MIR99AHG is upregulated in metastatic CRC cells and tissues
Highly metastatic KM12SM cells were isolated from spontaneous metastases of the liver in nude mice after multiple cycles of intracecal orthotopic injection of the parental KM12C cells . We confirmed the enhancement of cell migration, invasion and metastasis in KM12SM compared to KM12C cells (Figures S1A-D), but no difference was observed in cell proliferation or tumor growth (Figures S1E-J). To identify non-coding RNAs involved in CRC metastasis, RNA-seq and small RNA-seq were performed on KM12C and KM12SM cells. A total of 410 lncRNAs and 46 miRNAs were found to be differentially expressed between the two cell lines with a stringent filtering criterion (log2 fold change ≥ 4, p < 0.01; Fig. 1A). Notably, the most upregulated lncRNA in KM12SM cells was MIR99AHG, and the three most upregulated microRNAs (miRNAs) were miR-99a, let-7c and miR-125b-2, which are derived from a common host gene named MIR99AHG (Fig. 1B). We validated upregulation of MIR99AHG, as well as the precursor and mature transcripts of miR-99a, miR-125b and let-7c, in KM12SM cells (Figs. 1C, 1D and S2A). Since the roles of these miRNAs in CRC have been determined previously [26–29], we focused our investigation on the lncRNA MIR99AHG.
MIR99AHG was upregulated in most CRC cell lines compared to the normal colon mucosa cell line NCM460 (Fig. 1E). To investigate the clinical relevance of MIR99AHG, its expression level was analyzed by in situ hybridization (ISH) in tissue microarrays (TMAs), which contained 48 pairs of primary CRC tissues with matched adjacent normal tissues and lymphatic or distant metastatic tissues. Enhanced MIR99AHG expression was observed in 40 (83.3%) cases of metastatic tissues, compared to the matched adjacent normal tissue and primary CRC tissue (Fig. 1F). Statistically, the MIR99AHG expression level was positively correlated with metastasis and AJCC stage in the TMA cohorts (Fig. 1G). Further analysis using the TCGA data repository revealed that a higher MIR99AHG level was observed in advanced CRC samples (Fig. 1H) and was correlated with poor overall and disease-free survival in CRC patients (Fig. 1I). Similar expression patterns were observed in bladder cancer and mesothelioma (Figure S2B). These results indicate that MIR99AHG is pathologically and clinically associated with CRC metastasis and patient outcome.
MIR99AHG promotes migration, invasion and metastasis in CRC cells
To determine the role of MIR99AHG in CRC metastasis, we silenced MIR99AHG by transfecting antisense oligonucleotides (ASOs) into KM12SM and HuTu80 cells (Figure S3A). Inhibition of MIR99AHG significantly attenuated cell migration and invasion (Figs. 2A and S3B). In contrast, overexpression of MIR99AHG by transduction of a full-length MIR99AHG (MIR99AHG-FL) construct into KM12C and RKO cells (low endogenous MIR99AHG expression), as well as in DLD-1 cells (moderate endogenous MIR99AHG expression), promoted cell migration and invasion (Figs. 2B, S3C and S3D). We further generated MIR99AHG knockout (MIR99AHGKO) KM12SM cells using CRISPR/Cas9 gene-editing, which specifically deleted the longest exon, exon 8 (3,784 bp), of MIR99AHG (4,478 bp) without altering the expression of its neighboring miRNAs (Figures S3E and S3F). The MIR99AHGKO KM12SM cells exhibited impaired migration and invasion (Figure S3G), whereas reconstitution of MIR99AHG-FL, rather than its antisense counterpart, restored migration and invasion (Figs. 2C and S3H). Furthermore, tail vein injections of wild-type (WT) and MIR99AHGKO KM12SM cells into nude mice revealed that MIR99AHG knockout decreased the incidence of lung metastasis and the number of lung metastatic nodules in mice when compared to WT (Figure S3I). We established clones stably expressing MIR99AHG in KM12C and DLD-1 cells by lentiviral transduction, with overexpression of MIR99AHG remarkably increasing the metastatic potential of both cell lines (Fig. 2D). Notably, manipulation of MIR99AHG had no influence on cell proliferation or tumor growth (Figures S4A-E). These results indicate that MIR99AHG promotes migration, invasion and metastasis in CRC cells.
MIR99AHG regulates invadopodia formation in CRC cells
Gene Ontology (GO) and Gene Set Enrichment Analysis (GSEA) of RNA-seq data from KM12C and KM12SM cells revealed that a majority of the enriched differentially expressed genes (DEGs) were for cell membrane categories like “cell periphery”, “cell projection”, “anchoring junction” and “focal adhesion” (Figs. 2E and 2F), and for biological processes such as “locomotion”, “cell motility” and “cell migration” (Figure S5A). Reactome pathway analysis showed that DEGs involved in “ECM organization”, “degradation of the ECM” and “non-integrin membrane-ECM interactions” were also enriched (Figs. 2F and S5B). Consistent with these analyses, we observed that KM12C and KM12SM cells exhibited different morphological features in three-dimensional (3D) culture, with KM12C cells forming solid organized colonies and KM12SM cells forming disorganized masses with protrusions (Fig. 2G). Dense protrusions of KM12SM could be also observed at a single cell resolution (Fig. 2G). We then used electron microscopy to visualize these protrusions. Consistent with the 3D-culture studies, KM12C and KM12SM cells exhibited distinct morphological characteristics in term of membrane protrusions: there were few scattered finger-like projections at the edge of KM12C cells, but many podosome-like protrusion structures were observed in KM12SM cells (Fig. 2H). We speculated that the protrusions in KM12SM cells were invadopodia, which are present on invasive cancer cells and can degrade the ECM . In line with this, Rho GTPases, a family of proteins that modulates invadopodia dynamics , were enriched in KM12SM cells (Figure S5C). The expression of invadopodia components TKS5, N-WASP, MMP7 and MMP14 was remarkably increased, and the expression of tissue inhibitor of metalloproteinases 1 and 2 (TIMP1 and TIMP2) was decreased in KM12SM compared to KM12C cells (Fig. 2I). Although the total level of cortactin remained unchanged, phosphorylated cortactin at tyrosine 421 (p-cortactin) was increased (Fig. 2I), indicating hyperactivity of actin polymerization . Silencing of TKS5 or cortactin attenuated cell migration and invasion of KM12SM cells (Figures S5D and S5E), suggesting that invadopodia were related to the enhanced migration and invasion phenotypes. Additionally, we did not observe changes in markers of epithelial-to-mesenchymal transition between KM12C and KM12SM cells (Figure S5F).
We next investigated whether MIR99AHG is involved in invadopodia formation of CRC cells. Knockout of MIR99AHG not only diminished the density and shape of protrusive structures (Fig. 2J), but also reduced the expression of TKS5, p-cortactin and MMP14 in KM12SM cells (Fig. 2K). Overexpression of MIR99AHG exerted opposite effects in KM12C and DLD-1 cells (Figs. 2J and 2K). Notably, knockdown of TKS5 or cortactin largely attenuated cell migration and invasion mediated by MIR99AHG overexpression (Figs. 2L, S5G and S5H), indicating that MIR99AHG promotes migration and invasion in CRC cells via invadopodia formation.
MIR99AHG functions via its interaction with PTBP1 in CRC cells
We next investigated the mechanism by which MIR99AHG promotes CRC metastasis. We first determined which region of MIR99AHG is responsible for increasing metastatic capabilities. To this end, a series of truncated MIR99AHG fragments were generated (Fig. 3A), and the 1,261-nt sequence (termed MIR99AHG-F5 hereafter) near the 3’ end of MIR99AHG showed the closest equivalent to the effect of MIR99AHG-FL on promoting migration and invasion in KM12C, DLD-1 and RKO cells (Figs. 3B and S6A). A substantial fraction of lncRNAs exert functions by interacting with RNA binding proteins , hence we used biotin-labeled MIR99AHG-F5 as an RNA probe and conducted an RNA pulldown assay followed by mass spectrometric analysis to identify protein partners of MIR99AHG (Fig. 3C). Among the identified proteins, PTBP1 and U2AF2 (also known as U2AF65) are factors in regulating AS of pre-mRNAs , and their cellular distribution was consistent with the nuclear localization of MIR99AHG (Fig. 1D). Despite PTBP1 and U2AF2 being functionally related , there was no interaction between them (Figure S6B). Given that PTBP1 and U2AF2 levels were positively correlated with CRC progression [33, 34], we validated that MIR99AHG interacted with PTBP1 and U2AF2 individually (Figs. 3D and 3E). We next determined whether PTBP1 and/or U2AF2 mediated the pro-metastatic effects of MIR99AHG. Either PTBP1 or U2AF2 knockdown impaired cell migration and invasion in KM12SM and HuTu80 cells (Figures S6C and S6D). However, only knockdown of PTBP1, but not U2AF2, abolished the enhanced migration and invasion in MIR99AHG-overexpressing KM12C and DLD-1 cells (Figs. 3F, S7A and S7B), indicating that MIR99AHG promotes CRC migration and invasion mainly via PTBP1. Moreover, knockdown of PTBP1 reduced the levels of TKS5, p-cortactin and MMP14 in KM12SM cells (Fig. 3G), suggesting the functional involvement of PTBP1 in invadopodia formation.
PTBP1 contains four RBDs (RBD1-4), with the RBD3 and RBD4 exhibiting an interdomain interaction . To determine which RBD is required for interaction with MIR99AHG, full-length and truncated constructs of PTBP1 with a Flag tag were transduced together with MIR99AHG-F5 (Fig. 3H). RNA immunoprecipitation (RIP) using the antibody against Flag showed that RBD3 and RBD4 of PTBP1 were required for the association with MIR99AHG (Fig. 3H). To determine the region of MIR99AHG that bound PTBP1, nuclear extracts from KM12SM cells were subjected to a limited RNase T1 digestion, so that the RNA fragments protected by a bound protein would remain preferentially uncleaved. After RIP with the antibody against PTBP1, the enriched RNA fragments were identified by RT-qPCR analysis using primer sets that scanned the MIR99AHG transcript in overlapping ~ 150-nt-long segments. As shown in Fig. 3I, the 5’ end region of MIR99AHG-F5 was enriched by the RIP-based mapping assay, suggesting that the MIR99AHG transcript from nt 3,136 to 3,946 is the major region responsible for PTBP1 interaction.
Next, we explored the consequence of the interaction between MIR99AHG and PTBP1. Knockdown of MIR99AHG reduced PTBP1 expression in KM12SM and HuTu80 cells, while overexpression of MIR99AHG increased PTBP1 expression in KM12C and DLD-1 cells (Fig. 3J). Neither MIR99AHG knockdown nor overexpression affected PTBP1 mRNA levels (Figure S7C). We therefore speculated that MIR99AHG influences the stability of PTBP1. A higher portion of endogenous PTBP1 was retained in MIR99AHG-overexpressing cells than in control cells when treated with cycloheximide (CHX); when these cells were treated with MG132, a specific proteasome inhibitor, accumulated PTBP1 in MIR99AHG-overexpressing cells was comparable to that of the control cells (Fig. 3K). These results indicate that MIR99AHG contributes to PTBP1 stability in CRC cells.
MIR99AHG and PTBP1 regulate the AS of SMARCA1 pre-mRNA
Since PTBP1 mainly participates in the regulation of AS , we asked whether MIR99AHG influences PTBP1-regulated AS events. To this end, RNA-seq was employed to assess common AS events regulated by PTBP1 and MIR99AHG. In total, 1024 and 998 AS events were significantly altered (∣percent spliced in (PSI)∣ ≥ 0.3, p < 0.01) in KM12SM cells after PTBP1 and MIR99AHG knockdown, respectively (Figs. 4A and 4B). Among these AS events, we focused on cassette exon splicing (also known as exon skipping) because it is the most represented category and is known to be regulated by PTBP1 [6, 13]. MIR99AHG and PTBP1 regulated 20 common cassette exon events (12 skipping and 8 inclusion) in KM12SM cells (Fig. 4C; Table S1). We noticed that one of the repressive exons after MIR99AHG or PTBP1 knockdown is exon 13 of SMARCA1 (Fig. 4D). Since SMARCA1 encodes a core subunit of the ISWI chromatin remodelers  and depletion of SMARCA1 in HeLa cells increases cell migration and alters expression of genes related to cell locomotion , we selected SMARCA1 for further investigation.
The inclusion of exon 13 leads to a 12-aa insertion within the conserved catalytic domain of SMARCA1 (schematically depicted in Fig. 4E), which generates a long isoform (termed SMARCA1-L hereafter) that has no ATP-dependent chromatin remodeling activity . We confirmed that isoform switching occurred in KM12SM cells, which express higher levels of SMARCA1-L but lower levels of canonical SMARCA1 (termed SMARCA1-S hereafter) than in KM12C cells (Figs. 4E and S8A). This tendency was observed in a panel of CRC cell lines with increased endogenous MIR99AHG levels (Fig. 4F). Knockdown of PTBP1 inhibited inclusion of exon 13 in SMARCA1 pre-mRNA, which decreased SMARCA1-L and increased SMARCA1-S in KM12SM, HuTu80 and HCT116 cells (Figs. 4G and S8B). Similar results were observed after MIR99AHG knockdown (Figs. 4G, S8C and S8D). Conversely, overexpression of MIR99AHG-FL, but not its anti-sense counterpart, increased SMARCA1-L and decreased SMARCA1-S in KM12C, DLD-1 and HCT15 cells, as well as in MIR99AHGKO KM12SM cells (Figs. 4H, 4I and S8E). These results indicate that PTBP1 and MIR99AHG regulate the AS of SMARCA1 exon 13. However, we did not observe substantial changes in the level of SMARCA1 isoforms after PTBP1 overexpression (Figs. 4H, 4I and S8F). These results indicate that the effect of MIR99AHG on SMARCA1 splicing is independent of its role on stabilizing PTBP1 protein, suggesting a regulatory role for MIR99AHG in the AS process.
MIR99AHG modulates the binding position of PTBP1 on SMARCA1 pre-mRNA
We next sought to illustrate the mechanism by which MIR99AHG and PTBP1 regulates the AS of SMRACA1 exon 13. We first validated PTBP1 binding to the transcript fragment flanking SMARCA1 exon 13 (Fig. 5A). RIP assays further showed that only truncations containing RBD2 of PTBP1 maintained association with SMARCA1 pre-mRNA (Fig. 5B). These results, together with the data shown in Fig. 3H, indicate that individual PTBP1 RBDs bind to MIR99AHG or SMARCA1 pre-mRNA with different binding specificities. We then examined whether MIR99AHG interacts with SMARCA1 pre-mRNA. RNA antisense purification (RAP) assay using a biotin-labeled probe against the transcript fragment flanking SMARCA1 exon 13 showed that MIR99AHG strongly bound to SMARCA1 pre-mRNA, as compared to positive control U2 snRNA (Fig. 5C). Using chromogenic RNAscope ISH and IF staining, the colocalization of MIR99AHG, SMARCA1 pre-mRNA and PTBP1 protein was observed in the nucleus of KM12SM and HuTu80 cells (Figs. 5D and S9A). In addition, chromatin isolation by RNA purification (ChIRP) assays using biotin-labeled probes against MIR99AHG revealed an interaction of MIR99AHG with the SMARCA1 locus (Figure S9B). These results, together with the evidence that MIR99AHG interacts with PTBP1 (Figs. 3D, 3E, 3H and 3I), indicate a molecular basis underlying the mode of MIR99AHG/PTBP1-mediated AS regulation of SMARCA1 exon 13.
Next, we investigated how the AS of SMARCA1 exon 13 is regulated by PTBP1 and MIR99AHG. We mapped multiple potential PTBP1 binding sites across the flanking intronic sequences of SMARCA1 exon 13 using the RBPmap web server  according to the consensus sequences for PTBP1 (Fig. 5E). Further analysis of a crosslinking-immunoprecipitation sequencing (CLIP-seq) dataset of PTBP1 in HeLa cells  revealed that PTBP1 primarily binds near the 3’ splice site of SMARCA1 intron 12 (Fig. 5E). According to a position-specific effect proposed for PTBP1 , PTBP1 should promote exon skipping when it binds near the alternative exon but facilitate exon inclusion when the binding occurs near flanking constitutive exons. Thus, the CLIP-seq dataset suggests a suppressive role of PTBP1 on SMARCA1 exon 13 inclusion in HeLa cells, which appears to be the opposite of our observations in which PTBP1 increased SMARCA1 exon 13 inclusion in CRC cells (Fig. 4G). HeLa and KM12SM cells express PTBP1 at a comparable level, but MIR99AHG is expressed at an extremely low level in HeLa cells (Figure S9C), prompting us to speculate that MIR99AHG might be an important factor for determining regions within SMARCA1 pre-mRNA that are recognized and bound by PTBP1. To test this hypothesis, a RIP-based mapping assay was used to determine PTBP1-binding regions within SMARCA1 pre-mRNA in CRC cells. A total of 27 primer sets that scanned SMARCA1 exon 13 and its flanking introns were designed according to the potential PTBP1 binding sites (Fig. 5E). Surprisingly, the transcript segment near the 5’ splice site of SMARCA1 intron 12 (detected by the primer set #2) was strongly retrieved by the anti-PTBP1 antibody in KM12SM and HuTu80 cells, which contrasted with PTBP1 primarily binding near the 3’ splice site of SMARCA1 intron 12 in HeLa cells (Fig. 5F). The interaction between PTBP1 and this 5’ splice site segment was resistant to RNase T1 treatment (Figure S9D), indicating strong PTBP1 binding. Notably, knockout of MIR99AHG in KM12SM cells significantly weakened PTBP1 binding at the 5’ splice site but had a minimal effect on binding at the 3’ splice site (Fig. 5F). A similar reduction in enrichment was observed in KM12C compared to KM12SM cells (Fig. 5F). In addition, PTBP1 enrichment around the branch point sequence of SMARCA1 intron 12 (detected by the primer set #15) was also reduced after MIR99AHG knockout (Fig. 5F), suggesting the involvement of MIR99AHG in a multistep reaction for the AS of SMARCA1 exon 13. Moreover, we constructed a minigene which contains SMARCA1 exon 13 and its flanking introns and constitutive exons (Fig. 5G). Overexpression of MIR99AHG promoted SMARCA1 exon 13 inclusion (Fig. 5H). Deletion of the 5’ splice site largely diminished the effect of MIR99AHG on exon 13 inclusion, and deletion of the site near the branch point augmented this effect (Figs. 5H and S9E). These results indicate that PTBP1 promoted inclusion of SMARCA1 exon 13 in the presence of MIR99AHG where MIR99AHG acts as an address label for PTBP1 to locate its binding positions in the process of AS regulation.
SMARCA1-L loses the suppressive effects on cell migration, invasion and invadopodia formation
To evaluate whether SMARCA1 isoform switching affects CRC metastasis, we altered the expression of SMARCA1-L and SMARCA1-S using isoform-specific siRNAs in KM12C, DLD-1 and HCT15 cells (Figs. 6A and 6B). Knockdown of SMARCA1-S enhanced cell migration and invasion (Figs. 6A and S10A) and increased the levels of TKS5, p-cortactin and MMP14 (Fig. 6C), mimicking depletion of MIR99AHG or PTBP1 (Figs. 2K and 3G). Knockdown of SMARCA1-L had no substantial effect on cell migration and invasion or the expression of invadopodia-related proteins (Figs. 6B, 6C and S10B). Electron microscopy further showed that SMARCA1-S knockdown induced the formation of membrane protrusions (Fig. 6D). We then ectopically expressed either SMARCA1-L or SMARCA1-S vectors in KM12SM and HuTu80 cells. Overexpression of SMARCA1-S, but not SMARCA1-L, inhibited cell migration and invasion as well as reduced TKS5, p-cortactin and MMP14 levels (Figs. 6E, 6F and S10C). Moreover, we knockout SMARCA1 by CRISPR/Cas9 genome editing in DLD-1 cells (Figure S10D). Enhanced cell migration and invasion were observed in SMARCA1KO DLD-1 cells and this enhancement was attenuated after reconstitution of SMARCA1-S but not SMARCA1-L expression (Figs. 6G and S10E). Furthermore, overexpression of SMARCA1-S not only abolished the elevated migration and invasion in MIR99AHG-overexpressing KM12C and DLD-1 cells (Figs. 6H and S11A) but also attenuated the promotive effect of MIR99AHG on TKS5, p-cortactin and MMP14 levels (Fig. 6I), suggesting that loss of SMARCA1-S mediates the pro-metastatic roles of MIR99AHG. Additionally, manipulating SMARCA1-S expression showed minimal effects on proliferation or apoptosis in these CRC cells (Figures S11B-E). Since SMARCA1 is a subunit of the ISWI chromatin remodeling complexes with vital functions in gene expression , we performed RNA-seq to examine whether the phenotypes that resulted from depletion of SMARCA1-S were reflected at the level of gene transcription. A gene panel related to cell migration, invasion and invadopodia formation was found to be elevated upon SMARCA1-S knockdown in KM12C cells (Fig. 6J). Opposite changes for these genes were observed in KM12SM cells after knockdown of MIR99AHG or PTBP1 (Fig. 6J), which was shown to increase SMARCA1-S levels (Figs. 4G, S8B and S8D). These results indicate that SMARCA1-S mediates suppressive effects on cell migration, invasion and invadopodia formation and suggest that the isoform switching to SMARCA1-L promotes CRC metastasis.
Validation of MIR99AHG, PTBP1 and SMARCA1-L expression in CRC specimens
We finally examined the relationship between MIR99AHG, PTBP1 and SMARCA1-L in clinical CRC samples. We analyzed SMARCA1-L expression using chromogenic RNAscope ISH in TMAs containing 48 pairs of primary CRC tissues and their adjacent normal tissues and matched lymphatic or distant metastases. Consistent with the expression pattern of MIR99AHG, SMARCA1-L was significantly increased in metastatic tissues compared to matched primary tumor and adjacent normal tissue (Figs. 7A, 7B and S12A). Furthermore, multiplex IHC and multispectral imaging was performed and revealed substantial increases in PTBP1, TKS5, p-cortactin and MMP14 in metastatic tissues compared to matched primary tumor and normal tissue specimens (Figs. 7A, 7B and S12A). Notably, positive correlations between MIR99AHG and SMARCA1-L or PTBP1 were observed (Fig. 7C), and expression levels of MIR99AHG, SMARCA1-L and PTBP1 were highly correlated with the expression of invadopodia-related proteins (Figs. 7D, S12B and S12C). Collectively, these results support our pre-clinical findings and indicate that the MIR99AHG/PTBP1-mediated inclusion of SMARCA1 exon 13 occurs in the setting of patients with metastatic CRC.