293T, A549 and H1299 were provided by Cobioer Biosciences (Nanjing, China). Lung cancer cell lines were cultured in RPMI-1640 (Hyclone, Omaha, NE, USA). All media was supplemented with 10% fetal bovine serum (FBS; Gibco, Grand Island, NY, USA) and 100 U/mL penicillin/streptomycin (Hyclone, Logan, UT, USA). All cells were cultured at 37°C in an atmosphere containing 5% CO2. All cell lines were passaged for less than 8 weeks after thawing.

The short hairpin RNA (shRNA) plasmid and the over-expression plasmid were gifted from the Laboratory of Biliopancreatic Surgery, Tongji Hospital, Tongji Hospital, Huazhong University of Science and Technology. For construction of shRNA vectors, the shRNA primers of Las1l -S, Las1l -L, and Las1l were designed (Table S1). The forward and reverse primers were gradient annealed at 95°C to 25°C (-0.5°C/30 sec). Subsequently, double-strand oligonucleotides were inserted into the sites Age I (Thermo Fisher Scientific, USA) and EcoR I (Thermo Fisher Scientific, USA) of the double digestion pCDH-EF1-copGFP plasmid. For construction of over-expression vectors, pHAGE-puro plasmid was digested with MIuI (Thermo Fisher Scientific, USA) and BmtI (Thermo Fisher Scientific, USA), and the linearized plasmid was homologously ligated with the Las1l-L-Flag (ORF region) to construct a pHAGE-Las1l-L-Flag plasmid. The recombinant plasmid was transformed into DH5α (TsingKe, China) competent cells and overnight cultured on ampicillin-containing of plates. Correct plasmid construction was determined by nucleic acid sequencing. For construction pcDNA3.1-LAS1L-minigene vectors, the Las1l-minigene genomic sequences spanning exons 8 to 10 were synthesized and was cloned into the pcDNA3.1 (+) vector (TsingKe, China). All Plasmid were extracted by usingMolPure^® Endo-free Plasmid Mini Kit (Yeasen, China). All the primers used in this study were synthesized by WuHan Tsingke Biological Company.

To construct of the pcDNA3.1-Las1l-minigene mutant plasmid, mutant primers Las1l-M1-F and Las1l-M1-R (Las1l-M2-F/R or Las1l-M3-F/R) were designed. The pcDNA3.1-Las1l-minigene plasmid was used as template for PCR amplification reaction. Subsequently, 50 μL of PCR product was mixed with 1 μL Dpn I (Vazyme, USA) and then reacted at 37°C for 90min. Then, homologous recombination was performed on the mutant site of the digested product. All those reagents were from Mut Express II Fast Mutagenesis Kit V2 originated from Vazyme Biological Company. Then, 5 μL of the recombinant plasmid was transformed into DH5α competent cells and cultured on ampicillin-containing of plates overnight. Correct plasmid construction was extracted after verification by nucleic acid sequencing.

Together with recombinant plasmid, psPAX, and pMD2G plasmids, recombinant lentiviral vectors were transfected into 293T cells for 2 days. After the lentivirus is filtered, target cells were infected with supernatant for 6h. Then, the transfected cells were treated with puromycin for 2 weeks. After the efficiency of overexpression or depletion plasmids was confirmed, surviving cells were used for further experiments. The qPCR and western blot assay were used to confirm the efficiency of overexpression or depletion plasmids, stable transfected cells were used for further experiments.

The RIPA lysis buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, and 0.1% SDS) was used to lysis cells. Cell lysates were separated by 10% SDS-PAGE and transferred to the PVDF membrane (Merck Millipore, Darmstadt, Germany). Then, the PVDF membrane was blocked by TBST buffer (TBS containing 0.1% Tween 20 and 5% non-fat dry milk) for 60 min. Next, will be incubated with primary antibody overnight. HnRNPA1 primary antibody (D21H11), E-Cadherin primary antibody (24E10), N-Cadherin primary antibody (D4R1H), Vimentin primary antibody (D21H3) were purchased from Cell Signaling Technology Company. β-actin primary antibody (2D4H5) was purchased from Proteintech Company. The blots were then performed using an enhanced chemiluminescent detection system (Tanon 5200, Shanghai, China). Image J software (National Institutes of Health, USA) was used to quantify the grayscale of bands.

TRIzol reagent (Yeasen, China) was used to extract the total RNAs of cells or tissues, and the mRNA was reverse transcribed into cDNA using an cDNA Synthesis Kit (Yeasen, China). Semi-quantitative RT-PCR was performed with 2×Hieff^® HotStart PCR Genotyping Master Mix (Yeasen, China). The inclusion/exclusion forms of exon9 of LAS1L were detected by RT-PCR, and these two isoforms were depicted on 2% agarose gels. β-actin was used as internal control. RT-qPCR was performed using Hieff UNICON^® Power qPCR SYBR Green Master Mix (Yeasen, China) via Applied Biosystems (Vilnius, Lithuania). All primers used in this study were listed in Supplementary table 1 (Table S1 ).

For wound-healing assays, the cells were seeded in 6-well plates. When reached 100% fusion, the cells were scratched onto the plate using 200 μL pipette tip, and then cultured in serum-free 1640 medium. For each sample, 5 areas were randomly selected and observed under 10 x 10 magnification at 0 and 24 h, respectively. The relative distance of wound closure was used to evaluate cell migration ability.

For migration assays, about 8 × 10^4 of cells were seeded in the upper chambers of 24-well Transwell plates (Corning, USA) in FBS-free medium, and RPMI-1640 medium with 10% FBS was deposited in the lower chambers. For invasion assay, Transwell membranes were coated with Matrigel (Corning, USA) prior to seeding cells. After about 24-36 h, cells were fixed in 4% paraformaldehyde for 20 min and stained for 20 min in crystal violet. Swab the cells in the upper side of the chambers completely. The relative number of passed cells represented the abilities of migration or invasion. Multiple random fields were visualized under a wide-field fluorescent microscope (Carl Zeiss, Baden-Württemberg, Germany) at 200× magnification.

According to the manufacturer’s instructions, RIP experiments were performed using the Magna RIP Kit (17-701, Millipore) to clarified the binding ability of hnRNPA1 to Las1l pre-RNA.

RNA pull-down experiments were performed using Pierce™ Magnetic RNA-Protein Pull-Down Kit (Thermo Scientific, USA). The biotinylated Las1l RNA (Tsingke, China) probe was synthesized by WuHan Tsingke Biological Company. Then, 50 μl streptavidin magnetic beads were used to capture the biotin-labeled RNA. Incubate the streptavidin magnetic bead mixture with the protein lysate at 4° C for 45 min. The RNA-Binding Protein Complexes were eluted and analyzed by Western blot.

Statistical analyses were performed using Prism V6 software (GraphPad, La Jolla, CA, USA). The data were presented as the mean ± SEM. A two-sided Student’s t-test was performed to compare the difference between the two groups. The data in the bar graphs are presented as the mean ± SEM. Differences were considered statistically significant at p <.05. P < 0.05, P < 0.01 and P < 0.001 were marked as ‘*’, ‘**’, ‘***’.

HnRNPA1 has attracted significant attention as a promising tumor-related regulatory factors of alternative splicing. However, its crucial role in NSCLC metastasis remains controversial and to be further studied. To further explore whether hnRNPA1 affects lung cancer metastasis and its potential molecular mechanism, we firstly designed an independent shRNA targeted against hnRNPA1 ( Table S2 ), and verified the efficiency in the NSCLC cell lines of A549 and H1299 ( Figure 1A ). Then, we assessed the effects of hnRNPA1 depletion on cell migration. Wound-healing assays and Transwell cell migration assays all showed that knockdown of hnRNPA1 significantly promoted the migration of A549 and H1299 cells ( Figures 1B–D ). Similarly, Transwell cell invasion assays showed that knockdown of hnRNPA1 significantly promoted the invasion of A549 and H1299 cells ( Figures 1E–F ). Previous studies have shown that hnRNPA1 may regulate tumor metastasis by regulating EMT-related molecules, so we sought to determine whether hnRNPA1 regulates NSCLC metastasis through regulating the EMT process. Western blot assays ( Figures 1G–H ) showed that hnRNPA1 depletion inhibited the expression of E-cadherin ( Figure 1I ) and promoted the expression of N-cadherin ( Figure 1J ) and Vimentin ( Figure 1K ), which suggested that hnRNPA1 might promote tumor metastasis by accelerating the transition of mesenchymal phenotypic in NSCLC.

It has been clearly reported that hnRNPA1 has crucial roles in the regulation of many cellular AS events. In order to identify whether there are any other transcripts involved in the regulation of hnRNPA1 in NSCLC metastasis, we searched and analyzed the GSE34992 dataset (21), which focused on downstream AS events associated with hnRNPs in 293T cells. Spliced differential expression analysis was performed between hnRNPA1 knockdown groups and control groups in the GSE34992 dataset ( Figures 2A–B ). Then, the top 100 spliced differentially expressed genes were selected ( Table S3 ). To identify tumor-specific differential splicing genes, we compared the splicing events of these genes in the TCGA SpliceSeq database and only to found that the exon 9 skipping of LAS1L was specific in lung cancer compared to normal tissue ( Figure 2C ). Depending on exclusion/inclusion of exon 9, LAS1L can generate two transcript variants (LAS1L-S and LAS1L-L) regulated by splicing factors ( Figure 2D ). Further qPCR and agarose gel electrophoresis (AGE) assays showed that knockdown of hnRNPA1 significantly inhibited the LAS1L exon 9 skipping event in A549 and H1299 cell lines ( Figure 2E ). Consistent results were also confirmed in the TCGA dataset. Then, we downloaded the percent-splice-in (PSI) data of LAS1L exon9 from TCGA SpliceSeq ( 22 ). PSI is the ratio of normalized read counts indicating inclusion of a transcript element over the total normalized reads for that event (both inclusion and exclusion reads). Pearson correlation analysis of the expression level of hnRNPA1 and the PSI value of LAS1L exon9 also indirectly indicated that hnRNPA1 might promote the exon skipping of LAS1L exon9 (r = -0.31, P<0.001) ( Figure 2F ).

Similarly, Pearson’s correlation analysis was also performed between common 64 splicing proteins and the LAS1L PSI values ( Table S4 ). Then, the top 10 splicing proteins were selected for further verification ( Figure S1A ). The independent shRNA targeted against each splicing protein were designed ( Table S2 ) and verified ( Figure S1B ). To further verify the splicing ability of these proteins on LAS1L exon 9, qPCR assays were performed and the results showed that those proteins had no significant (fold change≥2 or ≤0.5) splicing regulatory ability on LAS1L exon 9 except hnRNPA1 protein ( Figure S1C ). To further clarify the expression of LAS1L exon 9 in NSCLC clinical samples, a total of 58 samples including 29 NSCLC tumor samples and 29 corresponding normal samples were randomly selected from the specimen bank. Semi-quantitative RT-PCR analysis and AGE assays were conducted to detect the expression ratio of LAS1L-S and LAS1L-L ( Figure S2 ). RT-qPCR was used to detect the mRNA expression level of hnRNPA1 ( Figure 2G ), and the pearson’s correlation between hnRNPA1 expression and LAS1L-S/L ratio was then calculated. The results also showed that hnRNPA1 may promote the exon skipping of LAS1L exon9(r = 0.57, P=0.0017) ( Figure 2H ) , which is consistent with the above results.

The above observations suggest that hnRNPA1 depletion promotes NSCLC metastasis and EMT, and promotes the expression of LAS1L-L, so we sought to determine whether the splicing variation of LAS1L exon 9 also affects metastasis and EMT of NSCLC. We firstly designed an independent shRNA to specifically knockdown the LAS1L-S isoform, and a shRNA targeted against exon 9 for isoform-specific knockdown of LAS1L-L ( Figure 3A ) , and then verified their efficiency ( Figures 3B–C ). We next assessed the effects of LAS1L isoform depletion on cell migration and invasion. Transwell cell migration/invasion assays showed that knockdown of LAS1L and specific knockdown of LAS1L-L significantly inhibited the migration and invasion of A549 and H1299 cells, while exist no significant effects when knocking down the LAS1L-S isoform ( Figures 3D–G ). Similarly, we posited that the two LAS1L isoforms caused by exon9 skipping might have different effects on the transition of mesenchymal phenotypic in NSCLC. Further western blots assays showed that LAS1L or LAS1L-L knockdown cells showed a significant reduction of mesenchymal markers N-cadherin ( Figure 3K ) and Vimentin ( Figure 3L ) and a more pronounced increase in the epithelial marker E-cadherin ( Figures 3H–J ).

The above results indicate that the LAS1L may regulate tumor metastasis ability through LAS1L-L, while LAS1L-S has no significant effect. To further verify the metastatic potential of LAS1L-L, we constructed LAS1L-L overexpression plasmid attached with FLAG fusion protein ( Figure 4A ) and then verified their protein expression efficiency in A549 and H1299 cell lines ( Figure 4B ). Next, we assessed the effects of LAS1L-L isoform on cell migration and invasion through transwell cell migration/invasion assays. The results showed that overexpression of LAS1L-L can promote the migration and invasion of A549 and H1299 cells ( Figures 4C–F ). In addition, overexpression of LAS1L-L can also promote the transition of EMT in NSCLC ( Figures 4G–K ). These data indicate that the proportion of LAS1L-L isoform is correlated with tumor metastasis ability.

The above observations suggest that hnRNPA1 can regulate the inclusion or exclusion of LAS1L exon 9. To further test whether hnRNPA1 regulates AS of LAS1L by directly binding pre-mRNA, we conducted RIP assays, and the results showed that hnRNPA1 could directly bind to the pre-mRNA of LAS1L ( Figure 5A ). To further illustrate whether hnRNPA1 regulates AS of LAS1L by directly binding to the region of exon 8 to exon 10 of LAS1L pre-mRNA, we constructed a minigene fragment corresponds to human LAS1L which contained the exons 8, 9, 10 and the flanking introns of exon 9 (350-nt upstream and 301-nt downstream) ( Figure 5B ). Semi-quantitative RT-PCR analysis and AGE assays showed that the ratio of LAS1L-S/L in minigene was significantly reduced in A549 and H1299 cells of knockdown hnRNPA1 compared with normal control cells ( Figures 5C–D ). Equally, the qPCR assays also indicated that knockdown of hnRNPA1 with or without the minigene significantly increased the mRNA levels of LAS1L-L, thus reduced the ratio of LAS1L-S/L ( Figures 5E–H ).

Previous studies have reported that hnRNPA1 has a more robust binding ability to UAGGGU, and a relatively moderate binding ability to UGGGGU and UAGAGU (23–25). To determine the specific binding site of hnRNPA1 on the pre-mRNA of LAS1L, we used bioinformatics tools (RBP map) to predict the potential hnRNPA1 binding sites on the sequence region of exon 8 to exon 10. As shown in Figure 5I , 3 integrated potential hnRNPA1 binding sites (WT1, WT2 and WT3) were eventually identified. Interestingly, these potential binding sites were all located in the intron 8 region rather than any other region. To determine the interactions between hnRNPA1 and those potential binding sites, we performed site-directed mutagenesis of the potential sequence based on the original pcDNA3.1-LAS1L-minigene plasmid vector ( Figure 5I ). The qPCR results indicated that the ratio of LAS1L-S/L was significantly reduced after WT1-segment mutation or WT3-segment mutation compared with the minigene-control group, and the WT1-mutation site was the most significant. However, there were no significant effects after WT2-segment mutation ( Figures 5J–K ). To further determine the direct binding ability between hnRNPA1 and the potential binding sites, RNA-pulldown assays were performed in A549 cells. The 3’biotin-labled RNA oligomers of the potential binding sequences (WT1, WT2 and WT3) and paired mutated binding sequences (M1, M2 and M3) were synthesized ( Figure 5I ). RNA pulldown assays showed that hnRNPA1 could bind with WT1 and WT3, but not WT2 or any other paired mutated sequences ( Figures 5L–Q ). It is noteworthy that hnRNPA1 has a more robust binding ability to WT1 sequences than WT3 sequences, which is consistent with the above results. Combined with the above results, we concluded that hnRNPA1 can directly bind to the WT1 and WT2 sites of intron 8 to regulate the AS of LAS1L exon 9.