Classifying patients for breast cancer by detection of autoantibodies against a panel of conformation-carrying antigens.

Patients with breast cancer elicit an autoantibody response against cancer proteins, which reflects and amplifies the cellular changes associated with tumorigenesis. Detection of autoantibodies in plasma may provide a minimally invasive mechanism for early detection of breast cancer. To identify cancer proteins that elicit a humoral response, we generated a cDNA library enriched for breast cancer genes that encode membrane and secreted proteins, which are more likely to induce an antibody response compared with intracellular proteins. To generate conformation-carrying antigens that are efficiently recognized by patients' antibodies, a eukaryotic expression strategy was established. Plasma from 200 patients with breast cancer and 200 age-matched healthy controls were measured for autoantibody activity against 20 different antigens designed to have conformational epitopes using ELISA. A conditional logistic regression model was used to select a combination of autoantibody responses against the 20 different antigens to classify patients with breast cancer from healthy controls. The best combination included ANGPTL4, DKK1, GAL1, MUC1, GFRA1, GRN, and LRRC15; however, autoantibody responses against GFRA1, GRN, and LRRC15 were inversely correlated with breast cancer. When the autoantibody responses against the 7 antigens were added to the base model, including age, BMI, race and current smoking status, the assay had the following diagnostic capabilities: c-stat (95% CI), 0.82 (0.78-0.86); sensitivity, 73%; specificity, 76%; and positive likelihood ratio (95% CI), 3.04 (2.34-3.94). The model was calibrated across risk deciles (Hosmer-Lemeshow, P = 0.13) and performed well in specific subtypes of breast cancer including estrogen receptor positive, HER-2 positive, invasive, in situ and tumor sizes >1 cm.