Potentially functional variants of CHMP4A and PANX1 in the pyroptosis-related pathway predict survival of patients with non-oropharyngeal head and neck squamous cell carcinoma

Abstract

Background:

Pyroptosis has been implicated in the advancement of various cancers. Triggering pyroptosis within tumors amplifies the immune response, thereby fostering an anti-tumor immune environment. Nonetheless, few published studies have evaluated associations between functional variants in the pyroptosis-related genes and clinical outcomes of patients with non-oropharyngeal head and neck squamous cell carcinoma (NON-ORO HNSCC).

Methods:

We conducted an association study of 985 NON-ORO HNSCC patients who were randomly divided into two groups: the discovery group of 492 patients and the replication group of 493 patients. We used Cox proportional hazards regression analysis to examine associations between genetic variants of the pyroptosis-related genes and survival of patients with NON-ORO HNSCC. Bayesian false discovery probability (BFDP) was used for multiple testing correction. Functional annotation was applied to the identified survival-associated genetic variants.

Results:

There are 8,254 single-nucleotide polymorphisms (SNPs) located in 82 pyroptosis-related genes, of which 202 SNPs passed multiple testing correction with BFDP<0.8 in the discovery and six SNPs retained statistically significant in the replication. In subsequent stepwise multivariable Cox regression analysis, two independent SNPs (CHMP4A rs1997996 G>A and PANX1 rs56175344 C>G) remained significant with an adjusted hazards ratios (HR) of 1.31 [95% confidence interval (CI)=1.09–1.57, P=0.004] and 0.65 (95% CI=0.51–0.83, P=0.0005) for overall survival (OS), respectively. Further analysis of the combined genotypes revealed progressively worse OS associated with the number of unfavorable genotypes (P_trend< 0.0001 and 0.021 for OS and disease-specific survival, respectively). Moreover, both PANX1 rs56175344G and CHMP4A rs1997996A alleles were correlated with reduced mRNA expression levels.

Conclusions:

Genetic variants in the pyroptosis pathway genes may predict survival of NON-ORO HNSCC patients, likely by reducing the gene expression, but our findings need to be replicated by larger studies.

Introduction

The head and neck squamous cell carcinoma (HNSCC) is one of the lethal cancers, accounting for over 90% of all head and neck malignancies, and HNSCC encompasses cancers originating in the oral cavity, oropharynx, larynx, and hypopharynx [1–2]. HNSCC stands as the eighth most common cancer globally. In 2020, there were approximately 878,348 new cases of HNSCC and an estimated 444,348 deaths [3]. The primary risk factors for HNSCC include tobacco use and alcohol consumption, with the emerging importance of human papillomavirus (HPV) infection that is mostly confined to oropharyngeal cancer [4]. Current treatment options for HNSCC include surgery, radiotherapy, and systemic therapies including chemotherapy, targeted therapy, and immunotherapy [5]. The majority of HNSCC patients present with a locally advanced disease at the time of the diagnosis. In the cases of advanced HNSCC that are unresectable or have experienced recurrence or metastasis, there remains some potential for curability, necessitating combined-modality treatments, and the targeted therapy or immunotherapy offers some new hope for the treatment of advanced HNSCC [6]. Unfortunately, most patients with recurrence or metastasis HNSCC do not respond to immune checkpoint inhibitors or do not respond well to immune checkpoint inhibitors, and thus there is an urgent need to identify additional survival-related predictive biomarkers to maximize the benefits of immunotherapy [7–9].

Pyroptosis is a form of programmed cell death mediated by gasdermin and regulated by inflammatory caspases, and is a product of continuous cell expansion until the cytomembrane ruptures, resulting in the release of cellular contents that can activate strong inflammatory and immune responses [10–12]. Activation of these inflammasomes can induce the maturation of caspase-1 or caspase-4/5/11 that both cleave gasdermin D to release its N-terminal domain, which can bind to membrane lipids and perforate the cell membrane, leading to changes in cell osmotic pressure and gradually swelling until the final cell membrane ruptures that induce pyroptosis [13–14]. Therefore, pyroptosis could fuel anti-tumor immunity, augment the antitumor immune response, and increase the efficacy of immune checkpoint blockade [12,15].

Given the vital role that pyroptosis proteins play in cellular immunity, our hypothesis centers on associations between genetic variants of the pyroptosis pathway genes and HNSCC survival. To test our hypothesis, we performed an association study by using the genotyping data from the published genome-wide association study (GWAS) of non-oropharyngeal head and neck squamous cell carcinoma (NON-ORO HNSCC) to examine the associations between genetic variants of the pyroptosis-pathway genes and survival of patients with NON-ORO HNSCC.

Material and Methods

Study populations

The present study included 985 NON-ORO HNSCC cases ascertained at the Head and Neck Surgery Clinic at The University of Texas MD Anderson Cancer Center (MDACC), Houston, Texas between December 1996 and July 2011 [16–18]. All cases were newly diagnosed with histologically confirmed and previously untreated HNSCC with exclusion of oropharyngeal cancer, because the latter had a different etiology of HPV infection. Blood samples were collected for genomic DNA extraction and genotyping was performed with Illumina HumanOmniExpress-12v1 BeadChip [16]. This HNSCC GWAS dataset contains about one million genotyped SNPs and are available in dbGaP (accession #: phs001173.v1.p1) [16]. Imputation was performed by using the Haplotype Reference Consortium (HRC) reference panel (Version r1.1 2016), consisting of 64,940 haplotypes for individuals of predominantly European ancestry and Minimac4 and on Michigan server (https://imputationserver.sph.umich.edu/). To have an internal validation, we randomly divided the patients into two groups: the discovery group of 492 patients and the replication group of 493 patient. The characteristics between the discovery dataset (n=492) and the replication dataset (n=493) for comparison are presented in Supplementary Table 1.

Gene and SNP selection

The genes involved in the pyroptosis-related pathway were retrieved by the Molecular Signatures Database (http://software.broadinstitute.org/gsea/msigdb/index.jsp) with the keyword “pyroptosis”, based on previous publications [19–21]. After removal of three genes in the X chromosome, 82 genes remained as the candidates for further analyses (Supplementary Table 2). SNPs located at the gene regions and their ±2kb flanking regions were extracted with the following criteria: an imputation quality score (r²) ≥ 0.8 (Supplementary Figure 1), a minor allelic frequency (MAF) ≥5%, and Hardy–Weinberg equilibrium (HWE) ≥1×10⁻⁵. As a result, a total of 8,254 SNPs (868 genotyped and 7,496 imputed) were extracted for further analyses.

For survival analysis, the time to event was from the date patient was diagnosed as having NON-ORO HNSCC to the date of patient’s death. The covariables to be adjusted in both discovery and replication sets included age at diagnosis, sex, smoking status, alcohol status, and tumor stage. In the single-locus analysis, we first used multivariable Cox proportional hazards regression analysis to assess the association between each of the 8,254 SNP and survival of patients with NON-ORO HNSCC in an additive genetic model, with adjustment for age, sex, smoking status, alcohol status and tumor stage, and the significant principal components (PCs) (Supplementary Table 3) by using the GenABEL package of R software [18]. For controlling multiple comparisons in the discovery, we chose the measure of Bayesian false discovery probability (BFDP), because of the implementation of imputation, to maximize the number of SNPs to be validated, and we used BFDP with a cut-off value of 0.80 for multiple testing correction to lower the probability of discovering potentially false positive results [22]. We assigned a prior probability of 0.10 to detect a hazards ratio (HR) of 3.0 for an association with variant genotypes or minor alleles of the SNPs with P<0.05. We also performed an inverse variance weighted meta-analysis to combine the results of both discovery and replication datasets. In the meta-analysis, Cochran’s Q-test and the heterogeneity statistic (I²) were used to assess the inter-study heterogeneity. If no heterogeneity was observed between the two datasets (P_het>0.10 and I²<50%), a fixed-effects model was implemented; otherwise, a random-effects model was applied. The summarized results of identified independent SNPs were visualized in Manhattan and the regional association plots.

Next, we used the combined genotypes to evaluate cumulative effects of the identified independent SNPs and the Kaplan-Meier curve to estimate overall survival (OS) and disease-specific survival (DSS) probability. We then used receiver operating characteristic (ROC) curve and time-dependent area under the curve (AUC) to illustrate prediction accuracy of the model integrating the effects of demographic, clinical and genetic variables on NON-ORO HNSCC survival [20]. To evaluate the correlations between SNPs and the corresponding mRNA expression levels, we performed expression quantitative trait loci (eQTL) analyses with a linear regression model. The mRNA expression data were obtained from three sources: 373 European individuals included in the 1,000 Genomes Project and 670 whole blood samples included in the genotype-tissue expression (GTEx) project [23,24] as well as data from The Cancer Genome Atlas Program (TCGA). Additional bioinformatics functional prediction for the identified SNPs were performed with SNPinfo [25], RegulomeDB [26] (http://www.regulomedb.org) and HaploReg [27] (http://archive.broadinstitute.org/mammals/haploreg/haploreg.php). Kaplan-Meier survival analysis was also performed to assess associations between the mRNA expression levels and survival probability. All statistical analyses were performed with a statistical significance of P<0.05 by using the R (version 4.2.1) and SAS software (version 9.4; SAS Institute, Cary, NC, USA), unless otherwise indicated.

Results

Associations between SNPs in the pyroptosis-related pathway genes and survival of patients with NON-ORO HNSCC

The flow chart of the present study is shown in Supplementary Figure 2. The basic characteristics of 492 NON-ORO HNSCC patients in the discovery group and 493 NON-ORO HNSCC patients in the replication group did not show any significant difference in the demographic and clinical variables of interest (all P>0.05) (Supplementary Table 1). With multiple testing correction by BFDP≤0.80, we identified 202 SNPs to be significantly associated with NON-ORO HNSCC OS in the discovery dataset, of which six SNPs remained statistically significant in the replication dataset (Table 1). The summarized results of these SNPs in both discovery and replication datasets are visualized in Manhattan plots (Supplementary Figure 3). Subsequently, we performed a combined-analysis of both discovery and replication datasets for these six newly identified SNPs and found that a better survival was associated with SNPs in PANX1, while a poorer survival was associated with SNPs in TSSK4 and CHMP4A, without heterogeneity (I² = 0) between the two groups (Table 1).

Table 1.

Associations of six validated significant SNPs with overall survival of patients with non-oropharyngeal head and neck squamous cell carcinoma in both discovery and replication datasets

SNP	Allele	Gene	Discovery set (n=492)			Replication set (n=493)			Combined-analysis (n=985)
			BFDP	HR (95% CI)	P a	BFDP	HR (95% CI)	P a	HR (95% CI)	P b	P het c	I 2
rs56175344	C>G	PANX1	0.75	0.68 (0.48–0.95)	0.025	0.35	0.60 (0.42–0.84)	0.003	0.64 (0.50–0.81)	2.90E-04	0.613	0
rs11607757	T>G	PANX1	0.79	0.69 (0.49–0.97)	0.032	0.35	0.60 (0.42–0.84)	0.003	0.64 (0.51–0.82)	3.50E-04	0.570	0
rs7117339	C>T	PANX1	0.79	0.69 (0.49–0.97)	0.032	0.35	0.60 (0.42–0.84)	0.003	0.64 (0.51–0.82)	3.50E-04	0.570	0
rs72970792	C>T	PANX1	0.75	0.68 (0.48–0.95)	0.025	0.34	0.58 (0.41–0.83)	0.002	0.64 (0.49–0.80)	1.90E-04	0.522	0
rs2236353	C>T	TSSK4	0.68	1.40 (1.07–1.83)	0.015	0.77	1.35 (1.04–1.75)	0.022	1.37 (1.14–1.66)	8.80E-04	0.849	0
rs1997996	G>A	CHMP4A	0.71	1.40 (1.07–1.84)	0.015	0.80	1.34 (1.04–1.74)	0.025	1.37 (1.13–1.65)	0.001	0.819	0

Abbreviations: SNP, single nucleotide polymorphisms; BFDP, Baysian false discovery probability; CI, confidence interval; HR, hazards ratio.

^aObtained from an additive genetic model with adjusted for age, sex, smoking status, alcohol status, stage, PC7;

^bMeta-analysis in the fix-effects model;

^cP_het: P value for heterogeneity by Cochrane’s Q test;

Independent SNPs associated with NON-ORO HNSCC survival

In the stepwise multivariable Cox regression analysis, only two SNPs remained to be significantly and independently associated with OS (Table 2). The regional association plots of these identified SNPs are shown in Figure 1. In the pooled analysis of both discovery and replication datasets, patients with the PANX1 rs56175344G allele had a decreased risk of death or a better survival [P_trend=0.001 for OS and P_trend=0.025 for DSS], while patients with the CHMP4A rs1997996A allele had an increased risk of death or a worse survival (P_trend=0.003 for OS and P_trend=0.078 for DSS) (Table 3). Specifically, compared with the CC genotype, the PANX1 rs56175344G variant genotypes were associated with a better survival (CG: HR=0.57, 95% CI=0.43–0.74, P<.0001 for OS and 0.55, 0.36–0.84, 0.005 for DDS; GG: 1.14, 0.50–2.57, 0.760 for OS and 1.27, 0.46–3.46, 0.646 for DSS; and CG+GG: 0.59, 0.46–0.76, <0.0001 for OS and 0.59, 0.40–0.88, 0.010 for DSS) (Table 3). In contrast, compared with the GG genotype, CHMP4A rs1997996 A variant genotypes were associated with a worse survival (GA: HR=1.36, 95% CI=1.09–1.70, P=0.006 for OS and 1.30, 0.93–1.82, 0.125 for DDS; AA: 1.89, 1.07–3.31, 0.027 for OS and 1.92, 0.89–4.14, 0.095 for DSS; and GA+AA: HR=1.40, 95% CI=1.13–1.73, P=0.002 for OS and 1.36, 0.99–1.87, 0.060 for DSS) (Table 3).

Table 2.

Two independent SNPs in multivariable Cox proportional hazards regression analysis with adjustment for other covariates for survival of patients with non-oropharyngeal head and neck squamous cell carcinoma in a combined analysis of both discovery and replication datasets

Variables	Category	Frequency	HR (95% CI)	P a
Total		985
Age,_median	≤60	374	1.00
	>60	611	1.04 (1.03–1.05)	<.0001
Sex	Male	665	1.00
	Female	320	1.02 (0.82–1.28)	0.830
Smoking status	Current	224	1.00
	Ever	279	1.29 (0.94–1.76)	0.111
	Never	482	1.36 (1.01–1.82)	0.043
Alcohol status	Current	276	1.00
	Ever	229	1.20 (0.90–1.60)	0.209
	Never	480	1.10 (0.84–1.44)	0.481
Stage	I	230	1.00
	II	185	1.37 (0.93–2.02)	0.108
	III	159	2.20 (1.52–3.20)	<.0001
	IV	411	3.31 (2.42–4.54)	<.0001
CHMP4A rs1997996G>A	GG/GA/AA	733/227/21	1.31 (1.09–1.57)	0.004
PANX1 rs56175344C>G	CC/CG/GG	783/189/13	0.65 (0.51–0.83)	5.0E-4

Abbreviations: SNP, single nucleotide polymorphisms; HR, hazards ratio.

^aStepwise analysis included age, sex, smoking status, alcohol status, stage, top significant principal components, and SNPs.

Figure 1.

regional association plots. Regional association plots for the two independent SNPs in the pyroptosis pathway genes in the 1000 Genome Project. Single nucleotide polymorphisms (SNPs) in the region of 50 kilobases up or downstream of PANX1 rs56175344 C>G (a) and CHMP4A rs1997996 G>A (b).

Table 3.

Associations between two independent SNPs and survival of patients with non-oropharyngeal head and neck squamous cell carcinoma in a combined analysis of both discovery and replication datasets

Genotype	Frequency	OSa			DSSa
		Death (%)	HR (95% CI)	P	Death (%)	HR (95% CI)	P
CHMP4A rs1997996 G>A b
GG	733	309 (42.16)	1.00		134 (18.28)	1.00
GA	227	109 (48.02)	1.36 (1.09–1.70)	0.006	46 (20.26)	1.30 (0.93–1.82)	0.125
AA	21	13 (61.90)	1.89 (1.07–3.31)	0.027	7 (33.33)	1.92 (0.89–4.14)	0.095
Trend test				0.003			0.078
Dominant
GG	733	309 (42.16)	1.00		134 (18.28)	1.00
GA+AA	248	122 (49.19)	1.40 (1.13–1.73)	0.002	53 (21.37)	1.36 (0.99–1.87)	0.060
Reverse
GA+AA	248	122 (49.19)	1.00		53 (21.37)	1.00
GG	733	309 (42.16)	0.71 (0.58–0.88）	0.002	134 (18.28)	0.74 (0.53–1.01）	0.060
PANX1 rs56175344 C>G b
CC	783	360 (45.98)	1.00		158 (20.18)	1.00
CG	189	68 (35.98)	0.57 (0.43–0.74)	<.0001	26 (13.76)	0.55 (0.36–0.84)	0.005
GG	13	6 (46.15)	1.14 (0.50–2.57)	0.760	4 (30.77)	1.27 (0.46–3.46)	0.646
Trend test				0.001			0.025
Dominant
CC	783	360 (45.98)	1.00		158 (20.18)	1.00
CG+GG	202	74 (36.63)	0.59 (0.46–0.76)	<0.0001	30 (14.85)	0.59 (0.40–0.88)	0.010
Reverse
CG+GG	202	74 (36.63)	1.00		30 (14.85)	1.00
CC	783	360 (45.98)	1.70 (1.32–2.19)	<0.0001	158 (20.18)	1.68 (1.14–2.50)	0.010

Abbreviations: SNP, single nucleotide polymorphisms; OS, overall survival; DSS, disease-specific survival; HR, hazards ratio.

^aAdjusted for age, sex, smoking status, alcohol status, stage, and PC7.

Combined effects of the two independent SNPs

We then assessed the combined effect of the two independent SNPs on both OS and DSS of patients with NON-ORO HNSCC. In the combined analysis, we combined the unfavorable genotypes (CHMP4A rs1997996 AA and PANX1 rs56175344 CC into an unfavorable genotype (NUGs) score. Patients with an increased NUG score had a worse OS and disease-specific survival (P_trend<0.0001 and 0.021 for OS and DSS, respectively) (Supplementary Figure 4a–4b). To further facilitate stratification analysis, we used the dichotomized NUG score to divide all the patients into two groups of 0–1 scores and 2 scores. Compared with the group of 0–1 scores, the group of 2 scores had a significantly worse survival (HR=1.50, 95% CI=1.20–1.87, P=0.0004 for OS and 1.57, 1.14–2.18, 0.007 for DSS) (Figure 2a–2b).

Figure 2.

Prediction of survival with combined unfavorable genotypes and eQTL analysis for SNPs in PANX1 and CHMP4A. Kaplan–Meier survival curves for OS in the MDACC dataset for the combined unfavorable genotypes (a); Kaplan–Meier survival curves for DSS in the MDACC dataset and for the combined unfavorable genotypes (b); CHMP4A rs1997996 A allele was associated with lower mRNA expression of CHMP4A in 373 Europeans from the 1000 Genomes Project (c) and whole blood from GTEx project (d); and PANX1 rs56175344G allele was associated with lower mRNA expression of PANX1 in whole blood from GTEx project (e). Abbreviations: NUG, number of unfavorable genotypes; MDACC, MD Anderson Cancer Center.

Stratified analysis for associations between NUGs and NON-ORO HNSCC survival

To evaluate whether the combined effect of the unfavorable genotypes on NON-ORO HNSCC OS and DSS were modified by other covariables, we evaluated possible interaction or effect modification with stratified analysis by age, sex, smoking status, alcohol status, tumor stage in the combined dataset. As a result, there was no obvious difference for the combined effects between the strata of these covariables (P>0.05, Supplementary Table 5), suggesting no interaction between the unfavorable genotypes and other covariables.

The ROC curves and time dependent AUC

We further assessed the predictive value of the two independent SNPs with time-dependent AUC and ROC curves for the three-, five-, and 10-year survival in the combined dataset (with the follow-up time between 0.00 and 184.00 months and the median follow-up time of 45.90 months). Compared with the model including age, sex, smoking status, alcohol status, tumor stage, and PC7, the addition of the independent SNPs did improve prediction performance for three-year overall survival (P=0.029), but not for the five-year and ten-year survival (For OS, P=0.131 and P=0.130, respectively; for DSS, P=0.144 and P=0.336, respectively). However, for both OS and DSS, the predictive performance for model including SNPs was better than model only including demographic and clinical variables alone (Supplementary Figure 5), which suggested that these newly identified independent SNPs contributed to the efficacy of survival prediction for NON-ORO HNSCC patients.

eQTL analysis

Subsequently, we performed the eQTL analysis to assess the correlations between genotypes of the newly identified two survival-predictive SNPs and their corresponding mRNA expression levels. In the RNA-Seq data of lymphoblastoid cell lines from 373 European descendants available in the 1,000 Genomes Project, the CHMP4A rs1997996 A alleles showed a correlation with decreased mRNA expression levels of the gene (P=0.047; Figure 2c), but this correlation was not seen for the PANX1 rs56175344G allele. Then, we performed eQTL by using the data of 670 whole blood samples from the GTEx project and found that the rs1997996 A and rs56175344 G allele remained to be correlated with lower expression levels of CHMP4A and PANX1 (P=0.015 and P<0.001 respectively) (Figure 2d–2e).

In addition, we also performed functional prediction for the identified SNPs [including three SNPs (i.e., rs11607757, rs7117339, and rs72970792) in high LD with rs56175344] using the online tools of SNPinfo, RegulomeDB, and Haploreg, and UCSC to predict their bioinformatics function. The results showed that PANX1 rs56175344 and CHMP4A rs1997996 were located at the potential promoter regions or enhancer regions. The details on the prediction results are summarized in (Supplementary Table 5 and Supplementary Figure 6).

Differential mRNA expression analysis

Finally, we assessed the mRNA expression levels of the two genes identified by the SNPs in 43 pairs of tumor and adjacent normal tissue samples in HNSCC obtained from the TCGA database (http://ualcan.path.uab.edu/), and used Kaplan-Meier curves to describe the association between mRNA expression levels and survival. As shown in Supplementary Figure 7a and 7b, PANX1 expression was up-regulated in tumor tissues compared with the adjacent normal tissues, and patients with increased mRNA expression levels of PANX1 had a poor survival.

Discussion

In the present study, we identified two independent SNPs (i.e., PANX1 rs56175344 C>G and CHMP4A rs1997996 G>A) to be associated with survival of patients with NON-ORO HNSCC in Caucasian populations. In SNP-mRNA expression correlation analysis, we found that low death-risk-associated rs56175344G alleles was associated with lower mRNA expression levels of PANX1 in lymphoblastoid cell lines, while the high death-risk-associated rs1997996A alleles was associated with lower mRNA expression levels of CHMP4A in whole blood tissue. Further functional prediction showed that the two SNPs were located at the potential promoter or enhancer regions.

PANX1 is located on human chromosome 11q14.3 in a 700 kb interval between the genes CRSP6 and MRE11, and the pannexin 1 (PANX1) channel is an important ATP-releasing pathway [28–29]. PANX1 is overexpressed in a variety of cancers, promoting the tumor immune microenvironment via the adenosine triphosphate (ATP) release channels [30–31]. One of the major features of the tumor microenvironment (TME) is an increased extracellular ATP levels that play essential roles in tumor progression via a variety of molecular mechanisms, including cell growth, cell differentiation, energy metabolism, and intercellular signaling [32–33]. The expression of PANX1 was positively correlated with cancer-associated fibroblast (CAF), macrophage, and neutrophils in most cancer types, but it was also negatively correlated with T cell NK. Extracellular ATP, an important biochemical component of TME, is also implicated in immune cell recruitment and activation, leading to tumor proliferation and metastasis [34–35]. Therefore, PANX1 may enhance tumor proliferation, invasion, and metastasis by regulating extracellular ATP concentration in the TME. For example, when P2 receptors are activated, extracellular ATP increases tumor cell survival and proliferation [36–38].

Our findings suggest that PANX1 may function as a potential oncogene in NON-ORO HNSCC, because PANX1 mRNA expression levels were lower in normal tissues than in tumor tissues, and a lower level of PANX1 expression was associated with a better survival. Also, PANX1 was overexpressed in most cancers, compared with that in normal tissues. The high expression of PANX1 was associated with a poor prognosis in multiple tumors, possibly by promoting cell proliferation and tumorigenic properties, such as in pancreatic adenocarcinoma, kidney renal papillary cell carcinoma, lung adenocarcinoma (LUAD), uterine corpus endometrial carcinoma [39], melanoma cells testicular cancer, and breast cancer [40–41]. Therefore, PANX1 is likely a cancer-promoting gene.

In contrast, our findings suggest that CHMP4A is more likely to be a potential suppresser gene in NON-ORO HNSCC, because we found that lower levels of CHMP4A mRNA expression were associated with a worse survival in patients with NON-ORO HNSCC, which is consistent with the findings in TCGA that CHMP4A mRNA expression was higher in patients with head and neck cancer, and such a high expression of CHMP4A was associated with a poor OS. CHMP4A is one of the components of mammalian ESCRT-III endosomal sorting complex and is involved in the degradation of membrane protein and the resistance to cell death [42–44], which promotes mitotic cytokinetic, midbody abscission and multivesicular body organization [45–47]. Also, CHMP4A acts as the modulator to increase the expression levels of hypoxia-inducible factor-1 α protein (HIF-1α), promoting tumor progression by a hypoxia-associated pro-tumor mechanism [48]. These molecular findings are inconsistent with our results of the risk effect from the CHMP4A rs1997996 A allele being associated with a lower mRNA expression and a higher risk of death in patients with NON-ORO HNSCC.

CHMP4A, located on chromosome 14q11.2 and encompassing six exons, is the member of the chromatin-modifying protein/charged multivesicular body protein (CHMP) family [49]. The expression and function of CHMP4A have been demonstrated in various solid and hematological tumors, including low-grade gliomas, hepatocellular carcinoma and cancers of the breasts, colorectum, prostate, and ovaries [50–55]. For example, high expression levels of CHMP4A were associated with a better clinical outcome of patients with breast cancer [56]; downregulation of CHMP4A was associated with a worse OS in low-grade gliomas (LGGs) [57]; the levels of CHMP4A were markedly increased in men with high Gleason score (GS) prostate cancer and in patients with advanced high-grade serious ovarian cancer (HGSOC); and the high expression levels of CHMP4A were associated with a poor prognosis in the hepatocellular carcinoma [58].

These results showed that the identified survival-associated loci in the pyroptosis-related genes were likely to have an effect on prognosis of NON-ORO HNSCC patients. Overall, our findings suggest that functional genetic variants in the pyroptosis-related pathway genes may have played roles in NON-ORO HNSCC progression, possibly through a mechanism of modulating the expression of the genes, such as PANX1 and CHMP4A, which provide some new scientific insights into the management and treatment of NON-ORO HNSCC patients, and these loci could be used as potential biomarkers of prognosis for clinically monitoring the outcomes of the treatment, if validated in additional investigations.

There are several limitations in the present study. Firstly, although several genetic variants backed up with in silico functional evidence in the pyroptosis-related genes were found to be associated with NON-ORO HNSCC survival, the exact molecular mechanisms of these SNPs underlying the observed associations still remain unclear. Secondly, the MDACC dataset consists of Caucasian populations; therefore, our results may not be generalizable to other ethnic populations. Thirdly, though some clinical factors were available in the analysis of the MDACC dataset, there are still some other information, such as performance status and nutritional status, as well as specific treatments, such as immunotherapy, that was not available for further adjustments and stratification analysis. Therefore, additional validation by studies with larger sample sizes are needed to confirm our findings.

Abbreviations:

MDACC

MD Anderson Cancer Center

SNPs

single nucleotide polymorphisms

GWAS

Genome-Wide Association Study

OS

overall survival

DSS

disease-specific survival

LD

linkage disequilibrium

BFDP

Bayesian false discovery probability

eQTL

expression quantitative trait loci

TCGA

the Cancer Genome Atlas

ROC

receiver operating characteristic

CHMP4A

Charged Multivesicular Body Protein 4A

PANX1

Pannexin 1

EAF

effect allele frequency

HR

hazards ratio

CI

confidence interval

AUC

area under the receiver operating characteristic curve

ROC

receiver operating characteristic curve

NON-ORO HNSCC

non-oropharyngeal head and neck squamous cell carcinoma