The clinical significance of the novel lncRNA TENM3-AS1 in GC metastasis

To identify lncRNAs playing a role in GC metastasis, we performed the whole RNA-sequencing in four paired primary (Pri group) and metastatic tumors (Pe-Met group) from four GC patients with peritoneal metastasis in the SYSUCC cohort. Compared to primary GC tumors, 708 lncRNAs were significantly upregulated, while 732 lncRNAs were downregulated in metastatic peritoneal tumors (Fig. 1A, Fig. S1A, and Tab. S5). Additionally, the transcriptome matrix and clinicopathologic data of 36 non-tumoral normal samples and 412 GC samples were retrieved from the TGCA-STAD cohort. Tumor samples were further divided into 365 M0-staged tumors (non-metastatic group; non-Met) and 29 M1-staged tumors (metastatic group; Met), excluding samples with undefined M stages. This analysis identified 228 significantly upregulated lncRNAs and 8 downregulated lncRNAs in the metastatic group compared to the non-metastatic group (Fig. 1A, Fig. S1A, and Tab. S6). After overlapping the up-regulated lncRNAs from the SYSUCC cohort and the TCGA-STAD cohorts, ten shared lncRNAs were selected for further analysis (Fig. 1A). Among these, four lncRNAs (LINC00565, TENM3-AS1, LINC00639, and LINC01303) were significantly upregulated in GC samples from the TCGA-STAD cohort compared to non-tumoral samples, while two lncRNAs (LINC01697 and LINC01783) were significantly reduced in tumoral samples (Fig. 1B). We next performed univariate Cox regression analysis for OS of GC patients on the basis of expression of the four upregulated lncRNAs in the TCGA-STAD cohort, and identified TENM3-AS1, an antisense RNA of the Teneurin transmembrane protein 3 located on chromosome 4q34, as the only independent predictor of the impaired prognosis in GC patients (Fig. 1C).

Fig. 1

Identification of the novel lncRNA TENM3-AS1 and its clinical significance to GC metastasis. (A) Overview of workflows to identify clinically significant lncRNAs in GC patients of the SYSUCC and the TCGA-STAD cohorts and select ten shared lncRNAs after overlapping the upregulated lncRNAs. (B) Expression levels of ten candidate lncRNAs in non-tumoral normal samples and GC samples in the TCGA-STAD dataset. (C) Univariate Cox regression analysis for overall survival based on the expression levels of four eligible lncRNAs in the TCGA-STAD cohort. (D) The fold changes of TENM3-AS1 expression in tumor samples versus non-tumoral normal samples across different cancer types in the TCGA database. (E) Representative images of RNA-FISH assays in human GC tissues from M0-staged and M1-staged patients of the SYSUCC cohort using TENM3-AS1 probe and DAPI. Scale bar means 20 μm. (F) Expression level of TENM3-AS1 in GC patients with low- or high-grade T stage (left) and AJCC Stage (right) from the TCGA-STAD cohort. (G) Kaplan-Meier curves displaying that GC patients with high expression of TENM3-AS1 gain the poor overall survival (left) and progression-free survival (right). Error bars represent mean ± SD. * means p < 0.05, ** means p < 0.01, and **** means p < 0.0001

Similar to gastric cancer (STAD), TENM3-AS1 was also significantly upregulated in other solid tumors compared to the corresponding non-tumoral tissues in the TCGA database, like head and neck cancer (HNSC), kidney papillary cell carcinoma (KIRP), colon cancer (COAD), Sarcoma (SARC), and lung squamous cell carcinoma (LUSC) (Fig. 1D). FISH assay was utilized to probe TENM3-AS1 in the primary tumors of 8 GC patients, half staged as M0 and half patients from Pe-Met group in the SYSUCC cohort. Our results revealed that TENM3-AS1 expression was markedly higher in advanced tumors compared to non-metastatic tumors (Fig. 1E). To investigate the association between TENM3-AS1 expression and various clinicopathologic factors, including T stage and American Joint Committee on Cancer (AJCC) stage, we compared TENM3-AS1 expression levels in GC patients with advanced versus lower stages. This analysis revealed a positive correlation between TENM3-AS1 expression and advanced T and AJCC stages in GC patients (Fig. 1F). In addition, high expression of TENM3-AS1 predicted the cacoethic OS and progression-free survival (PFS) in GC patients (Fig. 1G). The secondary structure of TENM3-AS1 was obtained via RNAfold software (Fig. S1B) and the coding potential of TENM3-AS1 was calculated through the CPC2 web tool and PhyloCSF software, which illustrated that TENM3-AS1 had no coding capability similar to other widely-recognized lncRNAs (Fig. S1C-D). In summary, TENM3-AS1 is shown to be predominantly elevated in metastatic or advanced tumors and correlated with a poorer prognosis in GC patients.

TENM3-AS1 is transcriptionally activated by the transcriptional factor EGR1

The expression levels of TENM3-AS1, in the human gastric mucosal epithelial cell line GES-1 and a variety of GC cell lines, were determined by RT-qPCR assay. Results demonstrated the overexpression of TENM3-AS1 in a majority of GC cell lines, in particular HGC27 and BGC823 cells (Fig. 2A), which conformed to the results above from non-tumoral samples and GC samples. In contrast, TENM3-AS1 expression was slightly changed in MKN28 cells compared to GES-1 cells (Fig. 2A). Since mRNA or lncRNA are mostly controlled by transcription factors (TFs) [23], we investigated potential TFs involved in the transcriptional activation of TENM3-AS1. Analysis of the expression levels of TFs in the SYSUCC cohort revealed 116 obviously upregulated TFs in the Pe-Met group contrasting to the Pri group (Fig. 2B and Tab. S7). We then identified candidate TFs positively correlated with TENM3-AS1, setting the criteria as Spearman’s ρ value > 0.2 and p-value < 0.5 (Tab. S8). Additionally, we explored TFs potentially binding to the promoter of TENM3-AS1 using the JASPAR website (Tab. S9). After intersecting these results, 23 eligible TFs were screen out (Fig. 2C). Of note, the top ten TFs identified in the JASPAR result were considered as the highest potential regulators of TENM3-AS1 expression (Fig. 2D). Furthermore, the public single-cell RNA sequencing data of thirteen gastric antral mucosa biopsies from nine patients with non-atrophic gastritis or GC in GSE134520 dataset was acquired from GEO database. The UMAP analysis of ten aforementioned TFs revealed that only EGR1 was evidently expressed in GC cells compared to stromal or immune cells (Fig. 2E and Fig. S2A-B). The expression levels of EGR1 were compared between GC patients with M0 and M1 stage in TCGA-STAD cohort, which unfortunately showed that there was no difference of EGR1 expression between these two groups (Fig. S2C). Therefore, IHC experiments were conducted to stain EGR1 protein in primary tumor tissues from 4 GC patients staged as M0 and 4 GC patients staged as M1 in the SYSUCC cohort. The results indicated that EGR1 expression was significantly elevated in GC patients with metastatic tumors (Fig. 2F and Fig. S2D). Furthermore, EGR1 expression was increased in GC samples of higher-risk AJCC stages in the TCGA-STAD cohort when comparing EGR1 expression between GC patients diagnosed as Stage I and II-IV (Fig. 2G). However, there was no significant difference of EGR1 expression between GC patients with Stage I-II and III-IV (Fig. S2E). High expression of EGR1 was discovered to have a detrimental impact on PFS, OS, and disease-specific survival (DSS) in GC patients (Fig. 2H and Fig. S2F). Our data showed that EGR1 actually serves as a detrimental TF in GC progression.

Fig. 2

TENM3-AS1 is transcriptionally regulated by EGR1. (A) Relative expression levels of TENM3-AS1 in GES-1 cell and GC cell lines. (B) Volcano plot showing the differently expressed TFs in metastatic tumors versus primary tumors in mRNA-sequencing data of the SYSUCC cohort. (C) Venn diagram to identify the potential TFs modulating TENM3-AS1 transcription. (D) Binding probability predicted in JASPAR database to find the highest-potential TFs. (E) Heatmap displaying expression levels of ten eligible TFs in malignant, stromal, and immune cells of GSE134520 dataset. (F) Representative images of IHC assays in human GC tissues from M0-staged and M1-staged patients of the SYSUCC cohort to indicate EGR1 expression. The scale bar means 100 μm in 10X field and 20 μm in 40X field. (G) Expression levels of EGR1 in GC patients with the low- or high-grade AJCC stage in TCGA-STAD cohort. (H) Kaplan-Meier curve showing GC patients with high expression of EGR1 attain a worse progression-free survival in TCGA-STAD cohort. (I) Correlation of TENM3-AS1 and EGR1 expression levels in TCGA-STAD cohort. (J) Validation of knocking down or overexpressing EGR1 in GC cells (top), and the corresponding relative expression changes of TENM3-AS1 (bottom). (K) ChIP-sequencing result from ENCODE database verifying EGR1 binds with the promoter region of TENM3-AS1 in H1 and Ishikawa cells. (L) The validated binding motif of EGR1. (M) The predicted binding sites (BS1-3) of EGR1 on the promoter of TENM3-AS1 and the designed fragments (#1, #2, and #3) for further ChIP analysis. (N) Result of ChIP assays using antibodies against EGR1 and IgG in GC cells. (O) Result of promoter dual-luciferase reporter assay performed by co-transfecting 293T cells with the plasmid pEZX-FR01 containing the TENM3-AS1 promoter region, along with either siRNAs targeting EGR1 or a synthesized plasmid overexpressing EGR1. Error bars represent mean ± SD. ** means p < 0.01, *** means p < 0.001, and **** means p < 0.0001

In addition, the certainly positive association between EGR1 and TENM3-AS1 in the TCGA-STAD dataset was drawn in Fig. 2I. The relationship was further explored in several GC cell lines using small interfering RNAs to knock down EGR1 (referred to as siEGR1 #1 and #2) and a plasmid to overexpress EGR1 (referred to as EGR1). The results demonstrated that TENM3-AS1 expression was significantly reduced alongside the downregulation of EGR1, while TENM3-AS1 expression increased in tandem with the elevated EGR1 expression (Fig. 2J and Fig. S2G). ChIP-sequencing results of H1 and Ishikawa cells using EGR1 antibody, visualized in the ENCODE database, suggested that EGR1-binding pools peaked in several sites, which were located in the promoter (2000 bp upstream from the transcription start site, namely TSS) of TENM3-AS1 (Fig. 2K). Using the JASPAR prediction tool and the validated binding motif of EGR1 (Fig. 2L), three predicted binding sites (BS1-3) were identified at 600 bp upstream from TSS in the promoter of TENM3-AS1 (Fig. 2M). Corresponding primers for different promoter segments (referred to as #1, #2, and #3) of TENM3-AS1 were synthesized for further experiments. ChIP experiments using an EGR1 antibody, along with RT-qPCR assays to validate the specific binding sites, were executed in HGC27 and BGC823 cells. The results determined that EGR1 was reproducibly bound to all the predicted sites within the promoter region of TENM3-AS1 (Fig. 2N and Fig. S2H-I). Finally, a promoter dual-luciferase reporter assay was finalized by co-transfecting 293T cells with the plasmid pEZX-FR01 containing the TENM3-AS1 promoter region, along with either siRNAs targeting EGR1 or a synthesized plasmid overexpressing EGR1. The corresponding control non-targeting siRNA and the blank vector plasmid were used as controls. Our results showed that luciferase activity was significantly declined when endogenous EGR1 expression was inhibited, whereas it was increased following exogenous EGR1 overexpression (Fig. 2O). Together our data demonstrated that EGR1 stimulates the transcriptional activation of TENM3-AS1.

TENM3-AS1 potentiates the migratory and invasive abilities of GC cells in vitro and tumor growth and liver metastasis in vivo

To explore the biological function of TENM3-AS1 in GC development and metastasis, functional assays were performed both in vitro and in vivo using HGC27 and BGC823 cells. These cells were stably transfected with short hairpin RNAs targeting TENM3-AS1 (designated as shTENM3-AS1 #1 and #2), a synthetic plasmid exogenously overexpressing TENM3-AS1 (designated as TENM3-AS1), or the corresponding controls (designated as shCtrl and Vector) (Fig. 3A and Fig. S3A). Colony formation assays confirmed that inhibiting TENM3-AS1 expression reduced the colony-forming competence of GC cells (Fig. 3B-C). In contrast, overexpressing TENM3-AS1 had the opposite effect, enhancing colony formation in GC cells (Fig. S3B-C). Similarly, wound-healing and transwell assays demonstrated that reducing TENM3-AS1 expression hindered the migratory and invasive abilities of GC cells (Fig. 3D-G), while increasing TENM3-AS1 expression enhanced the mobility of GC cells (Fig. S3D-G).

Fig. 3

TENM3-AS1 knockdown curbs the migration and invasiveness of GC cells in vitro and tumorigenesis and liver metastasis in vivo. (A) RT-qPCR result displaying the relative TENM3-AS1 expression after stably knocking down TENM3-AS1 by shRNAs in GC cells. (B-C) Representative graphs (B) and quantitative data (C) of colony formation assays in GC cells stably down-regulating TENM3-AS1. (D-E) Representative graphs (D) and quantitative data (E) of wound-healing assays in GC cells stably decreasing TENM3-AS1 expression. (F-G) Representative graphs (F) and quantitative data (G) of transwell assays in GC cells stably down-regulating TENM3-AS1. (H-J) Representative images of xenograft tumors morphology (H), quantitative data of tumors growth (I), and weights of xenograft tumors (J) in nude mice (5 mice per group) injected subcutaneously into groin region using the stably transfected GC cells. (K) Representative images of optical luciferase imaging assays in vivo in mouse models with liver metastasis. (L) HE staining assay showing the metastatic nodules in livers. (M-N) Quantitative data of in vivo optical luciferase imaging assays (M) and metastatic foci in liver tissues (N). Scale bars in this figure all signify 100 μm. Error bars represent mean ± SD. ** means p < 0.01, *** means p < 0.001, and **** means p < 0.0001

Next, we used animal experiments to investigate the role of TENM3-AS1 on GC tumor growth and metastasis in vivo. These results of subcutaneous xenograft tumor models in nude mice showed that tumors with TENM3-AS1 knockdown were significantly smaller, grew more slowly, and had reduced weights contrasting to the control group (Fig. 3H-J). In contrast, subcutaneous tumors overexpressing TENM3-AS1 exhibited faster growths and increased weights with comparison to the vector group (Fig. S3H-J). Furthermore, liver metastasis models via splenic vein injection in nude mice revealed that decreasing TENM3-AS1 expression significantly inhibited liver metastasis of GC cells and the number of the visible liver metastatic nodules decreased in shTENM3-AS1 groups when compared to the shCtrl control group (Fig. 3K-N). Conversely, overexpressing TENM3-AS1 obviously promoted liver metastasis of HGC27 cells and increased the number of the visible liver metastatic nodules (Fig. S3K-N). In a nutshell, TENM3-AS1 promotes GC cell migration and invasion in vitro, and tumorigenesis and liver metastasis in vivo.

TENM3-AS1 increases FASN expression and induces fatty acid biosynthesis in GC metastasis

To better understand the underlying mechanism by which TENM3-AS1 regulated GC metastasis, we analyzed differentially expressed genes (DEGs) by comparing RNA-sequencing data between Pri and Pe-Met groups. These studies revealed 1538 upregulated and 1443 downregulated DEGs in the Pe-Met group (p < 0.05 and|log Fold change| ≥ 1) (Fig. 4A and Tab. S7). Next, we performed functional pathways analysis via Gene Ontology Biological Process (GO-BP) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases in DAVID website. Results demonstrated that a series of metabolic processes were critical for GC metastasis formation, especially fatty acid (FA) metabolism (Fig. 4B-C). Among the processes of FA metabolism, including biosynthesis, transport, catabolism, and oxidation of FA, only the biosynthetic process of FA was significantly increased in GC patients with high expression of TENM3-AS1 in the TCGA-STAD cohort (Fig. 4D and Fig. S4A). To test whether TENM3-AS1 participated in the modulation of FA metabolism, the total content of cellular FAs was measured, revealing that silencing TENM3-AS1 significantly reduced cellular FA levels, while overexpressing TENM3-AS1 led to a marked increase in FA accumulation (Fig. 4E). Additionally, intracellular lipid droplets stained with the lipophilic fluorescent dye BODIPY 493/503 and Oil Red O demonstrated that lower TENM3-AS1 expression inhibited lipid storage, whereas higher TENM3-AS1 expression was associated with lipid accumulation in GC cells (Fig. 4F-I).

Fig. 4

TENM3-AS1 promotes fatty acid biosynthesis by increasing FASN in GC metastasis. (A) Heatmap exhibiting the differently expressed genes (DEGs) in metastatic (Pe-Met) and primary (Pri) tumors in the SYSUCC cohort. (B-C) Functional enrichment assays using GO-BP (B) and KEGG (C) pathways based on the DEGs in the SYSUCC cohort. (D) GSVA score of FA biosynthetic process in GC patients with high- or low-expression of TENM3-AS1 in TCGA-STAD cohort. (E) Quantitative data of FA detection assays in GC cells stably knocking down TENM3-AS1 (left) or overexpressing TENM3-AS1 (right). (F-G) Representative images (F) and statistical data (G) of BODIPY staining assays. The scale bar indicates 20 μm. (H-I) Representative images (H) and statistical data (I) of Oil Red O staining assays. The scale bar indicates 50 μm. (J) Venn diagram identifying the FA-biosynthesis-process-related gene potentially influenced by TENM3-AS1. (K) RT-qPCR result revealing the relative RNA expression of FASN, SCD1, CPT1A, and CD36 in stably transfected HGC27 cells. (L) Result of Western blot experiments performed in stably transfected GC cells to validate protein expression of FASN, SCD1, CPT1A, and CD36. Error bars represent mean ± SD. ns means not significant, * means p < 0.05, ** means p < 0.01, *** means p < 0.001, and **** means p < 0.0001

To identify FA biosynthesis-related genes potentially regulated by TENM3-AS1, genes that positively correlated with TENM3-AS1 (Tab. S8), genes upregulated in the Pe-Met group of the SYSUCC cohort (Tab. S7), and genes involved in the FA biosynthetic process (GO: 0006633) were intersected. This analysis revealed ten key genes (Fig. 4J), among which FASN appeared to be the most promising candidate. Previous study identified that FASN catalyzed the transformation of acetyl-CoA and malonyl-CoA into palmitate and other long-chain saturated FAs, which was linked to GC progression and metastasis [24]. To unveil how TENM3-AS1 influenced FA biosynthesis, we examined the expression levels of pivotal genes related to FA biosynthesis (FASN and SCD1), β-oxidation (CPT1A), and lipid transport (CD36) in GC cells. Our results indicated that TENM3-AS1 knockdown was correlated with the reduced mRNA and protein expression levels of FASN and SCD1, while minimal impact on CPT1A and CD36 expression was observed (Fig. 4K-L and Fig. S4B-D). Conversely, overexpression of TENM3-AS1 was associated with an increased expression of FASN and SCD1 at both mRNA and protein levels, while CPT1A and CD36 levels remained unaffected (Fig. 4K-L and Fig. S4B-D). Collectively, TENM3-AS1 enhances FA biosynthesis and accumulation in GC metastasis in part by upregulating FASN expression.

TENM3-AS1 promotes GC migration and invasiveness via interacting with hnRNPK and enhancing its deubiquitination

It is well accepted that the function of lncRNAs depends upon the proper subcellular localization. Subcellular fractionation assays and RT-qPCR experiments, along with RNA-FISH analysis in GC cells demonstrated that TENM3-AS1 was predominantly localized in the cytoplasm (Fig. 5A-B). On the basis of the results suggesting a regulatory relationship between TENM3-AS1 and FASN, a RIP assay using an anti-FASN antibody was performed but unexpectedly revealed that TENM3-AS1 did not directly interact with the FASN protein (Fig. S4E). To identify potential candidates that might be involved in TENM3-AS1 regulation of FASN, we performed an RNA pull-down assay and Mass Spectrometry (MS) analysis comparing the biotin-labeled TENM3-AS1 sense probe and the control anti-sense probe (Fig. 5C). Among the proteins pulled down by TENM3-AS1, we identified hnRNPK (Fig. 5D), an RNA-binding protein relevant to GC progression [25]. Interaction between TENM3-AS1 RNA and hnRNPK protein was verified by RNA pull-down assay and Western blot experiment (Fig. 5E) as well as in vitro RIP assay in GC cells (Fig. 5F and Fig. S4F). To delineate the region(s) of TENM3-AS1 responsible for interacting with hnRNPK, RNA pull-down assays using various exon probes labeled by biotin spanning TENM3-AS1 sense sequences were performed. These studies showed that the exon 5 segment of TENM3-AS1 was involved in hnRNPK binding (Fig. 5G). Previous studies have shown that the K homology (KH) domains of hnRNPK, consisting of KH1, KH2, and KH3, are crucial for its RNA-binding activity [26]. Meanwhile, the K interactive (KI) domain, located between KH2 and KH3, plays an intricate role in modulating RNA-protein interactions [27]. We constructed several Flag-tagged plasmids with various truncations of hnRNPK and performed in vitro RIP assays. Our studies confirmed that hnRNPK interacted with TENM3-AS1 through the KI domain (Fig. 5H and Fig. S4G).

Fig. 5

TENM3-AS1 interacts with RNA-binding protein hnRNPK and increases its deubiquitination in GC cells. (A) Subcellular fractionation assays revealing the relative abundance of TENM3-AS1 in cytoplasm and nucleus of GC cells. GAPDH and U6 are employed as indicators for the cytoplasmic and nuclear region, respectively. (B) Representative images (left) and statistical data (right) of RNA-FISH assays using TENM3-AS1, 18S ribosomal RNA, and U6 probes. 18S ribosomal RNA and U6 are used as positive controls for cytoplasm and nucleus, respectively. The scale bar symbolizes 20 μm. (C) Schematic workflow of the biotinylated RNA pull-down assay and Mass spectrometry to clarify TENM3-AS1-binding proteins. (D) Representative hnRNPK peptides binding with TENM3-AS1 detected by Mass spectrometry. (E-F) Result of RNA pull-down analysis (E) and RIP assay (F) validating the interaction between TENM3-AS1 and hnRNPK. (G) Schematic diagram of full-length sense or anti-sense and specific exons fragments spanning TENM3-AS1 (top) and result of RNA pull-down analysis using different biotinylated RNA fragments in HGC27 cell (bottom). (H) Schematic diagram of full domains and constructed truncations of hnRNPK (left), validation of biosynthesized Flag-tagged plasmids with various truncations of hnRNPK (right and top) and result of RIP assay using antibody against Flag in HGC27 cell (right and bottom). (I) Western blot experiment revealing the protein expression of hnRNPK following TENM3-AS1 knockdown or overexpression. (J) RT-qPCR assay indicating the RNA expression of hnRNPK in GC cells diminishing or increasing TENM3-AS1 expression. (K) GSVA score of protein deubiquitination process in GC patients with high- or low-expression of TENM3-AS1 in the TCGA-STAD cohort. (L) Result of CHX-chase assay to observe the protein stabilization of hnRNPK in GC cells stably knocking down TENM3-AS1 using cycloheximide (CHX, 50 ug/mL) treatment at different time. (M) Western blot experiments to detect hnRNPK protein in stably transfected GC cells treated with CHX at 6 h and with or without MG132 (25 μM). (N) Result of in vitro ubiquitination assay revealing hnRNPK deubiquitination in 293T cell. (O) Representative images (top) and statistical data (bottom) of rescued transwell assay for migration in GC cells. Scale bars symbolize 100 μm. Error bars represent mean ± SD. ns means not significant, * means p < 0.05, ** means p < 0.01, *** means p < 0.001, and **** means p < 0.0001

Furthermore, Western blot and RT-qPCR analyses illuminated that TENM3-AS1 knockdown reduced the expression of hnRNPK at the protein level, while overexpression of TENM3-AS1 increased hnRNPK protein expression, without significantly affecting its mRNA expression (Fig. 5I-J and S4H). Studies have shown that lncRNA can modulate protein expression through post-translational modification such as ubiquitination [28]. Next, we performed GSVA analysis using GO-BP gene sets and found that the protein deubiquitination process was enriched in GC patients with high expression of TENM3-AS1 in the TCGA-STAD cohort (Fig. 5K). We hypothesized that TENM3-AS1 might increase hnRNPK protein expression by enhancing its deubiquitination. CHX-chase assay and treatment with MG132 confirmed that TENM3-AS1 knockdown led to destabilization of the hnRNPK protein, resulting in an increased degradation rate in GC cells (Fig. 5L-M and S4I-J). An in vitro ubiquitination assay showed that hnRNPK ubiquitination was increased along with TENM3-AS1 knockdown, whereas TENM3-AS1 overexpression was associated with reduced hnRNPK ubiquitination (Fig. 5N and S4K). These results suggested that TENM3-AS1 upregulated hnRNPK protein expression by facilitating hnRNPK deubiquitination and increasing protein half-life. To further investigate whether TENM3-AS1 promoted GC progression via its interaction with hnRNPK, we performed rescue experiments by simultaneously knocking down TENM3-AS1 and exogenously upregulating hnRNPK in GC cells. These studies revealed that hnRNPK could restore the reduced migratory and invasive capabilities caused by TENM3-AS1 knockdown (Fig. 5O and Fig. S5A-B). Likewise, in vivo rescue experiments of liver metastasis models unraveled that exogenously overexpressing hnRNPK reinforced the restrained ability to metastasize to the liver triggered by TENM3-AS1 knockdown in GC cells (Fig. S5C-D). Overall, these data demonstrated that TENM3-AS1 interacts with hnRNPK to enhance its deubiquitination, thereby increasing hnRNPK protein levels and promoting GC progression.

Activation of the TENM3-AS1/hnRNPK/FASN axis reprograms fatty acid metabolism

It has been widely recognized that RNA-binding proteins exerts important effects to stabilize RNA [29,30,31], which prompted us to test whether TENM3-AS1 might increase FASN expression through hnRNPK-mediated mRNA stabilization. Results of eCLIP-sequencing with hnRNPK antibody in HepG2 cells derived from the ENCODE database suggested that hnRNPK probably bound to FASN mRNA (Fig. 6A). Indeed, RNA pull-down and RIP assays confirmed the interaction between hnRNPK protein and FASN mRNA (Fig. 6B-C and Fig. S6A). Next, RT-qPCR analysis manifested that mRNA expression of FASN was noticeably reduced after hnRNPK knockdown by siRNAs, while its expression was elevated when exogenous hnRNPK was overexpressed in HGC27 cells (Fig. 6D). Furthermore, RNA stabilization assay using actinomycin D (Act-D) demonstrated that hnRNPK knockdown significantly decreased the half-life of FASN mRNA (2.846 h versus 0.753 h and 1.333 h), whereas hnRNPK overexpression prolonged its half-life in HGC27 cell (2.391 h versus 2.930 h) (Fig. 6E). Rescue expression experiments using shTENM3-AS1 #1 and an hnRNPK-overexpressing plasmid demonstrated that overexpression of hnRNPK effectively restored the mRNA and protein expression of FASN caused by TENM3-AS1 knockdown (Fig. 6F-G). Interestingly, the reduced half-life of FASN mRNA induced by TENM3-AS1 knockdown was significantly extended by hnRNPK overexpression (Fig. 6H). Additionally, we performed Western blot assays to assess the protein expression levels of hnRNPK and FASN in metastatic tumors from in vivo rescue experiments executed in Fig. S5C-D, which showed that hnRNPK overexpression rescued the attenuated expression of FASN generated by TENM3-AS1 knockdown (Fig. S6B). Furthermore, TENM3-AS1 knockdown led to a reduction in FA storage, which was rescued by overexpressing hnRNPK in GC cells (Fig. 6I-J). To confirm whether FASN participated in the regulatory process of TENM3-AS1-induced GC cells migratory functions, rescue experiments were conducted by co-transfecting shTENM3-AS1 #1 with a plasmid overexpressing FASN. The results revealed that FASN overexpression restored the migratory and invasive capabilities of GC cells, whereas TENM3-AS1 knockdown had the opposite effect (Fig. 6K-L and Fig. S6C). To summarize, TENM3-AS1 rewires FA metabolism and intensifies the aggressiveness of GC cells by hnRNPK-stabilized FASN mRNA and the following upregulation of FASN.

Fig. 6

TENM3-AS1 reprograms FA metabolism via hnRNPK-stabilized FASN mRNA. (A) eCLIP-sequencing result from ENCODE database validating the combination of anti-hnRNPK protein and FASN mRNA in HepG2 cell. (B-C) Results of RNA pull-down assay using sense and anti-sense probes of FASN (B) and RIP assay using anti-hnRNPK or IgG antibodies (C) confirming the interaction between hnRNPK protein and FASN mRNA. (D) RT-qPCR result unraveling the relative RNA expression levels of hnRNPK and FASN in GC cells knocking down (left) or upregulating (right) hnRNPK. (E) Result of RNA stabilization assay in transiently transfected HGC27 cell with 5 µg/mL actinomycin D (Act-D) treatment at different times to examine RNA stabilization of FASN. (F-G) Results of rescued RT-qPCR experiment (F) and Western blot assay (G) by co-transfecting GC cells with shRNA targeting TENM3-AS1 and a hnRNPK plasmid to observe the expression levels of TENM3-AS1, FASN, and hnRNPK. (H) Rescued RNA stabilization assay validating RNA stabilization of FASN in co-transfected HGC27 cell. (I-J) Representative images (I) and statistical data (J) of rescued BODIPY staining assay in GC cells. Scale bar means 20 μm. (K-L) Representative images (K) and statistical data (L) of rescued transwell assays for migration and invasion performed in GC cells co-infecting shRNA targeting TENM3-AS1 and a FASN plasmid. Scale bars represent 100 μm. Error bars represent mean ± SD. * means p < 0.05, ** means p < 0.01, *** means p < 0.001, and **** means p < 0.0001

TENM3-AS1/hnRNPK/FASN axis plays a significant role in GC metastasis

To disclose the clinical significance of hnRNPK and FASN, we compared their expression between non-tumoral and GC samples in TCGA-STAD. Our analysis revealed that both hnRNPK and FASN were considerably upregulated in GC samples when compared to non-tumoral samples (Fig. 7A). Importantly, high expression of FASN predicted a worsened OS and the shortened time to first progression in GC patients (Fig. 7B). However, the expression of hnRNPK was not significantly associated with the prognosis of GC patients (Fig. S6D). Additionally, IHC experiments in the SYSUCC cohort including 4 non-metastatic and 4 metastatic gastric tumors revealed that hnRNPK and FASN proteins were increased in the metastatic GC tissues, which coincided with high expression of TENM3-AS1 (Fig. 7C). Besides, the expression levels of FASN and hnRNPK were compared between GC patients with M0 and M1 stage in TCGA-STAD cohort, which showed that there was negligible difference of FASN and hnRNPK expression between two groups (Fig. S6E). To confirm the importance of the TENM3-AS1/hnRNPK/FASN regulatory axis in vivo, additional IHC experiments were conducted on xenograft tumor samples described in the Fig. 3F model. The results illuminated that hnRNPK and FASN protein levels were significantly diminished in the shTENM3-AS1 group compared to the control group (Fig. 7D), consistent with the in vitro findings. Ultimately, the proposed mechanism of TENM3-AS1 in gastric cancer (GC) metastasis is illustrated in the schematic diagram in Fig. 7E. This model depicts that the EGR1/TENM3-AS1/hnRNPK/FASN signaling axis can serve as a novel curative target for metastatic GC.

Fig. 7

Verification of clinical significance of TENM3-AS1/hnRNPK/FASN axis in GC metastasis. (A) Expression levels of hnRNPK (left) and FASN (right) in non-tumoral normal samples and GC samples in the TCGA-STAD dataset. (B) Kaplan-Meier curves illustrating that GC patients with high expression of FAS