A Non-small Cell Lung Cancer Tissue Microarray to Analyze Protein Expression

迄今为止，我们已经开发了一种全面的肺组织微阵列，包括300例非小细胞肺癌和183例神经内分泌肿瘤。我们对肺癌诊断和预后的候选蛋白标志物进行了系统的评估。组织微阵列技术的发展和应用，使得对数百种人体组织标本中基因表达状态的有效评估成为可能。然而，客观评估放置在阵列上的单个组织核心上的蛋白质染色通常是耗时的，并且可能是可变的。为了克服这一限制，我们探索了一种计算机辅助半自动评分（CASA）方法，该方法可以快速可靠地对包含150例肺腺癌（AD）和150例鳞状细胞癌（SCC）的肺组织微阵列上组织核心的免疫组化染色状态进行评分。TMA切片采用Aperio ScanScope TM GL系统扫描，并使用Aperio ImageScope软件（版本7.3.36）查看。将图像传输到Aperio- tma实验室图像分析软件（版本7.0.1.235）中进行评分，使用Aperio的Positive Pixel Count算法进行分析，该算法基于每个组织核心样本中染色信号的分布和强度。我们优化了评分算法，使我们能够定量地确定单个组织核心的染色状态。为了评估该方法的稳健性，我们首先将CASA方法与由训练有素的病理学家使用聚类分析评分的12种染色体重塑蛋白的IHC染色数据进行了比较。CASA法对蛋白质因子的聚类结果与人工评分法几乎相同。这表明我们的计算机辅助评分方法与人工评分方法相当。使用这种方法，我们评估了近100种已知与癌症发展相关的蛋白质的表达。统计分析使我们能够确定其共表达与肺癌患者临床生存密切相关的基因。此外，我们在使用遗传标记进行肺癌诊断方面取得了实质性进展。我们与康奈尔大学的临床研究人员合作，成功开发了一种利用遗传标记在细针抽吸活检中识别肺癌的方法。采用荧光原位杂交（FISH）技术，以1q32、3q26、5p15和8q24染色体拷贝数增加作为FNA活检组织肿瘤状态的标记。我们用了两组肺样本。训练集包括6个石蜡包埋的非癌肺组织和33个经福尔马林固定的肺肿瘤活检，以建立肺癌检测的最佳固定和FISH评分标准。测试组包含40例混合肿瘤亚型和良性肺部疾病的FNA活检常规石蜡切片，用于评估FISH方法的敏感性和特异性。我们的研究结果表明，仅使用四种标记物，在可分析的样本中，无论肿瘤亚型、分期和大小如何，FNA活检的肿瘤状态都可以分别以100%的特异性和94%的灵敏度识别。遗传诊断在FNA活检中具有高度特异性，并且可能在肺癌筛查诊断和随访中补充常规细胞学和放射学。我们现在已经修改了这种高灵敏度和特异性的方法，以使用显色底物，因此遗传拷贝数的变化可以与组织形态学和病理学一起确定。

To date, we have developed a comprehensive lung tissue microarray that included 300 NSCLC cases and 183 neuroendocrine tumors. We have engaged in a systematical evaluation of candidate protein markers for lung cancer diagnosis and prognosis. The development and use of tissue microarray have made it possible to effectively evaluate gene expression status in hundreds of human tissue specimens. However, objective assessment of protein staining on the individual tissue core placed on the array is often time consuming and potentially variable. To overcome this limitation, we explored a computer-aided semiautomated scoring (CASA) method that allows fast and reliable scoring of the immunohsitochemistry staining status of the tissue cores on our lung tissue microarray containing 150 lung adenicarcinoma(AD) and 150 squamous cell carcinoma(SCC) cases. TMA slides were scanned by Aperio ScanScope TM GL System and viewed by Aperio ImageScope soft ware (version 7.3.36). For scoring, the images were transferred to Aperio-TMA lab image analysis software (version 7.0.1.235) and analyzed using Aperio's Positive Pixel Count Algorithm which is based on the distribution and intensity of staining signal in each tissue core sample. We optimized the scoring algorithm which allowed us to quantitatively determine the staining status of the individual tissues cores. To evaluate the robustness of this approach, we first compared the CASA approach to IHC staining data for 12 chromosome remodeling proteins scored by a trained pathologist using clustering analysis. The clustering of protein factors by CASA approach was nearly identical those obtained by manual scoring. This demonstrates that our computer-aided scoring method is comparable to the manually scoring method. Using this approach, we evaluated the expression of nearly 100 proteins commonly known to be associated with cancer development. Statistical analysis allowed us to identify genes whose co-expression is strongly associated with clinical survival of the lung cancer patients. In addition, we have made substantial progress toward using genetic markers for lung cancer diagnosis. In collaboration with clinical researchers at Cornell University, we have successfully developed a method using genetic markers to identify lung cancers in fine needle aspirate (FNA) biopsies. Chromosome copy number gains in 1q32, 3q26, 5p15 and 8q24 were used as markers to determine the neoplastic state of FNA biopsies using fluorescent in situ hybridization (FISH). We used two sets of lung of samples. The training set included six paraffin embedded non-cancer lung tissues and 33 formalin-fixed biopsies of lung tumors to establish optimal fixation and FISH scoring criteria for lung cancer detection. The testing set contained 40 routine paraffin sections of FNA biopsies with mixed tumor subtypes and benign lung diseases for used to evaluate the sensitivity and specificity of the FISH method. Our results showed that using only four markers, the neoplastic state of FNA biopsies can be identified with 100% specificity and 94% sensitivity, respectively, in analyzable samples regardless of tumor subtype, stage, and size. Genetic diagnosis is highly specific for lung identification in FNA biopsy and could potentially complement routine cytology and radiology in lung cancer screening diagnosis and follow up. We have now modified this highly sensitive and specific method to the use of chromogenic substrate so that genetic copy number changes can be determined together with tissue morphology and pathology.