CRLM1 is overexpressed in colorectal liver metastases and associated with poor survival
To conduct a thorough examination of the lncRNA expression landscape in CRLM tissues, 54 whole transcriptome sequencing data (RNA-seq) from 18 CRC samples, comprising normal colon, primary CRC, and liver metastasis tissues, were downloaded [21]. To discover lncRNAs linked with CRC metastasis, an analysis pipeline integrating different screening methodologies was used (Fig. 1A). The RNA-seq data was used to predict novel lncRNAs and lncRNA functions (Additional file 2: Fig. S1A). By overlapping the expressed lncRNAs in these three groups, it was discovered that there was a high overlap for both known and novel lncRNAs (Additional file 2: Fig. S1B). The first component of the principal component analysis (PCA) of the expressed lncRNAs and mRNAs revealed a significant difference between normal and tumor tissues. Still, the second component revealed a clear but not very different split between primary and metastatic tissues (Additional file 2: Fig. S1C). Weighted gene co-expression network analysis (WGCNA) was used to investigate metastasis-associated lncRNAs, and it discovered a co-expression module (MEgreen, including 309 lncRNAs) that was strongly related to metastatic tissues in a positive manner but not with normal/primary tissues (Fig. 1B, Additional file 2: Fig. S1D, E, and Additional file 7: Table S4). Concurrently, DEG analysis was performed to identify lncRNAs that were abnormally up-regulated in metastatic tissues (Fig. 1C, Additional file 2: Fig. S1F, and Additional file 7: Table S5), yielding 486 lncRNAs. The dysregulated lncRNAs identified by WGCNA and DEG methods were overlapped to restrict the search targets, yielding 89 credible candidates (Fig. 1D).
Among these candidates, lncRNA-linc01767, later called colorectal liver metastasis one (CRLM1), was explored further. CRLM1 was expressed at much higher levels in metastatic tissues than in normal and primary tumor tissues (Fig. 1E). By categorizing CRC patients based on CRLM1 expression levels (TCGA dataset), it was discovered that CRLM1 expression was negatively associated with patient survival time (Fig. 1F). It is well known that lncRNAs can control the expression of nearby genes in a cis-acting way. CRLM1 is found on chromosome 1 between the numbers 56,414,931 and 56,420,384. It has 4 exons that make up a 545-nt lncRNA and 3 short introns. The expression correlation between CRLM1 and its adjacent gene phospholipid phosphatase 3 (PPAP2B/PLPP3) was investigated, and it was discovered that these two neighboring genes were strongly correlated in metastasis tissues but not in normal or primary tissues (Fig. 1G, H), implying that CRLM1 may be an important regulator of PPAP2B expression. The relevance of phospholipid phosphatase in cancer metastasis is currently unknown.
CRLM1 inhibits CRC cell apoptosis and promotes metastasis in vitro and in vivo
CRLM1 overexpression (CRLM1-OE) plasmids and siRNAs targeting CRLM1 were transfected into HCT116 and SW620 cell lines to further investigate the biological activities of CRLM1 in CRC cells. Using qRT-PCR, the researchers discovered that the expression of CRLM1 was significantly up-regulated or down-regulated in CRC cells transfected with the corresponding vectors or siRNAs (Fig. 2A, B). Apoptosis was shown to be considerably enhanced in CRLM1 knockdown cells (Fig. 2C, D). To examine the migratory and invasive characteristics of CRLM1 overexpression or knockdown CRC cells, wound-healing and transwell assays were performed. The results showed that up-regulation of CRLM1 considerably boosted the migratory and invasive abilities of CRC cells, whereas down-regulation of CRLM1 significantly reduced them (Fig. 2E–K). To investigate the role of CRLM1 in CRC metastasis in vivo, we implanted stable CRLM1-OE or sh-CRLM1 SW620 cells into the distal tip of the spleens of Balb/c nude mice. Compared to the control group, the mice in the CRLM1-OE group developed more widespread liver metastasis (Fig. 2L, M), while the sh-CRLM1 group showed reduced liver metastasis (Fig. 2N, O). These findings added to evidence that CRLM1 inhibits CRC cell apoptosis and increases metastasis in vitro and in vivo.
CRLM1 regulates the expression of CRC metastasis-related genes
A whole transcriptome sequencing experiment (RNA-seq) was done on three biological replicates of HCT116 cells transfected with CRLM1-OE, antisense-OE, and free plasmid control to reveal the underlying molecular mechanism for CRLM1 enhanced metastasis of CRC to the liver. CRLM1-regulated genes were identified utilizing DEG analysis with antisense-OE control and free plasmid control. The comparison revealed 1141 and 968 up-regulated genes for the antisense and free plasmid controls, respectively, and 887 and 365 down-regulated genes, with the bulk of these DEGs shared (Fig. 3A and Additional file 7: Table S6). The DEGs that emerged from the comparison with the free plasmid control were studied further.
Several cancer metastasis-related gene ontology (GO) biological process (BP) terms, including inflammatory response, cell–matrix adhesion, neuron migration, and cell adhesion, were discovered by functional analysis of CRLM1 up-regulated genes (Fig. 3B and Additional file 3: Fig. S2A). CRLM1 down-regulated genes were also enriched in cancer-related BP terms such as inflammatory response, hypoxia response, negative regulation of cell migration, and angiogenesis (Fig. 3C and Additional file 3: Fig. S2B). Similar enriched GO BP terms were detected in up-regulated and down-regulated genes compared to CRLM1-anti samples (Additional file 3: Fig. S2C, D). These findings showed a link between CRLM1 and cancer metastasis.
DEGs induced by CRLM1 were also enriched in transcription factors. Transcriptional factors such as HMOX1, FOS, FOXD3, and FOXF1 that have been up-regulated may contribute to the metastatic transcriptome formed by CRLM1-OE (Fig. 3D). Additionally, CRLM1 boosted the expression of classic EMT transcription factors, including Snails, ZEB1 and ZEB1-AS, which positively regulates ZEB1 expression in CRC cells, which should be done contribute significantly to the metastatic transcriptome as well. The EMT marker vimentin expression was greatly elevated, whereas the epithelial marker CDH1 (E-cadherin) expression was not altered. Recently, it was shown that GALNT1 promotes metastasis. Its expression is regulated by the lncRNA SNHG7 [15]. However, it was revealed that CRLM1 drastically lowered GALNT1 expression while increasing GALNT2 and GALNT3 expression (Fig. 3E). To verify the results of RNA-seq, qRT-PCR was performed to detect the expression of above genes in HCT116 cells transfected with CRLM1-OE. The results showed that HMOX1, FOS, FOXD3, FOXF1,SNAI1, SNAI2, SNAI3, ZEB1, ZEB1-AS1, VIM, GALNT2 and GALNT3 mRNA level in CRLM1-OE cells was significantly higher than in control cells, whereas GALNT1 was significantly lower. And there was no significant difference in CDH1 expression between CRLM1-OE cells and control cells (Additional file 3: Fig. S2E, F). These findings were similar to the RNA-seq results.
A co-expression study of 54 RNA-seq datasets revealed that CRLM1 expression was linked with 520 genes, 31 of which overlapped with CRLM1-OE DEGs in CRC cells. The CRLM1-regulated and CRLM1-coexpressed genes were compared to metastasis-related genes (1938) in the human cancer metastasis database (HCMDB) [22], yielding 107 and 70 overlapped genes, respectively (p = 0.002 and 4.48123e-10, hypergeometric test, Fig. 3F). These three datasets shared only five genes, including ADAMTS13, CD14, CXCL2, HMOX1, and HPN (Fig. 3F). Heatmap analysis revealed that these five genes were overexpressed in metastatic samples (Additional file 3: Fig. S2G), while three (ADAMTS13, CD14 and HMOX1) were up-regulated and two (CXCL2, and HPN) were down-regulated in CRC cells by CRLM1-OE (Additional file 3: Fig. S2H). Consistent with the positive correlation between CRLM1 and its neighboring gene PPAP2B in patients, it was demonstrated that CRLM1 boosted PPAP2B expression in vitro (Additional file 3: Fig. S2I), indicating that CRLM1 has a regulatory effect on PPAP2B expression.
CRLM1-chromatin binding is weakly associated with CRLM1-regulated gene expression
Numerous lncRNAs regulate gene expression through their association with specific chromatin sites. To elucidate how CRLM1 regulates gene expression in HCT116 cells, the ChIRP approach was used to identify putative CRLM1–chromatin interaction sites by sequencing the chromatin DNA fragments associated with CRLM1 (ChIRP-seq) (Additional file 4: Fig. S3A). One thousand three hundred and one peaks were obtained after aligning the quality-filtered data to the human genome and calling Model-based Analysis of ChIP-Seq 2 (MACS2) peaks (Fig. 4A and Additional file 7: Table S7). When compared to the genomic length distribution, there were more CRLM1 binding sites (peaks) in the TSS upstream region than in the gene body (excluding the intron) (Fig. 4B). Six hundred sixty-three genes having CRLM1 peaks in the gene body region or the upstream 5 kb region were identified. Notably, CRLM1-associated genes were also considerably enriched in biological processes associated with metastasis, such as cell adhesion and response to DNA damage stimuli (Fig. 4C and Additional file 4: Fig. S3B).
However, only 15 and 7 of these genes overlapped with CRML1 up-and down-regulated genes (Fig. 4D). These findings imply that CRLM1 binding and CRLM1-regulated gene expression are not well correlated. The global expression pattern of genes containing one or more CRLM1-associated chromatin locations was studied using heatmap clustering. Around half of the CRLM1-associated genes exhibited distinct CRLM1-regulated gene expression patterns, with more up-regulated genes than down-regulated genes in CRLM1-OE samples (black frames, Fig. 4E). The activities of these CRLM1-associated genes were also investigated. They were found to be strongly correlated with cancer metastasis-related terms, including adhesion junction, cell cycle, apoptosis, and choline metabolism in cancer (Fig. 4F). Notably, these genes were significantly enriched in mucin-type O-glycan biosynthesis and CRC (Fig. 4F). Changes in O-glycan biosynthesis were previously observed in human CRC tissues [23]. The three CRC genes were SMAD2, MAPK9, and GSK3B. When the expression levels of these genes and three mucin-type O-glycan biosynthesis genes, GALNT5, GALNT13, and GALNT18, were plotted, we discovered that the expression change for all of these CRLM1-binding genes was low or negligible (Fig. 4G). qRT-PCR was performed to verify the mRNA expression of these genes of HCT116 cells transfected with CRLM1-OE. Similar results were obtained from qRT-PCR (Additional file 4: Fig. S3C). Interestingly, the expression patterns of these genes were largely similar in patient metastatic tissues (Fig. 4H) to those in CRLM1-OE CRC cells, implying the functional significance of the CLRM1 binding-associated gene expression changes, albeit minor. It was observed that the link between CRLM1 and chromatin was poor. Except for CRLM1 itself, most of the CRLM1 binding sites had poor ChIRP-seq signals (Additional file 4: Fig. S3D–F).
CRLM1 physically interacts with hnRNPK and promotes its nuclear localization
LncRNAs regulate cells function by interacting with proteins. ChIRP and mass spectrometry (ChIRP-MS) were used to identify proteins that interact with CRLM1. There were 76 proteins discovered in total (Additional file 7: Table S8). By clustering CRLM1-interacted proteins using STRING, it was discovered that CRLM1 interacts preferentially with a large number of histone proteins and a few RNA binding proteins (RBPs), shedding light on the potential processes by which CRLM1 interacts with chromatin (Fig. 5A). The transcriptional regulatory role of RBPs was discovered using a large-scale genome-wide chromatin immunoprecipitation (ChIP) approach. Among the RBPs that interact with CRLM1, heterogeneous nuclear ribonucleoprotein K (hnRNPK) has been investigated for decades as a DNA-binding transactivator that recruits transcription factors that promote cancer development and metastasis. As a result, it was suggested that CRLM1 might interact with hnRNPK to perform its metastasis-promoting functions.
Western blotting was used to confirm the physical interaction between CRLM1 and hnRNPK using protein components recovered from the CRLM1 ChIRP experiment. hnRNPK was found in the CRLM1 ChIRP sample but not in the LacZ negative control sample, and both ChIRP samples were devoid of ACTIN (Fig. 5B). To identify the unique 483 fragment capable of engaging with hnRNPK, we used an enhanced RNA immunoprecipitation (RIP) assay employing antibodies against hnRNPK to capture its associated RNA fragments. PCR amplification of the related CRLM1 RNA resulted in five distinct truncation segments with approximately 80 ~ 100 nt overlap between neighboring fragments (Fig. 5C). As a positive control, a well-characterized lncRNA-P21 was chosen that interacts with hnRNPK [24]. As negative controls, antisense strands were amplified. The results of qRT-PCR indicated that the first fragment (position 1–180) had the strongest binding signal, but the second fragment (position 83–260) had a very weak signal, implying a significant binding site in an 82-nt CRLM1 fragment positioned at the 5′ end (Fig. 5C). Similarly, a study of the binding signal difference between the fifth (position 362–545) and fourth (position 262–440) CRLM1 fragments showed the presence of a second strong binding site for hnRNPK in a 105-nt CRLM1 fragment positioned at the 3′ end of CRLM1 (position 441–545). Previous research established that hnRNPK binds to C-rich motifs and that this binding is connected with lncRNA nuclear enrichment [25, 26]. The C-rich and Alu elements in CRLM1 sequences were subsequently examined, and a putative hnRNPK binding site was identified in a large loop located at the 3′ end (position 483 ~ 490, Fig. 5D), consistent with RIP-qPCR results.
To further investigate the physical interaction between CRLM1 and hnRNPK, fluorescence in situ hybridization (FISH) tests were performed to confirm their cellular location. Under negative control, CRLM1 was both nuclear and cytoplasmic, whereas hnRNPK was more cytoplasmic than the nucleus. However, it was discovered that CRLM1 and hnRNPK were mostly localized in the nucleus with some colocalization under the CRLM1-OE condition. Surprisingly, hnRNPK's cytoplasmic distribution was decreased (Fig. 5E). To verify the results of FISH experiment, western blotting was performed to analyse the nuclear/cytosol fractionation and quantification of CRLM1-dependent hnRNPK in HCT116 and SW620 cells transfected with CRLM1-OE. The results showed that CRLM1-OE cells exhibited significantly higher levels of nuclear hnRNPK protein, compared with the control cells, as well as cytosolic hnRNPK protein level was significantly lower than in control cells (Fig. 5F). This finding was consistent with the FISH results. Taken together, our results indicate that CRLM1 interacts with and supports the nuclear localization of hnRNPK.
CRLM1 strongly promotes hnRNPK occupancy at promoter regions to regulate the expression of genes
Given that CRLM1 may interact with hnRNPK and increase its nuclear localization in trans, the question is how CRLM1 affects the hnRNPK-chromatin relationship (Fig. 6A). To test this idea, chromatin immunoprecipitation and sequencing (ChIP-seq) for hnRNPK was done in HCT116 cells overexpressing hnRNPK under CRLM1-OE or control conditions. The confident hnRNPK ChIP peaks were mostly situated in intergenic regions, significantly more than the genomic fraction of intergenic areas (Fig. 6B, Additional file 7: Tables S9 and S10). Only 34 genes had at least one confident hnRNPK peak within their promoter regions (Fig. 6C), consistent with the low hnRNPK concentration in the nucleus. The CRLM1-OE condition significantly enhanced the hnRNPK peaks within 10 kb upstream of the TSS and in the gene body regions, decreasing intergenic regions. There are now 170 genes that contain at least one confident hnRNPK peak (Fig. 6B, C). One hundred and fifty-one were unique to the CRLM1-OE condition, 25 were duplicated with previously published hnRNPK-bound genes (1177) [27] (Additional file 5: Fig. S4A), and 21 were overlapped with metastasis-related genes (Fig. 6d). Functional enrichment analysis revealed that hnRNPK-bound genes were strongly enriched in apoptotic pathways and processes, as well as gene expression regulation and RNA metabolic processes (Additional file 5: Fig. S4B).
By plotting the peak reads surrounding transcriptional start sites, the effect of CRLM1 on triggering hnRNPK redistribution at transcriptional start sites was further investigated. It is demonstrated that the association between hnRNPK and promoter was poor in the absence of CRLM1 but dramatically increased in CRLM1 overexpression (Fig. 6E). Figure 6f and Additional file 5: Fig. S4C, D illustrate examples of CRLM1-mediated hnRNPK promoter occupancy, consistent with CRLM1-mediated nuclear localization of hnRNPK.
The nuclear localization and promoter binding of CRLM1-promoted hnRNPK are consistent with our hypothesis that CRLM1-hnRNPK regulates gene expression. To demonstrate this point, we overexpressed hnRNPK in HCT116 cells with and without stable CRLM1 overexpression (Additional file 6: Fig. S5A). Transcriptome sequencing (RNA-seq) was used to discover genes controlled by CRLM1 and hnRNPK. Pearson's correlation coefficients (PCCs) analysis revealed that samples containing the majority of hnRNPK-OE samples were well separated from other samples except for the third replicate, and the two outlier samples were excluded during the further study (Additional file 6: Fig. S5B). Functional enrichment analysis revealed that hnRNPK and CRLM1 co-regulated genes were strongly enriched in apoptotic processes, cell adhesion, as well as transcription regulation and metabolic processes (Additional file 6: Fig. S5C–F). Overexpression of either hnRNPK or CRLM1 resulted in significant alterations in gene expression, with more up-regulated genes than down-regulated genes (Fig. 7A, Additional file 7: Tables S11, S12, and S13). Surprisingly, many genes controlled by hnRNPK or CRLM1 overlapped (Fig. 7B), showing that they target the same collection of genes. When hnRNPK was overexpressed in cells expressing CRLM1, its ability to regulate gene expression was significantly reduced, particularly for up-regulated genes (Fig. 7A). This observation could be explained by the fact that genes regulated by hnRNPK are likewise regulated similarly by CRLM1. The hierarchical clustering heatmap revealed that hnRNPK overexpression altered the expression of the majority of hnRNPK-bound genes. Regardless of the degree of CRLM1 expression, hnRNPK overexpression promoted and suppressed the expression of genes in Cluster 1 and Cluster 3, respectively (Fig. 7C). In contrast, hnRNPK and CRLM1 inhibited the expression of genes in Cluster 2, and an additive inhibitory impact was observed for the majority of these genes (Fig. 7C). BBC3 was found to be a Cluster 2 gene (Fig. 7D, E). These findings demonstrate that CRLM1-mediated hnRNPK-chromatin association plays a significant role in gene expression regulation, with a subset of these genes exhibiting transcriptional suppression by both CRLM1 and hnRNPK.
CRLM1-hnRNPK cooperatively regulates the expression of metastasis-related genes and promotes CRC cell metastasis
The occupancy of the CRLM1-hnRNPK promoter by a collection of CRLM1-hnRNPK co-repressed genes represents the direct target of the CRLM1-hnRNPK-chromatin relationship. It was hypothesized that CRLM1-hnRNPK indirectly regulates other genes, which may play a role in metastasis and apoptosis regulation. It was discovered that CRLM1 and hnRNPK regulate the same genes cooperatively in the same direction. These genes were up-or down-regulated (52 and 49 genes, respectively) by both CRLM1 and hnRNPK. Double overexpression accentuated the up-or-down-regulation (Fig. 8A, B, Additional file 7: Tables S14 and S15). These CRLM1-hnRNPK co-regulated genes were found to enhance functional pathways such as the TGF- signaling pathway, a well-known mechanism directing EMT (Fig. 8C), and other cancer development activities such as negative regulation of cell proliferation and apoptosis (Fig. 8D). Three genes involved in TGF-β signaling pathways regulated by CRLM1-hnRNPK were down-regulated SMAD6 and BMP7, and up-regulated SMURF2. Additionally, when hnRNPK was knocked down, the stimulatory effect of CRLM1 on SMURF2 expression and the inhibitory effect on BBC3, SMAD6, and BMP7 expression was knocked down (Fig. 8E). The change in mRNA levels of these genes correlated with the change in protein levels (Fig. 8F). To determine if CRLM1 functions biologically via hnRNPK, rescue tests in CRC cells were done following co-transfection of si-hnRNPK or hnRNPK-OE with CRLM1-OE or si-CRLM1. The results indicated that knockdown hnRNPK could significantly reverse the pro-migratory and pro-invasion roles of CRLM1 up-regulation in CRC cells, whereas hnRNPK overexpression attenuated the inhibitory effects of CRLM1 down-regulation in CRC cells using wound healing and transwell tests (Fig. 8G–N). Additionally, in vivo rescue experiments were conducted to confirm that si-hnRNPK reversed the tumor metastasis promoter effect of CRLM1 in CRC cells. We found that knockdown of hnRNPK could effectively suppress liver metastasis of CRLM1-OE CRC cells in the liver metastasis model (Fig. 8O, P). These findings indicate that CRLM1 and hnRNPK work in concert to induce CRC cell metastasis.