Dual roles of WISP2 in the progression of hepatocellular carcinoma: implications of the fibroblast infiltration into the tumor microenvironment

The dismal outcome of hepatocellular carcinoma (HCC) patients is attributable to high frequency of metastasis and. Identification of effective biomarkers is a key strategy to inform prognosis and improve survival. Previous studies reported inconsistent roles of WISP2 in carcinogenesis, while the role of WISP2 in HCC progression also remains unclear. In this study, we confirmed that WISP2 was downregulated in HCC tissues, and WISP2 was acting as a protective factor, especially in patients without alcohol intake using multiple online datasets. In addition, we reported that upregulation of WISP2 in HCC was related to inhibition of the malignant phenotype in vitro, but these alterations were not observed in vivo. WISP2 also negatively correlated with tumour purity, and increased infiltration of fibroblasts promoted malignant progression in HCC tissues. The enhanced infiltration ability of fibroblasts was related to upregulated HMGB1 after overexpression of WISP2 in HCC. The findings shed light on the anticancer role of WISP2, and HMGB1 is one of the key factors involved in the inhibition of the efficiency of WISP2 through reducing the tumour purity with fibroblast infiltration.


INTRODUCTION
Hepatocellular carcinoma (HCC) has high incidence rates in China, which accounts for more than 50% of the total number of liver cancer cases and deaths in the world [1]. Even though many advanced strategies, including liver transplantation, molecular targeted therapies and immune-based treatments, the general prognosis of patients with HCC is still unsatisfactory [2]. Therefore, continued identification of new molecules for the development of combining targeted therapy is still urgently needed.
Cellular communication network (CCN) family are scaffolding proteins that may govern and balance the interconnection among individual signaling pathways.
In the present study, we analysed expression of WISP2 and the prognostic correlation in HCC patients using Oncomine, Kaplan-Meier plotter, and Gene expression profiling interactive analysis 2 (GEPIA2). We explored the role of WISP2 in HCC using the Cancer Cell Line Encyclopedia (CCLE) and gene microarrays and then assessed the correlation between WISP2 and stromal cells in tumour tissues using the Tumor IMmune Estimation Resource (TIMER). Finally, the role of WISP2 and its relationship with tumour purity and fibroblast infiltration were examined both in vitro and in vivo. Our studies revealed the anticancer role for WISP2 was conditional in HCC, and the efficiency was influenced by fibroblast infiltration in the tumor microenvironment (TME).

mRNA expression, gene correlation, tumour purity, and immune infiltrate analysis in TIMER
TIMER is a comprehensive resource for analysis of immune infiltrates across diverse cancer types (https://cistrome.shinyapps.io/timer/) [14]. The expression levels of WISP2 between tumour and adjacent normal tissues in different types of cancer were identified across all TCGA tumours via 'Diff Exp' module. The correlation between WISP2 expression and immune infiltration, including B cells, CD4 + T cells, CD8 + T cells, neutrophils, macrophages, and dendritic cells, as well as a pairwise gene (CD34, PECAM, VCAM1, NT5E, ESM1, S100A4, VIM, and ACTA2) correlation in liver cancer were explored via 'Gene' and 'Correlation' modules, respectively.

mRNA expression and survival analysis in GEPIA2
GEPIA2 is resource for analyzing the RNA sequencing expression data of 9,736 tumours and 8,587 normal samples from the TCGA and the GTEx projects [15]. WISP2 tumour/normal differential expression and the correlation between WISP2 expression and survival in diverse cancer types are analysed in GEPIA2.

mRNA expression in cancer cell lines in the CCLE
WISP2 expression levels in cancer cell lines from diverse cancer types were examined using the CCLE (http://www.broadinstitute.org/ccle), which provides public access to genomic data, analysis, and visualization for more than 1,100 cell lines [16].

Survival analysis in Kaplan-Meier plotter
The correlation between WISP2 expression and survival in liver cancer was analysed using the Kaplan-Meier plotter (http://kmplot.com/analysis/) [17]. The Kaplan-Meier plotter can assess the effects of 54,000 genes on survival in 21 cancer types. Gene expression data and RFS and OS information were downloaded from GEO, EGA, and TCGA.

Protein-protein interaction (PPI) analysis in GeneMANIA
GeneMANIA is an online tool that predicts the function of genes and gene sets, including protein and genetic interactions, pathways, co-expression, co-localization, and protein domain similarity, in GeneMANIA [18] (https://genemania.org). WISP2 and CTGF were used as queries to predict PPIs, and the prediction output graphically shows a network that depicts the relationships between genes in the list.

Clinical features analysis in LinkedOmics database
The relationship between WISP2 expression and clinical features in liver cancer patients was analysed using the LinkedOmics database (http://www.linkedomics.org), which is a publicly available portal that includes AGING multi-omics data from all 32 TCGA Cancer types [19]. Based on the platform, a statistical analysis of the correlation between WISP2 expression and clinical features of HCC was performed.
cDNA microarray analysis cDNA microarrays were performed using the Human OneArray ® (Phalanx Biotech Group, San Diego, CA, USA) to evaluate the alterations of expression profiling. Total RNA was extracted from Hep3B-WISP2 and Hep3B-Vector cells and the isolations and microarray analyses were performed in triplicate according to the manufacturer's instructions. All data was uploaded to the Gene Expression Omnibus (GSE134563).
Patients and tissue microarray analysis, cell lines culture, vector construction and transfection, various functional assays of WISP2 in vivo (animal models with subcutaneous xenografts) and in vitro (migration, invasion, and proliferation) were all exhibited in Supplementary Materials.

Statistical analysis
Graphics were drawn using GraphPad Prism version 6 (GraphPad Software, La Jolla, CA, USA). Statistical analyses were performed using SPSS 15.0 for Windows (SPSS). A p-value <0.05 was considered statistically significant.

Ethics approval
Animal protocols were approved by the Medical Experimental Animal Care Commission of Northwest Polytechnical University, and all methods were performed in accordance with the relevant guidelines.

WISP2 mRNA levels was lower in tumour tissues of HCC compared with normal tissues, but the antitumor role of WISP2 is conditional
Using the TIMER and GEPIA2 database to evaluate the RNA-seq data, expression of WISP2 was found to be significantly downregulated in most human cancers, including HCC, compared with the associated normal tissues. Only in kidney renal clear cell carcinoma (KIRC) and kidney renal papillary cell carcinoma (KIRP) were WISP2 expression significantly higher than that in adjacent normal tissues ( Figure 1A, 1B). To explore the role of WISP2 on prognosis, 33 human cancers were included using the GEPIA2 database. The role of WISP2 on prognosis varied in different types of cancers. In HCC, WISP2 had a protective role, as high expression was associated with better prognosis for this cancer type, although there was no significant statistical difference ( Figure 1C).
To better understand the role of WISP2 in HCC prognosis, the Kaplan-Meier plotter was used. The HCC patients in the WISP2-high group had longer OS than the WISP2-low group, although there was no significant difference (HR = 0.7, p = 0.067). The patients in the WISP2-high group had a significantly longer RFS compared with those in the WISP2-low group (HR = 0.62, p = 0.011; Figure 1D). These results revealed WISP2 was acting as a protective factor and the level of WISP2 was decreased in HCC.
We evaluated the effects of clinicopathologic characteristics on the prognosis of HCC patients with different WISP2 expression levels. High expression of WISP2 was associated with longer OS in female patients (HR = 0.52, p = 0.026). Specifically, high expression of WISP2 mRNA was correlated with longer OS in stage 2 (HR = 0.37, p = 0.01) and longer RFS in stage 1 (HR = 0.5, p = 0.013) in HCC patients. High WISP2 expression was correlated with longer OS in grade 1 (HR = 0.3, p = 0.023) and 3 (HR = 0.45, p = 0.017) patients. In addition, high WISP2 expression was correlated with longer RFS in patients with no vascular invasion (HR = 0.57, p = 0.028), T1 (HR = 0.49, p = 0.0078), and T2 (HR = 0.36 p = 0.018) according to the criteria of the American Joint Committee on Cancer (AJCC). These results suggest that WISP2 expression levels can significantly influence the prognosis in most human cancers, but the role of WISP2 in HCC is conditional ( Table 1).

Upregulation of WISP2 in HCC is related to inhibited malignant phenotype in vitro
To determine the precise function of WISP2, we first evaluated the expression of WISP2 in different human tumour cell lines from the CCLE, revealing the significant differences in the expression of diverse tumors ( Figure 2A). We further analysed the WISP2 expression in 25 HCC cell lines from the CCLE, and found that the expression of WISP2 was low in 76% HCC cell lines, especially in Hep3B and HepG2 cells ( Figure 2B).
We also evaluated markers related to the epithelialmesenchymal transition (vimentin, Snail, and E-cadherin) as a measure of migration and invasiveness and Ki67 as a proliferation marker at the protein level in HCC cells. The expression of Ki67, vimentin, and Snail were significantly downregulated, and the epithelial cell surface marker E-cadherin was upregulated in Hep3B and HepG2 cells that overexpressed WISP2 ( Figure 2G). All of these results suggest that WISP2 expression is acting as a protective factor in vitro.  Bold text means the difference has statistical significance (p < 0.05).

The increased infiltration of fibroblasts plays a negative feedback role in inhibition of the efficiency of WISP2 in vivo
In this section, the role of WISP2 was evaluated in vivo. While, in nude mouse models, subcutaneous tumour growth after 4 weeks did not differ in mice injected with Hep3B-WISP2 and Hep3B-Vector (1.23 ± 0.15 mm vs. 1.33 ± 0.12 mm, p = 0.2354; Figure 3A, 3a). And subcutaneous tumour growth also did not differ in mice injected with HepG2-WISP2 and HepG2-Vector (1.40 ± 0.14 mm vs. 1.58 ± 0.15 mm, p = 0.103; Figure 3A, 3b). This finding was in direct contrast to our in vitro results. Additionally, we classified the patients into subgroups according to alcohol intake history, the major aetiologies of liver cancer in western countries. In patients without alcohol intake history, patients with high WISP2 expression showed significantly longer OS and RFS compared with those with low WISP2 expression ( Figure 3B, 3a). While, there was no significant difference in OS and RFS between high and low WISP2 expression in patients with alcohol intake history ( Figure 3B, 3b). It is now clear that alcohol consumption is closely related to liver fibrosis [20]. Thus, we sought to examine tumour microenvironmental influences on the WISP2 effects.
Additionally, we further examined the co-expression of WISP2 and α-SMA in subcutaneous tumour tissues. Increased numbers of fibroblasts and fibro-collagen deposition were positive correlated with the expression of WISP2 in HCC ( Figure 3E). Human liver cancer tissue microarrays confirmed a positive correlation between WISP2 and α-SMA ( Figure 3F). These results suggested that WISP2 played a negative role in tumour purity by promoting fibroblast infiltration into the tumour microenvironment.

Upregulation of WISP2 in HCC cells significantly alters gene expression profiles
Gene expression profiles were significantly altered in Hep3B cells that overexpressed WISP2 compared to vector control cells, with 415 differentially expressed genes ( Figure 4A). We analysed these altered genes using Gene Ontology (GO, Figure 4B) and Kyoto Encyclopedia of Genes and Genomes (KEGG, Figure  4C) pathway enrichment. According to GO analysis, 25 genes involved in wound healing, were significantly changed, and 80% of these were inhibited by WISP2 overexpression. The downregulated genes included KLK8, MGLL, ENO3, ALOX5, BLNK, CFI, IRAK2, GPR68, and APOL3 ( Figure 4D, 4a). With respect to anti-apoptosis factors, 11 genes were significantly altered by WISP2 overexpression (Figure 4D, 4b).
Twenty five genes related to the cell cycle were significantly changed, and 8 core genes were significantly downregulated ( Figure 4D, 4c). Of the 10 drug resistance-related genes that were altered, nine were downregulated by WISP2 overexpression ( Figure  4D, 4d). KEGG pathway enrichment analysis revealed significant alterations in seven important pathways, including Staphylococcus aureus infection, Mineral absorption, Biosynthesis of amino acids, Metabolic pathways, Arginine and proline metabolism, Toxoplasmosis, and HTLV-I infection. High-mobility group protein box1 (HMGB1) is a pivotal factor in the development and progression of many types of tumours, which is closely correlated with tumour-mediated inflammation microenvironment [21]. Interestingly, HMGB1 was significantly upregulated after WISP2 overexpression in HCC cells.

Enhanced fibroblast infiltration is related to upregulated HMGB1 in the presence of WISP2 overexpression in HCC
As determined in the gene expression profiles following WISP2 overexpression in HCC cells, HMGB1 was significantly upregulated, and this upregulation was found to be strongly correlated with cirrhosis in our previous study [21]. In this study, we first validated these gene array assays at the protein level via immunoblotting and confirmed the upregulation of HMGB1 in the presence of WISP2 overexpression ( Figure 5A). Next, we generated a Hep3B-WISP2 cell line with stably downregulated expression of HMGB1 (Hep3B-WISP2-shHMGB1; Figure 5B). Hepatic stellate LX2 cells treated with conditioned medium (CM) from Hep3B-WISP2-shHMGB1 exhibited inhibited proliferation ability ( Figure 5C, 5D). While, LX2 cells treated with CM from Hep3B-WISP2-shHMGB1 exhibited enhanced migration ability (3.17 ± 1.17 vs. 33.83 ± 12.16, p = 0.0001; Figure 5E). We then produced subcutaneous tumours in nude mice via transplantation of Hep3B-Vector, Hep3B-WISP2-Mock, and Hep3B-WISP2-shHMGB1 cells. After 4 weeks, weights of the subcutaneous tumours from Hep3B-WISP2-Vector and Hep3B-WISP2-Mock cells were also not significantly different (2.78 ± 0.12 g vs. 3.02 ± 0.16 g, p = 0.0819), while the weights of tumours from Hep3B-WISP2-ShHMGB1 cells were significantly decreased (0.42 ± 0.11 g vs. 2.78 ± 0.12 g, p < 0.0001; Figure 5F). We also found increased expression of α-SMA and fibro-collagen deposition in tumour tissues generated from Hep3B-WISP2-Mock cells, while tumours generated from the Hep3B-WISP2-ShHMGB1 cells exhibited significantly decreased α-SMA expression and fibro-collagen deposition ( Figure 5G). Thus, HMGB1 is involved in the AGING AGING increased fibroblast infiltration that results from WISP2 overexpression ( Figure 6).

DISCUSSION
HCC is the fourth leading cause of cancer death worldwide according to global cancer statistics 2018 [22], and alcohol abuse are considered to be pathogenic factors for HCC in western countries [23,24]. Although early diagnosis and surgical resection are primary antitumor strategies, the prognosis of cancer patients remains generally dismal, with a 5-year overall survival rate of only 50-70%, and the unfavorable outcomes attributed to the high frequency of tumor recurrence, metastasis, and therapeutic resistance [25]. Therefore, continued identification of new molecules for early survival prediction and the development of molecular targeted therapy are still urgently needed.
Cellular communication network (CCN) family are scaffolding proteins that may govern as localized multitasking signal integrators in tumours and the AGING associated TME [26]. CCN proteins are involved in many vital biological functions, including angiogenesis, fibrosis, and tissue regeneration and repair [14]. In human cancers, the expression levels of CCN proteins are closely correlated with in regulating tumor cellular function and the TME [27]. As CCN5 lacks a CT domain, this striking difference in structure compared with other CCN family members may allow it to have unique functional roles. Like its family members, however, previous studies reported inconsistent roles of AGING WISP2 in carcinogenesis. In HCC, the role of WISP2 in tumor progression also remains unclear. Previously, no significant difference in WISP2 expression was identified between HCC tumours and matched normal liver samples [13], indicating that the role of WISP2 in HCC tumour progression remains unresolved. In the present study, data from TIMER and LinkedOmics databases revealed that levels of WISP2 were lower in tumour tissues than in normal liver tissues, and high expression of WISP2 was associated with better prognosis in HCC patients, as revealed by the Kaplan-Meier plotter, indicating WISP2 is a protective factor in HCC. In addition, in HCC patients without history of alcohol intake, high expression of WISP2 indicated significantly better prognosis, indicating that the effects of WISP2 may be influenced by the microenvironment.
In vitro studies also showed that upregulation of WISP2 in HCC is related to inhibition of the malignant phenotype, although such inhibition of proliferation in vivo was not obvious.
Because the in vivo and in vitro studies were inconsistent, we sought to explore potential influences in the tumour microenvironment. Solid tumours are complex entities, as they are surrounded by a heterogeneous array of extracellular matrix and various stromal cells that play important roles in cancer progression. Specifically, tumour-infiltrating lymphocytes are an independent predictor of survival in cancers [28]. In the present study, WISP2 expression was negatively correlated with tumour purity in HCC. And we found WISP2 expression was weakly correlated with infiltrating lymphocytes, including CD8 + T cells, CD4 + T cells, B cells, neutrophils, dendritic cells, and macrophages, while WISP2 was positively correlated with the number of fibroblasts in the TME. The co-expression of WISP2 and the fibroblast marker α-SMA in subcutaneous tumour tissue and HCC tissue microarrays confirmed this relationship. Therefore, we hypothesized that the inconsistency of results in vitro and in vivo was due to fibroblast infiltration in the TME.
Accumulating evidence supports the concept that cirrhosis is one of the key factors that promotes HCC [29]. And according to the reports, the role of WISP2 appeared to have paradoxical effects in stromal cells. Grünberg et al. [30] showed that WISP2 activates the canonical WNT pathway and increased β-catenin levels via targeting LRP5/6 phosphorylation in mesenchymal cells. In the myocardium, WISP2 can reverse established cardiac fibrosis via inhibition of enhanced apoptosis of myofibroblasts [31]. In the current study, we demonstrated that CM from an HCC cell line that was engineered to overexpress WISP2 significantly inhibited proliferation of LX2, and after downregulation of HMGB1 in these cells, an additional significant decrease in proliferation was observed. While, it is surprised that we proved the CM from an HCC cell line that was engineered to overexpress WISP2 significantly increase the migration ability of LX2, and the trends was reversed in the HCC cell line with stably downregulated expression of HMGB1.
HMGB1 plays a role in several cellular processes, including inflammation, cell differentiation and migration [32]. In our previous study, the upregulation of HMGB1 was found to be strongly correlated with cirrhosis in HCC [21]. According to our screening results, HMGB1 was one of the significantly upregulated genes after WISP2 overexpression in HCC cells, we take HMGB1 as a follow-up research target. In the follow-up study, the weights of subcutaneous tumours generated from these engineered HCC cells in nude mice were also significantly decreased after downregulation of HMGB1 expression in the presence of WISP2 overexpression, indicating that HMGB1 is one of the key factors involved in reducing the anticancer efficiency of WISP2. Additionally, we observed decreased expression of the fibroblast biomarker α-SMA in tumour tissues with low expression of HMGB1. These results are in line with previous reports that upregulation of HMGB1 is associated with inflammatory pathogenesis, with enhanced local inflammation and fibrosis [33].

CONCLUSIONS
We propose that as high expression of WISP2 is associated with better prognosis in HCC, WISP2 can serve as a prognostic biomarker and the prediction efficiency is influenced by tumour purity with fibroblast infiltration. And the enhanced infiltration of fibroblasts is related to upregulated expression of HMGB1, that results from WISP2 overexpression and weakens the anticancer effects of WISP2 ( Figure 6). The findings shed light on the dual roles of WISP2 in HCC, suggesting that WISP2 up-regulation combined with HMGB1 inhibition may serve as an effective therapeutic strategy for better prognosis in HCC. While our study was only limited to liver cancer, and lack of deeper mechanistic knowledge of the regulatory relationship between HMGB1 and chemokines from HCC. Therefore, several fundamental questions remain to be answered concerning the mechanism of WISP2 regulating HMGB1 and chemokines in the further study.

Availability of data and material
The datasets used and/or analyzed and materials developed during the current study are available from the the repository of TCGA (https://portal.gdc.cancer.gov) and corresponding author by reasonable request.

Vector construction, transfection, and lentivirus transduction
Human full-length WISP2 cDNA was obtained from GeneCards (Shanghai, China) and cloned into the pCDH lentiviral expression vector (System Biosciences, Palo Alto, CA, USA). The amplified fragment was inserted into the pCDH plasmid (between XbaI and EcoRI sites) using the In-Fusion HD Cloning Kit (Takara, Tokyo, Japan). High-mobility group protein 1 (HMGB1) shRNA and scramble plasmids were obtained as a gift from the Institutes of Biomedical Sciences at Fudan University (Shanghai, China), and each was in the lentiviral expression plasmid PLKO.1. Stable cell lines were created by antibiotic selection with puromycin for a week, beginning 72 h after transduction.

Animal models
A total of 21 male BALB/c nu/nu mice (aged 4-6 weeks and weighing approximately 20 g) and raised in a controlled environment with 25°C under standard pathogen-free conditions and a natural light/dark cycle (morning 8:00; afternoon 8:00), and were provided with water and standard diet. Hep3B-Vector, Hep3B-WISP2, HepG2-Vector, HepG2-WISP, and Hep3B-WISP2-Mock, and Hep3B-WISP2-shHMGB1 cells were implanted subcutaneously into the upper left flank region of mice to establish subcutaneous xenografts. Tumour weights were evaluated 4 weeks after the treatments. The intraperitoneal injection of pentobarbital (5 mg/kg) combined with cervical spondylolisthesis was used for euthanasia of the mice after the study. The study protocol was approved by the Medical Experimental Animal Care Commission of Northwest Polytechnical University, and all methods were performed in accordance with the relevant guidelines.

Migration and invasion assays
Migration of Hep3B-WISP2 and Hep3B-Vector were evaluated in Boyden chambers containing membranes with 8.0-μm pores in 24-well plates (Corning, Tewksbury, MA, USA). Cells were seeded into the upper chamber of each well in serum-free DMEM (6 × 10 4 cells/well). DMEM containing 2% FBS was added to the lower chamber of each well. After 24 h, cells on the underside of the membrane were stained with Giemsa (Sigma Chemical Company, St. Louis, MO, USA), counted, and photographed at 200× magnification. Cell invasion assays were performed similarly, except that 80 μL Matrigel (BD Biosciences, San Jose, CA, USA) was added to each well 6 h before cells were seeded onto the membrane.