely, knockdown of miR-147a using antisense miRNA decreased the level of CAPN2 mRNA in the Ago2-IP material (Fig. 5f). In addition to Ago2-IP, luciferase reporter vectors containing the wild-type and mutant MRE of miR-147a were constructed to confirm the direct binding of miR-147a to the 3'UTR of CAPN2 mRNA. Overexpression of miR-147a inhibited luciferase activity in wild-type vector, whereas it did not affect the expression of luciferase in mutant vector (Fig. 5g).
Next, we tested the effect of hnRNPK and LINC00263 silencing on CAPN2 expression. Knockdown of hnRNPK and LINC00263 significantly decreased the level of CAPN2 protein and mRNA (Fig. 5h). CAPN2 mRNA was enriched in Ago2-IP following knockdown of hnRNPK or LINC00263 (Fig. 5i). These results indicated that decrease in hnRNPK strengthens the function of miR-147a by reducing LINC00263, a ceRNA of miR-147a. To validate whether CAPN2 is involved in the regulation of malignant phenotypes by hnRNPK/LINC00263/miR-147a, the effect of CAPN2 silencing on invasiveness, proliferation, and clonogenicity was examined. Introduction of CAPN2-specific siRNA into HeLa cells markedly decreased CAPN2 expression (Fig. 5j). As expected, knockdown of CAPN2 decreased the number of invading cells (Fig. 5k), inhibited cell proliferation (Fig. 5l), and suppressed colony forming ability (Fig. 5m). Collectively, we demonstrated that CAPN2 was responsible for the oncogenic function as a target of hnRNPK/LINC00263/miR-147a.
Extracellular signal-regulated kinase (ERK) and p70S6K pathways are involved in the function of hnRNPK/ LINC00263 /miR-147a/CAPN2
The proteome profiler human p-kinase array was used to identify common signaling pathways of the hnRNPK/LINC00263/miR-147a/CAPN2 axis. Phosphorylation of ERK and p70S6K was found to be diminished in HNRNPK- or LINC00263-silenced cells compared to the controls (Supplementary Fig. 8a and 8b). To verify this observation, the levels of p-ERK and p-p70S6K were assessed. Western blot analysis indicated that knockdown of hnRNPK or LINC00263 reduced phosphorylated ERK and p70S6K. Further, decreased expression of CAPN2 using miR-147a or siRNA inhibit the activation of ERK and p70S6K (Supplementary Fig. 8c). These results demonstrate that ERK and p70S6K pathways are closely involved in the function of hnRNPK/LINC00263/miR-147a/CAPN2.
Repression of malignant capabilities is restored by miR-147a inhibition or CAPN2 overexpression
From the above results, we found that hnRNPK-regulated LINC00263 decoys miR-147a and thus increases CAPN2 expression. To verify our findings, we performed rescue experiments by downregulating miR147-a using anti-miR-147a. HeLa cells were transfected with HNRNPK or LINC00263 siRNA and with anti-miR-147a. RT-qPCR analysis results showed the level of miR-147a was efficiently decreased following introduction of anti-miR-147a. The level of miR-147a was significantly decreased not only in the control but also in hnRNPK- or LINC00263-silenced cells where miR-147a is upregulated by lowering its ceRNA, LINC00263 (Fig. 6a). Whereas knockdown of hnRNPK or LINC00263 increased the level of CAPN2 mRNA in Ago2 IP, the inhibition of miR-147a by anti-miRNA lowered the enrichment of CAPN2 mRNA in miRISC, indicating that miR-147a is responsible for the repression of CAPN2 in hnRNPK- and LINC00263-silenced cells (Fig. 6b). Subsequently, we analyzed the level of CAPN2 protein and mRNA in transfected cells. Inhibition of miR-147a using anti-miRNA reversed the decrease in CAPN2 protein and mRNA caused by the knockdown of hnRNPK and LINC00263 as well (Fig. 6c). Next, we tested the effect of anti-miR-147a on invasiveness and clonogenicity. Consistent with the recovery of reduced CAPN2 expression, invasiveness and colony forming abilities were restored by anti-miR-147a (Fig. 6d and 6e, respectively). These results demonstrate that miR-147a is closely involved in the regulation of CAPN2 expression, and thus plays an important role in the gain of malignant phenotypes by hnRNPK/LINC00263.
In addition to inhibition of miR-147a, we examined whether the ectopic expression of CAPN2 reverses the lowered malignant capabilities resulting from the knockdown of hnRNPK and LINC00263. The appropriate concentration of overexpression vector was determined by introducing various concentrations of Flag-CAPN2 vector (Supplementary Fig. 3d). For the rescue experiment, HeLa cells were simultaneously transfected with Flag-CAPN2 vector and HNRNPK- or LINC00263-targeting siRNA. Western blot analysis showed that knockdown of hnRNPK and LINC00263 decreased CAPN2 expression and that ectopic CAPN2 did not affect the expression of hnRNPK (Fig. 6f). Invasive and clonogenic abilities were also examined under the same conditions. As observed earlier, invasiveness was significantly decreased following the knockdown of hnRNPK and LINC00263. However, following ectopic overexpression of CAPN2, the invasive ability was restored (Fig. 6g). Consistent with the results of invasion assay, the colony forming assay revealed that ectopic CAPN2 restored the clonogenic ability that was reduced in the hnRNPK- and LINC00263-silenced cells (Fig. 6h). Collectively, we concluded that CAPN2 is a major effector of the oncogenic function of hnRNPK/LINC00263/miR-147a.
hnRNPK/ LINC00263 /miR-147a/CAPN2 axis is applicable to various types of cancer cells
To generalize our findings to various types of cancer cells, the regulatory action of hnRNPK/LINC00263/miR-147a/CAPN2 was examined in two lung cancer cells (H460 and H1299). LINC00263 was recently reported to be abnormally regulated and play an important role in the progression of lung cancer cells (13). Therefore, we compared the level of LINC00263 in two GSE datasets (Supplementary Fig. 9a and 9b). LINC00263 was highly expressed in non-small cell lung cancer tissues compared to non-malignant tissues (GSE81089) and in tumor tissues compared to normal (GSE40419) tissues. To confirm the observations of the GSE dataset, we compared the level of HNRNPK mRNA and LINC00263 in two lung cancer cells (H460 and H1299) with those in non-cancerous WI-38 cells (Fig. 7a). Compared to that in WI-38 cells, the expression of HNRNPK and LINC00263 was significantly increased in both the lung cancer cells. Further, the expression level of HNRNPK mRNA and LINC00263 was positively correlated. Interestingly, H1299 cells showed higher invasive ability than H460 cells (Supplementary Fig. 9c), indicating that the higher the invasiveness, the greater the increase of HNRNPK mRNA and LINC00263. Consistent with the previous results, knockdown of hnRNPK and LINC00263 induced a decrease of CAPN2 mRNA in both the lung cancer cells (Fig. 7c). In addition to HNRNPK and LINC00263 siRNA, introduction of pre-miR-147a and CAPN2 siRNA into H460 and H1299 cells also decreased the expression of CAPN2 (Fig. 7d). As expected, the invasive and clonogenic abilities were diminished following knockdown of hnRNPK and LINC00263 (Fig. 7e). Since H460 cells exhibited relatively lower expression of LINC00263, we tested whether overexpression of LINC00263 potentiates invasive ability (Supplementary Fig. 9d). We found that LINC00263 increased the number of invading cells in a dose-dependent manner. The number of colonies was also decreased in HNRNPK- and LINC00263-silenced cells (Fig. 7f). Overexpression of miR-147a by introducing pre-miR-147a lowered invasive and colony forming abilities of both the lung cancer cells (Fig. 7g and 7 h, respectively). Conversely, inhibition of miR-147a using anti-miR-147a induced an increase in the invasiveness and clonogenicity. The oncogenic function of CAPN2 was verified by assessing the number of invading cells and colonies in CAPN2-silenced cells. From these results, we confirmed that hnRNPK/LINC00263/miR-147a/CAPN2 regulatory axis is very closely related to the malignant phenotype of lung cancer cells.
Next, the role of hnRNPK/LINC00263/miR-147a in regulation of CAPN2 expression was verified in various other cancer cells including DLD1 and LoVo (colon cancer), A375 (melanoma), T98G (neuroblastoma), and A172 (astrocytoma) cells. All the cells tested showed suppression of CAPN2 expression as observed in HeLa and lung cancer cells. Briefly, CAPN2 expression was decreased in hnRNPK- or LINC00263-silenced cells. In addition, overexpression of miR-147a also resulted in suppression of CAPN2. From above results, we confirmed that our findings are applicable to various types of cancers.