Late simian virus 40 (SV40) mRNA contains eight different cap structures which we have previously identified and mapped on the viral genome. As reported here, 5'-cap heterogeneity is a common feature to both the early and the late SV40 mRNA's. methyl-3H-labeled viral mRNA was purified from cells infected at 41 degrees C with SV40 mutant tsA209. Three different cap cores were identified: m7GpppGm, m7GpppCm, and m7GpppAm. An average of three to four m6A residues per mRNA molecule was found. RNase T2-resistant 32P-labeled early caps from tsA209-infected cells isolated and characterized. Six distinct cap I structures were identified: m7GpppCmpU (30%), m7GpppGmpC (24%), m7GpppAmpG (18%), m7GpppGmpU (13%), m7GpppGmpG (12%), and m7GpppAmpU (3%). A similar 5'-end heterogeneity was observed in early SV40 mRNA from BSC-1 cells infected with wild-type SV40 strain 777 in the presence of cytosine arabinoside and in the SV40 UV-transformed permissive line C-6. Five of these capped dinucleotides are complementary to DNA sequences at 0.66 map unit in a region previously identified by the primer extension method (Reddy et al., J. Virol. 30:279-296, 1979; Thompson et al., J. Virol. 31:437-438, 1979) as the 5' end of the early message. DNA sequences upstream from this region contain the TATTTAT (Hogness-Goldberg box), which is missing from upstream of the 5'-cap sites of late SV40 mRNA. Thus, 5'-end heterogeneity is not necessarily related to the presence or the absence of this putative transcriptional "initiation signal." When the possibility that SV40 5' caps represent transcriptional initiation sites is considered, the data also suggest that, on SV40 DNA, eucaryotic RNA polymerase II initiates transcription at multiple nucleotide sequences, including pyrimidines.