We analyzed 178 B. henselae isolates/strains from various sources (Table 1): 156 (88%) feline isolates/strains, 21 (11%) from diseased humans, and 1 isolate from a sick dog. The number of alleles varied from 7 (BHV-E) to 22 (BHV-B). Most of the European isolates (all but 1 of feline origin) (2,4,6) and of the American isolates/strains (North Carolina and California) (5,14), of which 85% were of feline origin, belonged to genotype II (89% and 64.6%, respectively). The Asian isolates (all but 1 of feline origin) (7–9) and the Australasian isolates (60% of human origin) (12), mainly belonged to genotype I (89.6% and 65%, respectively).

Ninety-nine different MLVA profiles were observed (Table 1), corresponding to an average number of isolates per profile of 1.8 (Table 2). Sixty-nine of these profiles were found in only 1 isolate or strain (67%), and 30 were observed in >1 isolate. Among these, none was shared by genotype I and genotype II isolates. Diversity index (DI) was 0.98 (Table 1). Diversity was observed in both genotypes because genotype-specific DIs were almost identical (Table 1).

MLVA profiles appeared location-specific because only 4 (13%) of the 30 profiles observed in >1 isolate/strain were present in >1 continent (Table 2). Within continents, no marked dominance of a given profile was observed, and continent-specific DIs were similar (Table 1).

Of the 99 B. henselae profiles, 12 were obtained from the 21 human isolates/strains and 1 from the dog, whereas 92 profiles were obtained from the 156 feline isolates. Five profiles were common to 5 human and 11 feline isolates. Among the 30 profiles observed in >2 isolates, 23 were observed only in feline isolates (Table 2). The proportion of genotype I profiles was significantly higher in human-specific profiles than in cat-specific profiles (p = 0.01, by Fisher test).

For BHV-A, only 2 alleles (14 and 15 copies) were found in isolates from humans, whereas all 8 identified alleles were observed in cat isolates. The number of repeats differed significantly between sick humans and healthy cats (p = 0.02, by Fisher test).

Relationships between the 99 MLVA profiles were analyzed by unweighted pair group method with arithmatic mean (UPGMA), using a categorical distance, with a B. koehlerae isolate used as an outgroup. To take into account that UPGMA is sensitive to taxa entry order, we computed the majority-rule consensus tree of 500 dendrograms built with random taxa entry order. MLVA profiles were grouped into 2 main groups named A and B (Appendix Figure). Group A (26 profiles), was exclusively constituted by genotype II feline isolates. Group B (73 profiles), to which all human isolates belonged, further divided in 2 subgroups, Ba and Bb. Subgroup Ba (38 profiles) was exclusively composed of genotype I isolates, including the reference strain Houston I and a homogenous subgroup, Ba1, containing 84% of the Asian isolates. Finally, 83% of subgroup Bb isolates belonged to genotype II (29/35 profiles).

The utility of MLVA for molecular epidemiologic analysis of clusters was tested using isolates from California cats and their owners (14). Five human–cat groups of B. henselae isolates were analyzed. For 1 cat-human pair of isolates, which belonged, respectively, to genotype II and genotype I, major profile differences were observed, as expected. The 4 other cat-human groups, which possessed the same genotype, also had the same MLVA profile with the 5 tested BHV, as well as with the 6 additional BHV (F–K) and variant alleles for BHV-A and/or B (6). Sequencing confirmed these results.

Our results confirm that VNTRs are excellent molecular markers for confirming or excluding the responsibility of a given cat in the transmission of B. henselae to a human. In California, the profile identity observed within 4 clusters further supports the hypothesis that all these humans acquired infection from their respective domestic cat contacts.

MLVA enabled a clear separation between genotypes I and II, because no profile was shared between both genotypes. The dendrogram showed a high level of discrimination between 16S rDNA genotypes in the B. henselae population tested. Interestingly, the groups and subgroups delineated by MLVA were the same as those defined by MLST, a standard method for phylogenetic analysis (12). The same was observed with MST (13). The isolates of the subgroup Bb appeared divergent and distant from each other and from subgroup Ba that contains almost all genotype I profiles (98%). Moreover and despite possible clustering for some of the isolates, none of the 21 human isolates was present in group A. Interestingly, as for most of the human patients, the isolate obtained from the ill dog also belonged to genotype I.

These observations suggest that all genotype I isolates could be phylogenetically derived from genotype II isolates located in group B but not in group A, as already suggested using MLST (15). This observation could mean that genotype II isolates belonging to group B are closer to genotype I isolates than to genotype II isolates belonging to group A; it also raises an important clinical question: Are feline genotype II isolates belonging to group A nonpathogenic for humans? Genotype I isolates could represent the most pathogenic isolates for humans within a group of potentially zoonotic isolates, all belonging to group B and could represent an ultimate evolutionary step toward human infection. Additionally, within group B, the differences in the number of BHV-A repeat units observed between isolates from patients (humans, dog) versus cat isolates suggest that this specific VNTR could constitute a marker for the ability to cross the species barrier from reservoir cats to susceptible species, independent of the 16S rDNA genotype.