Distinctive Prognostic Value and Cellular Functions of Osteopontin Splice Variants in Human Gastric Cancer

Background: Osteopontin (OPN) splice variants are identified as predictors of tumour progression and therapeutic resistance in certain types of solid tumours. However, their roles in gastric cancer (GC) remain poorly characterized. The current study sought to assess the prognostic value of the three OPN splice variants (namely OPN-a, OPN-b, and OPN-c) in gastric cancer and their potential functions within gastric cancer cells. Methods: RNA extraction and reverse transcription were performed using our clinical cohort of gastric carcinomas and matched normal tissues (n = 324 matched pairs). Transcript levels were determined using real-time quantitative PCR. Three OPN splice variants overexpressed cell lines were created from the gastric cancer cell line HGC-27. Subsequently, biological functions, including cell growth, adhesion, migration, and invasion, were studied. The potential effects of OPN isoforms on cisplatin and 5-Fu were evaluated by detecting cellular reactive oxygen species (ROS) levels in the HGC-27-derived cell lines. Results: Compared with normal tissues, the expression levels of three splice variants were all elevated in gastric cancer tissues in an order of OPN-a > OPN-b > OPN-c. The OPN-a level significantly increased with increasing TNM staging and worse clinical outcome. There appeared to be a downregulation for OPN-c in increasing lymph node status (p < 0.05), increasing TNM staging, and poor differentiation. High levels of OPN-a and OPN-b were correlated with short overall survival and disease-free survival of gastric cancer patients. However, the low expression of OPN-c was significantly associated with a poor prognosis. Functional analyses further showed that ectopic expression of OPN-c suppressed in vitro proliferation, adhesiveness, migration, and invasion properties of HGC-27 cells, while the opposite role was seen for OPN-a. Cellular ROS detection indicated that OPN-a and OPN-c significantly promoted ROS production after treatment with 5-Fu comparing to OPN-vector, while only OPN-a markedly induced ROS production after treatment with cisplatin. Conclusion: Our results suggest that OPN splice variants have distinguished potential to predict the prognosis of gastric cancer. Three OPN variants exert distinctive functions in gastric cancer cells. Focusing on specific OPN isoforms could be a novel direction for developing diagnostic and therapeutic approaches in gastric cancer.


Introduction
Gastric cancer (GC) is one of the most commonly diagnosed cancers and the leading cause of death worldwide, especially in Far East Asia [1]. According to the statistics provided by the World Health Organization, approximately one million new cases of gastric cancer and more than 8% of deaths worldwide from cancer occurred in 2018. Despite multiple treatment methods including surgery, chemotherapy, radiotherapy, and targeted therapy which have been applied over the past few decades [2], the prognosis of patients with gastric cancer remains poor with survival rates around 30% worldwide [3]. The conventional predictive factors, such as TNM stage, age, sex, and histological type are valuable, but molecular prognostic factors have been an area of intense research. Therefore, it is necessary to exploit novel biomarkers that could accurately predict the outcome to guide clinical management and improve the prognosis of patients with gastric cancer.
As a phosphorylated glycoprotein, Osteopontin (OPN/SPP1) is secreted by various tissues and cells, and has been implicated in physiological as well as pathological processes. Over the last few years, accumulating evidence has shown that aberrant OPN expression is closely associated with tumourigenesis and metastasis in several tumours, including gastric cancer [4,5]. Multiple malignant gastric cancer cell lines overexpress the metastatic related gene osteopontin, and the transfection of this gene into benign tumorigenic gastric epithelial cell lines conveys invasive behaviour. Although several studies have explored the association between OPN expression and the clinical outcome and prognosis of gastric cancer patients, the results regarding the ability of OPN to predict disease progression in patients with gastric cancer emerged controversial [6].
The biological functions of tumour-associated gene products are extensively regulated on posttranscriptional levels. Alternative splice-generation of multiple mRNA products from a single gene is a critical mechanism for generating proteomic diversity. Consistently, the OPN precursor-mRNA (pre-mRNA) is subject to alternative splicing, which generates three splice variants, OPN-a (consists of all exons), OPN-b (lacks exon 5), and OPN-c (lacks exon 4). Recent studies have shown that OPN splice variants are differentially expressed and may have functional heterogeneity in a tumor-specific manner [7][8][9]. For example, OPN-b expression dominates in gliomas, but is hardly detected in breast cancer tissues. Overexpression of OPN-c promotes tumour metastasis in ovarian cancer, but prevents both cell migration and invasion in hepatocellular carcinoma. Based on these findings, we postulate that a specific OPN isoform might become more valuable as a cancer target than OPN itself, since its diverse roles are dependent on the cancer type. However, up to date, few studies have focused on three OPN splice variants in gastric cancer.
In the present study, we aimed to investigate the clinical significance of the OPN splice variants by taking advantage of our gastric cancer cohort. The roles of the three OPN splice variants in gastric cancer cells and in cells' response to chemotherapeutic agents were also assessed.
Absorbance was determined at a wavelength of 540 nm using a spectrophotometer (Elx800; Bio-Tek, Bedfordshire, UK).

In Vitro Invasion Assay
This was carried out as previously reported and modified in our laboratory [12]. Transwell inserts with an 8-µm pore size were coated with 50 mg Matrigel and air-dried. The Matrigel was rehydrated before use. A total of 25,000 cells were seeded to each well and after 48 h, cells that had migrated through the matrix and pores were fixed with 4% formalin, stained in crystal violet, and counted.

In Vitro Cell-Matrix Adhesion Assay
A total of 40,000 cells were added to each well of a 96-well tissue-culture plates previously prepared by coating with Matrigel (5 mg per well). The cells were incubated at 37 • C in 5% CO 2 for 40 min and the medium was then discarded. Nonadherent cells were washed off using PBS buffer. The remaining cells were then fixed in 4% formaldehyde for 5 min. After further washing, cells were stained with crystal violet, and the number of adherent cells was then counted.

Electrical Cell Substrate Impedance Sensing (ECIS)-Based Cell Migration Assay
This method has been widely implemented by our laboratories as previously reported [13]. Briefly, 96-well W96E1 microarrays were used on the ECIS ® Zθ (theta) instrument (Applied Biophysics Ltd., Troy, New Jersey, USA). Cells were added to the wells of the array. Then, confluent lung cancer monolayers in the arrays were electrically wounded (2000 mA for 20 s each), after which the migration of the cells was immediately tracked, again over a range of frequencies (1000 Hz-64,000 Hz). All the experiments were conducted in triplicate.

Measurement of ROS Production
Cellular ROS generation was measured with a Cellular Reactive Oxygen Species Detection Assay Kit (Red Fluorescence) (Abcam, UK). HGC-27 (2.5 × 10 4 cells). Cells which overexpressed OPN splice variants were seeded into 96-well plates and treated with the drugs (cisplatin and 5-Fu, respectively), incubated for 24 h in the dark and ROS Red Dye Working Solution was then added. After incubating at 37 • C for 30 min, the fluorescence intensity increase was monitored at an excitation wavelength of 650 nm and an emission wavelength of 675 nm with the bottom read mode.

Statistical Analysis
Data of gene expression from our gastric cancer cohort were analysed using Minitab (version 14. Minitab Ltd., Coventry, UK). Mann-Whitney U tests or Kruskal-Wallis tests were used to compare gene expression between patient groups. The association of the OPN splice variants with the patient survival was evaluated using Kaplan-Meier survival analysis. The prognostic values of the factors were undertaken by univariate and multivariate analysis of variance. The in-vitro cellular function data were analysed by R programming (version 4.0.5. https://www.r-project.org, accessed on 28 April 2021). Following normality test, Student's t-test was used for the comparison of two groups while ANOVA was used for multiple groups. A p < 0.05 was regarded as statistically significant unless otherwise indicated. The statistical asterisks are: * p < 0.05, ** p < 0.01, and *** p < 0.001.

The Prognostic Value of OPN Splice Variants in Gastric Cancer Patients
The Kaplan-Meier survival curve indicated that gastric cancer patients with high OPN-a expression had worsened OS and DFS rates than gastric cancer patients with low OPN-a expression, although the statistical analysis did not show significance among these groups. OPN-b showed a similar trend to OPN-a. In contrast, a higher OPN-c mRNA level was significantly associated with a better prognosis in patients with gastric cancer (Figure 1).
analysis. The prognostic values of the factors were undertaken by univariate and multivariate analysis of variance. The in-vitro cellular function data were analysed by R programming (version 4.0.5. https://www.r-project.org). Following normality test, Student's t-test was used for the comparison of two groups while ANOVA was used for multiple groups. A p < 0.05 was regarded as statistically significant unless otherwise indicated. The statistical asterisks are: * p < 0.05, ** p < 0.01, and *** p < 0.001.

The Prognostic Value of OPN Splice Variants in Gastric Cancer Patients
The Kaplan-Meier survival curve indicated that gastric cancer patients with high OPN-a expression had worsened OS and DFS rates than gastric cancer patients with low OPN-a expression, although the statistical analysis did not show significance among these groups. OPN-b showed a similar trend to OPN-a. In contrast, a higher OPN-c mRNA level was significantly associated with a better prognosis in patients with gastric cancer ( Figure  1). While decreased OPN-c expression was significantly associated with adverse pathology in gastric cancer (C and F).

Association of OPN Splice variants' Expression with Clinicopathological Features in Gastric Cancer
As shown in Table 2, the mRNA expression levels of the OPN splice variants were significantly higher in gastric cancer tissues than in normal tissues. No significant rela-

Association of OPN Splice variants' Expression with Clinicopathological Features in Gastric Cancer
As shown in Table 2, the mRNA expression levels of the OPN splice variants were significantly higher in gastric cancer tissues than in normal tissues. No significant relationship was found between OPN splice variants' expression and gender. Both the expression of OPN-a and OPN-c were found to increase with differentiation and in tumours with poor clinical outcomes, although these did not reach statistical significance. In addition, increased OPN-a levels were found to be significantly associated with tumour stage, when comparing N0 to N1 + 2 + 3, M0 to M1, and TNM1 + 2 to TNM3 + 4, respectively. However, the expression levels of OPN-c decreased with increasing lymph node status and this reached statistical significance when comparing N0 with N1 + 2 + 3. The multivariate analysis unveiled that OPN-c was significantly correlated with both OS (p = 0.014) and RFS (p = 0.013) (Tables 3 and 4). Likewise, OPN-b was significantly correlated with both OS (p = 0.026) and RFS (p = 0.046) indicated by multivariate analysis. However, OPN-a did not show a correlation with either OS or RFS by multivariate analysis. The univariate analysis showed similar trends.

Effects of OPN Splice Variants on In Vitro Functions of Gastric Cancer Cell Line HGC-27, Respectively
To investigate the expression pattern of OPN isoforms in gastric cancer, we firstly analysed their mRNA levels in gastric cell lines by RT-PCR. We found that OPN splice variants were upregulated in AGS cells, but barely detected in HGC-27 cells (Figure 2A). The expression levels of OPN isoforms were considerably increased in OPN isoforms construct transfected cells, compared with vector control cells in HGC-27 ( Figure 2B). To further clarify the contribution of each OPN splice variant in gastric carcinogenesis, a series of in vitro cell function assays were performed.   Figure 2B).

Effect of OPN Splice Variants on the Proliferation of HGC-27 Cells
In the cell proliferation assay, cells with upregulated individual OPN-a and OPN-b both enhanced cell proliferation as compared with the corresponding vector control, especially OPN-b. Compared with vector control cells, overexpression of OPN-c resulted in an inhibition of cell growth ( Figure 3A).

Effect of OPN Splice Variants on the Proliferation of HGC-27 Cells
In the cell proliferation assay, cells with upregulated individual OPN-a and OPNb both enhanced cell proliferation as compared with the corresponding vector control, especially OPN-b. Compared with vector control cells, overexpression of OPN-c resulted in an inhibition of cell growth ( Figure 3A).

Effect of OPN Splice Variants on the Invasion of HGC-27 Cells
Overexpression of OPN-a in HGC-27 cells showed a remarkable increase in invasion ability. A similar situation was also observed in OPN-b overexpressed cells, while overexpression of OPN-c could significantly decrease the number of migrated cells ( Figure  3D).

Effect of OPN Splice Variants on ROS Generation of HGC-27 Cells Following Treatment with 5-Fu and Cisplatin, Respectively
To confirm that OPN splice variants in HGC-27 cells, following treatment with 5-Fuand cisplatin, induced apoptosis mediated by ROS generation, we used the ROS detection assay kit to examine ROS production. Treatment with 5-Fu caused an increase in the ROS level in HGC-27 cells. Compared with OPN-vector, both OPN-a and OPN-c promoted

Effect of OPN Splice Variants on the Adhesion of HGC-27 Cells
As shown in Figure 3B, OPN-a overexpression in HGC-27 increased in vitro cellular adhesion, while OPN-c overexpression resulted in a remarkable reduction in the adhesion ability of HGC-27 cells.

Effect of OPN Splice Variants on the Migration of HGC-27 Cells
Overexpression of OPN-a in HGC-27 cells showed an increase in post-wound migration as indicated by the ECIS system; however, this did not reach statistical significance. OPN-c overexpression resulted in a slight reduction in the migratory ability of HGC-27 cells ( Figure 3C).

Effect of OPN Splice Variants on the Invasion of HGC-27 Cells
Overexpression of OPN-a in HGC-27 cells showed a remarkable increase in invasion ability. A similar situation was also observed in OPN-b overexpressed cells, while overexpression of OPN-c could significantly decrease the number of migrated cells ( Figure 3D).

Effect of OPN Splice Variants on ROS Generation of HGC-27 Cells Following Treatment with 5-Fu and Cisplatin, Respectively
To confirm that OPN splice variants in HGC-27 cells, following treatment with 5-Fuand cisplatin, induced apoptosis mediated by ROS generation, we used the ROS detection assay kit to examine ROS production. Treatment with 5-Fu caused an increase in the ROS level in HGC-27 cells. Compared with OPN-vector, both OPN-a and OPN-c promoted ROS production after treatment with 5-Fu. However, cisplatin showed a diverse effect on ROS generation comparing to 5-Fu, and only OPN-a significantly decreased the level of ROS. Cisplatin and 5-Fu did not show an effect on ROS level in the OPN-b group and its vector-only control (Figure 4).

Effect of OPN Splice Variants on Protein Structural Features
To investigate how the splice variants affect the protein structure, we first aligned the three variants. As shown in Figure 5A, the mismatch mainly occurred in the N-side of the protein sequences. OPN-b and OPN-c have different missing fragments compared to OPN-c which has the full sequence. The protein structure prediction showed that in the secondary structure, OPN-a forms 4 helixes and 8 strands ( Figure 5B), OPN-b forms 5 helixes and 6 strands ( Figure 5C), while OPN-c forms 3 helixes and 6 strands at different sites ( Figure 5D). ROS production after treatment with 5-Fu. However, cisplatin showed a diverse effect on ROS generation comparing to 5-Fu, and only OPN-a significantly decreased the level of ROS. Cisplatin and 5-Fu did not show an effect on ROS level in the OPN-b group and its vector-only control (Figure 4). To investigate how the splice variants affect the protein structure, we first aligned the three variants. As shown in Figure 5A, the mismatch mainly occurred in the N-side of the protein sequences. OPN-b and OPN-c have different missing fragments compared to OPN-c which has the full sequence. The protein structure prediction showed that in the secondary structure, OPN-a forms 4 helixes and 8 strands ( Figure 5B), OPN-b forms 5 helixes and 6 strands ( Figure 5C), while OPN-c forms 3 helixes and 6 strands at different sites ( Figure 5D).

Discussion
OPN is one of the known proteins that have been considered to correlate with tumourigenesis and progression of cancers, including gastric cancer [14][15][16][17][18]. Many recent studies show that OPN splice variant expression in malignancy is tissue-type specific and may have functional heterogeneity [19][20][21]. However, very few specific reports have elaborated the relationship between the expression of OPN isoforms and gastric cancer. In order to characterize the three OPN isoforms in gastric cancer, we firstly focused on the prognostic value of OPN splice variants in gastric cancer patients. Our results demonstrate that high OPN-a or OPN-b expression correlates with poor prognosis, which is opposite to OPN-c. Moreover, the increased expression of OPN-a, but not of OPN-b and OPN-c, is associated with adverse clinicopathological features. Together with these findings in gastric cancer tissues, we postulate that the abnormal alternative splicing, which results in elevated OPN-a expression, may intensively occur during the progression of GC, while OPN-c may prevent adverse clinical outcomes. A previous study also shows that OPN-a is highly expressed in gastric cancer, which is in line with our findings [22]. However, that study also suggests that OPN-c is highly expressed in gastric cancer, which is in contrast with our result. The inconsistency of the findings may be due to sampling sources, measurement sensitivity, or data normalization, which require further investigation using a large-scale cohort and more robust measurement and data analysis approach.
We next clarified the diverse malignant phenotypes of three OPN splice variants through in vitro cell function assays. We found that over-expression of OPN-a was linked to the elevation of proliferation, adhesion, invasion, and migration of HGC-27 cells in vitro. In contrast, following genetic overexpression of OPN-c, the properties of cellular proliferation, adhesion, invasion, and migration in HGC-27 cells were reduced. The results indicate that OPN splice variants may have functional heterogeneity in HGC-27 cells. As the heterogeneity of OPN splice variants has previously been described in other cancer types [23][24][25], our study further extends the observation of this phenomenon into gastric cancer.
It has been demonstrated that OPN can promote cell survival by negatively regulating apoptosis in response to stress conditions, including exposure to anticancer agents; OPN splice variants might play distinct roles in this regard. A previous study suggests that acquired up-regulation of OPN-b and c isoforms might prevent conventional chemotherapy drug (Daunorubicin, Cytarabine and Idarubicin)-induced apoptosis in acute myeloid leukemia (AML) cells by upregulating the expression of AKT, VEGF, STAT3, CXCR4, and IL-6 [26]. In PC3 prostate cancer (PCa) cells, a study showed that OPN-c or OPN-b overexpression mediates resistance and cell survival features in response to docetaxel-induced cell death [27]. However, little evidence indicates the connection between OPN splice variants and chemotherapy drug-induced apoptosis in gastric cancer cells. Reactive oxygen species (ROS) production is considered a critical stressor that causes cell death, especially through the induction of apoptosis [28,29]. Herein, we investigated the association between the expression of three main OPN isoforms and chemotherapeutic drug-induced apoptosis in the HGC-27 cell line, which was evaluated by detecting ROS generation. Our study indicates that both OPN-a and OPN-c can effectively promote ROS production which is induced by 5-Fu when compared with corresponding controls, especially OPN-a. Meanwhile, in HGC-27 cells treated with cisplatin, overexpressed OPN-a significantly inhibits ROS production and OPN-c does not show an effect on ROS level.
It is worth noting that the changed trends of ROS in the two groups of drugs are the opposite. In Recent studies suggest that ROS seems to be a double-edged sword in the mediation of cancer [30][31][32]. On the one hand, ROS generation may promote tumor initiation and progression by activating oncogenes or inactivating tumor suppressor genes, regulating VEGF signal transduction pathways to promote angiogenesis, and activating AKT and PKC to induce cell invasion or metastasis [33,34]. However, on the other hand, ROS can increase the sensitivity of cancer cells to various death-inducing signalling pathways (such as apoptosis, autophagy, necrosis, etc.), thereby inhibiting tumor progression [35]. Therefore, the biological role of ROS in cancer is complicated. At present, our preliminary experimental results show that the expression of OPN-a and OPN-c can enhance the anti-tumor effect of 5-Fu on HGC-27 cells, while the expression of OPN-a can enhance the anti-tumor effect of cisplatin on HGC-27 cells. In the process, they may affect the balance between cellular ROS production and clearance. In the future, we will further elucidate possible molecular mechanisms by expanding cell lines, setting drug concentration gradients, and detecting signalling transduction.
Differential effects of the three splice variants may be due to the distinctive functions of the spliced exons that individual OPN isoforms contain. For example, exon 4 encodes two glutamine residues essential for transglutaminase crosslinking. As exon 4 is absent in OPN-c, but not in OPN-a and OPN-b, OPN-c has no specific functions exerted by exon 4. In addition, deletion of exon 4 may alter the pattern of post-translational modifications (PTMs), resulting in functional modifications [36]. Protein prediction also suggests that the three OPN splice variants form different secondary structures. That might explain why these isoforms of OPN have cell-type-specific patterns.
Although the present study, by using a large gastric cancer cohort, has shown the importance and the differential role of the three OPN variants in this rather lethal cancer, the study has its limits. First, the study was unable to carry out a protein-based investigation, due largely to the lack of suitable antibodies that can clearly distinguish the three variants. Fortunately, the minor differences of the variants can be distinguished by gene amplification studies such as the one carried out here. It is expected that antibodies of suitable quality will be developed in the near future to allow such studies to be carried out. Secondly, the study was unable to provide experimental information from a suitable in vivo tumour model or models. Information from in vivo models would provide an important link between the in vitro and clinical findings. It would be vital in the future to unveil the crystal structures of the OPN protein isoforms. This would allow in depth mechanistic investigation into the reasons behind the distinctive roles of the OPN variants in cancer and, importantly provide a suitable platform in search for druggable compounds. Furthermore, as OPN is considered an inflammatory sensing factor, it would be interesting to explore how OPN isoforms are involved in inflammation-mediated tumorigenesis in the stomach. With the development of antibodies specific to individual OPN isoforms, it may be worth examining whether the levels of the secretory OPN isoforms in patient plasma could be used to predict the progression of gastric cancer [37]. Although it is a challenge, appropriate in vivo gastric cancer models would be a prerequisite before we attempt to pinpoint the actual mechanisms of differential functions of OPN isoforms in vivo. Collectively, the study has provided some vital findings for the highly interesting variants of OPN and demonstrated that some of the variants are important clinical, prognostic indicators in gastric cancer. The study has laid the foundation for future work to be carried out, including searching for in-depth mechanisms of distinctive actions and for potential therapeutic targeting methods.

Conclusions
In conclusion, we believe that this is the first work that shows that OPN splice variants have distinctive clinical and biological roles in gastric cancer. Exploring the implications of OPN splice-derived protein isoforms in gastric cancer could open up a novel avenue for the development of diagnostic and therapeutic approaches in patients with tumor malignancies.
Author Contributions: W.G.J., Y.C. and J.J. conceived the study, S.J. and J.J. sourced the cohort materials, C.H. and Y.C. carried experimental work, W.G.J., J.L. and Y.C. conducted statistical analyses, all participated in manuscript preparation. All authors have read and agreed to the published version of the manuscript. Data Availability Statement: Data reported in this study can be obtained from the authors upon reasonable request.

Conflicts of Interest:
The authors declare no conflict of interest.