In the present studies we analyzed the Ag specificity, VH gene structure, and cellular origin of three IgM mAb-producing cell lines established in vitro from bioptic specimens of three AIDS patients with BL. We found that (i) both HBL-2 and HBL-3 IgM mAbs were cold agglutinins highly specific for the i blood group determinant, a self Ag the expression of which is dominant in the early stages of life, and both mAbs used somatically point-mutated VH 4-21 segments; (ii) HBL-1 IgM mAb, the Ag-specificity of which has not been determined, used a putatively mutated member of the VHIII family; and (iii) both HBL-1 and HBL-2, but not HBL-3, cells expressed CD5 mRNA, suggesting a B-1 cell origin. The utilization of VH4-21 by the HBL-2 and HBL-3 cold agglutinins is consistent with the usage of this gene segment by all the reported pathogenic except the naturally occurring cold agglutinins. This restricted VH gene usage may reflect an inherent affinity of the germline VH4-21 gene product for the i/I carbohydrate structure, and, perhaps, an overrepresentation of VH4-21 in the human early and late B-cell repertoire. Consistent with both an early and late developmental expression of the VH4-21 gene is the B-1 and B-2 cellular origin of the two VH4-21+ cold agglutinins reported here. Thus, the two cold agglutinin autoantibodies possibly emerged at different stages of the natural history of the B-cell repertoires of these patients and might display a different temporal relationship, as discussed below, to the time of emergence of the respective tumoral cells. The somatically mutated status of the HBL-2 and HBL-3 mAb VH segments was suggested by the monomorphism of the human VH4-21 gene, the extension of the nucleotide differences to the, in general, highly conserved JH segment; and it was formally proved in HBL-3 mAb. Positive selection by Ag of the R mutations in the HBL-2 and HBL-3 mAb VH segments was suggested by the differential R:S mutation ratios in the CDRs and FRs (HBL-2 mAb, 5.0 and 1.1, respectively; HBL-3 mAb, 2.2 and 0.3, respectively) but not substantiated by appropriate statistical analysis according to the binomial distribution model.(ABSTRACT TRUNCATED AT 400 WORDS)