All genospecies display a probable Asian origin
For both B. afzelii and B. garinii the oldest node separates a clade containing only Asian isolates from a clade containing isolates from both continents (Figure 2A,C) suggesting that the Asian population is ancestral for both genospecies. Borrelia bavariensis displays a deep branching between the two continents, and European isolates are characterized by a low divergence and almost clonal expansion as previously described (Becker et al., 2020; Gatzmann et al., 2015) (Figure 2B). Our original analysis did not include RussianB. afzelii isolates. A single Russian B. afzelii isolate exists with a full chromosome in GenBank, Tom3107 (Accession Number: NZ_CP009212.1). We re-ran the phylogeny utilizing all B. afzeliichromosome sequences including Tom3107 and PBi as an outgroup to root the tree (see Suppl. Met.). Tom3107 was basal to the monophyletic European B. afzelii clade (Figure S1) suggesting a stepwise colonization from far-east Asia through Russia into Europe, which was not observed in the other two genospecies.
Higher genetic diversity (π (Nei, 1987)) was found in AsianB. bavariensis and B. garinii in comparison to their European counterparts (Table 1). Genetic diversity was similar between Asian and European B. afzelii isolates (Table 1). In all cases, the Borrelia populations showed negative Tajima‘s D(Tajima, 1989) values (Table 1) as expected for bacteria due to the influence of population expansion (Gatzmann et al., 2015; Tajima, 1989). The European samples always showed more negative values (Table 2), suggesting a more recent expansion into Europe. Borrelia bavariensis displayed the largest difference in Tajima’s D and also had the largest absolute divergence value (Dxy (R. R. Hudson et al., 1992)) in comparison to the other two genospecies hinting that B. bavariensisbranching is potentially the oldest and that B. afzelii is the youngest with the lowest absolute divergence and difference in Tajima’sD (Table 1).
Our dataset, as with many others, includes non-randomly sampled isolates which could lead to biased estimates of population level statistics (Nei, 1987). As our data set includes randomly sampled isolates as well (see Text S1 and Tables S3, S4, S5) we were able to test for potential sampling biases. Interestingly, we did not observe strong bias in any of these statistics (π , FST ,DXY , Tajima’s D ) when calculated on datasets containing random and non-random samples (Text S2).