Establishment and validation of a gasdermin signature to evaluate the immune status and direct risk‐group classification in luminal‐B breast cancer

Dear Editor, We developed a gasdermin (GSDM) signature score to evaluate immune status and predict outcomes for luminalB breast cancer (BRCA). BRCA has become the most commonly diagnosed cancer (11.7%) and is the leading cause of cancer-relatedmortality in women.1 Various studies proposed models for prognosis prediction and subtype classification in BRCA,2 yet more precise models are needed. GSDMs regulate pyroptosis, an inflammatory form of cell death,3–5 and is associated with tumour progression,6,7 but an integrative study of different GSDMs in BRCA is still lacking. The current study explores the role of GSDMs in BRCA by interrogating a Chinese patient cohort from National Cancer Center (NCC) and The Cancer Genome Atlas (TCGA) dataset. The study design is shown in Figure S1. We first studied the differential regulation of GSDMs between BRCA and normal tissues using themulti-omics data from the TCGA.Different subtypes of BRCA exhibited different GSDM mutations (Figure S2A), but the luminal-B subtype had the most diversity (Figure 1A). We next analysed the transcriptional changes of GSDMs and found that GSDMA, GSDMC and GSDMD were up-regulated in cancer compared with normal tissues, while the contrary was true for GSDMB and GSDME (Figure 1B). Correlation of their mRNA expressions were also analysed (Figure S2B). The protein levels of GSDMs were detected by immunochemical staining and quantified, which also revealed aberrant expression in cancer (Figure 1C,D). Since the luminal-B subtype had the most significant genetical changes in GSDMs, we aimed to investigate the impact of GSDMs on outcomes of this subtype. The Cox regression model was performed to evaluate each GSDM in the NCC cohort. Univariate and adjusted-multivariate analyses revealed GSDMB, GSDMC and GSDMD as independent prognostic factors (Table S1). We used the multivariate model to construct a GSDM signature score, which

was prognostic independent of other clinical factors in both cohorts (Table S2 and S3). Next, we stratified the patients into the high and low score groups using the median as a cutoff. The two groups had comparable clinical features (Table S4 and S5), while the high score group showed significantly worsened outcomes ( Figure 1E-1H, Figure S3A,C). The receiver operator curve (ROC) revealed excellent prediction power of the GSDM signature score ( Figure 1I,J). The exact GSDM expressions in each group are shown in Figure S3B,D.
To understand the biological differences, we explored the differentially expressed genes (DEGs) between the high and low score groups using edgeR and limma packages ( Figure S4A,B). The representative pathways and molecular functions were then analysed using Gene Set Enrichment Analysis (GSEA) algorithm. The high score group was enriched for cell adhesion and DNAbinding transcription activation which were associated with tumour invasiveness (Figure 2A). The low score group was enriched for immune and defense responses (Figure 2A). Gene Set Variation Analysis (GSVA) analysis confirmed that the low score group had significantly higher enrichment in interferon responses, yet the high score group was significantly enriched for transforming growth factor-β (TGF-β) signaling ( Figure 2B). Transcription factor (TF) and kinase enrichment analysis revealed key regulators of the DEGs ( Figure 2C,D). We next constructed a protein-protein interaction (PPI) network to summarise the physical relationships between the DEGs coded proteins ( Figure 2E). Seven central clusters were derived by MCODE analysis as the hub genes ( Figure S4C). Enrichment analysis for the 31 hub genes was conducted based on their interacting targets, and the enriched terms were largely in line with the previous analysis based on all DEGs ( Figure 2F), suggesting that the hub genes are highly representative. To identify potential small chemicals and drugs targeting the hub genes, we also investigated the   Figure S4D,E).
GSDMs could modulate immune microenvironment through the release of cytokines and inflammatory contents. 4 We therefore aimed to study the immunological behaviours of BRCA in high and low score groups. Three sets of immune-related genes (immunoinhibitors, immunostimulators, and major histocompatibility complex (MHC) molecules) were compared ( Figure 3A,B, Figure S5A,C). The majority of immune-related genes were negatively correlated with the GSDM score, suggesting active immune responses in the low score group. Besides the immune-related genes, we also evaluated the immune cell components within the tumour tissue. The immune score and stromal score were calculated by ESTIMATE algorithm. The GSDM score was significantly correlated with lower immune cell abundance but not the stromal cells ( Figure 3C-F). We used CIBERSORTx to further calculate the lymphocyte abundance and found that the TCGA cohort showed significantly increased cytotoxic CD8 T cells, regulatory T cells and monocytes in the low score group ( Figure 3H, Figure S5D). Similarly, the low score group in the NCC cohort also had significantly increased cytotoxic CD8 T cells ( Figure 3G, Figure S5B).
To determine the driver mutations regulating the different biological behaviours between the two groups, we analysed the somatic mutations using the whole exosome sequencing data from the TCGA cohort. The high score group had 84.72% of patients carrying somatic mutations, while the low score group had 73.61% ( Figure 4A,B). Interestingly, the high score group had a significantly higher mutation rate in GATA-Binding Protein 3 (GATA3), mainly in the zinc finger (ZnFn) segments ( Figure 4C). Most of the GATA3 mutations were frame shift insertions or deletions. Apart from these mutations, the two groups had comparable tumour mutation burden ( Figure 4D). The cooccurrence of somatic mutations in the high and low score groups was determined ( Figure 4E,F).
The immune checkpoint inhibitors (ICIs) have revolutionized the treatment for various tumours, yet their applications are limited in BRCA and require further understandings. 8 Different forms of cell death have a major impact on the anti-tumour immunity and the responses to ICIs. Pyroptosis is one of such 'immunogenic cell deaths' which can dramatically alter the tumour immunological landscape. 4,9 In the current study, we established a GSDM signature score for patient risk classification and presented evidence of differentially regulated intrinsic cellular processes as well as tumour immune status in patients with different GSDM expression profiles. Our findings shed light on harnessing pyroptosis in enhancing anti-tumour immunity and developing immunotherapies.

A C K N O W L E D G E M E N T S
The authors give special thanks to Yao Zhang for providing support during this study.

C O N F L I C T O F I N T E R E S T
The authors declare that they have no competing interests.