Requirement of splicing factor hnRNP A2B1 of hnRNPs for tumorigenesis of melanoma stem cells

Background: Cancer stem cells play essential roles in tumorigenesis, thus being the important targets for tumor therapy. The hnRNP family proteins, the important splicing factors, are found to be associated with tumor progression. However, the inuence of hnRNPs on cancer stem cells has not been extensively explored. Methods: Quantitative real-time PCR and Western blot were used to examine the gene expression level. RNA immunoprecipitation assay and RNA sequencing were conducted to identify the RNAs interacted with hnRNP A2B1 on a genome-wide scale. The in vivo assays were performed in nude mice. Results: In this study, the results showed that hnRNP A2B1 of 19 hnRNPs was signicantly upregulated in melanoma stem cells compared with non-stem cells, suggesting the important role of hnRNP A2B1 in cancer stem cells. The hnRNP A2B1 silencing triggered the cell cycle arrest in G2 phase, leading to apoptosis of melanoma stem cells. The results revealed that hnRNP A2B1 could bind to the precursor mRNAs of pro-apoptosis genes (DAPK1, SYT7 and RNF128) and anti-apoptosis genes (EIF3H, TPPP3 and DOCK2) to regulate the splicing of these 6 genes, thus promoting the expressions of anti-apoptosis genes and suppressing the expressions of pro-apoptosis genes. The in vivo data indicated that hnRNP A2B1 was required for tumorigenesis of melanoma stem cells in vivo by affecting the splicing of TPPP3, DOCK2, EIF3H, RNF128, DAPK1 and SYT7, thus suppressing apoptosis of melanoma stem cells. Conclusions: HnRNP A2B1 was required for tumorigenesis of melanoma stem cells. Therefore our ndings presented novel molecular insights into the roles of hnRNPs in cancer stem cells. To address this issue, the expression levels of hnRNPs were examined in melanoma stem cells. The results showed that hnRNP A2B1 of 19 hnRNPs was signicantly upregulated in melanoma stem cell compared with melanoma non-stem cells. The further investigations revealed that hnRNP A2B1 took great effects on tumorigenesis of melanoma stem cells.


Background
Cancer is known as a major cause of death due to its high morbidity and mortality [1]. A lot of investigations have suggested that only a small proportion of cancer cells have tumorigenic capability, which is named as cancer stem cells (CSCs) [2]. CSCs are proved to play essential roles in predicting the biological aggressiveness of cancers. They are correlated with recurrence, metastasis and poor survival in solid tumors [3], thus being responsible for tumorigenesis, tumor differentiation, tumor maintenance, tumor spread, and tumor relapse. CSCs have the ability of unlimited growth and self-renewal [4].
Therefore, elimination of CSCs is di cult but essential to cure cancer completely [5]. At present, it is believed that CSCs origin from abnormal expressions of a series of genes. Gene expression regulation plays an essential role in the biogenesis of CSCs. The regulation of gene expression depends on many factors, including DNA sequence, transcriptional regulation, post-transcriptional modi cation, translational control and protein modi cation [6,7]. Post-transcriptional regulation is the most important regulation process, because it can in uence RNA stability, localization, translation e ciency and sequence [8]. It also contributes to shaping tissue-type speci c proteomes [9]. Post-transcriptional regulation includes transcript stability, binding of RNAs to RNA-binding proteins (RBPs), alternative splicing and regulations by microRNAs (miRNAs) [10], thus providing the exibility for cells to adapt to a wide range of physiological conditions. Among post-transcriptional regulations, alternative splicing is one of the crucial mechanisms for gene expression [11].
It is well known that alternative splicing often results in differential intron and exon retention or skipping and this process can produce different mRNA and protein isoforms from one gene. Studies have shown that over 90% of human RNAs are alternatively spliced, leading to a cellular mRNA accumulation that can encode four to ve fold more proteins than protein-coding genes in the genome [12]. Alternative splicing occurs in a large ribonucleoprotein (RNP) machine called spliceosome [13]. The spliceosome contains ve different small nuclear RNP (snRNP) subunits (U1, U2, U4, U5 and U6) along with more than 200 associated protein cofactors [14]. The splicing process involves 70-250 splicing factors depending on the species, which can interact in multiple steps with RNA and proteins [15]. Among splicing factors, serine/arginine (SR) proteins and heterogenous ribonuclear proteins (hnRNPs) are the crucial families of splicing factors [16,17]. At present, 12 classical SR proteins and 17 canonical hnRNPs are identi ed [18].
Most SR proteins act as splicing activators by binding precursor mRNAs (pre-mRNAs) at exonic splicing enhancers. SR proteins often compete with splicing repressors, such as hnRNPs, whose binding to exonic or intronic splicing silencers can inhibit the selection of splicing site, thereby modulating alternative splicing [19]. By binding to different splicing factors, pre-mRNAs are able to be alternatively spliced in different ways [20]. As reported, some hnRNPs, such as hnRNP A1, K and A2B1, are aberrantly higher expressed in lung cancer, breast cancer or hepatocellular carcinoma [21][22][23], suggesting the important roles of hnRNPs in tumorigenesis. However, the in uence of hnRNPs on cancer stem cells has not been extensively explored.
To address this issue, the expression levels of hnRNPs were examined in melanoma stem cells. The results showed that hnRNP A2B1 of 19 hnRNPs was signi cantly upregulated in melanoma stem cell compared with melanoma non-stem cells. The further investigations revealed that hnRNP A2B1 took great effects on tumorigenesis of melanoma stem cells.
Quanti cation of mRNA with real-time PCR Total RNAs were extracted from cells using an RNA Isolation Kit (Ambion, USA). The reverse transcription reaction was conducted with PrimeScript RT Reagent Kit (TaKaRa, Japan). Quantitative real-time PCR was performed using 2×ChamQ SYBR qPCR Master Mix (Vazyme, USA). The PCR reaction mixture (10 μL) contained Rox reference Dye, cDNA, ChamQ SYBR qPCR Master Mix (Vazyme) and primers (Table   S1). GAPDH (glyceraldehyde-3-phosphate dehydrogenase) was included for normalization.  method was used to calculate the relative fold change of mRNA expression [26]. PCR was conducted by maintaining the reaction at 95℃ for 30 s, and then alternating for 40 cycles between 95℃ for 5 s and 60℃ for 30 s.

Northern blot
Total RNAs were extracted using a cell/tissue genomic DNA extraction kit (Generay Biotech, China) according to the manufacturer's manual. After electrophoresis of 30 μg RNAs on a 2% agarose gel, the RNAs were transferred to a nylon membrane (Amersham Biosciences, Sweden) for 1 h, followed by ultraviolet cross-linking. The membrane was prehybridized in prehybridization solution (Roche, Switzerland) for 30 min. Subsequently, the membrane was hybridized with a DIG-labeled probe (Table S2) at 42 0 C overnight. The membrane was rinsed and then blocked in a blocking solution (Roche) for 1 h at room temperature. After incubation with the antibody (10 μl) against DIG-labeled alkaline phosphatase (Roche) for 2 h at room temperature, the signals of the membrane were detected with the substrate BCIP/NBT solution (Roche).

RNA immunoprecipitation (RIP) assay
Cells were treated with paraformaldehyde for 10 min at room temperature. After washes with ice-cold phosphate buffer saline (PBS), the cells were incubated with PBS containing 0.125 M glycine for 5 min and then with hypotonic buffer [10 mM N-2-hydroxyethylpiperazine-N-ethane-sulphonicacid (HEPES), 1.5 mM MgCl 2 , 10 mM KCl, 0.4% Nonidet P-40, pH 7.9] on ice for 15 min, followed by centrifugation at 3,000 ×g for 7 min. The pellet was resuspended in sonication buffer [10% sodium dodecyl sulfate (SDS), 0.5 M ethylene diamine tetraacetic acid (EDTA), 1 M Tris-HCl, pH 8.0] and then subjected to ultrasonication. The sample was centrifuged at 12,000 ×g for 20 s and the supernatant was incubated with antibody-coupled Protein G magnetic beads (70μl) (Bio-Rad Laboratories, USA) at 4℃ overnight. The beads were washed with PBS. Subsequently the RNAs were extracted using RNA Isolation Kit (Ambion, USA).

RNA-seq and data analysis
The extracted RNAs were subjected to RNA-seq using an Illumina Hiseq 2500 system by Novogene Corporation (China). Brie y, the rRNAs were removed by ribo-zerotm kit (Epicentre, France). Subsequently fragmentation buffer was added to break the RNA into short segments of 250-300bp, followed by the synthesis of cDNAs with random hexamers. After puri cation with AMPure XP beads, the double-stranded cDNAs were added with A tails and the connection of sequencing joints. The cDNA library was enriched by PCR. Sequencing was performed. After assembly of RNA-seq data, the raw data was processed to remove the sequences of adapter, ploy-N and low-quality reads. The clean reads were aligned to the genome reference consortium human reference 38 (hg38) by BWA (Burrows Wheeler Aligner) and IGV (Integrative Genomics Viewer). Based on the counts of reads, the gene expression pro le was obtained.

Kyoto encyclopedia of genes and genomes (KEGG) analysis
The coding sequences of transcripts were extracted and used as queries to search the protein sequences collected in the GO (gene ontology) database with the blast E-value of less than 1×10 -5 . The best hit GO identities were assigned to the transcripts. The p-values were corrected for false discovery rate. Deduced genes with homologues in other organisms were used to map to conserved biological pathways.
Semi-quantitative reverse transcription (RT)-PCR Total RNAs were extracted from cells using a cell/tissue genomic DNA extraction kit (Generay Biotech, China) and then quanti ed by NanoDrop ND-1000 spectrophotometer. The complementary DNA was synthesized using HiScript III 1st Strand cDNA Synthesis Kit (+gDNA wiper) (Vazyme, USA) following the manufacturer's instructions. Subsequently PCR was conducted with sequence-speci c primers (Table  S3). β-tubulin was used as a loading control.

Analysis of caspase 3/7 activity
Caspase-Glo 3/7 assay (Promega) was used to evaluate the activity of caspase 3/7 according to the manufacturer's protocol. Cells at a density of 1×10 4 /well were plated onto a 96-well plate. Subsequently 50 mL of caspase-Glo 3/7 reagent (Promega) was added to each well. After incubation in the dark at room temperature for 1 h, the luminescence of cells was measured.

Cell cycle analysis
Cell cycle analysis was conducted with ow cytometry. Cells were xed in ice-cold ethanol overnight. Then the cells were incubated with DNase-free RNase A (20 mg/mL) for 30 min. After centrifugation at 500×g for 5 min, the cells were stained with propidium iodide (PI, 50 mg/mL). The uorescence intensity of 1×10 4 cells was measured with a ow cytometer at an excitation wavelength of 488 nm.

Silencing and overexpression of gene in cells
To silence the gene expression in cancer stem cells, RNA interference (RNAi) assay was conducted using gene-speci c siRNA (Table S4). The melanoma stem cells (1×10 5 ) were transfected with 50 nM of siRNA using Lipofectamine 2000 (Invitrogen, USA). All the siRNAs were synthesized by Shanghai GenePharma Co., Ltd. At different time after transfection, the cells were harvested for later use.
To overexpress a gene in cancer stem cells, the gene was ampli ed using PCR with sequence-speci c primers (Table S5), followed by cloning into pcDNA3.1 (+) vector. The recombinant plasmid was transfected into melanoma stem cells using Lipofectamine 2000 (Invitrogen). At different time after transfection, the cells were collected for later use.
Melanoma stem cells transfected with hnRNP A2B1-shRNA or hnRNP A2B1-shRNA-scrambled were suspended in physiological saline. Matrigel (Becton, Dickinson and Company, USA) was added to the cell suspension at the nal concentration of 33%. Subsequently, 100 μL of the cell suspension was subcutaneously injected into each BALB/c mouse to induce tumor growth. The tumor volume was measured every week. Six weeks later, the mice were sacri ced. The tumor sizes and tumor weights were determined. All procedures conducted on mice in this study were performed in accordance with the protocols approved by the Institutional Animal Care and Use Committee (IACUC). All the methods were carried out in accordance with the approved guidelines.

Immunohistochemical analysis
Tumor tissue was cut into 5 μm pieces and mounted onto a slide. The slide was dewaxed and hydrated in 100%, 95% and 80% ethanol for 5 min, respectively. Subsequently the primary antibody was incubated with the slide. After washing with PBS, the slide was incubated with the secondary antibody for 10 min at room temperature. Streptavidin peroxidase was added to the slide, followed by incubation for 10 min at room temperature. Then 3-amino-9-ethylcarbazole (AEC) buffer and AEC chromogen (Santa Cruz Biotechnology, Santa Cruz, CA, USA) were mixed and added to the slide. The slide was incubated for 10 min at room temperature. The proteins and the nucleus were labeled with diaminobenzidine (DAB) (Sigma, USA) or 4',6-diamidino-2-phenylindole (DAPI), respectively.

Statistical analysis
The numerical data were analyzed by one-way analysis of variance (ANOVA). The differences between different treatments were analyzed by Student's t test. All data were presented as mean ± standard deviation.

Differentially expressed hnRNPs in melanoma stem cells and non-stem cells
To explore the roles of hnRNPs in cancer stem cells, the expression levels of 19 hnRNP family genes were characterized in two types of melanoma stem cells and non-stem cells. The melanoma stem cells were previously sorted in our laboratory from melanoma cell lines MDA-MB-435 and A375 [24,25], respectively. The results of quantitative real-time PCR revealed that compared with the controls, 3 hnRNPs (A2B1, I and L) and 8 hnRNPs (A2B1, C, E1, H1, H2, I, K and L) were signi cantly upregulated in MDA-MB-435 stem cells and in A375 stem cells compared with melanoma non-stem cells (Fig. 1A), respectively. These data indicated that hnRNP A2B1, hnRNP I and hnRNP L were highly expressed in two types of melanoma cancer stem cells. As shown in Fig. 1B, the expression pro les of hnRNP I and hnRNP L proteins were similar in MDA-MB-435 melanoma stem cells and non-stem cells, while the hnRNP A2B1 protein was signi cantly upregulated in melanoma stem cells compared with non-stem cells (Fig. 1B). Western blot revealed that the content of hnRNP A2B1 was much higher in A375 melanoma stem cells than that in non-stem cells (Fig. 1C). Therefore, hnRNP A2B1 was further characterized in melanoma stem cells.

Requirement of hnRNP A2B1 for melanoma stem cells
To investigate the role of hnRNP A2B1 in melanoma stem cells, the hnRNP A2B1 expression was knocked down or rescued, followed by the evaluation of stem cell viability. Northern blots showed that the hnRNP A2B1 mRNA was knocked down in melanoma stem cells by hnRNP A2B1-siRNA compared with the control (Fig. 2A). To con rm the e ciency of hnRNP A2B1-siRNA, another siRNA speci cally targeting hnRNP A2B1 (hnRNP A2B1-siRNA-2) was transfected into melanoma stem cells. The results indicated that the e ciency of hnRNP A2B1-siRNA-2 was similar to that of hnRNP A2B1-siRNA ( Fig. 2A). To rescue the expression of hnRNP A2B1 in hnRNP A2B1-silenced melanoma stem cells, the stem cells were cotransfected with hnRNP A2B1-siRNA and the plasmid expressing hnRNP A2B1. The data of Northern blot analysis revealed that the expression of hnRNP A2B1 was rescued in the hnRNP A2B1-siRNA-treated melanoma stem cells ( Fig. 2A). Western blot yielded essentially the similar results (Fig. 2B). These data indicated that hnRNP A2B1 was silenced in melanoma stem cells or rescued in the hnRNP A2B1-silenced melanoma stem cells.
The results of MTS assays indicated that the hnRNP A2B1 silencing led to a signi cant decrease of cancer stem cell viability compared with the control, while the cell viability of the rescue of hnRNP A2B1's expression was similar to that of the control (hnRNP A2B1-scambled) (Fig. 2C). At the same time, the apoptotic activity of cancer stem cells transfected with hnRNP A2B1-siRNA or hnRNP A2B1-siRNA-2 was signi cantly increased (Fig. 2D), indicating that the hnRNP A2B1 silencing promoted apoptosis of cancer stem cells. The rescue of hnRNP A2B1's expression generated the similar result to the control (Fig. 2D). The analysis of cell cycle showed that the hnRNP A2B1 knockdown induced the cell cycle arrest in G2 phase of melanoma stem cells, while the rescue of hnRNP A2B1's expression yielded the similar result to the control (Fig. 2E). These data revealed that the hnRNP A2B1 silencing triggered the cell cycle arrest, leading to apoptosis of melanoma stem cells.
Underlying mechanism of hnRNP A2B1 on the stemness of melanoma stem cells To reveal the hnRNP A2B1-mediated regulatory mechanism of the stemness of melanoma stem cells, RNA immunoprecipitation (RIP) assay and RNA sequencing were conducted in melanoma stem cells (MDA-MB-435) to identify the RNAs interacted with hnRNP A2B1 on a genome-wide scale. The RNA sequencing data were deposited in National Center for Biotechnology Information (NCBI) with an accession no. PRJNA658448. RNA-seq results demonstrated that the expression levels of 36 and 69 mRNAs bound to the hnRNP A2B1 protein were signi cantly increased and decreased in melanoma stem cells compared with melanoma non-stem cells, respectively ( Fig. 3A and Table 1). Due to their importance, the upregulated mRNAs bound to the hnRNP A2B1 protein were further characterized. To con rm the results, the expression levels of six mRNAs, selected from the 36 upregulated mRNAs, were examined. The quantitative real-time PCR data showed that the six mRNAs bound to hnRNP A2B1 were signi cantly upregulated in melanoma stem cells (MDA-MB-435 and A375) (Fig. 3B), con rming the RIP data. Table 1 The differential mRNAs bound to the hnRNP A2B1 protein in melanoma stem cells and non-stem cells.  ANXA4, ADD3, ADAMTSL1, B3GAT2, ERI3, TNRC6C, XRCC1,  PKMYT1, CYB5R4, NIN, MPRIP, ST8SIA5, PLEKHD1, YEATS2,  HBE1, ADRA1B, FOXP1, ZCCHC12, NMNAT3, JTB, TAF15, INTS13,  TP53BP1, B4GALNT3, UGP2, TRPM2, KBTBD8, ANKRD33B,  RNF165, EML5, FGF14, FBLN1, MSRB3, TCF12, HBG2, FOXN2,  KRTAP8-1, LOXHD1, NUP88, RTN1, L34079.1, TDP1, UBE2NL,  PPP1R9B, TAL1, STXBP4, LAMA5, VWC2L, TMEM106C, CDH8,  KCNK10, GRB10, GRIK2, TMEM132C, CDH2, ERC2, GYG2, XG KEGG analysis indicated that the upregulated genes interacting with the hnRNP A2B1 protein were involved in cellular pathways, including apoptosis (Fig. 3C). The above results showed that the hnRNP silencing led to apoptosis of melanoma stem cells. Thus apoptosis was further investigated. Among the 36 upregulated RNAs in melanoma stem cells, TPPP3 (tubulin polymerization promoting protein family member 3) [27], DOCK2 (dedicator of cytokinesis 2) [28], EIF3H (eukaryotic translation initiation factor 3 subunit H) [29], RNF128 (ring nger protein 128) [30], DAPK1 (death-associated protein kinase 1) [31] and SYT7 (synaptotagmin 7) [32] were reported to be associated with apoptosis. Therefore, the effects of hnRNP A2B1 on the splicing of these six genes were explored. The RIP data analysis revealed that the intron sequences of all 6 genes were found in the RIP products (Fig. 3D), showing that the precursor mRNAs of 6 genes were bound to hnRNP A2B1 and then spliced in the hnRNP A2B1 complex. The analysis of the eCLIP (cross-linking immunoprecipitation) data in the encyclopedia of DNA elements (ENCODE) project (https://www. encodeproject.org) using USCS tool (Gene Interaction) revealed that hnRNP A2B1 was interacted with EIF3H, which was consistent with our results. The results showed that the sequences of 6 genes bound to hnRNP A2B1 were different in melanoma stem cells and non-stem cells (Fig. 3D), suggesting the difference of hnRNP A2B1-mediated splicing between melanoma stem cells and non-stem cells. To con rm the hnRNP-mediated splicing of TPPP3, EIF3H, DOCK2, DAPK1, RNF128 and SYT7 in melanoma stem cells, melanoma stem cells were transfected with hnRNP A2B1-siRNA or hnRNP A2B1-siRNA-scrambled,followed by the extraction of total RNAs and then the detection of a randomly selected intron of a gene with semi-quantitative RT-PCR. The results demonstrated that the intron content was obviously decreased in the hnRNP A2B1-siRNA-transfected melanoma stem cells compared with the hnRNP A2B1-siRNA-scrambled treatment (Fig. 3E). These data indicated that hnRNP mediated the splicing of TPPP3, EIF3H, DOCK2, DAPK1, RNF128 and SYT7 in melanoma stem cells.
To explore the in uence of hnRNP A2B1 on the expressions of TPPP3, EIF3H, DOCK2, DAPK1, RNF128 and SYT7 in melanoma stem cells, hnRNP A2B1 was silenced or overexpressed, followed by the examination of gene expression. Western blot data indicated that the hnRNP A2B1 silencing led to signi cant downregulations of TPPP3, EIF3H and DOCK2 and upregulations of DAPK1, RNF128 and SYT7 in melanoma stem cells compared with the control (Fig. 3F). On the other hand, the hnRNP A2B1 overexpression resulted in signi cant upregulations of TPPP3, EIF3H and DOCK2 and downregulations of DAPK1, RNF128 and SYT7 in melanoma stem cells (Fig. 3G). These data indicated that promoted the expressions of EIF3H, TPPP3 and DOCK2 and inhibited the expressions of DAPK1, SYT7 and RNF128 in melanoma stem cells.
Collectively, it could be concluded that hnRNP A2B1 promoted tumorigenesis of melanoma stem cells via regulating the splicing of the precursor mRNAs of TPPP3, DOCK2, EIF3H, RNF128, DAPK1 and SYT7.
Roles of TPPP3, DOCK2, EIF3H, RNF128, DAPK1 and SYT7 in apoptosis of melanoma stem cells To reveal the roles of TPPP3, DOCK2, EIF3H, RNF128, DAPK1 and SYT7, which were bound to with hnRNP A2B1, in apoptosis of melanoma stem cells, the expression levels of these genes were examined. Northern blot data indicated that TPPP3, DOCK2 and EIF3H were signi cantly upregulated in melanoma stem cells compared with cancer non-stem cells, while RNF128, DAPK1 and SYT7 were downreglated in melanoma stem cells (Fig. 4A). Western blots essentially yielded the similar results (Fig. 4B). These data suggested that TPPP3, DOCK2, EIF3H, RNF128, DAPK1 and SYT7 played important roles in apoptosis of melanoma stem cells.
To explore the effects of TPPP3, DOCK2, EIF3H, RNF128, DAPK1 and SYT7 on apoptosis of melanoma stem cells, the expressions of these genes were respectively silenced by sequence-speci c siRNAs, followed by caspase 3/7 detection. The results of quantitative real-time PCR and Western blots showed that the expression of TPPP3, DOCK2, EIF3H, RNF128, DAPK1 or SYT7 was silenced by sequence-speci c siRNA compared with the control (Fig. 4C and 4D). The knockdown of EIF3H, TPPP3 or DOCK2 promoted apoptosis of stem cells (Fig. 4E). However, the RNF128, DAPK1 or SYT7 silencing had no effect on apoptosis of melanoma stem cells (Fig. 4E). On the other hand, the overexpression of RNF128, DAPK1 or SYT7 led to a signi cant increase of apoptosis of melanoma stem cells, while the TPPP3, DOCK2 or EIF3H overexpression had no effect on apoptosis of melanoma stem cells ( Fig. 4F and G).

Role of hnRNP A2B1 in tumorigenesis of melanoma stem cells in vivo
To evaluate the impact of hnRNP A2B1 on tumorigenesis of melanoma stem cells in vivo, melanoma stem cells (A375 and MDA-MB-435) transfected with hnRNP A2B1-shRNA or hnRNP A2B1-shRNAscrambled were injected into nude mice, followed by tumor examination (Fig. 5A). The expression of hnRNP A2B1 could be stably silenced by hnRNP A2B1-shRNA. The results showed that the tumor sizes were signi cantly reduced in mice treated with hnRNP A2B1-shRNA compared with those in mice treated with hnRNP A2B1-shRNA-scrambled (Fig. 5B), indicating that the hnRNP A2B1 shRNA suppressed tumor development in vivo. The data of tumor volume revealed that the knockdown of hnRNP A2B1 signi cantly decreased tumor volume compared with the control (Fig. 5C). At the same time, the hnRNP A2B1 silencing led to signi cant decreases of tumor weights compared with the control (Fig. 5D). Western blot data revealed that the hnRNP A2B1 knockdown signi cantly decreased the expressions of stemness genes sox2, oct4 and ALDH1 in solid tumors (Fig. 5E). These results demonstrated that hnRNP A2B1 could promote tumorigenesis of melanoma stem cells in vivo.
To evaluate whether the suppression of tumorigenesis resulted from the downregulation of hnRNP A2B1 and its target genes, the expression levels of hnRNP A2B1 and its targets in solid tumors of mice treated with hnRNP A2B1-shRNA or hnRNP A2B1-shRNA-scrambled were examined. The quantitative real-time PCR analysis showed that the hnRNP A2B1 RNA expression was signi cantly reduced in the solid tumors of the hnRNP A2B1-shRNA-treated mice compared with the control (Fig. 5F). Western blots yielded essentially the similar results (Fig. 5G).
To explore the effects of hnRNP A2B1 silencing on its targets (TPPP3, DOCK2, EIF3H, RNF128, DAPK1 and SYT7) in vivo, the protein levels of 6 genes in solid tumors were examined. Western blot results showed that the expressions of TPPP3, EIF3H and DOCK2 were signi cantly decreased in the solid tumors of mice treated with hnRNP A2B1shRNA compared with the controls, while the SYT7, DAPK1 and RNF128 protein levels were signi cantly increased (Fig. 5H). These data indicated that hnRNP A2B1 could affect the splicing of TPPP3, DOCK2, EIF3H, RNF128, DAPK1 and SYT7 in vivo.
To evaluate the in uence of hnRNP A2B1 knockdown on apoptosis in vivo, immunohistochemistry analysis of solid tumors was conducted. The results revealed that the hnRNP A2B1 silencing resulted in the decreased expression of the proliferation marker Ki-67 and the increased expression of caspase 3 compared with the control (Fig. 5I), indicating that the hnRNP A2B1 silencing triggered apoptosis of melanoma stem cells in vivo.
Taken together, the ndings revealed that hnRNP A2B1 played a positive role in tumorigenesis of melanoma stem cells in vivo by affecting the splicing of TPPP3, DOCK2, EIF3H, RNF128, DAPK1 and SYT7, thus suppressing apoptosis of melanoma stem cells.

Discussion
Apoptosis or programmed cell death, an evolutionarily conserved process in organisms, plays very important roles in tumorigenesis [33]. At present, apoptosis evading is known as a hallmark of cancers [34]. Tumor cells have developed various strategies to evade apoptosis in the key modulators of apoptosis pathways [35]. As well known, the process of apoptosis is controlled by many genes that are differentially expressed in cancers. Alternative splicing is one of the crucial causes for the differential expression of multiple genes [11]. It is reported that the formation of hnRNP assemblies on pre-mRNA promotes the following alternative splicing events [36,37]. Many hnRNPs have been shown to bind to some speci c RNAs and then regulate the post-transcriptional processing of RNAs, thus affecting the expression of RNAs. HnRNP L can directly target the immediate downstream 5' splice-sites of its binding sequence, and subsequently regulate a further downstream exon 5' splice-sites selection [38]. HnRNP A1 can bind to the stem loop of pri-miRNA-18a and modify the secondary structure of this RNA, thereby generating a more favorable cleavage site for Drosha [39]. During the process of in uenza A virus (IAV) infection, host hnRNP K binds to virus mRNA and promotes U1 snRNP recruitment, resulting in mRNA mis-splicing to prevent IAV replication [40]. In the past decades, some hnRNPs are found to be involved in the RNA splicing. However, the relationship between hnRNP-mediated RNA splicing and apoptosis of cancer stem cells has not been explored. In this study, based on the analysis of 18 human hnRNPs, it was found that hnRNP A2B1, upregulated in melanoma stem cells, could suppress apoptosis of melanoma stem cells via post-transcriptional regulation. HnRNP A2B1 takes effects on mRNA stability, mRNA transport, mRNA alternative splicing, cellular senescence and telomere stability by binding singlestranded DNA [41][42][43]. However, the function of hnRNP A2B1 in cancer stem cells remains unclear. As reported, hnRNP A2B1 is required for the alternative splicing of apoptosis-related genes, such as the tumor suppressor BIN1 and the anti-apoptotic gene CFLAR (c-FLIP) [44]. In glioblastoma, breast cancer and pancreatic cancer, hnRNP A2B1 acts as an oncogene by regulating the splicing of apoptosisassociated genes to inhibit apoptosis of cancer cells [44][45][46]. These results are consistent with our ndings. In this context, our ndings provided novel insights into the role of hnRNP-mediated RNA splicing in apoptosis of melanoma stem cells.
As a splicing factor, hnRNPs can bind to RNAs to mediate the mRNAs' transcriptional processing, thereby regulating the expression levels of multiple genes [47]. It can regulate mRNA localization [48], mRNA stability [49] and mRNA deadenylation [41]. At present, however, the mechanism of hnRNP-mediated apoptosis of cancer stem cells remains unknown. In this study, the results showed that hnRNP A2B1 suppressed apoptosis of melanoma stem cells through post-transcriptional regulation of apoptosisrelated DAPK1, SYT7, RNF128, EIF3H, TPPP3 and DOCK2 genes. As reported, TPPP3, DOCK2 and EIF3H act as oncogenes through suppressing apoptosis of cancer cells [50][51][52], while RNF128, DAPK1 and SYT7 function as tumor suppressors via promoting apoptosis of many types of cancer cells [53,54]. Our ndings revealed that the hnRNP A2B1-mediated splicing triggered the upregulation of TPPP3, DOCK2 and EIF3H, and the downregulation of RNF128, DAPK1 and SYT7 in melanoma stem cells, leading to the suppression of apoptosis of cancer stem cells. It has been found that the hnRNP A2B1-mediated splicing can increase or decrease the expression levels of its target genes by interacting with other factors [55,56].
In this context, our study revealed a novel mechanism of hnRNP A2B1-mediated suppression of apoptosis of melanoma stem cells. HnRNP A2B1 protein might be a target for melanoma therapy. Except for the apoptosis-associated genes, the other genes of our RNA immunoprecipitation assay, which were upregulated in melanoma stem cells, merited to be further investigated. The proteins interacted with hnRNPA2B1 could be characterized in the future work.

Conclusion
Our ndings revealed that the hnRNP A2B1-mediated splicing triggered the upregulation of TPPP3, DOCK2 and EIF3H, and the downregulation of RNF128, DAPK1 and SYT7 in melanoma stem cells, leading to the suppression of apoptosis of cancer stem cells. HnRNP A2B1 protein might be a target for melanoma therapy.