The long non-coding RNA CYTOR drives colorectal cancer progression by interacting with NCL and Sam68

Two cohorts of clinical samples from the tissue bank in our laboratory were used in this study according to protocols approved by the Internal Review Board of Zhejiang University. A total of 138 pairs of CRC tissues and matched normal tissues were collected to measure CYTOR levels by qRT-PCR. Each experiment was performed at least three times. Another cohort, including 144 CRC FFPE tissue samples, was used for the detection of NCL and Sam68 expression by immunohistochemistry (IHC), and the IHC scores were evaluated by at least two pathologists. The pathological diagnoses of all samples were evaluated by pathologists. All patients with familial adenomatous polyposis, hereditary non-polyposis CRC, or inflammatory bowel disease were excluded. All tissue samples were obtained from colorectal adenocarcinoma patients without any radiotherapy, chemotherapy, or other adjuvant treatment prior to surgery and diagnosis.

The transcriptome expression profiling of CRC and relevant clinical information in this study were identified by searching public databases online. The key words “colon cancer or colorectal cancer and expression profiling by array” were used to search the GEO databases. In addition, the data from the TCGA database were also included in our study to ensure that the data on CYTOR expression in CRC were not missed.

CRC cell lines were cultured and transfected with either CYTOR shRNA or overexpression vectors. The proliferation and metastasis capacity of the cells were assessed by CCK-8, colony assays and transwell migration/invasion assays in vitro and by subcutaneous tumor growth in nude mice or mouse tail-vein injection assays in vivo. A revised ChIRP protocol and mass spectrometry (MS) were used to pull down and identify CYTOR RNA-binding proteins. RNA immunoprecipitation (RIP) was carried out to verify the interaction between CYTOR and its binding proteins. CYTOR EXON-1-deleted and EXON-4-deleted cells were constructed by combining CRISPR/Cas9 and the donor vector. Co-immunoprecipitation and RIP were performed to confirm the interaction motif of CYTOR. A luciferase reporter system and an immunoblotting assay were used to investigate the downstream regulatory targets and interacting proteins of CYTOR.

All further information can be found in the Additional file 1 section.

Our previous expression profile data [11] showed that CYTOR was up-regulated in CRC and tumor budding cells (Fig. 1a). We further examined the expression of CYTOR in 138 pairs of matched CRC and normal tissues by qRT-PCR, which revealed that CYTOR was up-regulated in CRC tissues (Fig. 1b). Moreover, there was a significant correlation between CYTOR expression and overall patient survival, and patients with higher CYTOR expression showed worse clinical outcome (Fig. 1c). To validate our results, we evaluated the expression levels of CYTOR in the matched pairs of CRC and normal tissue samples using the TCGA RNA-seq database (Fig. 1d), GEO GDS2947 database (Fig. 1g), GEO GSE31737 database (Fig. 1h), GEO GSE32323 database (Fig. 1i) and GEO GSE41328 database (Fig. 1j). Consistent with our observation, the results from these datasets showed significantly up-regulated CYTOR expression in tumor tissue samples. More interestingly, overall survival data showed a poorer prognosis with higher CYTOR levels in the TCGA database (Fig. 1e), GEO GSE38832 database (Fig. 1k), GEO GSE39582 database (Fig. 1m), GEO GSE17536 database (Additional file 2: Figure S1A), GEO GSE17537 database (Additional file 2: Figure S1C), and GEO GSE29621 database (Additional file 2: Figure S1H). Similarly, the expression of CYTOR in CRC was also negatively correlated with disease- or recurrence-free survival in the TCGA database (Fig. 1f), GEO GSE38832 database (Fig. 1l), GEO GSE39582 database (Fig. 1n), GEO GSE17536 database (Additional file 2: Figure S1B), GEO GSE24549-GPL5175 database (Additional file 2: Figure S1J), and GEO GSE24550-GPL11028 database (Additional file 2: Figure S1K).

However, no such correlation was observed between CYTOR and overall, disease-free or recurrence-free survival in some other databases, such as GEO GSE17537 (Additional file 2: Figure S1D), GEO GSE56699 (Additional file 2: Figure S1E and F), GEO GSE16125 (Additional file 2: Figure S1G), GEO GSE24549-GPL11028 (Additional file 2: Figure S1I), GEO GSE24550-GPL5175 (Additional file 2: Figure S1L), GEO GSE31595 (Additional file 2: Figure S1M) and GEO GSE33113 (Additional file 2: Figure S1N). To address this issue, we performed a meta-analysis to systematically evaluate the association between CYTOR and CRC survival risk, combining our data with all the other online CYTOR data. When the cutoff value of CYTOR expression was set according to the receiver operating characteristic (ROC) curve, the patients with higher expression of CYTOR had significantly poorer overall survival (OS; pooled HR, 1.86; 95% CI, 1.50–2.30; Additional file 3: Figure S2A) and disease- or recurrence-free survival (DFS; pooled HR, 1.67, 95%CI, 1.33–2.09; Additional file 3: Figure S2B). When the cutoff value of CYTOR expression was set as P50, CYTOR expression was also negatively associated with overall survival (OS; pooled HR, 1.22; 95% CI, 1.01–1.48; Additional file 3: Figure S2C) and disease- or recurrence-free survival (DFS; pooled HR, 1.30, 95%CI, 1.07–1.58; Additional file 3: Figure S2D). Funnel plots further showed no bias among these databases, which confirmed that CYTOR was a risk factor for survival in colorectal cancer (Additional file 4: Figure S3). Taken together, these clinical data revealed the strong association between CYTOR expression and CRC development/prognosis.

Next, we examined the CYTOR levels in CRC cell lines, and higher expression was found in RKO, SW480 and SW620 cells than in HCT116 or HCT8 cells (Fig. 2a). When we stably knocked down CYTOR in RKO, SW480 and SW620 cells (Fig. 2b), the colony-forming potential of RKO and SW620 was inhibited (Fig. 2c), and no colonies of SW480 cells were observed. In addition, shRNA-mediated CYTOR knockdown significantly decreased migration and invasion compared with the scramble control in these cell lines (Fig. 2d). To avoid off-target effects of shRNA, we designed and synthesized another two siRNAs to knockdown CYTOR in RKO, SW480 and SW620 cells (Additional file 5: Figure S4A). The results showed knockdown of CYTOR by siRNAs also inhibited the potential of colony-forming (Additional file 5: Figure S4B), migration and invasion in RKO (Additional file 5: Figure S4C), SW480 (Additional file 5: Figure S4D) and SW620 (Additional file 5: Figure S4E) cell lines, which were consistent with the results from shRNA. Furthermore, ectopic expression of CYTOR in the HCT116 and HCT8 cell lines (Fig. 2e) not only promoted colony formation (Fig. 2f) but also enhanced the capacity for migration and invasion (Fig. 2g). More interestingly, when the ectopic expression of CYTOR was inhibited by specific siRNAs in the transfected HCT116 cells (Fig. 2h), the enhanced migration and invasion were also decreased by these siRNAs (Fig. 2i). These results demonstrated that CYTOR could promote the anchorage-independent growth, migration and invasion of CRC cells in vitro. As tumor budding cells are considered a histological phenomenon of EMT, which contributes to tumor metastasis [19‐21], we also evaluated the relationship between CYTOR and EMT markers. In SW480 and SW620 cells, knockdown of CYTOR increased E-cadherin expression while decreasing Vimentin expression (Fig. 2j). On the other hand, overexpression of CYTOR inhibited E-cadherin expression while increasing Vimentin expression in HCT8 and HCT116 cells (Fig. 2k). Analysis of the current GEO database (GSE29621 and GSE38832) showed that CYTOR expression was negatively correlated with the epithelial marker E-cadherin and positively correlated with mesenchymal markers including Vimentin, N-cadherin, FN1, Twist, and MMP9 (Additional file 6: Figure S5).

We also investigated whether CYTOR was functionally involved in CRC progression in mouse xenografts. When RKO cells transfected with CYTOR shRNA were subcutaneously injected into nude mice, the volume (Fig. 3a) and weight (Fig. 3b) of the xenograft tumors were significantly decreased compared with the scramble control group. Moreover, tumor growth was also repressed by CYTOR knockdown as shown in the tumor growth curve (Fig. 3c). By contrast, overexpression of CYTOR increased the volume (Fig. 3d) and weight (Fig. 3e) of the xenograft tumors and promoted tumor growth in vivo (Fig. 3f). To explore whether CYTOR also promotes CRC metastasis in vivo, we intravenously injected luciferase-labeled control or CYTOR knockdown RKO cells into NOD/SCID mice and subjected them to bioluminescent imaging to monitor metastasis. The results showed that the whole-body luminescence signals in the CYTOR knockdown group were significantly reduced compared with those of the control group after 30 days (Fig. 3g). Overall, our data indicated that CYTOR could promote EMT and CRC progression.

To explore the detailed mechanism whereby CYTOR regulates CRC progression, we first investigated the distribution of CTYOR in RKO cells and found it distributed in both cytoplasm and nucleus (Additional file 7: Figure S6A). Then, we designed a set of specific probes labeled with biotin to pull down the proteins that directly bind to CYTOR [8, 22]. The efficiency and specificity of the probes were confirmed by PCR (Fig. 4a). We isolated the proteins from the pull-down complex by SDS-PAGE electrophoresis (Additional file 7: Figure S6B) and identified NCL and Sam68 as the CYTOR-binding proteins by MS assays (Additional file 7: Figure S6C) in RKO cells. Furthermore, the interactions between CYTOR and NCL, CYTOR and Sam68 were confirmed by immunoblotting (Fig. 4b). In addition, RIP assays revealed that both NCL (Fig. 4c) and Sam68 (Fig. 4d) could pull down CYTOR directly.

The results above demonstrated that NCL and Sam68 could bind to CYTOR; next, we investigated whether a complex could be formed between NCL and Sam68. The co-IP assay showed that NCL was able not only to pull down Sam68 (Fig. 4e) but also to be immunoprecipitated by Sam68 (Fig. 4f) in live cells. Then, we simultaneously transfected the FLAG-tagged NCL (NCL-FLAG) and HA-tagged Sam68 (Sam68-HA) expression vectors into 293 T cells, and the harvested cell lysates were treated with either RNase inhibitor or RNaseA, followed by dual co-IP assays with either anti-FLAG or anti-HA antibody. The results showed that NCL and Sam68 protein could be co-immunoprecipitated reciprocally in the RNase-inhibitor-treated samples (Fig. 4g and h, 3rd lanes) but not in RNaseA-treated samples (Fig. 4g and h, 6th lanes). These results suggested that the interaction between NCL and Sam68 depended on the existence of RNA.

The NCL protein consists of six main domains, including the N-terminal domain, four RNA binding domains (RBDs) and the GAR domain [23]. To further identify the sites on NCL that interact with CYTOR, we constructed four deletion mutants of FLAG-tagged NCL (Fig. 4i). It was shown that in addition to full-length NCL (FL), the mutants that included the 3rd and 4th RBDs and the GAR domain could pull down CYTOR (Fig. 4j), indicating that these domains might be the key sites of interaction with CYTOR. Moreover, we also constructed KH-domain-deleted (ΔKH), N-terminus-deleted (ΔN) and C-terminus-deleted (ΔC) HA-tagged Sam68 expression vectors [24] (Fig. 4k). RIP results revealed that the ΔC mutant could bind CYTOR, while the ΔKH and ΔN mutants could not (Fig. 4l), demonstrating that the KH and N-terminal domains were the sites of interaction with CYTOR. More intriguingly, co-IP assays revealed that the 3rd and 4th RBDs and the GAR domain of NCL were required for binding to Sam68 (Fig. 4m), and the KH and N-terminal domains of Sam68 were indispensable to its interaction with NCL (Fig. 4n). Taken together, the above data indicated that CYTOR plays an essential part in forming a complex between NCL and Sam68 by interacting with specific motifs on the two proteins.

We next intended to identify the motifs in CYTOR that are essential for its interaction with NCL and Sam68. The secondary structure of CYTOR was predicted using the RNA fold web server [25]. It was found that only the EXON1 deletion (ΔEXON1) and not the other EXON deletion mutants (ΔEXON2, ΔEXON3 and ΔEXON4) of CYTOR could induce a dramatic change in the secondary structure relative to wild-type (WT) CYTOR (Fig. 5a). Thus, we used the CRISPR/Cas9 method (Fig. 5b) to construct ΔEXON1 (Fig. 5c) and ΔEXON4 RKO cells (Fig. 5d). The mutant cells were screened and confirmed as homozygous EXON1-deleted (Fig. 5e) and EXON4-deleted (Fig. 5f) clones by sequencing. As shown in Fig. 6a, the colony-forming potential of ΔEXON1 cells was reduced, but no changes were observed in ΔEXON4 cells. Similarly, deleting EXON1 of CYTOR inhibited cell migration and invasion, but deleting EXON4 had no such effect (Fig. 6b). Interestingly, a ChIRP assay showed that CYTOR-specific probes could pull down NCL and Sam68 in wild-type and ΔEXON4 RKO cells but not in ΔEXON1 RKO cells (Fig. 6c). Additionally, an RIP assay was performed to investigate the endogenous interactions between NCL, Sam68 and CYTOR in these mutant cells, and the results showed that while both NCL and Sam68 could bind to CYTOR in ΔEXON4 RKO cells, neither could pull down CYTOR in ΔEXON1 RKO cells (Fig. 6d). To determine whether complex formation between NCL and Sam68 also requires EXON1 of CYTOR, we evaluated their interaction in ∆EXON1 and ∆EXON4 cells using co-IP. The results showed that while NCL and Sam68 could be immunoprecipitated reciprocally in ∆EXON4 cells (Fig. 6f), no such interaction could be observed in ∆EXON1 cells (Fig. 6e).

To further verify the above results, we co-transfected 293 T cells with exogenous expression vectors for FLAG-tagged NCL, CYTOR EXON1-deletion (CYTOR-Δ1) or EXON4-deletion (CYTOR-Δ4) mutants. An RIP assay with anti-FLAG antibody revealed that NCL could bind to CYTOR-∆4 but not CYTOR-∆1 (Fig. 6g). Similarly, HA-tagged Sam68 could also pull down CYTOR-∆4 but not CYTOR-∆1 (Fig. 6h). Fortunately, we found two NCL-specific binding motifs [26] at the 125th nt of EXON1 and the 602nd nt of EXON4 in the CYTOR RNA sequence, as well as two Sam68-specific binding motifs [27] at the 60th nt of EXON1 and the 786th nt of EXON4 (Fig. 6i). Therefore, we constructed a series of CYTOR expression vectors with mutations in these sites (Fig. 6i). When FLAG-tagged NCL and CYTOR mutants with mutations in the NCL-specific sites (CYTOR-m1 for mutation in EXON1 or CYTOR-m2 for mutation in EXON4) were co-transfected into 293 T cells, an RIP assay showed that NCL could bind to CYTOR-m2 but not to CYTOR-m1 (Fig. 6j). Similarly, while HA-tagged Sam68 could not pull down the CYTOR mutant with the mutation in its specific-binding site on EXON1 (CYTOR-m3), it could bind to the CYTOR mutant with the mutation on EXON4 (CYTOR-m4) (Fig. 6k). Taken together, these data indicated that the formation of heterotrimeric complex of NCL, Sam68, and CYTOR requires their specific interacting sites in CYTOR EXON1.

Although it was confirmed that NCL and Sam68 could directly interact with CYTOR to form a complex, their function in CRC progression was still unknown. The results from online databases showed that NCL (Fig. 7a and Additional file 8: Figure S7A) and Sam68 (Fig. 7b and Additional file 8: Figure S7B) were both up-regulated in tumor samples compared with paired normal samples. In addition, the expression levels of CYTOR, NCL, and Sam68 were positively correlated (Fig. 7c). Our IHC results also revealed that higher NCL (Fig. 7d) and Sam68 (Fig. 7e) expression levels were associated with poorer survival rates. These results suggest that both NCL and Sam68 are involved in CRC progression.

In an effort to decipher the detailed functions of both proteins, we knocked down NCL or Sam68 using siRNAs (Fig. 7f), and it was shown that the proliferation of SW620 (Fig. 7g) and RKO (Additional file 8: Figure S7C) cells was inhibited under those conditions. Consistently with the CYTOR knockdown results, loss of NCL and Sam68 also repressed the migration and invasion activity of SW620 (Fig. 7h) and RKO (Additional file 8: Figure S7D) cells. Furthermore, the epithelial marker E-cadherin was increased and the mesenchymal marker Vimentin was decreased by NCL and Sam68 siRNAs (Fig. 7i). To verify these observations, we investigated the relationship between the expression of NCL/Sam68 and EMT markers in the GSE38832 database, and the results showed that both NCL and Sam68 expression were negatively correlated with the epithelial marker E-cadherin and positively correlated with the mesenchymal marker Vimentin, N-cadherin, Snail, and Twist (Additional file 9: Figure S8A for NCL and Additional file 9: Figure S8B for Sam68). Collectively, NCL and Sam68 may function as oncogenes to promote CRC EMT and progression.

The NF-κB pathway has been recognized as a key player in CRC progression, being involved in the EMT process [28, 29]; therefore, we evaluated the effect of the CYTOR-NCL-Sam68 complex on the NF-κB signaling pathway. The results showed that both NCL and Sam68 knockdown caused a dramatic decrease in phosphorylated P65 in CRC cell lines (Fig. 8a) and attenuated the transcriptional activity of NF-κB as shown by a luciferase promoter assay (Fig. 8b). Similarly, CYTOR knockdown also significantly decreased the level of phosphorylated P65, while overexpression of CYTOR increased the expression of phosphorylated P65 (Fig. 8c). Moreover, a luciferase promoter assay showed decreased NF-κB promoter activity when CYTOR expression was inhibited by shRNA (Fig. 8d). These results demonstrated that the NCL-CYTOR-Sam68 complex might promote CRC progression by activating the NF-κB signaling pathway.

To further evaluate the clinical significance of the CYTOR-NCL-Sam68 complex, we performed ROC analysis for CYTOR, NCL and Sam68, either separately or combined, in the TCGA, GSE17536 and GSE17537 databases. The results showed that the AUC for combined CYTOR, NCL, and Sam68 was greater than the individual AUC for CYTOR, NCL or Sam68 (Fig. 8e). Interestingly, in the 3D curve assay, the area with high expression of CYTOR, NCL and Sam68 represented samples from recurrent patients (Fig. 8f). The overall survival data from the GSE17536 database also showed that the patients with high expression of all three molecules (group 4) displayed the worst prognosis (Fig. 8g). These clinical data suggest that the NCL-CYTOR-Sam68 complex can be used as a biomarker for CRC recurrence and prognosis.