Genetic alteration landscape of pyroptosis-related genes in cancer
In this study, we identified thirty-three genes that play critical roles in regulating pyroptosis by reviewing previous studies (Fig. 1). The lists of 33 pyroptosis-associated genes are provided in Additional file 1: Table S2. To determine the patterns of dysregulation of pyroptosis genes in cancer, we examined genomic data, including genetic variation, somatic copy number alteration, messenger ribose nucleic acid (mRNA) expression, and DNA methylation data of tumour and normal tissues from 32 cancer types. The overall alteration levels of pyroptosis genes ranged from 0.7% to 7.7%. Although the DNA alteration level was relatively low, 50% of tumours had at least one type of alteration (Fig. 2A). Mutations, amplifications and homozygous deletions accounted for the majority of pyroptosis gene alterations. Among these 33 genes, GSDMC showed the highest alteration frequency (7.7%), with amplification being the most frequent (5.5%).
The alteration rates of GSDMD and NLRP3 were similar (6.3% and 6.0%). The alteration rates of other genes ranged from 01 to 3%, and CASP6 had the lowest alteration level (0.7%). We further analysed the alteration patterns among all cancer types and found differences in alterations among different cancer types. UCEC showed alterations in all pyroptosis genes, while the alteration levels of all genes related to kidney chromophobe (KICH) and mesothelioma (MESO) were very low (Fig. 2A). Individual pyroptosis genes also showed differential amplification or deletion patterns among cancer types. Almost all cancers exhibited a high heterozygous amplification frequency, while the heterozygous amplification frequency was relatively low in thymoma (THYM), acute myeloid leukaemia (LAML) and thyroid carcinoma (THCA). The same trend was observed regarding heterozygous deletions in the three types of cancers. However, GSDMC and GSDMD had high homozygous amplification frequencies in ovarian serous cystadenocarcinoma (OV), oesophageal carcinoma (ESCA), uterine carcinosarcoma (UCS), breast invasive carcinoma (BRCA), stomach adenocarcinoma (STAD), liver hepatocellular carcinoma (LIHC), and uveal melanoma (UVM). The high homozygous deletion rates were scattered in most cancers (Additional file 2: Fig. S2A, B).
The gene mutations mainly consisted of missense mutations, nonstop mutations and multihit mutations. The mutation levels of these genes ranged from 0 to 3%. NLRP3 showed the highest mutation level (3%), and NLRP7, NLRP1, CASP8 and NLRC4 showed the same mutation frequencies (2%). The remaining genes had mutation frequencies of only 1% (Additional file 2: Fig. S2C). For cancer types, the mutation frequencies of pyroptosis genes are relatively low in all cancers and relatively high in skin cutaneous melanoma (SKCM). Furthermore, NLRP3, which encodes a pyrin-like protein containing a pyrin domain, functions as an upstream activator of nuclear factor (NF-kappa B) signalling and regulates inflammation, the immune response, and apoptosis, showed relatively high mutation frequencies in several cancers, including uterine corpus endometrial carcinoma (UCEC), SKCM, colon adenocarcinoma (COAD), LUAD, and lung squamous cell carcinoma (LUSC) (Additional file 2: Fig. S3). NLRP3 mutation was significantly related to the PFS, OS and DSS of UCEC patients. Other genes also showed significant associations with survival prognosis in multiple cancers (Additional file 1: Table S3).
Aberrant expression of pyroptosis-related genes among cancers
We performed differential expression analysis of pyroptosis genes among cancers except MESO and UVM without normal tissues. Our results indicated that all pyroptosis genes were differentially expressed in at least one type of cancer. Some pyroptosis genes exhibited consistent expression patterns in multiple cancers. GSDMC, NLRP7, CASP5, PYCARD, IL18, IL1B and GSDMA were significantly upregulated in 22, 18, 18, 18, 18, 16 and 17 types of cancers, respectively (Additional file 2: Fig. S4). Protein kinase CAMP-activated catalytic subunit alpha (PRKACA), elastase neutrophil expressed (ELANE), NLRP1, pejvakin (PJVK), and CASP9 were significantly downregulated in 25, 24, 22, 25, and 23 types of cancers, respectively. Several pyroptosis genes showed cancer type-specific patterns. ELANE was significantly downregulated in almost all cancers but obviously upregulated in GBM (log2FC = 2.07) and LAML (FC = 12.15). CASP8 mutations are associated with increased risks of cancer, and low expression of CASP8 is closely associated with poor prognosis in patients with cancer; however, CASP8 expression was significantly upregulated in glioblastoma multiforme (GBM) and pancreatic adenocarcinoma (PAAD) [18]. IL6 seems to promote the development of cancer [19] and is significantly upregulated in lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), GBM, PAAD, testicular germ cell tumours (TGCTs), and THYM but downregulated in adrenocortical carcinoma (ACC), BLCA, BRCA, kidney chromophobe (KICH), kidney renal clear cell carcinoma (KIRP), LAML, and LUAD. We also found that IL6 exhibited the opposite expression pattern in some subtypes of brain and kidney cancers. These results indicated that pyroptosis genes may differentially function in different cancers.
Copy number variation (CNV) is known to play a role in tumorigenesis development, and we further evaluated the association between CNV and pyroptosis gene expression. Pearson correlation analysis indicated that most pyroptosis genes were correlated with CNV in most cancers (Fig. 2B). For example, TIRAP, which is involved in the Toll-like receptor (TLR4) signalling immune system pathway, was significantly correlated with CNV in 29 types of cancers. PRKACA, which participates in cellular processes, including differentiation, proliferation, and apoptosis, was significantly associated with CNV in 27 cancers. These results showed that abnormal copy numbers of pyroptosis genes are common in most cancers and affect gene expression levels.
We also assessed the methylation levels of pyroptosis genes in tumour and normal tissues. We found that pyroptosis genes exhibited complex methylation patterns in the 14 types of cancers (Fig. 2C), and only ELANE showed hypermethylation in 12 types of cancers. We observed that NLRP7 (n = 8), AIM2 (n = 10), CASP8 (n = 6), GSDMA (n = 5), GSDMB (n = 8), and GSDMC (n = 12) mainly showed hypomethylation in most cancers, and PLCG1 (n = 11), NLRP6 (n = 13), ELANE (n = 11), CASP6 (n = 5), NLRC4 (n = 12), and PYCARD (n = 8) showed hypermethylation in most cancers. The methylation levels of pyroptosis genes differed significantly in 14 cancers (Additional file 1: Table S4). Spearman correlation analysis indicated a negative relationship between gene expression and overall methylation levels (Fig. 2D and Additional file 2: Table S5). These results suggested that DNA methylation regulates the expression of pyroptosis genes in cancers.
Estimated modelling of pyroptosis levels and their association with prognosis among cancers
To further explore the role of pyroptosis in the development of tumours and understand pyroptosis-related biological processes, we built an estimated model of pyroptosis levels in all cancers based on enrichment scores by single-sample GSEA. We observed that LAML had the highest pyroptosis level, while pheochromocytoma and PGL had the lowest pyroptosis levels (Fig. 3A). We further compared the pyroptosis levels between tumours and normal tissues. We observed that the pyroptosis levels were significantly increased in oesophageal carcinoma (ESCA), head and neck squamous cell carcinoma (HNSC), KIRC, KIRP, and THCA (Fig. 3B–F), while the pyroptosis levels were significantly decreased in LIHC, LUSC and PRAD (Fig. 3G–I).
We further performed univariate Cox regression to evaluate the associations between pyroptosis levels and four survival outcomes, OS, DSS, DFI and PFI. The pyroptosis score was significantly associated with OS in eight types of cancer (Fig. 4A), KIRC (P < 0.001), SKCM (P < 0.001), LGG (P < 0.001), PAAD (P = 0.002), UVM (P = 0.006), BLCA (P = 0.029), THYM (P = 0.043), and SARC (P = 0.048), while the pyroptosis score was significantly associated with DSS in seven types of cancers (Fig. 4B), KIRC (P < 0.001), SKCM (P < 0.001), LGG (P < 0.001), PAAD (P = 0.011), UVM (P = 0.039), and UCEC (P = 0.044). However, the pyroptosis score was significantly associated with DFI in only one cancer (Fig. 4C), COAD (P = 0.030). A significant correlation between the pyroptosis score and PFI was also observed in seven cancers (Fig. 4D), KIRC (P < 0.001), LGG (P < 0.001), GBM (P < 0.001), THYM (P = 0.002), PAAD (P = 0.002), SKCM (P = 0.017), and BLCA (P = 0.029). Among these cancers, the elevated pyroptosis score was associated with poor survival outcomes for patients with twelve types of cancer (Additional file 2: Fig. S5A), ESCA, GBM, HNSC, KIRC, LAML, LGG, LUSC, PAAD, THYM, UCES, UCS, and UVM, while the elevated pyroptosis scores favoured survival for patients with eight types of cancer (Additional file 2: Fig. S5B), BRCA, KICH, MESO, SARC, SKCM, STAD, THCA and BLCA. The pyroptosis score had different clinical effects in some cancers. For example, pyroptosis was shown to be unfavourable for KIRC but favourable for KICH.
We also evaluated the associations between pyroptosis genes and tumour risks. Overall, the NLRP3, PJVK, TIRAP, IL18, NLRP1 and NLRP6 genes were thought to protect against cancer (Additional file 2: Fig. S6A, B), while the rest of the pyroptosis genes seemed to be correlated with cancer risk. Some genes showed different risk patterns. For example, IL6 is a risk gene in multiple cancers but plays a protective in only sarcoma (SARC). A similar result was also observed for PRKACA in several cancers. In contrast, TIRAP is a protective gene in KIRC, rectum adenocarcinoma (READ), and STAD but a risk gene in only BRCA. These results indicated that pyroptosis genes may play different roles in tumours.
Pyroptosis-related pathways and immune signatures among cancers
To evaluate the associations between pyroptosis levels and pathways, we calculated Spearman correlation coefficients between pyroptosis scores and other genes and pathways using GSEA in all cancers. As shown in Fig. 5, IL-6/JAK/STAT3 signalling, allograft rejection, inflammatory response, IL2/STAT5 signalling, tumour necrosis factor (TNF-A) signalling via nuclear factor kappa B (NF-kB), apoptosis KRAS signalling and the P53 pathway were enriched in tumours with high pyroptosis levels, indicating that pyroptosis was positively associated with these pathways. Spermatogenesis (29 cancers), pancreatic beta cells (20 cancers), oxidative phosphorylation (24 cancers), hedgehog signalling (22 cancers), Wnt-beta catenin signalling (22 cancers), peroxisomes (20 cancers), and the G2/M checkpoint (22 cancers) were enriched in most tumours with low pyroptosis levels, which indicated that pyroptosis was negatively associated with these pathways. Other common pathways, such as the reactive oxygen species pathway, hypoxia, epithelial-mesenchymal transition (EMT), PI3K/Akt, and some metabolism-related pathways, were also enriched in multiple cancers. These pathways showed a positive association with the pyroptosis level. We further performed GSEA of six cancers (significantly associated with the pyroptosis level) based on significant survival analysis (Additional file 2: Fig. S7A–F). We observed that multiple immune-related pathways, such as the innate immune system, cytokine signalling in the immune system, and the adaptive immune system, were enriched in BRCA, KIRC, LUSC and PAAD.
Considering the important function of the immune response process in tumorigenesis, we explored the correlation of pyroptosis with the immune microenvironment in cancers. The results showed that the immune score and stromal score were positively associated with the pyroptosis score, while the pyroptosis score was negatively associated with tumour purity (Fig. 6A). Furthermore, we investigated the associations between the pyroptosis score and immune-related pathways, matrix/metastasis-related pathways, and DNA damage repair pathways. The results showed that the pyroptosis score was positively associated with the immune checkpoint, CD_8_T effector, and antigen processing machinery pathways in almost all cancers. DNA damage repair was negatively associated with the pyroptosis score in most types of cancers, especially HNSC, TGCT, ESCA, SARC, LAML, CEUS, GBM, and PCPG. The pyroptosis score was positively associated with EMT2 in cancers, with pan_F TBRs in 14 cancers, and with EMT3 in 16 cancers, while EMT1 showed a positive correlation with pyroptosis in 11 cancers (Fig. 6A). To better understand the correlations of pyroptosis with immunotherapy, we calculated the Spearman correlation coefficients of the pyroptosis score and immune cell infiltration and found that the pyroptosis score was positively associated with the immune cell infiltration score in almost all cancers except THYM and DLBC (Fig. 6B). The pyroptosis score was positively correlated with most T cells, such as Tc, Tfh, Tex, Th1, iTreg, CD8_T, CD4_T, and Tr1 cells, in most types of cancers. Positive associations were observed between the pyroptosis score and macrophages, DCs, NK cells, and T cells in most cancers (Additional file 2: Fig. S8A-B). In contrast, the pyroptosis score showed negative associations with naïve CD8 T cells, neutrophils, and Th17 cells. Some immune cells had individual patterns. For example, Th1 and Th2 cells were only negatively associated with pyroptosis levels, while Th17 cells showed a positive association with the pyroptosis level in DLBC.
We also investigated the associations between the pyroptosis level and MHC genes (Fig. 6C), immunosuppressive genes (Fig. 6D), chemokines (Fig. 6E) and their receptors (Fig. 6F). The results showed that MHC genes, immunosuppressive genes, chemokines and their receptors were positively associated with the pyroptosis level in most types of cancers. Pyroptosis was negatively correlated with these immune-related genes in some cancers. For example, human leukocyte antigen (HLA-DMA), HLA-DOB, and HLA-DQB1 were negatively associated with pyroptosis in DLBC, and KDR (an immune suppressor gene) was negatively associated with the pyroptosis level in TGCT. Similarly, chemokines (CXCR4) showed a negative association with pyroptosis in ACC. Some chemokine receptors, such as cCCL27 (8 cancers), CCL28 (7 cancers), CCL16 (5 cancers), CCL17 (5 cancers), and CCL15 (four cancers), were negatively associated with pyroptosis in several cancers. CCL11 showed a negative relationship with the pyroptosis level in only DLBC. These results indicated a close association between pyroptosis and the immune microenvironment in cancers, but more research is required to fully elucidate the details.
We further evaluated the correlations of the pyroptosis level with microsatellite instability (MSI) and the tumour mutation burden (TMB), which were suggested to be associated with the prognosis of multiple cancers after immunotherapy. We observed that the pyroptosis level was positively associated with MSI in COAD, STAD, THCA, and PRAD, while negative relationships were observed in LIHC, KIRP, OV, PADD, TGCT and DLBC (Additional file 2: Fig. S9A). For TMB, the pyroptosis level showed a positive association in COAD and STAD but a negative association in LUAD, PCPG, TGCT and CHOL (Additional file 2: Fig. S9B).
To better understand the correlation of pyroptosis with immunotherapy, we investigated the effect of the pyroptosis level on prognosis using three GEO cancer datasets (GSE13507: primary bladder cancer; GSE32894: urothelial carcinoma; GSE61676: non-squamous non-small cell lung cancer). The results showed that a high pyroptosis level was associated with poor OS in primary bladder cancer (Fig. 7A), with DFS in urothelial carcinoma (Fig. 7B), and with OS in non-squamous non-small cell lung cancer after immunotherapy (Fig. 7C). These results implied that pyroptosis might affect immunotherapeutic efficacy in some cancers.
Identification of potential compounds targeting pyroptosis-related genes
To further understand the association between pyroptosis and drug sensitivity, we calculated the correlation coefficients between pyroptosis genes and drug sensitivity (evaluated by the percent viability curve approach) using the CTRP and GDSC datasets. We selected the top 30 compounds targeting pyroptosis-related genes (|r|> 0.3). The results showed that the expression levels of IL-6, IL18 and GSDME were positively associated with these compounds, while CASP4 was positively correlated with sensitivity to 26 cancer drugs (Fig. 7D). AIM2, CASP3, NLRC4, TNF, and NLRP6 may be associated with tumour drug resistance (Fig. 7E). The other results for the two datasets are presented in Additional file 1: Table S6 and Additional file 1: Table S7. These results indicated that pyroptosis might be associated with cancer sensitivity to multiple drugs.