We have previously shown that Sindbis virus RNA polymerase requires an N-terminal aromatic amino acid or histidine for wild-type or pseudo-wild-type function; mutant viruses with a nonaromatic amino acid at the N terminus of the polymerase, but which are otherwise wild type, are unable to produce progeny viruses and will not form a plaque at any temperature tested. We now show that such mutant polymerases can function to produce progeny virus sufficient to form plaques at both 30 and 40 degrees C upon addition of AU, AUA, or AUU to the 5' terminus of the genomic RNA or upon substitution of A for U as the third nucleotide of the genome. These results are consistent with the hypothesis that (i) 3'-UA-5' is required at the 3' terminus of the minus-strand RNA for initiation of plus-strand genomic RNA synthesis; (ii) in the wild-type virus this sequence is present in a secondary structure that can be opened by the wild-type polymerase but not by the mutant polymerase; (iii) the addition of AU, AUA, or AUU to the 5' end of the genomic RNA provides unpaired 3'-UA-5' at the 3' end of the minus strand that can be utilized by the mutant polymerase, and similarly, the effect of the U3A mutation is to destabilize the secondary structure, freeing 3'-terminal UA; and (iv) the N terminus of nsP4 may directly interact with the 3' terminus of the minus-strand RNA for the initiation of the plus-strand genomic RNA synthesis. This hypothesis is discussed in light of our present results as well as of previous studies of alphavirus RNAs, including defective interfering RNAs.