Alternative splicing in human cancer cells is modulated by the amiloride derivative 3,5‐diamino‐6‐chloro‐N‐(N‐(2,6‐dichlorobenzoyl)carbamimidoyl)pyrazine‐2‐carboxide

Alternative splicing (AS) is a process that enables the generation of multiple protein isoforms with different biological properties from a single mRNA. Cancer cells often use the maneuverability conferred by AS to produce proteins that contribute to growth and survival. In our previous studies, we identified that amiloride modulates AS in cancer cells. However, the effective concentration of amiloride required to modulate AS is too high for use in cancer treatment. In this study, we used computational algorithms to screen potential amiloride derivatives for their ability to regulate AS in cancer cells. We found that 3,5‐diamino‐6‐chloro‐N‐(N‐(2,6‐dichlorobenzoyl)carbamimidoyl)pyrazine‐2‐carboxamide (BS008) can regulate AS of apoptotic gene transcripts, including HIPK3, SMAC, and BCL‐X, at a lower concentration than amiloride. This splicing regulation involved various splicing factors, and it was accompanied by a change in the phosphorylation state of serine/arginine‐rich proteins (SR proteins). RNA sequencing was performed to reveal that AS of many other apoptotic gene transcripts, such as AATF, ATM, AIFM1, NFKB1, and API5, was also modulated by BS008. In vivo experiments further indicated that treatment of tumor‐bearing mice with BS008 resulted in a marked decrease in tumor size. BS008 also had inhibitory effects in vitro, either alone or in a synergistic combination with the cytotoxic chemotherapeutic agents sorafenib and nilotinib. BS008 enabled sorafenib dose reduction without compromising antitumor activity. These findings suggest that BS008 may possess therapeutic potential for cancer treatment.

Alternative splicing (AS) is a process that enables the generation of multiple protein isoforms with different biological properties from a single mRNA. Cancer cells often use the maneuverability conferred by AS to produce proteins that contribute to growth and survival. In our previous studies, we identified that amiloride modulates AS in cancer cells. However, the effective concentration of amiloride required to modulate AS is too high for use in cancer treatment. In this study, we used computational algorithms to screen potential amiloride derivatives for their ability to regulate AS in cancer cells. We found that 3,5-diamino-6-chloro-N-(N-(2,6dichlorobenzoyl)carbamimidoyl)pyrazine-2-carboxamide (BS008) can regulate AS of apoptotic gene transcripts, including HIPK3, SMAC, and BCL-X, at a lower concentration than amiloride. This splicing regulation involved various splicing factors, and it was accompanied by a change in the phosphorylation state of serine/arginine-rich proteins (SR proteins). RNA sequencing was performed to reveal that AS of many other apoptotic gene transcripts, such as AATF, ATM, AIFM1, NFKB1, and API5, was also modulated by BS008. In vivo experiments further indicated that treatment of tumor-bearing mice with BS008 resulted in a marked decrease in tumor size. BS008 also had inhibitory effects in vitro, either alone or in a synergistic combination with the cytotoxic chemotherapeutic agents sorafenib and nilotinib. BS008 enabled sorafenib dose reduction without compromising antitumor activity. These findings suggest that BS008 may possess therapeutic potential for cancer treatment.

Introduction
In eukaryotes, alternative splicing (AS) is a crucial mechanism for generating transcriptome diversity and can result in single-gene coding for multiple proteins. Currently, 20 000 coding genes are estimated to be able to produce nearly 80 000 proteins and probably many more through AS in humans (Hahn et al., 2015). Protein isoforms that are generated from the same gene through AS may have the same, similar, or even opposite biological properties (e.g., protein-protein interactions, subcellular localization, or catalytic ability) (Tazi et al., 2009). Furthermore, splicing can be modulated through the interplay between RNAbinding proteins (RBPs) and splicing regulatory sequence elements (SREs) in mRNA. Numerous RBPs, which function to regulate AS, have been studied, and the well-known regulators of splice site selection are serine/arginine-rich (SR) proteins and heterogeneous RNPs (hnRNPs). In general, serine/arginine-rich proteins (SR proteins) bind to exonic splicing enhancers through their RNA-binding domain and promote exon inclusion by recruiting spliceosome components. By contrast, hnRNPs often bind to exonic splicing silencers or intronic splicing silencers to cause exon exclusion. SR proteins and hnRNPs generally display an antagonistic function toward each other in SRE selection (Fu and Ares, 2014). Therefore, AS can be determined by a balance between positive and negative regulation, and the modulation of RBPs may alter the final outcome of the splicing reaction. In cancer progression, the AS process is generally interrupted, leading to the creation of both functional and nonfunctional end products. Cancer cells often take advantage of these specific splicing events to produce oncogenes or aberrant tumor suppressors that promote growth and survival (Sveen et al., 2016;Venables, 2004). This implies that shifting splicing toward the appropriate isoform may offer therapeutic potential in cancer treatment.
Amiloride is a potassium-sparing diuretic that was first approved for use in 1967. Its function is to help the kidneys to remove excess sodium and water and retain potassium in the body, and it is usually used to treat hypertension, heart failure, liver cirrhosis, and other diseases that result in edema and ascites (Bull and Laragh, 1968;Loffing and Kaissling, 2003). In our previous studies, we have determined a novel biological action of amiloride, namely the modulation of AS in human cancer cells. Amiloride was able to modulate the AS of various apoptotic genes, including BCL-X and homeodomain-interacting protein kinase 3 (HIPK3), with a concomitant effect in many splicing factors, such as SRSF1, hnRNP A1, hnRNP A2/B1, and hnRNP C1/C2, finally resulting in apoptosis (Chang et al., 2011b). However, the effective concentration of amiloride is too high to limit its applications in cancer treatment. Therefore, in this study, we used computational algorithms to predict amiloride derivatives and attempted to find efficient amiloride derivatives to determine possible therapeutic applications.

Preparation of protein structures and screening databases
To define the binding sites, the RNA/DNA-bound structures of snRNP70 (PDB code 4PKD, Kondo et al., 2015), hnRNP I (PDB code 2AD9, Oberstrass et al., 2005), and U2AF2 (PDB code 4TU8, Agrawal et al., 2014) were aligned to the structure 4PKD using the structure alignment tool CE (Guda et al., 2004)  bound RNA/DNA. We selected compounds from the public database of the National Cancer Institute to generate the SiMMap. The total number of selected compounds for screening was nearly 50 000.

Computational screening and establishment of site-moiety maps
The approximately 50 000 compounds were docked into the binding sites of snRNP70, hnRNP I, and U2AF2 using an in-house docking tool, GEMDOCK (Yang and Chen, 2004), with docking parameters optimized as per virtual screening protocols. Subsequently, the top 2% of compounds (approximately 1000) with the greatest interaction energies for each protein were selected to establish the SiMMap, which described the interaction preferences between binding pockets and moieties. First, protein-compound interaction profiles were generated based on piecewise linear potential as calculated by GEMDOCK. The interaction profiles described the interactions, including electrostatic (E), hydrogen-bonding (H), and van der Waals (vdW) interactions, between the compounds and the target protein residues. A matrix (M) of size C 9 R was used to present the profile, where C is the number of docked compounds and R is the number of interacting residues of a protein. An entry (M ij ), representing the interacting energy between compound i and residue j, was considered significant if it had a Z-score of ≥ 1.65. Then, spatially neighboring interacting residues and their interactive moieties with statistical significance were assigned as an anchor (Figs 1C and 2). Finally, the SiMMap of each target protein was constructed.

Cell culture and gene knockdown
A549, LoVo, and Huh-7 cell lines were obtained from the Bioresource Collection and Research Center, Taiwan. Ba/F3 transfectant (expressing Bcr-Abl with kinase domain point mutations T315I) was provided by Michael W. Deininger (O'Hare et al., 2005). Cells were maintained in DMEM (solid tumor-cell lines) and RPMI-1640 medium (leukemia cell lines) supplemented with 10% FBS and antibiotics (100 UÁmL À1 penicillin and 100 lgÁmL À1 streptomycin) at 37°C in a humidified atmosphere of 5% CO 2 . To perform snRNP70 or hnRNP I knockdown, antisense oligonucleotide sequences of snRNP70, hnRNP I, or a scrambled control (Table S1) were transfected into cells using Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol.

Cell growth inhibition assay
Cells were prepared in a 96-well plate, and 10 lL of MTT solution (5 mgÁmL À1 ) was added to each well. After 2 h, the medium was removed and cells were lysed with 100 lL of DMSO. The absorbance at 565 nm was measured using a microplate reader, and the results are presented as a percentage of the control results.

Reverse transcription PCR analysis
Next, mRNA was obtained using a TurboCapture 8 mRNA kit (Qiagen, Chatsworth, CA, USA) and then converted into cDNA using High-Capacity cDNA Reverse Transcription Kits (Applied Biosystems) according to the manufacturer's instructions. PCR was performed using specific pairs of primers (Table S2).

Cell cycle distribution
Cells were collected and fixed in 70% (v/v) ethanol at 4°C overnight and then stained with 1 mL of PI staining buffer (0.1% Triton X-100, 100 lgÁmL À1 RNase A, and 500 lgÁmL À1 PI in phosphate-buffered saline) at 37°C for 30 min. Data were collected and analyzed using a BD FACSAria TM Fusion flow cytometer (BD Biosciences).

RNA sequencing
Samples were prepared using an mRNA-seq sample kit (Illumina, San Diego, CA, USA) according to the manufacturer's instructions. The raw sequencing data were deposited in the National Center for Biotechnology Information's Gene Expression Omnibus (GEO), and they can be obtained through GEO Series with the accession number GSE110059.
2.11. Huh-7 xenograft model BALB/cAnN.Cg-Foxn1 nu /CrlNarl mice (male, 4-6 weeks old) were purchased from the National Laboratory Animal and Research Center (Taipei, Taiwan); 5 9 10 6 Huh-7 cells were subcutaneously injected into the right flank of the mice. Therapy started after 3- 4 weeks when the tumors had reached an average volume of 300 mm 3 . The mice were assigned to groups of six and subcutaneously injected daily with 0.1 mL of solvent (final volume ratio of DMSO/Kolliphor EL/ PBS was 4.5 : 4.5 : 91) for the control group or 30.9 lg/10 g of BS008 for the treatment group. The longest axis (a), shortest axis (b), and thickness (c) of the tumor were measured with calipers, and the tumor volume (V) was evaluated using the formula V = pabc/ 6. All animal experiments were performed in accordance with the guidelines established by the Institutional Animal Care and Use Committee (IACUC) of China Medical University (CMU). All animals were housed in the Laboratory Animal Center of CMU under a 12-h light/dark (08:00/20:00) cycle with free access to food and water. The mice were sacrificed using CO 2 , and the tumor tissues were subsequently harvested. All breeding and subsequent use of animals in this study, including sacrifice, was approved by the IACUC of CMU (approval number 103-264-B).

Statistical analysis
Differences between the control and treatment were analyzed using Student's t-test, with a probability of less than 5% (P < 0.05) considered significant.

Identification of target proteins
We applied Homopharma (Chiu et al., 2014), similar compound structures often targeting proteins with similar binding sites and sharing interactions, to identify potential target proteins of amiloride. First, we collected known target proteins (Table S3) of amiloride and proteins (Fu and Ares, 2014) that are relative to the functions/pathways of AS (Table S4). Subsequently, we searched the similar compound structures of amiloride and their target proteins recorded in the BindingDB database (Gilson et al., 2016). On the basis of these proteins, we used a structure alignment tool (CE, Guda et al., 2004) to identify the proteins with similar binding environments, such as snRNP70 (PDB code 4PKD), hnRNP I (PDB code 2AD9), and U2AF2 (PDB code 4TU8) (Fig. 1A). Their root-meansquare derivations were 1.14 and 1.35 A when hnRNP I and U2AF2 aligned with snRNP70, respectively. Notably, these three proteins were cocrystalized with mRNA, and their aromatic amino acids (F213 and F171 in snRNP70; F98 and H62 in hnRNP I; and F304 and F262 in U2AF2) consistently formed strong vdW contacts with the rings of RNA (Agrawal et al., 2014;Kondo et al., 2015;Oberstrass et al., 2005) After identifying these potential target proteins, we integrated molecular docking (GEMDOCK, Yang and Chen, 2004) and statistically analyzed thousands of docked poses for site-moiety maps (SiMMaps, Chen et al., 2010) to further investigate and evaluate their binding environments. Our previous studies have shown that GEMDOCK has a similar performance to other docking methods (Ewing et al., 2001;Kramer et al., 1999) and was successfully used to identify novel inhibitors and binding sites (Chin et al., 2010;Hsu et al., 2013Hsu et al., , 2015Yang et al., 2007). An anchor of a SiMMap possesses three properties: a binding subpocket with interacting residues, the moiety composition of screening compounds, and the pocket-moiety interaction type (electrostatic, hydrogen-bonding, or vdW). In this study, we docked approximately 50 000 compounds collected from the public database of the National Cancer Institute using GEMDOCK.
Subsequently, the SiMMaps of snRNP70, hnRNP I, and U2AF2 were established by statistically analyzing the top 1000 docked compounds (approximately 2% of the 50 000 compounds; Figs 1C and 2). The SiM-Maps of snRNP70 (four anchors) and hnRNP I (four anchors) are similar, and their anchors can be well aligned. Therefore, the anchors V1, H1, H2, and V2 of snRNP70 were aligned to the anchors V1, V2, H1, and V3 of hnRNP I, respectively. The docked aromatic moiety of amiloride occupied the anchor (V1 in snRNP70 and V1 in hnRNP I) and imitated the ribose of RNA, forming p-p vdW with the rings of F171 and W252 in snRNP70 as well as residues H62 and R52 in hnRNP I. Conversely, the SiMMap position of U2AF2 differs from that of snRNP70 and hnRNP I because the aromatic residues of these three proteins consistently form strong vdW forces with the ribose of RNA (Fig. 1B,C). These results indicate that a SiM-Map provides a comprehensive analysis and its anchors describe the relationship between the moiety preferences and physicochemical properties of the binding site, as determined by the interaction profiles between target proteins and docked compounds.

Site-moiety map for amiloride optimization
Based on the SiMMaps of these three target proteins, we utilized their anchor distributions and moiety preferences to guide the lead optimization of amiloride (Figs 1C and 2). The aligned SiMMap of snRNP70 and hnRNP I showed that amiloride occupied two anchors in snRNP70 (V1 and H1) as well as in hnRNP I (V1 and V2) (Fig. 2). The positions and moiety preferences of the two nonoccupied anchors (i.e., V2 and H2 in snRNP70) in the aligned SiMMap could be utilized for designing amiloride derivatives and improving its potency. For example, the majority of the 550 compounds (approximately 55%) located in the V2 anchor formed vdW forces with anchor residues (F213, E244, and N161 in snRNP70) through their ring-based moieties. To design amiloride derivatives, a dichlorobenzene was considered as a potential moiety to mimic the ribose of RNA to form vdW and hydrogen bonds with anchor residues (F213, E244, and N161 in V2 in snRNP70; H62, F98, and Q96 in V3 in hnRNP I). The docked poses (blue) of 3,5-diamino-6-chloro-N-(N-(2,6-dichlorobenzoyl)carbamimidoyl)pyrazine-2-carboxamide (BS008) occupied four anchors in snRNP70 and hnRNP I (Fig. 1D). For the H2 anchor in snRNP70 and V2 in hnRNP I, the anchor residues (E244 and R247 in snRNP70; H62 and N132 in hnRNP I) formed vdW and hydrogenbonding forces with the carboxamide moiety of BS008.
According to the aligned SiMMap, U2AF2 lost aligned anchors (i.e., H1 and V1 in snRNP70) to form a sandwich conformation interacting with BS008. These results show that BS008 inhibited both snRNP70 and hnRNP I but not U2AF2 in validation in vitro.

Effect of BS008 on cell viability
According to the predictions of the computational algorithms, BS008 (Fig. 3A) was synthesized and provided by Yang-Chang Wu (Graduate Institute of Natural Products, Kaohsiung Medical University, Kaohsiung, Taiwan). First, we used an MTT assay to assess the growth-inhibitory activity of BS008 in human hepatocellular carcinoma (HCC) Huh-7 cells. Huh-7 cells were treated with various concentrations of BS008 for 24 h, and the results showed that the BS008 treatment exhibited a dose-dependent loss of cell viability, with the half maximal inhibitory concentration (IC 50 ) of BS008 being 0.128 mM (Fig. 3B). This inhibitory effect of BS008 on cell survival demonstrated in the Huh-7 cells implied that BS008 might have potential for use in cancer treatment.

BS008 regulates the AS of apoptotic gene transcripts
The abnormal AS of apoptotic gene transcripts has been reported to possibly be connected to carcinogenesis (Lin, 2017). Since we have previously reported that amiloride regulates the AS of BCL-X and HIPK3 pre-mRNA in various cancer cell lines including human HCC cell line Huh-7 (Chang et al., 2011a(Chang et al., ,2011b, it would be of interest to investigate if BS008 had the same effects on AS in Huh-7. Due to the importance of second mitochondria-derived activator of caspases (SMAC/DIABLO) in determining the sensitivity of cancer cells to apoptotic death, it was also chosen to investigate whether its splicing variants could be regulated by BS008. Huh-7 cells treated with BS008 showed a decrease of the antiapoptotic splice variants, BCL-X L and HIPK3 U(+), while relative maintenance of the proapoptotic splice variants BCL-X S and HIPK3 U(À). Western blot analysis further confirmed a reduced level of BCL-X L protein after the treatment with BS008 (Fig. 9D). Besides BCL-X and HIPK3, we also found SMAC was regulated by BS008 with an increase in the proapoptotic splice variant SMAC-3 (Fig. 4A). Similar splicing patterns were also observed in A549 lung cancer cells and LoVo colon cancer cells (Fig. 4B,C). We also found that BS008 exerts an effect on AS of these gene transcripts in K562 leukemia cells by detecting an increase in the proapoptotic splice variants BCL-X S , SMAC-3, and HIPK3 U(À) with a concomitant decrease in the antiapoptotic splice variant HIPK3 U(+). Notably, using highly imatinib-resistant BaF3/Bcr-Abl T315I cells, we observed that BS008 slightly increased the proapoptotic splice variants BCL-X S and SMAC-3, but not HIPK3 U(À). On the contrary, we detected no alteration in the BCL-X, HIPK3, and SMAC splicing patterns in normal mononuclear cells from healthy individuals (Fig. S1).
These results suggested that BS008 might regulate the apoptosis-related pathway through modulating the AS of apoptotic transcripts in cancer cells. Transient transfection of exon 10-12 spliced HIPK3 U(+) cDNA (Fig. S2A) had no effects on the BS008-modulated HIPK3 splicing. However, transient transfection of HIPK3 U(+) intron containing mRNA resulted in the protection of AS of HIPK3 U(+) regulated by BS008, which might be a cause for the competition between pre-mRNAs for the splicing machinery (Fig. S2B). Cell viability assays further confirm that inhibition of the antiapoptotic HIPK3 U(+) splicing could partially rescue the cell death induced by BS008 (Fig. S2B).

BS008 affects the expression levels of SR proteins and hnRNPs
Because BS008 could modulate the AS of pre-mRNA and the computational algorithms predicted that BS008 might influence the splicing factors hnRNP I, snRNP70, and U2AF2, we used western blotting to analyze the expression levels of these three splicing factors. The results showed that the expression levels of hnRNP I and snRNP70 were downregulated, but U2AF2 did not alter significantly after BS008 treatment (Fig. 5A). This suggested that BS008 might modulate AS through decreasing the expression of hnRNP I and snRNP70. Hence, knockdown experiments were performed to investigate the probable modulation mechanism of BS008. Huh-7 cells were transfected with 20 nM hnRNP I and/or si-snRNP70 for 72 h, and the efficiency of the reduction of the hnRNP I and snRNP70 was validated by western blot (Fig. 5B). We discovered that hnRNP I knockdown decreased the proportion of HIPK3 U(+) variants, but it had no significant effect on the AS of BCL-X and SMAC transcripts. Furthermore, snRNP70 knockdown increased the proportion of HIPK3 U(+) and SMAC-3 variants, but it had no significant effect on the AS of BCL-X transcripts. The combination of hnRNP I and snRNP70 knockdown increased the proportion of SMAC-3 variants, but it had no significant effect on the AS of BCL-X and HIPK3 transcripts in the Huh-7 cells (Fig. 5C). The results of the combined knockdown were not entirely consistent with those of BS008 treatment, which suggested that other splicing factors might be involved in BS008-induced splicing regulation. Therefore, we explored the expression levels of SRSF1, SRSF3, hnRNP A1, hnRNP A2/B1, and hnRNP C1/C2 in BS008 treatment. The results showed that the phosphorylation levels of SRSF1 and SRSF3 were decreased, and the expression levels of hnRNP C1/C2 were downregulated, but hnRNP A1 and hnRNP A2/B1 did not alter significantly after BS008 treatment (Fig. 5A). These results implied that BS008 could modulate AS through changes in the phosphorylation levels or expression levels of various splicing factors. Moreover, the splicing factors have been reported to possibly be subject to AS (Lazar and Goodman, 2000;Robinson and Smith, 2006). Thus, we also evaluated the AS of the splicing-related gene transcripts that changed at the protein level in BS008 treatment. We found that BS008 increased the proportion of hnRNP C Delta Ex2 and hnRNP I Delta Ex2 variants, but it did not significantly affect the AS of SRSF1, SRSF3, and snRNP70 transcripts in the Huh-7 cells (Fig. 5D). Similar splicing patterns were observed in K562 cells and T315I cells (Fig. S3).

Effect of BS008 on AKT kinase, SRPK1, SRPK2, and PP1
During AS, the kinases and phosphatases can be involved in regulating the phosphorylation status of splicing factors (Naro and Sette, 2013). We also explored the effect of BS008 on AKT kinase, PP1, and SRPKs in Huh-7 cells. The results showed that BS008 reduced the expression levels of SRPK1 and SRPK2, activated AKT kinase by increasing the phosphorylation of AKT at Ser473, and activated PP1 by decreasing the phosphorylation of PP1 at Thr320 (Fig. 6A). To explore the possible role of AKT kinase and PP1 in the modulation of AS in BS008 treatment, the AKT and PP1 inhibitors were used. We pretreated Huh-7 cells with 10 lM AKT inhibitor for 1 h before treatment with 0.1 mM BS008, and the efficiency of the reduction of the phosphorylation level of AKT (Ser473) was validated by western blot (Fig. 6B). The results showed that AKT inhibitor had no significant effect on the AS of BCL-X, HIPK3, and SMAC transcripts in the BS008-treated Huh-7 cells (Fig. 6C). By contrast, we pretreated Huh-7 cells with 20 nM PP1 inhibitor okadaic acid (OA) (Cohen et al., 1989) for 1 h before treatment with 0.1 mM BS008. Although there was no significant effect of OA on the AS of BCL-X transcripts, OA pretreatment could relieve the effects of BS008 on HIPK3 and SMAC splicing in Huh-7 (Fig. 6E), as well as effects of BS008 on the dephosphorylation of SRSF3 (Fig. 6D). These results suggested that other splicing factors such as hnRNPs might be involved in the BCL-X splicing process, and PP1 played an important role in modulating phosphorylation of SRSF3 and hence the modulation of HIPK3 and SMAC in Huh-7 cells after BS008 treatment. Pretreatment of the leukemic K562 cells with OA to inhibit PP1 phosphatase activity abrogated the effects of BS008 on BCL-X, HIPK3, and SMAC splicing (Fig. S4). Taken together, these results suggested that PP1 phosphatase rather than AKT kinase was associated with the dephosphorylation of SRSF3, which in turn played a role in the BS008-modified AS process. In this regard, BS008 appears similar to amiloride, which regulates the AS through a PP1-mediated splicing mechanism (Chang et al., 2011a(Chang et al., ,2011b.

BS008 affects histone-tail post-translational modifications
Several studies have revealed the correlation between histone-tail post-translational modifications (PTMs) and AS. Specific histone marks can affect the splicing outcome by recruiting splicing factors to the site of transcription (Podlaha et al., 2014;Zhou et al., 2014). Moreover, we have demonstrated that amiloride reversed aberrant histone modification patterns, resulting in the disruption of the association of splicing complex with the transcripts . To investigate whether histone-tail PTMs are involved in the BS008-regulated splicing, we analyzed the effects of BS008 on histone-tail PTMs in Huh-7 cells. The results showed that BS008 increased the expression levels of H3K4me1, H3K4me2, H3K4me3, H3K27me1, H3K27me2, H3K36me3, and H3K79me2, except H3K27me3 and H3K79me1 (Fig. 7A). We further investigated whether BS008 increased the expression levels of histone methylation through affecting the expression of histone-modifying enzymes. Western blot Fig. 6. Effect of BS008 on AKT kinase, SRPK1, SRPK2, and PP1. (A) Huh-7 cells were treated with the indicated concentrations of BS008 for 24 h. The proteins were analyzed using western blot. (B, C) Huh-7 cells were treated with 10 lM AKT inhibitor for 1 h and then exposed to 0.1 mM BS008 for 24 h. (D, E) Huh-7 cells were treated with 20 nM OA for 1 h and then exposed to 0.1 mM BS008 for 24 h. The proteins and mRNA were analyzed through western blot and RT-PCR, respectively. Western blot: Equal amounts of whole-cell lysates (20 lg) were separated using SDS/PAGE and immunoblotted with various antibodies as indicated. Tubulin is shown as the internal standard. RT-PCR: mRNA was extracted and detected for the AS of BCL-X, HIPK3, and SMAC transcripts. M, marker. and PCR analysis confirmed that BS008 reduced the protein levels of histone demethylases, KDM4A, KDM5A, KDM5C, and KDM7A (Fig. 7C), through modulating the AS of these enzymes (Fig. 7B). These results implied that alterations in histone-tail PTM patterns after BS008 treatment might involve in BS008-induced splicing regulation.

BS008 affects the genome-wide AS detected by RNA sequencing
The ability of BS008 to modulate histone-tail PTM patterns and the phosphorylation and expression levels of splicing-related proteins implied that it might have a wide range of effects on the AS of pre-mRNA. Therefore, we used whole-RNA sequencing to evaluate global genes in BS008-treated Huh-7 cells and found that BS008 could affect the AS of various gene transcripts and the frequency of AS (Fig. 8A,B). Top 500 altered transcripts regulated by BS008 could be classified into functional categories involving pathways in cancer, spliceosome, cell cycle, and apoptosis. We randomly selected five apoptotic genes (AATF, ATM, AIFM1, NFKB1, and API5) and further validated the changes in AS. As shown in Fig. 8C, the AS patterns of these genes were similar to the results of RNA sequencing in BS008 treatment. These results indicated that BS008 had a genome-wide effect on the AS of pre-mRNA.

BS008 arrested the cell cycle at the G 2 /M phase and induced apoptosis
Our results showed that the AS of apoptotic genes was affected by BS008 treatment. To further investigate the effects of BS008 on the inhibition of cell growth, we analyzed the effect of BS008 on cell cycle distribution using flow cytometry; the results showed that administration of BS008 could arrest Huh-7 cells at the G 2 /M phase (Fig. 9A,B) and increase the quantity of apoptotic sub-G 1 -phase cells (Fig. 9C). Moreover, western blot results showed that BS008 increased the expression levels of BAX and the active form of CASPASE-3, while it decreased the expression levels of BCL-X L (Fig. 9D). These data indicated that BS008 could affect cell cycle progression and induced cell apoptosis.

Inhibitory effect of BS008 on tumor growth in a xenograft animal model
Because we observed the inhibition of cancer cell viability by BS008 in vitro, we investigated whether these observations could be translated into an animal model system in vivo. Huh-7 cells were inoculated subcutaneously at the right flank of six male immunodeficient athymic mice. When the tumors reached a size of approximately 300 mm 3 , the mice received subcutaneous injection of BS008 or control solvent every other day, and the sizes of the tumors were monitored every week. After 3 weeks of treatment, tumor growth was significantly suppressed under the influence of BS008 treatment. Representative picture of tumor growth in xenograft nude mice administered solvent (left) and BS008 (right) is shown in Fig. 10A. The six average tumor sizes in the BS008-treated and control groups were 1484.87 and 2362.8 mm 3 , respectively (Fig. 10B). At the end of treatment, the tumor tissue was isolated and weighed. The tissue extracted from the BS008treated group weighed considerably less than that extracted from the control group (0.73 vs. 1.79 g; Fig. 10C). These findings demonstrated that BS008 administration significantly inhibited Huh-7 xenograft growth in vivo.

BS008 potentiates the growth-inhibitory effect of sorafenib and nilotinib
Pharmaceutical combinations have been widely used in clinical applications, displaying advantages over the use of individual product . Currently, sorafenib is the only first-line treatment approved for HCC by the US Food and Drug Administration, but the complete response rate to sorafenib in HCC is relatively low (0.7-3%) (Arao et al., 2013). To investigate if BS008 can potentiate the growth-inhibitory effect of sorafenib, Huh-7 cells were treated with BS008, sorafenib, or both. We found BS008 and sorafenib alone caused a dose-dependent loss of cell viability. A combination of BS008 at a concentration of 0.025 mM and sorafenib at a concentration of 20, 30, and 40 lM inhibited Huh-7 viability to 0.52, 0.32, and 0.14, respectively (Fig. 11A). These data indicated that combination of BS008 and sorafenib produced synergistic effects. As amiloride exhibited the synergistic effect in combination with cytotoxic chemotherapeutic agents on T315I cells (Chang et al., 2011b), we further investigated whether BS008 exerted a therapeutically beneficial effect when administered in combination with nilotinib. As shown in Fig. 11B,C, BS008 exhibited inhibitory effect alone or in a synergistic combination with nilotinib. Notably, BS008 showed a much more potent effect than amiloride when combined with nilotinib (Fig. 11C).

Discussion
The phosphorylation state of splicing factors has been reported to affect almost all steps of AS (Misteli, 1999). In general, AKT can directly phosphorylate SR proteins (Bavelloni et al., 2014;Shultz et al., 2010) or indirectly act on splicing regulation by SRPKs, leading to the translocation of SRPKs into the nucleus and thereby modifying SR protein activities (Zhou et al., 2012). Thus, BS008 decreasing the expression levels of SRPKs may cause AKT to fail to phosphorylate SR proteins through the AKT-SRPK-SR pathway. However, increased phosphorylation levels of AKT were also found in BS008-treated cells, which may be a compensatory activation that attempts to 'normalize' the phosphorylation state of SR proteins. By contrast, pretreatment with the PP1 inhibitor OA could partially restore the effects of BS008 on the phosphorylation levels of SRSF3 as well as the AS of HIPK3 and SMAC transcripts. These results are similar to those of other reports that revealed that 4b-hydroxywithanolide E (Lee et al., 2017), ceramide (Chalfant et al., 2002;Massiello et al., 2004), emetine (Boon-Unge et al., 2007;Pan et al., 2011), and epigallocatechin-3gallate/ibuprofen can modulate (Kim, 2008) the AS of apoptotic gene transcripts through a PP1-mediated splicing mechanism. These publications imply that PP1 plays a crucial role in regulating the AS of apoptotic gene transcripts. However, we found PP1 mediated the effects of BS008 on the AS through the dephosphorylation of SRSF3 proteins, and it mediated the effects of amiloride through the dephosphorylation of SRSF1 proteins. This might be the different functional group of amiloride derivatives, making the interaction between the compounds and proteins differ and leading to diverse regulation of AS. The results showed that BS008 could regulate the AS of hnRNP C, hnRNP I, KDM4A, KDM5A, KDM5C, and KDM7A transcripts. The lost exon of hnRNP C Delta Ex2 will cause the loss of the start codon, and the lost exon of hnRNP I Delta Ex2, KDM4A Delta Ex13, KDM5A Delta Ex18, KDM5C Delta Ex2 and Delta Ex2&3, and KDM7A Delta Ex2 will cause frameshifts, which may contribute to the decline of protein levels in BS008treated cells. Changes in the expression levels of KDM4A, KDM5A, KDM5C, and KDM7A are correlated with alterations in H3K4me1, H3K4me2, H3K4me3, H3K27me1, H3K27me2, and H3K36me3 levels (Kooistra and Helin, 2012) as well as the transcripts of their target RNA. For example, KDM5B is enriched nearby alternative exons, and the reduction of KDM5B causes altered levels of H3K4 methylation in alternative splice exon regions (He and Kidder, 2017); furthermore, knocking down KDM4A increases H3K36me3 levels and increases intron retention (Sen Fig. 9. BS008 arrested the cell cycle at the G 2 /M phase and induced apoptosis. Huh-7 cells were treated with BS008 at the indicated concentrations for 24 h and then stained with PI. (A) DNA content was analyzed using flow cytometry. (B) Histogram of the cell cycle distribution. Student's t-tests (unpaired, two-tailed) were used to compare the cell cycle G 2 /M distribution between BS008 treatment and control (*P < 0.05). (C) Histogram of the apoptotic sub-G 1 population. Data are presented as mean AE SD from three independent experiments (*P < 0.05 using unpaired two-tailed Student's t-test). (D) Huh-7 cells were treated with the indicated concentrations of BS008 for 24 h and then harvested. Equal amounts of whole-cell lysates (20 lg) were separated using SDS/PAGE and immunoblotted with various antibodies as indicated. Tubulin is shown as the internal standard. et al., 2017). BS008 could alter the AS of the genes of splicing factors and histone modification enzymes, and the proteins translated by these genes affect the regulation of AS. Their causal or feedback mechanism needs to be further explored.
In addition to the effects of the AS of BCL-X, HIPK3, and SMAC transcripts, the results of RNA sequencing showed that other apoptotic genes such as AATF, AIFM1, and ATM could be regulated by BS008. The transcriptional cofactor AATF is an RNA polymerase II and a central regulator of the p53-driven DNA damage response. Evidence has indicated that increased expression levels of AATF were found in various cancerous tissues and are negatively correlated with patient survival in neuroblastomas (Fanciulli et al., 2000;Hopker et al., 2012). Moreover, studies have shown that the loss of AATF inhibits proliferation and promotes apoptosis in MG-63 cells by decreasing the level of mutant p53 . The present study showed that an increased proportion of AATF Delta Ex5 was found in BS008-treated cells. The lost exon of AATF Delta Ex5 led to a frameshift and a premature stop at codon 298, which suggested that the deviated AATF that was translated from AATF Delta Ex5 might be involved in BS008-induced apoptosis. AIFM1 is found in the inner mitochondrial membrane and has a dual role as an NADH oxidoreductase and regulator of apoptosis. A defect in AIFM1 can lead to a loss of prosurvival function associated with the assembly/stabilization of the mitochondrial electron transfer chain, resulting in promotion of the translocation of AIFM1 from mitochondria to the nucleus, and finally causing large-scale DNA cleavage (Ardissone et al., 2015;Polster, 2013). Cell-free systems also showed that AIFM1 could cause mitochondrial membrane permeabilization and trigger the release of cytochrome c, indicating that AIFM1 may be involved in a positive feedback loop of the apoptosis pathway (Susin et al., 1999). Our results showed that the proportion of AIFM1 Delta Ex2 was increased in BS008-treated cells, and the lost exon 2 of AIFM1 caused a frameshift, wherein the shifted frame did not encounter a new stop codon. This implied that the AIFM1 deficiency of BS008-treated cells may result in the loss of respiratory elements and, therefore, impaired mitochondrial oxidative phosphorylation, eventually leading to apoptosis. ATM is a serine/threonine protein kinase that can be recruited and activated by DNA double-strand breaks (DSBs) or genotoxic stress, thereby acting as a DNA damage sensor. In response to DNA damage, ATM phosphorylates various downstream proteins, such as CHK2, histone H2AX, and TP53, to start a complex signaling cascade, and it then induces cell cycle arrest, increases Data are presented as mean AE SD from six independent experiments (*P < 0.05 using unpaired two-tailed Student's t-test).
DNA repair, and inhibits apoptosis (Hollingworth and Grand, 2015;Spriggs and Laimins, 2017). Studies have also indicated that ATM-deficient cells derived from ataxia-telangiectasia patients are distinctly sensitive to radiation and defective in DSB repair (Marechal and Zou, 2013). We found that BS008 increased the proportion of ATM Delta Ex9, which is able to cause a frameshift and a premature stop at codon 372. Thus, this improper ATM may not only have severe consequences in repairing DNA damage, but also affect cell cycle distribution and apoptosis in BS008treated cells. Based on these findings, BS008 probably has an effect on apoptosis signaling pathways through regulating the AS of various apoptosis-related gene transcripts, making it potentially useful in cancer treatment.
Sorafenib is a multikinase inhibitor that represses tumor-cell proliferation and angiogenesis and Fig. 11. Treatment with BS008 enhances the sensitivity of cancer cells to sorafenib or nilotinib. (A) Huh-7 cells were treated with BS008, sorafenib, or both at the indicated concentrations for 24 h. (B) T315I cells were treated with BS008 at the indicated concentrations for 24 h. (C) T315I cells were treated with a constant concentration of BS008 or amiloride in combination with nilotinib for 24 h. MTT assays were used to assess cell viability. Data are presented as mean AE SD from three independent experiments.
promotes tumor-cell apoptosis by inhibiting the Raf/ MEK/ERK signaling pathway and VEGF receptor tyrosine kinase. Although sorafenib is the standard first-line systemic drug for HCC, primary and acquired resistance to sorafenib results in limited benefits. Primary resistance of HCC to sorafenib is caused by genetic heterogeneity and may be associated with overexpression of EGFR and abnormal activation of its downstream molecules Ras/Raf/MEK/ERK. Acquired resistance is possibly related to activation of the compensatory pathways, such as cross talk involving JAK-STAT pathways (Zhu et al., 2017). In this study, the results of RNA sequencing showed that BS008 affected not only the AS of EGFR and ERK transcripts, but also the AS of STAT3, Mcl-1, and cyclin D1 transcripts (data not shown). This suggested that BS008 might potentiate the growth-inhibitory effect of sorafenib through modulation of these gene transcripts. However, detailed examination is still required to explore sorafenib resistance-related genes in BS008induced AS.

Conclusions
We found that BS008, an amiloride derivative, is more effective than amiloride in cancer treatment. BS008 has an effect not only on histone-tail PTMs but also on the expression and phosphorylation state of splicing factors, resulting in the genome-wide effects on the AS of gene transcripts (Fig. 12). In addition, BS008 can sensitize cancer cells to sorafenib and nilotinib, indicating that a combination of BS008 and sorafenib or nilotinib is a promising therapeutic strategy for clinical application in the treatment of cancer.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Effects of BS008 on alternative splicing of (A) BCL-X, (B) HIPK3, and (C) SMAC RNAs. Messenger RNA was extracted and detected using RT-PCR for the AS in K562, T315I and normal mononuclear cells. K562 cells were transfected with the expression plasmids as indicated for 24 h prior to BS008 treatment. RT-PCR validation of endogenous HIPK3 alternative splicing (upper panel). Cell viability was analyzed using MTT assay followed by 24-h treatment with BS008 (lower panel). Data are presented as mean AE standard deviation from six independent experiments (*P < 0.05 using unpaired two-tailed Student's t-test). Fig. S3. Effects of BS008 on alternative splicing of (A) snRNP70, (B) SRSF1, (C) SRSF3, (D) hnRNP C, and (E) hnRNP I RNAs. Cells were treated with BS008 for 24 h and then harvested for RT-PCR analysis. Fig. S4. Effects of PP1 phosphatase on BS008-induced alternative splicing. RT-PCR results from K562 cells pretreated with (+) or without (-) okadaic acid and then exposed to BS008 for 24 h. Table S1. Information about the Small interfering RNA used in the study. Table S2. Information about the primers used in reverse transcription polymerase chain reaction. Table S3. Ten known target proteins of amiloride collected from BindingDB. Table S4. Alternative splicing-related proteins.