A Novel Pyroptosis-Associated Gene Signature to Predict Prognosis in Patients with Colorectal Cancer

Background Pyroptosis is a form of cell death characterized by cell swelling and plasma membrane bubbling in association with inflammatory and immune responses. To date, the association between pyroptosis and colorectal cancer remains unclear. We aimed to establish a novel pyroptosis-associated model for the prognosis of colorectal cancer. Methods Pyroptosis-related genes were extracted using Gene Set Enrichment Analysis. A least absolute shrinkage and selection operator regression model was constructed to identify a pyroptosis-related gene signature using the Cancer Genome Atlas and Gene Expression Omnibus databases. Then, Kyoto Encyclopedia of Genes and Genomes and Gene Ontology and GSEA were performed to better understand the potential mechanisms and the functional pathways associated with pyroptosis involved in colorectal cancer. The relationship between the pyroptosis-related signature and immune infiltration was investigated using Cell-Type Identification by Estimating Relative Subsets of RNA Transcripts and MCPcounter. Results A 12 pyroptosis-related gene signature was identified. Then, patients were classified into high- and low-risk groups. Kaplan–Meier and receiver operating characteristic analyses confirmed that the high-risk groups showed worse overall survival, progression-free survival, or relapse-free survival probability. Functional enrichment analysis showed that pyroptosis was associated with extracellular matrix-related pathways. Furthermore, the pyroptosis risk score was associated with immune infiltration. The low-risk group exhibited a higher percentage of plasma cells, CD4 T cells, activated dendritic cells, and activated mast cells. M2 macrophages and M0 macrophages were positively related to the risk score. Conclusion Our research yielded a novel pyroptosis-related prognostic signature for colorectal cancer that was related to immune cell infiltration, and it provided an immunological perspective for developing personalized therapies.


Introduction
Colorectal cancer (CRC) is the third most common cancer and the second major cause of cancer death worldwide in 2020 [1]. e cornerstones of therapy for CRC include surgery for primary and metastatic disease, neoadjuvant radiotherapy for rectal cancer, and auxiliary chemotherapy for phase III/IV and high-risk phase II colon cancer [2,3].
Nevertheless, these new treatment options have limited influence on cure rates and long-term survival. During the past two decades, the interconnections between patient prognosis and therapeutic response as well as the molecular mechanisms underlying the development of CRC have become increasingly apparent. In this regard, developing novel prognostic gene signatures for CRC is urgently required.
Programmed cell death includes apoptosis, necroptosis, ferroptosis, and pyroptosis [4]. Among the intrinsic forms of cell death, pyroptosis has recently received increasing attention. Pyroptosis is initiated by the gasdermin (GSDM) family and characterized by cell swelling and plasma membrane bubbling accompanied by inflammatory and immune responses [5]. Except for PJVK, the members of the GSDM superfamily (GSDMA, GSDMB, GSDMC, GSDMD, GSDME, and PJVK) harbour an N-terminal pore-generating domain and a C-terminal repressor domain as well as a linker connecting them [6]. e association between pyroptosis and cancer is complicated, and various tissues and genetic backgrounds exhibit varied effects of pyroptosis on cancer. On the one hand, the occurrence and growth of tumours can be inhibited by pyroptosis [7]; on the other hand, as a type of proinflammatory death, pyroptosis can generate a proper microenvironment for tumour cell development and thus drive tumour development [8]. erefore, it is essential to identify the effect of pyroptosis on the prognosis of CRC.
Currently, some studies have reported prognostic gene signatures associated with pyroptosis in lung adenocarcinoma [9], ovarian cancer [10], gastric cancer [11], and CRC [12][13][14][15][16] based on data obtained from public databases. Our study provides a new method to identify a pyroptosis-associated prognostic gene signature for CRC. e above articles excluded genes with no significance based on the Cox regression model, which may ignore valuable information. In the current study, a prognostic model based on least absolute shrinkage and selection operator (LASSO) analysis was built using pyroptosis-associated genes without Cox regression model screening. We also determined the association of pyroptosis with the tumour immune microenvironment in CRC. Our findings may offer extra evidence on prognostic biomarkers and therapeutic targets for CRC.
We present the following article in accordance with the TRIPOD reporting checklist.

Pyroptosis-Related Signature in CRC.
All of the pyroptosis-related genes extracted from GSEA were included in the LASSO regression model in the TCGA database. LASSO was performed to identify the key signatures and the related parameters using the glmnet R package.

Kyoto Encyclopedia of Genes and Genomes (KEGG), Gene
Ontology (GO) Analysis, and GSEA. Sangerbox tools, a free online platform for information analysis, were employed to perform KEGG, GO analysis, and GSEA. Briefly, samples were categorized into high-and low-risk groups, and GSEA was performed to identify the possible mechanisms associated with pyroptosis. Here, |NES| >1, NOM p value <0.05, and FDR q-value <0.25 were used as the thresholds of significance. e R software package clusterProfiler was applied for GO and KEGG functional enrichment analyses to identify pathway enrichment between the low-risk and high-risk groups.

Estimation of Tumour Microenvironment. Cell-Type
Identification by Estimating Relative Subsets of RNA Transcripts (CIBERSORT) and MCPcounter was employed to forecast immune cell infiltration. e relative proportion of 22 immune cells was estimated with CIBERSORT, and MCPcounter scores evaluated the composition of 8 immune cells based on another algorithm. en, correlation analysis was performed using Pearson's test.

Statistical Analysis.
e survival time difference was measured by Kaplan-Meier analysis.
e "time ROC" package in R was adopted to perform receiver operator characteristic (ROC) analyses. A p value <0.05 was considered statistically significant, and all p values were two-tailed.

Identification of a Pyroptosis-Related Prognostic
Signature. To date, 27 genes associated with pyroptosis have been identified from GSEA. To understand the function of the overall pyroptosis-related genes, univariate Cox analysis was not performed. None of the pyroptosis-related genes were excluded from the following analysis. LASSO regression analysis was performed using 27 pyroptosis-related genes, and 12 genes with the best capacity were identified using the LASSO regression model (Figures 1(a) and 1(b)). e description and parameters of the 12 pyroptosis-related genes are shown in Table 1.
e association network containing the 12 pyroptosisassociated genes and risk scores is displayed in Figure 2(a) (red: positive associations; blue: negative associations). For further exploration of the mutual effects of these pyroptosisassociated genes, PPIs between the 12 pyroptosis-related genes were analysed by using the STRING platform ( Figure 2(b)).
Univariate and multivariate Cox analyses were performed using data from the TCGA cohort (Table 2). Univariate analysis indicated that risk score, age, and pathologic stage were related to the survival time of CRC patients. Multivariate analysis identified the risk score, age, and pathologic T stage as independent factors associated with CRC prognosis.

Landscape of Genetic Variation of Pyroptosis-Associated
Genes in CRC. At the genetic level, pyroptosis-related genes were altered in 51 (23%) of the 220 patients/samples (Figure 3(a)). Deep deletion was the most common variant classification, and C > T is ranked as the top mutation. Of these, TP63, IRF2, and CASP3 showed the highest frequency of mutations, and only 1 sample (approximately 0.5%) harboured IL1A missense mutations. Figure 3(a) also presents the mutation count, survival status, survival time, and expression heatmaps of these genes.
In the TCGA cohort, IL1A, HMGB1, and CHMP4C were enriched in CRC samples, whereas the remaining pyroptosisrelated genes all showed decreased expression in CRC samples (Figure 3(b)). For more effective verification of our findings, the mRNA levels of pyroptosis-associated genes among 32 pairs of tumour tissues and normal nearby tissue samples were tested. Similarly, these pyroptosis-related genes were downregulated in tumours (Figure 3(c)).

e Prognostic Capacity of the Pyroptosis-Associated Gene
Signature. CRC patients were separated into high-and lowrisk groups based on their median risk score. Figures 4(a)-4(c) show the risk score allocation, survival status, and gene profiles of the 12 genes in the training (TCGA) and verification (GSE39582 and GSE87211) cohorts. As the risk score increased, the patients' risk of death increased, and the survival time declined. e heatmaps showed that CHMP2A, CHMP3, GSDME, HMGB1, IRF2, and TP63 were overexpressed, whereas downregulation of CHMP2B, CASP3, CHMP4C, CHMP6, CHMP7, and IL1A was observed in high-risk cases.
Next, the value of the risk score was further determined by forecasting the prognosis of patients. Patients were classified into high-and low-risk score groups based on the best cut-off value. According to the Kaplan-Meier curve, CRC patients with a high-risk score showed a worse total survival, progression-free survival, or relapse-free survival compared with those with a low-risk score (Figures 5(a)-5(c)). e number of patients at risk is displayed in the lower portion of each graph.
Evidence-Based Complementary and Alternative Medicine e sensitivity and specificity of the prognostic model were assessed with time-dependent ROC analysis, and area under curve (AUC) in the TCGA, GSE39582, and GSE87211 cohorts predicting overall survival reached 0.78, 0.55, and 0.7 for 5-year survival (Figures 6(a)-6(c)). AUC values predicting relapse-free survival or progression-free survival were also significant at >0.5 (Figures 6(a)-6(c)).

Functional Enrichment Analysis of Pyroptosis-Associated
Genes. To clarify the function of the pyroptosis-related genes between the subgroups classified by the risk model, we identified 512 differentially expressed genes (DEGs) with an absolute value of log10 FDR <1 and |log2-fold change| <1 between the low-and high-risk groups in the TCGA cohort.
e KEGG pathway analysis and GO enrichment analysis were then conducted based on these 512 DEGs.
We found that these DEGs were mostly involved in the ECM-receptor interaction, vascular smooth muscle contraction, TGF-beta signalling pathway, dilated cardiomyopathy (DCM), and axon guidance based on KEGG pathway analysis (Figure 7(a)). Moreover, GO analysis suggested that these DEGs were mainly correlated with extracellular matrix, extracellular structure organization, and extracellular matrix structural constituent (Figures 7 Subsequently, GSEA was performed between the highand low-risk groups. On the basis of the outcomes, the highrisk group expressed all enriched gene sets. ese gene sets were involved in mechanisms associated with aminoacyl tRNA biosynthesis, base excision repair, citrate cycle TCA cycle, DNA replication, proteasome, homologous recombination, terpenoid backbone biosynthesis, and nucleotide excision repair (Figure 8).

Identification of the Association between Pyroptosis-Related Genes and the Tumour Immune Microenvironment.
To better assess how pyroptosis-related genes interact with the immune microenvironment, the CIBERSORT algorithm   was adopted, and integrated comparisons with the risk scores were conducted. Figures 9(a) and 9(b) show the relative content allocation of tumour-infiltrating immune cells in the TCGA cohort and the association between tumour-infiltrating immune cells.
Next, whether the risk score would have guiding value for clinical treatment, especially immunotherapy, was investigated. e infiltration of tumor microenvironment (TME) cells was analysed in patients with various risk scores (Figure 10(a)). Patients in the low-risk group showed a higher percentage of antitumoural immune cells, such as plasma cells, CD4 T cells, activated dendritic cells (DC), and activated mast cells. Higher scores were closely related to M2 macrophages and M0 macrophages. Similarly, further analysis showed that the risk score was positively related to M2 macrophages and M0 macrophages but negatively related to plasma cells, CD4 T cells, activated dendritic cells, and activated mast cells (Figures 10(b)-10(h)). Furthermore, Kaplan-Meier analysis indicated that plasma cells conferred a significant survival advantage (Figure 10(i)). e MCP-counter score showed the association between the risk score and the monocytic lineage, fibroblasts, B lineage, and endothelial cells in the TCGA cohort (Figures 11(a)-11(d)). Conversely, Kaplan-Meier analysis (Figures 11(e) and 11(f )) and univariate Cox regression (Figure 11(g)) highlighted that CRC patients benefited from myeloid dendritic cells, whereas fibroblasts impaired CRC patient prognosis.

Discussion
is research established a pyroptosis-related prognostic signature for CRC patients. TCGA cohort data were assessed using the LASSO regression model, and a 12-gene signature associated with the outcome of CRC was produced. ROC curve and Kaplan-Meier analyses among patients with regard to high-and low-risk scores were validated in the GEO cohort. Genetic variation analysis showed that TP63, IRF2, and CASP3 presented the highest frequency of mutations. In the following functional analysis, KEGG, GO, and GSEA results indicated that the high-risk groups were  Evidence-Based Complementary and Alternative Medicine related to extracellular matrix, TGF-beta signalling pathway, and DNA replication. Immune cell infiltration data demonstrated that the low-risk group exhibited more activated immune cell infiltration, which may result in a better response to immunotherapy. is work will promote future pyroptosis studies in CRC.
Consistent with our observations, other researchers also identified pyroptosis-associated models and characterized tumour microenvironment infiltration in CRC. Generally, genes with no significance were excluded using the Cox regression model before lasso analysis. However, the Cox regression model failed to take advantage of genes that had important biological functions without statistical significance. Many valuable pieces of information, such as biological characteristics, have been ignored. In our study, the threshold of Cox analysis was abolished. Based on the overall trend of the dataset, our study provided a more reasonable pyroptosis-related prognosis signature.
We identified a 12 pyroptosis-associated gene signature for the prognosis of colorectal cancer, including CAS P3, CHMP2A, CHMP2B, CHMP3, CHMP4C, CHMP6, CHMP7, GSDME, HMGB1, IL1A, IRF2, and TP63. Activated caspase 3 could specifically cleave GSDME [17], which generated pores in the plasma membrane and induced pyroptosis. In certain cells, the switch from apoptosis to pyroptosis could be induced by knocking out GSDME [17]. GSDME expression was silenced in colorectal, gastric, and breast cancer [18], indicating that GSDME may be a tumour suppressor gene. Furthermore, GSDME-induced pyroptosis was an underlying mechanism involved in the resistance to the toxicity of chemotherapy drugs [17]. CHMP2A, CHMP2B, CHMP3, CHMP4C, CHMP6, and CHMP7 are endosomal sorting complexes necessary for transport (ESCRT)-III subunit [19]. ESCRT-III plays a key role in the degradation of transmembrane proteins in lysosomes, midbody abscission during cytokinesis, and retroviral budding. In the pyroptosis process, ESCRT-III is required for damaged plasma membrane repair [20]. Given that EGFR is a cell membrane protein, inactivation of the ESCRT-III machinery impairs EGFR degradation [21], which may induce tumorigenesis. Furthermore, CHMP4C has been frequently overexpressed in human cancer tissue [22]. As a response to damage-associated molecular patterns (DAMPs), the nuclear-associated protein HMGB1 and cytokine IL1A are released following exposure to inflammatory stimuli [23]. HMGB1 overexpression was confirmed in melanoma, colon cancer, prostate cancer, pancreatic cancer, and breast cancer [24]. During pyroptosis, caspase-1, which is activated by inflammasomes, promotes the secretion of HMGB1 and IL1A [25,26]. Subsequently, HMGB1 and IL1A promoted the expression of vascular endothelial growth and tumorigenesis. IRF-2 is a member of the interferon regulatory factor (IRF) family, which was originally identified based on its effects on innate and adaptive immunity. IRF-2 targeted the interferon-stimulated response element and was involved in the regulation of the interferon system [27]. IRF2, which is overexpressed in pancreatic cancer [28] and oesophageal squamous cell carcinoma [29], acts as an oncogene. IRF2 sequentially activates caspase-4 and GSDMD to induce pyroptosis [30]. TP63 is a member of the p53 family and exhibits diverse biological functions, such as cellular proliferation, differentiation, stem cell maintenance, cell death and survival, DNA damage response, and metabolism. Furthermore, multiple isoforms of TP63 played opposite roles in different human tumours [31,32]. Numerous studies have confirmed that TP63 can induce apoptosis and DNA damage [33]. In CRC, knockdown of TP63 could cleave GSDME to induce pyroptosis [34].        More recently, immunotherapy and the tumour microenvironment have attracted increasing attention. Immune cells, such as T lymphocytes and B lymphocytes, infiltrated the tumour microenvironment and influenced tumour progression. In our study, pyroptosis risk scores were negatively related to plasma cells, CD4 T cells, activated dendritic cells, and activated mast cells and positively related to M2 macrophages and M0 macrophages. Generally, M2 macrophages secrete cytokines, such as IL-10 and TGF-β, which suppress the immune response and promote tumorigenesis [35]. Conversely, high proportions of infiltrating CD4 T cells and dendritic cells, which are optimally equipped antigen-presenting cells (APCs), were associated with a favourable prognosis. Upon activation by tumour antigens, DCs can internalize, process, and later show processed epitopes to T cells and cause cytotoxic T lymphocyte (CTL) immune responses [36]. CD4 T cells were mainly helper T cells ( cells) in our study. cells play a role in antitumour immunity by assisting CD8 effector T cells and functioning cytotoxic T cells [37,38].

Conclusion
In our study, a new, reasonable pyroptosis-associated prognostic signature for CRC was observed and an extensive regulatory mechanism by which they affect the prognosis was revealed. Genetic variation analysis presented the high mutation frequency of pyroptosis-related genes. Moreover, we identified that immune cell infiltration in the low-risk group contributes to CRC patient prognosis. Our findings highlight the crucial implications of pyroptosis-related genes, which may be developed as a precise indicator for individualized clinical prognostication.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.

Disclosure
Juji Dai and Shuangdong Chen are co-first authors.

Conflicts of Interest
e authors declare that they have no conflicts of interest.