肺癌是癌症相关死亡最常见的原因。随着分子靶向治疗和免疫治疗的发展，非小细胞肺癌（non-small cell lung cancer，NSCLC）患者的预后和生存得到了显著的改善。然而，仍有部分患者对目前的治疗药物不敏感或在治疗后出现耐药，导致疾病进展。因此，寻找非小细胞肺癌靶向治疗及免疫治疗的疗效标志物，将有助于患者风险分层及耐药进展监测，为耐药机制的研究提供分子靶点，指导耐药患者的后续治疗。本论文的研究内容分为两个部分：第一部分以EML4-ALK阳性NSCLC患者为研究对象，基于Luminex液相悬浮芯片以及数据非依赖性采集质谱技术，对基线及动态血浆样本中的蛋白预后标志物进行筛选及验证，并在细胞和动物实验中对CCL20诱导耐药与安罗替尼逆转耐药的分子机制进行了阐述。第二部分以NSCLC免疫治疗患者为研究对象，通过bulk以及单细胞基因组学、转录组学和蛋白质组学的综合分析，构建基于肿瘤相关成纤维细胞的预后模型，并在临床队列中对FBLIM1在免疫治疗中的预后效能进行了验证。

第1部分 EML4-ALK阳性非小细胞肺癌靶向治疗耐药与预后标志物研究

第1章 趋化因子介导的血管生成在EML4-ALK阳性非小细胞肺癌克唑替尼耐药及安罗替尼逆转耐药中的作用

在间变性淋巴瘤激酶（anaplastic lymphoma kinase，ALK）阳性晚期非小细胞肺癌患者中，识别与临床疗效和耐药进展相关的早期血浆生物标志物对于患者的风险分层具有重要意义。此外，抗血管生成药物安罗替尼是否能够逆转ALK-酪氨酸激酶抑制剂（ALK-tyrosine kinase inhibitor，ALK-TKI）克唑替尼的耐药尚不清楚，目前暂无研究探讨安罗替尼联合克唑替尼对ALK阳性NSCLC的抗肿瘤作用。

在本研究中，使用Luminex液相悬浮芯片对61例接受克唑替尼治疗的ALK阳性NSCLC患者的76例基线和进展血浆样本进行了40种趋化因子的检测。RNA测序（RNA sequence，RNA-seq）用于鉴定H3122和H3122CR（克唑替尼耐药株）细胞之间的差异表达基因（differentially expressed genes，DEGs）。采用血管形成实验探究趋化因子对血管生成的影响。构建H3122CR小鼠移植瘤模型，体内验证安罗替尼联合克唑替尼的有效性和安全性。

基线和进展期血浆样本检测表明，趋化因子CCL20的表达在预测和监测克唑替尼的临床反应中起着至关重要的作用（无进展生存期风险比：2.27 (1.13-4.58)；总生存期风险比：2.7(1.23-5.8)）。H3122和H3122CR细胞的RNA-seq结果显示，趋化因子的高表达和血管生成途径参与了克唑替尼耐药。随后，体外实验表明，CCL20可能通过JAK2/STAT3-CCL20-VEGFA/IL6轴激活血管生成从而诱导克唑替尼耐药。进一步体内外研究发现，抗血管生成小分子抑制剂安罗替尼可以通过抑制趋化因子诱导的血管生成来逆转克唑替尼耐药，安罗替尼联合克唑替尼具有比单药治疗更好的抗肿瘤效果。

综上所述，CCL20介导的血管生成参与EML4-ALK阳性NSCLC克唑替尼耐药，使用安罗替尼可以逆转该耐药。安罗替尼与克唑替尼联合治疗对ALK-TKIs耐药患者来说是一种具有前景的治疗策略。

第2章 EML4-ALK阳性非小细胞肺癌靶向治疗动态血浆蛋白质组学分析

ALK-TKIs对ALK阳性的非小细胞肺癌患者具有显著的治疗效果。鉴定可靠的疗效预测及耐药监测标志物有助于增强对ALK-TKI耐药机制的理解，并指导耐药患者的后续治疗。

使用数据非依赖性采集质谱（data-independent acquisition-mass spectrometry，DIA-MS）分析了来自63例ALK阳性NSCLC患者的159例靶向治疗前和治疗过程中的血浆样本，共鉴定出737种血浆蛋白。采用一致性聚类算法识别具有不同生物学特征的亚型。利用LASSO-Cox方法构建血浆预后模型。通过Mfuzz分析对治疗过程中血浆蛋白的动态变化模式进行聚类。采集另一个独立ALK-TKI治疗队列的52例基线血浆样本，通过ELISA对潜在的预后标志物进行验证。

基线血浆蛋白质谱分析共鉴定出三种ALK阳性NSCLC亚型，分别具有不同的生物学特征和临床疗效。其中，亚组1的患者体液免疫和炎症反应激活，正向急性时相反应蛋白表达增加，治疗后易出现耐药进展，预后最差。随后，基于5种血浆蛋白（SERPINA4、ATRN、APOA4、TF和MYOC）的基线表达水平构建ALK-TKI疗效预测模型。接着，对治疗期间血浆蛋白表达的动态变化进行研究，鉴定出四种不同的变化模式（Cluster 1-4），发现急性时相反应蛋白在治疗过程中的动态改变可以反映患者的治疗状况与耐药进展。最后，在另一个接受ALK-TKI治疗的独立NSCLC队列中进一步验证了基线血浆CRP、SAA1、AHSG、SERPINA4和TF的预后预测效能。

综上，本研究通过血浆样本鉴定出与ALK-TKI疗效及耐药相关的生物标志物，为耐药机制的深入探讨和后续治疗的选择提供了新的见解。

第2部分 整合bulk和单细胞RNA-seq分析鉴定与NSCLC免疫治疗预后有关的肿瘤相关成纤维细胞标志物

肿瘤相关成纤维细胞（cancer-associated fibroblast，CAF）在肿瘤发生和肿瘤免疫抑制微环境（tumor microenvironment，TME）中起着关键作用，但有关CAFs在非小细胞肺癌中的临床意义和生物学功能的研究仍然较少。

本研究旨在通过bulk以及单细胞基因组学、转录组学和蛋白质组学的综合分析，鉴定非小细胞肺癌的CAF预后标志物。利用加权基因共表达网络分析（weighted gene co-expression network analysis，WGCNA）中鉴定到的CAF标志基因，我们在4个独立的NSCLC队列中构建并验证了基于CAF的预后风险模型。该模型将患者划分为两个预后不同的亚组。与低评分组相比，高评分组患者表现出更高的CAFs丰度、免疫细胞浸润减少、上皮-间质转化（epithelial-mesenchymal transition，EMT）增加、转化生长因子β（transforming growth factor beta，TGFβ）信号激活以及较差的生存率。基于高评分组的免疫抑制表型，我们推测这部分患者对免疫治疗的临床反应较差，并在两个接受免疫检查点抑制剂（immune checkpoint blockades，ICBs)）治疗的NSCLC队列中得到了验证。此外，通过单细胞RNA测序数据集进一步探讨高评分组患者侵袭性及免疫抑制表型的分子机制。我们发现风险模型中的基因FBLIM1（filamin binding LIM protein 1）主要表达于成纤维细胞。与正常组织成纤维细胞相比，FBLIM1在CAFs中的表达上调。FBLIM1阳性CAF亚型与TGFβ表达增加、间质标志物水平升高以及免疫抑制肿瘤微环境相关。最后，我们在临床样本中证实FBLIM1可能是免疫治疗预后不良的标志物。

综上，我们构建了一种新的基于CAFs的预后模型，对NSCLC患者以及接受ICBs治疗的患者具有预后预测价值。单细胞转录组分析发现FBLIM1阳性CAFs是一类侵袭性的CAFs，具有高表达TGFβ、诱导EMT和免疫抑制的表型。

Lung cancer is the leading cause of cancer-related death. With the development of molecular targeted therapy and immunotherapy, the prognosis and survival of non-small cell lung cancer (NSCLC) patients have been significantly improved. However, some patients remain insensitive to current therapeutic drugs or develop resistance after treatment, leading to disease progression. Therefore, identifying therapeutic markers for targeted therapy and immunotherapy in NSCLC will contribute to the risk stratification of patients and monitoring of drug resistance, provide molecular targets for studies of drug resistance, and guide the follow-up treatment for drug-resistant patients. This thesis consists of two parts: The first part focuses on EML4-ALK positive NSCLC patients. Based on Luminex liquid suspension chip and data-independent acquisition-mass spectrometry technology, we screened and validated prognostic protein markers in baseline and longitudinal plasma samples. Additionally, we elucidated the molecular mechanisms of CCL20-induced resistance and its reversal by anlotinib in cell and animal experiments. The second part focuses on NSCLC patients treated with immunotherapy. Through integrated analysis of bulk and single-cell genomics, transcriptomics, and proteomics profiles, we constructed a cancer-associated fibroblasts-related prognostic model and validated the prognostic efficacy of FBLIM1 in a clinical cohort treated with immunotherapy.

Part 1: Study on drug resistance and prognostic markers of targeted therapy for EML4-ALK positive non-small cell lung cancer

Chapter 1: Role of chemokine-mediated angiogenesis in resistance towards crizotinib and its reversal by anlotinib in EML4-ALK positive NSCLC

The identification of early plasma biomarkers for clinical outcomes and drug resistance has key importance for risk stratification in anaplastic lymphoma kinase (ALK)-positive advanced NSCLC patients. Moreover, it remains unclear whether the anti-angiogenic drug anlotinib can reverse the resistance of ALK-tyrosine kinase inhibitor (ALK-TKI) crizotinib, and no research has explored the effect of anlotinib combined with crizotinib on ALK-positive NSCLC.

In this study, 76 baseline and longitudinal plasma samples from 61 ALK-positive NSCLC patients receiving crizotinib treatment were analyzed by Luminex liquid suspension chip for 40 chemokines. RNA sequence (RNA-seq) was used to identify differentially expressed genes (DEGs) between H3122 and H3122-crizotinib resistant (H3122CR) cells. Tube formation assay was performed to investigate the effect of chemokines on angiogenesis. And H3122CR-derived xenograft model was constructed to validate the efficacy and safety of anlotinib combined with crizotinib in vivo.

Baseline and progression plasma samples detection suggested that CCL20 played a crucial role in monitoring and predicting the clinical response of crizotinib (hazard ratio for progression-free survival: 2.27 (1.13-4.58); for overall survival: 2.7 (1.23-5.8)). RNA-seq results for H3122 and H3122CR cells showed that high expression of chemokines and angiogenesis pathways were involved in crizotinib resistance. Subsequently, in vitro experiments indicated that CCL20 may induce crizotinib resistance by activation of angiogenesis via JAK2/STAT3-CCL20-VEGFA/IL6 axis. We further found that anti-angiogenic TKI anlotinib could reverse crizotinib resistance by inhibiting chemokines-induced angiogenesis, and anlotinib combined with crizotinib has a better antitumor effect than monotherapy in vitro & in vivo.

Overall, CCL20-mediated angiogenesis is involved in crizotinib resistance and could be overcome by using anlotinib in EML4-ALK positive NSCLC. The combination of anlotinib and crizotinib is a promising strategy for patients resistant to ALK-TKIs.

Chapter 2: Longitudinal plasma proteomic profiling of EML4-ALK positive lung cancer receiving ALK-TKIs therapy

The introduction of ALK-TKIs has demonstrated remarkable therapeutic effect in ALK-positive NSCLC. Identifying robust prognostic and drug-resistance biomarkers can enhance our understanding of the mechanism underlying drug resistance and guide the next-line therapies for patients.

In this study, we profiled 737 plasma proteins from 159 pre-treatment and on-treatment plasma samples of 63 ALK-positive NSCLC patients using data-independent acquisition-mass spectrometry (DIA-MS). Consensus clustering algorithm was used to identify subtypes with distinct biological features. A plasma-based prognostic model was constructed using LASSO-Cox method. We performed Mfuzz analysis to classify the patterns of longitudinal changes in plasma proteins during treatment. 52 treatment-naive plasma samples from another independent ALK-TKI treatment cohort were collected to validate the potential prognostic markers using ELISA.

We identified three subtypes of ALK-positive NSCLC with distinct biological features and clinical efficacy. Patients in subgroup 1 exhibited activated humoral immunity and inflammatory responses, increased expression of positive acute-phase response proteins, and the worst prognosis. Then we constructed and verified a prognostic model that predicts the efficacy of ALK-TKI therapy using the expression levels of five plasma proteins (SERPINA4, ATRN, APOA4, TF, and MYOC) at baseline. Next, we explored the longitudinal changes in plasma protein expression during treatment and identified four distinct change patterns (Clusters 1-4). The longitudinal changes of acute-phase proteins during treatment can reflect the treatment status and tumor progression of patients. Finally, we validated the prognostic efficacy of baseline plasma CRP, SAA1, AHSG, SERPINA4, and TF in another independent NSCLC cohort undergoing ALK-TKI treatment.

This study contributes to the search for prognostic and drug-resistance biomarkers in plasma samples for ALK-TKI therapy and provides new insights into the mechanism of drug resistance and the selection of follow-up treatment.

Part 2: Integrative analyses of bulk and single-cell RNA-seq identified cancer-associated fibroblasts-related signature as a prognostic factor for immunotherapy in NSCLC

An emerging view regarding cancer-associated fibroblast (CAF) is that it plays a critical role in tumorigenesis and immunosuppression in the tumor microenvironment (TME), but the clinical significance and biological functions of CAFs in NSCLC are still poorly explored.

Here, we aimed to identify the CAF-related signature for NSCLC through integrative analyses of bulk and single-cell genomics, transcriptomics, and proteomics profiling. Using CAF marker genes identified in weighted gene co-expression network analysis (WGCNA), we constructed and validated a CAF-based risk model that stratifies patients into two prognostic groups from four independent NSCLC cohorts. The high-score group exhibits a higher abundance of CAFs, decreased immune cell infiltration, increased epithelial-mesenchymal transition (EMT), activated transforming growth factor beta (TGFβ) signaling, and a limited survival rate compared with the low-score group. Considering the immunosuppressive feature in the high-score group, we speculated an inferior clinical response for immunotherapy in these patients, and this association was successfully verified in two NSCLC cohorts treated with immune checkpoint blockades (ICBs). Furthermore, single-cell RNA sequence datasets were used to clarify the molecular mechanisms underlying the aggressive and immunosuppressive phenotype in the high-score group. We found one of the genes in the risk model, filamin binding LIM protein 1 (FBLIM1), is mainly expressed in fibroblasts and upregulated in CAFs compared to fibroblasts from normal tissue. FBLIM1-positive CAF subtype was correlated with increased TGFβ expression, higher mesenchymal marker level, and immunosuppressive tumor microenvironment. Finally, we demonstrated that FBLIM1 might serve as a poor prognostic marker for immunotherapy in clinical samples.

In conclusion, we identified a novel CAF-based classifier with prognostic value in NSCLC patients and those treated with ICBs. Single-cell transcriptome profiling uncovered FBLIM1-positive CAFs as an aggressive subtype with a high abundance of TGFβ, EMT, and an immunosuppressive phenotype in NSCLC.