Identification of a Prognostic Pyroptostic-Related Model for Head and Neck Squamous Cell Carcinoma Based on LASSO-Cox Regression Analysis

Pyroptosis is associated with the biological behavior of the tumor and with tumor immunity. We investigated the effect of pyroptosis on the tumor microenvironment and tumor immunity in head and neck squamous cell carcinoma (HNSCC). RNA sequencing data and clinical information of HNSCC were downloaded from TCGA. Differentially expressed pyroptosis-related genes in HNSCC were identified between HNSCC and normal tissue. Pyroptosis-related classification of HNSCC was conducted based on consensus clustering analysis. LASSO-Cox regression analysis was used to construct a prognostic risk model-based pyroptosis-related gene. Evaluation of the immune microenvironment was conducted in prognostic risk signature based on pyroptosis-related genes. Total 22 differentially expressed pyroptosis-related genes were identified in HNSCC. Six prognostic-related genes were included to construct a LASSO regression model with a prognostic risk score = (0.133 ∗ GSDME (DFNA5) + 0.084 ∗ NOD1 + 0.039 ∗ IL6 + 0.003 ∗ IL1B + 0.084 ∗ CASP3 + 0.028 ∗ NLRP2). Higher fraction of resting memory CD4+ T cells and macrophages M1 was infiltrated in the high-risk group compared with the low-risk group in HNSCC. Furthermore, the PI3K-Akt signaling pathway and the IL-17 signaling pathways were identified to be involved in the development of high-risk HNSCC. Our study constructed a prognostic risk signature based on pyroptosis-related genes, which emphasizes the critical importance of pyroptosis in HNSCC and provided a novel perspective of HNSCC therapy.


Introduction
Head and neck squamous cell carcinoma (HNSCC) ranks as the seventh most common cancer worldwide, which leads to 450,000 deaths in 2018 [1,2]. Tobacco use, alcohol consuming, and HPV infection are the main risk factors, especially HPV-positive associated HNSCC, which is increasing among young people in the western world [1]. Multiple therapeutic interventions, including surgery, radiation therapy, chemotherapy, and immunotherapy, are recommended for HNSCC according to the anatomical sites and the degree of severity of HNSCC [1]. Despite tremendous eforts to therapy intervention for HNSCC, low therapeutic response to HNSCC remains a global challenge [3]. Te prognosis in HNSCC varies dramatically depending on risk factors, perineural invasion and extensive of invasion [4]. In addition, recurrent and metastatic HNSCC still have a poor prognosis [5]. Terefore, understanding the tumor biology of HNSCC will help us predict the prognosis and identify novel therapeutic targets for HNSCC. Recently, pyroptosis in cancer cell death induced by targeted therapy ad chemotherapy has gained more and more attention in cancer management [6,7]. Pyroptosis, a kind of regulated cell death pathway [8], is characterized by a distinct infammatory outcome, accompanied by the release of many infammatory cytokines release. It presents with cellular swelling and rapid rupture. Te Gasdermin (GSDM) superfamily, including GSDMA, GSDMB, GSDMC, GSDMD, and GSDME (DFNA5), are activated and cleaved during pyroptosis, which is regulated by infammatory caspases [8]. Several studies have demonstrated that pyroptosis increases cancer immune therapy [9][10][11][12]. However, the role of pyroptosis in HNSCC remains unclarifed. In this study, our objective was to identify the signature to predict the prognosis of HNSCC and help provide a novel therapy strategy for HNSCC.

Data Preprocess.
RNA sequencing data and clinical information of HNSCC was downloaded from TCGA (https://tcga-data.nci.nih.gov) database. Total 542 HNSCC and 44 normal human samples were included in this study. Te gene expression profles in HNSCC were normalized to fragment per kilobase million (FRKM).
Diferentially expressed pyroptosis-related genes (DEPG) between HNSCC and normal tissues were identifed by "limma" package. Te heatmap and bar plots were constructed to depict the expression of DEPG. Furthermore, the correlation among 33 genes in HNSCC was conducted based on Pearson's correlation analysis. A P value less than 0.05 was considered signifcance.

Pyroptosis-Related Classifcation of HNSCC Based on
Consensus Clustering Analysis. To further identify the pyroptosis-related subtype of HNSCC, consensus clustering analysis was performed to justify the classifcation of HNSCC based on the resampling method. A consensus matrix was clustered to determine the optimal k value. In addition, the clinicopathological characteristics (including age, sex, grade, stage of TNM, metastasis, and alcohol use) of HNSCC were included in the heatmap. To further evaluate the efect of pyroptosis-related classifcation on survival time in HNSCC, survival analysis was performed between two pyroptosis-related clusters in HNSCC. Te survival rate was estimated on the basis of the Kaplan-Meier method. Comparison of the survival rate between two clusters was implemented with the log-rank test. A P value less than 0.05 was considered signifcance.

Construction of a Risk Signature of HNSCC in TCGA.
To further identify the prognostic genes of pyroptosis, univariate Cox regression analysis was applied to identify the prognostic genes of pyroptosis-related genes. Subsequently, the prognostic pyroptosis-related genes were enrolled in the LASSO-Cox regression analysis to avoid overftting the model. Te contraction penalty term (λ) was introduced into the model to optimize the regression model. Five-fold cross-validation was used to determine the optimal λ by choosing the logarithm of the minimum mean squared error of the lambda. LASSO-Cox-regularized regression analysis was performed using "glmnet" package in R. Te ROC curve was conducted to evaluate the efcacy of the prognostic risk model of HNSCC with pyroptosisrelated genes.

Evaluation of the Immune Microenvironment in a Prognostic Risk Signature Based on Pyroptosis-Related Genes.
To further evaluate the association between the immune microenvironment and prognostic risk model based on pyroptosis-related genes in HNSCC, cell-type identifcation by estimating relative subsets of RNA transcripts (CIBERSORT) and the ESTIMATE method were applied to analyze the infltration of immune and stroma cells [13]. CIBERSORT (https://cibersort.stanford.edu/) calculates 22 immune cells based on the deconvolution method. ESTI-MATE (https://bioinformatics.mdanderson.org/estimate/) was applied to calculate the scores of stromal and immune cells.

Functional Enrichment of Diferentially Expressed Genes between High Risk and Low Risk Group at HSNCC.
Diferentially expressed genes (DEG) between the high-risk and low-risk groups in HSNCC were calculated using "limma" package with absolute log2(|fold change|) >1 and P value <0.05. To further understand the potential mechanism in DEG between the high-risk and low-risk groups in HNSCC, functional enrichment of Gene Ontology (GO) (https://geneontology.org/) and Kyoto Encyclopedia of Genes and Genomes (KEGG) (https://www.kegg.jp/) was applied to identify the biological function of hub genes in the signifcant module.

Identifcation of Pyroptosis-Related Subtype of HNSCC Based on Consensus Cluster Analysis.
Te pyroptosis-related subtype of HNSCC was clustered into two groups based on consensus clustering analysis (k � 2) (Figures 2(a) and 2(b)). A higher level of 33 pyroptosis-related genes was clustered in cluster 2 ( Figure 2(c)). In addition, a higher survival rate was detected in cluster 1 of HNSCC compared with cluster 2 (P < 0.001) (Figure 2(d)).

Relationship between Prognostic Risk Model and Clinical
Characteristics in HNSCC Based on Six Genes. Te HNSCC patients were divided into 2 groups based on the above risk score (high-risk and low-risk group). Te HNSCC patient in the high-risk group tended to have a shorter survival time (P < 0.001) (Figure 3(e)). In addition, the expression of 6 pyroptosis-related genes (GSDME, IL6, IL1B, NOD1, CASP3, and NLRP2) in the prognostic risk model was elevated in the high-risk group. Compared with the low-risk group, a high-risk group manifested in higher grade with female predominance (P < 0.05). However, no diferences in alcohol use were observed between the high-risk group and the low-risk group (P > 0.05). (Figure 4(a)).

Efects of Pyroptosis-Related Genes on Immune Microenvironment in HNSCC.
Immune cell infltration varied between high and low prognostic risk signature based on pyroptosis-related genes (P < 0.05) (Figure 4(b)). A higher fraction of resting memory CD4+ T cells and macrophages M1 was infltrated in the high-risk group (P < 0.05) (Figure 4(b)). Interestingly, a higher fraction of plasma cells, native B cells, native CD4+ T cells, follicular helpers of T cells, and monocytes were identifed in the low-risk group (P < 0.05). Furthermore, compared with the low-risk group, a higher immune score was detected in the high-risk group of HNSCC (P < 0.05) (Figure 4(c)).

Functional Enrichment of DEGs between the High-Risk and Low-Risk Groups in HSNCC.
Higher expression of (GSDME (DFNA5), PXN, ACTN1, DSE, LAMC2, PROCR, UBASH3B, ITGA3, LIMA1, and MYO1B) was identifed in the high-risk group compared with the low-risk group ( Figure 5(a)). GO and KEGG analyses were performed to evaluate the functional enrichment in DEGs. In KEGG analysis, cytokine-cytokine receptor interaction pathway, phosphatidylinositol 3-kinase (PI3K)-Akt signaling pathway, focal adhesion, and IL-17-signaling pathway were mostly enriched ( Figure 5(b)). In the GO analysis, biological processes (including extracellular structure organization and extracellular matrix organization) were mostly enriched ( Figure 5(c)). Te extracellular matrix, especially for collagen-containing extracellular matrix, was the leading cellular component ( Figure 5(d)). In addition, the molecular function among DEGs varies in receptor regulator activity, receptor ligand activity, and endopeptidase activity ( Figure 5(e)).

Discussion
Although pyroptosis is essential for host defense infection and hazardous signals, its role in tumor activity remains ambiguous. Emerging evidence shows that pyroptosis inhibits tumor growth and migration [7]. However, several lines of evidence support that a possible harmful role of pyroptosis in HNSCC [7]. In consistent with our study, high expression of NLR family pyrin domain family members was detected in the HNSCC tissue. Increased NLRP3 is positively correlated with the growth and metastasis of oral squamous cell carcinoma (OSCC) [14,15]. Knockdown of Nlrp3 expression inhibits the proliferation, migration, and invasion of OSCC by decreasing the cleavage of Caspase-1 and lL-1β release in vitro and in vivo [14]. In addition, high expression of NLRP3 implied a poor prognosis and minor diferentiation in histopathological grading in patients with OSCC. Increased   Log (λ) Partial Likelihood Deviance 9 9 9 9 9 7 7 7 7 7 6 5 5 5 4 1  expression of NLRP3 enhances the 5-FU resistance of oral squamous cell carcinoma (OSCC). Knockout of Nlrp3 and Caspase1 inhibited the tumor growth in mice with 5-FU [15].
NLRP3 promotes the epithelial-mesenchymal transition (EMT) in colon cancer [16]. Similarly, activation of the NLRP3 infammasome accelerates tumor proliferation and migration in lung cancer [17]. In addition, the NLRP3 infammasomes are involved in radiotherapy resistance in glioblastoma [18]. Hence, pyroptosis might play an important role in tumorigenesis in HSNCC.

Te Caspase-3/GSDME Signaling Pathway in Head and
Neck Cancer. Gasdermin family proteins, as the efectors of pyroptosis, play a critical role in lytic-programmed cell death (pyroptosis) [19,20]. GSDME-mediated is involved in the regulation of tumor immune microenvironment and antitumor immunity [21]. Consistent with our result, high levels of GSDME have been detected in many cancers, including lung cancer, melanoma, osteosarcoma, digestive cancers, and head and neck cancer [6,22]. GSDME, cleaved by Caspase 3, releases the pore-forming domain (N-terminal) in the cell membrane to punch holes, which enables the  Grade*  alcohol  alcohol  NO  Yes  Grade*  G1  G2  G3  G4  Stage  I  II  III  IV  M  M0  M1  N  N0  N1  N2  T  T1  T2  T3   secretion of IL-1B and leading to cell swelling, permeabilization, rupture, and death [22,23]. In addition, GSDME, the downstream of Caspase 3, is involved in the switch from apoptotic cell death to secondary pyroptosis. Interestingly, GSDME, in turn, enable us to enhance the intrinsic apoptotic pathway in cancer cells by forming pores in the mitochondria with GSDME-N and liberating proapoptotic factors (including Cyt c and HtrA2) [24]. A recent study has indicated that GSDME ablation impaired the tumor-suppressive activity in head and neck cancer [6].
Consistently, LASSO regression model with prognostic risk in our study demonstrated that higher GSDME and CASP3 expression was inclined to a poor prognosis, suggesting a compensatory increase in Caspase-3-mediated cleavage of GSDME in HNSCC. Not only does GSDME restrains the tumor growth, but augments the function of tumor-infltrating immune cells [21]. Fewer fractions of tumor-infltrating lymphocytes and tumor-associated macrophages were detected in the tumor microenvironment when GSDME was knocked out in mice [21]. Similarly, higher number of macrophages and CD4+ T cells was computed from CIBERSORT analysis in our study. In mouse melanoma, the activation of pyroptosis increases the infltration of tumor-associated T cells and damps dendritic cell infltrates [25]. In addition, immune microenvironment disorder might induce pyroptosis. Recent studies have demonstrated that serine protease granzymes (Gzmes), transferred by perforin and released by NK cells and cytotoxic T lymphocytes, act to cleave GSDMB and GSDME and lead to pyroptosis [7].

Pattern Recognition
Receptors are Involved in the Regulation of Pyroptosis. NOD1 and NLRP2, as the nucleotidebinding oligomerization domain-like receptor (NLPs) members, function as the pattern recognition receptors (PRR) to recognize the pathogen-associated molecular patterns (PAMP) or damage-associated molecular pattern (DAMP) from external stimuli and activate the downstream of caspases through the classical pyroptosis-signaling pathway [26]. Our study demonstrated that NLRP2 was highly expressed in the high-risk group of HNSCC. Similarly, NOD1 and NLRP2 function as risk genes in pyroptosis-related prognostic gene signature related to pyroptosis of lung adenocarcinoma [11]. Recent reports found that NLRP2 elevated the level of profbrotic mediators, but reduced the expression of proinfammatory cytokines and reduced the apoptotic cell rate in proximal tubular epithelial cells [26].
Although the association between pyroptosis and antitumor immunity has been discussed, there is no consensus on the molecular mechanism. PI3K-Akt signaling pathway impacts the tumor growth, migration, survival, and metabolism [27]. Recent research has demonstrated that pyroptosis is induced by ischemia-reperfusion through the PI3K-Akt signaling pathway in neurons [28]. In addition, the activation of the IL-17signaling pathway augments pyroptosis with GSDMD cleave and IL-1 1β and IL-18 release in pneumonia-induced sepsis [29]. Tis evidence supports our results on the role of the IL-17 and PI3K-Akt signaling pathways in high-risk HNSCC.

Conclusion
Our study constructed a prognostic risk signature based on genes and suggested a role of the PI3K-Akt signaling pathway and the IL-17 signaling pathways in poor survival of HNSCC, which emphasized the critical importance of pyroptosis in HNSCC and provides a novel perspective of the HNSCC therapy.

Conflicts of Interest
Te authors declare that they have no conficts of interest.