A core genome MLST scheme for Borrelia burgdorferi sensu lato improves insights into the evolutionary history of the species complex

Summary

Multi-locus sequence typing (MLST) based on eight genes has become the method of choice for Borrelia typing and is extensively used for population studies. Whole-genome sequencing enables studies to scale up to genomic levels but necessitates extended schemes. We have developed a 639-loci core genome MLST (cgMLST) scheme for Borrelia burgdorferi sensu lato (s.l.) that enables unambiguous genotyping and improves the robustness of phylogenies and lineage resolution within species. Notably, all inner nodes of the cgMLST phylogenies had consistently high statistical support. Analyses of the genetically homogeneous European B. bavariensis population support the notion that cgMLST provides high discriminatory power even for closely related isolates. While isolates differed maximally in one MLST locus, there were up to 179 cgMLST loci differences. Thus, the developed cgMLST scheme for B. burgdorferi s.l. resolves lineages at a finer resolution than MLST and improves insights into the evolutionary history of the species complex.

Graphical abstract

Highlights

• •A core genome multi-locus sequence typing scheme for Borrelia burgdorferi sensu lato
• •The scheme allows unambiguous genotyping and enables robust, high-resolution phylogenies
• •We demonstrate high discriminatory power within species and closely related isolates
• •The cgMLST scheme is publicly available on the Borrelia PubMLST website

Motivation

Multi-locus sequence typing (MLST) analyses based on eight chromosomal housekeeping genes (clpA, clpX, nifS, pepX, pyrG, recG, rplB, uvrA) are extensively used for Borrelia typing, population genetics, and evolutionary analyses. However, we still encounter limitations with regard to statistical support of inner nodes in phylogenies and the resolution of lineages within species. As whole-genome sequencing has become available, MLST can be extended to the genomic scale, analyzing a higher number of loci and grouping them in an extended scheme. We aimed to gain higher reliability, robustness, and resolution and have developed a 639-loci core genome MLST (cgMLST) scheme for B. burgdorferi sensu lato (s.l.) that is applicable to whole-genome data. The schemes are available on the Borrelia PubMLST database (https://pubmlst.org/organisms/borrelia-spp).

Hepner et al. develop a 639-loci core genome multi-locus sequence typing (cgMLST) scheme for Borrelia burgdorferi sensu lato that extends the traditional 8-loci MLST to a genomic scale. The cgMLST scheme gains higher reliability, robustness, and resolution, enabling improved insights into the evolutionary history of the species complex.

Introduction

Lyme borreliosis is the most prevalent tick-borne disease in the Northern Hemisphere, including North America and the temperate regions of Eurasia. The causative agents of the disease are spirochetes of the Borrelia burgdorferi sensu lato (s.l.) species complex.^1,2,3 The complex contains 22 species, including the six human pathogenic species that can cause Lyme borreliosis: Borrelia afzelii, B. bavariensis, B. burgdorferi sensu stricto (s.s.), B. garinii, B. mayonii, and B. spielmanii.^1,4,5,6 The bacterium is maintained in transmission cycles between tick vectors of the Ixodes ricinus-persulcatus species complex and vertebrate reservoir hosts.^1,6,7

The genome of B. burgdorferi s.l. is relatively small (∼1.5 Mb) but remarkable for bacteria in that it consists of a linear chromosome and numerous linear and circular plasmids.^8,9,10 Sequence typing has become the state-of-the-art method for bacterial characterization and for the analyses of ecological and evolutionary processes that influence population structure and dynamics. Chromosomal as well as plasmid-located loci have been used for B. burgdorferi s.l. typing, but early analyses were mostly based on single-locus analyses (16S rRNA gene [rrs], flaB, ospA, ospC, rrs-rrlA intergenic spacer [IGS]).^11,12

In 1998, a novel and portable bacterial typing technique was introduced: multi-locus sequence typing (MLST).¹³ This approach used nucleotide sequences of multiple core genes, and it was suspected that MLST can be applied to almost all bacterial species. An advantage of MLST is that the method can be based on PCR amplification of the appropriate loci; thus, in the case of vector-borne bacteria, it can be applied to environmental samples (vector, host, or patient) and does not necessarily rely on cultured material.¹² Today, this approach has become the method of choice for many organisms and is extensively used in studies of population biology and public health surveillance of bacterial pathogens.^14,15,16,17 The sequence data are shared worldwide via MLST databases such as the PubMLST database (https://pubmlst.org/) or the Institute Pasteur MLST databases (https://bigsdb.pasteur.fr/).^17,18

The Borrelia MLST scheme is based on eight chromosomal housekeeping genes (clpA, clpX, nifS, pepX, pyrG, recG, rplB, uvrA).^12,19 The major advantage of analyzing multiple loci in comparison to single loci is that a skewed evolutionary picture can be mitigated. Additionally, it allows the detection of, and can buffer against, recombination. Furthermore, the use of multiple loci leads to a higher resolution of lineages within species and improved epidemiological delineations.^12,15 Since 2015, the PubMLST.org website has been the home for Borrelia MLST/MLSA data²⁰ and currently contains over 78,000 unique allele sequence definitions and information for more than 3,900 isolates (https://pubmlst.org/organisms/borrelia-spp, June 26, 2024).

Even though MLST was a significant improvement over single-locus analyses, we still encounter some limitations—mainly due to limited statistical support of inner nodes in phylogenies, which may become more pronounced if some loci fail to produce typing results.^15,21,22

Rapid and high-throughput sequencing methods by next-generation sequencing (NGS) technologies have opened the door for accurate whole-genome data, opening up new possibilities in terms of typing methods.¹⁵ To take full advantage of this, new platforms for data storage and bioinformatics analyses as well as extended typing schemes are required. The Bacterial Isolate Genome Sequence Database (BIGSdb) is incorporated in the PubMLST.org website and allows the hosting of all levels of sequence data (single sequences up to whole-genome data) and comparative genome analyses.^17,18,20 This allows extending the principle of MLST to a higher number of loci that can be grouped into an extended typing scheme, known as core genome MLST (cgMLST).^15,18 It has been shown in other bacteria that cgMLST can lead to insights in the emergence and evolution of pathogens and enables standardized molecular surveillance, including a replicable typing method (for example, Neisseria,^23,24 Listeria monocytogenes,²⁵ Mycobacterium tuberculosis,²⁶ Staphylococcus,^27,28 Campylobacter,²⁹ Salmonella,^30,31,32 and Klebsiella pneumonia^33,34). To extend the advantages of MLST to a genomic scale for Borrelia, we developed a 639-loci cgMLST scheme for B. burgdorferi s.l. that enables unambiguous genotyping and reliable phylogenetic analysis with high resolution. The scheme is publicly accessible through the Borrelia PubMLST.org website.

Results

The cgMLST scheme for B. burgdorferi s.l. contains 639 core loci

We have used 174 high-quality B. burgdorferi s.l. genomes of unique strains and 17 species (Tables 1 and S1; see STAR Methods for details) and 815 chromosomal coding sequences (CDS) of B. burgdorferi s.s. B31 from GenBank (GenBank: NC_001318.1/AE000783.1, November 15, 2023) to develop the cgMLST scheme.

Table 1.

Overview of number of isolates per genospecies that was used to develop and validate the cgMLST scheme

Species	No. of isolates in the development genome set	No. of isolates in the validation genome set	No. of isolates in total
B. afzelii	6	31	37
B. americana	1	1	2
B. bavariensis	27	22	49
B. bissettiae	2	1	3
B. burgdorferi s.s.	86	34	120
B. californiensis	1	–	1
B. carolinensis	1	–	1
B. chilensis	1	–	1
B. garinii	37	29	66
B. japonica	1	–	1
B. kurtenbachii	1	–	1
B. maritima	1	–	1
B. mayonii	2	–	2
B. spielmanii	1	–	1
B. turdi	3	–	3
B. valaisiana	2	2	4
B. yangtzensis	1	–	1
Total	174	120	294

See also Table S1.

The 174 genomes were analyzed for core loci that are present and designated in ≥95% of the genomes (with 97% identity over 99% alignment length). We identified 639 chromosomal core loci and included them in the Borrelia cgMLST scheme (see Table S2 for description and length information). A schematic overview of the cgMLST development is shown in Figure 1.

Figure 1.

Schematic overview of the B. burgdorferi s.l. cgMLST scheme development

174 B. burgdorferi s.l. genomes were scanned for the presence of 815 chromosomal CDSs from B31 reference, resulting in 639 cgMLST loci. An additional 120 B. burgdorferi s.l. genomes were added to validate the scheme, resulting in 294 genomes.

cgMLST phylogeny has high statistical support and is not affected by recombination

Comparison of the unrooted maximum likelihood (ML) phylogenetic trees (n = 174 genomes) based on the 8-loci MLST vs. the developed 639-loci cgMLST scheme (Figures 2A and 2B, respectively) clearly showed that in both trees, samples clustered according to species and formed two major clades (an “American” and a “Eurasian” clade), as previously described.^35,36 A notable difference between MLST and cgMLST phylogenies was the clustering of B. chilensis and B. maritima.

Figure 2.

Unrooted ML trees of 174 and 294 genomes

Trees were generated with IQ-TREE 2.2.2.7³⁷ (A, B, and D) or RAxML v.8.2.12³⁸ (C). Isolates are labeled according to genospecies. The scale bars for (A), (B), and (D) denote substitutions per site, whereas the scale bar for (C) refers to SNPs.

(A) ML tree is based on 8-loci MLST of 174 isolates. Bootstrap values of internal nodes are shown in colored points (green: 100, yellow: <100 and ≥95, red: <95). The enlarged inset show details of bootstrap values <95.

(B) ML tree based on 639-loci cgMLST of 174 isolates. All internal nodes are colored green, indicating bootstrap values of 100.

(C) ML tree based on SNPs present in non-recombinant cgMLST regions of 174 isolates. No difference was found compared to the tree shown in (B).

(D) ML tree is based on 639-loci cgMLST of 294 isolates. The additional 120 samples are marked with red asterisks. All internal nodes are highly supported with bootstrap values of 100 (colored green).

In the MLST ML tree (Figure 2A) the American clade encompassed B. americana, B. bissettiae, B. burgdorferi s.s., B. californiensis, B. carolinensis, B. kurtenbachii, and B. mayonii. The Eurasian clade included B. afzelii, B. bavariensis, B. garinii, B. japonica, B. spielmanii, B. turdi, B. valaisiana, and B. yangtzensis. In this tree, the species B. chilensis and B. maritima (labeled in red in Figure 2A) clustered within the Eurasian clade as sister clades to B. afzelii and B. spielmanii. All of the internal nodes of the American clade were well supported (bootstrap values ≥ 95; colored yellow and green in Figure 2A). While some nodes of the Eurasian clade also showed high bootstrap values of 100 (colored green in Figure 2A), there were several internal nodes with lower bootstrap support (bootstrap values ranged from 67 to 87, colored red in Figure 2A, see the enlarged inset for details).

The phylogenetic tree based on the cgMLST scheme (Figure 2B) also showed clustering according to species, as well as the division into an American or Eurasian clade. However, the two strains belonging to the species B. chilensis and B. maritima (labeled in red in Figure 2B) did not cluster within the Eurasian clade; instead, they clustered between the Eurasian and American clades, appearing as sister clades to each other and to the Eurasian and American isolates, respectively. All internal nodes were well supported with bootstrap values of 100 (colored green in Figure 2B).

To analyze whether recombination events within cgMLST loci bias the phylogenetic reconstruction, we compared the cgMLST phylogeny to a phylogeny based on non-recombinant cgMLST regions. For this, the cgMLST alignment of the 174 development genomes was used to identify SNPs in non-recombinant regions. Based on these SNPs, an ML tree was generated (Figure 2C) and compared to the cgMLST ML tree (Figure 2B). Both trees showed the same clustering (according to species) and the same topology. The concordance of the cgMLST ML tree and the tree based on SNPs in non-recombinant regions of the cgMLST loci indicate that recombination does not affect the phylogeny.

To validate the scheme, a further 120 genomes were included, resulting in a total number of 294 genomes belonging to 17 B. burgdorferi s.l. species (see Tables 1 and S1 for detailed isolate information). The resulting cgMLST ML tree showed species-specific clustering in high resolution (high support of all internal nodes with bootstrap values of 100, colored green in Figure 2D; Figure 2D: unrooted ML tree; Figure 3: midpoint rooted circular ML tree). The cgMLST ML tree topology of 294 isolates (Figure 2D) is congruent with the topology of the cgMLST ML tree of the development genome set (174 isolates, Figure 2B), the division into an American or Eurasian clade, and B. chilensis and B. maritima clustering between the two major clades.

Figure 3.

Midpoint rooted circular ML tree based on 639-loci cgMLST using the genome set of 294 isolates

Tree was generated with IQ-TREE 2.2.2.7.³⁷ The scale bar refers to nucleotide substitutions per site. Samples belonging to the validation genome set are marked with red asterisks. The different colors of the first inner circle represent the genospecies, and the second circle represents the ST (based on the 8-loci MLST scheme). The symbols in the third circle represent the cgST (based on the 639-loci cgMLST scheme): an empty field represents a missing value, unfilled red squares indicate that two samples of the same species share the same cgST, and filled red squares represent unique cgST values. The different colors in the outer circle represent the different countries of origin. The branch of the European B. bavariensis isolates is highlighted in yellow.

cgMLST improves lineage resolution within species

The distance matrices give information about the allelic differences between the isolates. For species for which at least two isolates were included in the validation, the minimum and maximum allelic differences between isolates of the same species are listed in Table 2. For species with more than five isolates (B. afzelii, B. bavariensis, B. burgdorferi s.s., B. garinii), the allelic differences ranged from 0 to 8 based on the 8-loci MLST scheme and from 0 to 637 allelic differences based on the 639-loci cgMLST scheme.

Table 2.

Minimum and maximum ADs within isolates of the same genospecies based on the 8-loci MLST and 639-loci cgMLST schemes

Species	No. of isolates	MLST		cgMLST
		Min. allelic differences	Max. allelic differences	Min. allelic differences	Max. allelic differences
B. afzelii	37	0	8	0	603
B. burgdorferi s.s.	120	0	8	1	616
B. garinii	66	0	8	0	637
B. bavariensis
Total	49	0	8	0	637
Asian	30	0	8	3	635
European	19	0	1	0	179
B. valaisiana	4	0	7	331	591
B. bissettiae	3	6	8	492	594
B. turdi	3	4	8	483	492
B. americana	2	8	8	636	636
B. mayonii	2	1	1	5	5

B. bavariensis is additionally distinguished into isolates belonging to the Asian and European populations. ADs, allelic differences. See also Tables S4–S7.

Based on the MLST and cgMLST distances matrices, minimum spanning trees (MSTs) were generated and visualized in GrapeTree (MLST: Figures 4A and 4B, cgMLST: Figures 4C and 4D). The MLST MSTs show clustering according to species (Figure 4A) and several isolates with identical MLST allele profiles clustered together with 0 AD (show as pie charts in Figures 4A and 4B), resulting in identical sequence type (ST) assignments (Figure 4B). In the cgMLST MSTs, isolates also clustered according to species (Figure 4C), but only two B. afzelii, two B. bavariensis, and two B. garinii isolates had the same cgMLST allele profile, while the other isolates had unique profiles, resulting in the assignment of different core genome STs (cgSTs) (Figures 4D and 3).

Figure 4.

Minimum spanning trees of the 294 genomes.

The MSTs are based on the MLST (A and B) and cgMLST (C and D) typing schemes. The MSTs are colored according to species (A and C), ST (B), or cgST (D). Samples with missing STs (B) or cgSTs (D) are shown in white. The European B. bavariensis isolates are highlighted yellow.

To group the cgSTs into a core genome cluster, single-linkage clustering was applied using thresholds of 100, 50, 25, 10, and 5 allelic differences. The cg clusters are designated with identifier “Bb_cgc_” followed by the allelic difference thresholds used, e.g., Bb_cgc_100 for a threshold of 100 allelic differences. This resulted in 116 clusters of Bb_cgc_100 and up to 184 clusters of Bb_cgc_5. Table 3 shows the number of clusters depending on the applied allelic mismatch threshold.

Table 3.

Number of core genome clusters (Bb_cgc) resulting from various thresholds

Core genome cluster name	Allelic mismatch threshold	No. of clusters
Bb_cgc_100	100	116
Bb_cgc_50	50	140
Bb_cgc_25	25	154
Bb_cgc_10	10	169
Bb_cgc_5	5	184

Figures S2A–S2E show the MSTs colored according to the Bb_cgc clusters with the various thresholds (Figure S2A: Bb_cgc_100, Figure S2B: Bb_cgc_50, Figure S2C: Bb_cgc_25, Figure S2D: Bb_cgc_10, and Figure S2E: Bb_cgc_5).

Example: B. bavariensis European population

The species B. bavariensis is divided into two populations, a genetically heterogenic Asian population and a genetically homogeneous, almost clonal European population.^39,40 The clonal genetic characteristics of the European population are utilized to investigate how the new 639-loci cgMLST scheme impacts the differentiation of isolates of this closely related group.

The circular ML tree (Figure 3) shows the separation of B. bavariensis (colored red in the first inner circle of Figure 3) into an Asian and a European population (the branch of the European population is highlighted in yellow in Figure 3). The average nucleotide identity (ANI) values for the whole batch of B. bavariensis isolates ranged from 95.357 to 99.997, while values ranged from 96.431 to 99.997 for the heterogenic Asian and from 98.968 to 99.993 for the homogeneous European B. bavariensis isolates, respectively. ANI values for each pairwise comparison of the B. bavariensis isolates are shown in Table S3. The genetic characteristics can also be observed in the allelic differences of isolates within the two populations (Table 2). The Asian isolates (n = 30) show high variation in the MLST and cgMLST allelic differences, varying from 0 to 8 (MLST) and from 3 to 635 (cgMLST) (for details, see Tables S4 and S5). While the homogeneous European isolates (n = 19) differ by a maximum of one MLST allele giving rise to just two STs (ST84 and ST85), the cgMLST scheme led to a higher resolution with a maximum of 179 allelic differences (for details, see Tables S6 and S7). One isolate (B. bavariensis A104S [Borrelia PubMLST: ID3756]) was not assigned to a cgST (>2% of loci were missing), and the other 18 isolates were assigned to 17 different cgSTs (two isolates had an identical cgST: B. bavariensis PBN [Borrelia PubMLST: ID3759] and B. bavariensis PNi [Borrelia PubMLST: ID2712]) (Figures 3 and 4D), showing that cgMLST has higher discriminatory power than MLST. Various allelic mismatch thresholds were applied to group the cgSTs (Figure S2, European B. bavariensis isolates are highlighted in yellow). Using a threshold of 100 allelic differences (meaning that isolates with cgMLST profiles differing in 100 loci or fewer will be assigned to the same Bb_cgc_100 cluster), the 18 European B. bavariensis isolates with assigned cgSTs shared the same Bb_cgc_100 cluster. By decreasing the threshold of allelic differences, the 18 isolates are separated into 4 (Bb_cgc_50), 5 (Bb_cgc_25), 6 (Bb_cgc_10), or 9 clusters (Bb_cgc_5).

Availability of the B. burgdorferi s.l. cgMLST scheme

The developed and validated 639-loci cgMLST scheme for B. burgdorferi s.l. is publicly available on the Borrelia PubMLST.org website (https://pubmlst.org/organisms/borrelia-spp).¹⁷ Numerous analysis tools are listed in the STAR Methods.

Discussion

Accurate whole-genome sequences have become available for many organisms. In order to build on the success of MLST and at the same time make use of the power of whole-genome data, we have developed a cgMLST scheme for B. burgdorferi s.l. While the 8-loci MLST scheme already enables unambiguous Borrelia genotyping,¹² we demonstrate here that the developed 639-loci cgMLST scheme has the same capabilities but with higher reliability and more discriminatory power.

Phylogenetic analyses showed that 174 isolates clustered in both ML trees (8-loci MLST and 639-loci cgMLST) according to species and separated into an American (B. americana, B. bissettiae, B. burgdorferi s.s., B. californiensis, B. carolinensis, B. kurtenbachii, and B. mayonii) or Eurasian clade (B. afzelii, B. bavariensis, B. garinii, B. japonica, B. spielmanii, B. turdi, B. valaisiana, and B. yangtzensis) (with the caveat that B. bissettiae and B. burgdorferi s.s. also occur in Europe). Bootstraps values ranged from 67 to 100 in the MLST ML tree, and low bootstrap values (<95) are likely due to the limited number of analyzed loci. The developed cgMLST scheme used 639 core loci and provided a higher resolution and consistently well-supported internal nodes (bootstrap values of 100 for all inner nodes).

In previous studies based on the 8-loci MLST using 20 B. burgdorferi s.l. strains, B. maritima clustered within the Eurasian clade as a sister to B. afzelii and B. spielmanii,⁴¹ which is in accordance with the 8-loci MLST results of this study. However, another study analyzing 37 single-copy genes of 114 B. burgdorferi s.l. strains and one relapsing fever species was in accordance with the data presented here using the 639-loci cgMLST, where B. maritima was a sister clade of the American clade.³⁵ High bootstrap values of internal nodes of 100 in the cgMLST phylogenies support the proposition of B. maritima being a sister clade of American species. Thus, it can be assumed that the cgMLST phylogeny represents a robust topology.

As recombination events can impact and bias phylogenetic reconstructions, we compared the phylogeny of the cgMLST ML tree and the ML tree based on SNPs of non-recombinant regions of the cgMLST alignment. As both ML trees showed the same clustering and topology, we suggest that recombination did not affect the phylogeny of the cgMLST ML tree. Evidence for chromosomal recombination has been suggested in previous studies on B. burgdorferi s.s. (the recombination-to-mutation ratio was >1 to 3)^42,43; however, it did not affect the phylogenic analyses in those studies, tying in with our results presented here.^42,44

Using a dataset of 294 genomes of the B. burgdorferi s.l. complex, phylogenetic analyses confirmed species-specific clustering of the cgMLST ML tree with consistently highly supported nodes (bootstrap values of 100 for all internal nodes). Furthermore, no difference in topology of the cgMLST ML trees (174 and 294 isolates) was noticed, with B. chilensis and B. maritima clustering in both trees between the Eurasian and the American clade. These data are in accordance with the phylogenetic analyses based on 37 single-copy genes conducted by Margos et al.³⁵ These results underscore the robustness of the developed 639-loci cgMLST typing scheme.

The cgMLST scheme also impacts the discriminatory power within species as exemplified using the species B. bavariensis. This species has genetically unique characteristics, as the it is divided into a heterogeneous Asian population and a homogeneous European population. The genetic bottleneck in the European populations is supposed to be the result of a vector switch (from I. persulcatus to I. ricinus), which enabled the species to invade Europe.^39,40 The genetic clonality of the European B. bavariensis isolates provided a suitable model to investigate whether the cgMLST scheme may allow the differentiation of strains of this population. Indeed, the 639-loci cgMLST scheme separated 18 isolates with assigned cgSTs into 17 different cgSTs. As expected, these data show that the cgMLST scheme enables population genetic analyses at an even finer scale than MLST.

As most of the isolates were assigned to a unique cgST, various allelic mismatch thresholds (100, 50, 25, 10, and 5) were used to group the cgSTs. We observed that the European B. bavariensis were grouped in the identical Bb_cgc_100 cluster (threshold: 100 allelic differences) but separated into different clusters based on stricter thresholds. This indicates that the threshold of 100 allelic differences may be a biologically meaningful threshold to group Borrelia isolates. It needs to be noted that the cg clusters are assigned group numbers that are not stable and therefore should not be used for nomenclature. Due to the single-linkage clustering, adding new genomes can lead to cg cluster merging, which in turn may lead to new group numbers.

In conclusion, we have extended the advantages of MLST to a genomic scale and have developed a 639-loci cgMLST scheme for B. burgdorferi s.l. that enables an improved insight into the evolutionary history of the species complex. Apart from unambiguous genotyping, the scheme allows fine-scale population structure analyses in high resolution and reliability even for genetically clonal samples as the European B. bavariensis population.

Limitations of the study

The 639-loci cgMLST scheme was established for whole-genome data and therefore requires cultivation of the bacteria. While the generation of sequence data is getting cheaper, faster, and, most importantly, more accurate, cultivating Borrelia is still fastidious, and thus MLST will still have a place in investigating environmental samples. The MLST scheme as well as the newly developed cgMLST scheme include conserved chromosomal genes and do not include plasmid genes, which may be under greater selective pressure than conserved chromosomal loci. Therefore, to address the question of the evolutionary relationship of Borrelia, conserved chromosomal loci are better suited as plasmid-encoded genes. Thus, all the analyses of this manuscript are only based on chromosomal loci and do not include any plasmid analyses. cgMLST enables improved insights in the evolutionary history of the species and strains and is a prerequisite to understand the biological meaning and/or clinical relevance of highly variable Borrelia plasmids, their presence, and their gene content. So far, the cgMLST scheme has been developed and validated for species belonging to the B. burgdorferi s.l. complex, and further studies are required to test its suitability for relapsing fever- and reptile-associated Borrelia species. This will require the adoption or expansion of the scheme as necessary.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Sabrina Hepner (sabrina.hepner@lgl.bayern.de).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •This paper analyzes existing, publicly available genome data of the Borrelia PubMLST database. The link to the database is listed in the key resources table. PubMLST ID information is listed in Table S1.
• •This paper does not report any original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

The authors gratefully acknowledge Samuel K. Sheppard for his generous and expert support of this work. We would like to thank Nicholas H. Ogden, Shary Tyson, and Robert E. Rollins for providing the sequence data that were uploaded to the Borrelia PubMLST webpage. We would like to thank all members of the German National Reference Centre for Borrelia for technical support. The National Reference Center for Borrelia was funded by the Robert-Koch-Institut, Berlin. The sequencing was funded by the ESGBOR ESCMID Study Group. A.S., V.F., and G.M. are members of the ESGBOR ESCMID Study Group. PubMLST is funded by a Wellcome Trust Biomedical Resource Grant (218205/Z/19/Z).

Author contributions

Conceptualization, S.H., G.M., V.F., and K.A.J.; project administration, S.H., G.M., and V.F.; supervision, G.M., V.F., K.A.J., E.M., A.D., A.W., J.H., and A.S.; investigation, S.H., G.M., K.A.J., and S.C.-R.; formal analysis, S.H., G.M., K.A.J., and S.C.-R.; validation, S.H., G.M., K.A.J., and S.C.-R.; visualization, S.H.; data curation, S.H., G.M., E.M., and K.A.J.; writing – original draft, S.H. and G.M.; writing – review & editing, S.H., G.M., V.F., A.D., K.A.J., and E.M.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Deposited data
B. burgdorferi s.s. B31 genome data (extraction of chromosomal coding sequences; extraction date: November 15, 2023)	GenBank	NC_001318.1/AE000783.1
Genome set (development and validation of cgMLST scheme, see Table S1 for PubMLST id information)	Borrelia PubMLST database	https://pubmlst.org/organisms/borrelia-spp
Software and algorithms
BIGSdb software	Jolley et al.17	https://github.com/kjolley/BIGSdb
MAFFT (used for “Sequence Export” BIGSdb tool)	Katoh et al.45	https://mafft.cbrc.jp/alignment/software/linuxportable.html
IQ-TREE v2.2.2.7	Nguyen et al.37 Hoang et al.46 Kalyaanamoorthy et al.47	https://github.com/iqtree/iqtree2
iTOL v6.8.1	Letunic et al.48	https://itol.embl.de/
Gubbins v3.2.1	Croucher et al.49	https://github.com/nickjcroucher/gubbins
rapidnj v2.3.2	Simonsen et al.50	https://github.com/somme89/rapidNJ
RAxML v8.2.12	Stamatakis38	https://github.com/stamatak/standard-RAxML
GrapeTree (built-in BIGSdb)	Zhou et al.51	https://achtman-lab.github.io/GrapeTree/MSTree_holder.html
PYANI v0.2.12	Pritchard et al.52	https://github.com/widdowquinn/pyani

Methods details

Development of the cgMLST scheme

Loci set

For the development of the scheme chromosomal coding sequences (CDS) (n = 815) of B. burgdorferi s.s. B31 from GenBank: NC_001318.1/AE000783.1 were extracted (extraction date: November 15, 2023) and defined in BIGSdb using built-in functionality. Unique identifiers in the format BORRxxxx were assigned with original designations defined as aliases. This independent locus nomenclature allows for the future addition of new loci not present in B. burgdorferi s.s. B31.

Genome set

High quality genome assemblies (see criteria below) of unique strains and available through the Borrelia PubMLST website (December 28, 2023, Table S1) were used as the development genome set (n = 174). Published results have shown that most genes on the chromosome are core genes.^{10,53,54,55,56,57} Insufficient DNA quality and low coverage can result in low quality genomes that are missing a noticeably high number of loci (unpublished data). To identify high and low quality genomes all available genomes from the Borrelia PubMLST database (December 28, 2023; 177 genomes belonging to 17 B burgdorferi s.l. species, Table S1) were scanned against the 815 chromosomal B31 CDS using the BIGSdb “Gene Presence” analysis tool with default settings (70% min identity, 50% min alignment, BLASTN word size of 20). Our analyses showed that 175 genomes were missing <4%, which were considered high quality genomes. The remaining two genomes (B. bavariensis Tmsk976-2013 [Borrelia PubMLST: ID2722] and B. garinii Tmsk1193-2013 [Borrelia PubMLST: ID2751]) missed >14% of the loci and thus, were considered low quality genomes and were excluded from the genome set. Two of the 175 high quality genomes were identified as duplicates (B. garinii PBr; Borrelia PubMLST: ID2723 and ID2733) of which one (Borrelia PubMLST: ID2733) was excluded from the dataset. Thus, the development genome comprised 174 genomes and included samples of 17 B burgdorferi s.l. species (Table 1, see Table S1 for detailed isolate information).

cgMLST scheme set up and refinement

The 174 genomes were annotated at the defined loci using successive rounds of automated BIGSdb allele assignment using thresholds of 97% identity and over 99% alignment length compared to the closest matching allele. Auto-assigned alleles were complete coding sequences with consensus start codons, no internal stop codons, and an in-frame stop codon. Some manual curation was also necessary to assign exemplary alleles that were slightly different lengths to those found in B31. Loci that were present and designated in ≥ 95% of the genomes were included in the cgMLST scheme.

Core genome sequence types and clustering

Core genome sequence types (cgSTs) were assigned for samples with ≤ 2% missing loci (13 loci). Missing cgMLST loci were assigned as “N” in profiles and ignored in pairwise comparisons. This can result in some isolates potentially having more than one cgST where profiles are identical apart from the presence of missing loci (see example in Figure S1). To group the cgSTs, single linkage clustering was applied using various allelic mismatch thresholds (100, 50, 25, 10, and 5 loci). Core genome clusters were designated with “Bb_cgc” indicating B. burgdorferi s.l. core genome cluster followed by the allelic mismatch threshold used, although it should be noted that these cluster groups are not stable due to merging as data are added and therefore should not be used for nomenclature.

Phylogenetic and GrapeTree analyses

Phylogenetic analyses based on the MLST scheme and the developed cgMLST scheme were performed with the 174 genomes and results were compared. A MAFFT⁴⁵ alignment of the concatenated MLST and cgMLST sequences was generated on the Borrelia PubMLST.org Website using the “Sequence Export” tool and the options “MAFFT aligner”⁴⁵ and “concatenate in frame”. Maximum likelihood (ML) trees were generated with IQ-TREE 2.2.2.7,³⁷ 1000 ultrafast bootstrap replicons (UFBoot)⁴⁶ using the substitution models GTR+F+I+R3 for MLST and GTR+F+I+R6 for cgMLST phylogenies, respectively. ModelFinder as part of IQ-TREE⁴⁷ was used for model selection. The ML trees were visualized and edited using the online tool iTOL (Interactive Tree of Life) v6.8.1.⁴⁸

To analyze if recombination in the cgMLST loci affected the phylogeny, the cgMLST ML tree was compared to an ML tree based on SNPs found in non-recombinant regions of the cgMLST loci. A phylogeny without recombination was constructed using Gubbins v3.2.1.⁴⁹ For the first phylogeny rapidnj v2.3.2⁵⁰ (model: JC) was used, followed by a subsequent iteration of an ML tree using RAxML v8.2.12³⁸ (model: GTRGAMMA) and 7 iterations.

An additional 120 genomes (validation genome set, Tables 1 and S1), uploaded to the Borrelia PubMLST website between December 28, 2023 and March 25, 2024, were included in the analyses (294 genomes in total). For phylogenetic analyses, a MAFFT⁴⁵ alignment of the concatenated cgMLST loci was generated by the built-in BIGSdb iTOL tool and uploaded to the iTOL website.⁴⁸ As before, the ML tree was generated with IQ-TREE 2.2.2.7³⁷ using same settings.

Distance matrices based on the MLST or cgMLST scheme were generated using BIGSdb “Genome Comparator” with default settings (70% min identity, 50% min alignment, BLASTN word size of 20, pairwise ignoring missing values). The output also included information about the sequence type (ST) (based on MLST) and the cgST (based on cgMLST). GrapeTree⁵¹ was used to generate and visualize a minimum spanning tree (MST) using the built-in BIGSdb GrapeTree tool and selecting the MLST or cgMLST scheme.

Average nucleotide identity (ANI) analyses

The ANI values between B. bavariensis isolates were calculated using PYANI v0.2.12⁵² choosing BLAST+ method (ANIb) to align 1020 nt fragments of the input sequences.

Example analysis tools available on the BorreliaPubMLST.org website

• (1)Sequence Export: enables the download of loci sequences and also of the MAFFT⁴⁵ alignment of the concatenated scheme loci of up to 200 isolates from the database.
• (2)Third party analyses tools: such as GrapeTree,⁵¹ PhyloViz⁵⁸ (generating minimum spanning trees and distance matrices), iTOL⁴⁸ and Microreact⁵⁹ (generating neighbor joining trees and the MAFFT⁴⁵ alignment of the concatenated scheme loci).
• (3)Genome Comparator: generates distance matrices, alignment of the concatenated scheme loci, splits graph and generates several further outputs. The tool provides the possibility to include external users’ genomes that are currently under investigation but not published yet.

Detailed information about the website and their applications can be found in Jolley et al.¹⁷

Quantification and statistical analysis

For the validation of the developed cgMLST scheme phylogenetic, GrapeTree and ANI analyses were conducted. For the phylogenetic analyses ML trees were generated using IQ-TREE 2.2.2.7,³⁷ 1000 ultrafast bootstrap replicons (UFBoot)⁴⁶ and ModelFinder as part of IQ-TREE⁴⁷ for model selection. In order to evaluate the statistical support of inner nodes, the bootstrap values were displayed as colored points in the ML trees (green: 100, yellow: <100 and ≥ 95, red: <95). Additionally, phylogeny without recombination was constructed using Gubbins v3.2.1,⁴⁹ rapidnj v2.3.2⁵⁰ and RAxML v8.2.12.³⁸ For the ML trees visualization and editing the online tool iTOL v6.8.1⁴⁸ was used. The BIGSdb “Genome Comparator” was used to generate distance matrices and the built-in BIGSdb GrapeTree⁵¹ tool was used to generate and visualize the MSTs based on the distance matrices. The ANI values were calculated using PYANI v0.2.12.⁵² Additional details are provided in the method details section, results and figure legends.

Published: December 18, 2024

Supplemental information

Table S1. Isolate information, sequence typing, assembly statistics, and genome quality assessment, related to Table 1

Excel file containing additional data too large to fit in a PDF.

Table S2. cgMLST loci, related to Figure 1

Excel file containing additional data too large to fit in a PDF.

Table S3. ANI values of B. bavariensis isolates, related to Figures 3 and 4

Excel file containing additional data too large to fit in a PDF.