William Bennett M.D.

LUNGevity Foundation

Lay Description

The immune system detects invaders, and serum antibodies have been used to diagnose viral and bacterial infections for decades. Useful tumor-specific antigens were harder to find, but modern molecular protocols identified serum auto-antibodies that accurately detect lung and prostate cancers. Meanwhile, cellular immunity showed little promise as a diagnostic, until cancer immunologists dissected the defenses deployed by tumors to evade immune surveillance. Most of their work concerns tumor-infiltrating lymphocytes, and their main goal is to re-activate the anti-tumor cytolytic response with exogenous cytokines and growth factors. However, dozens of reports document abnormal expression of receptors and signaling molecules in peripheral blood mononuclear cells (PBMCs). These findings show that (1) tumors alter the expression of hundreds of genes in immune cells, (2) many of the changes are evident in PBMCs, and therefore (3) gene expression markers are potential diagnostics. The last point was affirmed in 2008, when two groups reported accurate discrimination of lung cancer cases and controls by profiling messenger RNA (mRNA) expression in unfractionated peripheral blood mononuclear cells (PBMCs). We are pursuing a similar strategy, except that we are measuring DNA methylation as a proxy for mRNA expression. Our rationale is (i) expression of most human genes is regulated by promoter region methylation, and (ii) DNA is more stable than RNA, particularly in blood samples collected under routine clinical conditions.

Our proposal is based on recent work, which showed that solid tumors release cytokines, growth factors and antigens that alter gene expression in PBMCs. The implications for cancer detection are clear, but largely unexplored. We will pursue this neglected line of investigation by identifying relevant pathways and genes with CpG islands overlapping transcription start sites. Essentially, we aim to detect lung cancers by finding and measuring their disruptions of epigenetic profiles in blood-borne immune cells. Our plan is a significant departure from traditional biomarkers, which are molecules from tumor cells, e.g., prostate specific antigen (PSA) and carcinoembryonic antigen (CEA). However, proof of principle is provided by our preliminary data and recent reports, which document significant differences in PBMC methylation in cancer cases and controls. If successful, our strategy will be simpler and more practical than competing nucleic acid assays because: (1) DNA methylation markers are more stable and reliable than messenger RNA or micro-RNA molecules, (2) blood samples are more accessible than sputum, alveolar lavage, or bronchial biopsies, (3) DNA from PBMCs is abundant, while serum DNA is scarce, and (4) DNA methylation of PBMCs can be measured by standard methods, while detecting rare circulating cells or DNA from tumors requires complex technologies.

The cancer immunology literature suggests that we will find many promising candidate markers. If our evaluation studies demonstrate utility, then we may be able to develop a novel “blood test” for cancer. Given the sensitivity of the immune system, it is possible that such a test would identify cancers, of many types, at early stages of development. Considering the rapid build-up and demobilization of the cellular immune response, quantitative markers might also provide insights into prognosis, response to therapy and disease progression.

Scientific Abstract

Background-Lung neoplasia has been the leading cause of cancer death among American women and men, for decades. Although surgical resection of small tumors provides long-term survival, most lung cancers are found at late, incurable stages. Currently, computed tomography (CT) scans provide the best hope of finding treatable, early stage disease. However, annual CT screening is unattractive because (i) it is cumbersome, (ii) serial scans compound tobacco smoke damage with radiation injury, and (iii) there is little evidence that CT screening reduces mortality rates for lung cancer. Accordingly, non-invasive molecular biomarkers are an appealing alternative. A cancer “blood test” has been a holy grail for generations of researchers, but recent advances may bring it within reach.

Recent Advances-Cellular immunologists have shown that tumors evade immune surveillance by releasing cytokines, growth factors and antigens, which deflect or subvert the immune response. These findings show that (i) tumors alter the expression of hundreds of genes in immune cells, (ii) many of the changes are evident in peripheral blood cells, and (iii) gene expression markers are potential diagnostics. The last point suggests DNA methylation markers because (i) expression of most human genes is regulated by DNA methylation, and (ii) DNA is more stable than RNA, particularly in blood samples collected under routine clinical conditions. DNA methylation is a promising new class of biomarkers because (i) epigenetic changes are more common than genetic lesions (i.e., mutations, deletions, and amplifications), and (ii) methylation patterns differ in cancer vs. normal cells.

Novel Strategy-We aim to detect lung cancers by finding and measuring their disruptions of epigenetic profiles in blood-borne immune cells. Our plan is a significant departure from traditional biomarkers, which are molecules from tumor cells, e.g., prostate specific antigen (PSA) and carcinoembryonic antigen (CEA). However, proof of principle is provided by our preliminary data, and by recent reports that show significant differences in PBMC methylation in cancer cases and controls. If successful, our approach will be simpler and more practical than competing nucleic acid assays because: (1) DNA methylation markers are more stable and reliable than messenger RNA or micro-RNA molecules, (2) blood samples are more accessible than sputum, alveolar lavage, or bronchial biopsies, (3) DNA from PBMCs is abundant, while serum DNA is scarce, and (4) DNA methylation of PBMCs can be measured by standard methods, while detecting rare circulating cells or DNA from tumors requires complex technologies.

Principal Hypothesis- Quantitative methylation analysis of PBMC DNA samples can identify lung cancer patients. More specifically, we propose that lung tumors produce cytokines and growth factors that alter methylation of CpG islands in circulating immune cells.

Specific Aims: [I] Discover novel methylation markers by a candidate gene strategy. Cancer immunologists have identified many pathways that are dysregulated in immune cells from cancer patients. Pathway components with CpG islands overlapping promoter regions will be prioritized for experimental analysis by literature review. [II] Evaluate candidate biomarkers for diagnostic utility (i.e., can they accurately identify lung cancer patients?). Promising markers will be quantified in 146 PBMC DNA samples from lung cancer cases and controls, and the data will be analyzed to build a diagnostic panel. The accuracy of the diagnostic panel will be evaluated in an independent Validation Study, which is beyond the scope of this two-year proposal.

Samples & Laboratory Methods-This project will discover and evaluate DNA methylation markers using buffy coat samples from 146 subjects (74 cases and 72 controls). Candidate markers will (i) occupy genetic pathways that are dysregulated in immune cells from cancer patients, and (ii) have CpG islands overlapping promoter regions. Markers will be screened by PCR amplification and melting curve analysis, and promising candidates will be further analyzed by a quantitative methylation assay. The resulting data will be analyzed to define a Diagnostic Marker Panel (DMP) that accurately detects cancers. Eventually, the DMP will be tested in a Validation Study, but that is beyond the scope of a two-year proposal.

Data Analysis-A two-sample T-test or nonparametric Wilcoxon rank sum test will be used to rank candidate markers on their discriminatory ability (in descending order). Multiple biomarkers will be added to build shrunken centriods classifiers. To take into account potential DNA methylation interactions, a recursive partitioning and regression trees will be used to identify informative biomarkers, and to classify lung cancer cases and controls. A 10-fold cross-validation method will be used to get accurate estimates of the misclassification rate. The most promising markers, i.e., those yielding the lowest misclassification rate, will be combined in a Diagnostic Marker Panel.