We have analysed the cDNA coding for the envelope glycoprotein (E1) gene and the terminal non-translated regions (NTRs) of the avirulent Semliki Forest virus (SFV) A774 (A7) variant. The E1 gene exhibited 98.5% identify to the SFV prototype strain L10 (WT) sequence at the nucleotide level. Of the 34 single base substitutions, six led to a change in the deduced amino acid sequence. The 3' NTR of A7 consisted of a 101 nucleotide sequence, not found in WT, followed by five tandemly arranged sequence motifs, two of which were truncated forms of the others. One full-length and one truncated repeat are found at the 3' NTR of WT. The repeats of A7 were followed by a non-repeating sequence, very similar to the equivalent region in WT. Owing to the unique sequence motif and the tandem repeats, the 3' NTR of A7 is 334 nucleotides longer than that of WT. Each of the repeats had an internal 12 nucleotide motif complementary to a conserved sequence in the 5'-terminal non-structural protein 1-encoding region, thought to be important in alphavirus RNA replication. In the 5' NTR, three point mutations were found. The conserved sequence binding to the repeated 3' motifs was identical in A7 and WT.