In vitro and clinical data analysis of Osteopontin as a prognostic indicator in colorectal cancer

Abstract Osteopontin (OPN) has been shown to promote colorectal cancer (CRC) progression; however, the mechanism of OPN‐induced CRC progression is largely unknown. In this study, we found that OPN overexpression led to enhanced anchorage‐independent growth, cell migration and invasion in KRAS gene mutant cells but to a lesser extent in KRAS wild‐type cells. OPN overexpression also induced PI3K signalling, expression of Snail and Matrix metallopeptidase 9 (MMP9), and suppressed the expression of E‐cadherin in KRAS mutant cells. In human CRC specimens, a high‐level expression of OPN significantly predicted poorer survival in CRC patients and OPN expression was positively correlated with MMP9 expression, and negatively correlated with E‐cadherin expression. Furthermore, we have found that 15 genes were co‐upregulated in OPN highly expression CRC and a list of candidate drugs that may have potential to reverse the secreted phosphoprotein 1 (SPP1) gene signature by connectivity mapping. In summary, OPN is a potential prognostic indicator and therapeutic target for colon cancer.


| INTRODUCTION
Colorectal cancer (CRC) is the third most common malignancy worldwide, it has been reported that about 1.3 million cases of CRC are diagnosed every year. 1 The tumour development involves multi-steps process over years and mostly occurs in the ageing population. 2 Despite traditional surgery, chemotherapy and radiotherapy, the prognosis of CRC remains poor. Some specific gene mutation leads to tumour aggression in CRC. KRAS gene mutation is commonly observed in the early stages of the adenoma-adenocarcinoma sequence in human CRC development. 3,4 KRAS mutations occur in 30%-40% of all cases of CRC. However, to date, specific therapeutic agents against KRAS-mutated CRC have not been developed. 4 The potential correlation between KRAS and other oncogene remains to be further investigated. Osteopontin (OPN) is a 34 kD integrin-binding glycophosphoprotein, which plays an important role in cancer progression and is considered as a potential biomarker for cancer prognosis. 5,6 It is encoded by SPP1 gene that has five isoform variants, which include OPN-a, OPN-b, OPN-c, isoform 4 and isoform 5. 5 OPN has been previously found to be highly expressed in colon cancer cells or tissues than that in normal intestinal epithelial cell line or corresponding normal colon tissues. 7 Up-regulation of OPN increased cell motility in vitro, tumorigenesis and angiogenesis in vivo, 8 and predicted patient survival in colon cancer. 9 Therefore, OPN has been recognized as a potential target for human colon cancer. 10 However, the upstream and downstream effectors in OPN overexpression (OPN OE) of colon cancer are still largely unknown.
In this study, we investigated the role of OPN overexpression related to phenotypic changes using KRAS mutant and KRAS wildtype colon cancer cells. We further identified potential downstream targets involved in OPN overexpression-mediated colon cancer progression. In addition, we also carried out the connectivity mapping analysis to identify potential therapeutic drug candidates for OPNoverexpressing colon cancer.

| Cell culture
Isogenic pair of Duke Stage C colorectal cancer cells DLD1 with DKS8 were selected. DLD1 expresses heterozygous Kras G13D mutation and the isogenic cell, DKS8 had the Kras G13D mutation knocked out at its endogenous locus. 11 The parental (DLD1), normal epithelial (FHC) and Phoenix-AMPHO cell lines purchased from ATCC, while the isogenic cell lines were gifts from Prof. Senji Shirasawa's Laboratory. These cell lines were grown in cell culture media DMEM/F-12 (Gibco: Cat. no. 1320033) with 10% foetal bovine serum and 1% penicillin/streptomycin. FHC medium prepared with additional 10 mmol/L HEPES, 10 ng/mL cholera toxin, 5 ng/mL transferrin, 5 ng/mL insulin and 100 ng/mL hydrocortisone.

| Retroviral Infection and stable cell line transfection
The p3XFLAG-CMV-7.1 (Sigma, E7533) was ligated to the N-terminal of OPN-c gene of pcDNA3-OPN-V5, which was a gift from Steven Johnson (Addgene plasmid # 11617). 12 Then it was inserted into pBABE-puro which was a gift from Hartmut Land, Jay Morgenstern and Bob Weinberg (Addgene plasmid # 1764). 13 Phoenix-AMPHO cells were used to produce the retrovirus to transduce the DLD1 and DKS8. The expression level of OPN in these stable cell lines was determined with Western blotting.

| Total protein quantification and Western blotting
The cell line samples were homogenized with ice-cold RIPA lysis buffer that was added with protease inhibitor (Complete EDTA-free, #10634200, Roche) and phosphatase inhibitor (PhosSTOP, #04906837001, Roche) and centrifuged at 20 0009 g for 30 minutes at 4°C. Supernatants were collected and kept in À80°C. BCA method was used to quantify the protein concentration required for Western blot sample loading (Pierce BCA Protein Assay Kit, #23225). All samples were dissolved in LDS sample buffer and reducing agent (Life Technologies) and heated for 5 minutes at 95°C. An equal concentration of proteins was electrophorized and separated with TGX Stain-Free TM FastCast TM Acrylamide Kit, 12% and transferred to a 0.2 lm nitrocellulose membrane (#IB401002) membranes. Blocking was carried out with 5% non-fat milk, 0.1% Tween 20 in PBS for non-specific binding for an hour at room temperature.

| Soft agar colony formation assay
Soft agar assays were carried out in a 6-well plate, each sample in triplicate and colonies were counted on day 15 after plating. The base layer consisted of 2 mL with a final concentration of culture medium and 0.6% low melting temperature agarose (Lonza). Next, the respective cells were seeded with culture medium containing 0.6% agar to result in a final concentration of was 0.3% agarose with 5 x 10 4 cells.
A further 500 lL culture medium only was added above the congealed middle agar layer on the next day. The colonies were captured at five random fields at 409 magnification with inverted phase microscope (Leica), measured using the Leica software. The scale was set at ImageJ to select colony size more than 0.45 mm.

| Proliferation assay
Cell proliferation was measured using a bright field image label-free high-content time-lapse assay system (IncuCyte Zoom system; Essen BioScience) according to the manufacturer's instructions. In brief, for the proliferation assay, equal numbers of cells (1 9 10 3 cells/well) were seeded on to 96-well plate in the 200 lL culture medium with supplements or agents, and per cent cell confluence was then continuously measured using the IncuCyte system over a 5-day period.

| Wound healing assay
Each cell line group consists of OPN OE and control (pBABE puro).
The initial seeding density was 2 9 10 4 and 4 9 10 4 cells/100 lL growth medium in a 96-well plate (Image Lock plates, Essen Bioscience), respectively, for DLD1 and DKS8 cells, and left to reach fully confluence before wounding with a wound maker (Essen Bioscience). The plates were replaced with the new medium (200 lL) before further incubated and visualized for 60 hours at 37°C in the IncuCyte. Wound width was measured every two hours and images were captured at set time 4009 magnification. This experiment was repeated three times.

| Extraction of clinical and microarray gene expression data from colon cancer patient datasets
Two colon cancer patient datasets available in the Gene Expression Omnibus database were included in this study as previously described. 14 Microarray gene expression data were retrieved from the data matrices and R scripting was used to extract the expression values of genes of interest and the clinical data from the data matrices.
The median expression value was used as a cut-off point for high and low levels expression for survival analysis. The details of the data extraction method and R script referred to the previous publication. 15

| Connectivity mapping
Gene expression connectivity mapping was performed using the algorithms in the framework of Statistically Significant Connection's Map (sscMap) 16,17 to identify candidate small molecule compounds that may inhibit the expression of genes that are co-regulated in OPN-overexpressing colon cancer. OPN and 15 co-regulated genes (see Section 3.5 below) formed a gene signature for connectivity mapping. The compiled gene signature was then fed to the Java application sscMap as a query signature, and its association with a large collection of reference gene expression profiles were compared. In this study, we used the reference gene expression profiles recently released with the QUADrATiC system, 18 covering data for over 1000 FDA approved drugs from the LINCS database (https://c lue.io/). The gene signature perturbation procedure, 19 which increases the specificity of the output results, was applied. All the small molecular compounds, that were negatively associated with the OPN Gene Expression Signature, were sorted and ranked by their P-value, perturbation stability and standardized connection score. As we are only interested in inhibitory compounds, one-tailed P-values were used. To effectively control false positive discoveries, a stringent P-value threshold was set as 1/N = 1/1432 = .0007, where N = 1432 is the number of FDA drugs screened here. Any drugs with a P-value smaller than the set threshold are considered statistically significant (those above the blue line in Figure 1). These criteria control the expected number of falsely significant drugs at 1, as previously described and justified for multiple testing issues. 20,21 Given the number of significant drugs as 95 in this study, the overall false discovery rate of our results is approximately 1% (1/95). The top 20 small molecules were searched together with colon cancer using the PubMed search engine to identify research articles that have described their effects on the treatment of colon cancer.  Table 1. The full results of connectivity mapping to all 1432 drugs can be found in Table S1 closure assay ( Figure 2C) and transwell migration assay ( Figure 2D cells also significantly increased its ability to form colony in soft agar assay ( Figure 2F).

| Overexpression of OPN in colon cancer cells with wild-type KRAS
As DLD1 contains mutant KRAS while it has an isogenic KRAS wildtype DKS8 generated previously, 22 FLAG-tag OPN was overexpressed in DKS8 cells ( Figure 3A) and its phenotype was investigated. As shown in Figure 3, overexpression of OPN again did not significantly increase cell proliferation in DKS8 cells ( Figure 3B).
OPN overexpression did not significantly increase cell migration in wound closure assay ( Figure 3C), but slightly and significantly increased cell migration in transwell migration assay ( Figure 3D) and cell invasion in transwell invasion assay ( Figure 3E). Overexpression of OPN did not significantly increase anchorage-independent growth of DKS8 cells ( Figure 3F). F I G U R E 2 Phenotype changes of OPN overexpression in DLD1. A, Western blot analysis of FLAG-tag OPN level in OPN OE and control group, b-actin was used as a control. B, Cell proliferation was recorded for 7 d using the IncuCyte instrument, mean AE SD (error bars, n = 9). C, Cell migration was measured by wound closure over 72 h using the IncuCyte, mean AE SD (error bars, n = 9). D, The migration rates determined and analysed after 48 h with Boyden Chamber assay. Images are representative sections of one well at 20x. E, The invasion rates determined and analysed after 48 h with Boyden Chamber assay. Images are representative sections of one well at 20x. F, Cells were seeded in a 0.6% agar/growth medium layer and its colony formation was measured after 15 d of cultivation. The graphs indicate the mean AE SD, of triplicate wells counted using ImageJ software and the images are representative sections of one well at 40x. "*," "**" and "***" indicate P < .05, P < .01 and P < .001, respectively, and statistically analysed with T-test

| The prognostic significance of OPN overexpression in colon cancer
To investigate whether OPN expression is a prognostic marker in colon cancer, two colon cancer datasets in GEO database previously described 13,14 with available data on OPN expression, disease-specific survival were included in this study, namely GSE14333 23 and GSE17538. 24 In GSE14333 colon cancer patient cohort with available clinical information (n = 226), we found that patients whose tumour expressed a high-level expression of OPN had a mean survival time of 74.4 months (95% CI = 64.6-84.2 months), which was significantly shorter compared to those patients whose tumour expressed a low level of OPN, who had a mean survival time of 120.8 months (95% CI = 110.3-131.3 months, P = .004; Figure 5A).
In GSE17538 colon cancer patient cohort with available clinical information (n = 232), we found that patients whose tumour expressed a high-level expression of OPN had a mean survival time of 78.6 months (95% CI = 66.2-91.0 months), which was significantly shorter compared to those patients whose tumour expressed a low level of OPN, who had a mean survival time of 113.5 months (95% CI = 102.6-124.5 months, P < .001; Figure 5B). Our results confirm that OPN is a potential prognostic marker in colon cancer patients. F I G U R E 3 Phenotype changes of OPN overexpression in DKS8. A, Western blot analysis of FLAG-tag OPN level in OPN OE and control group, b-actin was used as a control. B, Cell proliferation was recorded for 7 d using the IncuCyte instrument, mean AE SD (error bars, n = 9). C, Cell migration was measured by wound closure over 72 h using the IncuCyte, mean AE SD (error bars, n = 9). D, The migration rates determined and analysed after 48 h with Boyden Chamber assay. Images are representative sections of one well at 20x. E, The invasion rates determined and analysed after 48 h with Boyden Chamber assay. Images are representative sections of one well at 20x. F, Cells were seeded in a 0.6% agar/growth medium layer and its colony formation was measured after 15 d of cultivation. The graphs indicate the mean AE SD, of triplicate wells counted using ImageJ software and the images are representative sections of one well at 40x. "*," "**" and "***" indicate P < .05, P < .01 and P < .001, respectively, and statistically analysed with T-test we also found that OPN expression was negative correlated with Ecadherin expression (GSE14333: r = À.355, P < .001 and GSE17538: r = À.165, P = .012) and was positive correlated with MMP9 expression (GSE14333: r = .442, P < .001 and GSE17538: r = .484, P < .001), consistent with our findings in vitro in DLD1 cell line.

| Identification of small molecules that can reverse OPN gene signature
Using connectivity mapping, OPN and the top 15 OPN co-regulated genes, we have identified 53 small molecules that could potentially reverse colon cancer OPN gene (SPP1) signature passing the perturbation stability test (Table 1) among 1432 candidate drugs. The full results of all 1432 drugs are attached in " Table S1." The top five small molecules are bendroflumethiazide, calcitriol, benzoic acid, memantine and sibutramine.

| DISCUSSION
In this study, we established OPN stable overexpressing cells using  the most active metabolite of vitamin D, has significant anti-neoplastic activity in preclinical models. There are some well-studied mechanisms such as growth inhibition and accumulation in G0-G1, associated with transcriptional activation of CDK inhibitors p27Kip1 and/or p21Waf1 as well as some other mitogenic signals, induction of apoptosis, and inhibition of invasiveness and angiogenesis have also been reported. 34 Some reports indicated that calcitriol would enhance radiation sensitivity in colorectal cancer regulated by the epithelialmesenchymal transition, for example, in vitro experiment trial of DLD1 and HCT116, 24 hours calcitriol pre-treatment enhanced the radiation sensitivity by 2.3-and 2.6-fold. 35 Pre-clinical and epidemiological studies have promoted the anti-tumour effect of calcitriol, particularly against colorectal cancer which involved in anti-proliferation, pro-differentiation, pro-apoptosis, anti-angiogenesis, immune modulation and miR regulation. 36 A number of genes are recognized to contain functional vitamin D response elements, which also include OPN. 34,37 Survival from colorectal cancer has been reported to be positively associated with vitamin D status and OPN expression level, respectively. 5,38 The potential effect of calcitriol on OPN gene signature in colorectal cancer development warrant further investigation.

This study was supported by the University of Macau Multi-Year
Research Grant (MYRG2015-00065-FHS) to HFK.

CONFLI CT OF INTEREST STATEMENT
The authors declare no competing financial interests.