Molecular Genetic Epidemiology of Primary Hepatocellular Carcinoma

大多数癌症表现为复杂的表型，并通过基因-基因和/或基因-环境相互作用表现出来。研究人类复杂癌症表型的理想范例是原发性肝细胞癌（HCC）。肿瘤遗传改变的分子研究已经确定p53是HCC中常见的肿瘤抑制基因。流行病学研究已经确定慢性乙型肝炎病毒感染（HBV）和黄曲霉毒素B1 （AFB1）暴露作为环境危险因素的作用。然而，大多数暴露于HBV和AFB1的个体不会发展为HCC。遗传分析正被用于评估基因在确定原发性肝细胞癌（HCC）的良好描述途径中的作用。这种方法结合了基因定位和候选位点研究，包括候选通路的所有成员。每个感兴趣的基因在其内部或附近被多个多态性位点“标记”，以确定在暴露于AFB1的人群中调节发生HCC风险的遗传因素。每个家族的个体成员（GSTA1、GSTM1、GSTM3、GSTP、GSTT1、GST12、EPHX1、EPHX2、GSTA4、GSTT2、GSTZ1、STP、COMT、ESD、DTD、CYP、MGST1）都被标记上了新的或已发表的多态性，并在巢式病例对照人群中研究了它们在HCC风险中的作用。基因座GSTM1、GSTP、GSTT1、EPHX1与HCC风险显著相关，而EPHX2位点与发病年龄相关。当根据病例的HBV状态对结果进行分层时，GSTM1和GSTT1仅在HBV（+）病例中相关，而GSTP在HBV（-）病例中相关。这些结果表明，这些基因是更详细的功能和遗传分析的候选者。目前正在一项大型病例对照研究（550例和550例对照）中检查15个候选癌症易感位点的候选基因变异。复杂性状分析中的遗传信息可以从癌症的遗传变异和体细胞（肿瘤）变异的联合研究中获得。使用Affymetrix HuSNP芯片上的全基因组简单串联重复多态性（STRP）标记、候选位点和1,300个单核苷酸多态性（snp）对HCC肿瘤/正常配对进行检测。对这些数据进行分析，以确定杂合性缺失（LOH）区域，并与使用含有12,000个特征基因的Affymetrix HG-U95A芯片从相同样品中收集的基因表达数据进行关联。在22条染色体上产生了16个以上的HCC LOH特征。我们发现，相对于非LOH区域，LOH区域的肿瘤基因（肿瘤基因和肿瘤抑制基因）数量显著高于非LOH区域。此外，通过系统发育重建研究，我们证明了这些LOH特征与基因表达结果显著相关；并鉴定了两个LOH特征，4q13.3和17q11.2，它们可能对产生HCC LOH特征很重要。该研究现已扩展到包括使用Affymetrix hd - u133芯片（45,000个探针集）的表达数据和使用Affymetrix Mapping 10K阵列（10,000个SNP）改进LOH区域的SNP数据。使用内部算法和Affymetrix CCNT工具也生成了数据来研究染色体拷贝数和杂合性损失的关系。正在使用最新的Affymetrix SNP6.0阵列进行其他实验。每个SNP Array 6.0有180多万个遗传变异标记（包括90多万个SNP和94多万个拷贝数探针）用于遗传分析。使用Affymetrix SNP6.0阵列对550例病例和550例对照进行基因分型分析。此外，在同一平台上分析了20对肿瘤/正常肝组织。基因型的估计总数为11亿。目前。我们将执行以下分析，其中涉及查询数据库、存储结果和基于分析结果改进查询的迭代过程。这些分析包括：a)病例和对照的比值比，以确定疾病相关snp。在此查询中，我们可能只包括具有高调用率的snp（例如大于或等于85%的样本具有基因型调用；最小等位基因频率超过10%；基因型质量超过一定阈值）。这个查询可能会在所有11亿个基因型行上执行。b)鉴定高关联snp的相关基因；获得这些基因的基因型，构建单倍型和LD bin，进行更结构化的分析，如单倍型覆层。c)确定snp之间的等位基因相互作用。这将需要使用多个snp作为风险评估的单个单位来评估遗传风险。d)测试和验证分析。初始关联研究将使用2/3的样本作为训练数据，以确定疾病关联变异。剩下的1/3的样品将测试结果。作为测试和验证的样本应具有可比的遗传谱和临床特征。测试和验证可以使用单SNP或多SNP作为疾病预测因子。e)对于肿瘤/正常配对的肝脏组织，我们将识别遗传异常，包括丢失杂合性和拷贝数变异。这些数据将与我们生成的表达数据进行比较，以评估基因改变与表达变化之间的相关性。收集到的基因表达、候选基因座和体细胞等位基因丢失数据将通过表达数据的分层聚类以及结果聚类与候选易感基因变异的相关性进行整合。这些信息将用于开发、测试和验证癌症/正常细胞通路模型的实验室策略。为了验证通路模型，将进行sirna实验来敲除这些通路中涉及的目标基因。

The majority of cancer presents as a complex phenotype and is manifest through gene-gene, and/or gene-environment interactions. An ideal paradigm for the investigation of complex cancer phenotypes in humans is primary hepatocellular carcinoma (HCC). Molecular studies of genetic alterations in tumors have identified p53 as a tumor suppressor gene commonly altered in HCC. Epidemiologic studies have firmly established the role of chronic hepatitis B virus infection (HBV) and aflatoxin B1 (AFB1) exposure as environmental risk factors. However, the majority of individuals exposed to HBV and AFB1 do not develop HCC. Genetic analysis is being used to assess the role of genes in well-described pathways in determining primary hepatocellular carcinoma (HCC). This approach merges gene mapping and candidate locus studies by including as candidates all the members of a pathway. Each gene of interest is "tagged" with multiple polymorphic sites, in or near it, to identify genetic factors modulating the risk of developing HCC among populations exposed to AFB1. The individual members of each family (GSTA1, GSTM1, GSTM3, GSTP, GSTT1, GST12, EPHX1, EPHX2, GSTA4, GSTT2, GSTZ1, STP, COMT, ESD, DTD, CYP, MGST1) have been tagged with new or published polymorphisms, and their role in HCC risk examined, in a nested case-control population. The loci GSTM1, GSTP, GSTT1, EPHX1 showed significant association with HCC risk while the EPHX2 locus was associated with age of onset. When results were stratified by the HBV status of the case, GSTM1 and GSTT1 were associated only in the HBV(+) cases, while GSTP was associated in the HBV(-) cases. These results indicate that these genes are candidates for more detailed functional and genetic analysis. Candidate gene variation at the 15 candidate cancer susceptibility loci are currently being examined in a large case-control study (n=550 cases and 550 controls). Genetic information in complex trait analysis may be accessible from the joint study of heritable variation and somatic (tumor) variation in cancer. HCC tumor/normal pairs were examined using a collection of genome-wide simple tandem repeat polymorphism (STRP) markers, candidate loci, and the 1,300 single nucleotide polymorphisms (SNPs) present on the Affymetrix HuSNP chip. This data was analyzed to identify regions of loss of heterozygosity (LOH), and was correlated with gene expression data collected from the same samples using Affymetrix HG-U95A chips containing 12,000 characterized genes. More than 16 LOH signatures of HCC were generated across 22 chromosomes. We found that the number of cancer genes (tumor genes and tumor suppressor genes) was significantly higher in regions of LOH relative to regions of non-LOH. In addition, through phylogeny reconstruction studies we demonstrated that these LOH signatures correlate significantly with gene expression results; and identified two LOH signatures, 4q13.3 and 17q11.2 that may be important in generating the HCC LOH signature. This study has now been expanded to include expression data using the Affymetrix HG-U133 chips ( 45,000 probe sets) and SNP data for refining the regions of LOH using the Affymetrix Mapping 10K Array (10,000 SNPS). Data has also been generated to investigate the relationship of chromosome copy number and loss of heterozygosity using in-house algorithms and the Affymetrix CCNT tool. Additional experiments are being carried out using the latest Affymetrix SNP6.0 arrays. Each single SNP Array 6.0 has over 1.8 million total markers for genetic variation (including more than 900,000 SNPs and more than 940,000 copy number probes) for genetic analysis. Using Affymetrix SNP6.0 arrays, we generated genotyping data from 550 cases and 550 controls. In addition, there are 20 pairs of tumor/normal liver tissues analyzed on the same platform. The estimated total number of genotypes is 1.1 billion. Currently. We will be performing the following analysises that involves an iterative process of querying the database, storing the results and refining the query based on the analytical results. These analysis include: a) Odds-ratio of case and control to identify disease association SNPs. We will probably include only SNPs with a high call rate (for example greater than or equal to 85% of the samples have genotype calls; minimum allele frequency exceeds 10%; genotype quality exceeds certain threshold) in this query. This query is likely to be performed across all 1.1 billion genotype rows. b) Identify genes underlying the high-association SNPs; obtain genotypes in these genes to construct haplotypes and LD bin for more structured analysis like haplotype clad. c) Identify allelic-interaction across SNPs. This would require evaluation of genetic risk using multiple SNPs as a single-unit for risk assessment. d) Test and validation analysis. The initial association study will be performed using 2/3 of the samples as training data to identify disease association variations. The results will be tested in the remaining 1/3 of samples. The samples serving as the test and validation shall have the comparable genetic profile and clinical features. The test and validation can be performed using single SNP or multi-SNP as a disease predictor. e) For tumor/normal paired liver tissues, we will identify genetic abnormality including loss-heterozygosity and copy-number variation. This data will be compared against the expression data that we have generated to evaluate the correlation between genetic alteration and expression change. Data collected on gene expression, candidate loci, and somatic allele loss will be integrated via hierarchical clustering of expression data, and correlation of the resulting clusters with variation at candidate susceptibility loci. This information will be used to develop, test, and validate laboratory strategies for pathway models of the cancer/normal cell. To validate the pathways model, siRNAs experiments will be carried out to knock down targeted genes involved in these pathways.