【背景】中国是肝癌大国，贡献了全球每年新发病例的50%以上。80%的患者在首诊时已处于晚期，失去了根治性手术的机会。近年来，靶向治疗和免疫检查点抑制剂（ICI）治疗在多种恶性肿瘤中取得了重大进展，并在肝癌中表现出较好的疗效。但也存在如低应答率、高肿瘤耐受性等问题。焦亡是一种促炎的程序性细胞死亡形式，已成为肿瘤免疫微环境的关键调节因子。在执行焦亡的gasdermin家族中，gasdermin E（GSDME）已被证明通过释放炎症因子来促进抗肿瘤免疫。然而，GSDME 介导的细胞焦亡在肝癌中的功能作用仍然知之甚少。鉴于肝癌对靶免治疗的临床反应有限且异质性高，揭示细胞焦亡如何塑造肿瘤免疫微环境可能揭示增强靶免治疗疗效的新策略。

【方法】我们通过整合多篇高质量文献筛选获得了57个焦亡相关基因（PGs），并分析了TCGA和GEO数据库中肝癌肿瘤与邻近非瘤组织之间的表达差异。通过共识聚类、主成分分析（PCA）和生存分析对患者进行分群。进一步结合功能富集分析、免疫浸润分析（TIMER、TIDE）、m6A修饰状态、药物敏感性预测（GDSC）以及预后模型（bPGs）构建，评估焦亡预后模型（bPGs）在肝癌中患者预后预测中的价值临，并在ICGC队列中进行了外部验证。随后我们整合了8例靶免治疗前肝癌组织的单细胞RNA测序（scRNA-seq）数据和来自癌症基因组图谱（TCGA）的bulk RNA测序数据，以探索焦亡相关基因在靶免治疗效果预测方面的潜力。同时描绘焦亡相关的单细胞免疫微环境景观使用基于关键焦亡基因表达的主成分分析构建焦亡评分。通过在原代肝癌细胞、肝癌细胞系和小鼠模型中的实验进一步研究了仑伐替尼通过GSDME诱导焦亡的作用机制。采用免疫荧光、流式细胞术和 ELISA 等检测手段反洗免疫细胞募集和细胞因子产生。在原位肝癌小鼠模型中使用抗PD-1治疗评估治疗新方案。

【结果】共鉴定出43个在肝癌中差异表达的PGs，其中30个上调、13个下调。PPI网络和Cox单因素回归分析筛选出16个与预后显著相关的基因。基于PGs表达水平将肝癌患者分为两种焦亡状态，分组间在临床病理特征及生存预后方面存在显著差异。高焦亡组患者表现出细胞周期、ECM重构及免疫活化增强，尤其是CD8⁺T细胞浸润增加，同时伴随髓源性抑制性细胞（MDSC）增多及免疫检查点高表达（PD-1、CTLA4等）；而低焦亡组则呈现代谢下调及免疫“荒漠”状态，CAF和M2型巨噬细胞浸润增加。药物敏感性分析显示，高焦亡组对索拉非尼、卡博替尼和吉非替尼更敏感。我们进一步构建了六焦亡基因预后模型bPGs（GSDME、GZMA、IL1A、NLRC4、NLRP6、TRIM21），在TCGA和ICGC队列中均能有效预测患者的总生存期（OS）和无进展生存期（PFS）。

单细胞测序分析根据PGs确定了两种具有不同免疫表型的焦亡亚型。高焦亡亚型表现出CD8⁺ T细胞的浸润但伴随着髓源性抑制细胞（MDSC）的同步浸润，CD8⁺ T细胞免疫检查点表达升高，但这一亚型表现出对靶免治疗的高度反应。靶向药物仑伐替尼在肝癌中能够通过caspase3-GSDME诱导肝癌细胞焦亡，导致MIP-1β/CCL4 和 IL-8/CXCL8等趋化因子的表达增加。功能测定证实，GSDME介导的焦亡促进MDSC募集，并通过趋化因子信号通路诱导CD8⁺ T细胞耗竭。通过使用仑伐替尼诱导焦亡并联合抗PD-1治疗和抗MDSC招募治疗，成功治愈Hepa1-6原位肝癌小鼠。

【结论】本研究揭示了焦亡相关基因在肝细胞癌中的关键作用。所构建bPGs模型可作为肝癌患者预后与治疗反应的有效预测工具。我们进一步揭示了以GSDME为代表的焦亡基因在驱动肝癌形成炎性免疫环境的作用。靶向药物仑伐替尼能够通过GSDME诱导肝癌焦亡，并通过趋化因子介导的免疫抑制细胞募集来重塑肿瘤微环境，限制了免疫疗法的疗效。靶向焦亡-趋化因子-MDSC轴可能代表一种有前途的新策略，可以克服免疫逃避并改善接受检查点阻断治疗的肝癌患者的预后。

Background China accounts for over 50% of the newly diagnosed hepatocellular carcinoma (HCC) cases globally each year. Approximately 80% of patients are already in advanced stages at their initial diagnosis, losing the opportunity for curative surgery. In recent years, targeted therapy and immune checkpoint inhibitors (ICIs) have made significant advances in various malignancies and have demonstrated promising efficacy in HCC. However, challenges such as low response rates and high tumor resistance remain. Pyroptosis, a pro-inflammatory form of programmed cell death, has emerged as a key modulator of the tumor immune microenvironment (TME). Among the gasdermin family proteins that execute pyroptosis, gasdermin E (GSDME) has been shown to promote antitumor immunity by releasing inflammatory cytokines. Nevertheless, the functional role of GSDME-mediated pyroptosis in HCC remains poorly understood. Given the limited and heterogeneous clinical response of HCC to targeted and immunotherapy, uncovering how pyroptosis shapes the TME may reveal novel strategies to enhance therapeutic efficacy.

Methods We curated 57 pyroptosis-related genes (PGs) from multiple high-quality publications and analyzed their differential expression between tumor and adjacent non-tumor tissues in the TCGA and GEO datasets. Patients were stratified using consensus clustering, principal component analysis (PCA), and survival analysis. We further conducted functional enrichment analysis, immune infiltration analysis (via TIMER and TIDE), m6A modification status evaluation, drug sensitivity prediction (based on GDSC), and construction of a prognostic model (bPGs) to assess the predictive value of pyroptosis in HCC prognosis. External validation was performed using the ICGC cohort. Subsequently, we integrated single-cell RNA sequencing (scRNA-seq) data from 8 pre-treatment HCC tissue samples and bulk RNA-seq data from TCGA to explore the predictive potential of PGs in response to targeted and immunotherapy. A pyroptosis score was constructed using PCA based on the expression of key PGs to characterize the single-cell immune landscape associated with pyroptosis. Mechanistically, we investigated the role of lenvatinib in inducing pyroptosis through GSDME in primary HCC cells, HCC cell lines, and mouse models. Immune cell recruitment and cytokine production were assessed by immunofluorescence, flow cytometry, and ELISA. In an orthotopic HCC mouse model, anti-PD-1 therapy was used to evaluate the efficacy of the proposed therapeutic strategy.

Results We identified 43 differentially expressed PGs in HCC, with 30 upregulated and 13 downregulated. Protein-protein interaction (PPI) network analysis and univariate Cox regression identified 16 genes significantly associated with prognosis. Based on PG expression, HCC patients were stratified into two pyroptosis-related phenotypes with distinct clinicopathological features and survival outcomes. The high-pyroptosis group exhibited enhanced cell cycle activity, ECM remodeling, and immune activation—especially increased CD8⁺ T cell infiltration, along with elevated myeloid-derived suppressor cells (MDSCs) and high expression of immune checkpoints (PD-1, CTLA4, etc.). In contrast, the low-pyroptosis group showed metabolic downregulation and an immune “desert” phenotype, with increased infiltration of cancer-associated fibroblasts (CAFs) and M2 macrophages. Drug sensitivity analysis revealed that the high-pyroptosis group was more sensitive to sorafenib, cabozantinib, and gefitinib. We further constructed a six-gene prognostic signature bPGs (GSDME, GZMA, IL1A, NLRC4, NLRP6, TRIM21), which effectively predicted overall survival (OS) and progression-free survival (PFS) in both TCGA and ICGC cohorts.
 Single-cell transcriptomic analysis identified two pyroptosis-related subtypes with distinct immune phenotypes. The high-pyroptosis subtype displayed robust CD8⁺ T cell infiltration, but concurrently elevated MDSCs and immune checkpoint expression on CD8⁺ T cells. This subtype showed a strong response to targeted and immunotherapy. Lenvatinib was found to induce pyroptosis in HCC via the caspase3-GSDME pathway, leading to increased expression of chemokines such as MIP-1β/CCL4 and IL-8/CXCL8. Functional assays confirmed that GSDME-mediated pyroptosis promotes MDSC recruitment and induces CD8⁺ T cell exhaustion through chemokine signaling. Combination therapy with lenvatinib-induced pyroptosis, anti-PD-1 antibody, and MDSC recruitment blockade successfully cured orthotopic HCC in mouse models.

Conclusion This study highlights the critical role of pyroptosis-related genes in HCC. The constructed bPGs model serves as a robust tool for predicting patient prognosis and response to therapy. Furthermore, we demonstrate that GSDME-driven pyroptosis fosters an inflammatory immune microenvironment in HCC. Lenvatinib induces pyroptosis through GSDME, which, via chemokine-mediated recruitment of immunosuppressive cells, reshapes the TME and limits immunotherapeutic efficacy. Targeting the pyroptosis–chemokine–MDSC axis may represent a promising new strategy to overcome immune evasion and improve outcomes for HCC patients receiving immune checkpoint blockade therapy.