Fresh isolates of Actinobacillus actinomycetemcomitans (Aa) bind avidly to surfaces in vitro, but existing in vivo studies of the adherence of Aa are limited. This study had two goals: (1) to compare the oral colonization of two isogenic strains of Aa-CU1010, a clinical isolate that expresses the adherent phenotype, and CU1012, a minimally adherent laboratory variant-and (2) to check for phenotypic reversion of these strains in a clinical setting. Rifampicin-resistant strains, developed for tracking in Sprague-Dawley rats, were tested in vitro to determine their stability and binding. In study 1, after antibiotic suppression, six rats (group I) received CU1010 in their feed. The eight rats in group II received CU1012 in their feed and four were supplemented by oral swabbing and four by gastric gavage. Group III consisted of three sham-inoculated controls. All rats were inoculated for 4 days. Microbiological data were collected at 1, 4 and 8 weeks after inoculation. Supporting data were supplied by antibody titres and clinical measures of alveolar bone loss. Study 2 consisted of six rats in each of three groups as above, but tagged strains of Aa were delivered by food alone. At all time-points in both studies, Aa was absent before inoculation and controls had no Aa or antibody to Aa. In study 1, all six rats in group I yielded positive cultures for Aa at 8 weeks. In group II, five of eight had positive cultures for Aa at 1 week, two of eight at 4 weeks and none had Aa at 8 weeks (P < or =0.001). All six rats in group I had serum anti-Aa titres compared to group II, where titres were seen in four of eight rats (P < or =0.015). In vitro data paralleled those found in vivo. No phenotypic reversion of either strain was seen in vivo. In study 2, four of six rats in group I showed Aa and had titres to Aa, while no other animals showed Aa at any time. The model provides convincing evidence that, unlike laboratory variants, clinical isolates colonize, persist and integrate into an already established, albeit reduced, econiche.