第一部分：食管癌中HLA-A+三级淋巴结构及肿瘤浸润淋巴细胞与免疫治疗疗效的关系研究

目的：免疫检查点抑制剂（immune checkpoint blockade, ICB）疗法在多种实体瘤治疗中展现出显著临床疗效，然而，在食管鳞状细胞癌（esophageal squamous cell carcinoma, ESCC）中的总体响应率仍低于30%。在中国，ESCC占食管癌病例的90%，且预后较差，五年生存率仅20-30%。目前，ESCC缺乏可靠的ICB治疗疗效预测生物标志物，肿瘤微环境（tumor microenvironment, TME）中调控免疫应答的关键机制也尚未明确。三级淋巴结构（tertiary lymphoid structures, TLS）虽已被证实与多种肿瘤ICB治疗响应相关，但在ESCC中，TLS能否预测ICB治疗的临床疗效尚不明确，TLS微环境对肿瘤浸润淋巴细胞（tumor-infiltrating lymphocytes, TILs）再激活的影响同样未知。本研究旨在探究ESCC患者肿瘤组织内TLS的异质性及其内部TIL的功能状态，揭示不同标志物表达状态下TLS及其细胞组分与ICB 治疗响应的相关性，为优化ESCC免疫治疗策略提供理论依据。

方法：本研究整合多组学技术，对42例初治ESCC肿瘤组织基因表达展开系统分析，筛选与免疫治疗响应相关的差异表达基因、TIL亚群，重点关注抗原呈递体系（antigen presenting machinery, APM）等免疫相关基因标记。运用多色免疫组化（multiplex immunohistochemistry, mIHC）技术，定位识别HLA-A+ TLS，并借助该技术，结合肿瘤组织形态学特征，以CD23+生发中心的出现为标志，判定TLS成熟状态，分析驻留细胞HLA-A表达情况，明确TLS内HLA-A+细胞类型。同时，借助NanoString GeoMx的空间多靶标分析系统（Digital Spatial Profiler, DSP），从空间维度解析HLA-A+ TLS的分布、成熟状态与细胞组成。进一步纳入60例接受ICB治疗的ESCC患者样本，运用单细胞RNA测序（scRNA-seq）和NanoString GeoMx DSP技术，剖析TLS内肿瘤浸润T细胞（tumor-infiltrating T lymphocytes, TIL-T）和肿瘤浸润B细胞（tumor-infiltrating B lymphocytes, TIL-B）的转录组动态，最后通过机器学习建模，验证HLA-A+ TLS及其关键细胞组分的临床预测价值。

结果：与ICB非响应组相比，ICB治疗响应组中TIL-B以及APM等免疫相关基因特征表达显著性升高。通过mIHC鉴定出HLA-A⁺ TLS的存在，且随着TLS逐渐成熟，以CD23+生发中心的出现为标志，驻留在TLS中的细胞HLA-A表达量显著增加，且多数HLA-A+细胞为TIL-T或TIL-B。对60例食管鳞状细胞癌肿瘤组织，开展空间上不同的TIL-T淋巴细胞的空间转录组测序，结合单细胞RNA测序数据进行分析，发现随着TLS成熟，TLS内表达CXCL13的耗竭性TIL-T被重新激活，APM特征表达进一步升高。此外，DSP同样印证这一变化趋势。最后，通过综合分析证实，HLA-A+ TLS及其主要细胞成分TIL-T和TIL-B，与食管鳞状细胞癌ICB治疗的临床获益紧密相关。这些结果为食管鳞状细胞癌的免疫治疗提供了重要的理论依据和潜在的治疗靶点。

结论：HLA-A+ TLS存在于食管鳞状细胞癌肿瘤组织中。驻留在TLS中、APM 特征表达升高的TIL-T被重新激活。HLA-A+ TLS及其主要细胞成分TIL-T和 TIL-B作为接受ICB治疗的食管鳞状细胞癌患者的疗效响应生物标志物。这一发现对于食管鳞状细胞癌的精准治疗具有重要意义，有望为临床治疗方案的制定提供新的思路和方向。

第二部分：基于CRISPR和10x Genomics linked-read测序对宫颈癌MHC区域进行靶向伪单倍型分辨组装及体细胞突变特征分析研究

目的：主要组织相容性复合体（major histocompatibility complex，MHC）区域中的基因是适应性免疫系统的重要组成部分，并在肿瘤细胞免疫治疗中发挥着关键作用，但其高度的遗传多态性和复杂的结构给解析工作带来了巨大挑战。在宫颈癌领域，尽管免疫治疗为患者带来了新的希望，但较低的应答率表明亟需开发新的生物标志物以优化治疗方案。本研究旨在借助先前研究开发的基于CRISPR分离富集MHC区域与10×Genomics linked链读长测序技术平台，对高精度伪单倍型分辨组装的分析流程进行深度优化，实现深入探究MHC区域的遗传变异与宫颈癌免疫逃逸机制的关联，为免疫治疗的精准分型提供理论依据。

方法：对宫颈癌队列的肿瘤组织和白细胞，开展了一系列基于体外CRISPR技术的 MHC靶向区域富集实验：对肿瘤组织和白细胞样本分别实施细胞固定处理，随即开展CRISPR切割操作，运用脉冲场电泳技术，分离出MHC靶向区域的DNA片段。针对富集所得的白细胞MHC靶向区域DNA，采用10×Genomics linked链读长测序技术进行分析；对于肿瘤组织来源的MHC靶向区域DNA，选择全基因组测序（WGS）技术开展检测。以白细胞MHC靶向区域测序数据为基础，构建MHC区域靶向伪单倍型（pseudo-haplotype）组装流程。对10×Genomics linked 测序数据，严格执行数据过滤、比对以及组装操作。基于高质量的MHC伪单倍型组装数据，借助HiFiHLA 工具，并参考IMGT v3.55数据库，开展HLA分型研究。运用MUMmer4、razors3、MuSiCal 等多种生物信息学工具，对MHC区域的片段化缺失、体细胞突变等遗传特征展开分析。同时，预测新抗原与HLA特异的杂合性丢失（loss of heterozygosity，LOH）之间的相关性，解析MHC区域的遗传变异特征。

结果：本研究对71例宫颈癌患者的肿瘤组织和白细胞完成MHC靶向区域富集，并对白细胞开展MHC伪单倍型高精度组装，分析后获得以下研究发现：（1）数据质控与组装：成功获得MHC区域靶向伪单倍型组装，在肿瘤组织和白细胞样本中均展现出良好的有效性。样本MHC区域富集倍数大于20×，测序深度100×及以上的比例大于95%，组装大小均在4Mb以上，组装长度主要集中在4.3M。（2）HLA分型和种系分析：获得32个HLA基因的分型结果。基于分型结果进行种系分析发现宫颈癌队列中DQB1*02:02, DRB1*01:01, DQA1*01:05, DQA1*03:02, DQA1*03:03, DRB1*15:01, DRB1*15:02等等位基因与宫颈癌发病风险相关，是宫颈癌相关的致病性等位基因，并解析了其本质是由氨基酸变体决定的。（3）片段化缺失：提高了MHC区域LOH片段识别的分辨率，达到基因水平，发现LOH集中于Class Ⅰ的HLA-A附近、HLA-B和HLA-C附近，以及Class Ⅱ的DRB1、DQA1、DQB1附近和DPB1、DPB2附近区域，并且与自身肿瘤新抗原亲和度更高的等位基因拥有更高的LOH可能，证明了LOH为肿瘤免疫逃逸的重要机制 （4）突变特征：以高准确率组装为参考得到宫颈癌体细胞突变，其特征与COSMIC数据库中SBS5高度相似，说明SBS5为宫颈癌的特征突变，并推测与HPV持续诱导有关。MHC区域的突变主要分布于Ⅰ类和Ⅱ类HLA基因区域，并且在经典HLA基因中的突变显示了高度的正向选择，证明了体细胞突变同样是肿瘤逃避免疫系统的重要机制。

结论：本研究开发的新技术通过宫颈癌患者队列MHC区域的高精度组装进行分析，揭示了该区域的遗传变异特征及其与宫颈癌免疫逃逸的关系，为宫颈癌免疫治疗的精准分型和患者筛选提供了理论依据和新的研究视角。

Part I: Study on the Relationship Between HLA-A+ Tertiary Lymphoid Structures, Tumor-Infiltrating Lymphocytes and Immunotherapy Efficacy in Esophageal Cancer

Objectives: Immune checkpoint blockade (ICB) therapy has demonstrated significant clinical efficacy in various solid tumors. However, in esophageal squamous cell carcinoma (ESCC), the overall response rate remains below 30%. In China, ESCC accounts for 90% of esophageal cancer cases and has a poor prognosis, with a five-year survival rate of only 20-30%. Currently, reliable biomarkers for predicting ICB response in ESCC are lacking, and the key mechanisms regulating immune responses in the tumor microenvironment (TME) remain unclear. Tertiary lymphoid structures (TLS) have been associated with favorable responses to ICB therapy in multiple tumors. However, whether TLS can predict ICB response in ESCC remains uncertain, as does their role in the reactivation of tumor-infiltrating lymphocytes (TILs). This study aims to investigate the heterogeneity of TLS and the functional state of TILs within ESCC tumors, elucidate the association between different marker-defined TLS states and ICB response, and provide theoretical support for optimizing immunotherapy strategies in ESCC.

Methods: A multi-omics approach was applied to systematically analyze gene expression in tumor tissues from 42 treatment-naïve ESCC patients, identifying differentially expressed genes and TIL subpopulations associated with ICB response, with a particular focus on antigen-presenting machinery (APM)-related immune markers. Multiplex immunohistochemistry (mIHC) was used to locate HLA-A⁺ TLS, identify their cellular components, and assess TLS maturation based on the presence of CD23⁺ germinal centers. NanoString GeoMx spatial transcriptomics was employed to analyze the spatial distribution, maturation status, and cellular composition of HLA-A⁺ TLS. Additionally, tumor samples from 60 ESCC patients receiving ICB therapy were analyzed using single-cell RNA sequencing (scRNA-seq) and digital spatial profiling (DSP) to characterize transcriptional dynamics of tumor-infiltrating T (TIL-T) and B (TIL-B) lymphocytes within TLS. Finally, machine learning models were applied to validate the clinical predictive value of HLA-A⁺ TLS and their key cellular components.

Results: In ICB responders, TIL-B and APM-related gene signatures were significantly upregulated. mIHC identified the presence of HLA-A⁺ TLS, with progressive TLS maturation marked by the emergence of CD23⁺ germinal centers. This process was accompanied by increased HLA-A expression in resident TLS cells, which were predominantly TIL-T and TIL-B cells. Spatial transcriptomic analysis of 60 ESCC tumor tissues, combined with scRNA-seq, revealed that as TLS matured, CXCL13-expressing exhausted TIL-T cells were reactivated, coinciding with increased APM expression. DSP analysis further confirmed these trends. Ultimately, comprehensive analyses demonstrated that HLA-A⁺ TLS and their primary cellular components, TIL-T and TIL-B, were closely associated with clinical benefits from ICB therapy in ESCC, providing critical insights and potential therapeutic targets for ESCC immunotherapy.

Conclusion: HLA-A⁺ TLS are present within ESCC tumor tissues and exhibit characteristic changes in ICB responders. TIL-T cells within TLS undergo reactivation in conjunction with increased APM-related gene expression. HLA-A⁺ TLS and their major cellular components, TIL-T and TIL-B, serve as potential biomarkers for predicting ICB treatment response in ESCC patients. These findings hold significant implications for the precision immunotherapy of ESCC and may provide novel directions for clinical treatment strategies.

Part II: Targeted Pseudo-Haplotype-Resolved Assembly of the MHC Region and Analysis of Somatic Mutation Signatures in Cervical Cancer Based on CRISPR and 10x Genomics Linked-Read Sequencing

Objective: The major histocompatibility complex (MHC) region plays a crucial role in adaptive immunity and tumor immunotherapy. However, its highly polymorphic and complex genomic structure poses significant challenges for analysis. In cervical cancer, despite the promise of immunotherapy, the low response rate highlights the urgent need for novel biomarkers to optimize treatment strategies. This study aims to deeply optimize the analysis process for the assembly of high-precision pseudo-haplotypes with resolution by leveraging the technical platform developed in previous research, which is based on the separation and enrichment of the MHC region using CRISPR and the 10×Genomics linked-read sequencing technology. The goal is to conduct an in-depth exploration of the association between the genetic variations in the MHC region and their association with immune evasion in cervical cancer, thereby providing a theoretical basis for precision immunotherapy stratification.

Methods: Tumor tissues and leukocytes from a cervical cancer cohort were subjected to MHC-targeted enrichment using an in vitro CRISPR-based approach. Both sample types underwent cell fixation, followed by CRISPR cleavage, and MHC-targeted DNA fragments were isolated using pulsed-field gel electrophoresis. The enriched leukocyte-derived MHC DNA underwent 10× Genomics linked-read sequencing, while whole-genome sequencing (WGS) was performed for tumor-derived MHC DNA. A pseudo-haplotype assembly workflow was established based on the leukocyte MHC sequencing data. Stringent filtering, alignment, and assembly steps were applied to the linked-read data. High-quality pseudo-haplotype assemblies were further analyzed using HiFiHLA and the IMGT v3.55 database for HLA typing. Various bioinformatics tools, including MUMmer4, razors3, and MuSiCal, were employed to assess structural variations, such as segmental deletions and somatic mutations. The correlation between neoantigens and HLA-specific loss of heterozygosity (LOH) was also evaluated to elucidate genetic variation patterns in the MHC region.

Results: A total of 71 cervical cancer patients underwent MHC-targeted enrichment and high-precision pseudo-haplotype assembly of leukocyte-derived MHC sequences. The key findings include: (1) Data quality control and assembly, The pseudo-haplotype assembly demonstrated high accuracy and efficiency in both tumor and leukocyte samples. MHC enrichment exceeded 20×, with sequencing depths of ≥100× in over 95% of cases. The assembled MHC region exceeded 4 Mb, predominantly around 4.3 Mb. (2) HLA typing and germline analysis—HLA genotyping identified 32 HLA genes. Risk allele analysis revealed that DQB1*02:02, DRB1*01:01, DQA1*01:05, DQA1*03:02, DQA1*03:03, DRB1*15:01, and DRB1*15:02 were significantly associated with cervical cancer susceptibility, primarily due to amino acid variations. (3) Segmental deletions—The resolution of MHC region LOH detection was improved to the gene level, with LOH events predominantly occurring near HLA-A, HLA-B, and HLA-C in Class I, and DRB1, DQA1, DQB1, DPB1, and DPB2 in Class II. Notably, alleles with higher binding affinity to tumor neoantigens exhibited a higher likelihood of LOH, supporting LOH as a key mechanism of tumor immune evasion. (4) Mutation characteristics—Somatic mutation analysis revealed a mutation pattern highly similar to COSMIC SBS5, suggesting a link to persistent HPV-induced mutagenesis. Mutations in the MHC region were predominantly located in Class I and II HLA genes, with classical HLA genes showing strong positive selection, further validating somatic mutations as a crucial mechanism of immune evasion in cervical cancer.

Conclusion: This study establishes a novel high-precision MHC assembly method and provides a comprehensive analysis of genetic variations in the MHC region in cervical cancer. The findings highlight the relationship between MHC genetic variations and immune evasion, offering new insights and a theoretical basis for precision immunotherapy stratification and patient selection in cervical cancer treatment.