研究目的：非小细胞肺癌（Non-small cell lung cancer, NSCLC）是全球癌症相关死亡的首要病因，新辅助免疫治疗显著提高可手术NSCLC患者的生存率，但疗效异质性显著且缺乏精准预测工具。因肿瘤免疫治疗耐药的机制复杂，现有生物标志物（如PD-L1、TMB）预测效能不足，限制了新辅助免疫治疗的临床应用。肿瘤微环境（Tumor microenvironment，TME）中的代谢重编程特征和细胞异质性是介导免疫治疗耐药的重要因素及预测新辅助免疫治疗疗效的潜在生物标志物。本研究通过多组学方法系统探索NSCLC新辅助免疫治疗的疗效预测生物标志物及潜在耐药机制，旨在为优化个体化治疗策略提供科学依据。

研究方法：本研究从无创临床检测指标、基于肿瘤组织的转录组标志物和单细胞转录组标志物三个维度，逐层深入，探究预测NSCLC新辅助免疫治疗疗效的新型生物标志物。在无创临床检测生物标志物层面，本研究回顾性纳入104例接受新辅助免疫治疗的I-IIIB期NSCLC患者，系统分析基线及治疗后¹⁸F-FDG PET-CT代谢参数（SUVmax、ΔSUVmax%）、外周血炎症指标（NLR、SII）对主要病理缓解（Major pathological response, MPR）的预测效能。在组织样本的普通转录组层面，本研究整合了来自TCGA、GEO数据库的1,463例NSCLC手术样本的转录组数据，通过无监督聚类鉴定乏氧-糖酵解亚型，构建了17基因HGscore评分系统，并基于NSCLC新辅助免疫治疗的独立队列转录组数据，验证其预测MPR和患者预后的临床价值。在单细胞层面，本研究对193例NSCLC患者新辅助免疫治疗后的手术组织样本的单细胞转录组测序数据进行了系统分析，探究新辅助免疫治疗耐药相关的单细胞生物标志物和潜在的细胞生物学机制。

研究结果：通过对无创临床检测生物标志物的系统分析，研究发现ΔSUVmax%预测NSCLC新辅助免疫治疗后MPR的曲线下面积（Area under curve，AUC）在免疫单药治疗组高达1.00，免疫联合治疗组为0.94，其预测准确性显著优于PD-L1和TMB。基于组织样本的转录组数据，本研究通过无监督聚类鉴定出NSCLC两种乏氧亚型和两种糖酵解模式，系统解析了乏氧-糖酵解轴与免疫微环境的交互作用。此外，本研究还构建了基于17基因的HGscore评分系统；分析结果显示，高HGscore患者兼具“热”免疫表型与代谢重塑特征，对新辅助免疫治疗响应更佳。HGscore在独立的新辅助免疫治疗转录组队列中展现出优异的预后及疗效预测价值。在单细胞层面，本研究发现，LRRC15⁺ HaCAF的浸润水平可作为预测NSCLC新辅助免疫治疗疗效的新型生物标志物，同时分析结果揭示了LRRC15⁺ HaCAF通过促进Treg浸润介导新辅助免疫治疗耐药的潜在机制，靶向LRRC15⁺ HaCAF的联合治疗策略有望逆转免疫治疗耐药。

研究结论：本研究基于多组学数据从无创临床检测指标、肿瘤组织转录组和单细胞转录组三个层面系统探究了NSCLC新辅助免疫治疗疗效预测生物标志物及潜在耐药机制。¹⁸F-FDG PET-CT代谢参数是无创动态评估NSCLC新辅助免疫治疗疗效的优选工具。转录组层面上，本研究创新性地建立了HGscore评分系统，其基线预测疗效、术后评估复发风险的双时间点动态评估的优异效能有望为新辅助免疫治疗的个体化应用提供有效指导。基于单细胞数据的分析揭示，LRRC15⁺ HaCAF是新辅助免疫治疗的重要标志物和潜在的联合用药靶点，为进一步改善NSCLC新辅助免疫治疗疗效提供了理论依据。

Objective: Non-small cell lung cancer (NSCLC) remains the primary contributor to global cancer-related mortality. While neoadjuvant immunotherapy has markedly enhanced survival rates in operable NSCLC patients, its clinical benefits exhibit substantial heterogeneity, and reliable predictive tools remain elusive. The limited predictive accuracy of current biomarkers (e.g., PD-L1 expression, tumor mutational burden [TMB])—attributable to the intricate mechanisms underlying immunotherapy resistance—has hindered the broader clinical adoption of neoadjuvant immunotherapy. Emerging evidence implicates metabolic reprogramming and cellular heterogeneity within the tumor microenvironment (TME) as pivotal mediators of treatment resistance and promising predictors of therapeutic response. This study employs a multi-omics approach to systematically identify predictive biomarkers and elucidate resistance mechanisms in NSCLC neoadjuvant immunotherapy, ultimately advancing personalized therapeutic strategies.

Methods: Adopting a multi-tiered analytical framework, this investigation explores predictive biomarkers across three dimensions: non-invasive clinical parameters, bulk tumor transcriptomics, and single-cell resolution profiling. First, a retrospective cohort of 104 stage I-IIIB NSCLC patients undergoing neoadjuvant immunotherapy was analyzed to evaluate the predictive utility of ¹⁸F-FDG PET-CT metabolic indices (baseline SUVmax, ΔSUVmax%) and systemic inflammatory markers (neutrophil-to-lymphocyte ratio [NLR], systemic immune-inflammation index [SII]) for major pathological response (MPR). Second, leveraging transcriptomic data from 1,463 NSCLC surgical specimens (TCGA/GEO databases), hypoxia-glycolysis subtyping was performed via unsupervised clustering, culminating in the development of a 17-gene hypoxia-glycolysis score (HGscore). This signature was subsequently validated in independent neoadjuvant immunotherapy cohorts for MPR prediction and prognostic stratification. Third, single-cell RNA sequencing was conducted on post-treatment specimens from 193 patients to delineate cellular biomarkers and mechanistic drivers of immunotherapy resistance.

Results: Non-invasive biomarker analysis demonstrated exceptional discriminative capacity of ΔSUVmax%, achieving AUC values of 1.00 (monotherapy) and 0.94 (combination therapy) for MPR prediction, surpassing conventional biomarkers (PD-L1/TMB). Transcriptomic profiling revealed two hypoxia subtypes and distinct glycolysis activation patterns, with the hypoxia-glycolysis axis showing dynamic interplay with immune microenvironment remodeling. The HGscore system effectively stratified patients, with high-scoring individuals characterized by a "hot" immune phenotype coupled with metabolic reprogramming, correlating with superior immunotherapy response. External validation confirmed HGscore’s prognostic power for recurrence risk assessment. Single-cell analyses uncovered LRRC15⁺ hypoxic cancer-associated fibroblasts (HaCAFs) as a novel resistance biomarker, mechanistically linked to Treg recruitment and immunosuppressive niche formation. Combinatorial targeting of LRRC15⁺ HaCAFs emerged as a potential strategy to overcome resistance.

Conclusion: This multi-omics investigation systematically deciphers predictive biomarkers and resistance networks in NSCLC neoadjuvant immunotherapy across clinical, bulk tissue, and single-cell resolution. ¹⁸F-FDG PET-CT metabolic dynamics emerged as a robust non-invasive modality for early response assessment. The HGscore system demonstrated robust performance in dual-phase dynamic assessment (baseline efficacy prediction and postoperative risk stratification), offering actionable insights for treatment personalization. Crucially, single-cell insights position LRRC15⁺ HaCAFs as a nexus of resistance biology, highlighting its dual role as a predictive biomarker and therapeutic target. These findings collectively establish a framework for refining patient selection and developing mechanism-based combination therapies in NSCLC neoadjuvant immunotherapy.