Analysis of the Borreliaceae Pangenome Reveals a Distinct Genomic Architecture Conserved Across Phylogenetic Scales

Abstract

The family Borreliaceae contains arthropod-borne spirochetes that cause two widespread human diseases, Lyme disease and relapsing fever. Lyme disease is a subacute, progressive illness with variable stage and tissue manifestations. Relapsing fever is an acute febrile illness with prominent bacteremia that may recur and disseminate, particularly to the nervous system. Clinical heterogeneity is a hallmark of both diseases. While human clinical manifestations are influenced by a wide variety of factors, including immune status and host genetic susceptibility, there is evidence that Borreliaceae microbial factors influence the clinical manifestations of human disease caused by this family of spirochetes. Despite these associations, the spirochete genes that influence the severity and manifestations of human disease are, for the most part, unknown. Recent work has identified lineage-specific expansions of lipoproteome-rich accessory genome elements in virulent clones of Borrelia burgdorferi. Using publicly available genome assemblies, it is shown that all Borreliaceae lineages for which sufficient sequence data are available harbor a similar pattern of strongly structured, lineage-specific expansions in their accessory genomes, particularly among lipoproteins, and that this pattern holds across phylogenetic scales including genera, species, and genotypes. The relationships among pangenome elements suggest that infrequent episodes of marked genomic change followed by clonal expansion in geographically and enzootically structured populations may account for the unique lineage structure of Borreliaceae. This analysis informs future genotype–phenotype studies among Borreliaceae and lays a foundation for studies of individual gene function guided by phylogenetic patterns of conservation, diversification, gain, and/or loss.

Spirochetes of the family Borreliaceae cause two major clinical syndromes: relapsing fever (RF) and Lyme disease (LD). RF is an acute, febrile disease with prominent and often recurrent spirochetemia. High densities of spirochetes are often found in the bloodstream during RF [1]. LD, in contrast, is a subacute, progressive illness with variable, organ-specific manifestations, a consequence of the tropism of the spirochete to various host tissues [2]. LD and RF also have different primary tissue tropisms: RF spirochetes are found at the highest abundance in the bloodstream [3], whereas most LD spirochetes are found at the highest concentrations in tissues and occur only rarely, and at low levels, in the bloodstream [2].

Relapsing febrile syndromes have been recognized since antiquity, but it wasn't until the studies of Obermeier in 1873 that spirochetes were recognized as a cause of this syndrome (reviewed in [4]). The spirochetes described by Obermeier were initially called Spirochaeta obermeieri and later named Borrelia (see references in [4]). LD was described more than a century later. Although reports of erythema chronicum migrans were known in the European literature [5], it was not until the 1970s that Steere and coworkers [6, 7] systematically described the syndrome that later came to be known as LD. The causative spirochete was isolated by Burgdorfer and coworkers in 1982 [8] and subsequently named Borrelia burgdorferi in his honor.

The clinical divergence between LD and RF is associated with a deep phylogenetic split between RF and LD spirochetes. In 2003, a third monophyletic clade within Borreliaceae was identified and found to contain spirochetes that infect reptiles, particularly Varanus lizards and other vertebrates [9–11]. These reptile-associated (REP) Borrelia are not known to cause human disease but given their recent recognition, human pathogenicity and many other details of these taxa are not fully understood. Subsequent characterization of RF and LD spirochetes revealed new lineages that differed in their genetic content and phenotype. Many RF species have been described and delineated on the basis of their geographic distribution (“Old World” vs “New World”) and/or host associations (soft-tick vs hard-tick RF agents) [4]. After the initial description of B burgdorferi, extensive diversity among B burgdorferi spirochetes was recognized by several authors [12, 13] using newly developed serological reagents. With the recognition of these genospecies came early evidence suggesting that infection by different members of the species complex was correlated with geography and distinct clinical manifestations [14], foreshadowing that the complex clinical syndrome of LD was due, in part, to extensive microbial genetic diversity. These findings were later confirmed and extended (reviewed in [15]). Genetic lineages within B burgdorferi have been referred to as genospecies, genomic species, or subspecies, and the overall complex of LD-causing Borreliaceae is referred to as B burgdorferi sensu lato (Bbsl) or Borreliella. (Recent phylogenetic analysis supports separation of RF and LD agents into separate genera [16]. Adeolu and Gupta [16] proposed revising the taxonomic assignments, designating Bbsl as Borreliella while retaining Borrelia for RF spirochetes based on historical priority. Although the proposed taxonomic revision is controversial [17–19], this article follows NCBI and uses Borrelia to refer to RF spirochetes and Borreliella to refer to LD spirochetes. The article parenthetically adds Bbsl for clarity and because these terms by GenBank have also been criticized [20] and are not universally adopted by taxonomic databases. The goal is clarity and consistency with the data source, rather than to take a position on the nomenclature.) All Borreliella/Bbsl spirochetes cause LD, but the clinical manifestations of LD vary based on the infecting genotype. Borrelia burgdorferi is strongly associated with Lyme arthritis (LA) [21]; Borrelia garinii and Borrelia bavariensis are strongly associated with Lyme neuroborreliosis [22–24], and Borrelia afzelii is associated with acrodermatitis chronica atrophicans [22–24] and Borrelial lymphocytoma [25, 26].

Extending the paradigm that phenotypic differences within LD are linked to genetic differences, lineages within a given genospecies, B burgdorferi sensu stricto (hereafter referred to as B burgdorferi) were identified, classified by various typing schemes—including serotyping, ribosomal spacer type (RST), and outer surface protein C (OspC) type [27–29]—and found to differ in their phenotypes. At the genotype level, which is best studied for B burgdorferi, OspC type A/RST 1 genotypes of B burgdorferi disseminate in humans at higher rates [30–33] and cause higher rates of symptoms [34], greater inflammation [34, 35], and more severe LA [34].

Thus, the divergence of clinical disease in humans correlates with genetic divergence across phylogenetic scales in Borreliaceae. The largest phenotypic differences are observed between RF and LD spirochetes; lesser but still notable clinical variation is observed between species within a genus, and more subtle yet meaningful variation occurs at the genotype level.

Despite strong evidence that microbial genetic factors influence clinical manifestations, the specific genes that contribute to these phenotypic differences are incompletely understood. Many genes lack annotated functions; in other cases, where some or all functional activities are well-characterized, including systems for antigenic variation, serum resistance, and tissue adhesion (reviewed in [36–38]), gaps remain in understanding how these functions work together to influence disease manifestations. This is in part due to the complexity of Borreliaceae genomes, which contain numerous plasmids that are challenging to sequence and annotate [39, 40], and the relative rarity of human clinical isolates of LD and RF, which require specialized culture conditions not typically performed in routine clinical practice [41].

Identifying genes that contribute to distinct clinical manifestations of LD and RF remains an important scientific goal because of its potential to reveal fundamental insight into Borreliaceae pathogenesis and open new opportunities for therapy and prevention. There are opportunities to do so using comparative phylogenetic approaches, in which the genetic content of related lineages is cataloged and compared, as well as microbial genome-wide association study approaches, in which genetic elements are tested for their association with a phenotype of interest. Recent work has used these approaches to identify the genetic elements associated with disseminated LD [33], but the genetic and phenotypic heterogeneity of Borreliaceae spirochetes means that these and related approaches are in their infancy. The concept of a pangenome, representing a “core” of genes shared by all isolates along with a “shell” or “accessory” genome of genes present in some isolates, was introduced in 2005 [42] and has become one of the most widely used methods in comparative bacterial genomics. Understanding pangenome relationships among strains and lineages can extend comparative genomic and genotype–phenotype studies to Borreliaceae linages; prioritize individual genes for functional follow-up; and foster an understanding of genomic diversity and evolution within the family. Here, publicly available genome sequence data are used to construct pangenomes and characterize lineage-specific genomic events from across Borreliaceae.

MATERIALS AND METHODS

Borreliaceae genome assemblies and associated metadata were downloaded from the National Center for Biotechnology Information (NCBI). Assemblies were filtered to be >1 Mb (excluding chromosome-only assemblies) and <2 Mb. Assemblies were annotated using Prokka [43]. Pangenomes were constructed using Roary [44], with a minimum BLAST threshold of 80% homology and without splitting of paralogous groups, and PIRATE [45] using default parameters. Trees were constructed using FastTree [46] and IQ-TREE2 [47] with 1000 ultrafast bootstrap replicates. The pipeline was executed using Snakemake [48]. Lipoproteins were annotated using SpLip [49]. The resulting pangenomes, phylogenetic trees, and metadata files were analyzed and plotted using R [50], ggplot [51], and ggtree [52]. The genomic position of pangenome homology groups was annotated by aligning pangenome elements against Borreliaceae reference genomes using minimap2 [53]. Because chromosome sequences are easier to assemble than plasmid sequences in Borreliaceae [39, 54], and therefore presumably more accurate if assemblies are unfinished, the following approach was used: If pangenome elements aligned to the chromosome sequences, they were annotated as provisionally chromosomal. If they aligned to reference sequences but not to chromosome sequences, they were annotated as provisionally plasmid-encoded. Lineage-specific associations of individual pangenome elements were identified using logistic regression to predict the lineage based on the pangenome homology group, as implemented in R using the glm() command.

Ethical Approval

The analysis of microbial genome sequences from deidentified patient data was approved by the Massachusetts General Hospital Institutional Review Board under protocol 2019P001846.

RESULTS

To characterize the pangenome relationships among Borreliaceae, all publicly available Borreliaceae assemblies were downloaded from NCBI GenBank. The assemblies were annotated using Prokka [43], pangenomes were constructed by efficient all-versus-all BLAST implemented in Roary [44], and maximum-likelihood phylogenetic trees consisting of core genome sequences were constructed using FastTree [46] and/or IQ-tree2 [47]. As of 26 December 2023, NCBI GenBank contained 702 Borreliaceae genome assemblies from 20 Borreliella/Bbsl taxa and 18 Borrelia taxa from both the RF group and the “reptilian” (REP) group [55]. The core genome tree constructed from these assemblies is shown in Figure 1, with overrepresented taxa downsampled to a maximum of 25. Consistent with prior phylogenetic analyses, the tree reveals a well-supported, deep division into 2 monophyletic branches consisting of Borrelia spp and Borreliella spp/Bbsl spirochetes. Borrelia spp contain 2 monophyletic sister clades, RF spirochetes, and the REP group.

Figure 1.

Core genome phylogenetic relationships among Borreliaceae. Radial tree of Borreliaceae with tips annotated by species. Nodes with bootstrap support >90% are colored with a dot, demonstrating robust phylogenetic support for the genus- and species-specific divisions on the tree. The clade of Lyme disease (LD)–causing spirochetes, Borreliella/Borrelia burgdorferi sensu lato, has been highlighted in red; relapsing fever (RF)–causing spirochetes and the reptile-associated group are shaded in blue and orange, respectively. Old World and New World subgroupings of RF spirochetes are shown with a dotted line. Within the LD spirochetes, the clades associated with distinct manifestations are shown. The scale bar denotes nucleotide substitutions per site. Abbreviations: ACA, acrodermatitis chronicum atrophicans; LA, Lyme arthritis; LD, Lyme disease; LNB, Lyme neuroborreliosis; REP, reptile-associated group; RF, relapsing fever.

The Borreliaceae pangenome contains 299 core open reading frames (ORFs) at a protein BLAST homology threshold of 80% (Figure 2). The noncore (accessory) pangenome among Borreliaceae contains 5684 ORF homology groups, although this estimate is probably inflated due to fragmented assemblies [56]. The accessory genome is strongly structured by lineage, with divisions in the accessory genome following the core genome phylogeny. The deep phylogenetic division between Borrelia spp and Borreliella spp/Bbsl is mirrored in the divergence of the accessory genomes. Borrelia spp and Borreliella spp/Bbsl possess additional genus- and species-specific core genomes that are readily apparent in Figure 2. Annotated pangenomes can be used to define genus- and species-specific pangenomes and lipoproteomes (Supplementary File 1). This analysis was repeated using PIRATE [45], which allows for the construction of pangenomes across multiple homology values, confirming that this accessory genome structure was not an artifact of a single homology threshold.

Figure 2.

Accessory genome content among Borreliaceae is strongly structured by lineage. Patterns of open reading frame presence are shown according to a phylogenetic tree produced from core genome sequences. A matrix showing the presence or absence of pangenome homology groups in each strain is shown at right. Pangenome homology groups that are present are shaded in solid color, and groups that are absent are uncolored. Each row corresponds to an individual strain, and each column corresponds to individual homology groups. The rows are ordered according to a phylogenetic tree, with tips annotated by the Borreliaceae genus and species. Relevant taxa are also labeled. As in Figure 1, the Borreliaceae that cause Lyme disease (Borreliella/Borrelia burgdorferi sensu lato) are shaded and labeled in red; the relapsing fever and reptile-associated groups are shaded and labeled in blue and orange, respectively. The columns are clustered with hierarchical clustering to reflect similarity in the patterns of gene presence among lineages. The tree scale bar denotes nucleotide substitutions per site. Abbreviations: Bbsl, Borrelia burgdorferi sensu lato; LD, Lyme disease; ORF, open reading frame; REP, reptile-associated group; RF, relapsing fever.

The structure of accessory genome elements consists of a lineage-specific “block” pattern in which groupings of accessory genome presence are almost perfectly correlated with lineage-defining subdivisions in the core genome tree (Figure 2). This pattern closely resembles the pattern observed for the B burgdorferi pangenome [33] but is present across phylogenetic scales. Within the Borreliella spp/Bbsl, there are lineage-specific groups of accessory loci, for example among B burgdorferi (B burgdorferi), B garinii, B bavariensis, and B afzelii genomes. Borrelia spp similarly possess lineage-specific changes such that there are tightly structured core genomes for Old World RF agents (including Borrelia crocidurae and Borrelia duttonii), hard-tick RF agents (Borrelia miyamotoi), and New World RF agents (including Borrelia turicatae, Borrelia parkeri, and Borrelia hermsii). Within sublineages, a similar pattern occurs. For example, B miyamotoi also shows evidence of distinct genotypes, with subsets of B miyamotoi diverging based on geographic origin [57]. Russian and Asian B miyamotoi isolates possess a slightly enlarged pangenome and a distinct block of genetic elements compared to those from Western Europe and North America, each of which forms a distinct clade (Supplementary Figure 1). Among B burgdorferi sublineages, OspC type A/RST1 genotypes possess a lineage-core group of ORFs [33].

Lipoproteins are particularly important for Borreliaceae immune evasion, tissue tropism, and other aspects of vertebrate pathogenesis [2, 58]. To characterize the relationships among the Borreliaceae pan-lipoproteome, lipoproteins were annotated using SpLip [49] and analyzed the phylogenetic patterns of lipoprotein-annotated homology groups (Figure 3). The strong lineage-specific pattern of the pangenome (Figure 2) is reproduced almost exactly in the lipoproteome (Figure 3), except that all lipoproteins are lineage-variable at the homology thresholds used.

Figure 3.

The Borreliaceae lipoproteome is strongly structured by lineage. Patterns of open reading frames annotated as putative lipoproteins, as classified by SpLip [49], are shown according to a phylogenetic tree produced from core genome sequences. A matrix showing the presence or absence of probable lipoproteins in each strain is shown. Each row corresponds to an individual strain and each column corresponds to a homology group classified as a probable lipoprotein. The rows are ordered according to a phylogenetic tree, with tips annotated by the Borreliaceae genus and species. Relevant taxa are also labeled. As in Figure 1, the Borreliaceae that cause Lyme disease (Borreliella/Borrelia burgdorferi sensu lato) are shaded and labeled in red; the relapsing fever and reptile-associated groups are shaded and labeled in blue and orange, respectively. The columns are clustered with hierarchical clustering to reflect similarity in the patterns of gene presence among lineages. The tree scale bar denotes nucleotide substitutions per site. Abbreviations: Bbsl, Borrelia burgdorferi sensu lato; LD, Lyme disease; ORF, open reading frame; REP, reptile-associated group; RF, relapsing fever.

The spatial organization of homology groups within Borreliaceae genomes was then assessed. Because many assemblies are incomplete due to the difficulty of generating complete assemblies through short-read sequencing [39, 40, 59], a provisional analysis of draft genome organization was performed by aligning a representative of each pangenome homology group to well-annotated reference genomes. Lipoproteins are preferentially encoded on plasmids throughout Borreliaceae (Figure 4). There was a strong association between lipoprotein status and chromosomal versus plasmid genomic location (Fisher exact test, P < 2.2 × 10⁻¹⁶); a similar association was observed when lipoprotein status was evaluated using 2 categories (not a lipoprotein vs probable lipoprotein) or 3 categories (not a lipoprotein, possible lipoprotein, or probable lipoprotein). Core genome elements were almost exclusively encoded on the chromosome (Supplementary Figure 2). Accessory genome elements are preferentially encoded by plasmids (Supplementary Figure 3), including all Borreliella accessory genome elements and many Borrelia accessory elements. One notable difference between Borrelia spp and Borreliella spp/Bbsl is that in Borrelia spp portions of the species-specific accessory genome are encoded on the chromosome (Supplementary Figure 2), whereas in Borreliella/Bbsl nearly all species-specific accessory genome elements are extra-chromosomal (Supplementary Figure 3). This suggests that the chromosome in Borrelia spp has evolved to carry lineage-specific accessory gene content, whereas, in Borreliella spp/Bbsl, lineage-specific variation in gene content is almost exclusively encoded on plasmids. Thus, the pattern of lineages defined by plasmid-encoded accessory elements enriched in lipoproteins is present but less absolute in Borrelia compared to Borreliella/Bbsl. Together, these analyses enable the identification of lineage-specific lipoproteomes (Supplementary Figures 4 and 5), an approach that can highlight more subtle lineage-specific genomic patterns such as the unique sequences associated with OspC type A/RST1 genotypes among B burgdorferi (Supplementary Figure 4B and 4C) [33].

Figure 4.

Borreliaceae lipoproteins are preferentially encoded on plasmids. The genomic location of pangenome elements is shown. Counts are stratified by whether they are classified by SpLip [49] as probable lipoproteins, possible lipoproteins, or not lipoproteins. Test of association by Fisher exact test, P < 2.2 × 10⁻¹⁶.

By summing the number of homology groups in an individual genome, pangenome analysis also provides an estimate of the size of the coding genome and lipoproteome. Borrelia genomes have slightly more ORF homology groups than Borreliella/Bbsl genomes (Figure 5A), an effect that is driven by soft-tick Borrelia RF spirochetes (Figure 5B), particularly among lipoproteins (Figure 5C and 5D). Differences are seen among arthropod vectors. The hard-tick RF spirochete B miyamotoi has fewer ORF homology groups and probable lipoproteins than soft-tick RF spirochetes. Despite being an RF spirochete, its number of ORFs and probable lipoproteins are similar to LD spirochetes. Among LD Borreliella/Bbsl, B garinii and B bavariensis have fewer ORF homology groups than B burgdorferi; however, there is marked variation in genome size within B burgdorferi (Figure 5) [33] and other species (Figure 5B and 5D).

Figure 5.

Borrelia pangenomes contain an expanded number of open reading frames (ORFs) and putative lipoproteins. A, The number of ORF homology groups for each genome assessed, grouped by genus. B, The number of ORF homology groups for each genome assessed, grouped by species. C, The number of ORF homology groups annotated as probable lipoproteins for each genome assessed, grouped by genus. D, The number of ORF groups annotated as probable lipoproteins for each genome assessed, grouped by species. P values report the results of a Wilcoxon rank-sum test with a 2-sided alternative. Abbreviations: **P < .01; ****P < .0001.

DISCUSSION

The growing quantity of genomic data in publicly available databanks provides unique opportunities to characterize the evolutionary and genomic relationships among Borreliaceae. Such characterizations can help prioritize individual genes for further investigation and provide a basis for genome-wide studies of genotype-phenotype correlations. Recent work has shown that lineages of B burgdorferi contain groups of strongly linked loci accessory elements, particularly surface lipoproteins encoded on plasmids, and these blocks of loci are linked to clinical manifestations [33]. Using published assemblies of Borreliaceae, it is shown here that this pattern extends across phylogenetic scales and appears to be a general organizing principle of the Borreliaceae pangenome. The large number of genomes included from across the family is a strength of this analysis and builds on previous reports [33, 60], providing a panoramic view of Borreliaceae genome diversity. Limitations include a convenience sample with a lack of standardization among sequencing, assembly, and submission procedures; incomplete and fragmented assemblies that may inflate the number of pangenome elements [56] and do not easily permit an analysis of associated genomic context; and a focus on variation at the level of homology clusters rather than single genetic variants.

The strong block structure of accessory genome elements across phylogenetic scales is remarkable. The monophyletic nature of these blocks implies that most have a single common ancestor and that the set of genes included in the block has been stable over evolutionary time. This pattern suggests that extant variation in Borreliaceae is the consequence of a relatively small number of profound ancestral changes involving multiple genetic elements, and that these marked changes occurred simultaneously or in short succession. The abrupt genesis of major genomic changes is consistent with a “punctuated equilibrium” model [61] of evolution. This pattern would explain the strong statistical linkage between physically unlinked markers, for example, OspC and the B burgdorferi ribosomal spacer [62].

It is notable that the patterns hold for major and minor lineages across scales. The fractal nature of this pattern may provide clues into the underlying mechanisms that result in this pattern. Fractals occur when the same laws apply recursively across scales [63]. The process of lineage generation may be an example of such a recursive law and suggests the following mechanism: The creation of a new genotype begins with a rare genomic event that involves the gain, loss, and/or exchange of multiple genes, probably in an arthropod or vertebrate host infected with multiple genotypes and/or different species. Such a lineage may clonally expand in a new niche, leading to geographically and/or ecologically structured populations. Over time, these populations accumulate further genetic diversity and eventually become differentiated enough to be classified as a species or genus. Along the way, sublineages form recursively by the same mechanism.

Although this mechanism raises questions about how Borreliaceae could survive abrupt genetic shifts, complex genomic rearrangements are increasingly recognized in other contexts, especially in cancer, where the phenomenon is termed chromothripsis [64–66]. The unique genome organization of Borreliaceae [39, 67]—in which conserved functions are encoded by a single chromosome and genes that mediate host–pathogen interactions, particularly surface lipoproteins encoded on plasmids—probably facilitates organism survival following an episode of dramatic change by allowing loss, gain, or exchange of multiple genes on plasmids while insulating the core metabolic or housekeeping functions of the spirochete on the chromosome and certain conserved plasmids. Other mechanisms may also predispose to complex genetic rearrangements: Both RF and LD Borreliaceae genomes contain sophisticated machinery for inducing gene conversation to support antigenic variation [36, 68–70]. Regulation of this process is paramount to genome integrity. Genomic instability involving multiple linked loci is readily observable in B burgdorferi, in which routine in vitro culture is associated with spontaneous plasmid loss [71, 72] and double-stranded breaks appear to result in plasmid loss in vitro [73]. B burgdorferi also possesses a bacteriophage [74–76], raising the possibility that bacteriophage-mediated lateral transfer of multiple loci has contributed to the evolutionary history of Borreliaceae.

The genetic structure of Borreliaceae informs the search for microbial loci associated with human disease in several ways. Lineage-level associations aggregate multiple genes’ worth of effects, potentially increasing the effect size of the genotype–phenotype link, and the consistent genetic profile among strains within a given lineage reduces the variance of these effects because strains are nearly isogenic. Thus, lineage-level associations should be relatively easy to identify, which may be why several are known. On the other hand, locus-level effects will be more difficult to identify because individual loci are rarely observed alone. Large samples will be required, and statistical techniques, such as conditioning on lineage-level effects, may help. Experimental methods in appropriate disease models involving reverse genetics will likely be necessary to assign or exclude causal roles for individual genes. Functional genomic approaches capable of screening multiple loci, such as TnSeq [77], CRISPRi using dCas9 [78, 79], and gain-of-function screens [80, 81], have recently been developed and are well-suited to the task of identifying experimentally important loci from among a larger set of candidates. Lineage-level associations greatly narrow the list of candidates for functional characterization and reverse genetic methods.

In summary, this analysis of Borreliaceae pangenomes reveals a conserved genomic architecture in which lineages are defined by correlated blocks of accessory genome elements, primarily encoded on plasmids and enriched for lipoproteins, a pattern that holds across phylogenetic scales. This unusual fractal genetic structure provides clues into the origin and evolution of Borreliaceae lineages and establishes a basis for reductionist studies of gene function informed by family-wide phylogenetic relationships.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. The author gratefully acknowledges all of the sequence submitters and contributors who deposited Borreliaceae assemblies in GenBank (Supplemental File 2), as well as Klemen Strle, Ira Schwartz, and Allen Steere for helpful discussions.

Code availability. Analysis code is available at github.org/jacoblemieux/panborreliaceae.

Financial support. This work was supported by the National Institutes of Health (award number K99/R00AI148604).

Supplement sponsorship. This article appears as part of the supplement “Lyme Disease,” sponsored by the Lyme Disease Program at Massachusetts General Hospital and a generous gift from the Morse family.

Potential conflicts of interest. Author certifies no potential conflicts of interest.

The author has submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.