The roles and prognostic significance of ABI1-TSV-11 expression in patients with left-sided colorectal cancer

Abnormally expressed and/or phosphorylated Abelson interactor 1 (ABI1) participates in the metastasis and progression of colorectal cancer (CRC). ABI1 presents as at least 12 transcript variants (TSVs) by mRNA alternative splicing, but it is unknown which of them is involved in CRC metastasis and prognosis. Here, we firstly identified ABI1-TSV-11 as a key TSV affecting the metastasis and prognosis of left-sided colorectal cancer (LsCC) and its elevated expression is related to lymph node metastasis and shorter overall survival (OS) in LsCC by analyzing data from The Cancer Genome Atlas and TSVdb. Secondly, ABI1-TSV-11 overexpression promoted LoVo and SW480 cells adhesion and migration in vitro, and accelerated LoVo and SW480 cells lung metastasis in vivo. Finally, mechanism investigations revealed that ABI1-isoform-11 interacted with epidermal growth factor receptor pathway substrate 8 (ESP8) and regulated actin dynamics to affect LoVo and SW480 cells biological behaviors. Taken together, our data demonstrated that ABI1-TSV-11 plays an oncogenic role in LsCC, it is an independent risk factor of prognosis and may be a potential molecular marker and therapeutic target in LsCC.


Results
Elevated expression of ABI1-TSV-11 correlates with lymph node metastasis and predicts poor prognosis in patients with LsCC. To determine the clinical significance and underlying role of ABI1-TSVs in CRC and its subgroups (LsCC and RsCC), we first analyzed the correlations between the clinicopathological information (sex, age, ethnicity, TNM stage, etc.) and survival time (OS; disease-free survival, DFS) by Kaplan-Meier analysis and chi-squared test to evaluate the reliability and representativeness of data. As shown in Table 1, TNM stage was closely related to survival time (OS, DFS) and the selected cases are representative of the typical clinicopathological and prognostic characteristics of CRC patients.
Next, we found that increased ABI1-TSV-11 (Abelson interactor 1 transcript variant 11) expression correlated with lymph node metastasis (P = 0.050, Table 2) in LsCC, and there was no correlation between ABI1-TSV-11 overexpression and KRAS mutation. Notably, Kaplan-Meier analysis showed that, in the CRC group and LsCC group, the 10-year OS rates (32.1% and 44.3%, respectively) in patients with high ABI1-TSV-11 expression were significantly lower than in those with low ABI1-TSV-11 expression (53.8% and 51.5%, respectively; P = 0.020 and P = 0.004; Fig. 1a,b). There were no statistically significant differences in 10-year DFS rates observed between patients with high expression and low expression of ABI1-TSV-11 (Fig. 1d,e). In the RsCC group, while patients with high ABI1-TSV-11 expression showed 10-year OS and DFS rates lower than those with low ABI1-TSV-11 expression, the differences were not statistically significant (Fig. 1c,f).The results of Multivariate Cox proportional hazards models showed that ABI1-TSV-11 could serve as an independent risk factor in LsCC [hazard ratio (HR) = 3.060, P = 0.008; Table 3], but not an independent risk factor in CRC (Table 4). Taken together, these results indicate that ABI1-TSV-11 is a specific prognostic risk factor and may serve as a therapeutic target for patients with LsCC.
The construction of stable cell lines overexpressing ABI1-TSV-11. To determine the biological function of ABI1-TSV-11 in LsCC, we first analyzed the expression of ABI1-TSV-11 in a normal colorectal epithelial cell line (CRL-1541) and three CRC cell lines (LoVo, SW480, SW620) by next-generation sequencing (Fig. 2a), and LoVo and SW480 cell lines, which originate from LsCC and have low endogenous ABI1-TSV-11 expression and different invasiveness, were then selected, to construct the stable ABI1-TSV-11 overexpressed cell models. As shown in Fig. 2b-d, quantitative real-time PCR (qRT-PCR) and western blotting confirmed the www.nature.com/scientificreports/ successful construction of stable cell lines at the mRNA and protein levels. As shown in Fig. S1, EPS8 Knockdown were confirmed at the mRNA by qRT-PCR.

ABI1-TSV-11 overexpression promotes LoVo and SW480 adhesion, migration, and invasion in vitro.
To investigate the effects of ABI1-TSV-11 overexpression on cell adhesion and motility, we performed adhesion, Transwell migration, and invasion assays in vitro. For the adhesion, we firstly compared the adhesion ability of LoVo and SW480 cells under routine culture conditions (Fig. 3a) and found that the adhesion ability of LoVo cells was significantly weaker than that of SW480 cells on PBS-or 0.02% BSA-coated plates (P < 0.05). Next, we compared the adhesion ability of the two types of cell on fibronectin-, gelatin-, and collagen I-coated culture plates, and found that although their adhesion ability was significantly enhanced compared with that of PBS and 0.02% BSA (P < 0.05), their responses to stimulation with fibronectin, gelatin, and collagen I were not significantly different (Fig. 3a,b). As shown in Fig. 3c,d, ABI1-TSV-11 overexpression did not affect the adhesion of LoVo and SW480 cells on PBS-, 0.02% BSA-, collagen I-, and gelatin-coated plates. However, ABI1-TSV-11 overexpression significantly enhanced the adhesion of LoVo and SW480 cells on fibronectin-coated plates (Fig. 3c). On the fibronectin-coated culture plates, the LoVo and SW480 cells overexpressing ABI1-TSV-11 showed different responses to fibronectin. While the overexpression of ABI1-TSV-11 inhibited LoVo cell adhesion to fibronectin, it promoted that of SW480 cells (Fig. 3c,d). To address why ABI1-TSV-11 over-expression in the LoVo and SW480 cells induces distinct responses on adhesion to fibronectin, we analyzed the mRNA expression profile of integrins using RNA-seq data of LoVo and SW480 cells, and found that the differential expression of these integrins, especially ITGB1, in SW480 and LoVo cells induces distinct responses on adhesion to fibronectin (Fig. S2).
For migration and invasion, we found that ABI1-TSV-11 overexpression increased the migration and invasion of LoVo-ABI1-TSV-11 by 4.57-and 5.96-fold, respectively, compared with those of their parental control cells (both P < 0.05; Fig. 4g,o). ABI1-TSV-11 overexpression increased the migration of SW480-ABI1-TSV-11 cells by 17% (Fig. 4h, P < 0.05), but was not sufficient to stimulate the invasion of SW480-ABI1-TSV-11 cells (Fig. 4p, P > 0.05). To address why ABI1-TSV-11 over-expression in the LoVo and SW480 cells induces distinct responses on Matrigel invasion, we analysis the mRNA expression of MMP2 and MMP9, and found that ABI1-TSV-11 www.nature.com/scientificreports/  www.nature.com/scientificreports/ overexpression in LoVo cells could significantly increase the mRNA expression of MMP2 and MMP9, but not change the mRNA expression of them in SW480 cells. That means that it is these differences induces distinct responses on Matrigel invasion (Fig. S3).

ABI1-TSV-11 overexpression has no influence on the proliferation of LoVo and SW480 cell lines in vitro.
To test the role of ABI1-TSV-11 in LoVo and SW480 proliferation, we conducted Cell Counting Kit-8 viability assays. As shown in Fig. S4a and S4b, the viability of LoVo-ABI1-TSV-11 and SW480-ABI1-TSV-11 cell lines did not differ significantly at each time point (0, 24, and 48 h) compared with that of their corresponding controls (P > 0.05 at each time point). Therefore, ABI1-TSV-11 overexpression had no influence on the proliferation of LoVo and SW480 cell lines.

ABI1-isoform-11 affects actin dynamics of LoVo and SW480 cells by interacting with EPS8. As
shown in Fig. S5 and S5b, to determine how ABI1-TSV-11 promotes LsCC metastasis, we firstly proved its overexpression did not alter WAVE2, PI3K, EPS8, and N-WASP proteins levels, and ABI1-isoform-11 could only interact with EPS8, but not with WAVE2, PI3K and N-WASP in both LoVo-ABI1-TSV-11 and SW480-ABI1-TSV-11 cells. Moreover, cellular immunofluorescence analysis proved that ABI1-isoform-11 colocalized with EPS8 at the subcellular level in LoVo-ABI1-TSV-11 and SW480-ABI1-TSV-11 cells (Fig. 7a,b), which was direct evidence of the interaction between them. Meanwhile, we also observed the co-localization of ABI1-isoform-11 and F-actin (Fig. 7a,b). Taken together, the above results indicate that the interaction of ABI1-isoform-11, ESP8, To illustrate the relationship among ABI1-isoform-11, ESP8, and actin dynamics, we examined cell morphology, protrusions, and the co-localization of ABI1-isoform-11, EPS8, and F-actin when stimulated with PBS, fibronectin, and gelatin, respectively. We found that LoVo-ABI1-TSV-11 and SW480-ABI1-TSV-11 cells showed more protrusions and extensions when stimulated with fibronectin (Fig. 8b,e), while showing a round morphology and fewer protrusions when stimulated with PBS and gelatin (Fig. 8a,c,d,f). Moreover, the co-localization of ABI1-isoform-11, EPS8, and F-actin was mainly distributed on lamellipodium-like cellular protrusions and the direction of cell extension when simulated with fibronectin (Fig. 8b,e), while being mainly located at the position of cell-cell adhesion when stimulated with PBS and gelatin (Fig. 8a,c,d,f). Thus, we concluded that the interaction between ABI1-isoform-11 and EPS8 was consistent with actin dynamics in LoVo and SW480 cells. Finally, we analyzed the difference in co-localization of ABI1-isoform-11, EPS8, and F-actin in LoVo-ABI1-TSV-11 and SW480-ABI1-TSV-11 cells when stimulated with fibronectin. As shown in Fig. 8b,e, the co-localization of signals between ABI1-isoform-11 and F-actin was weaker and the co-localization between ABI1-isoform-11 and EPS8 was only in a limited punctate pattern in SW480-ABI1-TSV-11 cells compared with that in LoVo-ABI1-TSV-11 cells. (with over-expressed ABI1-TSV-11) cells were successfully constructed and then migration and invasion assays were performed. We found the EPS8 Knockdown attenuates the migration of LoVo-ABI1-TSV-11 (Fig. S9) and SW480-ABI1-TSV-11 cells but not the invasion (Fig. S10) of them. We also found that EPS8 Knockdown did not change the mRNA expression of MMP2 and MMP9 (Fig. S11). As shown in Fig. S12, the knockdown of EPS8 in ABI1-TSV-11-expressing LoVo or SW480 cells do affect ABI1-TSV-11-induced protrusions, and the co-localization of ABI1-isoform-11 with F-actin but not cell morphology. EPS8 knockdown in ABI1-TSV-11-expressing LoVo or SW480 cells can cause ABI1-TSV-11 not to co-locate with F-actin well (arrow indicated). In vitro cytology-based studies have been unable to achieve specific analysis of one or several ABI1-TSVs, so the existing studies could not fully elucidate the molecular mechanism by which ABI1 promotes CRC metastasis, and the application of early diagnosis, prognostic evaluation, and targeted intervention for CRC based on ABI1 is also greatly limited. RNA sequencing is the current gold standard for qualitative and quantitative analyses of transcription variants 46 . In this study, we identified for the first time that elevated expression of ABI1-TSV-11 was related to LsCC lymph node metastasis and shorter OS, and it functions as an independent risk factor of poor prognosis in LsCC. This is consistent with the conclusion in previous studies that ABI1 acts as an oncogene in CRC 8,25-27 . KRAS mutation functions as an important marker of CRC metastasis 47 . Our study found that the expression level of ABI1-TSV-11 has no correlation with KRAS mutation, which is consistent with the conclusion reached by Sebesty' en et al. 48 . It may be suggested that the effect of ABI1-TSV-11 on the metastasis of LsCC is independent of KRAS mutation. On the other hand, because the sample size available in TCGA database is relatively small (55 cases), our analysis cannot exclude the correlation between ABI1-TSV-11 and KRAS. In addition, we need to further explore the role of ABI1-TSV-11 in the cells carrying wild-type KRAS outside LoVo and SW480 cell lines 49 .
Here, we proved that ABI1-TSV-11 overexpression does not affect the proliferation ability of LoVo and SW480 cells. It has been shown that the effect of ABI1 on cell proliferation is mainly achieved through the formation of a protein complex with PI3K (p85 subunit) and activation of the PI3K pathway 17 . As Fig. 6 shows, ABI1-isoform-11 lacks HHR and PxxP proline-rich domain, it cannot structurally combine with the p85 subunit of PI3K, resulting in the failure to activate LoVo and SW480 cell-related proliferation signal pathways.
Cell adhesion mainly occurs in the forms of cell-extracellular matrix and cell-cell adhesion, and a change of cell adhesion is necessary for tumor cells to obtain the ability to metastasize 50 . Change of the adhesion to fibronectin can also directly affect the movement of various tumor cells 51,52 . Recently, Steinestel also proved that ABI1-Y435 site phosphorylation can promote the adhesion of CRC cell CDH1 to fibronectin 8 . Our study found that the overexpression of ABI1-TSV-11 can significantly increase the adhesion of SW480 cells to fibronectin and inhibit the adhesion of LoVo cells to fibronectin, but compared with PBS and 0.02% BSA, the overexpression of ABI1-TSV-11 can improve the adhesion of SW480 and LoVo cells to fibronectin. That is to say, the integrated www.nature.com/scientificreports/ effect of ABI1-TSV-11 overexpression is still to enhance the adhesion of SW480 and LoVo cells. It makes it easier to obtain the traction that drives cell movement, and facilitates retention and implantation in the lung. Migration and invasion represent the key cell biological events of tumor cell metastasis 3 . The increase and/ or phosphorylation of ABI1 can promote the migration and invasion of leukemia cells 9,53 , breast cancer cells 10,29 , ovarian cancer cells 11 and liver cancer cells 12 . In CRC, the knockdown of ABI1 significantly reduced the degradation of extracellular matrix in CRC CHD1 cells, while ABI1 Y435-phosphorylation promoted the formation of lamellar pseudopodia and the invasion of extracellular matrix in CRC cells 8 . Here, we found that, although ABI1-TSV-11 can enhance the migration ability of LoVo and SW480 cells, it can only enhance the invasiveness of LoVo cells without affecting that of SW480 cells (as shown in Fig. 4). LoVo cells originated from metastasis of the upper left clavicle, with high malignancy, strong metastasis, and strong invasion 54,55 , while SW480 cells originated from a primary tumor of rectal adenocarcinoma, with low malignancy and metastatic potential 56 . The difference in the way in which ABI1-TSV-11 overexpression affected the invasiveness of LoVo and SW480 cells also confirms the results obtained in our previous clinical studies, further indicating that ABI1-TSV-11 could be an important candidate target for the treatment of LsCC metastasis.
To date, in vivo study on the role of ABI1 in tumor metastasis has been mainly related to the study of leukemia, breast cancer, and liver cancer 9,10,12,57 , but there are no related reports on CRC. Here, we chose a lung metastasis model of nude mice to evaluate the effect of ABI1-TSV-11 on the metastasis of CRC cells in vivo. The results showed that ABI1-TSV-11 overexpression promoted the process of lung metastasis of LoVo and SW480 cells www.nature.com/scientificreports/ in vivo, which was consistent with the results of in vitro experiments, and further proved that the overexpression of ABI1-TSV-11 enhanced the metastatic ability of LoVo and SW480 cells. ABI1 can regulate actin aggregation and cytoskeleton reconstruction by forming complexes with WAVE2, PI3K, EPS8, and/or N-WASP 17,20,23,24 , and thus play an important role in the metastasis of various malignant tumors including CRC [8][9][10][11][12]27,29,37 . Now we proved that ABI1-isoform-11 interacted only with EPS8 in LoVo and SW480 cells (as shown in Figs. 7, 8), and the co-localization of ABI1-isoform-11 and EPS8 were consistent with the dynamic change of F-actin, specifically with the related trend of lamellar pseudopodium formation and cell extension. Actin dynamics and cytoskeleton changes are the basis of tumor cell adhesion, migration, and invasion 58,59 . The binding of ABI1 and EPS8 could not only affect the localization and/or activity of actin nucleation, regulate the reconstruction of the actin cytoskeleton, and promote the assembly of filopodium structure and actin filaments of human breast cancer cells and mouse melanoma cells 22 , also seal the hook end of EPS8, promote actin capping, and directly induce F-actin-rich structure formation 60 . As shown in Fig. S12, the knockdown of EPS8 in ABI1-TSV-11-expressing LoVo or SW480 cells do affect ABI1-TSV-11-induced protrusions, and the co-localization of ABI1-isoform-11 with F-actin but not cell morphology. Taken these findings together, we believe that ABI1-isform-11 can regulate actin dynamics and cytoskeleton remodeling by interacting with EPS8, thus promoting the adhesion and migration of LoVo and SW480 cells in vitro and accelerating the process of lung metastasis in vivo. However, TCGA database shows that multiple ABI1 TSVs are often expressed in CRC at the same time, and there are potential synergistic or antagonistic effects among different TSVs. Therefore, it is necessary to integrate them with changes in expression, abnormal phosphorylation, and differences in complex formation to systematically elucidate the exact molecular mechanism of ABI1 in CRC metastasis.
In summary, we believe that ABI1-TSV-11 is a specific prognostic factor and therapy target for LsCC, and the current results of this study open up a new field for us to use TSV-specific nucleic acid detection technology 61 and ASO (antisense oligonucleotide) target technology 62 for the diagnostic classification, prognostic evaluation, and targeted treatment of CRC metastasis. qRT-PCR. qRT-PCR was performed as previously described 9 . The primers were listed in Supplement Materials (Table S1) and the amplification was conducted and analyzed by CFX96 (Bio-Rad, Hercules, CA, USA).
Integrins mRNA expression profile of LoVo and SW480 cells. RNA  Establishment of colorectal cancer lung metastasis model. All animal care and handling procedures were were approved by the Ethics Committee of Peking University People's Hospital BALB/c nude mice (6-8 weeks old; Charles River, Beijing, China) were kept in SPF conditions and randomly assigned into four groups (six mice per group). LoVo-ABI1-TSV-11, LoVo-EV, SW480-ABI1-TSV-11, and SW480-EV (each 1 × 10 6 ) cells were injected into the tail veins of nude mice. At 16 weeks after injection, all tested mice were sacrificed with the signs of cachexia appearing, and qualitative and quantitative analyses of lung metastases were conducted by hematoxylin and eosin (H&E) staining, Nano-Zoomer Digital Pathology S360 (Hamamatsu Photonics, Shizuoka, Japan), and ImageJ (National Institutes of Health, USA).
Western blotting and co-immunoprecipitation. Western blot and immunoprecipitation analysis were performed as previously described 9 . The antibodies were listed in Supplement Materials (Table S2).
Statistical analysis. Statistical analysis and graph drawing were performed using SPSS 20.0 (IBM, Chicago, IL, USA) and GraphPad Prism 8.0 (La Jolla, CA, USA). X-tile 3.6.1 (Yale University School of Medicine, New Haven, CT, USA) was used to confirm the most suitable cut-off points to define low and high ABI1-TSV-11 expression. Chi-square test was used to compared two groups, and one-way ANOVA was used to compare three www.nature.com/scientificreports/ or more groups. Kaplan-Meier analysis and log-rank test were used to survival analysis. The Cox proportional hazards regression model was used to univariate and multivariate analysis to determine the effect of ABI1-TSV-11 on OS. P-values < 0.05 were considered to be statistic significant. All p-values correspond to two-sided significance tests.

Approval for animal experiments. The animal experiments were approved by the Ethics Committee of
Peking University People's Hospital and all animal care and handling procedures were performed according to the National Institutes of Health's Guide for the Care and Use of Laboratory Animals. The study was carried out in compliance with the ARRIVE guidelines 2.0 (https:// arriv eguid elines. org).