Genomic and structural analysis of major antigenic proteins' variability in the human relapsing fever spirochete Borrelia persica

Highlights

• •Borrelia spirochetes have a unique mechanism to evade host immune system.
• •Borrelia persica has one chromosome, 12 linear and one circular plasmids.
• •Telomeric hairpin motifs were detected in the chromosome and linear plasmids.
• •Plasmid profiles remained stable after one year of continuous in vitro culture.
• •A Vlp common variable region kept the same external position across structures.

Abstract

Antigenic variation enables immune evasion of pathogens, yet the mechanisms for the rapid change in their antigens without loss of structural and functional features are not fully known. Tick-borne relapsing fever (TBRF) spirochetes of the genus Borrelia can be used as a model because their variable major proteins (Vmps) drive antigenic variation followed by recurrent waves of bacteremia. In this study we generated the genome of Borrelia persica LMU-C01 and evaluated Vmp structure across relapsing fever Borrelia species. Borrelia persica was found to have a genome of 1.7Mb with one linear chromosome, 12 linear and one circular plasmids. Telomeric hairpin motifs characteristic of Borrelia linear replicons were detected in the chromosome and in the linear plasmids. Pulsed-field gel electrophoresis indicated that plasmid profiles remained identical after a year of continuous in-vitro cultivation. Genome annotation revealed 48 complete variable large proteins (Vlps) and one variable small protein (Vsp), all encoded in plasmids. Comparative structure modeling of B. persica Vlps together with 420 Vlps from twelve TBRF species revealed a conserved, surface-exposed variable region that occupies a similar external predicted position across structures, even as primary sequences diverge. This indicated that antigenic variation preserves topology while tolerating sequence plasticity, which may be shared with other immune evasive bacteria.

Graphical abstract

1. Introduction

Tick- and louse-borne relapsing fever (RF) is an acute infectious disease caused by spirochetes of the genus Borrelia (Dworkin et al., 2008). The RF Borrelia species are generally divided based on the type of vector involved in their transmission. Most species are transmitted by soft ticks in the genus Ornithodoros, while one pathogenic species, Borrelia miyamotoi is transmitted to humans by ixodid hard ticks (Guillamet et al., 2023; Hoogstraal, 1985; Jiang et al., 2018; Morel et al., 2019; Parola and Raoult, 2001; Smith et al., 1985; Talagrand-Reboul et al., 2018). Additionally, louse-borne RF caused by Borrelia recurrentis is transmitted by the human body louse Pediculus humanus (Warrell, 2019).

Borrelia persica, transmitted by the argasid tick Ornithodoros tholozani has been identified as a leading cause of human tick-borne RF in the Middle East, the Persian Gulf and Central Asia (Tang et al., 2024). Human infection with B. persica is endemic in Israel and has been reported in Iran, Egypt, Pakistan, Russia, China and several Central Asian countries including Uzbekistan (Colin de Verdiere et al., 2011; Davis and Hoogstraal, 1954; Masoumi Asl et al., 2009; Muigg et al., 2020; Safdie et al., 2010; Zhang et al., 2022). The clinical presentation of RF in human patients typically involves non-specific symptoms such as recurring fever episodes, headaches, lethargy, vomiting, muscle pain, and joint pain (Barbour and Hayes, 1986; Dworkin et al., 2008; Muigg et al., 2020). The mortality rate in untreated patients can be as high as 10 %, particularly when co-existing factors such as malnutrition, dehydration, and coinfections are present (Barbour, 2022). Although appropriate antibiotic therapy can reduce the mortality rate to less than 2 %, severe complications including aseptic meningitis, acute respiratory distress syndrome, and neonatal death have been observed in patients infected with B. persica (Yagupsky and Moses, 1985; Yossepowitch et al., 2012).

The genome of Borrelia spp. is relatively small for bacteria (approximately 1–1.5 Mb), however, it exhibits an exceptional structural complexity and segmentation and consists of 5 to 24 distinct replicons (Casjens et al., 2000; Fraser et al., 1997; Kneubehl et al., 2022). All Borrelia spp. harbor a single, linear 900 kb chromosome, alongside a variable assortment of 5–22 extrachromosomal replicons, ranging in size from 5 to 220 kb. These replicons exist as linear or circular plasmids (Casjens et al., 2010; Scott et al., 2005). Because linear chromosomes are exceptional among bacteria and uncommon among other genera in the spirochete family, this architecture, together with the abundance of linear plasmids, is a hallmark of the Borrelia genus (Scott et al., 2005). These plasmids encode crucial factors involved in host adaptation, immune evasion, and survival within both the vector and vertebrate host, thereby playing a pivotal role in pathogenicity and virulence (Norris et al., 2010; Stewart et al., 2005).

Borrelia spp. have developed virulence strategies that diverge markedly from those of Gram-negative pathogens, favoring dynamic genetic and antigenic variation over classical secretion systems (Ante et al., 2021; Costa et al., 2015; Fraser et al., 1997). Antigenic variation is essential for establishing and sustaining infection. RF Borrelia spp. rely on an extensive repertoire of proteins of the variable major protein (Vmp) family, comprising distinct groups of variable small proteins (Vsp) and variable large proteins (Vlp) (Saint Girons and Barbour, 1991). Although dozens (40–60) of vmp genes are located on the Borrelia linear plasmids, a single expression locus is present and only one vmp is expressed at a time, while the rest are considered “silent cassettes” (Barbour et al., 1991; Dai et al., 2006). Vmp alleles are involved in recombination events, enabling the generation of novel antigenic variants in response to host immune pressure, and providing the spirochete’s ability to escape from the host humoral immune response (Dai et al., 2006). Most of our understanding of antigenic variation has come from research on the New World species, Borrelia hermsii, and prior studies identified nuances in the genomic structure of this immune evasion strategy (Kneubehl and Lopez, 2023). Previous studies involving structural and genetic analyses of the main Vsp variants of another New World species, Borrelia turicatae, established that these proteins consist of highly conserved N- and C-termini forming a membrane-proximal "stem" and a distal "dome" composed of four short variable regions (Cadavid et al., 1997; Lawson et al., 2006). Consequently, generating complete chromosome and plasmid resolved genomes of RF spirochetes from other regions of the globe is important.

To date, 39 complete and assembled RF Borrelia spp. genomes are available through the Integrated Microbial Genomes & Microbiomes (IMG/M) database (https://img.jgi.doe.gov). Of these, 85 % are of the New World species B. hermsii, Borrelia parkeri, B. turicatae, Borrelia coriaceae, Borrelia puertoricensis, and of the hard tick-borne species B. miyamotoi. The remaining 15 % are of the Old-World species Borrelia crocidurae, Borrelia duttonii, Borrelia maritima, Borrelia anserina and the louse-borne species B. recurrentis (Chen et al., 2019), and there are no complete genomes of RF borreliae from the Middle East and Eurasia.

This study provides a comprehensive characterization of the B. persica LMU-C01 strain isolated from a sick cat in Israel (Schwarzer et al., 2015). We performed an extensive genomic characterization of B. persica using long and short read whole genome sequencing, complete genome assembly, and annotation. A structural analysis was done to investigate extrachromosomal elements and to characterize telomeric sequences. We also performed a cross-species structural analysis of the virulence associated Vlp proteins. Our findings indicate that in the Vlp family there is a primary variable region within the protein structure flanked by two conserved regions and the entire protein conformation does not need to change. Collectively, our findings provide the first high-resolution genomic and structural map of the human RF species B. persica. By establishing a cross-species Vlp structural analysis and comparison, we clarify how Borrelia spirochetes utilize remarkably stable protein architectures and adapt to facilitate persistence in mammalian hosts.

2. Materials and methods

2.1. Whole genome sequencing and data analysis

Total DNA was extracted from B. persica strain LMU-C01 (passage 3), isolated from an ill cat and cultured in Pett./LMU Bp medium (Schwarzer et al., 2015), using a commercial kit (MasterPure Complete DNA and RNA Purification Kit, Epicentre) following the manufacturer's protocol. Whole genome sequencing was performed using Illumina and PacBio sequencing to obtain both short and long read sequences for assembly optimization.

2.2. Illumina sequencing

Libraries for sequencing on Illumina platform were prepared applying the Nextera XT DNA Library Preparation Kit (Illumina, San Diego, USA) with modifications according to Baym et al. (Baym et al., 2015). Samples were sequenced on NextSeq™ 500.

2.3. PacBio sequencing

SMRTbell™ template library was prepared according to the instructions supplied by PacificBiosciences (Menlo Park, CA, USA), following the Procedure & Checklist – Preparing Multiplexed Microbial Libraries Using SMRTbell® Express Template Prep Kit 2.0. Briefly, for preparation of 10kb-libraries 1 µg genomic DNA was end-repaired and ligated to barcoded adapters applying components from the SMRTbell Express Template Prep Kit 2.0 from Pacific Biosciences (Menlo Park, CA, USA). Reactions were carried out according to the manufacturer´s instructions. Samples were pooled according to the calculations provided by the Microbial Multiplexing Calculator. Conditions for annealing of sequencing primers and binding of polymerase to purified SMRTbell™ template were assessed with the Calculator in SMRT®link, PacificBioscience (Menlo Park, CA, USA). 1/16 SMRT cell was sequenced on the SequelII (PacificBiosciences, Menlo Park, CA, USA) taking one 15-h or one 30-h run.

2.4. Genome assembly and polishing

Long-read genome assembly was performed with the “Microbial Assembly” protocol included in SMRTlink v10 using default parameters (expected genome size 1.75 Mb). Error-correction was performed by a mapping of Illumina short reads onto finished genome using Burrows-Wheeler Alignment bwa v0.6.2 in paired-end (sample) mode using default settings (Li and Durbin, 2009) with subsequent variant and consensus calling using VarScan v2.3.6 (Koboldt et al., 2012). Genome annotation was based on Prokka v1.8 (Seemann, 2014) with subsequent manual curation.

2.5. Plasmid similarity and redundancy assessment

Comparative analysis was performed using the Pyani tool (v0.2.14) (Pritchard et al., 2016). Nucleotide sequence alignment was carried out with the MUMmer software (v4.0.0) (Kurtz et al., 2004). Plasmid similarity was then quantified via the Hadamard score, defined as the product of the sequence identity percentage and the overlap between the two compared sequences.

2.6. DNA replicons separation by pulsed-field gel electrophoresis (PFGE)

Genomic DNA samples were extracted from 50 ml of B. persica LMU-C01 cultures, grown through 103 passages over 1 year, with a cell density of 10⁷/ml using phenol-chloroform extraction, as previously described, with some modifications (Appendix A) (Lopez et al., 2008; Simpson et al., 1990). To determine the integrity of the chromosome and plasmids, DNA from two different passages (6 and 103) was assessed using PFGE, as previously described (Porcella et al., 2005).

2.7. Plasmid topology determination

For plasmid topology determination, the assembled B. persica LMU-C01 genome was processed using Circlator pipeline (v1.5.5) to differentiate circular from linear replicons (Hunt et al., 2015). Long-read sequencing and assembly data were subjected to Circlator’s automated workflow. Raw reads were mapped back to the assembly using BWA-MEM (v0.7.17), and the alignments were handled with SAMtools (v1.10) to generate sorted and indexed files for downstream analysis (Li et al., 2009; Li and Durbin, 2009). BLAST+ (v2.13.0) and Nucmer from the MUMmer package (v4.0.0) were used to screen for significant overlapping terminal sequences (Altschul et al., 1990; Kurtz et al., 2004). Contigs displaying high-identity overlaps (>95 % identity across >500 bp) at their ends were flagged as candidates for circularization. The Minimus2 algorithm from the AMOS package (v3.1.0) was used to refine and merge these suspected circular plasmids (Sommer et al., 2007). To further standardize the circular replicon assemblies, a starting point correction was performed using Circlator’s fixstart module. This step reorients each circular contig so that its sequence begins at a biologically relevant marker, thereby ensuring a consistent coordinate system for comparative genomic analyses and downstream annotation.

2.8. Telomeric sequences determination

Determination of the telomeric sequences of the linear replicons was done using assembled PacBio contigs. Each linear replicon was first aligned against itself using FlexiDot (v1.1) to create dot plot analyses (Appendix B). Breakpoints were defined from the dot plots and for each alignment, nucleotide sequences spanning 200–1000 bp upstream and downstream of the identified breakpoint were extracted and assessed for the presence of palindromic (hairpin) sequences using the EMBOSS Einverted tool (v6.6.0) (Rice et al., 2000). In cases where dot plot analysis did not reveal palindromes within a contig, trimmed and preassembled PacBio reads were mapped onto the contig, and the segments spanning the left or right edges were similarly analysed.

2.9. Genome annotation and plasmid typing

Plasmids and their names were designated based on topology, using “LP” for linear plasmid and “CP” for circular plasmids, and size (rounded to the nearest kb), with sequential numbering (LP17_1, LP17_2) applied when sizes were identical. The assembly was first annotated using a local version of NCBI’s Prokaryotic Genome Annotation Pipeline (PGAP, v2023–08–01) to identify coding sequences and genome features, while InterProScan (v5.57–90.0) provided further protein classification (Tatusova et al., 2016). Transcripts extracted from the assembly FASTA via GFFread (v0.12.2) were classified using the Conserved Domains Database (CDD), Pfam, and Superfamily databases (El-Gebali et al., 2019; Marchler-Bauer et al., 2015; Pandurangan et al., 2019). The Universal protein knowledgebase UniProt was implemented as additional database to all annotation and subsequent protein analyses (The UniProt Consortium 2025).

Plasmid typing was determined by comparing paralogous gene families (PF32, PF49, PF50, and PF57/62) against well-annotated reference genomes of various Borrelia spp. (Casjens et al., 2012). Analysis was done by extracting all ORF’s, including pseudogenes identified by PGAP, from the assembly and from reference genomes (B. burgdorferi B31, B. afzelii PKo, B. duttonii Ly, B. hermsii HS1, B. miyamotoi CT13–2396, B. miyamotoi FR64b, and B. miyamotoi LB-2001), clustering these sequences at a 90 % identity threshold using CD-HIT (v4.8.1), and integrating the results with the HMM profiles constructed for all reference genomes and InterProScan analyses to assign protein family identities.

To standardize replicon orientation, each replicon was reoriented so that the PF57/62 plasmid partitioning gene appeared in the positive-sense, except for the chromosome, which does not contain the PF57/62 gene (Miller et al., 2013; Wilder et al., 2016). Following reorientation, the genome was re-annotated with PGAP and reanalyzed via InterProScan to ensure that the orientation adjustments were fully integrated into the final annotation. All subsequent analyses were done on the re-oriented genome.

2.10. Expression locus identification

To identify the vmp expression locus in the B. persica LMU-C01 genome, we first determined which plasmid corresponded to the reported lp23 expression plasmid of B. duttonii and B. recurrentis (Lescot et al., 2008). Whole-plasmid alignments were performed using progressiveMauve (v2.4.0), comparing all assembled B. persica plasmids with annotated vmp genes against the lp23 plasmid of B. duttonii (GenBank accession number CP000980.1). The conserved region shared between LP90 and B. duttonii lp23 was extracted and examined for known expression-locus signatures characteristic of Old World RF Borrelia, most notably a long poly-T tract associated with the active expression site (Vidal et al., 2002). As genome annotation found the vmp gene within this region (pgaptmp_001220) encoded on the negative strand, the corresponding genomic segment was reverse-complemented to generate a sequence oriented 5′ to 3′ in the direction of transcription.

Within the oriented sequence, we searched for hallmark promoter and translation initiation features of the vmp expression locus, including the long poly-T tract, sigma 70 like promoter consisting of -35 (TTGACA-like) and -10 (TATAAT-like) motifs located downstream of the poly-T tract and a ribosome-binding site (AGGAGG-like) approximately 8–12 bp upstream of the translational start codon (Vidal et al., 2002). Motif searches were performed using custom Python scripts (Python v3.9) and genomic coordinates were validated using samtools faidx (v1.17) (Li et al., 2009).

To identify the upstream homology sequence (UHS), a contiguous sequence block located immediately downstream of the promoter and upstream of the vmp coding sequence was extracted from the confirmed expression locus. This putative UHS was then used as a nucleotide query in BLASTN searches (BLAST+ v2.13.0) against the complete B. persica genome. To assess whether UHS copies were associated with other silent vmp cassettes, BLAST hits were intersected with annotated vmp coding sequences using BEDTools intersect (v2.30.0) (Quinlan and Hall, 2010). Distances between UHS hits and vmp CDS start positions were calculated, confirming that UHS elements consistently localize upstream of vmp genes across multiple plasmids.

2.11. Functional annotation by COG classification

Annotated proteins were classified to clusters of orthologous groups (COG) by comparison to previously defined COG using RPS-BLAST tool (E-value 0.01) against the COG database (Camacho et al., 2009; Tatusov et al., 2000).

2.12. Identification and phylogeny of vmp genes

Vmp genes identification was done by extraction of all ORF’s from B. persica LMU-C01 genome and subsequent analysis using InterProScan (v5.57–90.0) with a search against Pfam and SUPERFAMILY databases for Vlp (IPR000680 Borrelia_lipo, PF00921) and Vsp (IPR001800 Lipoprotein_OspC, PF01441) integrated with the results of HMM profiles constructed for vmp genes of other RF Borrelia spp. including B. duttonii Ly, B. hermsii HS1, B. miyamotoi CT13–2396, B. miyamotoi FR64b, and B. miyamotoi LB-2001 as reference genomes.

A phylogenetic tree was constructed with coding genes amino acid sequences using IQ-TREE (v2.4.0) following sequence alignment using Muscle (v5) (Edgar, 2004; Minh et al., 2020).

2.13. Structural and functional analysis of Vlp proteins

Vmp amino acid sequences were extracted from the B. persica LMU-C01 genome and from 12 other RF Borrelia species including B. crocidurae, B. coriaceae, B. recurrentis, Borrelia venezuelensis, Borrelia nietonii, B. turicatae, B. parkeri, B. miyamotoi, B. hispanica, B. hermsii, B. duttonii and B. puertoricensis. Vmp protein sequences were analysed to predict and generate high confidence three-dimensional structure using AlphaFold (v1.5.2) with 10 recycling iterations, generating 3 models per sequence, early stop tolerance of 0.5 and PTM based ranking (Jumper et al., 2021). Vmp proteins biological activity, as well as variable region analysis were done using Foldseek (v10) by applying a search against all available databases (BFVD 2023_02, AlphaFold/UniProt50 v4, AlphaFold/Swiss-Prot v4, AlphaFold/Proteome v4, BFMD 20,240,623 CATH50 4.3.0, MGnify-ESM30 v1, PDB100 20,240,101 and GMGCL 2204) for biological function identification and by FoldMason Multiple Structure Analysis (MSA) for variable regions analysis (van Kempen et al., 2022). In addition, the Vmp protein orientation within the spirochetal membrane was determined using the Positioning of Proteins in Membranes (PPM) database by the assessment of the energetic favorability of a protein position within the lipid bilayer, considering factors such as the distribution of hydrophobic and hydrophilic residues, membrane thickness, and the protein three-dimensional structure (Lomize et al., 2022, 2012). Vmp proteins visualization and hydrophobic residues parameters analyses were done using ChimeraX (v1.9) (Pettersen et al., 2021). Furthermore, immune epitope properties of the common variable region (CVR) found in all Vlp proteins were analysed using the B cell epitope prediction tool in the Immune Epitope Database (IEDB) (Vita et al., 2019).

3. Results

3.1. B. persica LMU-C01 genome comprises 14 distinct replicons and remains stable during prolonged in vitro cultivation

To aid in studying the genome assembly of plasmids, we performed pulsed-field gel electrophoresis (PFGE) of genomic DNA isolated from two culture passages of the B. persica LMU-C01 strain passaged 6 and 103 times, which showed the separation of the genome into 9 linear segments and the appearance of high molecular weight bands representing the presence of circular plasmids (Fig. 1) (Porcella et al., 2005). Co-migration of the chromosome and the large megaplasmid is assumed, as well as for equivalent size fragments (<1 Kbp difference) such as the linear plasmids (LPs) LP46, LP45 and LP26, LP25 (Kneubehl and Lopez, 2023). Interestingly, no visible changes were noticed between the two different passages, suggesting low genome structural changes within the strain, even after one year of continuous in-vitro cultivation (Fig. 1). While PFGE cannot detect plasmid rearrangements, the absence of band loss across replicons over 12 months is inconsistent with the plasmid loss commonly seen in Lyme Borrelia (Biškup et al., 2011).

Fig. 1.

Pulsed-field gel electrophoresis (PFGE) of B. persica strain LMU-C01. The numbers to the left of the gel image correspond to DNA size in kilobases. P6 and P103 correspond to the culture passage number. Chr/LMP correspond to the mutual migration of the chromosome and linear megaplasmid, respectively. CP refers to circular plasmids.

The whole-genome sequencing of B. persica strain LMU-C01 was performed using both the PacBio and Illumina platforms, yielding a high-quality assembly. PacBio sequencing generated 1.8 Mb subreads (96.84 % mapped) with alignments averaging 5.2 Kbp in length and 13.4 Kbp polymerase reads (median 107 Kbp). A subreads length cutoff of 8 Kbp was applied during preassembly. For preassembly (assuming a 1.75 Mb genome), filtering produced 24.4 Kbp subreads (median 7.2 Kbp) totaling 175 Kbp (100 × coverage). Illumina sequencing provided an average coverage depth of 209.56 × with 99.9 % coverage breadth. The final assembly comprised 14 contigs totaling 1,647,793 bp (range 17–950 Kbp) with no gaps. Genome completeness was assessed as 99.2 % with only 0.4 % fragmented and 0.4 % missing, underscoring the robustness of the assembly.

To further validate the presence of 14 different replicons without technical assembly induced redundancy, the Pyani tool was used to assess plasmid uniqueness within the B. persica LMU-C01 genome. Redundant or duplicated contigs were defined to have <0.5 % nucleotide difference across the entire length. The resulting analysis showed distinct blocks of high identity only along the diagonal, with all off-diagonal signals remaining below the redundancy threshold, supporting 14 distinct replicons (Fig. 2).

Fig. 2.

Hadamard‐based heatmap created by Pyani analysis of the 14 putative replicons identified in the B. persica LMU-C01 genome. Each row and column corresponds to one contig, with hierarchical clustering shown at the top and left. The color scale indicates the degree of pairwise identity. Red diagonal blocks reflect high within‐contig identity, while minimal off‐diagonal signals indicate that each contig is distinct and not an assembly‐induced redundancy.

The final composition of the B. persica LMU-C01 genome consisted of one chromosome and thirteen extrachromosomal DNA segments (plasmids) with a total length of 1.65 Mbp and total gene content of 1452 (Table 1). Out of 13 plasmids, one was found to be circular in both SMRT and Circlator analyses (Hunt et al., 2015), whereas the remaining 12 were found to be linear plasmids. The average GC content was 29.4 % (range 25.5–31.7 %), similar to the GC content of other RF Borrelia species (Barbour, 2014). All tRNA and rRNA genes were found on the chromosome.

Table 1.

Borrelia persica LMU-C01 genome composition and gene content across replicons.

Contig name	Length (bp)	Topography	CDS (no.)	tRNA (no.)	rRNA (no.)	Gene (no.)	Pseudogene (no.)	GC_Content ( %)
Chromosome	951,069	Linear	856	32	3 (5S, 16S and 23S)	870	23	28.53
LP165	164,843	Linear	147	0	0	146	1	28.63
LP111	111,003	Linear	107	0	0	83	24	30.45
LP90	89,970	Linear	94	0	0	73	21	30.9
LP67	67,813	Linear	89	0	0	78	11	29.01
LP46	46,458	Linear	28	0	0	23	5	28.7
LP45	45,809	Linear	37	0	0	21	16	30.62
LP31	31,273	Linear	28	0	0	16	12	30.67
CP29	29,103	Circular	44	0	0	38	6	27.88
LP27	27,740	Linear	31	0	0	31	0	31.78
LP26	26,035	Linear	24	0	0	15	9	31.78
LP25	25,579	Linear	22	0	0	12	10	31.51
LP17_1	17,907	Linear	25	0	0	25	0	25.55
LP17_2	17,084	Linear	22	0	0	21	1	26.97

CDS, Coding DNA Sequence; tRNA, transfer RNA; rRNA, ribosomal RNA; LP, linear plasmid; CP, circular plasmid

Analysis of the terminal regions of the assembled chromosome and plasmids resulted in the identification of 5 terminal nucleotide sequences patterns, known as the hairpin telomeric sequences in Borrelia spp. in 10/13 linear replicons (Fig. 3) (Barbour and Garon, 1987; Casjens et al., 1997; Hinnebusch et al., 1990; Kuleshov et al., 2020). The well conserved “box 3″ was found in 10/10 segments, whereas plasmid typing based on the presence of “box 1″ patterns showed the presence of “box 1″ type 1 pattern (TATAAT) in LP165 and LP45 and the absence of “box1” (type 3) in the chromosome, LP111, LP90, LP46, LP31, LP27, LP26 and LP25. Furthermore, the more rarely reported short patterns “box 2, 4 and 5” were found in 10/10 segments (Casjens et al., 1997). In the three plasmids in which no telomeric sequences were found, no breaking point was present in the dot plot analysis, and no palindromic sequences were found using raw PacBio reads as well. The presence of intact telomeres demonstrated the relatively high completeness of the B. persica LMU-C01 genome assembly.

Fig. 3.

Telomeric sequences of the chromosome and linear plasmids of B. persica LMU-C01. The sequences are oriented with the hairpin bend positioned to the left. The sequence motif "box 1″ is highlighted by purple color and marked 1 in the top of the figure. The sequence motif "box 2″ is highlighted by red color and marked 2 in the top of the figure. The sequence motif "box 3″ is highlighted by blue color and marked 3 in the top of the figure. The sequence motif "box 4″ is highlighted by green color and marked 4 in the top of the figure. The sequence motif "box 5" is highlighted by yellow color and marked 5 in the top of the figure. Plasmid typing based on the presence of “box 1″ pattern marked at the right side of the figure. LP refers to linear plasmid.

3.2. B. persica chromosome shows high level of conservation, whereas plasmids are enriched in partitioning and recombination-associated genes

Classification of annotated proteins into clusters of orthologous groups (COG) showed high degree of conservation of the chromosome, with 81.4 % of proteins assigned to 19 COG categories. Plasmid proteins were classified in 9/13 segments with classification percentage range of 7.8–28 %. High percentage of COG classification was found on LP17_2, LP17_1, LP, LP27 and LP90 with 28,27.2,19.3 and 18 % of the proteins classified, respectively. These plasmids, having a high percentage of COG-classified proteins, are thought to carry vital genes which encode proteins involved in basic metabolic processes. These results are consistent with knowledge about other relapsing fever Borrelia spp., where the chromosome is enriched for core COG functions, whereas the plasmids contain surface lipoprotein paralogs (Vmps) and plasmid maintenance/partition loci (PF32, PF49, PF50, PF57/62) (Kneubehl et al., 2022).

Plasmid mapping was based on the identification of partitioning genes related to plasmid replication and maintenance. Five Borrelia paralogous gene families of putative partitioning genes were previously found: PF32, PF49, PF50 and PF57/PF62, which were grouped together under the assumption that they have similar function and are specific to Borrelia genomes (Casjens et al., 2012; Chaconas and Kobryn, 2010). Partitioning gene analysis showed the presence of PF genes in the B. persica LMU-C01 chromosome and all plasmids. More than one copy of PF genes was found in two plasmids, suggesting historical plasmid fusion events (Fig. 4) (Talagrand-Reboul et al., 2018).

Fig. 4.

Overview of plasmid repertoire and partitioning gene approximate locations. This figure illustrates the relative sizes of plasmids present in the B. persica LMU-C01 strain. Plasmids are designated based on topology, using “LP” for linear plasmid and “CP” for circular plasmids, and size (rounded to the nearest kb), with sequential numbering applied when sizes were identical. Replicons were plotted proportionally according to assembled size, and a 10 kilobase scale bar is shown. The diagram shows the estimated positions and orientations of the plasmid partitioning genes (PF32, PF49, PF50, and PF57/62), with arrows pointing right indicating a positive sense orientation and those pointing left indicating a negative sense orientation. Created with BioRender.com.

Additional analyses were done to search for common reccombination involving elements such as tRNA’s and recombinase genes which can be accosiated with antigenic variation mechanism (Del Val et al., 2019; Pratt-Hyatt et al., 2006). We found 32 tRNA’s located exclusively on the chromosome, and 5 recombinase genes located on the chromosome, and the LP67, CP29 and LP17_1 plasmids.

3.3. B. persica LMU-C01 genome harbors 48 vlp genes and one vsp gene, and its Vlps cluster into four subfamilies

Interproscan analysis integrated with the Uniprot database and HMM profiles of vmp genes from other Borrelia spp. (Conserved Domains Database (CDD), Pfam, and Superfamily databases) revealed the presence of 98 vmp genes in the B. persica LMU-C01 genome (Fig. 5). Of these, 48 vlp genes and 32 pseudogenes were found, as well as one vsp gene and 17 pseudogenes. Translated vlp genes created an average amino acid sequence length of 352.6 (range 323–433 amino acids) and a vsp translated gene of 247 amino acid sequence. Vmp genes were found in 10/13 linear segments and were not found in the circular plasmid. Plasmid LP90 appears to have a significant role in the pathogenesis and virulence of the B. persica LMU-C01 strain involving immune escape strategies, by containing 16/48 (33.3 %) of complete vlp genes (Barbour, 2016).

Fig. 5.

Approximate locations of vmp genes in B. persica LMU-C01 genome. Replicons were plotted proportionally according to assembled size, and a 10 kilobase scale bar is shown. The replicons are named based on topology (LP = linear plasmid, CP = circular plasmid) followed by the size to the nearest kilobase. The diagram shows the estimated positions and orientations of the vmp genes (vlp and vsp), with arrows pointing right indicating a positive sense orientation and those pointing left indicating a negative sense orientation. The vmp in the expression site is indicated by an asterisk within the indicated ORF. Created with BioRender.com.

Phylogenetic analysis of the 48 vlp genes, based on amino acid sequence alignments, showed the clustering of B. persica LMU-C01 Vlp proteins into 4 previously described sub-families: alpha (11 genes), beta (18 genes), gamma (13 genes) and delta (6 genes; Fig. 6) (Dai et al., 2006; Hinnebusch et al., 1998). Of these, three sub-families clustered within each other with strong support, but the Vlp-β sub-family was divided into four clades, according to the closest reference sequence, with a relatively weak bootstrap support (<70 %).

Fig. 6.

Phylogenetic diversity of Vlp proteins in the B. persica LMU-C01 genome. Different node colors indicate different Vlp-subfamilies: red - Vlp-α, black - Vlp-β, blue - Vlp-γ and green - Vlp-δ. The tree was constructed based on multiple sequence alignment of amino acid sequences extracted from vlp genes which contain domains corresponding to the Borrelia lipoprotein family (IPR000680 and PF00921). Protein sequence alignment was done using Muscle (v5). The tree was constructed using IQ-TREE using the LG+I+G model and 1000 bootstrap options and visualized using iTOL (v7.5.1). Values higher than 70 % are indicated.

3.4. The putative vmp expression locus in B. persica is located on plasmid LP90

The putative expression locus of the active vmp gene was identified based on the known expression locus in other Old World RF species, e.g. B. duttonii and B. recurrentis (Fig. 7, Appendix C). Progressive mauve alignment of the known expression plasmid lp23 of B. duttonii and B. recurrentis and all linear plasmids of B. persica LMU-C01 with annotated vmp genes resulted in the identification of a 23 kb shared block located on the LP90 plasmid of B. persica starting approximately at position 68 kb (Appendix D). Analysis of the shared 23 kb block resulted in the identification of a 121 bp poly-T tract, followed by a downstream promoter -35 and -10 motifs with 16 bp spacers. A repetitive homologous sequence was found downstream to the promoter motifs in the expression locus and in 36 other vmp genes 0–100 bp upstream to the start codon position (Appendix C). This repetitive segment shares 82–100 % identity over 55–103 bp length between different vmp genes and therefore can be considered as part of the putative upstream homologous sequence (UHS). A ribosomal binding site (RBS) was found 10 bp upstream to the start codon of an annotated vlp gene.

Fig. 7.

Putative vmp expression locus in the Borrelia persica LMU-C01 genome. The expression locus located on plasmid LP90 is presented in a positive strand orientation (5′ to 3′). The expression locus is defined by a long poly-T tract followed by Sigma 70 like motifs (-35 and -10). A repetitive segment thought to be part of the putative UHS is located between the promoter region and the ribosome binding site (RBS) which located 10 bp upstream to the start codon of vlp-1220 gene. Created with BioRender.com.

3.5. Structural modeling supports a conserved Vmp structure and a shared position of the common variable region

The 3rd dimensional structure of all Vmp proteins from the B. persica LMU-C01 genome was successfully predicted using the AlphaFold platform (Fig. 8A-E). The Vmp protein structure consists of four long coiled coil stalks with parallel α-helices extending towards a top globular helical domain and bottom membrane anchors (Fig. 8F-G). The Vmp proteins consist of mainly hydrophilic residues, as shown in the surface hydrophobicity distribution diagram (Fig. 7H). The overall structure of the Vmp proteins is of anchored, outer membrane proteins, as would be expected for proteins that contribute to antigenic variability (Sparling et al., 1986).

Fig. 8.

Predicted 3rd dimensional structure of B. persica LMU-C01 Vmp proteins. One representative from each Vlp subfamily (A, alpha; B, beta; C, gamma; D, delta) and the Vsp protein (E) are presented. Protein structures were modeled using AlphaFold (v1.5.2) and visualized using ChimeraX (v1.9). Membrane orientation and surface rendering were modeled using PPM database. Different colors correspond to the pLDDT score (dark blue, very high pLDDT>90; light blue, high 70<pLDDT<90; yellow, moderate 70>pLDDT>50; orange, low pLDDT<50). The membrane attachment properties of the Vsp (F) and a Vlp representative (G) are shown. The two-colored plates represent the double layered bacterial membrane where the blue plate corresponds to the outer membrane. Hydrophobic properties of the Vmp proteins represented by Vlp alpha protein are presented on the surface structure of the protein (H). Light blue represents hydrophilic residues, whereas yellow color represents hydrophobic residues.

Multiple structural analysis (MSA) of the B. persica LMU-C01 Vlp proteins, revealed the presence of 4–5 variable regions of 5–10 amino acid length within every Vlp sub-family. Variable regions located in regions with primary pLDDT<90 were excluded. Nevertheless, only one variable region of 8–12 amino acids, designated as the Common Variable Region (CVR) was consistently mapped to a similar predicted position within the hydrophilic top globular helical structure in all Vlp proteins (Fig. 8; A1-A6). Interestingly, the CVR shows amino acid sequence variation both within and between Vlp sub-families but when extracting and isolating its specific structure within the protein, it retained a similar predicted position across all Vlp proteins (Fig. 9; B1). Furthermore, to assess confidence in the putative CVR region, AlphaFold confidence metrics were analyzed for residues overlapping the CVR region across all predicted Vlp structures. The CVR associated residues showed high local confidence, with a mean regional pLDDT of 95.09, and a mean predicted aligned error (PAE) of 2.51 Å. Furthermore, an average of 99.72 % of residues showed pLDDT score above 70. In contrast, although located in similar amino acid sequence positions, other variable regions did not show a similarly shared predicted spatial placement across the Vlp proteins (Fig. 9; B2).

Fig. 9.

Common variable region (CVR) representation across six different Vlp proteins of B. persica LMU-C01 strain (A1–6). Different colors represent the different Vlp subgroups (green, alpha; light blue, beta; purple, gamma; orange, delta). CVR are highlighted by green whereas the rest of the protein is colored by red. Variable regions were defined from sequence comparison within Vlp subfamilies, and regions with primary pLDDT < 90 were excluded from the structural comparison. After extraction and isolation of the CVR from all Vlp subfamilies, it retains a similar predicted position (B-1), in comparison to other variable region where no predicted position is preserved (B-2). Protein structures were modeled using AlphaFold (v1.5.2) and CVR visualization and superposition were done using ChimeraX (v1.9).

The presence of the CVR was found also in 12 other RF Borrelia spp., in which 420 Vlp proteins were analysed and compared to the findings revealed in B. persica LMU-C01. Morover, the presence of the CVR was found in a similar predicted position relative to the globular domainas in B. persica LMU-C01 and separate from other variable regions (Fig. 10). The CVR position comparison was conducted after the 3rd dimensional structure of the 420 Vlp proteins was predicted and regions of pLDDT<90 were excluded.

Fig. 10.

Common variable region (CVR) across different RF Borrelia species in comparison to B. persica LMU-C01 strain showing a similar predicted structure. Protein structures were modeled using AlphaFold (v1.5.2), and regions with primary pLDDT < 90 were excluded before comparison. Four out of 12 RF species are represented together with B. persica LMU-C01 Vlp representative (A, B. turicatae; B, B. miyamotoi; C, B. hermsii; D, B. duttonii). CVR visualization and superposition were done using ChimeraX (v1.9).

To investigate the potential involvment of the CVR in virulence and pathogenicity, the Vlp amino acid sequence epitope characteristics were analysed using a well established epitope prediction tool of the Immune Epitope Database (IEDB) with a threshold score of 0.6. Results showed the presence of a main epitope of approximately 40 amino acids across all the Vlp proteins, and minor epitopes of 10–30 amino acid length. The CVR was found to be part of the main epitope and covers approximately 1/4 of its length, suggesting a significant role of the CVR in the antigenic variation mechanism of RF Borrelia species (Fig. 11).

Fig. 11.

Illustration of a CVR-based antigenic variation event in Borrelia persica. A blood vessel containing Borrelia spirochetes and antibodies that target outer membrane proteins (A). Antibodies attach to Vmp proteins located on the spirochete outer membrane (B). The CVR is recognized by specific antibodies (yellow portion of the protein). Antigenic variation event occurs and the CVR is changed (red portion of the protein) and is no longer detected by the previously neutralizing antibody (C).

3.6. Functional analysis identifies no structural homologs outside Borrelia Vmp proteins

Screening of the Vmp proteins extracted from the B. persica LMU-C01 genome against all databases included in the Foldseek platform to search for potential bacterial biological funcion, beside host immune evasion mechanism, resulted without any match to other proteins except for Vmp’s of different Borrelia species. It is plausible though, that the biological function of the Vmp proteins is limited solely to their virulence properties in host immune response evasion by antigenic variations (Bankhead, 2016; Stone and Brissette, 2017).

4. Discussion

Although B. persica is one of the most widespread RF species causing human infection (Tang et al., 2024), its genome and the structural basis of its antigenic variation mechanism have remained undefined. In this study, we identified a unique common variable region within the Vlp protein that is retained in a conserved predicted position on the protein surface, regardless of amino acid sequence diversity between different species or subfamily classification. This 8–12 amino acid region, located within the hydrophilic top globular domain of the Vlp protein structure, provides new insights for understanding antigenic variation in RF Borrelia spp. The conservation of the CVR across 420 Vlp proteins from twelve different RF Borrelia species suggests a fundamental feature of the antigenic variation system that has been preserved throughout RF Borrelia evolution. CVR orientation confidence analyses showed a relatively low mean regional PAE of 2.51 Å and a high mean regional pLDDT of 95.09, suggesting that the putative CVR orientation is locally reliable in the majority of Vlp models (Guo et al., 2022). A previous study that examined the structural properties of the Vmp-like sequences (VlsE) of LD Borrelia spp. revelaed minimal global structure variation, although extensive amino acid variation occurred (Norris and Brangulis, 2024). Nevertheless, previous studies focused on RF Borrelia species established the "stem and dome" model in B. turicatae (Cadavid et al., 1997). The Vsp family in B. turicatae is defined by highly conserved N- and C-termini that form an inner helical "stem" proximal to the spirochete membrane, while the variable central residues constitute a distal "dome" exposed to the host environment. Rather than exhibiting random variation throughout the protein, four specific variable regions (VR1-VR4) separated by conserved intervening motifs that maintain the overall fold were found (Cadavid et al., 1997; Lawson et al., 2006). The localized nature of this variation allows a significant functional shift, such as the differential tissue tropism observed between B. turicatae Vsp1 and Vsp2. While Vsp1 expression is linked to neurotropism and central nervous system invasion, Vsp2 is associated with high-density spirochetemia and affinity for glycosaminoglycans (Lawson et al., 2006). Our structural analysis of the B. persica and 12 other RF species Vlp family reveals a robust architectural parallel to the "stem and dome" model which maintains a stable global conformation despite a primary variable region. These findings suggests that the Vlp family also balances structural integrity with the localized flexibility required to evade immune response pressure.

In the present study, the conserved spatial orientation of the described CVR, despite considerable sequence variation, suggests its pivotal role in antigenic variation and immune evasion, a finding that may contribute to our understanding of RF Borrelia pathogenesis. Our analyses suggest that this CVR, comprises approximetely 25 % of the main epitope within the Vlp structure, supports its important role as a key mediator of antigenic variation and immune escape. Moreover, the lack of detectable homologs for Vmp proteins outside Borrelia spp., and the failure to assign additional biological functions, reinforce the idea that these proteins are adapted to virulence and host immune response modulation. Our results suggest that RF Borrelia spp. have evolved a sophisticated system where targeted variation within a spatially conserved region can generate sufficient antigenic diversity to evade host immune responses while maintaining protein structural integrity. A study that examined the role of Vmp proteins in the relapse phenomenon in mice found that vmp knockout strains of B. hermsii showed impaired infectivity and lack in relapse induction (Raffel et al., 2014). It is therefore plausible that RF Borrelia spp. evolved unique virulence proteins that focus on antigenic variation as a defense mechanism to avoid the vertebrate host’s immune response. An illustration of a CVR based antigenic variation mechanism, according to the findings in the current study is shown in Fig. 11.

The 1.7Mb genome of B. persica LMU-C01 was characterized in this study by the integration of both PacBio long reads and Illumina short reads sequencing methods with various assembly and genome analysis tools. This genome is comprised of one linear chromosome, 12 linear and one circular plasmid, consistent with the typical genomic organization of relapsing fever Borrelia species (Casjens and Huang, 1993; Ferdows and Barbour, 1989). Most linear replicons share the unique telomeric structure containing short palindromic sequences providing binding sites to important replication mechanism enzymes such as the telomer resolvase ResT (Tourand et al., 2007, 2003). The presence of these telomeric sequences is a hallmark of Borrelia spp. genomic architecture, a feature which is rare among bacteria, and shared only with a single soil bacterium species, Agrobacterium tumefaciens, and some bacteriophages (Barbour and Garon, 1987; Casjens et al., 1997; Goodner et al., 2001; Kuleshov et al., 2020; Rybchin and Svarchevsky, 1999). The replication of these linear elements initiates centrally and proceeds bidirectionally until a dimer junction forms at the telomeres, where a specialized enzyme, the telomere resolvase, mediates the precise cleavage and hairpin resolution necessary for generating mature plasmid termini (Chaconas, 2010; Picardeau et al., 1999). Both Lyme disease (LD) and RF Borrelia spp. possess conserved palindromic sequences at the ends of their linear replicons, which not only facilitate the formation of these hairpin structures but also serve as binding sites for the telomer resolvase enzyme, and also serve as molecular markers for plasmid identification and validation (Casjens et al., 1997; Kuleshov et al., 2020). Genome annotation revealed a rich array of vmp genes localized exclusively on linear plasmids, including 48 complete vlp genes and one vsp, with LP90 emerging as a putative hotspot for vmp-mediated immune evasion.

In this study, we demonstrate the architecture of the vmp expression locus in the B. persica LMU-C01 isolate, providing the first characterization of this key antigenic variation feature in this species. The expression locus identified on plasmid LP90 mirrors the organization described in B. duttonii and B. recurrentis, including a long upstream poly-T tract, a Sigma-70 like promoter with conserved 35 and -10 motifs, and a single expressed vmp gene encoded on the negative strand (Lescot et al., 2008; Vidal et al., 2002). This conserved order supports a shared evolutionary origin and mechanistic framework for antigenic variation among Old World RF borreliae. In contrast, the vmp expression locus in New World RF species including B. hermsii and B. turicatae exhibits distinct features lacking a poly-T tract, with large variablity in other promoter motifs and a well characterized short (∼60 bp) UHS spaning the start codon (Dai et al., 2006; Stoenner et al., 1982). The UHS was not characterized in the same manner in Old World RF species and it is therfore plausible to find lineage variations in other regulatory expression features. We found that B. persica exhibits a distinct extended spacer separating the promoter elements from the ribosome binding site. This structural decoupling suggests that the strict UHS constraints observed in North American species are not universal, pointing instead to a divergent regulatory evolution in Old World lineages. Furhtermore, variations were also described in the expression locus position on the telomeric region of the linear expression plasmid. Althogh previous studies reported a conserved telomeric positioning of the expression locus in B. hermsii (Dai et al., 2006), other studies reported variations in this paradigm, with diverse positioning of the expression locus on internal segments of the plasmid, as shown in B. coriaceae, B. peurtoricensis, B. parkeri and B. turicatae (Kneubehl and Lopez, 2023). The presence of the expression locus of B. persica LMU-C01 on the negative strand reflects the initial genome reorientation based on the putative partitioning genes analysis. These findings position B. persica within the Old World RF antigenic variation paradigm, while revealing lineage-specific refinements to expression locus architecture. The precise mapping of the expression locus provides a critical framework for future functional studies addressing the regulation, recombination dynamics, and evolutionary constraints of antigenic variation in RF spirochetes.

In evaluating the molecular foundations of antigenic variation in B. persica, we investigated the presence of recombination-associated elements, such as tRNAs and recombinase genes, which are often linked to the genetic shuffling mechanisms in other Borrelia species (Pratt-Hyatt et al., 2006). The absence of these elements in proximity to the vmp loci suggests that there are other molucular mechanisms for the reccombination of silent vmp cassettes into the single expression site. These may include other unique enzymatic processes with nuclease or recombinase activity, as found in other infectious agents which use antigenic variations as a key virulence factor, such as the RAD51 enzyme of Trypanosoma brucei (McCulloch and Barry, 1999).

The structural computational analysis of Vmp proteins did not reveal additional biological functions beyond immune evasion, suggesting that these proteins are highly specialized for virulence rather than metabolic processes.

While this study provides comprehensive genomic and structural analysis of the RF species B. persica, several limitations should be noted. Future experimental studies are warranted to confirm the functional significance of the CVR in immune evasion and to determine whether targeting this region could serve as a therapeutic or preventative target. The identification of a single expression site requires further investigation to fully understand the molecular mechanisms of antigenic variation in B. persica.

5. Conclusions

This study represents the first comprehensive genomic analysis of B. persica and provides insights into the organization and evolution of RF Borrelia genomes. The discovery of the CVR in Vlp proteins combined with its apparent maintenance of the genomic framework, advances our understanding that RF Borreliae retain a relative stable core architecture alongside with balanced, high frequency antigenic diversity. The CVR represents a novel paradigm for immune evasion that may have broader implications for understanding host-pathogen interactions in antigenic variability systems. These findings provide a foundation for future research into RF Borrelia pathogenesis and may contribute to the development of improved diagnostic and therapeutic approaches for tick-borne RF.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

This Whole Genome project has been deposited at DDBJ/ENA/GenBank under the accession JBTFHO000000000. The version described in this paper is version JBTFHO010000000. Supporting data for this article is presented in the main manuscript or additional supplementary material.

CRediT authorship contribution statement

Dor Shwartz: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review & editing. Yaarit Nachum-Biala: Conceptualization, Methodology, Project administration, Validation. Reinhard K. Straubinger: Investigation, Supervision. Ofir Segev: Formal analysis, Software. Asaf Levy: Methodology, Supervision. Job E. Lopez: Investigation, Writing – original draft. Gad Baneth: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.