The first step in tumor formation by Agrobacterium tumefaciens is the site-specific binding of the bacteria to plant host cells. Transposon mutants of the bacteria which fail to attach to carrot suspension culture cells were isolated. These mutants showed no significant attachment to carrot cells with either microscopic or viable cell count assays of bacterial binding. The nonattaching mutants were all avirulent. When revertants of the mutants were obtained by enriching for bacteria which do bind to carrot cells, the bacteria were found to have regained the ability to bind to carrot cells and virulence simultaneously. These results suggest that the ability of the bacteria to bind to plant cells is required for virulence. Like the parent strain, all of the nonattaching mutants synthesized cellulose, but unlike the parent strain, they failed to aggregate carrot suspension culture cells. The transposon Tn5, which was used to obtain the mutants, was located on a 12-kilobase EcoRI fragment of the bacterial chromosomal DNA in all of the nonattaching mutants from strain C58. That the mutant phenotype was due to the Tn5 insertion was shown by cloning the Tn5-containing DNA fragment from the mutant bacteria and using it to replace the wild-type fragment in the parent strain by marker exchange. The resulting bacteria had the same mutant phenotype as the original Tn5 mutants; they did not attach to carrot cells, they did not cause the aggregation of carrot cells, and they were avirulent. No difference was seen between the parent strain and the nonattaching mutants in hydrophobicity, motility, flagella, fimbriae, beta-2-glucan content, size of lipopolysaccharide, or ability of the lipopolysaccharide to inhibit bacterial attachment to tissue culture cells. Differences were seen between the parent strain and the nonattaching mutants in the polypeptides removed from the bacteria during the preparation of spheroplasts. Three of the mutants were lacking a polypeptide of about 34 kilodaltons (kDa). One mutant was lacking the 34-kDa polypeptide and another polypeptide of about 38 kDa. The fifth mutant was lacking a polypeptide slightly smaller than the 34-kDa polypeptide missing in the other four mutants. These missing polypeptides all reappeared in the revertants of the mutants. Thus, bacterial binding to plant cells appears to require the presence of these polypeptides.