Robo2 Predicts a Better Prognosis and Inhibits Malignant Behavior in Vivo in Pancreatic Ductal Adenocarcinoma

Pancreatic ductal adenocarcinoma (PDAC) is a lethal malignancy with an extremely poor prognosis and a high mortality rate. Genome-wide studies have shown that the SLIT/ROBO signaling pathway plays an important role in pancreatic tumor development and progression. However, the effect and mechanism of ROBO2 in the progression of pancreatic cancer remains largely unknown. In this study, real-time polymerase chain reaction (RT-PCR) and western blot analyses were adopted to evaluate the expression level of ROBO2 and proteins in pancreatic cell lines. Cell migration and invasion and cell proliferation were conducted in AsPC-1 and MIA PaCa-2 cell lines. RNA sequencing and western blot were undertaken to explore the mechanisms and potential targeted molecules. ROBO2 expression in tumor tissues was evaluated by immunohistochemistry in 95 patients. in patients with pancreatic cancer. Our study indicates that ROBO2 acts as a tumor suppressor gene and can potentially be a therapeutic target for pancreatic cancer. Protein chip and tissue microarray of ROBO2 may be used to predict the survival of patients with PDAC.


Introduction
Pancreatic ductal adenocarcinoma (PDAC), commonly known as pancreatic cancer, is becoming the main cause of cancer-related death because of its poor prognosis [1,2]. The di culty of early diagnosis and the fact that nearly 80-85% of patients have no opportunity to undergo surgery at the time of diagnosis are major obstacles to long-term survival [3]. Carbohydrate antigen 19 − 9 is the most important tumor marker in PDAC [4]. Therefore, elucidation of the molecular mechanisms underlying the development and progression of PDAC and identi cation of additional promising biomarkers may improve the prognosis.
The roundabout (ROBO) family comprises single-pass transmembrane proteins belonging to the immunoglobulin superfamily. The extracellular domain of ROBO2 contains ve immunoglobulin-like domains followed by three bronectin domains as well as an intracellular region that consists of four conserved cytoplasmic domains. ROBO2 plays important role in many tissues during development and particularly in the nervous system. Recent studies have demonstrated that ROBO2 is dysregulated in various cancers such as gastric cancer [5], acute myeloid leukemia [6], colorectal cancer [7], prostate cancer [8], ovarian cancer [9], and breast cancer [10]. However, the exact effect of ROBO2 in PDAC remains unclear.
In the present study, we explored the functions of ROBO2 alterations in tumor progression and the possible underlying mechanisms. As far as we know, this is the rst study to explore the biological roles of ROBO2 in PDAC.

Materials And Methods
Clinical specimens and immunohistochemistry In the aggregate, 95 tumor tissues were obtained from Peking Union Medical College Hospital (PUMCH) from November 2008 to September 2015. This study was authorized by the ethics committee of PUMCH and all patients provided written informed consent. The tissue microarray was constructed as described previously [11]. ROBO2 expression level was evaluated by immunohistochemistry using rabbit antihuman ROBO2 polyclonal antibodies at 1:300 dilution (ab75014; Abcam, Cambridge, UK). Two pathologists who were blinded to the clinicopathological data evaluated the sections independently. In the case of discrepancies, joint evaluation was performed to reach a consensus. The H-score, a combination of the percentage of positively stained cells and the staining intensity, was applied for further evaluation. The H-score ranged from 0 to 300, and the optimal cutoff value for the H-score re ecting a high/low ROBO2 expression level was de ned by the largest Youden's index within the receiver operating characteristic curve. The clinicopathological data, including the patients' characteristics, tumor status, and complete prognosis date, are summarized in Table 1 Transfection MIA PaCa-2 and PANC-1 cells were transfected with siRNA targeting ROBO2 and a scramble control siRNA. Transient transfections were performed using Lipofectamine 3000 (Thermo Fisher Scienti c) according to the manufacturer's protocol. MIA PaCa-2 and ASPC-1 cells were seeded in 6-well plates and cultured until the con uence reached 60-70%. They were then transfected with an ROBO2-knockdown lentivirus (termed SH), scrambled control lentivirus (termed NSH), ROBO2 overexpression lentivirus (termed OE), or negative control lentivirus (termed NOE). Lentivirus constructions of ROBO2 knockdown and overexpression were purchased from GenePharma (Shanghai, China). Quantitative real-time polymerase chain reaction (qRT-PCR) and western blot assays were performed to evaluate the speci city and e ciency of the transfections.
Next, we washed the membranes four times for 10 minutes each in TBS-T at room temperature and then incubated with secondary antibodies for 1.5 hours in 5% non-fat milk at room temperature. Finally, the membranes were washed four times for 10 minutes each in TBS-T and visualized by exposure to an enhanced chemiluminescence system.
Wound healing and Transwell migration and invasion assays A wound healing assay was used to examine the ability of cell migration. After the cells reached 80% to 90% con uence in the six-well plates, they were wounded by scratching with a 200-μl disposable pipette tip and washed twice with phosphate-buffered saline (PBS). The wounds were photographed at 0, 24, and 48 hours with a DFC300 FX microscope (Leica, Wetzlar, Germany), and their widths were quantitated with ImageJ software (National Institutes of Health, Bethesda, MD, USA).
Transwell migration and invasion assays were performed in 24-well Transwell plates (Corning) inculuding a polycarbonate membrane (8-μm pore size). About 1 × 10 4 cells were plated in the top chamber after transfection. Medium containing 10% PBS was added to the lower chambers. After incubation at 37℃ for 36 hours, the non-invading cells were gently removed with a cotton swab, and the penetrated cells were xed in methanol for 20 minutes and stained with hematoxylin and eosin for 10 and 5 minutes, respectively. Cell numbers were counted under a DFC300 FX microscope (Leica).
Colonial formation 500 MIA PaCa-2 or AsPC-1 cells were plated in 6-well plates. And two weeks later, the cells were xed in methanol for 20 minutes and stained with hexamethylpararosaniline for 10 minutes.

RNA sequencing
Total RNA was extracted from samples using TRIzol, and the RNA quality was determined by examining A260/A280 with a Nanodrop™ One C spectrophotometer (Thermo Fisher Scienti c). The kit eliminates duplication bias in PCR and sequencing steps by using a unique molecular identi er of eight random bases to label the pre-ampli ed cDNA molecules. All data were analyzed according to the manufacturers' instructions by Seqhealth Technology Co., Ltd. (Wuhan, China). Differentially expressed genes were then identi ed by a false discovery rate-corrected p-value cutoff of 0.05 and fold-change cutoff of 2. Gene ontology analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis were performed.

Animal experiments
All animal experiments were approved by the PUMCH Animal Care and Use Committee. MiaPaCa-2 cells stably transfected with shROBO2 or shNC and AsPC-1 cells stably transfected with oeROBO2 or oeNC were subcutaneously injected into the left back of 5-week-old female BALB/c mice (5 × 10 6 cells in 150 μL of PBS). The tumor size was measured twice a week, and the volume was calculated using the following formula: volume (mm 3 ) = 1/2 × length × width 2 The animals were euthanized 4 to 5 weeks later.

Statistical analysis
Statistical analysis and graph representations were performed using SPSS v. 22.0 software (IBM Corp., Armonk, NY, USA) and GraphPad Prism 7 Software (GraphPad, San Diego, CA, USA), respectively. Each experiment was repeated at least three times. Continuous variables are presented as mean ± standard deviation. Categorical data were compared using the chi-square test. The level of statistical signi cance was set at a P value of <0.05.

Downregulation of ROBO2 expression in PDAC
The pancreatic cell line HPNE and six PDAC cell lines were used to investigate ROBO2 expression by western blot and RT-PCR ( Figures 1A, B). HPNE cells had the highest ROBO2 expression, followed by PANC-1 and MIA PaCa-2 cells. AsPC-1 and BxPC-3 had lower ROBO2 expression. Based on these results, PANC-1, MIA PaCa-2, and AsPC-1 cells were selected for further analysis. ROBO2 knockdown (ROBO2-SH) and overexpression (ROBO2-OE) cell lines were established by transfection of MIA PaCa-2 and AsPC-1 cells. Transfection e cacy in the ROBO2-SH and ROBO2-OE cells lines was compared with that in the control cell lines and con rmed by western blot and RT-PCR (Figures 1C, D).

Restraint of pancreatic cell proliferation in vitro
CCK-8 assays were applied to investigate the roles of ROBO2 on proliferation. The proliferation assay showed that ROBO2 downregulation promoted the cell proliferation capacity compared with NSH in MIA PaCa-2 stable cell lines, whereas ROBO2 overexpression signi cantly inhibited proliferation in AsPC-1 cells (Figure 2A). The clonal formation assays showed similar results ( Figure 2B).

Inhibition of pancreatic cancer cell migration and invasion in vitro
The migration and invasion abilities of ROBO2-SH and ROBO2-OE cell lines were evaluated by wound healing and Transwell migration and invasion assays in vitro. Wound healing assays revealed that ROBO2 knockdown signi cantly shortened the wound healing time in MIA PaCa-2 cell lines; however, ROBO2 overexpression increased migratory activity in AsPC-1 cell lines compared with control cells ( Figure 3A). Transwell migration and invasion assays were performed to further validate these results. ROBO2 knockdown signi cantly promoted the migration and invasion of MIA PaCa-2 cells, and ROBO2 overexpression decreased the migration and invasion ability of AsPC-1 cells (Figures 3B, C). We also further investigated the role of ROBO2 in epithelial-mesenchymal transition (EMT). ROBO2 downregulation increased the expression of vimentin and N-cadherin, and ROBO2 upregulation increased the expression of E-cadherin ( Figure 3D).

Expression level of ROBO2 in tumor tissues and its association with a better prognosis in patients with PDAC
We investigated the level of ROBO2 expression in 95 samples by immunohistochemistry (Figures 4A, B).
First, we evaluated the association between the expression of ROBO2 and clinicopathologic features. The clinicopathological parameters are summarized in Table 1. ROBO2 expression was not signi cantly correlated with age, sex, tumor size, location, or T or N stage.
Next, we evaluated the association between ROBO2 expression and the prognosis. Kaplan-Meier analysis showed that patients with high ROBO2 expression had signi cantly better disease-free survival and overall survival than those with low ROBO2 expression ( Figures 4C, D). Univariate analysis and multivariable analysis revealed that the ROBO2 expression level was positively associated with diseasefree survival and overall survival. Overall, our data indicated that the ROBO2 expression level might be an independent predictor of overall survival.

Inhibition of PDAC cell proliferation invivo
We used a xenograft tumor model to investigate the effect of ROBO2 on cell proliferation in vivo. Stably transfected MIA PaCa-2 and AsPC-1 cells were injected into nude mice. After observation for 4 weeks (MIA PaCa-2 group) and 5 weeks (AsPC-1 group), the results showed that ROBO2 knockdown promoted tumor growth while ROBO2 overexpression inhibited tumor growth ( Figures 5A, B, C). ROBO2 mRNA and protein expression were detected by RT-qPCR and immunohistochemistry ( Figures 5D, E). In summary, these PDAC cell line xenografts indicated that ROBO2 is remarkably correlated with the proliferation ability of PDAC cells in vivo.

RNA sequence and mechanism
To further investigate the potential mechanisms underlying how ROBO2 affects the biological behavior of pancreatic cancer, mRNA expression pro ling was performed in ROBO2-overexpression cell lines ( Figure   6A). Gene ontology analysis and KEGG analysis of differentially expressed genes were performed. Genes regulated by ROBO2 were involved in biological process, cellular component, and molecular functions ( Figures 6B, C, D). Western blot was performed to further investigate the effect of ROBO2 on the extracellular matrix (ECM), revealing that the level of ECM protein 1 (ECM1) in the ROBO2-OE was signi cantly lower than that in the ROBO2-NC, and vice versa in the ROBO2-SH ( Figure 6E). Overall, our super cial results suggest that ROBO2 might inhibit PDAC progression by inhibiting ECM1; however, further experiments are needed to con rm these ndings.

Discussion
Because of delayed diagnosis and early metastasis, pancreatic cancer remains an extremely malignant carcinoma. Novel targets and biomarkers are required for the treatment of pancreatic cancer. In the present study, we explored the expression and effect of ROBO2 in pancreatic cancer. We rst evaluated the mRNA and protein expression levels of ROBO2 in the normal pancreatic cell line HPNE and six pancreatic cancer cell lines and found that ROBO2 expression was lower in the pancreatic cancer cell lines. Clinicopathologic analysis showed that increased ROBO2 expression was correlated with a better prognosis. Downregulation of ROBO2 promoted proliferation and migration in vitro and accelerated tumor growth in vivo. mRNA sequencing showed that ROBO2 downregulated the expression of ECM1, promoting the migration and invasion of pancreatic cancer.
The SLIT/ROBO signaling pathway, which was initially identi ed through its effect on axon guidance [12], has shown dual roles in cancers by affecting both oncogenes and anti-tumor genes. In colorectal cancer, the levels of ROBO1 and ROBO4 are signi cantly higher in tumor tissues than para-tumor tissues. ROBO1 is expressed mainly in tumor cells, while ROBO4 is situated in endothelial cells of tumor vessels [13]. In breast cancer, the SLIT2/ROBO1 signaling pathway enhances the invasion and migration of cancer cells by promoting matrix metalloproteinase 9 [14]. Additionally, the SLIT/ROBO pathway has been shown to act on tumor suppressor genes in some special cancers. The SLIT2 gene is downregulated in colorectal cancer, and high SLIT2 expression can inhibit the migration capability [15]. Accordingly, overexpression of SLIT2 can inhibit migration and invasion of melanoma cells and immortalized epithelial cells by reducing Cdc-42 activation and preventing hepatocyte growth factor-induced dynamic responses [16]. In lung cancer cells, low expression of USP33 was associated with a poor prognosis, which may be attributed to the instability of the ROBO1 protein [17]. Thus, clari cation of this contradictory effect may be of great signi cance for the treatment of pancreatic cancer.
As a member of the immunoglobulin superfamily, ROBO2 consists of three parts: the extracellular domain, transmembrane regions, and the intracellular region. ROBO2 has been found to regulate different physiological functions and tumor development. Gonçalves et al. [18] revealed an important role of ROBO2 in lung development. Escot et al. [19] found that ROBO2 plays an important role in pancreas cell identity and plasticity. In the developing mouse embryo, simultaneous knockout of ROBO1 and ROBO2 resulted in a decreased pancreatic volume.
The functions and mechanisms of ROBO2 in tumor progression are gradually being recognized. Skuja et al. found that ROBO2 deletion was identi ed in more than half of patients with metastatic colorectal cancer [20]. However, through L1-targeted resequencing of colorectal tumors and matched normal DNA, Solyom et al. described ROBO2 as an oncogene [21]. Generally, ROBO2 suppresses WNT/β-catenin signaling activity and inhibits migration and invasion of cancer cells [22,23]. Biankin et al. revealed that high expression of ROBO2 was associated with a better prognosis, which is consistent with our results. Furthermore, transforming growth factor (TGF)-β activation promoted the progression of pancreatic cancer [24]. Pinho et al. reported that in acute pancreatitis mouse models, the absence of ROBO2 from epithelial cells may result in the activation of TGF-β and promote a strong anti-in ammatory response and that these effects can be reversed by the TGF-β inhibitor galunisertib [25]. Han et al. found that ROBO3 expression was upregulated in PDAC tissue samples and that the level of ROBO3 was positively correlated with the tumor stage. Moreover, the overexpression of ROBO3 promoted malignant biological behavior of pancreatic cancer both in vitro and in vivo and activated the WNT/β-catenin signaling pathway. However, their study did not show a signi cant relationship between the ROBO3 expression level and overall survival or disease-free survival [26].
ECM1 overexpression is related to strong malignant behavior and a poor prognosis in multiple malignancies, such as gastric cancer, breast cancer, laryngeal carcinoma, hepatocellular carcinoma, and others [27][28][29][30]. Various regulatory mechanisms are involved in the migration of malignant tumors, and EMT is one such mechanism of great importance [31]. The main features of EMT are loss of the epithelial phenotype and acquisition of the mesenchymal phenotype, resulting in enhanced tumor cell migration and invasiveness via impaired cell-matrix adhesion and remodeling of the ECM [32][33][34][35]. E-cadherin, Ncadherin, and vimentin, important markers of EMT, are associated with the metastatic phenotype. Our study also showed that ROBO2 expression was negatively associated with EMT. Lee et al. reported that ECM1 could advance EMT progression via stabilization of the β-catenin protein in metastatic breast cancer. Their study also demonstrated that ECM1 regulates chemotherapy resistance, sphere-forming ability, and cancer stem cell maintenance [36]. Wu et al. reported that ECM1 could interact with moesin to promote aggressive breast cancer phenotypes [37]. Another primary study showed that migration and metastasis of cancer cells may be induced by a simple change in the structure of the ECM [38]. These ndings are consistent with the results of our study, suggesting that ROBO2 has anti-cancer effects in pancreatic cancer.
We acknowledge that our study has some limitations. First, the pancreatic cancer tissues were obtained from surgical resection, and because of the low resection rate, bias in the association between ROBO2 expression and the TNM stage is inevitable. Additionally, the sample size was not large (<100 patients). Second, although RNA sequencing and bioinformatics were used to analyze the potential targeted genes of ROBO2, the mechanism by which ROBO2 downregulates the expression of ECM1 is still unknown.
Deeper and more direct proofs are needed to demonstrate the signaling pathway regulated by ROBO2 and the mechanism of ROBO2 in the development of pancreatic cancer. Third, although RNA sequencing and analysis is used to investigate many genes involved in cellular metabolism, we did not investigate the functions and effects of ROBO2 in pancreatic cancer metabolism. Further studies are needed to investigate the role and mechanism of ROBO2 in pancreatic tumor metabolism.
In summary, our results suggest that ROBO2 inhibits malignant biological behaviors, including cell proliferation and migration. We found relevance between the overexpression of ROBO2 and benign cancer behaviors and a better prognosis in patients with pancreatic cancer. Our study indicates that ROBO2 acts as a tumor suppressor gene and can potentially be a therapeutic target for pancreatic cancer.  The total patient number does not equal 95 for all variables, as patient information was not available in some cases. knockdown and expression con rmed by western blot and RT-PCR analyses. Data are presented as mean ± SD. **P < 0.01, ***P < 0.001, by Student's t-test. PDAC, pancreatic ductal adenocarcinoma; SH, ROBO2 knockdown; NSH, scrambled control; OE, ROBO2 overexpression; NOE, negative control for ROBO2 overexpression (n=3).