The Pyroptosis-Related 9 LncRNA Signature and LncRNA-miRNA-mRNA Regulatory Network in Breast Cancer: A Comprehensive Analysis Based On TCGA Database

Background: Pytoproptosis has been veried to participate in various malignancies. However, studies on pyroptosis-related lncRNAs in breast cancer and its effects on tumor immune micro-environment are still limited. Consequently, it was aimed in this study to construct a pyroptosis-related lncRNAs signature for prognostic prediction and explore the effect of the pyroptosis-related LncRNAs on tumor immune microenvironment through LncRNA-miRNA-mRNA regulatory network. Methods: The pyroptosis-related differentially expressed genes (DEGs) were discovered using differential expression analysis. The differentially expressed LncRNAs (DELncRNAs) associated with DEGs were discovered using correlation analysis. The function of DEGs was analyed using GO and KEGG analyses. The LncRNAs signature used as the prognostic model of breast cancer was constructed using univariate and multivariate Cox analysis, and the effectiveness was veried by K-M analysis and ROC curve. The risk score calculated using the prognostic model was proved as an independent factor by univariate Cox analysis, multivariate Cox analysis and PCA analysis, and used to predict patient prognosis through nomogram. The pathyways enriched in High risk group and Low risk group were analyzed by GSEA. The differences in immune cell distribution (B cell memory, T cell CD4+, T cell CD8+ among others) were analyzed using ssGSEA. The immune function (type I/II IFN response among others), immune checkpoint (ADORA2A among others) and m6A-related protein expression (FTO among others) of High risk group and Low risk group were compared. The regulatory network of pyroptosis-related LncRNA-miRNA-mRNA was constructed and the core network was extracted. The functions of the target genes of miRNA associated with DELncRNAs were explored using GO and KEGG analysis. Results: A 9 LncRNAs signature (LMNTD2-AS1, AL589765.4, AC079298.3, U62317.3, LINC02446, AL645608.7, HSD11B1-AS1, AC009119.1, AC087239.1) was constructed as the prognostic model of breast cancer. Signicant differences were discovered in immune cell distribution, immune function, immune checkpointand m6A-related protein expression between High risk group and Low risk group. The regulatory network of LncRNA-miRNA-mRNA was constructed and found to participate in the crosstalk among apoptosis, pyroptosis and necroptosis of breast cancer. Conclusions: The 9 lncRNAs signature was valuable for predicting breast cancer prognosis, and the pyroptosis-related lncRNAs inuenced tumor immune microenvironment of breast cancer through the LncRNA-miRNA-mRNA regulatory network. a few related LncRNAs were discovered as dysregulated in breast cancer , which lay theoretical basis for the application of LncRNAs in diagnosis, prognostic prediction and treatment of breast cancer. It was also reported that some LncRNAs participated in through LncRNA-miRNA-mRNA regulatory network .However, studies on LncRNAs which exert a regulatory effect on breast cancer of breast cancer through regulation of tumor immune microenvironment 27 . Lnc RNA and miRNA also participated in tumorgenesis and progression of breast cancer through mutual regulation, which has not been completely claried at present. In this study, 14 pyroptosis-related genes differentially expressed between tumor and normal samples and 213 LncRNAs associated with the pyroptosis-related DEGs were discovered using expression prole data of breast cancer patients in TCGA database. On the basis of these DELncRNAs, a novel and ecient signature of breast cancer containing 9 pyroptosis-related DELncRNAs was identied using multivariate Cox analysis and the risk score was veried as an independent prognostic factor through multivariate Cox analysis. The different pathways enriched in either high risk group or low risk group based on the signature were analyzed. Immune responses including immune function, immune checkpoint, m6A-related proteins, distribution of tumor inltrating immune cells, correlation between various immune cells were also analyzed. The regulatory network of pyroptosis-related LncRNA-miRNA-mRNA was constructed and the target genes were found to participate in the crosstalk among apoptosis, pyroptosis and necroptosis of breast cancer. The pyroptosis-related DEGs discovered in this study included BAK1, BAX, ELANE, GSDMD, IL18, TP63, AIM2, GSDMC, IL6, NLRP6, NLRP7, NOD2, PYCARD, GZMA. According to the GO analysis, the functions of DEGs were mostly involved in the regulation of cytokine (IL-1, IL-1β) and chemokine production, response to LPS, pyroptosis-related intracellular construction (inammasome, secretory granule and cytosolic granule) and molecular function related to cell death. It has been known that IL-1 β and pyroptosis-related inammsome exerted a pivotal role in various pathological process of breast cancer, such as cell multiplication and migration, tumor angiogenesis and invasion, which contributed to breast cancer recurrence 28 . In addition, IL-1 β was found to stimulate the expression of CCL2 in tumor cells and macrophages, and as a chemokine,


Introduction
Breast cancer (BC) is a kind of life-threatening malignancy which primarily affects breast tissue and exhibits high tendency of pulmonary and osseous metastasis.
According to the latest data of global cancer burden published by World Health Organization International Agency for Research on Cancer (IARC) by 2020 year-end, breast cancer has become the world No.1 cancer with 2.26 million patients, surpassing the former No.1 lung cancer with 2.2 million patients. In spite of great development in BC diagnosis and therapy in recent years, the prognosis of BC is still unsatisfactory due to lack of obvious symptoms during early stage. Consequently, it is urgently needed to develop innovative and e cient methods for diagnosis, treatment and prognostic evaluation.
Pyroptosis, known as the in ammatory programmed cell death induced by Gasdermin, was initially observed by Friedlander in 1986 as lytic death of mouse macrophages accompanied with non-speci c leackage of cell content induced by anthrax lethal toxin 1 . Based on studies over the years, the de nition of pyroptosis was nally amended by the Nomenclature Committee on Cell Death (NCCD) as the regulable cell death generated by plasma membrane pore formation induced by Gasdermin, which is frequently but not always completed by the activation of in ammatory Caspase 2 . Induced by pathogens or damages, Caspase-1 is activated by canonical in ammasome pathway, or Caspase 4/5/11 is activated by non-canonical in ammasome pathway, which subsequently activate Gasdermin and leads to cell membrane perforation and cell death 3 .
It has been discovered that pyroptosis is closely related to tumor progression and metastasis 4 as well as drug resistance 5 through regulation on tumor immune microenvironment 6 . Orita et al reported that GSDMD was up-regulated in normal tissues and down-regulated or even silenced in colorectal cancer 7 , and Wang et al reported that the proliferation of gastric cancer cells promoted by down-regulation of GSDMD was possibly related to the acceleration of cell cycle S/G2 transition 8 . Yu et al reported the transition from autophagy to pyroptosis induced by doxorubicin in melanoma cells was accomplished through the activation of eEF2k, indicating the existence of crosstalk between prognosis and other kinds of cell death 9 .
As for studies in breast cancer, a few natural products were also shown to induce pyroptosis. Draganov et al reported that ivermectin activated Caspase-1 and induced pyroptosis of breast cancer cells through the over-expression of P2X4/ P2X7 and release of ATP 10 . Pizato et al discovered that docosahexaenoic acid (DHA) inhibited proliferation of breast cancer cells by inducing pyroptosis through NF-κB/Caspase/GSDMD/IL-1β 11 . In addition, pyroptosis was also discovered to exert an essential effect on drug resistance of breast cancer cells. IL-1β, which is an important in ammatory material released during pyroptosis, was reported to induce tamoxifen resistance through down-regulation of ER-α 12 and doxorubicin through up-regulation of BIRC3 13 , indicating that the pyroptosis is a potential target for increasing e cacy of regular chemotherapeutic drugs.
Long non-coding RNA (LncRNA) is a kind of regulatory non-coding RNA with length of more than 200nt. LncRNAs participate in various physiological processes including cell differentiation regulation, cell cycle regulation, epigenetic regulation and dosage compensation effect. LncRNAs also act as the pivotal part in tumorgenesis and the dysregulation of which is closely associated with various tumorgenesis process including proliferation, metastasis 14 , invasion 15 and angiogenesis 16 . In recent years, a few related LncRNAs were discovered as dysregulated in breast cancer 17 , which lay theoretical basis for the application of LncRNAs in diagnosis, prognostic prediction and treatment of breast cancer. It was also reported that some LncRNAs participated in tumorgenesis through LncRNA-miRNA-mRNA regulatory network 18 .However, studies on LncRNAs which exert a regulatory effect on breast cancer through pyroptosis and the corresponding LncRNA-miRNA-mRNA regulatory network are still limited. Consequently, in this study we compared the expression pro le of samples from patients and healthy subjects included in TCGA, found pyroptosis-related LncRNAs signi canly associated with clinical manifestations and established an effective clinical model for prognostic prediction. Moreover, we also explored the possible mechanism of pyroptosis-related LncRNAs regulation through tumor immune microenvironment, and constructed the pyroptosis-related LncRNA-miRNA-mRNA regulatory network in order to provide new insights into the application of pyroptosis-related LncRNAs in diagnosis, prognostic prediction and treatment of breast cancer.

Collection of data
The RNA expression pro les of 1109 tumor samples and 113 normal samples were obtained from TCGA database. The information of patients with more than one invalid indexes were exluded from this study.

The identi cation of pyroptosis-related LncRNAs
In total 52 pyroptosis-related genes were collected from literature. Pearson correlation analysis was performed to determine the LncRNAs orrelated to pyroptosis-related genes. The association between LncRNAs and pyroptosis-related genes was considered valid with a correlation coe cient |R2 | > 0.4 and p < 0.001.

The identi cation and functional analyses of pyroptosis-related DEGs
The clinical indexes contained in the dataset included survival time, survival status, gender,age stage and TMN classi cation. The R package "limma" was used to screen DEGs from expression pro le data and the criteria of pyroptosis DEGs was |log2FC| ≥ 1 and FDR < 0.05. Subsequently, the R package "ggplot2" was used to perform Gene Ontology (GO ) functional analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of the obtained pyroptosisrelated DEGs.
The establishment and veri cation of pyroptosis-related LncRNAs signature Univariate Cox analysis of overal survival (OS) was performed on the pyroptosis-related DELncRNAs and multiple variate Cox analysis was performed on the obtained candidate DELncRNAs to construct the Cox regression model. The formula of risk score was de ned as follows: Risk score = (Coe cientLncRNA1 × expression LncRNA1) + (Coe cientLncRNA2 × expression LncRNA2) + … + (Coe cientLncRNAn × expression LncRNAn).
Subsequently, the patients were divided into High risk group and Low risk group based on the median value of risk scores. Kaplan-Meier analysis was performed on overall survival time of the two groups. Univariate and multiple variate Cox analysis of clinical indexes (age, gender, stage, risk score) were performed and the results were shown by forest plot. The R package "time ROC" was used to assess the accuracy of pyroptosis-related LncRNAs signature in prognostic prediction. In addition, a nomogram was also drawn to predict the survival time on the basis of clinical manifestations including gender, TMN classi cation, stage, age and risk score.

Analyses of immune cells and gene expressions
TIMER, CIBERSORT, CIBERSORT-ABS, QUANTISEQ,MCPCOUNTER,XCELL, EPIC were used to analyse the difference of immune cell distribution between Low risk group and High risk group. The correlation between various immune cell types were analyzed on the basis of results of CIBERSORT. The differences of immune function between Low risk group and High risk group were compared on the basis of ssGSEA analysis using the R package "GSVA". The expressions of checkpoint-related genes (collected from literature) were compared between Low risk group and High risk group. The expressions of m6A-related genes(collected from literature) were compared between Low risk group and High risk group. The PCA analysis of Low risk group and High risk group was performed using the R package "ggplot2". The relationship between pyroptosis-related mRNA and pyroptosis-related LncRNA and the effects on patients was demonstrated using Sankey gram by means of the R package "ggalluvial".
The construction of LncRNA-miRNA-mRNA regulatory network The targets miRNA of pyroptosis-related DELncRNAs were predicted using miRcode database(miRcode -transcriptome-wide microRNA target prediction including lncRNAs). Sebsequently, the target genes of the predicted miRNAs were searched in three databases including miRDB (miRDB -MicroRNA Target Prediction Database), miRTarBase(miRTarBase: the experimentally validated microRNA-target interactions database (cuhk.edu.cn)), TargetScan(TargetScanHuman 7.1). Target genes which existed in at least two of the three databases were considered valid. GO and KEGG analyses of these target genes were performed and the network among the pyroptosis-related DELncRNAs, the target miRNAs of these LncRNAs and the target genes of these miRNAs was drawn using cytoscape 3.7.2. The hub genes of the network were discovered using the plug-in module of cytoscape hubba.

Statistical analyses
Chi-squared test was used to analyze the differences in clinical features. Independent prognostic factors were identi ed using univariate and multivariate Cox analyses. Time-dependent ROC curve analysis was used to evaluate the accuracy of prognostic model in predicting OS. The proportion of tumor-in ltrating immune cells, the expressions of m6A genes and immune checkpoint genes were compared between the High risk and Low risk groups using Wilcox-test. The correlation between tumor-in ltrating immune cells was analysed using Spearman correlation analysis. R. version 4.0.7 was used to perform the statistical analyses in this study. The threshold value of statistical signi cance was set as P <0.05.

GO and KEGG analyses of the pyroptosis-related DEGs
In total, 14 pyroptosis-related DEGs (BAK1, BAX, ELANE, GSDMD, IL18, TP63, AIM2, GSDMC, IL6, NLRP6, NLRP7, NOD2, PYCARD, GZMA) and 97 pyroptosisrelated DELncRNAs were discovered. BP of the the pyroptosis-related DEGs was enriched in cytokine production, regulation of interleukin-1 beta production, interleukin-1 beta production, regulation of interleukin-1 production, interleukin-1 production, response to liposaccharide, positive regulation interleukin-1 beta production, chemokine production, pyroptosis. CC of the the pyroptosis-related DEGs was enriched in in ammasome complex, secretory granule lumen, cytoplasmic vesicle lumen, vesicle lumen among others. MF of the the pyroptosis-related DEGs was enriched in heat shock protein binding, protease binding, cytokine receptor binding among others (Fig.1A). KEGG was principally enriched in NOD-like receptor signalling pathway, salmonella infection, in uenza A, pathogenic Escherichia coli infection, cytosolic DNA-sensing pathway, tuberculosis,transcriptional misregulation in cancer, lipid and atherosclerosis, legionellosis among others (Fig.1B). Subsequently multiple variate Cox analysis was performed on these 21 candidate LncRNAs and 9 of them(LMNTD2-AS1, AL589765.4, AC079298.3, U62317.3, LINC02446, AL645608.7, HSD11B1-AS1, AC009119.1, AC087239.1) were nally used to construct the Cox model and the formula of risk score was as follows: The veri cation of the effectiveness of the LncRNAs signature Kaplan-Meier curve was drawn to determine the effect of the LncRNAs signature on overall survival of patients. The overall survival of the High risk group was signi cantly higher than that of the Low risk group (p < 0.001) ( Fig.2A), indicating the prognostic value of the signature in breast cancer. The risk plot of patients in High risk group and Low risk group was also drawn and the mortality increased along with the rising of risk score ( Fig.2B). ROC curve was drawn to evaluate the sensitivity and speci city of the LncRNAs signature in breast cancer prediction. The AUC of the LncRNAs signature was 0.615, indicating a moderate speci city and sensitivity (Fig.2C). The AUC at 1 year, 2 years and 3 years were 0.615, 0.626, 0.613, respectively (Fig.2D).
The results of univariate and multivariate Cox regression analyses indicated that besides age and stage, the LncRNAs signature also acted as an independent prognostic factor of overall survival in breast cancer patients ( Fig.3A, Fig.3B). The PCA analysis based on the 4 LncRNAs of the prognostic signature presented obvious seperation between the data from High risk group and Low risk group (Fig.3C). The nomogram used to predict the survival time on the basis of clinical manifestations including gender, TMN classi cation, stage, age and risk score was shown in Fig.3D. The immune cell classi cation of patients in High risk group and Low risk group was analyzed using 7 algorithms including TIMER, CIBERSORT, CIBERSORT-ABS, QUANTISEQ, MCPCOUNTER, XCELL, EPIC (Fig.5A). The barplot of immune cell distribution by CIBERSORT was shown in Fig.5B. The correlation between immune cell types was drawn on the basis of CIBERSORT results (Fig.5C).
The immune functions of patients in High risk group and Low risk group were analyzed through ssGSEA. The immune functions including Type_I_IFN_Response, Type_II _IFN_Response, APC_co_stimulation, APC_co_inhibition, Cytolytic_activity, Check-point, T_cell_co-stimulation, MHC_class_I, parain ammation, T_cell_inhibition, In ammation_promoting, HLA, CCR were signi cantly different between patients in High risk group and Low risk group (Fig.6A). The expression levels of check-point-related genes were analyzed and all veri ed as signi cantly different between patients in High risk group and Low risk group (Fig.6B). The expression levels of m6A-related genes were analyzed and 8 genes were veri ed as signi cantly different between patients in The interaction between pyroptosis-related DEGs and pyroptosis-related DELncRNAs.
The regulatory relationship between pyroptosis-related DEGs and pyroptosis-related DELncRNAs in High risk group and Low risk group was described using Sankey gram (Fig.7).
The pyroptosis-related LncRNA-mRNA-miRNA regulatory network The pyroptosis-related LncRNA-miRNA-target regulating network was constructed using cytoscape 3.7.2 (Fig.8A) and the hub genes were discovered using the plug-in module of cytoscape hubba (Fig.8B). GO and KEGG analyses of these target genes of the miRNAs associated with DELncRNAs were also performed.
According to the GO analysis results, the target genes of the miRNAs associated with DELncRNAs were mainly enriched in cytokine receptor binding, phospholipid binding, cysteine-type endopeptidase activity involved in apoptosis, tumor necrosis factor receptor superfamily binding among others (Fig.8C).
According to the KEGG analysis results, the target genes of miRNAs corresponding to the pyroptosis-related DELncRNAs were mainly enriched in necrotosis, lipid and atherosclerosis, pathways of neurodegeneration diseases, TNF signalling pathway, miRNAs in cancer, apoptosis and various infections, such as pathogenic Escherichia coli infection, human immunode ciency virus 1 infection and especially, Coronavirus disease -COVID 19 among others (Fig.8D).

Discussions
De ned as a kind of in ammagory programmed cell death induced by gasdermin in 2015, pyroptosis has been revealed to participate in various tumors such as melanoma 19 , colorectal cancer 20 , hepatocellular carcinoma 21 , lung cancer 22 , cervical cancer 23 , leukemia 24 and breast cancer 25 . In breast cancer, the overexpression of GSDMB which belongs to the gasdermin family and acts as the executor of pyroptosis, is associated with poorer prognosis and tumorous progression 26 . In addition, the in ammasome/IL-1 pathway in pyroptosis exerted an essential role in the prgression and invasion of breast cancer through regulation of tumor immune microenvironment 27 . Lnc RNA and miRNA also participated in tumorgenesis and progression of breast cancer through mutual regulation, which has not been completely clari ed at present.
In this study, 14 pyroptosis-related genes differentially expressed between tumor and normal samples and 213 LncRNAs associated with the pyroptosisrelated DEGs were discovered using expression pro le data of breast cancer patients in TCGA database. On the basis of these DELncRNAs, a novel and e cient signature of breast cancer containing 9 pyroptosis-related DELncRNAs was identi ed using multivariate Cox analysis and the risk score was veri ed as an independent prognostic factor through multivariate Cox analysis. The different pathways enriched in either high risk group or low risk group based on the signature were analyzed. Immune responses including immune function, immune checkpoint, m6A-related proteins, distribution of tumor in ltrating immune cells, correlation between various immune cells were also analyzed. The regulatory network of pyroptosis-related LncRNA-miRNA-mRNA was constructed and the target genes were found to participate in the crosstalk among apoptosis, pyroptosis and necroptosis of breast cancer.
The pyroptosis-related DEGs discovered in this study included BAK1, BAX, ELANE, GSDMD, IL18, TP63, AIM2, GSDMC, IL6, NLRP6, NLRP7, NOD2, PYCARD, GZMA. According to the GO analysis, the functions of DEGs were mostly involved in the regulation of cytokine (IL-1, IL-1β) and chemokine production, response to LPS, pyroptosis-related intracellular construction (in ammasome, secretory granule and cytosolic granule) and molecular function related to cell death. It has been known that IL-1 β and pyroptosis-related in ammsome exerted a pivotal role in various pathological process of breast cancer, such as cell multiplication and migration, tumor angiogenesis and invasion, which contributed to breast cancer recurrence 28 . In addition, IL-1 β was found to stimulate the expression of CCL2 in tumor cells and macrophages, and as a chemokine, CCL2 recruided macrohages and facilitated the metastasis of breast cancer 29 . Our results in GO analysis were consistent with these previous ndings.
According to the KEGG analysis, the pathways in which the DEGs participated mostly included NOD-like receptor signalling pathway, various bacterial and virus infection, cytosolic DNA-sensing pathway and transcriptional misregulation in cancer. NOD-like receptors (NLRs) acted as the sensors for the assembly of canonical in ammasomes, which induced canonical pyroptosis accompanied with the release of IL-1β and IL-18 and GSDMD cleavage 30 . In ammasomes are related to both infections and non-infectious diseases. Cytosolic pattern recognition receptors (PRRs) were activated by danger-associated molecular patterns (DAMPs), such as abnormal nucleic acids, and pathogen-associated molecular patterns (PAMPs), such as liposaccharide 31 . The corresponding downstream signalling pathway was activated, leading to the production and release of type I interferons and pro-in ammatory cytokines, the assembly of in ammasomes and nally pyroptosis 32 . Our KEGG analysis results in this study were consistent with these previous ndings.  33 . LINC02446 which targeted CASP1, CASP4, GZMB, IL18, IRF1, AIM2, NLRP1, GZMA was found to inhibit the metastasis of bladder cancer cells 34 and melanoma 35 , and act as a prognostic signature in bladder cancer 36 . HSD11B1-AS1 which targeted NLRP1 and NLRP3 was reported to participate in corticoid metabolism and relevant physiological function 37 . AC079298.3 targeted NLRC4, which assembled in ammasome in responding to bacterial products and exerted an essential role in pyroptosis and in ammation-induced tumorgenesis 38 . AL589765.4, AL645608.7 and AC087239.1 targeted BAX, which was traditionally considered as a key protein of apoptosis 39 and recently reported to participate in the induction of pyroptosis through reactive oxygen species (ROS) activated by iron 40 . Besides BAX, AC087239.1 also targeted GSDMB, CHMP2A, CHMP3, CHMP4A, CHMP6, GSDMD, GPX4, PYCARD. After being cleaved by GZMA from cytotoxic lymphocytes, GSDMB triggered pyroptosis of target cells 41 . GSDMD which was cleaved by in ammatory caspases induced cell pyroptosis by forming membrane pores 42 . PYCARD took part in in ammasome activation 43 and the up-regulation of PYCARD independently predicted unfavorable prognosis of glioma 44 . CHMP2A, CHMP3, CHMP4A, CHMP6 belong to the chromatin-modifying protein/charged multivesicular body protein (CHMP) family, which participated in the formation of endocytic multivesicular bodies (MVBs) and cell cycle progression. CHMP2A was found to be stably and universially expressed in buccal-mucosa tumors 45 , and signi cantly up-regulated in progressive Renal cell carcinoma (RCC) 46 . CHMP3 as part of endosomal sorting complex required for transport (ESCRT) 47 exerted an important role in epithelial-mesenchymal transition (EMT), which was associated with tumor progression and metastasis 48 . CHMP4A was found to be associated with hypoxia inducible factor 1 49 , and up-regulated signi cantly in early esophageal cancer 50 . CHMP6 regulated endosomal cargo sorting 51 and mediated the recycling of Ras to plasma membrane and promoted the signalling pathway of growth factor 52 .
Based on the risk scores calculated with the above-mentioned eight LncRNAs signature, patients were divided into High risk group and Low risk group. The risk score was veri ed as an independent prognostic factor by multiple variate Cox analysis. According to the Kaplan-Meier analysis, the overall survival time of High risk group was signi cantly different from that of Low risk group (p < 0.001), indicating the validity of the 8 LncRNAs signature in predicting prognosis of breast cancer patients. Furthermore, the speci city and sensitivity were determined using ROC curve and shown by DCA curve. The T stage distribution of patients in High risk group was signi cantly different from that in Low risk group, indicating that the eight LncRNAs and their target proteins were possibly involved in the expansion of tumor volume and affected range of tissues.
The pathways enriched in High risk group were mainly involved in glycan biosynthesis, wnt signalling pathway, adherens junction, ECM receptor interaction, focal adhesion, TGF-β signaling pathway. Adherens junction-related proteins such as vimentin and E-cadherin exerted an essential role in malignant progression of cancer through Epithelial-Mesenchymal Transition (EMT) 53 . Focal adhesion kinase genes were up-regulated in cancer and used as targtes in cancer therapies 54 . The ECM receptor interaction was found to participate in the proliferation and invasion of kidney cancer cells 55 . It was also reported that TGF-β suppressed pyroptosis in triple-negative breast cancer 56 , and wnt/β-catenin participated in the metastasis of breast cancer 57 .
The pathways enriched in Low risk group were mainly involved in cytotoxicity mediated by natural killer cells, Fc epsilon RI signaling pathway, antigen processing and presentation, interaction between cytokine and cytokine receptor, cell adhesion molecules (CAMs), chemokine signaling pathway. The abovementioned immune reactions were all known as important in uence factors of tumor progression and metastasis. The enrichment results indicated that in ltrating immune cells and tumor immune micro-environment exerted a pivotal role in restraint of tumor expansion and invasion, and was associated with the prognosis of breast cancer patients.
The relative quantity of in ltrating immune cells of High risk group and Low risk group was evaluated using 7 algoriths including TIMER, CIBERSORT, CIBERSORT-ABS,QUANTISEQ,MCPCOUNTER,XCELL and EPIC. According to the results, B cells, CD4+ T cells, CD8+ T cells, T cell follicular helper, regulatory T cells (Tregs), activated NK cells, monocytes, macrophage M1 cells, immune scores and microenvironment scores were enriched in Low risk group, indicating the contributions of above-mentioned immune cells to tumor suppression. Meanwhile, macrophage M2 cells, activated myeloid dentritic cells, mast cells, neutrophils, endothelial cells, cancer-associated broblasts, granulocyte-monocyte progenitor were enriched in High risk group, indicating the promoting effect exerted by these cells on progression and metastasis of breast cancer. Among these immune cells, macrophage M2 cells had been known to promote the progression and metastasis of carcinomas 58 , while the roles played by mast cells in tumor immune microenviroment were quite controversial. On one hand, mast cells were once considered as a marker of favourable prognosis 59 . While on the other hand, it was con rmed as responsible to chemotherapy resistance 60 , in ammation and angiogenesis 61 by recent studies, which supported the ndings of immune cell classi cation in this study. The dual role of mast cells in breast cancer needed further investigation. Furthermore, the quantity of myeloid dentritic cells was reported as higher in breast cancer patients than healthy people 62 . Neutrophils 63 , endothelial cells 64 , cancer-associated broblasts 65 were reported as markers of unfavorable tumor prognosis, which exhibited consistence with our results, indicating that risk scores calculated with the 9 LncRNAs signature acted as a satisfactory criterion to separate the tumor immune microenvironment and prognosis of breast cancer patients. Immune checkpoint pathways normally maintain self-tolerance and avoid excessive immune responses to microbial infection. However, under pathological situation, immune checkpoint can be used by cancer to escape from immune surveilliance, and thus become an innovative target of chemotherapies 66 . In this study, the immune checkpoint-related proteins of High risk group were signi cantly different from those of Low risk group, indicating that the division of breast cancer patients by risk score calculated using the 9 LncRNAs signature adequately presented the differences in tumor immune micro-environment.
As the most universial pattern of post-transcriptional modi cation of LncRNAs and eukaryotic mRNA, N6-Methyladenosine (m6A) exerts its roles such as cancer progression and metastasis principally by recruiting eIF3 and m6A-relevant proteins with YTH domain 67 . M6A proteins have been developed as targets of chemotherapies and biomarkers for prognosis prediction 68 . In this study, the expressions of FTO, ZC3H13, METTL3, METTL14, YTHDC1, YTHDC2 were found as signi cantly different between patients in High risk group and Low risk group. It was reported that FTO promoted the progression and invasion of breast cancer through regulating the m6A levels of target genes 69 . ZC3H13 70 and METTL14 71 acted as writers in m6A pathways and were reported as tumorsuppressor genes in breast cancer ,while in this study ZC3H13 was up-regulated in High risk group and associated with unsatisfactory prognosis, which seemed ambiguous and needs further investigation. In this study, METTL3 expression was signi cantly lower in High risk group than Low risk group, which was consistent with the ndings of Y Shi et al that the metastasis of triple-negative breast cancer was promoted by reduced expression of METTL3 72 . Known as writers of m6A pathways, YTHDC1 73 / YTHDC2 74 inhibited tumor proliferation and migration and acted as protective genes. In this study, YHDC1/2 were both up-regulated in patients of High risk group than those of Low risk group and inconsistent with previous ndings, which needed further study.
LncRNAs and miRNAs are both important regulatory factors of pyroptosis. Consequently, the pyroptosis-related LncRNA-miRNA-mRNA regulatory network was constructed using cytoscape 3.7.2 and the hub genes were discovered using the plug-in module of cytoscape hubba. GO and KEGG analyses of these target genes of the miRNAs associated with DELncRNAs were also performed.
Among the hub genes of the regulaory network, MIAT exhibited the highest score. MIAT participated promoted cancer progression, migration and metastasis 75 , and was signi cantly up-regulated in breast cancer tissues and correlated with diameter, molecular classi cation, axillary lymph node metastases and TNM stages of tumors 76 . hsa-miR-24-3p predicted as the potential target of MIAT was reported to be a tumor-suppressor miRNA and negatively correlated with TMN stage 77 . hsa-miR-17-5p was also predicted as the potential target of MIAT and reported previously to promote cancer through regulation on cytotoxic T lymphocytes and NK cells 78 . hsa-miR-129-5p was reported to be up-regulated in colorectal cancer tissues and positively correlated with patient overall suvival 79 . WDFY3-AS2 was also reported as a potential prognostic factor of triple negative breast cancer 80 . The relationship between these nodes in the network still needed further studies.
In comparison with pyroptosis-related DEGs, the GO analysis results of the miRNA target genes contained cysteine-type endopeptidase activity, DNA-binding transcription activator activity, death receptor binding and TNF receptor binding, indicating the inherent correlation between pyroptosis and apoptosis. As for the KEGG analysis results of the miRNA target genes, besides the common pathways of various cancers, microbial infections and apoptosis, the miRNA target genes were also enriched in non-infective diseases such as neurodegeneration, rheumatoid arthritis and non-alcholic fatty liver disease; cancer-related pathways such as miRNAs in cancer, proteoglycans in cancer, viral cancinogenesis, PD-L1 signaling pathway; necroptosis and related pathways such as TNFa and Toll-like signaling pathway; and immune -related pathways such as NK cell-mediated toxicity, T cell receptor signaling pathway and Th17 cell differentiation. The results indicated that the target genes of pyroptosis-related LncRNA-miRNA-mRNA regulatory network was universially interrelated with various infectious, non-infectious diseases and relevant immune responses through possible crosstalk among apoptosis, pyroptosis and necroptosis.

Conclusions
As a kind of programmed cell death discovered in recent years, pyrotosis has been reported to exert an essential role in tumorgenesis, progression and metastasis, and provides an innovative target for development of cancer therapy. However, the speci c mechanism of regulatory LncRNAs in pyroptosis of breast cancer has not yet been fully clari ed. In this study, a 9-LncRNAs signature was constructed to predict the prognosis of breast cancer patients, and the regulatory network of LncRNA-miRNA-mRNA in pyroptosis was constructed. The miRNAs regulated by the pyroptosis-related LncRNAs possibly mediated crosstalk among pyroptosis, apoptosis and necroptosis through their target genes. Our work provided an innovative insight into the development of cancer diagnosis tools and therapies. More efforts are needed to verify the conclusion of this work with experimental and clinical approaches.

Declarations Ethics approval and consent to participate
The data used in this study is open to public in TCGA database (Home | NCI Genomic Data Commons (cancer.gov)) Concent to publication Not applicable Availability of data and materials Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.

Competing interests
The authors delcared no competing interests.   Figure 1 GO and KEGG analyses of pyroptosis-related DEGs using R packages "ggplot2" and "org.Hs.eg.db"(adjust p value <0.05) The R package "org.Hs.eg.db" was used to convert gene symbols into entrezIDs by mean of mget function. The X-axis represented the generation which was de ned as the quantity of genes enriched in a GO function or a KEGG pathway /the quantity of total involved genes. The Y-axis represented every speci c function or pathway enriched in the involved genes. Larger diameter of circle indicated higher quantity of genes enriched in each item, and color of circle closer to red indicated lower adjust p value. BP is the abbreviation of biological process. MF is the abbreviation of molecular functions.CC is the abbreviation of cellular components. A GO analysis. B KEGG analysis.

Figure 2
Evaluation on the validity of the pyroptosis-related 9 LncRNAs signature A K-M analysis revealed that the overall survival of High risk group (red) was signi cantly lower than that of Low risk group (blue) (p < 0.001). B The upper gure described the classi cation of patients on the basis of risk scores calculated with the signature formula; the middle gure indicated that the mortality was increased along with the rising risk scores; the bottome gure represented the differential expressions of the 9 LncRNAs in High risk group and Low risk group. C The AUC of the signature was 0.589 at the rst year, which was valid for the prediction of breast cancer prognosis. D The AUC values at the 1st, 2nd and 3rd year.

Figure 3
Evaluation on the prognostic value of the pyroptosis-related 9 LncRNAs signature using univariate and multivariate Cox proportional hazard ratio. PCA analysis and nomogram for pronostic prediciton A The risk score was veri ed as an individual factor associated with overall survival by univariate Cox regression analysis. Green: hazard ratio in univariate Cox regression analysis. B The risk score was veri ed as an independent prognostic factor of survival by univariate Cox regression analysis (Hazard ratio = 1.231; 95%Cl: 1.091-1.158). Red: hazard ratio in multivariate Cox regression analysis. C Grouping of patients based on PCA analysis of the expressions of pyroptosis-related 9 LncRNAs D The nomogram for prognostic prediction of breast cancer The differences in immune function correlated with immune cell classi cation on the basis of ssGSEA, checkpoint-related proteins and m6A-related proteins between High risk group (red) and Low risk group (blue) A Immune functions B Checkpoint-related proteins C m6A-related proteins Sankey plots demonstrated the relationship between 9 pyroptosis-related LncRNAs and their target genes.

Figure 8
The pyroptosis-related LncRNA-miRNA-mRNA regulatory network, the hub genes of the network, and the GO and KEGG analysis of the target genes of miRNAs associated with pyroptosis-related DELncRNAs. A The pyroptosis-related LncRNA-miRNA-mRNA regulatory network B The hub genes of the regulatory network C GO analysis of the target genes of miRNAs associated with pyroptosis-related DELncRNAs D KEGG analysis of the target genes of miRNAs associated with pyroptosis-related DELncRNAs