Identification of the MMP family as therapeutic targets and prognostic biomarkers in the microenvironment of head and neck squamous cell carcinoma

Oncomine, as a bioinformatics database that can collect and analyze cancer transcriptome data, provides powerful genome-wide expression analysis [9]. The identification of key genomic biomarkers is conducive to the diagnosis, prognosis and treatment of diseases [10]. In this study, data was extracted to evaluate the expression of the MMP family in HNSC, where P < 0.05, FC ≥ 2, and the top 10% of genes were the significance thresholds, and t test was used to analyze the expression differences of the MMP family in HNSC.

By obtaining data from TCGA, UALCAN can be used not only to assess the expression of protein-coding genes, but also to conduct in-depth analysis of clinical data in 33 cancers [11]. In this study, expression data of the MMP family was obtained through “Expression Analysis”module and “KIRC” dataset, and t test was used for analysis, with P < 0.05 as the significance threshold.

GEPIA is an analysis tool based on TCGA and GTEx data, which contains RNA sequence expression data of 9736 tumors and 8587 normal tissue samples [12]. In this study, GEPIA single gene analysis was used to analyze the difference of mRNA expression between tumor tissue and normal tissue, pathological staging analysis and prognosis analysis of the MMP family. The HNSC dataset was used to analyze the MMP family by polygene comparison. T test analysis was used, with P < 0.05 as the significant threshold, and Kaplan–Meier curve was used for prognosis analysis.

cBioportal is a platform for exploring, visualizing, and analyzing multi-dimensional cancer genomic data. cBioportal contains over 200 cancer genomics studies from the TCGA database [13]. In this study, genetic alteration, co-expression and network modules of the MMP family were obtained from cBiopartal based on TCGA database. A total of 564 HNSC specimens were analyzed.

GeneMANIA, based on genomic, proteomic and gene functional data, aims to provide information on protein-genetic interactions, pathways, co-expression, co-localization, and similarity of protein domains of submitted genes [14].

STRING aims to collect, score, and integrate protein–protein interaction data from all publicly available sources, and to predict and supplement these data through potential function calculations [15]. In this study, PPI network analysis was performed on the different-expressed the MMP family to explore the interaction between this family and STRING.

DAVID6.8 provides a method for elucidating the biological functions of the submitted genes [16]. In this study, GO enrichment analysis and KEGG pathway enrichment analysis of the MMP family and adjacent genes were isolated from DAVID6.8, including BP, CC, MF.

TRRUST contains 8,444 TF regulatory relationships of 800 human transcription factors, which can provide how these interactions are regulated [17], and is an intuitive and reliable tool for human transcriptional regulatory networks.

TIMER provides a systematic evaluation of different immune cell infiltrates and their clinical effects [18]. In this study, the gene module was used to evaluate the correlation between the MMP family levels and immune cell infiltration, and the survival module was used to evaluate the correlation between clinical outcomes and immune cell infiltration and the MMP family.

Linkedomics contains a multiomics data analysis of 32 TCGA cancer types [19]. In this study, the biological analysis of the enrichment of the MMP family kinase target was carried out using LinkInterpreter. GSEA was used for at least 3 genes and 500 simulations in the HNSC datasets. Spearman correlation test was adopted, and P < 0.05 was the significant threshold.

We used Oncomine database to investigate mRNA expression of the MMP family in different tumor types and to detect their levels in different HNSC datasets. The database contained mRNA expression of 24 MMP family members in 20 tumors. MMP1 mRNA expression was significantly different in 77 studies. In 73 of 77 studies, 16 of 20 tumors were observed to have higher levels of mRNA expression than normal tissue, and only 2 of the remaining 4 studies had lower levels of mRNA expression than normal tissue; The mRNA expression levels of MMP2 and MMP7 in 13 of 20 tumors were higher than those in normal tissues. The mRNA expression levels of MMP3, MMP10, MMP12 and MMP16 in 11 tumors were higher. The mRNA expression levels of ILF3 and MMP9 were higher than those of normal tissues. Among 439 studies of MMP11, 107 studies showed significant differences and among 97 studies, 15 tumor mRNA expression levels were higher than normal tissues. MMP13 mRNA was highly expressed in 9 kinds of tumors. MMP14 was highly expressed in 12 tumors. MMP19 was highly expressed in 10 tumors. The expression of MMP15, MMP17, MMP24 and MMP28 were low in most of the 20 tumors. The mRNA expression levels of MMP8, MMP20, MMP21, MMP23B, MMP25, MMP26 and MMP27 were not significantly different among different tumors. In 24 HNSC data sets, MMP1, MMP3, ILF3, MMP7, MMP9, MMP10, MMP11, MMP12, MMP13, and MMP16 were expressed at higher levels in most tumor tissues than in normal tissues, while MMP15 was expressed at lower level (Fig. 1).

Detailed data on independent HNSC datasets with significant differences in mRNA expression was listed in Oncomine (Table 1) [20‐25]. From the table, MMP1 and MMP10 were highly expressed in oral squamous cell carcinoma. The expression of MMP3 and ILF3 was up-regulated in tongue squamous cell carcinoma. MMP7, MMP9, MMP11, MMP12 and MMP13 were all up-regulated in head and neck squamous cell carcinoma. Compared with normal tissues, MMP16 expression was up-regulate in thyroid papillary carcinoma. MMP15 was down-regulated in all 14 HNSC tumors, including Buccal Mucosa Squamous Cell Carcinoma, Floor of Mouth Squamous Cell Carcinoma, Gingival Squamous Cell Carcinoma, Glottis Squamous Cell Carcinoma, Hard Palate Squamous Cell Carcinoma, Lip Squamous Cell Carcinoma, Maxillary Sinus Squamous Cell Carcinoma, Oral Cavity Squamous Cell Carcinoma, Oropharyngeal Squamous Cell Carcinoma, Postcricoid Squamous Cell Carcinoma, Soft Palate Squamous Cell Carcinoma, Supraglottic Squamous Cell Carcinoma, Tongue Squamous Cell Carcinoma, Tonsillar Squamous Cell Carcinoma.

According to the UALCAN analysis, MMP2 (P = 3.81E−04), MMP3 (P = 1.62E−12), ILF3 (P = 1.62E−12), MMP8 (P = 3.33E−08), MMP9 (P = 1.62E−12), MMP10 (P = 9.30E−10), MMP12 (P = 1.62E−12), MMP14 (P = 1.62E−12), MMP15 (P = 1.28E−06), MMP16 (P = 1.35E−04), MMP17 (P < 1E−12), MMP19 (P = 3.33E−15), MMP20 (P = 6.96E−04), MMP23B (P = 1.66E−03), MMP25 (P = 3.90E−07) and MMP28 (P < 1E−12) mRNA transcription levels were higher than those of normal tissues. The transcription level of MMP27 (P = 5.64–03) was significantly decreased (Fig. 2). We also compared the relative expression levels of the MMP family in HNSC, and found that MMP1 and MMP14 were the highest relative expression levels among all the MMP family members in HNSC tissues (Fig. 3).

We then evaluated the correlation between the MMP family expression and pathological stage in HNSC patients, and found that the expression of MMP3 (F = 3.14, P = 0.025), MMP11 (F = 3.25, P = 0.025) and MMP25 (F = 5.32, P = 0.001) were significantly correlated with pathological stage (Fig. 4). With the development of HNSC, the expression of MMP3 and MMP11 increased, and the expression of MMP25 decreased significantly.

To assess the value of the MMP family in the progression of HNSC, we used GEPIA to evaluate the association of differential expression of the MMP family with clinical outcomes. The Kaplan–Meier curve and log-rank test analysis revealed that there was no significant correlation between high and low transcription levels of all the MMP family members with disease-free survival rate (Fig. 5). It was found that increased mRNA level of MMP1 (P = 0.045), When exploring the correlation between the MMP family expression and overall survival in HNSC patients. MMP8 (P = 0.005) and MMP25 (P = 0.002) were significantly correlated with overall survival (P < 0.05), while no significant difference was found in other MMP (Fig. 6).

The genetic changes of the MMP family gene were analyzed and described as the tumors with mutation, amplification, deep deletion, high mRNA level and multiple changes. MMP1, MMP3, MMP7, MMP8, MMP10, MMP12, MMP13, MMP20 and MMP27 all had ≥ 9% genetic changes. Furthermore, the amplification and depth deletion were greater, while the other MMP had only minor genetic changes (Fig. 7A). GeneMANIA was used to analyze the correlation of the MMP family and its adjacent genes at gene level, found that MMP1, MMP2, MMP3, ILF3, MMP7, MMP8, MMP9, MMP10, MMP11, MMP12, MMP13, MMP14, MMP15, MMP16, MMP17, MMP19, MMP20, MMP21, MMP23B, BMMP24, MMP25, MMP26, MMP27, MMP28 were closely related to HPX, CTB-96E2.2, PRG4, VTN, ASTL, MEP1B, MEP1A, MFAP2, ILF2, BSPH1, ELSPBP1, STRBP, IEF2R, TLL2, TLL1, ENDOU, ZFR2, ZFR, BMP1, SEL1L (Fig. 7B). The interaction at the expression level of MMP protein was determined by STRING analysis. In this analysis, other MMP proteins except ILF3, MMP19, MMP20, MMP21, MMP23B, MMP27, and MMP28 interacted (Fig. 7C).

For a deeper understanding of the MMP family, GO enrichment analysis and KEGG pathway enrichment analysis of the MMP family and adjacent genes were isolated from DAVID6.8 in this study. GO enrichment analysis included BP, CC, MF. As shown in Fig. 8, the MMP family and adjacent genes in BP were the most enriched in the process of collagen metabolism and collagen catabolism of extracellular structures and tissues of extracellular matrix, and the enrichment was most significant in gastral action (Fig. 8A). In MF, the number of genes enriched in the activities of endopeptidase, metallopeptidase and endometal peptidase increased and the enrichment was significant in the activity of exopeptidase (Fig. 8B). Genes in CC were enriched in the collagen-containing extracellular matrix (Fig. 8C). KEGG analysis showed that a large number of these genes were enriched in the synthesis and secretion of parathyroid hormone, and were significantly enriched in the transcriptional regulation of cancer (Fig. 8D).

MMP1, MMP2, MMP3, MMP7, MMP9, MMP10, MMP11, MMP12, MMP13, MMP14, MMP17, MMP20, and MMP28 were contained in TRUST. We found that 21 transcription factors were related to the regulation of the MMP family chemokines. JUN was a key transcription factor of MMP1, MMP2, MMP3, MMP7, MMP9, MMP12, MMP13, and MMP20; STAT3 was a key transcription factor of MMP1, MMP2, MMP7, MMP9, MMP10, and MMP14; ETV4 was a key transcription factor of MMP1, MMP2, MMP7, and MMP14; ETS1 was a key transcription factor of MMP1, MMP3, MMP9, MMP10, MMP13; RELA was a key transcription factor of MMP1, MMP2, MMP3, MMP9, MMP12, MMP13, MMP14; ETS2 was a key transcription factor of MMP1, MMP2, MMP3 and MMP9. MAZ was a key transcription factor of MMP1, MMP9 and MMP14. FOS was a key transcription factor of MMP1, MMP3, MMP7 and MMP9. NFKB1 was a key transcription factor of MMP1, MMP2, MMP3, MMP9, MMP13, MMP14; NFKBIA was a key transcription factor of MMP1, MMP3 and MMP9; SRF was a key transcription factor of MMP2, MMP9 and MMP14; NCOA3 was a key transcription factor of MMP7 and MMP10; KLF8 was a key transcription factor of MMP9 and MMP14; SNAI2 was a key transcription factor of MMP9 and MMP17; SP1 was a key transcription factor of MMP2, MMP9, MMP11 MMP14, MMP28; RUNX2 was a key transcription factor of MMP2 and MMP13; CTNNB1 was a key transcription factor of MMP7 and MMP14; YBX1 was a key transcription factor of MMP2 and MMP13. Both TWSIT1 and TP53 were key transcription factors of MMP1 and MMP2; PPARG was a key transcription factor of MMP1 and MMP9; HDAC1 was a key transcription factor of MMP9 and MMP28; TFAP2A was a key transcription factor of MMP2 and MMP9; STAT1 was a key transcription factor of MMP9 and MMP13 (Table 2).

We identified the first two kinase targets of the MMP family from the LinkedOmics database. ATR and ATM were the most common first two kinase targets in the MMP family. ATR and ATM kinase targets were found in the first two kinases of MMP1, MMP2, MMP3, ILF3, MMP9, MMP10, MMP11, MMP13, MMP15, MMP17, MMP23B and MMP24 (Table 3).

In this study, the gene module was used to evaluate the correlation between the MMP family levels and immune cell infiltration. The survival module was used to evaluate the correlation between clinical outcomes and immune cell infiltration and the MMP family. The expression level of MMP1 was negatively correlated with B cells and CD8 + T cells, positively correlated with neutrophil infiltration level, and not significantly correlated with CD4 + T cells, macrophages and dendritic cells infiltration level. The expression levels of MMP2, MMP7 and MMP11 had no significant correlation with the infiltration levels of CD8 + T cells, but had a significant positive correlation with the infiltration levels of other immune cells. The expression level of MMP3 was positively correlated with neutrophil infiltration level. It was found that ILF3, MMP9, MMP12, MMP19, MMP25 were significantly positively correlated with the levels of immune cells. MMP8 was significantly positively correlated with the levels of infiltration of CD4 + T cells, macrophages and dendritic cells. The expression of MMP10 was negatively correlated with B cells and CD8 + T. The expression level of MMP13 was negatively correlated with CD8 + T cells, positively correlated with the infiltration level of CD4 + T cells, macrophages, neutrophils and dendritic cells, but not significantly correlated with the infiltration level of B cells. MMP14 and MMP16 showed significant positive correlation with the infiltration levels of CD4 + T cells, macrophages, neutrophils and dendritic cells, but no correlation with other immune cells. MMP15 was negatively correlated with neutrophil infiltration level, positively correlated with CD4 + T cells and macrophages infiltration level, and had no correlation with B cells, CD8 + T cells and dendritic cell immune cells. MMP17 was negatively correlated with B cells, CD8 + T cells and dendritic cells, but had no correlation with other immune cells. MMP20 was negatively correlated with CD8 + T cells, positively correlated with B cells and CD4 + T cells, and had no correlation with other immune cells. There was no correlation between MMP21 and neutrophil infiltration, but a significant positive correlation between MMP21 and other immune cells. MMP23B was significantly positively correlated with the infiltration levels of B cells, CD4 + T cells, macrophages and dendritic cells, while the other immune cells had no correlation. MMP24 was negatively correlated with neutrophil infiltration level, positively correlated with B cell infiltration level, and had no correlation with other immune cells. There was no correlation between MMP26 and the levels of immune cell infiltration. MMP27 was not correlated with macrophages and neutrophils, but positively correlated with other immune cells. MMP28 was positively correlated with the infiltration levels of CD4 + T cells, neutrophils and dendritic cells, but not with other immune cells. The results showed that the influence of gene expression on the microscopic characterization of immune infiltration was extremely complex and variable, reflecting the heterogeneity and complexity of the immune microenvironment (Fig. 9).

Cox proportional risk model was used and the following confounders were corrected: B cells, CD8 + T, CD4 + T cells, neutrophils, dendritic cells, MMP1, MMP2, MMP3, ILF3, MMP7, MMP8, MMP9, MMP10, MMP11, MMP12, MMP13, MMP15, MMP17, MMP20, MMP21, MMP23B, MMP24, MMP25, MMP 26, MMP27, MMP28, macrophages, MMP14, MMP16 and MMP19 were significantly associated with clinical outcomes in HNSC patients (Table 4).