To identify the minimal structures that may be important for the creation of a synthetic and/or recombinant vaccine against whooping cough, human T cell clones were obtained against Bordetella antigens. Cloned peripheral blood T lymphocytes from an immune donor were grown in IL-2 and tested for proliferation in response to inactivated Bordetella species (B. pertussis, B. parapertussis, and B. bronchiseptica) and mutants deficient for the expression of virulence-associated antigens. All the T cell clones obtained were CD4+8- and recognized specifically the Bordetella antigens when presented by autologous B cells. On the basis of the responsiveness to the whole inactivated bacteria, it was possible to cluster the 12 clones obtained into four groups with the following specificity: (1) filamentous hemagglutinin (FHA); (2) B. pertussis-specific antigens; (3) virulence-associated Bordetella-specific antigens; and (4) nonvirulence-associated Bordetella-specific antigens. Using two new B. pertussis deletion mutants, clone 6 (representative of cluster 1) was found to recognize the COOH terminus of FHA. Furthermore, three out of four clones of cluster 3 were specifically stimulated by the soluble 69-kD protein from the outer membrane of B. pertussis. Surprisingly, none of the twelve clones obtained by stimulation in vitro with whole inactivated bacteria recognized pertussis toxin (PT), which is believed to be the most important protein to be included in an acellular vaccine. However, when a new generation of clones was obtained using soluble PT as the in vitro stimulus, it was observed that 11 clones of this group recognized this antigen. Thus, PT does not seem to be the most representative antigen on the whole inactivated bacteria, although T cell memory against PT exists in a donor who had the disease several years ago.