The nine ADAMs family members serve as potential biomarkers for immune infiltration in pancreatic adenocarcinoma

Background The functional significance of ADAMs family members in the immune infiltration of pancreatic adenocarcinoma (PAAD) awaits elucidation. Methods ADAMs family members with significant expression were identified among differentially expressed genes of PAAD based on The Cancer Genome Atlas (TCGA) database followed by a verification based on the Oncomine database. The correlation of ADAMs in PAAD was estimated with the Spearman’s rho value. The pathway enrichment of ADAMs was performed by STRING and GSEALite, respectively. The protein–protein interaction and Gene Ontology analyses of ADAMs and their similar genes were exanimated in STRING and visualized by Cytoscape. Subsequently, the Box-Whisker plot was used to show a correlation between ADAMs and different tumor grade 1/2/3/4 with Student’s t-test. TIMER was applied to estimate a correlation of ADAMs expressions with immune infiltrates and immune checkpoint blockade (ICB) immunotherapy-related molecules. Furthermore, the effect of copy number variation (CNV) of ADAMs genes was assessed on the immune infiltration levels. Result ADAM8/9/10/12/15/19/28/TS2/TS12 were over-expressed in PAAD. Most of the nine ADAMs had a significant correlation. ADAM8/12/15/19 expression was remarkably increased in the comparison between grade 1 and grade 2/3 of PAAD. ADAM8/9/10/12/19/28/TS2/TS12 had a positive correlation with almost five immune infiltrates. ADAM12/19/TS2/TS12 dramatically related with ICB immunotherapy-related molecules. CNV of ADAMs genes potentially influenced the immune infiltration levels. Conclusion Knowledge of the expression level of the ADAMs family could provide a reasonable strategy for improved immunotherapies to PAAD.


INTRODUCTION
Pancreatic adenocarcinoma (PAAD) represents 85% all pancreatic cancer and is frequently described one of the most lethal malignancies with a poor outcome, although a limited number of PAAD patients can undergo routine adjuvant therapies after surgery (Higuera et al., 2016;Neoptolemos, 2011). In recent decades, more in-depth studies empower the immunotherapy on the clinical treatment for patient with PAAD. The immunotherapy has been used for the PAAD treatment accounting for an immune regulatory in the tumor microenvironment (TME) (Fan et al., 2020). In the TME, immunocytes resemble the features of sampling the microenvironment and recognizing the antigens, which originate from tumor cell (Nazemi & Rainero, 2020). The fact that immunotherapy is successful in treatment for malignancies like PAAD pleas for a better understanding of the molecular mechanism by which tumor-associated antigens (TAAs) can be recognized and captured by immunocytes.
TAAs, often originated from cell surface, are predominantly produced by the processing of membrane-associated protease (Vigneron, 2015). It has previously been observed that cancer cells often express an elevated level of the membrane-associated metalloprotease, such as a-disintegrin-and-metalloproteinases (ADAMs) (Murphy, 2008). ADAMs is a family of transmembrane proteases, which release substrates by cleaving the ectodomains from cell surface proteins (Klein & Bischoff, 2011). ADAMs family members cover a wide spectrum of potential substrates, and various cleaved ectodomains are soluble and biologically active (Moss et al., 2016;Reiss & Bhakdi, 2017). The functional complexity of substrates enables ADAMs family as an important participant in extracellular interaction. Overexpression of ADAMs is not oncogenic, which was exemplified as ADAM9 harboring angiogenetic function without an impact on tumor cell proliferation (Oria et al., 2019), but due to a cancer characteristic of cleavage products (Herrlich & Herrlich, 2017). Some of the ADAMs family members have been reported as an important role in the progression of pancreatic cancer. For example, ADAM8 and ADAM28 were regarded as a potential therapeutic target in pancreatic cancer (Schlomann et al., 2015;Valkovskaya et al., 2007;Wei et al., 2019). ADAM10 is remarkably involved in the invasiveness and migration of pancreatic cancer cells (Sepult et al., 2019). In addition, ADAMs can serve as a specific resource to regulate the cell-cell or cell-environment interaction, such as immune infiltration. Therefore, an analysis of the ADAMs family members may offer new insights on regulation of immune infiltration in PAAD, which may optimize immunotherapeutic strategies for PAAD patient.
In the present study, we carried out an initial identification of nine ADAMs from differently expressed genes (DEGs) of PAAD based on both TCGA and Oncomine databases, followed by a correlation of analysis. The biological function and proteinprotein interaction (PPI) of all nine ADAMs and their similar genes was performed. After correlation analysis of tumor grade, we studied a correlation of ADAMs expression with immune infiltration in PAAD at molecular and immunocyte levels. Furthermore, we continued to examine the copy number variation (CNV) of ADAMs to explore the potential factors affecting the association between ADAMs and immune infiltration. This integrated analysis has facilitated biological insights of ADAMs family members in PAAD, contributing to influence PAAD clinical management, especially the immunotherapeutic strategies.

Identification of differently expressed genes in PAAD
GEPIA (http://gepia.cancer-pku.cn/index.html) is an interactive web tool for analyzing the RNA sequencing expression data of 9,736 tumors and 8,587 normal samples from the TCGA. It can provide a customizable function to analyze the expression profiling of genes between tumor and normal samples (Tang et al., 2017). In this study, GEPIA was used to analyze the expression of nine ADAMs family members in PAAD. Differentially expressed ADAMs were identified with higher |log 2 Fold change (FC)| and lower q values than a pre-set threshold (|log 2 FC|> 2 and q < 0.01). Oncomine database (https://www.oncomine.org) is a web resource to provide the different expression profile of genes from gene-wide expression analysis (Rhodes et al., 2004). In our study, we performed a summary view in PAAD to verify the differently expressed nine ADAMs family members from TCGA database. We set a threshold including P-value as 0.01, Fold change as 1.5, and gene rank as 10%.

GO and pathway enrichment analysis
STRING (https://string-db.org/cgi/input.pl) is a web-based tool to provide an enriched analysis of known and predicted proteins (Szklarczyk et al., 2019). Now the STRING database (version 11.0) currently covers 24584628 proteins from 5090 organisms. We obtained the GO enrichment of ADAMs and their similar genes in three categories including Biological Process (BP), Molecular Function (MF) and Cellular Component (CC).
GSEALite (http://bioinfo.life.hust.edu.cn/web/GSCALite/) is a web-based platform with an integration of cancer genomics data of 33 cancer types from TCGA (Liu et al., 2018). We used the Pathway Activity module to present the difference of ADAMs expression between pathway activity groups.

The protein-protein interaction of ADAMs and their similar genes in PAAD
Cytoscape is a bioinformatics software with multiple plugins to analyze and virtualize protein-protein interaction (PPI) and provide a molecular network connectivity (Shannon et al., 2003). We firstly obtained 20 similar genes of each ADAM gene in PAAD through an expression analysis in GEPIA 2.0 (http://gepia2.cancer-pku.cn/#index). Then we deleted duplicate gene and analyzed the PPI of ADAMs and their similar genes in STRING followed by a visualization processing in Cytoscape.

Tumor grade correlation with ADAMs
UALCAN (http://ualcan.path.uab.edu) is an online web to perform in silico validation of potential genes of interest and valuable information, such as clinico-pathologic factors (Chandrashekar et al., 2017). Here we explored expression profile of ADAMs based on tumor grade. Box-Whisker plot showed a correlation between ADAMs and different tumor grade 1/2/3/4. The significance of difference was estimated by Student's t -test considering unequal variance, and P < 0.05 was considered as statistically significant.

Tumor immune estimation of ADAMs
TIMER (https://cistrome.shinyapps.io/timer/) is a comprehensive resource for systematical analysis of immune infiltrates with six immune infiltrates (B cells, CD4+ T cells, CD8+ T cells, Neutrophils, Macrophages, and Dendritic cells) (Li et al., 2016;Li et al., 2017). We used a correlation module of TIMER to draw the expression scatterplots between a pair of ADAMs with a purity-corrected partial Spearman's rho value and statistical significance, as well as a pair of ADAMs with immune checkpoint blockade (ICB) immunotherapy-related molecules. These molecules have been previously reported as key targets of immune checkpoint inhibitors: programmed death 1 (PDCD1) and its ligand 1 (CD274, also known as PD-L1), and its ligand 2 (PDCD1LG2, also known as PD-L2), cytotoxic T-lymphocyte antigen 4 (CTLA4), T-cell immunoglobulin domain and mucin domain-containing molecule-3 (HAVCR2, also known as TIM-3), and indoleamine 2,3-dioxygenase 1 (IDO1). We then obtained a correlation of ADAMs expressions with immune infiltration levels. Subsequently, we used SCNA module to compare tumor infiltration levels with different somatic copy number alterations of ADAMs. The infiltration level for each SCNA category was compared with the normal using a two-sided Wilcoxon rank-sum test. We also used CNV module in GSEALite to analyze the copy number variations of ADAMs in PAAD. UCSC Xena (http://xena.ucsc.edu) is an online exploration tool for integrating the any functional genomics data within the Xena Browser. We used Xena platform to search copy number of ADAMs in PAAD patient.

Different expression of ADAMs family members in patients with PAAD
The first section aimed to identify different ADAMs family members in PAAD. Based on TCGA database, we extracted more than 2600 DEGs in PAAD. Among these DEGs, significantly higher expression of ADAM8/9/10/12/15/19/28/TS2/TS12 were identified with log 2 FC from 2.194-3.53 (all P < 0.001) (Table 1), which implied that all these ADAMs were over-expressed in PAAD (Fig. 1).
We further verified the expression profile of these nine ADAMs in Oncomine database. There was more than one dataset from Oncomine to exhibit the over expression of all nine ADAMs in PAAD (Table 2). These data established that nine ADAMs genes were aberrantly activated in patients with PAAD.

Pathways enrichment of ADAMs in PAAD
The pathways of the nine ADAMs were enriched with GSCALite. We found that 10 pathways were associated with the function of the nine ADAMs in PAAD. The expressions     of ADAM8/12/19/TS2/TS12 were in a remarkably high level in epithelial-mesenchymal transition (EMT) among 10 pathways (Fig. 2). The expressions of nine ADAMs were almost inhibited in DNA damage repair, cell cycle, and hormone androgen receptor (AR) related pathway. Taken together, these results showed the biological function and pathways of the nine ADAMs in PAAD, further suggesting ADAMs were implicated in extracellular physiological processes.

PPI and GO analysis of ADAMs and their similar genes in PAAD
After analyzing biological functions of ADAMs, we continued to investigate the similar genes of ADAMs in PAAD, which play synergistic roles in PAAD development. The network analysis of PPI illustrated that the similar genes of ADAMs had a dramatical connectivity with each other, which validated that there was a synergistic function of similar genes (Fig. 3). GO analysis revealed that similar genes of ADAMs were remarkably related to the extracellular matrix organization and cell adhesion (Table S1). And these similar  genes were significantly associated with cellular component, including extracellular matrix, extracellular region part, extracellular matrix component, collagen-containin, extracellular matrix, and extracellular region (Table S1). Our results provided more evidence for the role of ADAMs and their similar genes in extracellular biological function.

Correlation between ADAMs expression and tumor grade in PAAD
To further our understanding of the role of ADAMs in the development of PAAD, we assessed the expression of ADAMs based on tumor grade for PAAD. We found that varying grades of tumor were significantly correlated with 6 ADAMs, including ADAM8/10/12/15/19/TS12 (Fig. 4). The expression of ADAM8/12/15/19 was apparently increased in the comparison between grade 1 and grade 2/3. Interestingly, ADAM10 presented almost a grade-wide influence with statistical significance. In short, we concluded that some ADAMs played a potential role on tumorigenesis and development of PAAD.

Correlation between ADAMs expressions and immune infiltration levels in PAAD
Given the potential role of ADAMs in the cell-environment interaction, we performed a correlation of ADAMs expression with immune infiltration level in PAAD (Fig. 5). When there was no immune infiltrate in PAAD, 8 ADAMs had a different expression with no significance, except ADAMTS12 that was negatively related to purity state of tumor cell. The expression of 8 ADAMs positively correlated with B cell without ADAM15. Intriguingly, ADAM15 level had no significant variation with either of 6 immune infiltrates. Except for ADAM15, the other eight ADAMs had remarkably positive correlations with almost five immune infiltrates, especially dendritic cell. ADAM8 expressed differently with only dendritic cell. These results provided important insights into the correlation between ADAMs and immune infiltration in PAAD patients. The association between nine ADAMs and ICB therapy-related genes was performed to assess the possible role of ADAMs family members in the immunotherapy of ICB in PAAD (Fig. 6). ADAM 12/19/TS2/TS12 were positively related to all six ICB immunotherapyrelated molecules (cor > 0, P < 0.001), whereas ADAM8/9/15/28 had no relevance with either of six molecules. There was a forward trend of all six molecules with ADAMs, except ADAM15. Our findings implied a possible role of ADAMs in ICB therapy.

Immune infiltration level in PAAD with copy number variations of ADAMs
Regarding that different expressions of ADAMs involved in the regulation of immune infiltration, we investigated the somatic copy number variation of ADAMs in PAAD (Fig. 7). We found that an arm-level alteration of ADAM8/9/10/12/19/28/TS2 genes attenuated the immune infiltration of B cell and CD4+T cell. An arm-level alteration of ADAM8/12/19/TS2 genes also decreased the immune infiltration of neutrophil. The high amplification of ADAM15 gene influenced the infiltration levels of B cells, CD4+ T cells, CD8+ T cells, Neutrophils, and Macrophages. And an arm-level alteration of ADAM15 gene decreased the immune infiltration levels of CD4+ T cells, CD8+ T cells, and Neutrophil. Interestingly, dendritic cells had no influence by CNV of ADAMs genes. These results suggested that B cell and CD4+T cell infiltrates were particularly affected by CNV of ADAMs in patient with PAAD.
We next investigated the CNV of ADAMs in PAAD. An analysis of CNV% showed that a heterozygous variation of ADAM15 gene accounted for a most heterozygous amplification without deletion. ADAM9/28 genes presented a relatively high probability of heterozygous deletion (Fig. 8). We also used Xena to analyze copy numbers of ADAMs from 186 samples from TCGA (Fig. 9). In some PAAD samples, the copy number of ADAM9 increased more than 2.8 times, as well as those of ADAM 15/28. ADAM15 had a most increase of copy number and ADAM28 had a most decrease in PAAD samples, which were directly in line with the results from the CNV% analysis (Fig. 8). As a result, CNV of ADAMs family members held their own variations, which may influence immune infiltration level in PAAD.

DISCUSSION
The identification of ADAMs in a collection of data derived from TCGA and Oncomine databases confirms the involvement of these genes in development and progression of PAAD. Despite some ADAMs have been reported to play important roles in tumor, distinct role of ADAMs family members in PAAD awaits clarification. In this study, the expression of ADAMs was identified followed by an analysis of their biological functions. Given that ADAMs had a potential influence on the cell-environment interaction, the correlation between ADAMs and immune infiltration in PAAD was established and analyzed. The Infiltration Level ADAM8 Expression Level (log2 TPM) Infiltration Level ADAM10 Expression Level (log2 TPM) Infiltration Level ADAM12 Expression Level (log2 TPM) Infiltration Level ADAM28 Expression Level (log2 TPM) Infiltration Level ADAMTS2 Expression Level (log2 TPM)