Introduction
Colorectal cancer (CRC) is one of the most prevalent malignancies worldwide. It ranks 3rd in terms of incidence and 2nd in terms of mortality among all cancers [
1]. In 2020, more than 280,000 people died from CRC in China [
2], and the overall 5-year survival rate for patients with advanced CRC is still less than 15% due to metastasis and recurrence [
2]. Thus, exploring the molecular mechanism of CRC progression and identifying more specific molecular markers will be helpful in determining early prognosis and progression and will ultimately reduce the mortality caused by this malignancy.
N6-Methyladenosine (m6A), the most abundant posttranscriptional modification of mRNA, is emerging as an essential chemical marker that modulates mRNA splicing, maturation, stability and nuclear export and subsequently facilitates translation of mRNA or promotes its turnover [
3‐
8]. m6A mRNA methylation is dynamically catalyzed and regulated by three types of protein complexes: methyltransferases (writers), demethylases (erasers), and binding proteins (readers). “Writers” and “erasers” are responsible for m6A deposition and removal. “Readers” are responsible for recognizing and binding to m6A-contained mRNA and eventually determining the fate of the target mRNAs, thus modulating the physiological and pathophysiological processes that occur in cells. Disrupted “readers” that contribute to incorrect deciphering of the m6A code in mRNA have been implicated in diverse biological processes including tumorigenesis [
9]. For example, a group of YTH domain-containing proteins (YTHDFs) have been identified as classical m6A “readers”. High YTHDF1 expression is associated with poor prognosis in hepatocellular carcinoma (HCC) and CRC [
10,
11]. YTHDF2 inhibits the proliferation and growth of HCC cells by disrupting the stability of epidermal growth factor receptor mRNA [
12], and YTHDF3 negatively regulates the interaction between two long noncoding RNAs, growth arrest-specific 5 (GAS5) and yes-associated protein (YAP) and ultimately inhibits CRC progression [
13]. Therefore, the m6A “reader” functions as an oncogene or as an antioncogene in different scenarios that depend on decoding distinct target genes.
Serine/arginine-rich splicing factor 9 (SRSF9), a member of the serine/arginine (SR)-rich family of pre-mRNA splicing factors (SRSFs), constitutes part of the spliceosome and is critical for mRNA splicing [
14]. SRSFs also regulate mRNAs by modulating mRNA export from the nucleus and translation [
15]. SRSF9 was reported to act as a key regulatory factor by increasing the expression and nuclear accumulation of β-catenin, and a possible mechanism through which it stabilizes β-catenin mRNA and increases its translation was proposed [
16]. However, the function of SRSF9 in human cancers, including CRC, has not been studied extensively, and its mechanism of action is even less clear. In 2018, Huang et al. reported that SRSF9 binds to an artificial m6A consensus sequence in high-level [
9]. This information gives us a novel perspective from which to investigate the mechanism of action of SRSF9 in human tumors. Here, we focus on the role of SRSF9 in CRC progression and explore whether the underlying mechanism of its effect is related to the functional m6A-binding protein.
Materials and methods
Tissue samples and cell lines
Sixty-three paraffin-embedded human CRC tissues and corresponding normal mucosal tissues and sixteen fresh human CRC tissues and matched normal mucosal tissues were collected for this study. All the samples were obtained from the Department of Pathology, Nanfang Hospital, Southern Medical University, Guangzhou, China (from March 2018 to June 2019). All samples were obtained from patients who had been diagnosed with primary CRC by two pathologists based on microscopic analysis of tumor tissue, and none of the patients received radiotherapy or chemotherapy prior to surgical resection. Sixty-three cases, including 38 (60.3%) males and 25 (39.7%) females, were recruited for this study. The use of human materials was approved by the Medical Ethics Committee of Nanfang Hospital, Southern Medical University. Seven human CRC cell lines (HCT8, SW620, Caco2, HT29, HCT15, HCT116, and LOVO) and an immortal human intestinal epithelial cell line (NCM460) were donated by Guangdong Province Key Laboratory of Molecular Tumor Pathology, Southern Medical University and validated by short tandem repeat sequence analysis prior to the start of this project and cultured in Roswell Park Memorial Institute (RPMI) 1640 medium (Gibco, Grand Island, NY, USA) supplemented with 10% fetal bovine serum (FBS, Gibco). All cells were maintained in a cell incubator under 5% CO2 at 37 °C.
Immunohistochemistry (IHC)
IHC was performed according to standard protocols as previously described [
17]. After deparaffinization and rehydration, 4-μm-thick tissue slides were blocked with 3% H
2O
2 and boiled in ethylenediaminetetraacetic acid (EDTA) (0.01 M, pH 8.0) for antigen retrieval. After blocking with 10% normal goat serum PBS, the slides were incubated with primary antibody at 4 °C, followed by treatment with secondary antibody and diaminobenzidine (DAB). Information on the primary and secondary antibodies used in this study is provided in Additional file
1: Table S1. The staining intensity of each tissue section was determined by two experienced pathologists in a double-blind manner. We referred to published standards to define “low” and “high” expression of SRSF9 [
17]. Tumor cells were graded according to the following criteria: ① Staining intensity score: 0, no staining; 1, poor staining; 2, moderate staining; and 3, strong staining; ② Positive staining score: 0, < 10% positive; 1, 10–30% positive; 2, 30–50% positive; and 3, > 50% positive. The total score was calculated as ① Staining intensity score × ② Positive staining score. Samples with total scores ≥ 5 were defined as showing high expression; those with scores < 5 were defined as showing low expression.
RNA isolation and qRT-PCR
Total RNA was extracted from human CRC cells and tissues, and mRNA was reverse-transcribed into complementary DNA using the PrimeScript™ RT reagent kit (TaKaRa, Shiga, Japan) in accordance with the manufacturer’s recommendations. qRT-PCR was performed in a 7500 Fast Real-Time PCR System with SYBR Green qRT-PCR master mix (TaKaRa). Information on the primer sequences used in this study is provided in Additional file
2: Table S2. Each experiment was performed in triplicate, and expression of all genes was normalized to that of β-actin.
Western blotting and antibodies
Protein samples were separated by electrophoresis on 10% sodium dodecyl sulfate gels and transferred to polyvinylidene fluoride membranes (Millipore, MA, USA). The membranes were then blocked in 5% skim milk for 1 h at room temperature, followed by incubation with primary antibodies for 8 h at 4 °C. The membranes were incubated with HRP-conjugated secondary antibodies for 1 h at room temperature. Finally, the signal density was detected by chemiluminescence using an ECL reagent kit (FDbio Science, Hangzhou, China). Information on the primary and secondary antibodies used in this study is provided in Additional file
1: Table S1.
Cell transfection and lentiviral infection
For transient transfection, small interfering RNAs (siRNAs) directed against SRSF9 (#SIGS0008682-4, RIBOBIO, Guangzhou, China), DSN1 (#SIGS0013609-1, RIBOBIO), METTL3 (#SIGS00056532-1, RIBOBIO) and negative control RNAs (si-ctrl) were synthesized by RIBOBIO Company (Guangzhou, China). Transient transfection was performed using a Hieff Trans™ Liposomal Transfection Reagent kit (#40802, Yeasen, Shanghai, China) in accordance with the standard protocol. Cells were collected after 24 h for qRT-PCR and after 48 h for Western blotting and functional studies.
For lentiviral transfection, Flag-SRSF9 (ov-SRSF9), empty vector (Vector), shSRSF9 (sh1, sh2), and shNC were purchased from GeneChem Company (Shanghai, China). Caco2 cells and HT29 cells were used to establish stable SRSF9 overexpression models, and HCT116 cells and LOVO cells were used in the stable SRSF9 knockdown experiments. According to the manufacturer’s instructions, 4 × 104 cells per well were seeded and transfected with the indicated lentiviruses. The infected cells were screened using 5 μg/mL puromycin (Solarbio, Beijing, China) for 1 week or longer, and transfection efficiency was determined by qRT-PCR and Western blotting analysis.
Cell proliferation assays
For CCK-8 assays, treated cells were placed in 96-well plates at 1 × 103 cells/well in quintuplicate and cultured in RPMI-1640 medium containing 10% FBS for 4 h, 8 h, 16 h, 32 h, and 64 h at 37 °C in 5% CO2. CCK-8 detection was performed using a Cell Counting Kit-8 (CCK-8, Yeasen) according to the manufacturer’s instructions. The cells were cultured in the presence of 100 μL of CCK-8 (1:10 dilution) for 2 h at 37 °C, and the optical densities (ODs) at 450 nm of the cultures were then measured. Each sample was analyzed in triplicate.
For the colony-forming assay, treated cells were seeded in 6-well plates at 200 cells/well and incubated in RPMI-1640 medium containing 10% FBS for 2 weeks at 37 °C in 5% CO2. The cells were then washed twice with PBS and fixed in 4% paraformaldehyde (Leagene, Beijing, China) for 20 min at 4 °C. Staining with 0.1% crystal violet (Sigma, St. Louis, MO, USA) was then performed within 20 min. The stained cells were counted using a scanner (Bio-Rad, Hercules, CA, USA). Each sample was analyzed in triplicate.
Transwell migration and invasion assays
Transwell chambers (Corning, NY, USA) were used to observe cell migration and invasion. For migration, cells were collected and resuspended in serum-free medium. A total of 200 µL of cell suspension at 5 × 105 cells/mL was placed in the upper chamber, and 600 µL of RPMI-1640 medium containing 10% FBS was added to the lower chamber. For invasion, the upper chamber was covered with 2 mg/mL Matrigel (Corning) prior to cell seeding. After allowing time for migration and invasion, the cells were fixed in 4% paraformaldehyde (Leagene) for 20 min at 4 °C and stained with 0.1% crystal violet (Sigma) for 30 min. The cells in three randomly selected visual fields of each sample were photographed and counted under a light microscope at 200× magnification. Each sample was analyzed in triplicate.
Wound-healing assay was also performed to observe cell migration. Monolayers of cells were uniformly scratched using the narrow edge of a 10-μL pipette tip and cultured in serum-free RPMI-1640 medium. At 24 h and 48 h thereafter, the extent of healing of the scratches was observed under a light microscope at 200× magnification. Each sample was analyzed in triplicate.
Flow cytometry
Treated cells were washed three times with PBS and fixed in 70% cold ethanol overnight at 4 °C. The cells were then washed again, centrifuged, suspended and stained with 50 μg/mL propidium iodide and 1 mg/mL RNase in PBS. Cell cycle analysis was performed using a flow cytometer (BD Biosciences, NJ, USA) and analyzed by ModFit software (BD Biosciences). Each sample was analyzed in triplicate.
In vivo tumorigenesis assay
Animal experiments were approved by the Use Committee for Animal Care and performed in accordance with institutional ethical guidelines for animal experiments. Stable Caco2 and HCT116 cells (5 × 106 cells/injection) were resuspended in 200 μL of PBS and subcutaneously injected into 4-week-old female athymic BALB/c nude mice purchased from the Guangdong Medical Laboratory Animal Center. The resulting tumors were measured every 3–5 days, and their size was calculated according to the formula: volume = 1/2 * (width2 × length). After 4 weeks, the mice were sacrificed, and the xenograft tumors were excised, photographed, and fixed in formalin for histological analysis.
Downstream target prediction
Downstream gene data for the RNA-binding protein SRSF9 were obtained from the web tool POSTAR2 (
http://lulab.life.tsinghua.edu.cn/postar), which is a comprehensive database for exploring posttranscriptional regulation based on high-throughput sequencing data and provides information on a large number of binding sites for RNA-binding proteins [
18]. Data on the differentially expressed genes (DEGs) between 473 human CRC tissue sample data sets and 41 normal tissue sample data sets downloaded from The Cancer Genome Atlas (TCGA) database (
https://www.cancer.gov/about-nci/organization/ccg/research/structural-genomics/tcga) were analyzed using the edge R package [
19] of R software (Vienna, Austria). The DEGs between 17 human CRC tissue sample data sets and 17 normal tissue sample data sets obtained from the Gene Expression Omnibus (GEO) database (
https://www.ncbi.nlm.nih.gov/geo/) (GSE32323) were analyzed by the limma package [
20] of R software. Software default settings were utilized in the analyses. Downstream genes of SRSF9 identified in the POSTAR2 database and the DEGs from TCGA and GEO were screened using Venn diagrams to identify the shared genes. Potential downstream target genes had to meet the following requirements: (1) their mRNAs were binding targets of the human RNA-binding protein SRSF9; and (2) the differences in their expression between human CRC tissue samples and normal tissue samples had P values < 0.05 by statistical analysis and log
2(fold change) > 1.
Survival analysis in the CRC dataset
The association of SRSF9 and DSN1 expression with overall survival was analyzed using the web tool GenomicScape (
http://www.genomicscape.com/microarray/survival.php), which was established based on data obtained from GEO [
21,
22]. Fifty-five CRC individuals in the GEO database (GSE17538) were sorted according to the expression of SRSF9 and DSN1. All cases from GenomicScape were provided to the algorithm for survival analysis. Kaplan–Meier survival plots with hazard ratios (HRs) and log-rank P values were obtained using the webpage. P values < 0.05 were considered to indicate significant differences.
Methylated single-stranded RNA affinity assay
Single-stranded RNA oligonucleotide probes containing the m6A-binding consensus sequence GGACU with methylated (ss-m6A, 5′-biotin-CGUCUCGG(m6A) CUCGG(m6A)CUGCU-3′) or unmethylated (ss-A, 5′-biotin-CGUCUCGGACUC GGACUGCU-3′) adenosine were synthesized by RIBOBIO Company (Guangzhou, China) and validated by mass spectrometry. According to the standardized procedure used with the RNA pull-down kit (gzscbio, Guangzhou, China), each single-stranded RNA oligonucleotide probe was immobilized on streptavidin magnetic beads and coincubated with total proteins extracted from LOVO cells for 8 h at 4 °C. After two washes, the proteins combined with the single-stranded RNA oligonucleotide probe (methylated or unmethylated) were separated by electrophoresis on 10% sodium dodecyl sulfate gels and detected by silver staining and immunoblotting analysis.
Silver staining of protein gels
Proteins that had been separated on gels were stained using a Protein Fast Silver Stain Kit (BBproExtra, Guangzhou, China) according to the manufacturer’s recommendations, and the silver signal density was analyzed using a scanner (Bio-Rad).
Gene-specific m6A qRT-PCR
Total RNA was isolated from LOVO cells. According to the standard operating protocol for the EpiQuik™ CUT&RUN m6A RNA Enrichment Kit (Epigentek, NY, USA), RNA samples were fragmented and incubated with anti-m6A antibody (#A-1801, Epigentek) for 90 min at room temperature. A nonimmune IgG was used as a negative control. RT-qPCR assays with DSN1 primers were performed to quantify the enrichment of m6A-containing RNA. Information on the primer sequences used in this study is summarized in Additional file
2: Table S2. Each experiment was performed in triplicate, and all samples were normalized to β-actin.
Dual-luciferase reporter assay
cDNAs containing the SRSF9-binding region of DSN1 mRNA sequences (chr20:36773597–36773736) were cloned into a GV361 control reporter plasmid consisting of firefly luciferase (F-luc) and verified by DNA sequencing (GeneChem, Shanghai, China). In the mutant reporter plasmid, adenosine (A) on the m6A motif was replaced by thymine (T). The GV219 vector (GeneChem) formed the backbone of the SRSF9 expression plasmid, and DNA sequencing was performed for verification. 293T cells were seeded into 24-well plates followed by cotransfection with 500 ng of DSN1 luciferase reporter plasmid (GV361-DSN1-WT and GV361-DSN1-MUT) and 0 µg, 0.25 µg, or 0.5 µg of SRSF9 expression plasmid (GV219-SRSF9) using a Dual-Luciferase Assay kit (Promega, WI, USA). After 48 h, the cells were harvested, and luciferase activity was measured according to the recommended protocols. Each group was assayed in triplicate.
Measurement of mRNA stability
Stable LOVO cells and stable Caco2 cells were incubated with 5 μg/mL actinomycin D (APExBIO, TX, USA) for 0 h, 2 h, 4 h, 6 h, or 8 h, and RNA was then extracted from the cells. Analysis of the half-life of DSN1 mRNA was performed using qRT-PCR as described earlier [
9].
Statistical analysis
The results are presented as the mean ± SD. Student’s t test (two-tailed) and one-way or two-way ANOVA were used to evaluate quantitative data. The chi-squared test was used to analyze qualitative data. The Wilcoxon test was used to analyze rank data. The log-rank test was used to assess significant differences in overall survival. All of the analyses were performed in SPSS 24.0 (SPSS, Inc., IL, USA) and GraphPad Prism 6.0 (GraphPad, Inc., CA, USA). The level of statistical significance was defined as P < 0.05 (*P < 0.05; *P < 0.01; ***P < 0.001; ****P < 0.0001; NS: not significant).
Discussion
CRC has a high prevalence and accounts for approximately 1 in 10 cancer cases and deaths worldwide. In 2020, the incidence and mortality rate of CRC in China ranked third and fifth, respectively, among all malignant tumors [
2]. Therefore, it is urgently necessary to improve the early diagnosis and prognosis of CRC by identifying biomarkers and uncovering the molecular mechanisms that contribute to CRC. Epigenetic alterations have emerged as a widespread regulatory mechanism that controls gene expression in diverse pathological processes, especially in cancer pathogenesis and progression. In recent years, RNA modifications have become a hot topic in epigenetic regulation research, among which
N6-methyladenosine (m6A) modifications are thought to be the most prevalent and abundant modifications that occur in higher eukaryotic mRNAs [
24]. Roles for m6A in various diseases have been reported [
25,
26]. An increasing number of studies have proposed that m6A-binding proteins (so called “readers”) execute the recognition and interpretation of the target m6A-modified RNA and directly guide various biological processes [
27]. For example, m6A modifications in mRNA are recognized by proteins that use the YTH domain (including YTHDF1/2/3 and many more) to bind directly to the m6A motif, thereby promoting the degradation of target RNAs with m6A modifications [
28‐
31]. On the other hand, proteins that belong to the newly discovered category of “reader” proteins, which includes IGF2BP1/2/3, utilize KH domains to recognize m6A-containing transcripts and are responsible for the stabilization of m6A-containing target RNAs [
9]. The opposing roles of IGF2BPs and YTHDFs make it necessary for cells to possess alternative mechanisms that permit modulation of downstream target m6A-modified RNAs following m6A modification. In the “m6A-switch” mechanism, the local structure of target m6A-containing RNAs is altered in a way that affects the accessibility of the last type of “readers” to substrates, including HNRNPC, HNRNPG, and HNRNPA2B1. These changes thus regulate the processing of RNA molecules [
32,
33]. Together, the characterization of the three main categories of m6A readers deepens our understanding of the effects of m6A-binding proteins on genetic information flow. Thus, it will be interesting to investigate other potential m6A-binding proteins that are able to drive corresponding alterations in genetic messages.
SRSF9, also known as pre-mRNA splicing factor, is a member of the serine/arginine-rich protein family. It is a conserved RNA-binding protein that possesses two RNA recognition motifs (RRMs) and an arginine/serine (RS)-rich domain. The RRMs determine the specificity of the protein’s binding to target mRNAs, and the RS-rich domain is mainly involved in protein–protein interactions [
14]. SRSF9 is highly expressed in various malignancies. However, few studies have attempted to determine the biological role of SRSF9 in human cancer. In the present study, our data reveled that SRSF9 was frequently upregulated in CRC cells as well as in clinical CRC tissue samples, and its overexpression was significantly associated with lymph node metastasis, tumor progression and poor prognosis of patients with CRC. These clinical data suggest SRSF9 as a proto-oncogene in CRC, similar to its role in glioblastoma, squamous cell lung carcinoma, malignant melanoma [
16] and bladder cancer [
34]. Our in vitro and in vivo biological assays also confirmed that SRSF9 promotes CRC proliferation, migration and invasion. These findings further emphasize the core roles of SRSF9 in cancer progression.
Recently, SRSF9 has been shown to bind an artificial m6A consensus sequence with high affinity [
9], giving us a distinctive perspective from which to explore its role in human cancers, especially CRC. We attempted to enrich the amount m6A-binding proteins in CRC cells using methylated single-stranded RNA as bait, an efficient and facile test model, with the consensus sequence GG(m6A)CU for RNA pull-down. Among these m6A-binding proteins, SRSF9 was identified by immunoblotting and silver staining. Therefore, we confirmed that SRSF9 is an m6A-binding protein.
The m6A-binding protein (so called “reader”) plays an important role in genetic messages, and dysregulated reader could result in abnormal accumulation of oncogenic products and thereby support the malignant state of cancer. In this study, combining the data on DEGs with the enhanced crosslinking and immunoprecipitation data, we found that DSN1 appeared as the SRSF9 protein high-confidence downstream target in CRC. As the component of MIS12 kinetochore complex, DSN1 plays an important role in sister-mitochondrial interaction including maintaining the stability of the centromere structure, and is also involved in the proper segregation of chromosomes, which is essential for the accurate transmission of genetic material during mitosis [
35]. Recently, it has been shown that high expression of DSN1 is associated with chromosomal instability (CIN) [
36]. In colorectal cancer, DSN1 affects cell cycle progression and is tightly associated with clinical-pathological features [
35]. Furthermore, chromosomal instability, which can lead to abnormal chromosome structure and number, is a prominent feature in human cancers [
37]. DSN1 has already been reported to be upregulated and to act as a considerable risk factor in various tumors, and depletion of DSN1 can inhibit the cellular malignant phenotype [
35‐
38]. In our present study, the expression of DSN1 was significantly up-regulated in CRC cells and clinical tissue samples and indicated poor prognosis in CRC patients, suggesting that DSN1 contributes to maintaining the malignant phenotype of CRC. We further found an obvious positive correlation between DSN1 expression and SRSF9 expression both in vitro and in vivo. In addition, overexpression of SRSF9 upregulated the expression of DSN1, while downregulation of SRSF9 contributed to a decrease in the level of DSN1 in CRC cells. The above results indicate that there is an expression correlation between DSN1 and SRSF9 in CRC. Functionally, silencing of DSN1 impaired SRSF9-induced viability, proliferation, cell cycle progression, migration and invasion, indicating that DSN1 and SRSF9 have functional relevance in CRC. Collectively, our results imply that DSN1 is a critical downstream target of SRSF9 that facilitates CRC progression.
Intriguingly, Wu et al. recently reported that MIS12 mRNA, another component of the MIS12 complex that is homologous to DSN1, is highly enriched in m6A modification [
39]. In our study, the results of gene-specific m6A qRT-PCR assays confirmed that DSN1 mRNAs were also enriched in m6A modifications in CRC cells. By retrieving eCLIP data in POSTAR2, we found that DSN1 mRNA is the binding target of SRSF9 protein and the information on the binding sequences and distribution of m6A modification sites were also collected. These findings indicate that the interaction between SRSF9 protein and DSN1 mRNA is associated with m6A modification. SRSF9 has been proposed to act through pathways such as the wnt signaling pathway [
16] and the apoptosis pathway [
34] or through hypoxia-related signaling [
40]. However, the involvement of SRSF9 as an m6A-binding protein in modulating downstream targets in human cancer has not been investigated previously. To investigate whether SRSF9 modulates DSN1 in an m6A-related manner, we constructed mutant plasmids using a dual-luciferase reporter system in which the adenosine base (A) in the m6A consensus sites was replaced by a thymine base (T) to eliminate the effect of m6A motifs. Our data reveal that m6A motif within the SRSF9-DSN1-binding region is required for the binding of SRSF9 protein to DSN1 mRNA.
It is well known that methyltransferase-like protein 3 (METTL3) (so called “writer”) is a critical component of the methyltransferase complex, which catalyses methylation at
N6-adenosine [
9]. METTL3 is responsible for the vast majority of m6A sites in mRNA [
41]. For this reason, deletion of METTL3 has been used to document numerous m6A-dependent functions [
42,
43]. In this study, there is a positive correlation between the expression of METTL3 and SRSF9, METTL3 and DSN1 as well as SRSF9 and DSN1 in CRC. Interference with METTL3 expression in SRSF9-overexpressing cells significantly down-regulated the expression of DSN1 while the level of SRSF9 expression was unchanged. Therefore, we conformed that the expression regulation of DSN1 by SRSF9 is associated with RNA m6A modification. Notably, m6A-binding proteins are pivotal in the stability of their target mRNAs. We speculated that the turnover of DSN1 mRNA in CRC would be affected by SRSF9 in an m6A-related manner. Measurement of the stability of DSN1 mRNA supports our hypothesis. Taken together, these findings indicate that SRSF9 acts in an m6A-related manner in target recognition and regulation in CRC.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.