High-Throughput Analysis of Antigen Recognition by the B Cell Receptors of Malignant Lymphomas with High-Density Protein Microarrays.

Abstract 2945

Poster Board II-921

The etiology of indolent B cell lymphomas (iNHL) is largely unknown. However, systemic autoimmune diseases are associated with certain iNHL. Organ-restricted chronic inflammation, i.e. H. pylori-associated gastritis, Sjogren's syndrome, and others, plays an etiological role in extranodal marginal zone lymphoma (MZL). Finally, shared structural features of the B cell receptor (BCR) suggest antigen recognition in iNHL etiology, particularly in CLL. Indeed, recognition of common autoantigens has been shown for CLL BCR. However, there has been no unbiased comparative assessment of antigen binding by iNHL BCR.

We measured binding of lymphoma BCR simultaneously to 8000 human proteins displayed on high-density microarrays. BCR from 45 lymphomas, including 13 mantle cell lymphomas (MCL), 10 CLL, 5 nodal MZL, 5 diffuse large B-cell lymphomas (DLBCL), 4 follicular lymphomas (FL), 3 myelomas, 2 splenic MZL, and 2 LPL, were expressed as recombinant Fab fragments in E. coli. The most abundant among the 19 represented different VH segments were VH 4-34 (n=10), 1-69 (6), 3-30 (4), 3-21 (3), and 1-8 (3). Bound Fab was detected by 647AlexaFluor-goat-anti-human IgG. Z-scores and Z-factors (Zhang et al., 1999) were calculated for each Fab-protein interaction. Fab binding was defined as either Z-score >1.65 and Z-factor >0, or Z-score >1 and Z-factor >0.5. 108 robust Fab-Protein interactions were identified that involved 48 different proteins. 21 NHL BCR, derived from all lymphoma types, did not bind any protein. 12 BCR recognized one protein only. 9 BCR were highly polyreactive as defined by binding to ≥5 proteins (3 VH4-34-utilizing Fabs, 3 VH1-69 Fabs, 1 VH1-8 Fab, 1 VH3-21 Fab, and the only VH2-26 Fab). All Fab-protein interactions were analyzed by biclustering after adaptation of the Bimax algorithm used in gene expression studies to the protein microarray platform. 28 biclusters involving 11 Fabs and 21 proteins were identifed. 11 of the bicluster proteins have a nuclear localization; the cellular localization of 6 proteins is unknown. Subsequent hierarchical clustering of the biclusters distinguished three separate clusters. The DILIMOT algorithm identifed GxAxSxA as a potential consensus protein motif among 8 proteins clustering together (S_cons=5.29×10^-23, p=3.83×10^-8). Although the arrays used here carried the most comprehensive assembly of proteins available, they represent only a small fraction of the human proteome. Therefore, homologues to recognized proteins were identified by BLAST search and included, among others, the paraneoplastic neuronal autoantigen Ma1 as being potentially recognized by 4 lymphomas, including a primary CNS DLBCL, and 2 cell wall proteins of pathogenic bacteria. There was no preponderance of any particular NHL entity within the biclusters, except a group of 3 MZL Fabs. Each of these MZL BCR utilized VH1-69 and Vk3-20, and all 3 were classified as polyreactive with very similar protein recognition patterns, including calcium binding and coiled domain 1 and the autoantigens cardiolipin and Ro-60/SS-A.

This study of a broad selection of lymphoma types and BCR structures establishes protein microarrays as a novel and valuable platform to study antigen recognition by lymphoma cells in an unbiased and quantitative fashion, and thereby to deduct a comprehensive view of antigen stimulation in iNHL development. The currently available array generations permit unbiased identification of directly recognized candidate autoantigens and potentially bound homologues in appr. 60% of cases. The remaining cases may recognize carbohydrate and/or microbial antigens not represented on the array. In addition, the development of appropriate bioinformatic tools within this study permitted to define recurrent oligo- and polyreactivity patterns of antigen recognition by lymphoma BCR. With the possible exception of MZL, these patterns appear to operate across various lymphoma entities, generally suggesting two different components in lymphoma development: A requirement for BCR-mediated stimulation in the majority of cases which may occur during various scenarios and may be specific or follow definable patterns of cross-reactivity, and full malignant transformation of the stimulated cell by genetic alterations. The latter step would be expected to define the lymphoma type through the nature of the oncogenic event and the maturation stage of the cell of origin.