Epigenetic Dysregulation of the Expression of PRSS3 Splice Variants Increases the Heterogeneity of Transcripts and Functionality in Human Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC) is one of the most lethal human tumors with extensive heterogeneity. Serine protease 3 (PRSS3) is an indispensable member of the trypsin family and has been implicated in the pathogenesis of several malignancies including HCC. However, paradoxical effects of PPRSS3 on carcinogenesis impede the utilization of its biomarker potential. We hereby systematically dissected the expression of four known splice variants of PRSS3 (PRSS3-SVs) and their functional relevance to HCC. The expression and DNA methylation of PRSS3 transcripts and their associated clinical relevance in HCC were analyzed using several publicly available datasets and were validated using qPCR-based assays. Functional assays were performed on gain- and loss-of-function cell models, in which PRSS3 transcript constructs were separately transfected after PRSS3 expression was knocked out by CRISPR-Cas9 editing.

Human primary liver cancer is one of the most lethal tumors with a dismal prognosis, featuring extensive heterogeneity and aggressiveness in the context of genetic and epigenetic aberrations [1][2][3][4][5]. Regardless of many approaches developed for the management of liver cancer in the past decade, its incidence and mortality rate continue to increase worldwide [1]. Liver hepatocellular carcinoma (HCC or LIHC) accounts for approximately 75-85% of all primary liver cancers. Most HCCs (> 90%) develop from chronic in ammation-induced liver cirrhosis contributed by multiple risk factors such as hepatitis viruses, alcohol consumption, and non-alcoholic fatty liver disease, which trigger the molecular complexity of intratumor heterogeneity (ITH) increasing HCC phenotypic diversity and therapeutic resistance [1,2].
Large-scale bioinformatics datasets generated with next-generation sequencing technologies reveal a comprehensive landscape of genomic and epigenetic heterogeneity among HCC cell lines and tissue specimens [3][4][5][6][7]. These studies offer invaluable insight into the molecular basis of ITH to categorize HCC into proliferative and non-proliferative subclasses in favor of integrative molecular monitoring malignant transformation and management of HCC. However, aside from most genetic alterations occurred in passenger genes that may be associated with aging and pollution, most genetic variants such as driver mutations in TP53, TERT and CTNNB1 detected in HCCs are not clinically relevant, or are not potentially targetable for the existing drugs [2]. This gives rise to a growing drive to integrate non-genetic variations into ITH, and to distinguish between functional and non-functional ITH [7,8]. Pre-mRNA alternative splicing (AS), as a key co-and post-transcriptional process drives non-genetic phenotypic heterogeneity, disruption of which generates aberrant splice variants (SVs) that contributes to ITH and functional divergence, thus functionally important to carcinogenesis and oncotherapeutics resistance [9][10][11][12].
As SVs emerge as new candidates for diagnostic and prognostic biomarkers and therapeutic targets [10,11], we systematically investigated the expression and epigenetic alteration of PRSS3-SVs functionally in relation to HCC heterogeneity. We found that differential expression of PRSS3 in HCC was attributed to aberrant expression of divergent PRSS3-SVs, which was epigenetically dysregulated by site speci c abnormal CpG methylation. We also found different functionality and clinical relevance of PRSS3-SVs in HCC cells and tissues. Therefore, epigenetic dysregulation of expression of PRSS3-SVs may be the molecular basis of PRSS3 to exert paradoxical effects on hepatocarcinogenesis.

Cell lines
Human HCC cell lines include well differentiated (HepG2 and Huh7) and poorly differentiated (SK-Hep-1, SMMC-7721 and LM3) cell lines and were authenticated by STR pro ling. The origin and growth conditions of all cell lines used in the study were described previously [35,42]. The cells were split to low density (30% con uence) for overnight culture, and were then treated with 5 μM of 5-aza-2'-deoxycytidine (5-aza-CR, Sigma-Aldrich, St Louis, MO, USA) for 96 hours with the medium exchanged every 24 hours.
Colony formation HCC cells were seeded in 6-well tissue culture plates (100 cells/well) in triplicate. Colonies with more than 50 cells were counted after 2 weeks. The cells were xed with 75% ethanol for 30 minutes and stained with 0.2% crystal violet (Beyotime Ltd., Jiangsu Province, China) for 20 minutes.

Transwell invasion assay
Transwell apparatus was used with 8-μm polyethylene terephthalate membrane filters (Corning Inc.; Corning, NY, USA). The upper chambers were seeded with 200 µl of serum-free medium containing 1 × 10 4 of serum-starved cells. The lower chambers were lled with 500 µl of 10% FBS-DMEM. After 24 hours, cells that invaded to the lower chamber were xed and stained with 0.2% crystal violet (Beyotime) as previously described [35].

RNA isolation and RT-qPCR
Cells were harvested for RNA isolation using RNeasy Mini Kit (QIAGEN) and rst strand cDNA was synthesized with the Superscript First-Strand Synthesis System (Invitrogen). RT-qPCR was performed using primers as described [35]. The relative expression level of each mRNA was normalized by βactin using 2 -ΔΔCt method Methylation-speci c qPCR (MS-qPCR) DNA extraction, bisul te modi cation and MSP-PCR were performed as described [35,43]. Genomic DNA was extracted from tissues using the QIAamp DNA mini Kit (Qiagen) followed by quantitative analysis using NanoDrop 1000 spectrophotometer (Thermo Fisher Scienti c, Inc.). Bisul te modi cation of DNA was performed using Zymo DNA Methylation Kit (Zymo Research). The positive and negative template control were the Human Methylated & Non-methylated DNA Set (Zymo Research). MSP-qPCR was performed by using methylated or unmethylated primer pairs speci cally for PRSS3 [35] and β-actin [43].
The relative level of methylation and unmethylation of PRSS3 was normalized to β-actin using the 2-ΔΔCq method.

Statistical analysis
The data are expressed as means ± standard deviation (SD) of at least three independent experiments. PRSS3 transcripts expression, epigenetic alterations and associative clinicopathological correlation were analyzed by using the two-tailed Student's t-test, Wilcoxon rank sum test, one-way analysis of variance (ANOVA) with Tukey's post hoc test, Spearman rank test and Fisher's exact test, or χ 2 or Fisher's exact tests. Cancer-related survival was analyzed using Kaplan-Meier method, and was compared using logrank tests. Statistical significance was considered when P < 0.05. All statistical analyses were performed using SPSS version 23.0 (IBM Corp.).

Results
Differentially decreased PRSS3 expression in HCC attributable to disruption of PRSS3 transcripts The RNAseq data from the DepMap portal demonstrated differential expression of PRSS3 in 24 human HCC cell lines (Fig. 1A, Table S1). This was further shown in 81 HCC cell lines from the Cancer Model Repository (LIMORE) ( Figure S1A). PRSS3 expression in HCC cell lines determined by RT-qPCR revealed levels of very low (PRSS3 Low ) to very high (PRSS3 High ) as compared to human fetal liver L02 cells ( Fig. 1B). Comparative analysis using the TCGA RNAseq data from FIREHOSE Broad GDAC showed divergent features s of PRSS3 expression in HCC tissues compared to their matched non-tumor tissues (n = 50) (Fig. 1C, Table S2), which was further evidenced from analysis of more HCC tissue specimens (Tumor = 371) (Fig. 1D, Table 1). GEPIA portal combined TCGA with GTEx RNAseq datasets showed that PRSS3 expression was more varied in HCC tissues (n = 369) than in normal controls (n = 160) ( Figure  S1B) [37,41], albeit with no statistical signi cance. PRSS3 mRNA level was lower but with an extraordinary wide range in HCC tissues as compared to normal tissues, suggesting an aberrant and differential expression of PRSS3 expression in HCC. To explore the molecular basis of the divergent expression of PRSS3 in HCC, we dissected the expression of four identi ed PRSS3-SVs expressed in various tissues [15][16][17][18][19][20] (Fig. 2A). Analysis of the DepMap data revealed that in 24 HCC cell lines, PRSS3-V1 and -V2 were two major transcripts contributing to the expression of PRSS3 because PRSS3-V3 was poorly expressed while PRSS3-V4 was absent (Fig. 2B, Table S1). RT-qPCR showed that despite almost undetectable PRSS3-V4 and very low expression of PRSS3-V3 in all cell lines, PRSS3-V1 was expressed at low levels in L02 cells, whereas PRSS3-V1 and -V2 were minimally expressed in HepG2, SK-Hep-1 and SMMC-7721 cells but highly expressed in Huh7 and LM3 cells (Fig. 2C).
Through analysis of the expression level of PRSS3-SVs in 50 paired tissue samples, we found that PRSS3-V2 and also -V1 were predominantly present in both normal and tumor tissues (Fig. 2D, Table S2). Figure 2E showed that the expression of PRSS3-SVs was decreased in 371 HCC tissue samples in a bipolar pattern as compared to normal liver tissues. Co-expression analysis of both HCC cell line and tissues summarized in Table 2 further showed that the highest contribution of PRSS3-SVs to PRSS3 High was co-expressed PRSS3-V2 and -V1, not by either PRSS3-V2 or -V1 in the specimens, con rming PRSS3-V2 was the predominant transcript in PRSS3 High in HCC. Meanwhile, PRSS3 Low was also resulted from decreased expression of PRSS3-V2 and/or -V1 because the minimally expressed PRSS3-V3 minimally affected the eventual expression of PRSS3 despite PRSS3-V3 Low most frequently associated with PRSS3 Low in HCC. These results thereby revealed disruption of PRSS3 transcripts towards a bipolar expression contributing to aberrant and differential expression of PRSS3 in HCC, in which PRSS3-V2 was a dominant transcript leading to PRSS3 expression. We next assessed the contribution of DNA methylation to the expression of PRSS3-SVs based on the data available from the DepMap and the FIREHOSE [39,40] for three genomic regions in PRSS3. These were referred to as extended promoter region with 17 CpG sites de ned as CpG 1-17, upstream (-1000 bp upstream from the TSS of PRSS3-V1/V3 covering 2-7 of the 17 CpG sites) and extensive fragment containing 6 CpGs (CpG A-F) (Fig. 3A, Table S4). Association analysis demonstrated an inverse association between the upstream methylation and mRNA expression of PRSS3 and its transcripts PRSS3-V1, -V2 that were distinct between hypermethylation of PRSS3 Low (mPRSS3 Low ) and hypomethylation of PRSS3 High (umPRSS3 High ) in HCC cell lines (Fig. 3B) and tissues (Fig. 3C).
Given that epigenetic silencing of PRSS3 has been observed in several types of human cancer [32][33][34][35], we investigated DNA methylation in the extended promoter region of PRSS3 upon the data availability ( Fig. 3A). Unsupervised clustering combined with spearman correlation analysis of methylation states and expression of PRSS3 transcripts in HCC cell lines revealed that among 17 CpGs distributing in the extended promoter region, methylation occurred at CpG site 5-17 (-89 ~ 653 bp from the TSS of PRSS3-V1/V3) was reversely correlated with the mRNA expression level of PRSS3-V1, while methylation at CpG site 12-16 (522 to 564 bp to PRSS3-V1 TSS) was highly related to PRSS3-V2 expression (Fig. 3D, Figure  S2, Table S5). No associative comparison was conducted on PRSS3-V3 and -V4 due to their rare expression in HCC. Moreover, despite a positive association shown in CpG site F, methylation at CpG site A-E was negatively correlated with PRSS3 expression (Fig. 3E, Figure S3). CpG site methylation at the extensive fragment of PRSS3 was decreased at site A, increased at B, C and D, and then decreased at E and F in HCC tumors as compared to normal controls (Fig. 3F). These data suggest that an extended promoter region was important for epigenetic regulation of PRSS3 transcripts.
We then examined methylation-speci c effect on PRSS3 expression using qPCR-based assays (Fig. 3A).
MSP-qPCR showed hypermethylation in PRSS3 Low cell lines (L02, HepG2, SK-Hep-1) in contrast to hypomethylation in PRSS3 high Huh7 cells (Fig. 3G). Figure 3H revealed that treatment with DNA methyltransferase inhibitor 5-aza-CR caused signi cant upregulation of PRSS3 expression in PRSS3 Low cell lines, but had no effect on PRSS3 High Huh7 cells. Notably, a bipolar expression pattern was observed in PRSS3 Low cell lines upon 5-aza-CR treatment showing signi cant upregulation of PRSS3-V1 and -V3 opposite to downregulation of PRSS3-V2, eventually integrative to the upregulation of PRSS3, whereas the treatment had no effect on PRSS3 high Huh7 cells, actually due to integration between upregulation of PRSS3-V2 and downregulation of PRSS3-V1 or -V3. MeDIP-qPCR further showed that anti-5methylcytosine (5-mC) antibody enriched signi cantly less genomic DNA fragments in HepG2 cells but not in Huh7 cells upon 5-aza-CR treatment (Fig. 3I), suggesting that 5-aza-CR was effective in the expression of PRSS3 speci cally by altering the DNA methylation in this target region. Although the expression of PRSS3-V3 in L02 or PRSS3-V2 in HepG2 and SK-Hep-1 cells was too low to be taken into account its decreased signi cance level, these results consistent with bioinformatic analysis of HCC cell lines and tissues, as well as our previous report [35], suggest that methylation occurring at this region is more critical for epigenetically controlling PRSS3 transcript activities in HCC. As shown in the summarized table (Fig. 3J), PRSS3-SVs were divergently expressed and response to 5-aza-CR treatment associated with site-speci c CpG methylation that eventually determined the expression level of PRSS3 as a whole, suggesting that epigenetic dysregulation of the expression of PRSS3-SVs by site-speci c CpG methylation may mediate their functional differences in HCC.
PRSS3-V2 exerts oncogenic functions distinct from tumor-suppressive effects of PRSS3-V1 and -V3 in HCC cells The functional role of PRSS3-SVs was assessed by transfecting PRSS3-V1 to -V4 respectively into PRSS3 Low HepG2 and SK-Hep-1 cells (de ned as V1 to V4) (Fig. 4A). MTT assays showed that ectopic expression of PRSS3-V1 or -V3 signi cantly inhibited HCC cell proliferation in contrast to notably enhancing effect by ectopic PRSS3-V2 expression, or in addition to unfunctional PRSS3-V4 in HCC cell proliferation as compared to the vector controls (Fig. 4B). Moreover, the results of clone formation assay showed that overexpression of PRSS3-V1 or -V3 remarkably diminished the number of colonies of HCC cells compared with the control group, but PRSS3-V2 overexpression resulted in an increased number of colonies only effectively in HepG2 cells. However, ectopic PRSS3-V4 signi cantly reduced clone formation in SK-Hep-1 cells but had no effect in HepG2 cells (Fig. 4C). Transwell assay further showed an inhibitory effect of PRSS3-V1 or -V3 on HCC cell migration, opposite to PRSS3-V2 that showed an enhanced effect in the cells (Fig. 4D). These results suggest a tumor-suppressive effect of PRSS3-V1/V3 versus an oncogenic effect of PRSS3-V2 in HCC cells.
To further de ne the phenotypic properties of PRSS3-SVs in HCC cells, we established a PRSS3 KO + V cell model, in which each PRSS3 transcript construct was separately transfected after endogenous PRSS3 was knocked out through CRISPR/Cas9 system (Fig. 5A). RT-qPCR showed that all the detected PRSS3 transcripts were effectively knocked out and their constructs were stably expressed in Huh7 cells, respectively designated as PRSS3 KO + V1 to PRSS3 KO + V4 , or the vector control (PRSS3 KO + C ) (Fig. 5B). Functional assays as shown in Fig. 5C to 5E revealed that PRSS3 knockout in Huh7 cells facilitated cell proliferation, colony formation and migration, which were abolished by re-expression of PRSS3-V1 or -V3. Ectopic re-expression of PRSS3-V2 augmented the PRSS3-knockout effects on cell proliferation, colony formation, and remarkably, on migration of PRSS3 KO Huh7 cells. Unexpectedly, PRSS3-V4 re-expression did not affect Huh7 cell proliferation but resulted in signi cantly inhibition of PRSS3 KO Huh7 cell activity. These results demonstrate dual roles of PRSS-SVs in HCC cells and divergent disruption of PRSS3 transcripts may be integrated to establish their functional heterogeneity in HCC cells.
Epigenetic alteration of PRSS3-V2 is associated with clinical relevance in patients with early HCC To further explore the contribution of PRSS3 transcripts to tumor heterogeneity, we used TCGA dataset to analyze their clinical relevance. We found that the expression of PRSS3 and PRSS3-V2 was similarly downregulated but with a gradually increased tendency in HCC tumors compared with control tissues, following the progression of tumors stages (Fig. 6A) and pathological grades (Fig. 6B), in which PRSS3-V2 Low was signi cantly detected in tumors of early HCC patients in contrast to PRSS3-V2 High in advanced tumors. Kaplan-Meier (K-M) analysis revealed that PRSS3-V2 Low was a favorable factor for overall survival of HCC patients based on cancer stages (Fig. 6C) and grades (Fig. 6D), in which PRSS3-V2 Low patient groups with low-grade tumors showed signi cantly favorable outcome (P = 0.011). Moreover, divergent disruption of CpG site methylation (A to F) was shown throughout clinical progression of tumors but occurred more frequently and signi cantly in tumors of HCC patients with early-stage (Fig. 6E) and lower-grade tumors (Fig. 6F). In such tumors alteration in CpG methylation at site D was most reversely correlated with the expression of PRSS3 and PRSS-V2. Since the region located at site D was shown as an important regulatory region speci cally for epigenetic regulation of PRSS3 transcripts (Fig. 3), the data suggest that site-speci c epigenetic alteration of PRSS3-V2 in HCC tissues was distinct between mPRSS3-V2 Low in early HCC and umPRSS3 High in advanced HCC patients, in which early HCC patients with PRSS3-V2 Low tumors had better outcomes.

Discussion
Paradoxical effects of many genes have been observed during tumorigenesis [13,44,45]. In this study we explored the dysregulation of splicing variants expression functionally contributing to HCC heterogeneity. Protease PRSS3 is the rst to link the enzyme to prostate cancer leading to the development of a compound to stop PRSS3 from promoting metastasis [13,46]. Since the high similarity in both sequences and structures to different trypsinogen isoenzymes made it di cult to delineate their functionally associated transcripts distributed in different tissues [13,14], the protumor [21][22][23][24][25][26][27][28][29][30][31] or antitumor properties of PRSS3 [32][33][34][35][36] were deciphered relying on the cellular source and cancer microenvironment [13,14,35]. In this study, we found differentially expressed PRSS3 in HCC due to CpG methylation-mediated epigenetic dysregulation of its splice variants. Different PRSS3-SVs expressed in HCC showed a dual role in hepatocarcinogenesis that may increase phenotypic diversity. Our study uncovered an epigenetic-mediated PRSS3 transcript variance contributing to non-genetic phenotypic diversity of HCC [44]. To our best knowledge, this is the rst study of functional dissection of the expression of PRSS3-SVs in cancer and thus has important implications in HCC patient-tailored management.
PRSS3 is known as a digestive protease with restricted expression in pancreas. However, the preferential expression of PRSS3-SVs differs in human tissues suggests a tissue-selective expression manner. For instance, PRSS3-V2 was exclusively expressed in human pancreatic tissue and uid encoding MTG [16,47]. Canonical PRSS3-V1 was originally identi ed in human brain [17,47]. PRSS3-V3 shares a same TSS with PRSS3-V1 but has a different in-frame exon with deduced a 261-amino acid sequence (formerly named isoform B) [19]. PRSS3-V4 was cloned from keratinocytesand shown in participating keratinocyte terminal differentiation [20]. Our study showed the differential expression of PRSS3 as a DEG in HCC across a large expression range that could be used to phenotypically distinguish between PRSS3 Low and PRSS3 High HCC cells and tissues. Accordingly, we found divergent expression of PRSS3-SVs towards bipolarity following clinical progress from downregulation in early HCC to upregulation in advanced cancer, unveiling the molecular basis of PRSS3 in tissue-selective expression of its splice transcripts in HCC. Despite the infrequent or minimal expression of PRSS3-V3 and unexpressed PRSS3-V4, the divergent expression changes of PRSS3-V2 and/or -V1 were major contributors to the transcript heterogeneity of PRSS3 in HCC. Notably, the expression of PRSS3-SVs was dynamically altered following clinical progress from downregulation in early HCC to upregulation in advanced cancer. PRSS3 transcript heterogeneity was further evidenced by its divergent responses to 5-aza-CR treatment of HCC cells, distinguishing between upregulation of PRSS3-V1 or -V3 but downregulation of PRSS3-V2 in PRSS3 Low HCC and downregulation of PRSS3-V1 but upregulation of PRSS3-V2 in PRSS3 High HCC. The divergent expression of PRSS3 transcripts and their responding to 5-aza-CR prompted our consideration of the effects of non-genetic heterogeneity on chemotherapy-response, because this well-known anticancer drug has broad clinical applications and the mis-splicing regulation as non-genetic mechanisms is frequently linked to therapy escape [48][49][50]. For precise evaluation of the clinical effectiveness and drug resistance by using a DEG, its functional splice variants, rather than its overall expression, need to be taken into account. Nevertheless, it was clear that differentially expressed PRSS3 decreased as a whole was mainly attributable to its aberrant transcript variance expressed in HCC.
PRSS3 translocates from chromosome 7q34, the locus of PRSS1 and PRSS2, to chromosome 9p11.2, a region frequently containing alterations [13,51]. However, frequent genetic variations occurred in PRSS3 have not yet demonstrated disease-associated PRSS3 variants (https://www.nextprot.org/entry/NX_P35030/medical). Alternative splicing forms dynamic interactome offering precise therapeutic approaches to correcting cancer-speci c defects caused by mis-splicing regulation, in which epigenetics plays an essential role [9,11,12,49,[52][53][54]. Our previously study showed epigenetic silencing of PRSS3 in HCC [35], we reasoned epigenetic regulation of PRSS3-SVs contributing to non-genetic heterogeneity in HCC. The different TSSs and start codes in PRSS3 suggest that PRSS3, like the majority of protein-coding genes, tends to be regulated by multiple or alternative promoters, the usage of which provides a pre-transcriptional control of gene activity to express its different isoforms in a tissue-speci c manner [4,6,11,25]. Here, we found an extended promoter region covering the upstream and the intragenic region of PRSS3-V1/V3 and -V4, providing a site-speci c way to regulate the expression of PRSS3-SVs. Both HCC cells and tissues were phenotypically classi ed as mPRSS3 Low and umPRSS3 High based on CpG methylation in association with expression of PRSS3 transcripts. Compared to the consistent upstream hypermethylation, site-speci c CpG methylation in the intragenic region was found more associated with the expression of PRSS3-V1 and V2, suggesting that this extended promoter region played a central role in regulation of both PRSS3-V1 and V2. Given that epigenetic promoter alterations can change chromatin accessibility of transcription regulatory elements binding to transcription factors [9,12,44,52,[55][56][57], the upstream hypermethylation of PRSS3 may impact tissue-speci c cis-regulatory modules that may alter transcription activity of PRSS3-SVs in HCC. Dynamic disruption of different CpG site methylation within the extended promoter region may affect certain transcriptional regulators or splicing factors occupancy, resulting in an alternation in exon skipping to control the expression of PRSS3-V1 or -V3. Meanwhile, site-speci c epigenetic control of PRSS3-V2 suggests that the extended promoter may be a distal regulatory region in regulation of PRSS3-V2 through a very different epigenetic pathway [58]. Consistent with this, epigenetic silencing of PRSS3 was found in several cancer types and our previous study showed an intragenic DNA methylation within the extended promoter region contributing to PRSS3 downregulation in HCC [35]. This study was the rst to dissect epigenetic heterogeneity in regulation of PRSS3-SVs that may provide important implications for understanding epigenetic contributions to the genomic occupancy of transcription factors during transcription, in which many events may appear to be co-spliced with distant events[53, 55-57] .
Many transcript isoforms can exist per gene [9][10][11], most of which are thought not to be functionally relevant [59]. Recently, comprehensive gain-and loss-of function works had shown the functional importance of SVs in tumor heterogeneity by linking genetic variants to individual's phenotypes [52-54, 60, 61]. PRSS3 appears to be transcribed differentially to display heterogenous functions in cancer, in which a dual role or contradictory effects reported might be due to MTG (PRSS3-V2) to be functionally regarded as PRSS3 [13,14,21,22,24]. We hereby deciphered a functional difference among the PRSS3 isoforms by using constructed Huh7 cell model. Despite PRSS3-V2 /MTG-mediated an oncogenic effect in HCC in line with the pro-malignancy activities of MTG shown in other cancer types [13,14,21,22,24], PRSS3-V1 or -V3 were found as tumor-suppressors in HCC cells, while ectopic PRSS3-V4 showed an inhibitory effect on the PRSS3 ko Huh7 cells. PRSS3 ko resulting in pro-tumor effects in Huh7 cells suggests a tumor-suppressive role of PRSS3 played in HCC that was attributed to the co-expressed PRSS3-V1 and -V2, the two isoforms with opposite functionality. This is in line with our previous observations [35] and may explain some but not all cases of a similar phenotype with well-differentiated and/or low metastatic potential appearing in either PRSS3 Low (e.g. HepG2, SK-Hep1 cells) or PRSS3 High (Huh7 cells) live cancer cell lines, or a dual role of PRSS3 contradictorily shown in carcinogenesis. To support this, corresponding clinicopathological analysis of HCC specimens compared to the normal tissue controls uncovered that PRSS3-V1 and -V2 were main functional components of clinical relevance since PRSS3-V1 and -V2 were bipolarly presented in both PRSS3 Low or PRSS3 High tissues, thereby their abnormal co-expression could bring out functional heterogeneity including insigni cant or paradoxical clinical association. However, a signature pattern of epigenetic regulation of PRSS3 expression by site-speci c CpG methylation showed dynamically from mPRSS3 Low to umPRSS3 High through clinical progression, better matched to PRSS3-V2, suggesting PRSS3-V2 to be a more prevalent isoform functionally through clinical progression of HCC.
Accordingly, signi cant epigenetic downregulation of PRSS3-V2 was seen in early HCC with favorable patient outcome. This supports an oncogenic role of PRSS3-V2/MTG dominantly in HCC thus providing early diagnostic and prognostic value for HCC [14,21,22,24]. Thus, our study provides additional evidence for supporting the hypothesis of functional hepato-heterogeneity attributed to genetic and epigenetic factors [3][4][5][6].
Aberrant expression of SVs in cancer generates tumor functional heterogeneity conducting eventual cellular phenotype(s) or in uence cell fate determination [4,5,7,8]. In this regard, delineation of the heterogeneity of PRSS3 expression and epigenetic regulation is critical for clarifying the molecular basis of PRSS3 transcripts thus facilitating functional interpretation of the paradoxical effects PRSS3 exerting in cancer development. Functional classi cation and experimental dissection of PRSS3-SVs and their response to 5-aza-CR treatment distinct between PRSS3 Low and PRSS3 High HCC cells (such as Huh7 versus HepG2 cells) may be used as an experimental model for studying PRSS3 splicing-mediated functional heterogeneity during hepatocarcinogenesis. In contrast to permanent genetic mutations, epigenetic disruptions frequently occurred in early clinical stages and plays an important role in modulating cell malignancy in a progressive and reversible manner. Therefore, delineation of the precise molecular mechanisms underlying epigenetic regulation of PRSS3-SVs could contribute to molecular phenotypes of HCC.
This study on bioinformatic analysis of RNA sequencing data of PRSS3-SVs and their clinical relevance gave lots of insigni cantly divergent results. For instance, PRSS3 Low was shown in 50 paired HCC tissues, consistent with our previous observation [35] and the analyses showing in the TCGA and the UALCAN portal [37]. But its decrease was no longer statistically signi cant in more HCC tissue specimens, due to different statistical methods, or integration of the RNAseq data with different median cutoff values for extensively divergent expression of PRSS3-SVs in HCC specimens. Therefore, the conventional parameter such as the median cutoff values may need to be reevaluated for grouping a DEG with divergent expression levels. Moreover, the functional heterogeneity could be caused by microenvironment enhanced co-expression diversity of PRSS3-SVs. As a result, further studies with larger sample size of paired HCC specimens are warranted to validate our observations.

Conclusions
In summary, PRSS3 was aberrantly expressed in HCC due to epigenetic dysregulation that was integrated with divergent expression of PRSS3-SVs by site-speci c CpG methylation. The effects of oncogenic PRSS3-V2 and tumor-suppressive PRSS3-V1 in HCC cells may increase the molecular diversity and functional plasticity of hepatocarcinogenesis. Epigenetic dysregulation of PRSS3-V2 distinct between mPRSS3-V2 Low in early clinical stages and umPRSS3 High in advanced tumors has potential diagnostic value for patients with early HCC. Consent for publication: All authors read and con rmed that this work can be published.
Availability of data and materials: All data are publicly released from TCGA, GTEX, GDAC, GEPIA, FIREHOSE and CCLE databases and hyper-links including citations have been included in the "Materials and Methods" and "Result" section.
Competing interests: The authors declare that they have no competing interests.    Columns: HCC cell lines or tissue specimens. The statistical signi cance of correlation coe cients between CpG sites (red) and mRNA expression of PRSS3 transcripts were shown at the bottoms. * P < 0.05, ** P < 0.01, ** *P < 0.001 ( Figure S3, S4 and Table S5)    50 μM. One-way ANOVA with Tukey's post hoc test was calculated for the transfected cells compared with the vector control in (C-D). *P < 0.01** P < 0.01, versus control. Data is presented as mean ± SD of a representative of three independent experiments done in triplicate.