N 6‐methyladenosine‐modified circSTX6 promotes hepatocellular carcinoma progression by regulating the HNRNPD/ATF3 axis and encoding a 144 amino acid polypeptide

Abstract Background Circular RNAs (circRNAs) play a significant role in the initiation and progression of various cancers, including hepatocellular carcinoma (HCC). Circular syntaxin 6 (circSTX6, also known as hsa_circ_0007905) has been identified as a microRNA (miRNA) sponge in pancreatic adenocarcinoma. However, its full range of functions in terms of protein scaffold and translation remain largely unexplored in the context of HCC. Methods The expression of circSTX6 and its encoded protein was examined in HCC tumour tissues. N 6‐methyladenosine (m6A) on circSTX6 was verified and quantified by methylated RNA immunoprecipitation (Me‐RIP), RIP and dual luciferase reporter assays. The biological functions of circSTX6 and its encoded protein in HCC were clarified by in vitro and in vivo experiments. Mechanistically, the interaction between circSTX6 and heterogeneous nuclear ribonucleoprotein D (HNRNPD) was investigated by RNA pull‐down, RIP and fluorescence in situ hybridization (FISH)/IF. The regulatory effects of circSTX6 and HNRNPD on activating transcription factor 3 (ATF3) mRNA were determined by mRNA stability and RIP assays. Furthermore, the presence of circSTX6‐encoded protein was verified by mass spectrometry. Results CircSTX6 and its encoded 144 amino acid polypeptide, circSTX6‐144aa, were highly expressed in HCC tumour tissues and served as independent risk factors for overall survival in HCC patients. The expression of circSTX6 was regulated by METTL14 in an m6A‐dependent manner. Functionally, circSTX6 accelerated HCC proliferation and tumourigenicity and reinforced tumour metastasis in vitro and in vivo. Mechanistically, circSTX6 acted as a sponge for HNRNPD protein, facilitating its binding to ATF3 mRNA, consequently promoting ATF3 mRNA decay. Meanwhile, circSTX6‐144aa promoted HCC proliferation, migration and invasion independent of circSTX6 itself. Conclusion Collectively, our study reveals that m6A‐modified circSTX6 drives malignancy in HCC through the HNRNPD/ATF3 axis, while its encoded circSTX6‐144aa contributes to HCC progression independent of circSTX6. CirSTX6 and its encoded protein hold promise as potential biomarkers and therapeutic targets in HCC.

• CircSTX6 promotes proliferation and metastasis of hepatocellular carcinoma (HCC) by binding to HNRNPD to facilitate HNRNPD-mediated ATF3 mRNA degradation.

BACKGROUND
Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related deaths globally, with an estimated 830, 000 fatalities annually. 1Despite significant advancements in diagnostic methods, surgical techniques and the emergence of targeted therapies and immunotherapies in recent years, the long-term overall prognosis for HCC patients remains unsatisfactory. 2,3Challenges persist due to the high rates of metastasis and recurrence, largely attributed to an incomplete understanding of tumour pathogenesis. 4hus, there is an imperative need to identify novel therapeutic targets and gain a deeper insight into the molecular mechanisms governing HCC for enhanced treatment efficacy.Circular RNAs (circRNAs), a unique subclass of endogenous noncoding RNAs (ncRNAs) characterized by their covalently closed loop structures, have attracted considerable attention. 5CircRNAs exhibit evolutionary conservation, tissue-specific expression and enhanced stability owing to their circularized structure. 5,6A mounting body of evidence suggested that circRNAs actively participate in the tumourigenesis and metastasis of cancers, and its role as miRNA sponge has been widely reported in HCC since its discovery. 7,8For instance, circASAP1 promotes HCC pulmonary metastasis by sponging miR-532-5p and miR-326 that mediate MAPK1 and CSF-1 signalling pathways. 9Notably, increasing studies are focusing on cir-cRNAs beyond their function as miRNA sponges.Our previous study demonstrated that circCPSF6 competitively binds to an RNA-binding protein (RBP) to activate Yes-associated protein (YAP) signalling and drive HCC malignancies. 10Moreover, studies related to the "unconventional" roles of circRNAs regarding protein encoding are emerging.Recently, a novel secretory polypeptide encoded by circZKSCAN1 was reported to suppress HCC progression through the degradation of mammalian targets of rapamycin (mTOR). 11Thus, additional efforts are needed to unveil the roles of circRNAs in HCC beyond the miRNA sponge.
N 6 -methyladenosine (m 6 A) as one of the most prevalent epigenetic modifications in eukaryotic RNAs, is catalyzed by methyltransferases such as methyltransferase-like protein 3 (METTL3), METTL14, and Wilm's tumour 1-associated protein (WTAP), recognized by m 6 A binding proteins such as YTH domain-containing family proteins (YTHDF) and insulin-like growth factor 2 mRNA-binding proteins (IGF2BP) and removed by demethylases including fat mass and obesity-associated protein (FTO) and AlkB homolog 5 (ALKBH5). 12m 6 a modification regulates RNA metabolism in diverse ways, including export, splicing, degradation, stability and translation. 13Intriguingly, accumulating studies have established a relationship between circRNAs and m 6 A. 14 For example, m 6 A can regulate circRNA degradation through YTHDF2-mediated endoribonuclease cleavage. 15imilarly, ALKBH5-mediated m 6 A modification on circ-CCDC134 increases its stability, promoting gastric cancer tumour growth and metastasis. 16Moreover, circMAP3K4 prevents HCC apoptosis by encoding circMAP3K4-455aa protein, whose translation is driven by IGF2BP1 in an m 6 A-dependent manner. 17These findings underscore the essential role of m 6 A in circRNA regulation across multiple cancers, including HCC. 10,18,19 Nonetheless, the functions of m 6 A-regulated circRNAs in HCC and their underlying mechanisms demand further investigation, and it would be important to understand the metabolism and functions of circRNAs by revealing whether and how they are regulated by m 6 A modification as well as to investigate novel therapeutics.
In this study, we identified circSTX6 by taking the intersection of circRNAs associated with m 6 A and HCC.The expression of circSTX6 was regulated by METTL14mediated m 6 A modification, and circSTX6 was prevalently upregulated in tumour tissues, correlating with adverse outcomes in HCC patients.Functionally, circSTX6 promoted cell proliferation, migration and invasion in vitro and accelerated tumour growth and metastasis in vivo.Mechanistically, circSTX6 bound to heterogeneous nuclear ribonucleoprotein D (HNRNPD) to enhance mRNA decay of activating transcription factor 3 (ATF3), thus inhibiting its expression and facilitating HCC development.Furthermore, circSTX6 encoded a novel 144 amino acid (aa) polypeptide that accurately predicted poor prognosis in HCC.CircSTX6-144aa promoted tumour progression independent of circSTX6 itself.In sum, our study expands the knowledge of m 6 A-guided regulation of circSTX6, reveals its novel mechanisms in HCC tumourigenesis and demonstrates its diagnostic and therapeutic potential.
Lentivirus plasmids carrying short hairpin RNA sequences specifically targeting circSTX6 or negative control RNAs were constructed by RepoBio (Hangzhou, China) as well as lentiviruses expressing circSTX6 and empty vectors.These plasmids were packaged with pMD2G and psPAX2 and transfected into 293T cells.Lentivirus supernatants were harvested 48 h after transfection and were used for the transfection of target cells.The transfected cells were subject to the filtration of 3 μg/mL puromycin for 1 week and checked for efficiency by qPCR.Similarly, circSTX6 pcDNA3.1,circSTX6 with ATG mutation (circSTX6-MUT) pcDNA3.1 and linear circSTX6-144aa pcDNA3.1 plasmids were constructed.Detailed sequences are listed in the Supplementary Materials and Methods section.

Statistical analysis
GraphPad Prism V8.0.1 software and R V.4.2.2 software were used for statistical analyses, and the results are presented as the means ± standard deviation (SD).The χ2 test was used to assess the qualitative data, and the paired or unpaired two-tailed Student's t-tests were used for the comparison of quantitative data.The log-rank tests were used for Kaplan-Meier survival analysis and the Cox regression analysis was used for the identification of independent prognostic factors.The Pearson correlation analysis was conducted between two genes with linear regression.In this study, p < 0.05 was defined as statistically significant (*p < 0.05; **p < 0.01; ***p < 0.001).
The in vitro assays including colony formation, apoptosis, 5-Ethynyl-20-deoxyuridine (EdU), wound-healing and transwell assays were conducted in three replicates.The CCK-8 assay was carried out in five replicates for each time point.Another detailed methodology was described in the Supplementary Materials and Methods section.

Identification and characterization of circSTX6 in HCC
To identify m 6 A-regulated circRNAs in HCC, we analysed circRNA-seq data from three published studies, includ-ing datasets of m 6 A-IP-identified circRNAs, 20 upregulated circRNAs in HCC tumour tissues 8 and genome-wide mapping of m 6 A-regulated circRNAs. 21As demonstrated in the Venn plot, only one particular m 6 A-related circRNA was obtained (Figure 1A).Hsa_circ_0007905 is derived from exons 4, 5, 6 and 7 of the gene STX6 on human chromosome 1 (q25, 3) (Figure 1B).Divergent and convergent primers were designed, respectively, based on the junction sites and parental genes.The back-splicing sites between exon 4 and exon 7 were validated by Sanger sequencing in HCC cells using divergent primers (Figures 1C and Supplementary Figure S1A).The expression of circSTX6 was determined in two hepatocytes and nine HCC cell lines.MHCC97H, HCCLM3 and Huh7 cells were chosen for subsequent studies (Supplementary Figure S1B).Additionally, according to agarose gel electrophoresis results, circSTX6 could only be amplified by divergent primers from cDNA instead of gDNA (Figure 1D).
The characteristics of circSTX6 were verified by a series of assays.HCC cells were treated with ribonuclease R (RNase R) and actinomycin D to determine the stability of circSTX6.Results showed that cricSTX6 was more resistant to RNase R and more stable under actinomycin D treatment compared to its parental transcripts (Figure 1E and Supplementary Figure S1C,D).RNA fractionation (nuclear/cytosolic) assay followed by qPCR demonstrated the predominant cytoplasm distribution of circSTX6 (Figure 1F), and this result was further confirmed by fluorescence in situ hybridization (FISH) assays performed on HCC cells (Figure 1G) and tumour tissues (Figure 1H and Supplementary Figure S1E).These results confirm that circSTX6 is a genuine circRNA characterized by the expected biological features.
To further evaluate the aberrant expression of circSTX6 in HCC, qPCR was performed on HCC cohort-1 with 72 pairs of tumour and adjacent tissues.The results revealed higher expression of circSTX6 in tumour tissues compared to matched adjacent tissues (Figure 1I and Supplementary Figure S1F).Moreover, the ISH staining assay demonstrated that circSTX6 was more abundant in tumour tissues of HCC cohort-3 (Figure 1J and Supplementary Figure S1G), and Kaplan-Meier survival analysis revealed that upregulation of circSTX6 in HCC patients was associated with worse overall survival (OS; p = 0.0052; Figure 1K) and recurrence-free survival (RFS; p = 0.0094) (Supplementary Figure S1H).Notably, high circSTX6 expression might be associated with advanced tumour stage and higher pathological grade (Figure 1L and Supplementary Figure S1I).Consistently, receiver operating characteristic (ROC) curves showed that the area under curve (AUC) value of circSTX6 is 0.8565 (p < 0.0001), suggesting circSTX6 could distinguish between HCC tumour and normal tissues (Supplementary Figure S1J).Besides, the clinical pathological characteristics of HCC cohort-3 were collected and analysed in detail.It was revealed that cricSTX6 was also significantly correlated with tumour differentiation (p = 0.0061), stage (p = 0.0331) and tumour size (p = 0.0012) (Supplementary Table S3).Additionally, overexpression of circSTX6 (hazard ratio [HR] = 1.183, p = 0.015) as well as AFP (HR = 4.410, p < 0.001) could be considered as independent risk factors for HCC overall survival (Figure 1M).Moreover, Kaplan-Meier survival analysis based on the combination of circSTX6 expression and AFP levels further revealed that HCC individuals with high circSTX6 expression and high AFP levels had an even worse OS than any other groups (p < 0.0001) (Supplementary Figure S1K), especially compared with those of low circSTX6 expression and low AFP levels.Taken together, the upregulation of circSTX6 in HCC indicates a poor prognosis and it may serve as a potential biomarker.

3.2
CircSTX6 is regulated by METTL14-mediated m 6 A modification Since circSTX6 was identified by taking the intersection of two m 6 A-related datasets, we explored whether circSTX6 is regulated by m 6 A 'writers' or 'erasers'.Interestingly, the knockdown of METTL14 led to increased circSTX6 expression in HCC cells (Figure 2A).Whereas inhibition of METTL3, WTAP, ALKBH5 and FTO scarcely influenced the expression of circSTX6 (Supplementary Figure S2A-D).Thus, we speculated that METTL14 might be the major m 6 A regulator of circSTX6 expression.Then we used SRAMP 22 and RMBase v 2.0 23 to predict potential m 6 A sites on circSTX6.Two potential m 6 A sites were obtained by overlapping the results (Figure 2B).Notably, one site was endowed with high confidence (purple) by SRAMP, and the other with moderate confidence (red).The corresponding primers (two pairs of primers for each site) were designed for subsequent MeRIP assays (Figure 2C).Consistently, m 6 A-modified circSTX6 RNA fragments were significantly enriched by anti-m 6 A antibody in MHCC97H, HCCLM3 and Huh7 cells (Figure 2D and Supplementary Figure S2E).Next, we performed RIP assays with anti-METTL14 antibodies, and the results showed a direct interaction between METTL14 and circSTX6 RNA (Figure 2E).Furthermore, dual-luciferase reporter plasmids were constructed by mutating the adenosines into cytosines on circ-STX6 sequences (Figure 2F).The results showed relative luciferase intensity increased upon METTL14 silencing in the wild-type groups (Figure 2G).Intriguingly, the overall luciferase intensity increased in mutant groups compared with wild-type groups but was not changed upon METTL14 silencing (Figure 2G), suggesting that METTL14-mediated regulation of circSTX6 was m 6 A dependent.Furthermore, MeRIP assays indicated that m 6 A enrichment of circSTX6 was reduced upon METTL14 knockdown (Figure 2H).Overall, these results demonstrate that the expression of circSTX6 is at least partially regulated by m 6 A in a METTL14-dependent manner.
To clarify the clinical relevance between m 6 A and circ-STX6, MeRIP was conducted on three pairs of HCC tumours and adjacent normal tissues.Notably, one in three tumour tissues demonstrated decreased m 6 A levels of circSTX6 compared with normal tissues (Supplementary Figure S2F).Additionally, the expression of METTL14 in HCC cohort-1 was determined by qPCR.The qPCR results revealed that METTL14 was negatively associated with circSTX6 (r = −0.452,p < 0.001; Figure 2I).In addition, the ISH staining of circSTX6 and IHC staining of METTL14 were performed based on HCC cohort-2 (Figure 2J).ISH and IHC score analyses demonstrated that around 61.5% of specimens that highly expressed METTL14 exhibited weak staining of circSTX6, while about 77.8% of specimens with low expression of METTL14 exhibited strong circSTX6 staining (Figure 2K).Taken together, circSTX6 is closely associated with m 6 A and is negatively correlated with METTL14 in HCC tumour tissues.

CircSTX6 accelerates the proliferation of HCC in vitro and in vivo
To investigate the biological roles of circSTX6 in HCC, a variety of functional assays were performed.The efficiency of knockdown through transient and stable transfection was validated in MHCC97H, HCCLM3 and Huh7 cells using qPCR.The results confirmed that circSTX6 was effectively silenced without affecting STX6 expression, indicating precise targeting of siRNAs and shRNAs at the junction motifs of circSTX6 (Figure 3A and Supplementary Figure S3A).As observed in the colony formation assay, CCK-8 assay and flow cytometry, genetic ablation of circSTX6 suppressed the proliferative capacities of tumour cells and intensified apoptosis (Figure 3B-F and Supplementary Figure S3B).What is more, EdU assays revealed a reduced DNA replication rate in HCC cells upon circSTX6 knockdown (Figure 3G,H).To further evaluate the impact of circSTX6 on tumourigenicity, subcutaneous implantation experiments were conducted.Significant slacked tumour growth along with reduced tumour volumes and weights were observed in circSTX6-knockdown groups (n = 8) compared to that of the control group (n = 8) (Figure 3I,J).Moreover, the IHC staining of tumour markers, including Ki67, PCNA, E-cadherin and N-cadherin, indicated that the loss of circSTX6 inhibited the proliferation and metastasis of HCC (Supplementary Figure S3C).
Conversely, a plasmid for circSTX6 overexpression was successfully constructed and transfected into MHCC97H and HCCLM3 cells, which led to a significant increase in circSTX6 expression, rather than STX6, as detected by qPCR (Supplementary Figure S4A).Colony formation and CCK-8 assays revealed that the upregulation of circSTX6 promoted the proliferation of HCC cells in vitro (Supplementary Figure S4B-D), and subcutaneous implantation experiments (n = 8 in each group) also showed accelerated tumour growth in vivo (Supplementary Figure S4E-G).Taken together, these findings demonstrate that circSTX6 accelerates HCC proliferation both in vitro and in vivo.

CircSTX6 promotes migration and invasion of HCC in vitro and in vivo
Wound healing and transwell assays were conducted to assess the effects of circSTX6 on HCC cell migration.The results demonstrated that the downregulation of circ-STX6 restrained the migration capacities of MHCC97H, HCCLM3 and Huh7 cells (Figure 4A-C).In addition, transwell assays also indicated compromised invasiveness of HCC cells with circSTX6 deficiency (Figure 4C,D).Next, to investigate the in vivo function of circSTX6 in promoting migration and invasion, tail vein assays of tumour metastasis experiments were performed.The number of lung metastatic nodules (n = 6 in each group) was significantly reduced with the knockdown of circSTX6, and the results were confirmed by HE staining (Figure 4E,F).In contrast, overexpression of circSTX6 notably promoted the migration and invasion of HCC cells, as demonstrated by wound-healing and transwell assays (Supplementary Figure S4H-K).In conclusion, circSTX6 promotes the migration and invasion of HCC both in vitro and in vivo.

CircSTX6 directly interacts with HNRNPD protein in HCC cells
Since previous studies have reported circSTX6 as a miRNA sponge, 24 we mainly focused on the circSTX6-protein interaction in HCC cells in this study.RNA pull-down assay was performed using biotinylated circSTX6 probes to identify potential interacted proteins of circSTX6.Silver staining revealed a differential band at approximately 40 kDa between the negative control (NC) and circSTX6 probe groups (Figure 5A).Mass spectrometry (MS) analysis identified HNRNPD as the protein with the highest Sequest Scores and most abundant unique peptides.Subsequent Western blot further confirmed the interaction between circSTX6 and HNRNPD (Figure 5B).Consistently, the RIP assay verified that circSTX6 could be enriched by anti-HNRNPD antibody, as shown by qPCR results (Figure 5C).Besides, the FISH/IF assay also identified colocalization of HNRNPD and circSTX6 in the cytoplasm of HCC cells (Figure 5D).
HNRNPD possesses two nonidentical RNA recognition motif (RRM) domains, which are crucial for high-affinity RNA binding. 25To determine the precise domain responsible for RNA binding, HNRNPD was truncated and plasmids for flag-tagged mutants were constructed.Western blotting verified the transfection efficiency of truncated proteins and their molecular weights (Figure 5E).Next, RIP assay was carried out and qPCR results showed that circSTX6 could be efficiently enriched by wild-type and RRM1 domain of HNRNPD but not the RRM2 (Figure 5F).Additionally, three biotinylated circSTX6 segment probes were designed and the RNA pull-down results indicated that circSTX6 segment 1 was responsible for RNA binding with HNRNPD (Figure 5G).Collectively, these results suggest a direct interaction between HNRNPD and circSTX6.
Next, we examined whether circSTX6 or HNRNPD affected the expression of each other.The qPCR and Western blot results demonstrated that circSTX6 had little effect on the expression of HNRNPD at either the mRNA or protein level and vice versa (Supplementary Figure 5H,I and S5A).Interestingly, previous studies have reported that HNRNPD can promote the degradation of mRNAs such as p21 (CDKN1A), cyclin D1 (CCND1) and p16 (CDKN2A), which are known to regulate the cell cycle. 26,27low cytometry was performed to assess whether circ-STX6 influenced the cell cycle in an HNRNPD-dependent manner.Notably, cell cycle regulation was not altered upon circSTX6 silencing (Figure 5J).Therefore, it is highly conceivable that the other downstream targets might be regulated by circSTX6 and HNRNPD.

CircSTX6 promotes ATF3 mRNA decay in an HNRNPD-dependent manner
To identify potential targets regulated by circSTX6 and HNRNPD, RNA-seq was performed with circSTX6-knockdown and negative control MHCC97H cells.As expected, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses showed significant enrichment in pathways related to cancer, focal adhesion, Phosphatidylinositol 3-kinase (PI3K)-Akt kinase (AKT) signalling and Mitogen-activated protein kinase (MAPK) signalling (Supplementary Figure S5B).Next, mRNA expression profiles from HCC tumour tissues based on the Gene Expression Omnibus database were employed (Figure 6A).A total of 618 differentially expressed genes (DEGs) were obtained from the intersection of two datasets (GSE102083 and GSE41804) (Figure 6A), and 15 candidate genes were subsequently screened out (Figure 6B).The expression of these candidate genes was examined by qPCR in circSTX6 knockdown or overexpressing HCC cells.ATF3 was identified as a potential downstream target of circSTX6, as its expression was consistently upregulated upon circSTX6 silencing, and downregulated upon circSTX6 overexpression (Figure 6C and Supplementary Figure S5C).The expression of ATF3 was further checked using two siR-NAs of circSTX6 in HCC cells (Figure 6D), and the result was consistent.Thus, we identified ATF3 as a potential downstream target of circSTX6.
Next, we sought to determine the relationship between HNRNPD and ATF3.The qPCR and Western blotting assays showed that knockdown of either circSTX6 or HNRNPD increased the expression of ATF3 at both mRNA and protein levels (Figure 6E-G).Moreover, overexpressing circSTX6 in HCC cells reduced ATF3 expression (Figure 6H and Supplementary Figure S5D).These results suggested that circSTX6 and HNRNPD work synergistically to regulate ATF3.Intriguingly, a previous study showed that HNRNPD could promote the decay of ATF3 mRNA by binding to it. 28Therefore, we wondered whether circSTX6 influences ATF3 expression by facilitating the HNRNPD-mediated ATF3 mRNA degradation and thus drives HCC malignancies.To verify this speculation, we first performed Western blot and qPCR.Results showed that the knockdown of HNRNPD retrieved the downregulation of ATF3 triggered by circSTX6 overex-

F I G U R E 5
CircSTX6 interacts with HNRNPD protein in HCC cells.(A) RNA pull-down assay was performed in MHCC97H cells using biotinylated circSTX6 and NC probes.Differential bands on the gel were visualized by silver staining and subsequently excised and sent for MS analysis.HNRNPD, with the largest unique peptide numbers, was chosen as the most potential binding protein of circSTX6.(B) The pression (Figure 6I,J), indicating that circSTX6 regulates ATF3 expression in an HNRNPD-dependent manner.Moreover, silencing of circSTX6 also rescued the downregulation of ATF3 resulting from HNRNPD overexpression (Supplementary Figure S5E), indicating the indispensable role of circSTX6 in the degradation of ATF3 by HNRNPD.Next, we conducted RNA stability assays to assess the mRNA decaying rate of ATF3 upon circSTX6 silencing.Consistently, inhibition of circSTX6 contributed to a longer half-life of ATF3 mRNA (Figure 6K).In contrast, overexpression of circSTX6 reduced the stability of ATF3 mRNA and led to a decreased half-life, while this effect was abolished by HNRNPD knockdown (Figure 6L).Subsequently, an RIP assay was performed to identify the binding of HNRNPD with ATF3 mRNA (Figure 6M).Notably, the loss of circSTX6 attenuated HNRNPD binding to ATF3 mRNA (Figure 6M), indicating that the conjugation of circSTX6 and HNRNPD increased the binding affinity of HNRNPD to ATF3 mRNA.Conversely, overexpression of circSTX6 increased the binding of ATF3 mRNA to HNRNPD (Supplementary Figure S5F).Collectively, circSTX6 sponges HNRNPD to facilitate its binding with ATF3 mRNA, thus reducing ATF3 mRNA stability.
Additionally, we examined the presence of other potential targets coregulated by HNRNPD and circSTX6 that play a part in HCC pathogenesis.0][31] Notably, Interleukin 8 (IL8) was found to be stably regulated by circSTX6, and it has been established as an important tumourigenic factor in HCC. 32,33Our findings suggest that the interaction of circSTX6 and HNRNPD promotes HCC progression through diverse pathways, and ATF3 represents only one of the downstream targets.Nevertheless, we mainly focused on the circSTX6-HNRNPD-ATF3 axis in this study.

Dysregulated circSTX6-ATF3 axis drives HCC malignancies
To determine whether circSTX6 drives HCC tumourigenesis and progression in an ATF3-mediated manner, a series of functional rescue assays were conducted.Colony formation and CCK-8 assays revealed that the loss of circSTX6 reduced the proliferation capacities of HCC cells, while silencing of ATF3 reversed this phenomenon (Figure 7A,B and Supplementary Figure S6A).Similarly, phenotypes of the EdU assay were rescued (Figure 7C and Supplementary Figure S6B).Furthermore, the inhibition of ATF3 partially attenuated the impaired capabilities of invasion and migration in HCC cells caused by circSTX6 knockdown (Figure 7D,E and Supplementary Figure S7A-D).In sum, dysregulated ATF3 might account for the circSTX6-driven malignant behaviours of HCC.
Additionally, the clinical correlation of the circSTX6-ATF3 axis was assessed.The expression of circSTX6 and ATF3 in HCC cohort-1 was determined by qPCR.Pearson correlation analyses showed a negative correlation between circSTX6 and ATF3 (r = −0.4741,p < 0.0001; Figure 7F).Moreover, we examined the expression of ATF3 in the HCC cohort based on the TCGA database (n = 369) and ATF3 was significantly downregulated in tumour tissues compared with adjacent tissues (Supplementary Figure S7E).Besides, the costaining of circSTX6 and ATF3 in HCC cohort-2 revealed that approximately 85.2% of the circSTX6 highly expressed tumour tissues weakly expressed ATF3, whereas 76.9% of circSTX6 weakly expressed tumour tissues showed stronger ATF3 staining (Figure 7G,H).Consistently, ROC curves showed that the AUC value of ATF3 was 0.7921 (p < 0.001; Figure 7I), suggesting ATF3 could be used to distinguish between HCC tumour and normal tissues.Therefore, the expression of circSTX6 and ATF3 is negatively correlated in HCC and the circSTX6-ATF3 axis could be a promising therapeutic target.
interaction of circSTX6 and HNRNPD was confirmed by circSTX6 pull-down assay in HCC cells using Western blotting.(C) RIP assays followed by qPCR verified the enrichment of circSTX6 by HNRNPD in HCC cells.The PCR products were subject to agarose gel electrophoresis.(D) FISH/IF assay showed co-localization of circSTX6 and HNRNPD proteins in the cytoplasm of HCC cells.Scale bar, 10 μm.(E) Plasmids of HNRNPD truncations (RRM1 and RRM2, labelled with a flag) were transfected into HEK-293T cells, respectively.(F) CircSTX6 was enriched using anti-flag or anti-IgG antibodies in each truncation group with RIP assay.The exogenous enrichment level of circSTX6 was determined by qPCR.(G) The sequences of circSTX6 were truncated into three segments.Specific probes targeting each segment were designed, and RNA pull-down was performed in HCC cells utilizing full-length (FL) probes or segment probes, respectively.(H) The expression of HNRNPD upon circSTX6 knockdown and circSTX6 upon HNRNPD knockdown was determined by qPCR in HCC cells, respectively.(I) The protein level of HNRNPD upon circSTX6 knockdown was determined by Western blot.(J) Flow cytometry was conducted to determine the impact of circSTX6 knockdown on the cell cycle in HCC cells.*p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

3.8
CircSTX6 encodes a novel 144aa protein, circSTX6-144aa, that serves a good prognostic biomarker Furthermore, we sought to determine the protein translation capacity of circSTX6.As predicted by circRNADb software, 34 circSTX6 has two potential internal ribosome entry sites (IRES) at 245-391 nt and 239-342 nt, respectively, and one open reading frame (ORF) that encodes a putative 144aa protein (Figure 8A and Supplementary Figure S8A).Moreover, the start codon ATG and ORF sequences formed by circSTX6 back splicing were verified in ORF Finder (https://www.ncbi.nlm.nih.gov/orfnder/) (Supplementary Figure S8B). 35Herein, we termed the 144aa novel protein as circSTX6-144aa.To determine which IRES drove the translation of circ-STX6, dual-luciferase assays were performed (Figure 8B), which demonstrated that IRES1 (245-391 nt) significantly increased the luciferase activity compared with IRES2 (239-342 nt) (Figure 8C).Driven by IRES1 (245-391 nt), circSTX6-144aa shares the same sequence with wild-type STX6 at 101−230 and contains unique sequences at the C-terminal (Figure 8D).Next, to confirm the existence of circSTX6-144aa, we transferred the circSTX6 overexpression plasmid and circSTX6 ATG mutant plasmid (circSTX6-MUT) into HCCLM3 cells.We performed silver staining, and the results showed a distinct band at approximately 16 kDa, consistent with the predicted molecular weight (Figure 8E).Then the target band was excised and sent for MS analysis.MS analysis detected the unique GHER sequence.
To specifically detect circSTX6-144aa, we constructed a custom-designed antibody targeting partial sequences of circSTX6-144aa (Supplementary Table S6).Western blotting was conducted to examine the presence of translated endogenous circSTX6-144aa.Notably, upon circSTX6 silencing, the expression of the parental gene STX6 was not altered (Figure 8F), whereas circSTX6-144aa was significantly reduced in MHCC97H, HCCLM3 and Huh7 cells (Figure 8G and Supplementary Figure S8C).In addition, we constructed a linear circSTX6-144aa overexpression plasmid.Western blot showed overexpressed circSTX6-144aa in circSTX6-OE and linear circSTX6-144aa groups, rather than the circSTX6-MUT group (Supplementary Figure S8D), indicating antibody specificity.Next, we sought to identify the expression of circSTX6-144aa in HCC tumour tissues.Interestingly, Western blotting showed that circSTX6-144aa was overexpressed in 32/38 (84.21%) tumour tissues compared to that of adjacent tissues (Figure 8H and Supplementary Figure S8E).Moreover, circSTX6-144aa was positively correlated with the expression of circSTX6 in 72 pairs of HCC tumour tissues according to qPCR (Figure 8I).To evaluate the clinical relevance of circSTX6-144aa, we performed IHC staining using the custom-designed antibody in HCC cohort-3 (Figure 8J).Consistently, overexpression of circSTX6-144aa was observed in HCC tumour tissues (Figure 8K) and its upregulation was associated with a worse overall prognosis (Figure 8L).Furthermore, the expression of circSTX6-144aa was considered an independent risk factor for OS of HCC patients (Figure 8 M).Taken together, these results indicate that circSTX6-translated 144aa could be a potential biomarker for HCC.

CircSTX6-144aa promotes HCC proliferation, migration and invasion independent of circSTX6
To further confirm the biological functions of circSTX6-144aa and verify whether these effects are independent of circSTX6 itself, a series of rescue assays were F I G U R E 6 CircSTX6 facilitates HNRNPD-mediated ATF3 mRNA decay.(A) Venn diagram showing the overlap of 15 candidate transcripts.RNA-seq data were obtained by sequencing the MHCC97H cells (si-circSTX6#2 vs. si-Ctrl), including 248 differentially expressed genes (DEGs, average FPKM > 10, fold change > 1, p < 0.05).DEGs (fold change >1, p < 0.05) in GSE102083 and GSE41804 were obtained from sequencing data of HCC tumour tissues, containing 1167 and 1461 transcripts, respectively.(B) Clustered heatmap presenting the expression of 15 genes between groups of si-circSTX6#2 and si-Ctrl MHCC97H cells.White and red bars indicated upregulation and downregulation, respectively.(C) The expression of 15 genes was preliminarily examined by qPCR, and only ATF3 was stably regulated by circSTX6 knockdown or overexpression.(D) The expression of ATF3 was examined by qPCR in circSTX6-silenced MHCC97H, HCCLM3 and Huh7 cells.(E) The expression of ATF3 was examined by qPCR in HNRNPD-silenced HCC cells.(F and G) The expression of ATF3 was examined by Western blotting in HNRNPD-or circSTX6-silenced HCC cells.(H) The expression of ATF3 upon circSTX6 overexpression was examined by qPCR and Western blotting.(I and J) The expression of circSTX6, ATF3 and HNRNPD were detected by Western blotting and qPCR assays in circSTX6/HNRNPD rescue groups.(K) Relative RNA levels of ATF3 were detected by qPCR at 0, 2 and 4 h in circSTX6-knockdown or control HCC cells treated with actinomycin D (10 μg/mL).(L) RNA levels of ATF3 were detected by qPCR at 0, 2 and 4 h after actinomycin D (10 μg/mL) treatment in circSTX6 overexpression, circSTX6 overexpression in combination with HNRNPD silencing and negative control HCC cell groups.(M) HNRNPD-RIP assay was performed in circSTX6-silenced or negative control HCC cells, followed by qPCR to detect the expression of ATF3 mRNA and to assess the impact of circSTX6 on the affinity of HNRNPD protein binding to ATF3 mRNA.*p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.conducted.As expected, overexpression of either circSTX6 or linear circSTX6-144aa significantly promoted the proliferation of HCC cells by colony formation and CCK-8 assays (Figure 9A-C) as well as migration and invasion by woundhealing and transwell assays (Figure 9D-G).Notably, the overexpression of circSTX6 with ATG mutant, which abrogates the translation of circSTX6-144aa, could also increase the proliferation, migration and invasion capacity of HCC cells (Figure 9A-G), indicating that circSTX6-144aa promoted HCC malignancies independent of circ-STX6 itself.Furthermore, we examined whether circSTX6-144aa enhanced the migration and invasion of HCC cells through epithelial-mesenchymal transition.The Western blot results showed that the expression of Vimentin, Snail and N-cadherin increased upon circSTX6-144aa overexpression, but E-cadherin decreased (Figure 9H).These results reveal that both circSTX6 and circSTX6-144aa play an oncogenic role in HCC.Based on our results, the potential mechanisms are illustrated in Figure 10.

DISCUSSION
CircRNAs are involved in tumour biology by acting as miRNA sponges, serving as protein scaffolds and encoding proteins, in which miRNA sponge is most widely reported.
Significantly, mounting research is paying attention to the other two mechanisms, especially protein translation, which can be driven by IRES or m 6 A modification on circRNAs. 20In our current study, using genome-wide mapping of m 6 A-immunoprecipitated circRNAs, we identified circSTX6 which, by coincidentally, has been reported to function as a miRNA sponge to promote the progression of pancreatic adenocarcinoma. 24However, the full range of functions of circSTX6 in HCC beyond miRNA sponge is still unclear.Our results confirmed an increased expression of circSTX6 in HCC, and its high expression was associated with advanced tumour stages, higher pathological grades and worse prognoses.In addition, lossand gain-of-function assays demonstrated that circSTX6 played an oncogenic role by enhancing cell proliferation, migration and invasion capabilities in vitro and in vivo in HCC.Thus, circSTX6 could be a candidate prognostic and therapeutic target in HCC.Over the years, the interplay between m 6 A modification and ncRNAs, especially circRNAs, has gained increasing attraction. 36METTL14, an auxiliary activator of METTL3, although less reported, has been shown to participate in circRNA regulation through m 6 A-dependent manners. 14o be exact, Fan et al. revealed that METTL14-mediated m 6 A modification on circORC5 decreases its expression and inhibits gastric cancer development. 19Consistently, our results showed that circSTX6 expression was regulated by METTL14 in an m 6 A-dependent manner and such regulation also existed in HCC tumour and adjacent normal tissues, which extended our knowledge of m 6 A-modified circRNAs in tumour progression.Nevertheless, m 6 A modification is a complex biological process that requires coordinated operations of many regulators, including 'writers' of METTL3, WTAP, 'readers' of IGF2BPs and YTHDFs and 'erasers' of ALKBH5 and FTO, and our study only focused on the interaction between METTL14 and circSTX6.It remains unclear whether other regulators (such as METTL3 and FTO) that catalyze methylation/demethylations on m 6 A sites regulate circ-STX6 in different manners (such as cytoplasmic export or splicing).Besides, emerging evidence has proved the encoding capacity of circRNAs, which could be driven by m 6 A in a variety of malignancies. 11,37,38For instance, circMAP3K4 encodes a 455aa polypeptide by means of IGF2BP1-mediated m 6 A modification and promotes HCC progression. 17In this study, we established that circSTX6 could be translated into 144aa polypeptide, circSTX6-144aa and it possesses good performance in prognostic prediction.In addition, circSTX6-144aa could promote tumour progression independent of circSTX6 itself, serving an oncogenic role in HCC.Nevertheless, current evidence only supports the initiation of circSTX6 translation by IRES (Figure 8C) instead of m 6 A modification (data not provided) and the detailed mechanisms underlying circSTX6-144aa and m 6 A warrant further exploration.
The functional role of circRNAs is closely associated with their subcellular localization. 39For example, cir-cRNAs located in the nucleus are usually involved in transcriptional regulation or splicing events 40,41 whereas cytoplasmic circRNAs frequently serve as miRNA sponges or protein scaffolds. 24,42In the current study, circSTX6 was predominantly located in the cytoplasm, consistent with the previous research.Therefore, using circRNA pulldown, MS analysis, RIP and FISH/IF, we identified HNRNPD as a robust interactor with circSTX6.HNRNPD, also known as adenylate-uridylate-rich element/poly(U)binding/degradation factor 1 (AUF1), encompasses four isoforms generated by alternative splicing of a precursor mRNA (pre-mRNA). 25,29The four protein isoforms (p37, p40, p42 and p45) of HNRNPD all share several common structural elements that include two non-identical RRM domains, two tandems and eight amino acid glutaminerich sequences. 25,43Nevertheless, there is no consensus about the precise subcellular localization of each isoform.This and the lack of HNRNPD isoform-specific antibodies make it difficult to determine which isoform associates with circSTX6 in detail. 44Since previous studies have revealed that the two RRM domains are necessary for high-affinity RNA binding, we constructed corresponding truncation mutants. 45The RIP assay confirmed that the RRM1 domain of HNRNPD mediated its binding to circ-STX6.Conversely, the precise RNA sequences required for binding were also truncated and the RNA pull-down revealed that truncation 1 showed the highest affinity.Hence, our study provided compelling evidence that circ-STX6 serves as a scaffold for HNRNPD, elucidated their interaction and broadened the understanding of binding motifs on circRNAs.However, it is worth noting that more stringent experiments considering circRNA secondary structure and protein conformation are warranted in future exploration.
ATF3 is a stress-induced transcription factor that belongs to the ATF/cyclin adenosine monophosphate response element-binding protein family. 46It has been reported to serve as a tumour suppressor in HCC.Chen and his colleagues revealed that ATF3 inhibited HCC progression through the transcriptional regulation of cysteinerich angiogenic inducer 61 (CYR61). 47Notably, ATF3 can be regulated by ncRNAs through diverse mechanisms, such as miRNA-mediated transcriptional inhibition and circRNA-guided competing endogenous RNA patterns. 48,49Although the interaction between ATF3 and circRNAs has been explored in a few studies, the underlying post-transcriptional mechanisms remain elusive.A canonical but crucial function of HNRNPD is controlling the mRNA decay. 50Notably, Wang et al. found that HNRNPD could antagonize ferroptosis by destabilizing the ATF3 mRNA. 51Mechanistically, HNRNPD promoted ATF3 mRNA decay by directly binding to its 3′-untranslated region.Furthermore, Pan et al. also established that HNRNPD regulated ATF3 expression under stress conditions in the HCC cell line HepG2. 28The RIP assay by Pan and his colleagues demonstrated decreased HNRNPD binding to ATF3 mRNA along with elevated ATF3 expression and declined mRNA decaying.Being consistent with previous studies, our research showed that circSTX6 served as an enhancer for HNRNPD-mediated ATF3 mRNA decay based on mRNA stability assays.Silencing of circSTX6 slowed the degradation of ATF3 mRNA.Moreover, the RIP rescue assay indicated attenuated HNRNPD binding to ATF3 mRNA upon circSTX6 knockdown.Overall, we enriched the current understanding of ATF3 metabolism through circRNA-RBP sponge mechanisms.
Nevertheless, there are still several limitations in this work that require further investigation.First, whether the biogenesis (backsplicing) or translocation of circSTX6 is controlled by m 6 A and in which way remains largely unknown.Additionally, we only validated one essential target (ATF3) of circSTX6 and HNRNPD, while other potential targets (such as IL8) exist and also take part in the progression of HCC.Furthermore, the oncogenic   mechanisms of circSTX6-144aa are not fully revealed compared to circSTX6 itself.

CONCLUSION
In conclusion, our study revealed the crucial roles of circ-STX6 in HCC tumourigenesis and progression.METTL14 regulated circSTX6 expression in an m 6 A-dependent manner.CircSTX6 enhanced HNRNPD-mediated ATF3 mRNA degradation, thereby promoting HCC growth and metastasis.Additionally, circSTX6 could be translated into a 144aa polypeptide, circSTX6-144aa, that served as a good biomarker and promoted HCC progression independent of circSTX6 itself.Clinically, circSTX6 was significantly upregulated in HCC tissues and high circSTX6 expression indicated an unfavourable prognosis for HCC patients.In addition, the expression of circSTX6 in tumour tissues was negatively associated with METTL14 and ATF3 but posi-tively correlated with circSTX6-144aa.Taken together, our study illustrated that circSTX6 and its encoded circSTX6-144aa are potential biomarkers for HCC diagnosis and prognosis.Targeting circSTX6 could be an effective strategy for HCC treatment.

A C K N O W L E D G E M E N T S
We thank the Home for Researchers editorial team (www.home-for-researchers.com) for the language editing service.
METTL14, ATF3 and circSTX6-144aa were determined and correlation analysis was performed.HCC cohort-2 consisted of 40 HCC patients.Tumour tissues were used for in situ hybridization (ISH) and immunohistochemistry (IHC) staining to explore the correlation between circ-STX6, METTL14 and ATF3, respectively, and for Western blotting to determine the expression of circSTX6-144aa in HCC.HCC cohort-3 consisted of 80 HCC patients, with complete clinicopathological information and follow-up data.Tumour tissues were also used for constructing HCC tissue microarray, ISH and IHC.The clinical relevance of circSTX6 and circSTX6-144aa was analysed separately based on HCC cohort-3.Patient anthropometrics and demographics for HCC cohort-1, HCC cohort-2 and HCC cohort-3 are provided in Supplementary Tables S1, S2 and S3, respectively.

F I G U R E 1
Identification, validation and clinical relevance of circSTX6 in HCC.(A) Venn diagram showing the overlap of m 6 A-IP circRNAs (85 transcripts), m 6 A-related circRNAs (515 transcripts) and upregulated circRNAs (82 circRNAs, fold change ≥2) in HCC.Hsa_circ_0007905 (circSTX6) was the only gene intersected by three datasets.(B) Schematic illustration of the constitution of circSTX6.(C) The back-splicing junction site was verified by Sanger sequencing in HCC cells.(D) Agarose gel electrophoresis using the PCR products of circSTX6 and GAPDH amplified by divergent and convergent primers, respectively.(E) The abundance of circSTX6 and STX6 mRNA was detected by qPCR after RNase R (3U/ug) or actinomycin D (10 μg/mL) treatment, respectively.(F) Subcellular distribution of circSTX6 was determined by qPCR with nuclear/cytoplasmic fractionation.GAPDH and U6 were chosen for cytoplasmic and nuclear internal references, respectively.(G) FISH assay shows predominantly cytoplasmic localization of circSTX6 (Cy3 labelled, RED).Scale bar, 8 μm.(H) Representative FISH images of circSTX6 (Cy3 labelled, RED) in HCC tumour tissues.Scale bar, 100 μm.(I) The expression of circSTX6 in 72 pairs of HCC (HCC cohort-1) tumour tissues and adjacent normal tissues was determined by qPCR and is shown in histograms.(J) The overall ISH staining and representative images of circSTX6 in HCC cohort-3 are shown.Scale bar in the overall image, 200 μm; scale bar in representative images, 50 μm.(K) Kaplan-Meier prognostic analyses show the OS (n = 80) of HCC patients based on circSTX6 expression.(L) The expression of circSTX6 was detected in different pathological grades by ISH.Scale bar, 50 μm.(M) Forest plot shows the hazard ratio (HR) and 95% confidence intervals (CI) of risk factors in HCC cohort-3 based on Cox multivariate analysis.CircSTX6 high expression and AFP levels are independent risk factors.*p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.
CircSTX6 is regulated by METTL14-mediated m 6 A modification.(A) Expression of circSTX6 upon METTL14 knockdown was detected by qPCR.(B) Potential m 6 A sites of circSTX6 were predicted by SRAMP and RMBase v 2.0 software.Two sites were obtained from the intersection, one at 347 nt with high confidence (purple) and the other at 363 nt with moderate confidence (red).(C) MeRIP primers were designed to target each site and span the junction.(D) MeRIP assays were conducted in HCC cells and measured by qPCR.Relative enrichment by m 6 A-antibody was normalized based on input.Results of agarose gel electrophoresis using PCR products are shown.(E) RIP assays followed by qPCR verified the enrichment of circSTX6 by METTL14 in HCC cells.(F) Schematic illustration of the construction of dual luciferase reporter assays.Sequences of circSTX6 were cloned into the vector, with (mutant group) or without (wild-type group) mutating the adenosines into cytosines.(G) Wild-type or mutant luciferase plasmids and siMETTL14 RNAs were co-transfected into HCC cells, followed by measurement of the luciferase activities.(H) MeRIP assay results detected by qPCR in METTL14-knockdown or control HCC cells.(I) Correlation analyses between circSTX6 and METTL14 based on qPCR results in HCC cohort-1 (n = 72).(J) ISH staining of circSTX6 and IHC staining of METTL14 were conducted in HCC cohort-2 (n = 40).Representative images of high-and low-expression of circSTX6 and METTL14 are shown, respectively.Scale bar, 100 μm.(K) Statistical analyses based on the IHC/ISH score show the percentages of HCC specimens with high-or low circSTX6/METTL14 expression.*p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

F I G U R E 3
Knockdown of circSTX6 inhibits HCC proliferation in vitro and in vivo.(A) qPCR analysis detecting the knockdown efficiency of circSTX6 in MHCC97H, HCCLM3 and Huh7 cells.(B) Colony formation assay was performed to detect the proliferation of HCC cells upon circSTX6 silencing.(C) Flow cytometry was conducted to evaluate the apoptosis of HCC cells upon circSTX6 knockdown.HCC cells were stained with Annexin V and propidium iodide.(D and E) The percentages of apoptotic cells and relative colony numbers are quantified in diagrams, respectively.(F) Growth curves in the CCK-8 assay show the proliferation of HCC cells transfected with circSTX6 siRNAs.(G and H) Representative images of EdU assays with corresponding quantified bar charts.Scale bar, 200 μm.(I) Representative images of subcutaneously implanted tumors established with MHCC97H and HCCLM3 cells in nude mice (n = 8).(J) Statistical analysis of the growth curves and tumour weight of subcutaneous tumours in different groups.*p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.*p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

F I G U R E 4
Ablation of circSTX6 suppresses HCC migration and invasion in vitro and in vivo.(A) Migration capacity of HCC cells upon circSTX6 knockdown was determined by wound-healing assays in MHCC97H, HCCLM3 and Huh7 cells.Scale bar, 200 μm.(B) Percentages of wound-healed areas are quantified in histograms.(C) Migration and invasion capabilities of MHCC97H, HCCLM3 and Huh7 were detected by transwell assays.Scale bar, 200 μm.(D) Relative count of migration and invasion cells is quantified in histograms.(E) Representative images and hematoxylin and eosin (HE) staining of metastatic nodules in different groups (n = 6).Scale bar, 100 μm.(F) Statistical analysis of metastatic pulmonary nodules in control and circSTX6-knockdown models.*p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

F I G U R E 7
Inhibition of ATF3 rescues the effects of circSTX6 knockdown.(A) Colony formation assay was performed to assess the proliferation of MHCC97H, HCCLM3 and Huh7 cells with either ATF3 inhibition or circSTX6 knockdown.(B) The colony numbers of the colony formation assay were quantified in histograms.(C) Statistical analysis of percentages of cells in the S phase in EdU assays.(D) Wound-healing assay was employed in either circSTX6 silencing or ATF3-deficient HCC cells.Scale bar, 200 μm.(E) Relative migration rate of HCC cells is exhibited.(F) Expression of ATF3 and circSTX6 in HCC cohort-1 (n = 72) was detected by qPCR.Their expression was analysed with Pearson correlation analysis.(G) IHC staining of ATF3 and ISH staining of circSTX6 were respectively performed based on HCC cohort-2 (n = 40).Representative images of cricSTX6 with varying degrees of dyeing intensity and corresponding ATF3 images in the same sample are shown.Scale bar, 100 μm.(H) The percentages of high-or low-expression circSTX6 or ATF3 in HCC specimens are displayed in histograms.(I) Diagnostic ROC curve based on the IHC score of ATF3 in HCC cohort-2 (AUC = 0.7921, p < 0.001).*p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

F I G U R E 8
CircSTX6-144aa is a novel protein encoded by circSTX6.(A) Schematic illustration of potential IRES sequences (at 245-391 and 239-342 nt, respectively) and ORF sequences on circSTX6.(B) Schematic illustration of the construction of dual luciferase reporter plasmids.IRES sequences were cloned between the Renilla and firefly luciferase genes with AUG and UGA codons.(C) Luciferase activity was detected in HCC cells transfected with IRES1 or IRES2 plasmids, respectively.(D) Schematic illustration of STX6 and circSTX6-144aa peptide sequences.The unique peptides of circSTX6-144aa were marked red.(E) HCCLM3 cells were transfected with circSTX6 or circSTX6 with ATG mutant plasmids, respectively; silver staining showed differential bands, and the gel was excised and sent for MS analysis.The identified polypeptide contained unique sequences of circSTX6-144aa.(F) The expression of STX6 in circSTX6-knockdown or negative control HCC cells.(G) Expression of circSTX6-144aa in circSTX6 knockdown or negative control HCC cells.(H) Expression of circSTX6-144aa was detected in 18 pairs of HCC tumour and adjacent normal tissues.(I) Expression of circSTX6-144aa in HCC cohort-1 was determined by qPCR (n = 72).Correlation analyses between circSTX6-144aa and circSTX6 were performed.(J) The overall and representative images of IHC staining of circSTX6-144aa.Scale bar in the overall image, 200 μm; scale bar in representative images, 50 μm.(K) IHC scores in 73 pairs (HCC cohort-3) of HCC tumour and adjacent tissues were compared.(L) Kaplan-Meier prognoses analysis showing the OS (n = 73) and RFS (n = 73) of HCC patients based on circSTX6-144aa expression.(M) Forest plot shows the hazard ratio (HR) and 95% confidence intervals (CI) of risk factors in HCC cohort-3 based on Cox multivariate analysis.High expression of circSTX6-144aa and AFP levels were independent risk factors.*p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

F I G U R E 9
CircSTX6-144aa promotes HCC proliferation, migration and invasion independent of circSTX6.(A) Colony formation assay to evaluate the proliferation capacity of HCC cells with overexpression of circSTX6, circSTX6 with ATG mutant and linear circSTX6-144aa.(B) Colony numbers are demonstrated in histograms.(C) CCK-8 assay to evaluate the effect of circSTX6-144aa on the proliferation of HCC cells.(D-G) Wound healing and transwell assays to evaluate the effect of circSTX6-144aa on migration and invasion of HCC cells.The percentage area of wound healed and migration cell numbers are demonstrated by histograms.(H) The protein levels of N-cadherin, E-cadherin, Snail and Vimentin in HCCLM3 and MHCC97H cells were detected by Western blot.*p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

F I G U R E 1 0
Proposed model illustrating the underlying mechanisms of circSTX6 in HCC.METTL14-mediated m6 A modification on circSTX6 inhibits its expression.CircSTX6 promotes the proliferation and metastasis of HCC by binding to HNRNPD to facilitate HNRNPD-mediated ATF3 mRNA degradation.Driven by IRES1, circSTX6 encoded a 144 amino acid novel polypeptide, circSTX6-144aa, which promotes HCC progression independent of circSTX6 itself.
The authors appreciate the kind help from senior pathologist Danjing Guo and flow cytometer expert Rong Su.This work was supported by the Natural Science Foundation of China (82073144), Huadong Medicine Joint Funds of the Zhejiang Provincial Natural Science Foundation of China (LHDMD22H310005), the Science Technology Department of Zhejiang Province (2023C03063), the Health Commission of Zhejiang Province (JBZX-202004), and the Research Unit Project of Chinese Academy of Medical Sciences (2019-I2M-5-030).