An experimental scheme was developed for direct sequence analysis of Puumala virus-containing specimens from wild rodents (Clethrionomys glareolus). Total RNA isolated from rodent lung tissues was reverse-transcribed in the presence of a universal 11 nucleotide primer complementary to all three viral RNA segments followed by amplification in a PCR with gene-specific primers. A full-length PCR product of approximately 1800 bp from the S segment encoding the viral nucleoprotein and a product of approximately 900 bp from the M segment (encoding the C-terminal two-thirds of the G2 protein and including the 3' non-coding region) of Puumala virus (from C. glareolus trapped in Udmurtia) were prepared and sequenced. No pronounced differences to Vero cell-grown viruses were seen. The Udmurtia/894Cg/91 strain was more closely related to the Bashkiria/CG18-20/84 strain than to the Finnish prototype strain of Puumala virus, Sotkamo/V-2969/81. Thus there is a correlation with the geographic origin of the three strains. The results indicate the occurrence of genetic drift and different selection pressures leading to (i) clustering of mutations, (ii) a lower frequency of nucleotide substitutions in the coding than in the 3' noncoding regions and (iii) a higher frequency of amino acid substitutions in G2 than in the N protein.