The rotavirus nonstructural glycoprotein NSP4 may represent the first identified viral enterotoxin. We have sequenced reverse transcription-polymerase chain reaction (RT-PCR)-generated fragments of 16 NSP4 genes of human rotavirus (HRV) strains from six different countries, representing seven different G and P type combinations. Based on the amount of sequence divergence between these and 11 previously sequenced NSP4 genes of human and animal rotaviruses, three distinct genetic groups could be recognized. Most strains within a group were closely related to each other at the nucleotide (nt) and amino acid (aa) levels (usually <10% divergence) but more distantly related (maximum 30.0% nt divergence and 24.7% aa divergence) to members of the other groups. Intergroup variation occurred in two highly variable regions of NSP4 (aa 16-34 and aa 131-148). The NSP4 "toxic peptide" (aa 114-135) exhibited aa variation at its carboxy terminus both within and between genetic groups. The largest group (genetic group II) contained HRV strains of subgroup II specificity (including genotypes P[8]G1, P[8]G3, P[6]G3, and P[8]G5 and serotype P8[11]G9), and the smaller group (genetic group I) contained HRV strains of subgroup I specificity (genotype P[4]G2). The NSP4 sequence of the rhesus rotavirus vaccine strain was distinct from all other strains and formed the third group (genetic group III). The NSP4 genes of animal rotaviruses UK, NCDV, and SA11 (genetic group I) and YM (genetic group II) and two possible human-animal rotavirus reassortant strains, Brazilian P[8]G5 and Indian P[11]G9 (genetic group II), could also be classified into one of these groups, suggesting a close evolutionary relationship between human and animal NSP4 genes. These results will facilitate studies of the host immune response to NSP4, which may be relevant to future HRV vaccine design.