An integrated genome-wide approach to discover tumor-specific antigens as potential immunologic and clinical targets in cancer

Abstract

Tumor-specific antigens (TSA) are central elements in the immune control of cancers. To systematically explore the TSA genome, we developed a computational technology called heterogeneous expression profile analysis (HEPA), which can identify genes relatively uniquely expressed in cancer cells in contrast to normal somatic tissues. Rating human genes by their HEPA score enriched for clinically useful TSA genes, nominating candidate targets whose tumor-specific expression was verified by reverse transcription PCR (RT-PCR). Coupled with HEPA, we designed a novel assay termed protein A/G-based reverse serological evaluation (PARSE) for quick detection of serum autoantibodies against an array of putative TSA genes. Remarkably, highly tumor-specific autoantibody responses against seven candidate targets were detected in 4% to 11% of patients, resulting in distinctive autoantibody signatures in lung and stomach cancers. Interrogation of a larger cohort of 149 patients and 123 healthy individuals validated the predictive value of the autoantibody signature for lung cancer. Together, our results establish an integrated technology to uncover a cancer-specific antigen genome offering a reservoir of novel immunologic and clinical targets.

Figure 1. The gene expression profiles of 8 prototype tumor-specific antigens widely adopted as clinical targets

Gene expression profiles are analyzed using Affymetrix U133 plus 2.0 microarray datasets for 34 normal tissues and 28 cancer types (see Methods). The maximumHEPA score for each antigen is labeled in the right side.

Figure 2. HEPA analysis preferentially identifies clinically important tumor-specific targets

A. The heat map of HEPA scores for reported tumor-specific antigen genes across 9 tumor entities. TSA genes reported to have clinical use are marked with colored asterisks. B. Separation of tumor-specific antigengenes active in cancer diagnostics and therapeutics from the rest of the human genome by HEPA scoring. Left: all 16435 coding genes included in the Affymetrix U133 plus 2 arrays are ranked by their maximumHEPA score. Right: the HEPA scores of the top 100 TSA genes in distinct tumor entities. Each point represents the HEPA score of one gene in one cancer type. Different cancer types are depicted by different colors. The eight prototype TSAs are highlighted in red. The complete lists of related literature are summarized in Tables S3 and S4.

Figure 3. Tumor-specific gene expression profile of newly identified TSA genes

A. Heat-map of normalized gene expression data depicting the tumor-specific expression profile of these TSA genes based on the compendium of microarray datasets from normal and cancer tissues. B. Expression profile of TSA genes in 16 normal tissues (left) and tumor-normal pairs (right). GAPDH was used as a control.

Figure 4. The principle and feasibility of PARSE and immunoprecipitation (IP) assays

A. Schematic depiction of the principle of PARSE (left) and IP assays (right). B. Detection of autoantibodies against a known cancer-testis antigen, TSPY, by ELISA. Representative results are shown with serum samples from healthy individuals (N21, N22) and HCC patients (H89, H90, and H91). The anti-TSPY antibody served as a positive control, and PBS as a negative control. *, positive serum. C. Detection of TSPY-specific autoantibodies using the PARSE assay. D. Confirmation of TSPY-specific autoantibodies by Western blot. The specific band for TSPYis indicated with arrowhead. *, non-specific bands. E. Confirmation test for TSPY-specific autoantibodies by immunoprecipitation. Input, radio-labeled TSPY. B and D: error bars, SD; n=3; C and E: to verify the binding specificity, unlabeled TSPY protein was added in 50-fold excess to compete with ³⁵S -labeled TSPY (H90 + TSPY).

Figure 5. Tumor specific autoantibodies against candidate TSAs are detected by PARSE assay

A. PARSE assay reveals autoantibodies against seven newly identified TSAs in serum samples from cancer patients or healthy individuals. The results are normalized and presented as a heat map in red color scales. Anti-His antibody is used as a positive control and PBS as a negative control. Grey color shows that no serum sample is available at the specific number. B. A subset of positive and negative sera revealed by the PARSE assay was subjected to IP evaluation. PARSE-positive sera are highlighted in red. The sample numbers indicated below each gel picture are matched to those shown in panel A. C. Diagnostic values of the panel of the four tumor antigens in a cohort with 72 lung cancer patients and 70 healthy individuals. The area under the ROC curve (AUC) is indicated in each chart. D. Validation using an independent cohort of 149 lung cancer patients and 123 healthy individuals.