A systems immunology approach to the host-tumor interaction: large-scale patterns of natural autoantibodies distinguish healthy and tumor-bearing mice.

Traditionally, immunology has considered a meaningful antibody response to be marked by large amounts of high-affinity antibodies reactive with the specific inciting antigen; the detection of small amounts of low-affinity antibodies binding to seemingly unrelated antigens has been considered to be beneath the threshold of immunological meaning. A systems-biology approach to immunology, however, suggests that large-scale patterns in the antibody repertoire might also reflect the functional state of the immune system. To investigate such global patterns of antibodies, we have used an antigen-microarray device combined with informatic analysis. Here we asked whether antibody-repertoire patterns might reflect the state of an implanted tumor. We studied the serum antibodies of inbred C57BL/6 mice before and after implantation of syngeneic 3LL tumor cells of either metastatic or non-metastatic clones. We analyzed patterns of IgG and IgM autoantibodies binding to over 300 self-antigens arrayed on slides using support vector machines and genetic algorithm techniques. We now report that antibody patterns, but not single antibodies, were informative: 1) mice, even before tumor implantation, manifest both individual and common patterns of low-titer natural autoantibodies; 2) the patterns of these autoantibodies respond to the growth of the tumor cells, and can distinguish between metastatic and non-metastatic tumor clones; and 3) curative tumor resection induces dynamic changes in these low-titer autoantibody patterns. The informative patterns included autoantibodies binding to self-molecules not known to be tumor-associated antigens (including insulin, DNA, myosin, fibrinogen) as well as to known tumor-associated antigens (including p53, cytokeratin, carbonic anhydrases, tyrosinase). Thus, low-titer autoantibodies that are not the direct products of tumor-specific immunization can still generate an immune biomarker of the body-tumor interaction. System-wide profiling of autoantibody repertoires can be informative