SULF2 Is a Prognostic Biomarker and Correlated with Tumor Associated Macrophages in Gastric Cancer

Background: Recently, the mutual effects of tumor cells and the tumor immune microenvironment have been identied as key factors in promoting cancer progression. Sulfatase 2 (SULF2) encodes an extracellular endoglucosamine ‐ 6-sulfatase, which could remodel the highly sulfated domains of heparan sulfate. The abnormal expression of SULF2 is reported to play an important role in the carcinogenesis of many kinds of cancer. However, the prognostic value of SULF2 and its correlation with immune cell inltration in gastric cancer (GC) remain unclear. Results: SULF2 expression was signicantly increased in GC compared with gastric normal tissue, especially in the advanced stage GC. In addition, high SULF2 expression signicantly predicted an unfavorable prognosis in GC patients (overall survival P=0.0074), particularly who had metastatic lymph nodes. Besides, pathway analyses of SULF2 in GC revealed SULF2 may take part in extracellular structure organization, cell-cell adhesion and proteoglycans in cancer, etc. Importantly, the expression level of SULF2 was found to be positively correlated with the inltration levels of tumor associated macrophages (TAMs). Moreover, SULF2 expression in GC positively correlated with expression of several immune cell markers, including monocyte markers, TAMs markers and programmed cell death 1 ligand 1 (PD-L1), suggesting its role in regulating tumor immunity. Conclusion: This study identied distinct expression and prognostic values of SULF2 in GC using public databases. Signicantly, our ndings shed light on the role of SULF2 in GC progression and provided an underlying mechanism that SULF2 expression might modulate tumor immunity by regulating the inltration of TAMs in GC.


Background
GC is still the fth most common cancer and fourth leading cause of cancer related death in the world 1 .
Although a variety of treatment methods have been used in the past decades, the prognosis of patients with GC remains unsatisfactory, and the 5 year survival rate is about 30% around the world 2 .
Unfortunately, there is a lack of clinically useful biomarkers to inform prognosis or aid treatment strati cation for GC patients. Therefore, it is necessary to explore novel molecular prognostic factors and potential therapeutic targets for GC.
The tumor microenvironment (TME) consists of immune cells, in ammatory cells, broblasts and vascular endothelial cells, which could release various molecules to either directly activate the growth signaling or remodel extracellular matrix, favoring the growth and expansion of cancer cells 3 . The concept of GC treatment is changing with the deepening understanding of TME. Researchers are trying to explore various components and their interactions in the TME of GC and try to block some signal pathways, so as to improve the therapeutic effect of the tumor. Immunotherapy, such as nonspeci c immune enhancer 4 , cytokines 5 , adoptive therapy of immune cells 6 and immune checkpoint inhibitor [7][8][9] , has been associated with improved outcomes among a part of GC patients. However, the bene t population is too small and needs to be further screened. The biomarkers for immunotherapy are still lacking in GC. Sulfatase 2 (SULF2) gene at chromosome 20q13 encodes a protein with heparin-degrading endosulfatase activity. SULF2 could modify heparan sulfate glycosaminoglycan (HSGAG) chains by removing 6-O-sulfate groups from heparan sulfate disaccharide units, decreasing the a nity of Heparan sulfate proteoglycans (HSPGs) for heparan-sulfate binding ligands and releasing sequestered factors from storage sites 10 . In this way, SULF2 is involved in the modulation of ligand-receptor interactions and activation of downstream signaling pathways 11 . A detailed investigation of the association between SULF2 and in ltrating immune cells is needed.
In fact, promotion of tumor progression via the increased expression of SULF2 is established in several types of cancers, including hepatocellular carcinoma (HCC) 12 , non-small cell lung cancer (NSCLC) 13 , colorectal cancer 14 , cervical cancer 15 , breast cancer 16 , pancreatic ductal adenocarcinoma 17 , neuroblastoma 18 , and head and neck squamous cell carcinoma (HNSC) 19 . In HCC, SULF2 acts as an oncogenic protein by increasing Wnt3a and glypican-3 (GPC3) expression and activating the Wnt/βcatenin pathway, thus promoting growth of HCC cell lines and xenografts 20 . Besides, SULF2 induces the differentiation of hepatic stellate cells into carcinoma-associated broblasts (CAFs) through the TGF-β1/SMAD3 signaling pathway. Then SULF2-induced CAFs attenuated HCC apoptosis by activating the SDF-1/CXCR4/PI3K/AKT signaling pathway and induced epithelial-mesenchymal transition through the SDF-1/CXCR4/OIP5-AS1/miR-153-3p/SNAI1 axis 12 . Aside from that, in NSCLC, SULF2 activated TGF-β1/SMAD signaling pathway, which involved in the induction of migration and epithelial-mesenchymal transition 13 . Moreover, SULF2 expression increases VEGF-A release and activity in TME in colorectal cancer 21 . And SULF2 promotes tumorigenesis through the ERK/AKT signaling pathway in cervical cancer and colorectal cancer 14,15 . Besides, in breast cancer, SULF2 facilitated lymphangiogenesis by regulating VEGF-D and the AKT1-related signaling pathway was involved 16 . Therefore, SULF2 can serve as an independent risk factor and prognostic biomarker for different types of cancer. However, the prognostic and immunological signi cance of SULF2 in GC has not been elucidated.
In this study, we aimed to integrate a variety of bioinformatics methods to study whether SULF2 is involved in GC progression and immune in ltration. We found that the expression of SULF2 was signi cantly upregulated in GC tissues compared with normal tissues. The high expression of SULF2 was negatively correlated with the prognosis of GC patients. In addition, there was a signi cant relationship between the expression of SULF2 and the in ltration levels of tumor-associated macrophages (TAMs) in GC. Importantly, SULF2 seemed to affect the prognosis of GC patients partially through TAMs in ltration.

SULF2 expression is increased in stomach adenocarcinoma (STAD) patients
To determine differences of SULF2 expression in various cancer types, we analyzed mRNA expression between different cancers and normal tissues in multiple cancers using Tumor Immune Estimation Resource (TIMER) database. Compared with the corresponding normal tissues, we observed higher expression of SULF2 in breast invasive carcinoma (BRCA), cholangiocarcinoma (CHOL), colon adenocarcinoma (COAD), esophageal carcinoma (ESCA), glioblastoma multiforme (GBM), HNSC, kidney renal clear cell carcinoma (KIRC), kidney renal papillary cell carcinoma (KIRP), lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), rectum adenocarcinoma (READ), STAD and thyroid carcinoma (THCA) ( Figure 1A). To further verify the conclusion, we also consistently found that higher mRNA of SULF2 was expressed in STAD than in para-carcinoma tissues in the Gene Expression Pro ling Interactive Analysis (GEPIA) and UALCAN databases ( Figures 1B, C). Furthermore, the protein expression of SULF2 was investigated in STAD by immunohistochemical staining, and we found that the SULF2 protein levels were obviously increased in STAD tissues compared with para-carcinoma tissues ( Figures  1D). These ndings suggest that SULF2 expression is upregulated in multiple types of cancers, including STAD.

Correlation of SULF2 expression and clinical parameters of STAD patients
By using the UALCAN online tool, we then investigated SULF2 expression among groups of patients according to different clinical parameters. SULF2 expression was signi cantly upregulated in STAD samples from all races including Caucasian, African-American and Asian, compared to the corresponding normal controls (Figure 2A). According to gender, SULF2 expression was signi cantly upregulated in STAD samples from both males and females ( Figure 2B). In terms of age, the expression of SULF2 was dramatically elevated in the STAD tissues of patients from different age groups (41-60 years, 61-80 years and 81-100 years) ( Figure 2C). Regarding cancer stage, a signi cant increase in SULF2 expression was observed in STAD patients in stage 2, 3 and 4, compared to the corresponding normal controls or STAD patients in stage 1 ( Figure 2D). Based on nodal metastasis status, SULF2 expression was higher in patients with STAD classi ed as N0, N1 or N2 and N3 ( Figure 2E). Upregulation of SULF2 expression was observed in the STAD tissues of patients from different differentiation status, including well differentiation, moderate differentiation and poor differentiation ( Figure 2F). These results suggest that SULF2 expression may related to STAD progression.

Increased SULF2 expression correlates with poor prognosis in STAD patients
Because the expression levels of SULF2 are closely related to the progression of STAD, we further examined the prognostic value of the SULF2 gene. STAD patients with higher expression of SULF2 gene exhibited faster rst progression (FP) ( Figure 3A), poorer overall survival (OS) ( Figure 3B) and poorer post progression survival (PPS) ( Figure 3C) according to the Kaplan-Meier plotter database. These results indicate that SULF2 is signi cantly associated with the prognosis of STAD patients.

Validation of the prognostic value of SULF2 based on various clinicopathological features
In order to better understand the prognostic value and potential mechanism of SULF2 expression in STAD, we used the Kaplan-Meier plotter database to explore the relationship between SULF2 mRNA expression and OS/FP according to clinical characteristics. High SULF2 expression was signi cantly correlated with poor OS and FP in male but not female STAD patients ( Figure 3D). Regarding different cancer stages, high SULF2 expression was associated with poor OS and poor FP only in stage 3 STAD patients ( Figure 3D). A signi cant correlation between SULF2 expression and poor OS was observed in American Joint Committee on Cancer (AJCC) T3 STAD patients ( Figure 3D). Besides, high expression of SULF2 was associated with fast FP in T3 and T4 STAD patients ( Figure 3D). In addition, high SULF2 expression was associated with poor OS and FP in STAD patients with N1, N2, N1-3 nodal metastasis status ( Figure 3D). Upregulated SULF2 levels corresponded with poor OS and FP in M0 patients ( Figure  3D). Based on Lauren classi cation, SULF2 upregulation was correlated with poor OS and poor FP in diffuse type of STAD patients, and correlated with poor FP in intestinal type ( Figure 3D). Moreover, we found an association between SULF2 expression and unfavorable OS in poorly differentiated STAD patients ( Figure 3D). These results imply that SULF2 mRNA expression possesses prognostic value in STAD.

Identi cation of SULF2-interacting genes and proteins
We used the LinkedOmics database to identify RNA-seq genes co-expressed with SULF2 in STAD. We recognized the top 50 genes that positively or negatively interacted with SULF2 ( Figure 4A, B). And we used GeneMania to construct a SULF2 correlation network to determine the potential mutual effects between SULF2 and cancer-related targets. The results showed that the 20 most frequently altered genes were closely correlated with SULF2, including SULF1, NOTCH1, and TNIP1 ( Figure 4C). Functional analysis suggested that these genes were signi cantly associated with the sulfuric ester hydrolase activity ( Figure 4C). In addition, a protein-protein interaction network of SULF2 was generated using the STRING database. There were 35 edges and 11 nodes, including GPC3, HS2ST1 and GLCE ( Figure 4D). The results revealed that multiple differentially expressed genes were correlated with SULF2 expression.
Enrichment analysis of SULF2 functional networks in STAD Three independent ontologies (biological process, cellular component, and molecular function) were analyzed by gene set enrichment analysis. The results indicated that SULF2-related differentially expressed genes were involved in a variety of biological processes (extracellular structure organization, cell-cell adhesion via plasma-membrane adhesion molecules, cell junction organization, etc.), cellular components (extracellular matrix, cell-cell junction, receptor complex, etc.), and molecular functions (extracellular matrix structural constituent, glycosaminoglycan binding, protein tyrosine kinase activity, etc.).We then used KEGG pathway analyses to evaluate the differentially expressed genes associated with SULF2 for potential functional pathways (focal adhesion, cell adhesion molecules, axon guidance, proteoglycans in cancer, etc.) ( Figure 5). The enrichment analysis showed that the most important functional network of SULF2 is associated with extracellular matrix and intercellular interaction, which have been proved to play a critical role in tumorigenesis and cancer progression 22 . Correlation analysis between SULF2 expression and in ltrating immune cells To comprehensively investigate the role of SULF2 in STAD TME, we analyzed the correlation between SULF2 expression and six types of in ltrating immune cells, including B cells, CD4+ T cells, CD8+ T cells, macrophages, neutrophils and dendritic cells. The results showed that SULF2 expression levels had a signi cant positive correlation with the in ltration of CD4+ T cells, CD8+ T cells, macrophages and neutrophils, and no signi cant correlations with B cells and dendritic cells in STAD ( Figure 6A).

Correlation between SULF2 expression and various immune markers
To deepen our understanding of SULF2 crosstalk with the immune response, we used the TIMER database to verify the correlations between SULF2 expression and different immune signatures in STAD. The genes listed in Table 1 were used to characterize immune cells, including B cells, T cells, CD8+ T cells, monocytes, TAMs, M1 macrophages, M2 macrophages, neutrophils, natural killer (NK) cells and dendritic cells. In clinical cancer biopsies, tumor purity is an important aspect affecting the analysis of immune in ltration. After adjusting for tumor purity, SULF2 expression was signi cantly correlated with some immune markers of monocyte and macrophage in STAD (Table 1). We further investigated the interrelationship between SULF2 expression and famous T cell checkpoints, such as PD-L1, PD-1 and CTLA-4, in the GEPIA database. SULF2 expression was signi cantly associated with the expression of PD-L1 in STAD ( Figures 6B). These ndings further supported that SULF2 expression was signi cantly related to immune in ltration and indicated that SULF2 played a key role in immune escape in STAD.

Discussion
GC is thought to be caused by the interaction of host genetic factors and complex environmental factors, and the importance of local tumor-host cell interactions in cancer biology is increasingly recognized 23 . Importantly, SULF2 is at the interface between the cancer cell and the tumor microenvironment. SULF2 performs post-synthetic editing of 6-O-sulfation on HSGAG chains, which liberate the HS-binding proteins, including VEGF, FGF, or Wnt, therefore modifying the interactions between signaling ligands and their cognate receptors 21 . Several pro-angiogenic factors like VEGFB, MMP-2, MMP-9, PDGFA and PDGFB were also overexpressed in SULF2-expressing cells 24 . SULF2 has been shown to have a cancer promoting effect in a variety of tumors, especially in hepatocellular carcinoma, but its role in gastric cancer has not been fully clari ed. In GC, Hur et al 25 have previously reported that GC tissues showed higher expression of SULF2 (p = 0.001) compared to normal gastric mucosa with a small sample size, which was correlated with its promoter hypomethylation. And mice injected with SULF2-bearing cell lines showed a signi cantly higher tumor volume compared with controls at 10 weeks post-injection. However, the sample size is small and they didn't validate whether the expression of SULF2 is a prognostic in human GC patients. Besides, SULF2 CpG island methylation status may in uence GC sensitivities to some chemotherapeutics, such as platinum and irinotecan regimen 26,27 . Our study further veri ed that SULF2 was highly expressed in gastric cancer, and showed that the expression levels of SULF2 were related to GC patient prognosis. Recruiting TAMs to escape immune regulation may be one of the mechanisms.
In the present study, we showed that the expression of SULF2 in GC was higher than that in normal gastric tissue in multiple databases ( Figure 1A, B, C). These ndings were consistent with the previous report and suggested that SULF2 may act as an oncogene by promoting the development and progression of GC 25 . And we further veri ed that the protein expression levels of SULF2 in tumor tissue were higher than that in adjacent normal tissues ( Figure 1D, E). Subsequently, the clinical characteristics of SULF2 in GC patients was investigated. The results indicated that no matter which race, gender, age, lymph nodal metastasis status and histological grade are, SULF2 is highly expressed in GC compared to corresponding normal tissue (Figure 2). The expression levels of SULF2 in tumor tissues of stage 1 patients and that in normal tissues have no statistical signi cance, but different from that in advanced tumor tissues, suggesting that SULF2 plays a critical role in tumor progression instead of tumorigenesis.
Furthermore, Kaplan-Meier survival analyses showed that GC patients with high SULF2 expression exhibited a markedly worse survival rate than those with low expression ( Figure 3A, B, C). With high SULF2 expression, GC patients are more likely to have advanced stage, and the prognosis is worse after progression, especially in those GC patients with local lymph node metastasis but without distant metastasis ( Figure 3D). These results indicated that SULF2 may be a prognostic biomarker in GC and may facilitate the development of targeted precision oncology.
We also identi ed top 50 genes co-expressed with SULF2 ( Figure 4A). Several genes of them take part in GC progression, such as LATS2, GLI2 and SULF1. As the Hippo pathway transducer, LATS2, which is located in the centrosome and works for accumulation of γ-tubulin and formation of mitotic spindle, was reported to play a pivotal role in the promotion of GC cell migration 28 . The transcription factor GLI2, as a member of the Hedgehog signaling pathway, modulated several cytokine genes in the TME and also promoted GC tumorigenesis and progression 29,30 . Zheng et al have found that SULF2 Inhibitor 2,4-Disulfonylphenyl-tert-Butylnitrone has an antitumor effect via suppression of Hedgehog/GLI1 Signaling in HCC 31 . Besides, SULF1 was elucidated as a novel prognostic and lymph nodal metastasis predictive marker, and played an oncogenic role in GC 25 . A SULF2 correlation network was then constructed to identify potential interactions between SULF2 and cancer-related gene targets ( Figure 4C). The most important function of the identi ed gene was sulfuric ester hydrolase activity, which corresponded to the function of SULF2. Due to the fact that SULF2 performs post-synthetic editing of HSGAG, PPI network may be more valuable than gene network ( Figure 4D). In the PPI network, we can see GPC3, which was the most highly expressed HSPGs in HCC 32 . GPC3, released from HSPGs by SULF2, could exerts in ammatory activity by increasing the expression of NF-κB, CRP, TNF-α, IL-1β and IL-6. SULF2 expression, which was induced by p53, then promoted IL-6 expression by stabilizing β-catenin, followed by stimulation of the STAT3/Bcl-XL pathway 33 . Though SULF2 was con rmed to be a direct transcription target of p53 34 , SULF2 might also regulate prion-like behavior of p53 through remodeling HSGAG Sdomains in cancer 35 . This interaction between SULF2 and p53 may promote the progression of GC. In GC, IL-6 could increase the density of TAMs in TME 36 , which was further clari ed in the following text.
Consistently, the results of enrichment analysis showed that SULF2 played an important role in tumorhost cell interaction in extracellular matrix ( Figure 5).
The expression level of SULF2 was positively correlated with the in ltration of CD4 + T cells, CD8 + T cells, macrophages and neutrophils in GC ( Figure 6A). And the correlations between SULF2 and immune markers of different immune cells were further validated ( Table 1). The results indicated that TAMs were signi cantly related to SULF2, which tallied with the recognition that TAMs might predict a poor OS in GC 37 . As previously mentioned, we hypothesized that SULF2 could release some in ammatory cytokines, like IL-6, in the extracellular space by a direct or indirect means to recruit TAMs. And SULF2-mediated upregulation of STAT3 may promotes TAMs polarization to the M2 phenotype 38, 39 , which has been reported to have ability to promote GC invasion, migration and angiogenesis [40][41][42][43] . According to the results of GO and KEGG, SULF2 is signi cantly related to the structure and function of extracellular matrix. The reasons might be as following: (1) The desulfation function of SULF2 directly regulates the components and functions of extracellular matrix.
(2) TAMs recruited by SULF2 secrete matrix metalloproteinase 37 , thus acting on extracellular matrix indirectly. Besides, the expression of PDL1 was associated with SULF2 ( Figure 6B). And TAMs have been reported that it could increase the expression of PDL1 to help GC escape immune regulation 44 . These ndings con rmed that SULF2 may help GC cell escape immune regulation via TAMs in TME, so as to promote tumor progression.
Chrysin was reported to have the ability to inhibit GC cells invasion 15 . There is a study suggested that chrysin has the antiproliferative activity against HCC through the suppression of SULF2 45 . So chrysin antiproliferative action against GC cells might also be attributed to the suppression of SULF2. Inhibitors of SULFs are under study 12 , and, once these have been developed to the stage where they are suitable for cell-based and preferably in vivo studies, the therapeutic potential of SULF2 inhibition can be explored. It is possible that such compounds could counteract the effect of SULF2 by sequestering growth factors released in the extracellular and pericellular space 18 . At present, some studies have tried to improve the prognosis of GC patients by inhibiting TAMs in ltration, such as pexidatinib 46 and sophoridine 47 . We found a positive correlation between SULF2 and TAMs in ltration, so SULF2 may be used as a biomarker for targeted TAMs treatment in the future.

Conclusions
Our study found the high expression of SULF2 in GC compared to normal tissue. And its high expression was related to the poor prognosis of GC. The potential mechanism may be direct or indirect releasing in ammatory factors in TME upon the desulfation function of SULF2, and then recruit TAMs, which leads to tumor progression. Therefore, these ndings revealed that SULF2 may serve as a candidate prognostic for determining the prognosis of GC associated with immune in ltration.

Methods
TIMER TIMER (http://timer.cistrome.org/) is a comprehensive resource for systematical analysis of immune in ltrates across diverse cancer types 48 . In the present study, we used the "Gene_DE" module to analyze SULF2 expression in multiple types of cancer. In STAD, the correlation of SULF2 and immune cell in ltration (B cells, CD8+ T cells, CD4+ T cells, neutrophils, macrophages, and dendritic cells) was evaluated through the "Gene" module. We also applied TIMER to investigate the relationship between SULF2 expression and different gene marker sets of immune cells by using the "GENE_Corr" module.

Kaplan-Meier Plotter Database Analysis
The Kaplan-Meier Plotter (http://kmplot.com/analysis/) was used to analyze the prognostic value of SULF2 in STAD 52 . The patient samples were divided into high and low groups by median expression to analyze FP, OS and PPS with hazard ratios, 95% con dence intervals and logrank p-values.

LinkedOmics
The "LinkFinder" module of LinkedOmics (http://www.linkedomics.org/login.php) was used to identify differentially expressed genes related to SULF2 (N=415) in the TCGA STAD section 53 . The search and target datasets were obtained by RNA-seq, and the results were analyzed with the Pearson correlation coe cient. Enrichment analysis was performed for Gene Ontology and KEGG analyses through "LinkInterpreter" module.