A nucleotide fragment encoding amino acids 29 to 94 in the E-protein of 28 dengue-2 isolates of diverse geographic and host origins was examined by direct sequencing of a polymerase chain reaction (PCR)-amplified product, and compared to six previously published sequences. Nucleotide divergence ranged from 0 to 19.8% corresponding to a maximum of 9% divergence in the amino acid sequence. Taking a divergence of 6% between the nucleotide sequence as a cut off for genotype classification, six groups have been established. Southeast Asian and the Jamaican 1983 genotypes show a high rate of similarity (> 95.2%). Our results suggest that virus of this group is now circulating as the dominant topotype in Brazil (1990) and in French Guyana (1986-1991). African strains fall into two groups, one endemic group (1970-1990) and one epidemic group (1986-1987). The three other groups correspond to viruses from Sri Lanka (1982) and the Seychelles (1977), from Puerto Rico (1973) and from Tahiti (1975). Our approach appears to be valuable characterizing dengue isolates, easily and rapidly.