The hypoxanthine-guanine phosphoribosyl transferase (hprt) locus in 6-thioguanine (TG) resistant T-lymphocytes is a useful target for the study of somatic in vivo mutagenesis, since it provides information about a broad spectrum of mutation. Mutations in the hprt coding region were studied in 124 TG-resistant T-cell clones from 38 healthy, non-smoking male donors from a previously studied population of bus maintenance workers, fine-mechanics and laboratory personnel. Their mean age was 43 years (range 23-64) and their hprt mutant frequency was 9.3 +/- 5.2 x 10(-6) (mean +/- SD, range 1.4-22.6 x 10(-6)). Sequence analysis of hprt cDNA identified 115 unique mutations; 76% were simple base substitutions, 10% were +/-1 bp frameshifts, and 10% were small deletions within exons (3-52 bp). In addition, two tandem base substitutions and one complex mutation were observed. Simple base substitutions were observed at 55 (20%) of 281 sites known to be mutable in the hprt coding sequence. The distribution of these mutations was significantly different than would be expected based upon a Poisson distribution (P < 0.0001), suggesting the existence of 'hotspots'. All of the 87 simple base substitutions occurred at known mutable sites, but eight were substitutions of a kind that have not previously been reported at these sites. The most frequently mutated sites were cDNA positions 197 and 146, with six and five independent mutations respectively. Four mutations were observed at position 131, and three each at positions 143, 208, 508 and 617. Transitions (52%) were slightly more frequent than tranversions (48%), and mutations at GC base pairs (56%) more common than mutations at AT base pairs (44%). GC > AT was the most common type of base pair substitution (37%). The majority of the mutations at GC base pairs (78%) occurred at sites with G in the non-transcribed strand. All but one of eight mutations at CpG-sites were of the kind expected from deamination of methylated cytosine. Deletion of a single base pair (-1 frameshift) was three times more frequent than insertion of a single bp (+1 frameshift). Almost half (6/13) of the small (3-52 bp) deletions within the coding sequence clustered in the 5' end of exon 2. Short repeats and other sequence motifs that have been associated with replication error were found in the flanking regions of most of the frameshifts and small deletions. However, several differences in the local sequence context between +/-1 frameshift and deletion mutations were also noticed. The present results identify positions 197, 146 and possibly 131 as hotspots for base substitution mutations, and confirm previously reported hotspots at positions 197, 508 and 617. In addition, the earlier notion of a deletion hotspot in the 5'end of exon 2 was confirmed. The observations of these mutational cluster regions in different human populations suggest that they are due to endogeneous mechanisms of mutagenesis, or to ubiquitous environmental influences. The emerging background spectrum of somatic in vivo mutation in the human hprt gene provides a useful basis for comparisons with radiation or chemically induced mutational spectra, as well as with gene mutations in human tumors.