第一部分 外周血炎症生物标志物动态反映新辅助免疫治疗非小细胞肺癌患者的治疗反应并预测预后分析

背景和目的

新辅助免疫治疗已经显著改变了可手术切除非小细胞肺癌（Non-small cell lung cancer，NSCLC）患者的治疗方法。然而，既往基于预处理组织的静态生物标志物，如程序性死亡配体1（Programmed death ligand 1，PD-L1）和肿瘤突变负荷（Tumor mutational burden，TMB），并不能更好准确预测患者的治疗效果。既往研究显示肿瘤患者体内炎症细胞、免疫细胞和相关介质所构成的免疫微环境能够深刻改变肿瘤的发生、发展、增殖和转移。炎症生物标志物在晚期NSCLC患者免疫治疗中研究较多，但在新辅助免疫治疗NSCLC患者中的研究较少。本文系统地纳入外周血炎症生物标志物以及以前较少关注的嗜酸性粒细胞分数、预后营养指数（Prognostic nutritional index，PNI）和改良后格拉斯哥预后评分（Modified Glasgow prognostic score，mGPS），以全面分析它们在预测NSCLC患者的新辅助免疫治疗疗效和预后方面的潜力。

方法

本研究回顾性收集了189例接受新辅助免疫治疗的I-III期可手术切除的NSCLC患者的临床资料和外周血指标，并将这些患者分为训练组（Training cohort TC, n=94）和验证组（Validation cohort VC, n=95）。计算并记录接受新辅助免疫治疗NSCLC患者基线和治疗后4-6周的嗜酸性粒细胞分数（Eosinophil %）、中性粒细胞与淋巴细胞比值（Neutrophil-to-lymphocyte ratio，NLR）、血小板与淋巴细胞比值（Platelet-to-lymphocyte ratio，PLR）、全身免疫炎症指数（Systemic immune-inflammation index，SII）、单核细胞与淋巴细胞比值（Monocyte-to-lymphocyte ratio，MLR）、预后营养指数（Prognostic nutritional index ，PNI）及其变化（Δeosinophil, ΔNLR, ΔPLR, ΔSII, ΔMLR和ΔPNI）。并根据患者血清C-反应蛋白（CRP）和白蛋白水平计算改良后格拉斯哥预后评分（modified Glasgow prognostic score，mGPS）。根据新辅助免疫治疗NSCLC患者术后病理状态，分析这些炎症生物标志物与新辅助免疫治疗疗效的相关性，并分析了炎症生物标志物预测患者无事件生存期（Event-free survival，EFS）和总生存期（Overall survival，OS）的潜力。根据手术切除后的病理反应率来确定缓解组和无缓解组。手术切除后肿瘤残留≤10%的患者为缓解组（Major pathological response，MPR），其余患者为无缓解组（Non-MPR）。MPR组和非MPR组患者的分类和连续变量分别采用Fisher精确检验和非参数Mann-Whitney U检验进行比较。评估连续变量与免疫治疗MPR之间的Spearman相关系数。绘制ROC曲线以确定曲线下面积（AUC）值。取Youden系数最高值计算嗜酸性粒细胞分数、NLR、SII、PLR、MLR、PNI和mGPS的最佳临界值。采用Logistic回归分析确定影响MPR的因素。我们分别使用Kaplan-Meier（KM）生存分析和log-rank检验计算和比较患者的OS和EFS。采用单因素和多因素Cox回归分析确定影响新辅助免疫治疗队列患者预后的因素。

结果

在189例接受新辅助免疫治疗的NSCLC患者中，TC组和VC组患者性别为男性的分别占80.8%和87.4%，在VC中观察到有吸烟史患者显著多于无吸烟史患者，74.5%的TC患者和72.6%的VC患者病理类型为鳞状细胞癌。与肺腺癌相比，肺鳞状细胞癌患者的MPR率明显较高。在TC中，41例（43.6%）患者达到MPR，其中34例（占TC中所有MPR的82.9%）具有放射学PR/CR状态，在VC组中也观察到类似的结果。此外，27.7%的TC组患者和16.8%的VC组患者实现了完全病理缓解（CPR）。在主要病理反应（MPR）组患者中，治疗后嗜酸性粒细胞分数显著高，NLR、PLR、SII和MLR均显著低于非MPR组，无论是训练组还是验证组。受试者工作特征（ROC）曲线显示，治疗后的嗜酸性粒细胞分数和SII及其变化的AUC值居前两位。单因素和多因素logistic回归分析显示，治疗后嗜酸性粒细胞分数、SII、mGPS和ΔSII可以独立预测新辅助免疫治疗患者的MPR。生存分析显示，治疗后高NLR、PLR、SII、mGPS及其ΔNLR、ΔSII的升高与患者OS和EFS显著相关。

结论

本研究首次在接受新辅助免疫治疗可手术切除的NSCLC患者队列中系统性的分析了多个炎症生物标志物预测新辅助免疫治疗疗效和患者预后的价值。我们的研究结果表明，接受新辅助免疫治疗的NSCLC患者外周血中的炎症生物标志物是动态变化的，这些炎症生物标志物为预测患者新辅助免疫治疗疗效和预后提供了新的依据。

第二部分 基于血浆细胞外囊泡长链RNAs预测新辅助免疫治疗非小细胞肺癌患者的治疗反应和生存率分析

背景和目的

新辅助免疫治疗为非小细胞肺癌(NSCLC)患者带来了新的希望。然而，由于缺乏临床可行的标志物，在治疗前选择对新辅助免疫治疗反应良好的NSCLC患者和预测患者的临床结局仍然很困难。因此，寻找更亦获取的、非侵入性的和更可靠的生物标志物来预测新辅助免疫治疗NSCLC患者的疗效和预后是十分必要的。在本研究中，我们招募了78例接受新辅助免疫治疗的NSCLC患者作为三个独立的队列，并在新辅助治疗前收集每位患者的血浆样本。利用RNA-seq和qRT-PCR，我们确定了三个血浆来源的细胞外囊泡长链RNA（exLRs）作为潜在的预测因子，然后建立了一个预测NSCLC患者新辅助免疫治疗临床疗效的模型。

方法

在治疗前，我们从78名接受新辅助免疫治疗的NSCLC患者的三个队列(发现、训练和验证队列)中分离出血浆细胞外囊泡（extracellular vesicles，EVs）。为了鉴定差异表达exLRs，我们在发现队列中使用了RNA测序。随后，我们使用qRT-PCR在其他两个队列中建立并验证了新辅助免疫治疗NSCLC患者的预测模型。我们从27个应答者和无应答者之间差异表达的排名靠前的exLRs中鉴定出8个候选exLRs，并在训练队列中使用qRT-PCR检测其表达，根据差异表达及其验证结果，联合PD-L1表达构建预测新辅助免疫治疗NSCLC患者临床结果的预测模型。

结果

对EVs总RNA进行了RNA-seq测序显示，在新辅助免疫治疗NSCLC患者血浆中鉴定出包括mRNA、lncRNAs和circRNAs在内的不同类型RNA，其中mRNA最丰富。差异表达分析显示，circRNA中hsa-circ-0007765在无应答组中显著表达，而hsa-circ0001485在应答组中显著高表达。在应答组中，大多数排名最高的mRNA和lncRNA均有显著表达，包括PSMA4、H3C2和MALAT1。随后qRT-PCR验证显示，两种mRNA（H3C2， P = 0.029；RPS3， P = 0.0086）和一个lncRNA（MALAT1, P = 0.043）的表达水平在应答组中显著高于无应答组。单一exLRs基因H3C2（AUC: 0.716）预测患者免疫治疗效果优于另外两个基因。两个或三个exLRs作为不同的模型进行整合，发现三个exLRs联合的模型准确率（AUC: 0.777）高于其他模型。组织来源PD-L1（AUC: 0.697）的有效性低于血浆源性EVs的H3C2和RPS3，联合PD-L1和三个exLRs的模型AUC值为0.892，表现出最佳的预测效果。此外，我们的模型被证明是新辅助免疫治疗患者预后良好的预测因子，这表明我们的模型在临床实践中的可行性（P = 0.048）。

结论

我们的研究结果表明，我们建立并验证了基于外周血exLRs的模型，该模型可以将对新辅助免疫治疗有反应的NSCLC患者与无反应的NSCLC患者进行分层，可以准确预测NSCLC患者对新辅助免疫治疗的反应和预后。

第三部分 基于肿瘤组织和公共数据库综合分析CXCR4在非小细胞肺癌患者中的免疫学意义及预后价值

中文摘要
背景和目的

CXCR4 (C-X-C趋化因子受体4型)是恶性肿瘤中最常表达的趋化因子受体。然而，关于CXCR4在非小细胞肺癌（NSCLC）肿瘤免疫微环境中的研究，包括决定其免疫功效和预后潜力的研究仍然很少。因此，在本研究中，我们在来自中国国家癌症中心的两个独立队列中分别使用免疫组织化学染色和RT-PCR来确定CXCR4预测非小细胞肺癌免疫治疗反应和预后的能力。

方法

在这项研究中，我们分析了CXCR4在肺腺癌（LUAD）和肺鳞癌（LUSC）患者中的免疫和突变状态，并确定了CXCR4在中国国家癌症中心(NCC)大样本队列中的预后潜力。我们还在另一个新辅助免疫治疗队列中，基于CXCR4 mRNA表达和相应的抗PD-1免疫治疗反应，探索了CXCR4预测NSCLC患者免疫治疗疗效的能力。最后，我们在LUAD和LUSC中鉴定了30种CXCR4相关免疫调节基因，并基于CXCR4相关免疫调节因子和CXCR4相关突变基因构建了免疫预后模型。

结果

基于CIBERSORT程序从TCGA数据库中计算LUAD和LUSC患者的基因表达谱，得到22个免疫细胞在肿瘤组织和正常组织中的浸润比率是显著不同的，22个免疫细胞在LUAD和LUSC肿瘤中呈弱至中度相关性。DNAH8、PAPPA2、SPHKAP、XIRP2和ZNF804B的突变频率在LUAD队列CXCR4高表达组和低表达组中存在显著差异；在LUSC队列CXCR4高表达组和低表达组中，CSMD2、PCDH15、RELN、SI和ZNF804A的突变频率存在显著差异。CXCR4与LUAD中经典免疫检查点基因的表达呈显著正相关，且PD1、CTLA4、TIM3、TIGIT和PD-1LG2的相关系数均大于0.5，免疫检查点相关基因在CXCR4高表达组中显著上调；在LUSC中观察到了相同的趋势。在LUAD中，CXCR4高表达组的TIDE评分明显较低，T细胞功能障碍评分较高，T细胞排斥评分较低；在LUSC中，CXCR4高表达组TMB、IPS和T细胞功能障碍评分显著升高，而TIDE评分和T细胞排斥评分显著降低。在新辅助免疫治疗独立队列中，高CXCR4表达在很大程度上与更好的免疫治疗反应一致，ROC曲线显示，CXCR4表达在LUAD免疫治疗应答队列中的AUC值为0.7778，在LUSC免疫治疗应答队列中的AUC值为0.7231。此外，在另外一个大样本的NSCLC独立队列中的免疫组化和生存分析显示，CXCR4的高表达与LUAD（P<0.001）和LUSC患者的不良预后显著相关（P=0.016），多因素分析显示，年龄、肿瘤长度、分化程度、淋巴结转移和CXCR4表达是LUAD的独立预后因素；分化程度、T分期和TNM分期被确定为LUSC的独立预后因素。在LUAD中构建了包含12种免疫调节基因的风险模型显示，高危组的OS明显短于低危组（P<0.0001），高危组和低危组的风险评分和基于TNM分期的AUC值分别为0.786和0.683。结合风险评分和TNM分期计算的AUC为0.816。在LUSC中，构建了16种CXCR4相关免疫调节基因的模型，高风险评分与LUSC患者预后不良显著相关（P<0.0001）。风险评分的ROC-AUC为0.809，结合风险评分与TNM分期计算的AUC为0.817。nomogram模型联合计算3年和5年的生存概率发现该模型具有较强的预测能力。

结论

CXCR4预测NSCLC患者新辅助免疫治疗反应方面具有较高的性能，同时是一种有潜力的患者预后生物标志物。更重要的是，构建CXCR4相关免疫调节基因的预后模型可以更准确预测LUAD和LUSC患者的生存状态，CXCR4在NSCLC患者中具有重要的免疫学意义和预测预后的价值。

Part 1: Dynamic reflection of biomarkers in peripheral blood inflammation on treatment response and prognostic analysis of neoadjuvant immunotherapy for non-small cell lung cancer patients

Abstract

Background and purpose

Neoadjuvant immunotherapy has significantly changed the treatment of patients with surgically resectable non-small cell lung cancer (NSCLC). However, previous static biomarkers based on pretreated tissues, such as programmed death ligand 1 (PD-L1) and tumor mutational burden (TMB), do not better accurately predict patient outcomes. Previous studies have shown that the immune microenvironment composed of inflammatory cells, immune cells and related mediators in tumor patients can profoundly alter tumorigenesis, development, proliferation and metastasis. Inflammatory biomarkers have been studied more in immunotherapy for patients with advanced NSCLC, but less in patients with neoadjuvant immunotherapy for NSCLC. In this study, we systematically included peripheral blood inflammatory biomarkers as well as eosinophil score, prognostic nutritional index (PNI), and modified Glasgow prognostic score (mGPS), which have received less attention before, to comprehensively analyze their potential in predicting the efficacy and prognosis of neoadjuvant immunotherapy in NSCLC patients.

Methods

In this study, clinical data and peripheral blood indices were retrospectively collected from 189 patients with stage I-III surgically resectable NSCLC who received neoadjuvant immunotherapy and these patients were divided into a training cohort （TC，n=94） and a validation cohort （VC，n=95). Eosinophil fraction, neutrophil-to-lymphocyte ratio（NLR）, platelet-to-lymphocyte ratio（PLR）, systemic immunoinflammatory index（SII）, monocyte-to-lymphocyte ratio（MLR）, prognostic nutritional indices（PNI）, and their changes (Δeosinophil, ΔNLR, ΔPLR, ΔSII, ΔMLR, and Δ PNI) were calculated and recorded at baseline and 4-6 weeks post-treatment in patients with NSCLC who received neoadjuvant immunotherapy. The modified Glasgow prognostic score (mGPS) was also calculated based on patients' serum C-reactive protein (CRP) and albumin levels. Based on the postoperative pathological status of NSCLC patients treated with neoadjuvant immunotherapy, the correlation between these inflammatory biomarkers and the efficacy of neoadjuvant immunotherapy was analyzed, and the potential of inflammatory biomarkers to predict the event-free survival (EFS) and overall survival (OS) of patients was analyzed. The pathological response rate after surgical resection was used to determine the response group and the non-response group. Patients with tumor residual ≤10% after surgical resection were classified as the major pathological response group (MPR), and the rest were classified as the non-MPR group. The categorical and continuous data of patients in the MPR and non-MPR groups were compared using Fisher’s exact test and the non-parametric Mann-Whitney U test, respectively. Spearman's correlation coefficient was assessed between continuous variables and immunotherapy MPR. ROC curves were plotted to determine the area under the curve (AUC) values. The highest Youden’s index was used for the purpose of calculating the optimal cut-off values of eosinophil, NLR, SII, PLR, MLR, PNI and mGPS. Logistic regression analyses were conducted to identify the factors predicting MPR. We calculated and compared the OS and EFS of patients using the Kaplan-Meier (KM) survival analysis and log-rank tests, respectively. Univariate and multivariate cox regression analyses were used to determine the factors predicting the prognosis of patients in the neoadjuvant immunotherapy cohort.

Results

Among 189 NSCLC patients receiving neoadjuvant immunotherapy, 80.8% and 87.4% of patients in the TC and VC groups, respectively, were male in gender, significantly more patients with a history of smoking were observed in the VC than in those without a history of smoking, and the pathologic type of the disease was squamous cell carcinoma in 74.5% of the patients in the TC and 72.6% of the patients in the VC. Compared with lung adenocarcinoma, patients with squamous cell carcinoma of the lung had a significantly higher rate of MPR. In TC, 41 (43.6%) patients achieved MPR, of which 34 (82.9% of all MPR in TC) had radiologic PR/CR status, and similar results were observed in the VC group. In addition, 27.7% of patients in the TC group and 16.8% of patients in the VC group achieved complete pathologic response (CPR). In patients in the MPR group, the post-treatment eosinophil fraction was significantly higher, and the NLR, PLR, SII, and MLR were significantly lower than those in the non-MPR group, both in the training and validation groups. The receiver operating characteristic (ROC) curves showed the top two AUC values for post-treatment eosinophil fraction and SII and their changes. Univariate and multivariate logistic regression analyses showed that post-treatment eosinophil fraction, SII, mGPS, and ΔSII independently predicted MPR in patients undergoing neoadjuvant immunotherapy. Survival analysis showed that high post-treatment NLR, PLR, SII, mGPS, and elevation of their ΔNLR and ΔSII were significantly associated with patients' OS and EFS.

Conclusions

This study is the first to systematically analyze the value of multiple inflammatory biomarkers in predicting neoadjuvant immunotherapy efficacy and patient prognosis in a cohort of patients with surgically resectable NSCLC receiving neoadjuvant immunotherapy. Our results suggest that inflammatory biomarkers in peripheral blood of NSCLC patients receiving neoadjuvant immunotherapy are dynamically changing, and these inflammatory biomarkers provide a new basis for predicting the efficacy and prognosis of patients receiving neoadjuvant immunotherapy.

Part 2: Prediction of treatment response and survival rate of neoadjuvant immunotherapy in non-small cell lung cancer patients based on plasma extracellular vesicle long chain RNAs

Abstract

Background and purpose

Neoadjuvant immunotherapy offers new hope for patients with non-small cell lung cancer (NSCLC). However, selecting NSCLC patients who respond well to neoadjuvant immunotherapy before treatment and predicting their clinical outcomes remain difficult due to the lack of clinically feasible markers. Therefore, it is necessary to search for more accessible, non-invasive and reliable biomarkers to predict the efficacy and prognosis of patients with NSCLC with neoadjuvant immunotherapy. In this study, we recruited 78 NSCLC patients receiving neoadjuvant immunotherapy as three independent cohorts and collected plasma samples from each patient prior to neoadjuvant therapy. Using RNA-seq and qRT-PCR, we identified three plasma-derived extracellular vesicle long RNAs (exLRs) as potential predictors and then developed a model to predict the clinical efficacy of neoadjuvant immunotherapy in NSCLC patients.

Methods

Prior to treatment, we isolated plasma extracellular vesicles (EVs) from three cohorts (discovery, training, and validation cohorts) of 78 NSCLC patients receiving neoadjuvant immunotherapy. To identify differentially expressed exLRs, we used RNA sequencing in the discovery cohort. Subsequently, we used qRT-PCR in two other cohorts to establish and validate predictive models in patients with neoadjuvant immunotherapy NSCLC. We identified eight candidate exLRs from the top ranked 27 differentially expressed exLRs between responders and non-responders and used qRT-PCR to detect their expression in the training cohort, and constructed a prediction model for predicting the clinical outcomes of patients with neoadjuvant immunotherapy for NSCLC based on the differential expression and its validation in combination with PD-L1 expression.

Results

RNA-sequencing of EVs total RNA showed that different types of RNA, including mRNA, lncRNAs and circRNAs, were identified in the plasma of NSCLC patients with neoadjuvant immunotherapy, among which mRNA was the most abundant. 

Differential expression analysis showed that hsa-circ-0007765 in circRNA was significantly expressed in the no-response group, while hsa-circ0001485 was significantly highly expressed in the response group. In the response group, most of the highest ranking mRNAs and lncRNAs were significantly expressed, including PSMA4, H3C2, and MALAT1. Subsequent qRT-PCR showed that two mRNAs (H3C2, P = 0.029; The expression levels of RPS3, P = 0.0086) and one lncRNA (MALAT1, P = 0.043) were significantly higher in the responding group than in the non-responding group. A single exLRs gene, H3C2 (AUC: 0.716), predicted better immunotherapy outcomes than the other two genes. When two or three exLRs were combined as different models, it was found that the combined model accuracy (AUC: 0.777) of the three exLRs was higher than that of the other models. The efficacy of tissue-derived PD-L1 (AUC: 0.697) was lower than that of H3C2 and RPS3 of plasma-derived EVs, and the AUC value of the model combining PD-L1 and three exLRs was 0.892, showing the best predictive effect. In addition, our model was shown to be a predictor of good outcomes in patients with neoadjuvant immunotherapy, indicating the feasibility of our model in clinical practice (P = 0.048).

Conclusions

Our results show that we have established and validated a peripheral blood exLRs based model, which can stratify NSCLC patients who respond to neoadjuvant immunotherapy and those who do not, and can accurately predict the response and prognosis of NSCLC patients to neoadjuvant immunotherapy.

Part 3: Comprehensive analysis of the immunological significance and prognostic value of CXCR4 in non-small cell lung cancer patients based on tumor tissue and public databases

Abstract

Background and purpose

CXCR4 (C-X-C chemokine receptor 4) is the most commonly expressed chemokine receptor in malignant tumors. However, studies on CXCR4 in the tumor immune microenvironment in non-small cell lung cancer (NSCLC), including determining its immune efficacy and prognostic potential, remain scarce. Therefore, in this study, we used immunohistochemical staining and RT-PCR in two separate cohorts from the National Cancer Center in China (NCC) to determine the ability of CXCR4 to predict immunotherapy response and prognosis in non-small cell lung cancer.

Methods

In this study, we analyzed the immune and mutational status of CXCR4 in patients with lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) and identified the prognostic potential of CXCR4 in a large sample cohort from the NCC. We also explored the ability of CXCR4 to predict immunotherapy response in NSCLC patients based on CXCR4 mRNA expression and corresponding anti-PD-1 immunotherapy response in another neoadjuvant immunotherapy cohort. Finally, we identified 30 CXCR4-related immunomodulatory genes in LUAD and LUSC, and constructed an immune prognostic model based on CXCR4-related immunomodulators and CXCR4-related mutated genes.

Results

The gene expression profiles of LUAD and LUSC patients were calculated from the TCGA database based on CIBERSORT program. It was found that the infiltration ratio of 22 immune cells in tumor tissues and normal tissues was significantly different, and 22 immune cells showed weak to moderate correlation in LUAD and LUSC tumors. The mutation frequencies of DNAH8, PAPPA2, SPHKAP, XIRP2 and ZNF804B were significantly different between CXCR4 high expression group and low expression group in LUAD cohort. In the LUSC cohort, there were significant differences in the mutation frequencies of CSMD2, PCDH15, RELN, SI and ZNF804A between the high-expression and low-expression CXCR4 groups. CXCR4 was significantly positively correlated with the expression of classical immune checkpoint genes in LUAD, and the correlation coefficients of PD1, CTLA4, TIM3, TIGIT and PD-1LG2 were all greater than 0.5. The immune checkpoint related genes were significantly up-regulated in the CXCR4 high-expression group. The same trend was observed in LUSC. In LUAD, the TIDE score of CXCR4 high expression group was significantly lower, the T cell dysfunction score was higher and the T cell rejection score was lower. In LUSC, the scores of TMB, IPS and T cell dysfunction were significantly increased in the group with high CXCR4 expression, while the scores of TIDE and T cell rejection were significantly decreased. In the neoadjuvant immunotherapy independent cohort, high CXCR4 expression was largely consistent with better immunotherapy response, with the ROC curve showing an AUC of 0.7778 in the LUAD immunotherapy response cohort and 0.7231 in the LUSC immunotherapy response cohort. In addition, immunohistochemical and survival analyses in another large NSCLC independent cohort showed that high CXCR4 expression was significantly associated with poor outcomes in patients with LUAD (P<0.001) and LUSC (P=0.016). Multivariate analysis showed that age, tumor length, degree of differentiation, lymph node metastasis and CXCR4 expression were independent prognostic factors of LUAD. Differentiation, T stage and TNM stage were identified as independent prognostic factors for LUSC. A risk model with 12 immunomodulatory genes constructed in LUAD showed that the OS in the high-risk group was significantly shorter than that in the low-risk group (P<0.0001), and the risk scores and TNM-stage-based AUC values in the high-risk and low-risk groups were 0.786 and 0.683, respectively. The AUC calculated by combining risk score and TNM staging was 0.816. In LUSC, a model of 16 CXCR4-related immunomodulatory genes was constructed, and the high-risk score was significantly associated with poor prognosis in LUSC patients (P<0.0001). The ROC-AUC of risk score was 0.809, and the AUC calculated by combining risk score and TNM staging was 0.817. The combined calculation of 3-year and 5-year survival probability by nomogram model shows that this model has strong prediction ability.

Conclusions

CXCR4 has a high performance in predicting neoadjuvant immunotherapy response in NSCLC patients and is a potential prognostic biomarker for patients. More importantly, the construction of a prognostic model of CXCR4-related immunoregulatory genes can more accurately predict the survival status of LUAD and LUSC patients. CXCR4 has important immunological significance and prognostic value in NSCLC patients.