LncRNA-like MMP14 RNA facilitates colorectal cancer metastasis by suppressing H3K27cr at the STARD13 promoter region.

Introduction
Colorectal cancer (CRC) has emerged as a major contributor to cancer-related morbidity and mortality worldwide [1]. While early-stage CRC can be effectively treated with surgery and adjuvant therapy, advanced metastatic stages remain clinically intractable [2]. The 5-year survival rate of metastatic CRC is less than 15% [3]. Cancer metastasis is a complex, multi-step process that involves tumor cell detachment, vascular intravasation, immune evasion, parenchymal extravasation, and colonization at distant organs [4]. Although mechanisms such as epithelial–mesenchymal transition [5], tumor cell plasticity [6], and extracellular matrix proteolysis [7] have been linked to CRC metastasis, the key molecular drivers of metastasis are still poorly understood, hindering the development of effective therapeutic strategies.
Matrix metalloproteinase 14 (MMP14) is recognized as a pro-invasive and pro-tumorigenic gene [8, 9]. MMP14 protein contributes to cancer cell invasiveness through multiple mechanisms, including extracellular matrix degradation, invadopodia formation, protease shedding, and integrin- and metastasis-related gene expression regulation [10, 11]. Additionally, MMP14 protein affects the tumor microenvironment by inhibiting cytotoxic T-cell infiltration, promoting connective tissue proliferation, and facilitating tumor vascularization, all of which contribute to cancer progression [12, 13]. Given its critical role in tumor development, MMP14 protein has emerged as a promising target for therapeutic intervention. However, no pharmacological agents specifically targeting MMP14 protein have been successfully implemented in clinical oncology so far [14]. Evidences suggest that even small fragments of MMP14 RNA, such as those resulting from cytoplasmic tail deletion [15], the E240A active site mutation, and mutations at furin cleavage sites [16], can facilitate cancer metastasis. This raises a new question of whether MMP14 RNA contributes to tumor progression.
There are currently no reports demonstrating that MMP14 RNA promotes tumor progression independently of its protein. The present study uncovered that MMP14 RNA played a crucial role in facilitating CRC metastasis. Mechanistically, MMP14 RNA recruited SIRT3 to the STARD13 promoter, leading to the removal of H3K27cr, which repressed STARD13 expression and ultimately enhanced CRC metastasis. Moreover, silencing MMP14 RNA had a more significant inhibitory effect on tumor metastasis compared with inhibiting MMP14 protein. These discoveries suggest that MMP14 RNA functions as a crucial regulator of CRC metastasis, independent of its protein role.

Methods

Cell culture
Human CRC cell lines, including DLD-1 (TCHu 134), HCT116 (TCHu 99), LoVo (TCHu 82), SW480 (SCSP-5033), and SW620 (TCHu 101), were purchased from the Cell Bank of the Chinese Academy of Science (Shanghai, China). In addition, MMP14 knockout (MMP14KO) cells were purchased from Guangzhou Ubigene Biosciences Co., Ltd. (Guangzhou, China). The cells were cultured in Dulbecco’s modified Eagle medium (DMEM) (Bio-Channel, cat. no.: BC-M-005), supplemented with 1% antibiotics (Vicmed, cat. no.: VC2003) and 10% fetal bovine serum (FBS) (ExCell Bio, cat. no.: FSP500). All cultures were maintained in a humidified incubator at 37 ℃ with 5% CO2.

Transfection and lentivirus transduction
Plasmids and small interfering RNAs (siRNAs) were purchased from GenePharma (Shanghai, China), Hebio (Shanghai, China), GENEWIZ, Inc. (Suzhou, China), and YouBio (Hunan, China). The MMP14-ATGmut plasmid was designed to simultaneously mutate all 11 putative open reading frame (ORF) initiation codons within the MMP14 mRNA sequence, thereby restricting expression to MMP14 RNA without allowing translation. Transient transfection of plasmids and siRNAs was conducted using the liposomal transfection reagent (YEASEN, cat. no.: 40802ES08) and Lipofectamine® 2000 (Invitrogen, cat. no.: 1168019), respectively, in accordance with the manufacturer’s protocols. The lentivirus containing the MMP14-ATGmut plasmid was purchased from OBiO Technology Co., Ltd. (Shanghai, China). HCT116 cells were infected with the lentivirus to establish a stable MMP14 RNA overexpression cell line, and successfully transfected cells were selected using puromycin (Beyotime, cat. no.: ST551-10). The sequences of the siRNAs are detailed in Table S1.

Single-cell RNA sequencing (scRNA-seq) analysis
The Seurat package was employed to analyze single-cell transcriptomes. The functions FindIntegrationAnchors and IntegrateData were utilized to eliminate batch effects. The FindAllMarkers function was utilized to identify markers for each cell type, with cell types determined using the CellMarker 2.0 database. Gene set enrichment analysis (GSEA) of specific cell types was conducted using the fgsea package. The R code for GSE221575, GSE110009, and GSE225857 is available in the supplementary material of our previous study [17]. In addition, the R code for GSE178318 is provided in the Supplementary Material.

Transwell assay
Approximately 50 μl of diluted Matrigel (Corning, cat. no.: 356234) was pre-plated in each well of the chamber to determine the invasion levels, while Matrigel was omitted for migration level evaluation. HCT116 cells (6 × 104 cells/well) and LoVo cells (1 × 105 cells/well) underwent a 12 h starvation period in FBS-free DMEM before being seeded in the upper chamber (Corning, cat. no.: 3422) of the Transwell plate. DMEM supplemented with 20% FBS was added to the lower chamber. After a 24 h incubation, invading cells were fixed with 4% paraformaldehyde for 30 min and subsequently stained with crystal violet for an additional 30 min. Cell observations were conducted using an Olympus microscope, and data were analyzed using ImageJ software.

Western blot analysis
Total cellular proteins were extracted using RIPA lysis buffer (Beyotime, cat. no.: P0013B). Protein concentrations were determined using a BCA kit (Abbkine, cat. no.: KTD3001). The proteins were then separated by SDS-PAGE and transferred onto a nitrocellulose membrane. The membrane was blocked with 5% skim milk for 2 h at room temperature, followed by overnight incubation with primary antibodies at 4 ℃. Afterward, the membrane was incubated with secondary antibodies at room temperature for 2 h. Finally, protein bands were visualized with an ECL luminescent solution (Abbkine, BMU102-CN). Western blot analysis was performed using the following antibodies: anti-E-cadherin (E-Cad) (ABclonal, A20798, 1:2000), anti-N-cadherin (N-Cad) (Abclonal, A3045, 1:1000), anti-Vimentin (Proteintech, 10366-1-AP, 1:2000), anti-GAPDH (Proteintech, 60004-1-Ig, 1:10,000), anti-Snail1 (Proteintech, 13099-1-AP, 1:1000), anti-SIRT3 (Santa Cruz, sc-365,175, 1:200), anti-H3K27cr (PTM BIO, PTM-545RM, 1: 1000), anti-MMP14 (Affinit, AF0212,1:1000), anti-STARD13 (Proteintech, 21325-1-AP,1:1000), and anti-H3 (Proteintech, 17168-1-AP, 1:10,000).

Reverse transcription-polymerase chain reaction (RT-PCR) assay
Total DNA was extracted utilizing the DNA extraction kit (TIANGEN, cat. no.: DP304-02). The RT-PCR assay was conducted using the 2× PCR Master Mix (Beyotime, cat. no.: D7228). The PCR products were then separated using 1% agarose gel (Servicebio. GC205013-100 g). The gel containing the PCR products was excised for sequencing.

Quantitative real-time PCR assay
Total RNA was isolated from tissues or cells using TRIzol reagent (Vazyme, cat. no.: R401-01-AA). The extracted RNA was reverse-transcribed into cDNA using the SweScript RT I First Strand cDNA Synthesis Kit (Servicebio, cat. no.: G3330-50). qRT-PCR was performed with the 2× SYBR Green qPCR Master Mix (High ROX) (Servicebio, cat. no.: G3322-05). Gene expression levels were normalized to GAPDH and calculated using the 2−ΔΔCt method. The sequences of primers used are provided in Table S2.

Cell fractionation assay
CRC cells were separated using the nuclear and cytoplasmic extraction kit (Beyotime, cat. no.: P0027) following the manufacturer’s instructions. Briefly, cells were lysed on ice for 15 min with pre-cooled Reagents A and B. The cell lysate was then centrifuged at 14,000 rpm for 10 min at 4 ℃, and the supernatant was collected for cytoplasmic RNA extraction using TRIzol reagent. The sediment was then resuspended in reagents A and B and incubated on ice for 45 min. Subsequently, the sediment was washed thrice with cold phosphate-buffered saline (PBS) before the extraction of nuclear RNA using a TRIzol reagent.

Luciferase activity assay
A pGL3-enhancer plasmid containing 2 kb fragment of the STARD13 promoter was obtained from GENEWIZ, Inc. (Suzhou, China). CRC cells were co-transfected with this plasmid along with either MMP14-ATGmut or siMMP14 for 48 h. The firefly luciferase activity was determined using the Mono-Lumi firefly luciferase reporter gene assay kit (Servicebio, cat. no.: G1702). The fluorescence intensity was determined using a microplate reader.

RNA fluorescent in situ hybridization (RNA-FISH) and immunofluorescence microscopy
A cyanine 5-labeled MMP14 RNA probe was procured from Integrated Biotech Solutions (Shanghai, China). Cells were seeded onto coverslips and incubated overnight. Subsequently, the cells were washed thrice with cold PBS and then permeabilized on ice for 5 min using CSK buffer [100 mM NaCl, 300 mM sucrose, 3 mM MgCl2, 10 mM PIPES (pH 6.8)]. Following permeabilization, the cells were fixed on ice with 4% paraformaldehyde for 10 min. The cells were then dehydrated sequentially in ethanol at concentrations of 70%, 80%, 95%, and 100%, with each step lasting 3 min.
The MMP14 probe was diluted using 2× hybridization buffer [4× SSC, 40% (w/v) dextran sulfate, 2 mg/ml bovine serum albumin (BSA), 400 mM RVC], and incubated in a humid chamber at 37 °C overnight in the dark. The following day, the coverslips were washed thrice with freshly prepared 50% formamide and 2× SSC buffer at 42 °C for 5 min each. Subsequently, cells were blocked with 3% BSA for 45 min and then incubated with an anti-SIRT3 antibody (Bioss, Bs-6105R, 1:25) at room temperature for 1.5 h. The secondary antibody (Bioss, bs-0295D-AF488, 1:50) was then incubated at room temperature for 30 min. Afterward, the cells were stained with DAPI (Servicebio, cat. no.: G1012) at room temperature for 10 min. Finally, the coverslips were sealed onto the slide with anti-fluorescence quenching (Beyotime, cat. no.: P0126).

Chromatin isolation by RNA purification (ChIRP) assay
Biotin-labeled MMP14 probes were designed and synthesized by RiboBio Co., Ltd. (Guangzhou, China), with the LacZ probe serving as a negative control. The ChIRP protocol was applied as previously described [18]. Briefly, cells were cross-linked with 1% glutaraldehyde, after which the cross-linking reaction was terminated with 1.25 M glycine solution. The cells were then sonicated with a 30 s pulse and 30 s pause for a total of 20 cycles. Next, the lysate was centrifuged, and the supernatant was collected. Subsequently, hybridization buffer was added to the supernatant in a volume equivalent to twice that of the supernatant, along with the probes. Subsequently, 120 μl of BeyoMag™ streptavidin magnetic beads (Beyotime, cat. no.: P2151) was added to the mixture and incubated at 37 °C for 5 h with shaking. The beads were subsequently washed five times and resuspended using lysis buffer. The samples were divided into two portions: 100 μl for RNA extraction and 900 μl for DNA extraction. RNA and DNA were isolated using phenol–chloroform–isoamyl alcohol extraction, and their enrichment was quantified using the qRT-PCR assay.

Chromatin immunoprecipitation (ChIP-seq and ChIP-qPCR) assay
Chromatin immunoprecipitation was performed as previously described [19]. In brief, cells were cross-linked with 1% formaldehyde, and the cross-linking reaction was subsequently terminated with 1.25 M glycine solution. The cells were then harvested and incubated in 200 μl of ChIP lysis buffer. The samples were then sonicated with a 10 s pulse and 20 s pause for a total of ten cycles. After sonication, the lysates were centrifuged, and the supernatant was collected. To this supernatant, 1 μg of primary antibody was added and incubated at 4 °C for 2 h. Following this, 25 μl of protein A/G magnetic beads (MCE, cat. no.: HY-K0202) were added and incubated at 4 °C for an additional 2 h. The beads were washed three times with TE buffer. Finally, DNA was collected, and its enrichment was assessed using the qRT-PCR assay or sequenced by Shanghai Orizymes Biotech. Co., Ltd. (Shanghai, China).

RNA immunoprecipitation (RIP) assay
The RIP assay was conducted as previously described [17]. In brief, cells were harvested using polysome lysis buffer. After cell lysis, the lysate was centrifuged, and approximately 2 μg of primary antibody was added to the supernatant, followed by incubation at 4 °C overnight. The following day, 80 μl of protein A/G magnetic beads was added to the mixture and incubated at 4 °C for 4 h. The beads were then washed eight times. Afterward, the beads were resuspended in 100 μl of polysome lysis buffer with 0.1% SDS and 30 μg of Proteinase K, and incubated at 50 °C for 30 min. The supernatant was collected, and 10 μl of phenol–chloroform–isoamyl alcohol was added. The samples were then centrifuged and the supernatant was collected. RNA was subsequently precipitated using 5 μl of linear polyacrylamide (10 μg/ml), 12 μl of 3 M sodium acetate, and 250 μl of pre-chilled ethanol. Finally, RNA enrichment was determined by qRT-PCR.
The following antibodies were used: H3K27cr, SIRT3, H3K4me3 (Abcam, ab213224), H3K27ac (Abcam, ab177178), H3K27me3 (Abcam, ab192985), SIRT5 (Santa Cruz, SC-271635), SIRT1 (Proteintech, 13,161–1-AP), SIRT2 (Proteintech, 19,655–1-AP), SIRT6 (Proteintech, 13,572–1-AP), HDAC2 (Santa Cruz, SC-9959), EP300 (Santa Cruz, SC-48343), GCN5 (Affinity, DF3383), EED (Abcam, ab240650), SUZ12 (Abcam, ab175187), and EZH2 (Abcam, ab191250).

Immunohistochemistry (IHC)
Tissue samples were baked at 65 °C for 4 h. The tissue sections were subsequently deparaffinized by incubating them twice in xylene and anhydrous ethanol, followed by sequential incubation in graded alcohol solutions of 95%, 85%, 75%, and 50%. Antigen retrieval was then conducted using 0.01 M citrate buffer. Peroxidase activity was terminated using 3% hydrogen peroxide. The tissue sections were blocked with 10% BSA for 30 min and then incubated with primary antibodies at 4 °C overnight. Afterward, they were incubated at room temperature for 30 min with horseradish peroxidase (HRP)-conjugated secondary antibodies. The samples were then incubated with the DAB staining solution in a humidified chamber for 3 min, followed by counterstaining with hematoxylin for 1 min. Finally, the tissue sections underwent a graded alcohol series (75%, 80%, and 95%) for 5 min each, followed by dehydration in absolute ethanol and xylene for 10 min.

Hematoxylin and eosin (HE) staining
Tissue sections were stained with hematoxylin for 2 min, followed by successive incubation in 1% aqueous ammonia solution and eosin for 5 s each. After staining, the sections underwent treatment in a graded series of alcohol solutions: 85%, 95%, and 100% for 2 min each, followed by treatment with 100% alcohol for 4 min and xylene I and II for 10 min each. Finally, the tissue sections were sealed with neutral resin.

Animal assay
Five-week-old BALB/c nude mice were procured from GemPharmatech (Nanjing, China) and randomly divided into two groups, each consisting of six mice. HCT116 cells, either with or without stable overexpression of MMP14 RNA, were injected via the tail vein at a dose of 2 × 106 cells per injection. The mice were euthanized 5 weeks post-injection. Subsequently, the lungs were collected, photographed, and analyzed. This study was approved by the Ethical Committee for Animal Experimentation and the Ethics Committee of the Affiliated Hospital of Xuzhou Medical University (202306T019).

Statistical analysis
Data were presented as means ± SD. Differences between the two groups were analyzed using an unpaired two-tailed Student’s t-test, with a significance threshold set at P < 0.05. GraphPad Prism (version 8.0) was employed for graph generation and data analyses.

Results

Synonymous mutation of the MMP14 gene is associated with mortality and distant metastasis
To investigate CRC metastasis drivers, a comprehensive multi-cohort comparative analysis was conducted, revealing 42 genes that exhibited significant differential expression between distant metastatic and primary CRC tissues (Fig. 1A). Survival analysis in the The Cancer Genome Atlas (TCGA) CRC cohort identified 8 prognostically significant genes from the 42 genes (Fig. 1B). Single-cell transcriptomic analysis revealed that CRC-tissue-derived epithelial cells exhibited significant upregulation of MMP14, CLU, and SPP1 (Fig. 1C, D). qRT-PCR analysis revealed significantly elevated MMP14 expression in CRC tissues compared with normal tissues, consistent with TCGA cohort trends (Fig. 1E; Additional file 1: Fig. S1A–C). Overall, 2.89–7.17% of patients with CRC exhibited mutations in the MMP14 gene (Additional file 1: Fig. S1D). Notably, approximately 70% of MMP14 mutations in carriers were synonymous, a frequency disproportionately higher than in other top 20 mutated genes (Additional file 1: Fig. S1E). MMP14 synonymous variants correlated with elevated mortality and distant metastasis rates (Fig. 1F, G), indicating MMP14RNA-mediated CRC metastasis.

MMP14 RNA facilitates CRC cell invasion and migration
Subsequently, the functional role of MMP14 RNA in promoting CRC metastasis was examined. Transcriptomic profiling of MMP14 synonymous mutation carriers identified 178 downregulated and 29 upregulated genes (|log2FC|> 1.5, P < 0.05) compared with wild-type controls (Additional file 1: Fig. S2A). Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of these 207 genes demonstrated significant enrichment of cell adhesion molecules (Additional file 1: Fig. S2B), suggesting MMP14 RNA-mediated CRC metastasis. Next, two MMP14 expression plasmids were constructed: (i) a protein-coding plasmid containing the ORF sequence, and (ii) an RNA-specific plasmid harboring full-length MMP14 mRNA with an ATG-to-TTG mutation in the initiation codon (Additional file 1: Fig. S2C). Transient transfection of either construct significantly enhanced invasion and migration levels (Fig. 2A; Additional file 1: Fig. S2D). Stable expression of MMP14 RNA exhibited this pro-metastatic effect, whereas suppression of MMP14 RNA showed the opposite effect (Fig. 2B, C; Additional file 1: Fig. S2E, F). Cas9-mediated MMP14 protein knockout (with preserved RNA expression) combined with siMMP14 treatment further confirmed RNA-driven metastatic potential (Fig. 2D; Additional file 1: Fig. S2G–H). Immunoblotting analysis showed that overexpression of MMP14 RNA downregulated E-Cad while upregulating N-Cad, Vimentin, and Snail1, whereas MMP14 RNA silencing exhibited the opposite trend (Fig. 2E–H). These data collectively demonstrated that MMP14 RNA facilitated the invasion and migration levels of CRC cells.

MMP14 RNA facilitates cellular invasion and migration by inhibiting STARD13
To study the pro-invasive mechanism of MMP14 RNA, 18 downstream genes regulated by MMP14 RNA were identified (Additional file 1: Fig. S3A). Among these, STARD13, a Rho GTPase-activating protein (GAP) and tumor suppressor gene, exhibited a significant negative correlation with MMP14 RNA levels (Additional file 1: Fig. S3B). The qRT-PCR confirmed marked downregulation of STARD13 in CRC tissues (Fig. 3A), consistent with the trend observed in TCGA cohort (Additional file 1: Fig. S3C). MMP14 RNA overexpression reduced both RNA and protein levels of STARD13, which were increased by MMP14 silencing (Fig. 3B–G). Transient transfection of a STARD13 plasmid into MMP14-RNA-stabilized cells successfully rescued invasion and migration levels by restoring the expression of E-Cad and Snail1 (Fig. 3H, I). Collectively, these data demonstrated that MMP14 RNA drove invasion and migration via STARD13 suppression.

MMP14 RNA suppresses the transcriptional activity of the STARD13 promoter
Subcellular localization analysis revealed part of MMP14 RNA accumulated in the nucleus, providing mechanistic insight into its suppression of STARD13 expression (Fig. 4A). LongTarget, a tool for predicting lncRNA–DNA interactions, predicted MMP14 RNA binding to the STARD13 promoter (Fig. 4B), further validated by ChIRP assays showing significant MMP14 probe enrichment at the STARD13 promoter (Fig. 4C–G). Luciferase reporter assays revealed reduced STARD13 promoter activity in CRC cells transfected with MMP14-ATGmut plasmid, whereas MMP14 silencing enhanced activity (Fig. 4H, I). Altogether, these observations demonstrated that MMP14 RNA bound to the STARD13 promoter and suppressed its transcriptional activity.

MMP14 RNA reduces H3K27cr occupation on the STARD13 promoter
Subsequently, the mechanism by which MMP14 RNA suppressed transcriptional activity of the STARD13 promoter was studied. GSEA analysis revealed that MMP14 RNA expression was associated with H3K27cr-enriched promoters (Fig. 5A). Furthermore, the RIP assay demonstrated that H3K27cr significantly interacted with MMP14 RNA (Fig. 5B). Further investigation into MMP14-RNA-mediated regulation of STARD13 via H3K27cr revealed that croconic acid disodium salt (NaCr) rescued STARD13 expression suppressed by MMP14 RNA (Fig. 5C). Furthermore, ChIP-PCR analysis revealed that MMP14-ATGmut plasmid transfection in HCT116 cells impaired H3K27cr enrichment at the STARD13 promoter (Fig. 5D), suggesting that MMP14 RNA inhibited STARD13 expression via H3K27cr.
ChIP-seq analysis of H3K27cr in MMP14-ATGmut-transfected HCT116 cells identified 29,257 differential peaks, including 22,158 upregulated and 7099 downregulated peaks (Fig. 5E; Additional file 1: Fig. S4A; Additional file 2: Table S3). H3K27cr enrichment is predominantly found at the promoters of active genes [20]. Therefore, we analyzed the 20.15% of differential peaks localized to promoter regions, identifying 622 genes harboring at least two differential peaks within their promoters (Additional file 1: Fig. S4B). KEGG enrichment of these peaks highlighted metastasis-associated pathways (e.g., focal adhesion and adherens junction) (Additional file 1: Fig. S4C). These genes exhibited significant enrichment in distant metastatic sites (Fig. 5F) and had a predominant association with epithelial cells (Additional file 1: Fig. S4D). Metascape analysis of the 622 genes identified that the RHO GTPase cycle pathway was enriched, which showed significant enrichment in distant metastatic sites (Additional file 1: Fig. S5A, B). Single-cell transcriptomic analysis revealed RHO GTPase cycle pathway was enriched in LM-derived epithelial cells (Fig. 5G, H; Additional file 1: Fig. S5C). Notably, the enrichment was validated in the independent GSE225857 cohort, a previously analyzed cohort [19] (Additional file 1: Fig. S5D, E). These findings suggested that MMP14 RNA regulated CRC metastasis via the H3K27cr-mediated RHO GTPase cycle pathway.

MMP14 RNA regulates H3K27cr occupation on the STARD13 promoter by binding to SIRT3
To investigate how MMP14 RNA regulates H3K27cr occupancy at the STARD13 promoter, we focused on the crotonyltransferases and decrotonylases. The RIP assay revealed SIRT3 exhibited the strongest association among the identified histone-modifying enzymes (Fig. 6A). RNA-FISH demonstrated nuclear colocalization of SIRT3 protein and MMP14 RNA (Fig. 6B). Domain analysis revealed the N-terminal region of SIRT3 interacted with MMP14 RNA (Fig. 6C, D). ChIP-seq analysis in MMP14-ATGmut-transfected HCT116 cells identified 1652 upregulated and 21 downregulated SIRT3-binding peaks (Fig. 6E; Additional file 1: Fig. S6A; Additional file 3: Table S4). Among these differential peaks, 14.52% localized to promoter regions (Additional file 1: Fig. S6B), corresponding to 240 genes showing significant enrichment in distantly metastatic CRC tissues (Fig. 6F). These 240 genes demonstrated cell-type specificity for LM-derived epithelial cells, Tex, and Treg populations (Fig. 6G), with pathway analysis revealing the RHOA GTPase cycle as the most enriched signaling cascade (Additional file 1: Fig. S6C), which was enriched in LM-derived epithelial cells (Fig. 6H). H3K27cr levels at promoters of 8 genes within the 240-gene subset showed MMP14-RNA-dependent regulation (Fig. 6I). At the STARD13 promoter, MMP14 RNA overexpression triggered SIRT3 recruitment and concomitant H3K27cr reduction (Fig. 6J), with ChIP-PCR assays further demonstrating MMP14-RNA-mediated SIRT3 enrichment at this locus (Fig. 6K). Rescue experiments in HCT116 cells co-transfected with MMP14-ATGmut and siSIRT3 revealed that SIRT3 knockdown rescued MMP14-RNA-mediated STARD13 repression (Fig. 6L, M). These findings suggested that MMP14 RNA regulated the H3K27cr occupancy by binding to SIRT3 and facilitating its recruitment to the STARD13 promoter region.

MMP14 RNA facilitated CRC metastasis in vivo
Subsequently, the pro-metastatic function of MMP14 RNA was investigated in vivo. The result revealed MMP14 RNA overexpression promoted pulmonary metastatic nodule formation and concomitant increases in lung weight and metastatic area (Fig. 7A–F). Molecular analysis revealed MMP14-RNA-mediated downregulation of E-cadherin and STARD13 (Fig. 7G). The comparison of the effects of inhibiting MMP14 RNA and MMP14 protein on cell invasion revealed that inhibiting MMP14 RNA was more effective (Fig. 7H). Taken together, these findings suggest that MMP14 RNA facilitates CRC metastasis in vivo, and nucleic acid therapeutics targeting MMP14 RNA may provide greater anti-tumor efficacy than traditional protein inhibitors.

Discussion
This study aims to reveal a non-canonical, lncRNA-like mechanism of MMP14 RNA in CRC metastasis, distinct from its protein’s protease activity. Clinical data analysis shows that MMP14 RNA overexpression correlates with poor prognosis in CRC and drives metastasis. Mechanism investigation uncovers that MMP14 RNA interacts with SIRT3 and facilitates SIRT3 recruitment to the STARD13 promoter, reducing H3K27cr levels and suppressing STARD13 expression, finally promoting CRC metastatic progression (Fig. 7I). This RNA-driven mechanism may account for the limited effectiveness of MMP14 inhibitors in reducing tumor burden or improving patient survival [21]. The current study demonstrates that RNA silencing of MMP14 suppresses tumor metastasis more effectively than protein inhibition, particularly under MMP14 RNA overexpression conditions, indicating that protein-level monotherapy may be insufficient to halt tumor progression. In the future, nucleic acid therapeutics against MMP14 RNA may offer greater anti-tumor efficacy.
Increasing evidence indicates that lncRNAs play a crucial role in cancer progression [22, 23]. However, the distinction between coding and non-coding has become increasingly ambiguous [24]. At the beginning of this century, researchers identified the first bifunctional RNA, challenging the conventional classifications [25, 26]. Recently, more and more studies further demonstrate that lncRNAs regulate cancer progression by encoding short open reading frame (sORF)-derived peptides [27]. Conversely, some protein-coding genes have been shown to function as non-coding RNAs, although such examples are rare. Tuck et al. were the first to highlight a subset of “lncRNA-like” mRNAs, and our research identified another subset of mRNAs exhibiting lncRNA-like regulatory functions [28, 29]. Recently, NMRK2 mRNA was considered to function as a lncRNA-like scaffold enhancing mitochondrial respiration in NONO-TFE3 renal cell carcinoma [30]. lncRNA-like transcripts act as protein scaffolds and chromatin anchors, thereby regulating gene transcription [31, 32]. Our findings specifically reveal that MMP14 RNA binds to both the SIRT3 protein and the STARD13 promoter region, suggesting a novel lncRNA-like mechanism driving colorectal cancer metastasis. Further investigation is required to fully elucidate the tumor-regulatory roles and molecular pathways of these bifunctional RNAs.
The biological significance of synonymous mutations has long been contentious. Early studies regarded synonymous mutations as “neutral mutations” because they do not alter protein sequences [33]. However, recent research suggests that most synonymous mutations may be deleterious [34]. In cancer, synonymous mutations frequently act as driver mutations, with substitutions notably enriched in oncogenes rather than tumor suppressor genes [35]. Although these mutations do not change the encoded amino acids, they can influence gene expression and impact RNA stability, folding, and splicing, thus affecting the speed or accuracy of mRNA translation [36]. Moreover, mutations in lncRNAs have been implicated in tumorigenesis and tumor progression [37]. This phenomenon may be attributed to the ability of these mutations to alter RNA secondary structures, thereby influencing the binding affinity of proteins to RNAs [38]. Although the correlation between MMP14 expression and prognosis has been extensively validated [39], the association between its somatic mutations and clinical outcomes remains underexplored, likely owing to the low mutation frequency of MMP14 in tumors [40]. In the current study, we found that synonymous mutations in the MMP14 gene were significantly associated with increased mortality and distant metastasis. We proposed that these synonymous mutations might alter the secondary structures of MMP14 RNA, thereby altering its interaction with SIRT3. This altering could ultimately influence metastasis through the H3K27cr/STARD13 pathways. The lncRNA-like behavior of mRNA may provide a novel explanation for the deleterious effects of synonymous mutations.
Our previous research demonstrated that H3K27cr is elevated in CRC tissues and serves as a transcriptional activation marker at oncogene promoters (e.g., ETS1) to activate oncogene transcription [17, 19]. However, its role in tumor suppressor gene (TSG) remains poorly understood. Liu et al. showed H3K27cr-mediated TSG (e.g., p21) transcriptional repression by recruitment of the GAS41/SIN3A-HDAC1 repressive complex [41]. Here, we revealed H3K27cr enrichment was reduced at TSG promoters (e.g., STARD13) and suppressed gene expression, consistent with the idea that H3K27cr accumulates at the promoters of active genes [20]. The mechanisms maintaining spatial H3K27cr balance—elevated at oncogenes versus reduced at TSGs—remains poorly characterized. This study uncovered a spatially precise regulatory axis: MMP14 RNA scaffolds SIRT3 recruitment to the STARD13 promoter, catalyzing localized H3K27cr removal to dynamically modulate TSG transcription in CRC.

Conclusions
In summary, MMP14 RNA is highly expressed in CRC tissues and correlates with aggressive metastasis and poor clinical outcomes. MMP14 RNA operates through a non-canonical lncRNA-like mechanism, regulating chromatin remodeling by recruiting SIRT3 to reduce H3K27cr levels at the STARD13 promoter, which subsequently represses STARD13 expression and drives malignant progression. These findings elucidate a novel lncRNA-like function of MMP14 RNA in cancer progression, which may lead to novel cancer treatment strategies.

Supplementary Information