PO.MD01.01 · 分子诊断与数据
基于人工智能的局部肿瘤微环境空间分析与非小细胞肺癌中c-MET表达的关系
Artificial intelligence-based spatial analysis of the local tumor microenvironment in relation to c-MET expression in non-small cell lung cancer
作者与单位 Authors & Affiliations
摘要 Abstract
中文摘要
引言：MET是一个众所周知的致癌驱动因子，可导致多种基因组异常。抗体药物偶联物近期的批准扩展了c-MET表达型非小细胞肺癌（NSCLC）的治疗选择。然而，MET表达、肿瘤微环境（TME）与免疫治疗反应之间的关系仍不清楚。本研究探讨NSCLC中c-MET表达与空间TME特征之间的关联，以更好地理解其免疫学行为。
方法：本研究共分析了来自不同队列的25,674例NSCLC样本，包括AACR GENIE、TCGA、亚洲大学医学中心（AUMC）和Agilent Technologies。AI驱动的分析器（Lunit SCOPE IO和SCOPE uIHC）此前使用25种癌症类型的19,112张H&E和4,638张IHC全切片图像开发而成，这些切片用20多种不同抗体染色。这些平台能够对TME以及IHC染色检测到的亚细胞表达水平进行定量评估。
结果：MET改变发生于27例（2.9%）TCGA样本和909例（3.8%）GENIE样本，包括外显子14跳跃（n=388，1.5%）、扩增（n=380，1.5%）及其他（n=223，0.9%）。与野生型相比，MET改变的肿瘤具有更高的MET RNA表达（中位数：0.4对-0.2，p<0.001）。对高（Z评分≥2）和低RNA表达样本的TME分析显示，肿瘤浸润淋巴细胞（TIL）密度（/mm²，中位数：851对673，p=0.46）无显著差异。在来自AUMC和Agilent的640对H&E和IHC切片中，c-MET阳性（3+，≥50%）样本在整张切片上的TIL密度往往低于c-MET阴性样本（中位数，AUMC，78.5对79.2，p=0.16；Agilent，80.6对211.3，p=0.23）。这一趋势在细胞和亚细胞空间分析中变得显著。在表现出强（3+）c-MET表达的肿瘤细胞30μm范围内，TIL密度显著降低（中位数：85.5对121.6，p<0.001；132.2对162，p=0.013）。值得注意的是，在具有膜特异性c-MET表达的肿瘤细胞周围也观察到类似的TIL密度降低（中位数：106.6对122.3，p<0.001；144.9对165.8，p=0.002）。
结论：IHC的空间分析表明，在具有强c-MET表达或膜特异性定位的肿瘤细胞附近免疫细胞稀疏。这些发现提示c-MET过表达与免疫逃逸之间存在机制性联系，表明联合MET靶向治疗与免疫治疗的潜在获益。
查看英文原文 English abstract
Introduction: MET is a well-known oncogenic driver that confers various genomic aberrations. Recent approval of antibody drug conjugates has expanded therapeutic options for c-MET expressing non-small cell lung cancer (NSCLC). However, the relationship between MET expression, tumor microenvironment (TME), and immunotherapy response remains unclear. This study explores the association between c-MET expression and spatial TME features in NSCLC to better understand its immunologic behavior.
Methods: This study analyzed a total of 25,674 NSCLC samples from various cohorts, including AACR GENIE, TCGA, Ajou University Medical Center (AUMC), and Agilent Technologies. AI-powered analyzers (Lunit SCOPE IO and SCOPE uIHC) were previously developed using 19,112 H&E and 4,638 IHC whole slide images of 25 cancer types, stained with over 20 different antibodies. These platforms enabled the quantitative assessment of both the TME, and subcellular expression levels detected by IHC staining.
Results: MET alterations occurred in 27 (2.9%) of TCGA and 909 (3.8%) of GENIE, including exon 14 skipping (n=388, 1.5%), amplification (n=380, 1.5%), and others (n=223, 0.9%). MET-altered tumors had higher MET RNA expression compared with wild-types (median: 0.4 vs. -0.2, p<0.001). TME analysis of samples with high (Z-score ≥2) and low RNA expression showed no significant difference in tumor-infiltrating lymphocyte (TIL) density (/mm 2 , median: 851 vs. 673, p=0.46). In 640 pairs of H&E and IHC slides from AUMC and Agilent, c-MET positive (3+, ≥50%) samples tended to have lower TIL density across the whole slide compared to c-MET negative samples (median, AUMC, 78.5 vs. 79.2, p=0.16; Agilent, 80.6 vs. 211.3, p=0.23, respectively). This trend became significant with cell and subcellular spatial analysis. The density of TIL was markedly reduced within 30um of tumor cells exhibiting strong (3+) c-MET expression (median:85.5 vs. 121.6, p<0.001; 132.2 vs. 162, p=0.013, respectively). Notably, a similar reduction in TIL density was also observed around tumor cells with membrane-specific c-MET expression (median: 106.6 vs. 122.3, p<0.001; 144.9 vs. 165.8, p=0.002, respectively).
Conclusion: Spatial analysis of IHC demonstrated sparse immune cells near tumor cells with strong c-MET expression or membrane-specific localization. These findings suggest a mechanistic link between c-MET overexpression and immune evasion, indicating the potential benefit of combining MET-targeted and immunotherapy.
利益披露 Disclosure
J. Lee,
Lunit Employment.
J. Littrell,
Agilent Technologies, Inc. Employment.
C. Oum,
Lunit Employment.
T. Lee,
Lunit Employment.
S. Song,
Lunit Employment.
Y. Lim,
Lunit Employment.
C. Ahn,
Lunit Employment.
J. Christian,
Agilent Technologies, Inc. Employment.
S. Kim, None.
S. M. Ali,
Lunit Employment.