Mechanism of Borrelia immune evasion by FhbA-related proteins

Immune evasion facilitates survival of Borrelia, leading to infections like relapsing fever and Lyme disease. Important mechanism for complement evasion is acquisition of the main host complement inhibitor, factor H (FH). By determining the 2.2 Å crystal structure of Factor H binding protein A (FhbA) from Borrelia hermsii in complex with FH domains 19–20, combined with extensive mutagenesis, we identified the structural mechanism by which B. hermsii utilizes FhbA in immune evasion. Moreover, structure-guided sequence database analysis identified a new family of FhbA-related immune evasion molecules from Lyme disease and relapsing fever Borrelia. Conserved FH-binding mechanism within the FhbA-family was verified by analysis of a novel FH-binding protein from B. duttonii. By sequence analysis, we were able to group FH-binding proteins of Borrelia into four distinct phyletic types and identified novel putative FH-binding proteins. The conserved FH-binding mechanism of the FhbA-related proteins could aid in developing new approaches to inhibit virulence and complement resistance in Borrelia.

Introduction initial solution was obtained by molecular replacement with the published structure of FH19-20 (PDB 2G7I [35]) as a search model. We identified a single molecule of FH19-20 in the asymmetric unit with clear density for another polypeptide. Next, we assigned BhFhbA to the density by multiple rounds of manual building in Coot [36] and refinement with BUSTER [37]. The final solution showed a 1:1 complex of FH19-20 and BhFhbA in the asymmetric unit. The model was refined to a good overall geometry, and water molecules and ions were introduced to clear unassigned densities. The final model fits the observed diffraction data with the final R work /R free values of 19.7%/23.7% (Table 1), and all residues are in the most favored (98.0%) or allowed (1.65%) regions of the Ramachandran plot. The structure of BhFhbA (residues 44-202) reveals a compact, single-domain fold, which is composed solely of a bundle of nine α-helices ( Fig 1A). The first five α-helices (α1-α5) wrap around the core formed by α-helices 6-9 with the overall architecture of the fold resembling a curved arc or capital letter 'L'. Helices 6-9 are essentially perpendicular to the rest of the helices. α3, α6, α8 and α9 compose the inside face of the L, creating a cavity with a hinge-like loop on top, formed by the connecting region between α-helices 6 and 7.
We performed a structure-guided search from the Protein Data Bank to analyze the conservation of the protein. Interestingly, we could not find any proteins with a similar fold either from PDBeFold or the Dali server [38]. The structure of BhFhbA is thus a previously uncharacterized fold and further represents a novel FH-binding scaffold.
There are two possible FH19-20-BhFhbA interfaces in the crystal structure (S1 Fig). The primary and biologically relevant interface in the FH20 domain has a buried surface area of 1066 Å 2 , as determined by the PISA server [39]. The other interface, located at the tip of domain 20 of FH, arises from crystal packing. Further, the primary, biologically relevant interface contains more specific contacts between FH and FhbA, as demonstrated by the increased number of hydrogen bonds when compared to the other interface (18 vs. 6, S1 Table).
A careful analysis of the primary interface reveals that α-helices 6-9 of BhFhbA form a hydrophobic cavity in which FH20 sits (Fig 1B). Within the cavity, Trp1183 of FH20 surrounded by a sandwich like stack of aromatic residues sits between two BhFhbA residues: Phe154 from the hinge region and Tyr170 from helix α8. This tightly constrained binding pocket is further coordinated by the hydrophobic sidechains of α6 Met137 and Leu146 and α8 Ile171. Moreover, towards the C-terminus of FH20, Tyr199 from BhFhbA α9 forms van der Waals interactions with FH20 Val1200, with the closest approach being 3.8 Å (Fig 1B).
There are two crystal structures of microbial proteins complexed with FH19-20: OspE from Bo. Burgdorferi [22] and BhFhbA. We compared the binding mechanisms between the two proteins. Previous binding inhibition assays using 15 mutants of FH20 showed that the FH19-20 complex. BhFhbA forms a bundle of nine α-helices resembling an arc-like or L-shaped structure. α-helices 1-5 (in green) wrap around the core formed by α-helices 6-9 (in teal). The connecting loop of α-helices 7 and 8 create a hinge-like structure close to C-terminus of BhFhbA. Together the L-shaped cavity (α-helices 3, 8 and 9) and the hinge-loop coordinate the binding of FH20 (in blue). (b) Close-up of the binding site with key sidechains represented as sticks. FH20 Trp1183 and Val1200 (in blue, indicated by � ) coordinate binding by a group of hydrophobic residues in BhFhbA. Trp1183 of FH20 is inserted to the hydrophobic pocket formed by BhFhbA Phe154, Met137, I171 and Tyr170. Further, FH20 Val1200 is within van der Waals distance of BhFhbA Tyr170 and Tyr199. 2Fo-Fc density, contoured at 1 σ, is shown around the residues. https://doi.org/10.1371/journal.ppat.1010338.g001

PLOS PATHOGENS
Borrelia immune evasion binding-sites of BhFhbA and OspE overlap [30] , and that BhFhbA inhibited binding of OspE to FH19-20 and vice versa. We superimposed the FH19-20s of the two complex structures to analyze the similarities and differences in the binding. We identified five key amino acids in FH20 that form hydrogen bonds to both OspE and BhFhbA (S1 Table and S2A Fig). Thus, BhFhbA and OspE have completely unrelated folds, but nevertheless utilize the same surface and partially similar contacts to bind FH20 (S2A Fig). Comparison to the structure of FH19-20 in complex with sialic acid and C3d (PDBID: 4ONT [40]) revealed that the same surface on FH20 is also occupied by sialic acid and, moreover, that FH20 W1183 makes similar interactions with the sialic acid moiety as with the F154 of FhbA (S2B Fig). This supports our conclusion that the primary interface seen in the crystal structure is the biologically relevant one.

Mutagenesis and binding assays reveal important amino acids in the hydrophobic binding pocket
Next, we designed ten alanine point mutations located on the interface of BhFhbA that binds FH19-20. A group of mutations (Asn153Ala, Phe154Ala, Met137Ala, Ile171Ala) targeted at the hydrophobic binding pocket, whereas the rest (Phe85Ala, Asn88Ala, Lys91Ala, Glu178Ala, Phe181Ala, Glu198Ala) were targeted to the interface below the hydrophobic cavity (Fig 2A and 2B). The single alanine mutants of BhFhbA were expressed in E. coli as 6x-His fusion proteins and purified using Ni-NTA and size-exclusion chromatography.
We used affinity ligand binding immunoblot as an initial robust screen for the effects of mutations on the binding to FH19-20 (S3 Fig). Some mutations decreased binding, and one

PLOS PATHOGENS
Borrelia immune evasion (Phe154Ala) completely abolished it. Next, we used fluorescence-based MicroScale Thermophoresis (MST) to determine the binding affinities of the BhFhbA mutants to FH19-20 (Table 2 and S4 Fig). MST measurements showed~70-fold decrease in binding of the Phe154Ala mutant (K d = 6.0 μM) to FH19-20 compared to the wild type protein (K d = 0.082 μM) (Fig 2C and 2D). This was also confirmed by gel filtration chromatography, where no complex formed between the Phe154Ala mutant and FH19-20 (S5 Fig). In addition, we compared the binding of wild type BhFhbA and the Phe154Ala mutant to full length FH using MST. As with FH19-20, wild type FhbA bound FH with high affinity (K d = 30 nM) and Phe154Ala mutant showed no binding (S6A Fig). In the binding data, we observed a 12-fold and a 5-fold decrease in the affinities of two other mutants, Met137Ala and Ile171Ala for FH19-20, respectively (Table 2). In the complex, Trp1183 from FH20 is elegantly slide in between the coordinated aromatic binding stack of four hydrophobic residues Met137, Leu146, Phe154, and Ile171 (Fig 1B), providing a structural explanation for these biochemical results. However, most single alanine substitutions of charged residues had no effect on binding, suggesting that hydrophobic interactions dominate. To confirm the dominance of hydrophobic interactions further, we compared complex formation in high-salt conditions (PBS + 500 mM NaCl) and physiological environment (PBS) by gel filtration (S5 Fig). High ionic strength did not disrupt the binding of BhFhbA to FH19-20, supporting our proposal that the interaction is mainly hydrophobic, although involvement of other charged residues in the interaction cannot be excluded.
To exclude the possibility that the critical effect of the Phe154Ala mutation on binding to FH is due to protein misfolding, we performed CD-spectroscopy and demonstrated that wild type and Phe154Ala mutant proteins have identical secondary structure profiles, indicating that both are correctly folded (S7 Fig).

FhbA interacts with FH mainly via binding to FH20
B. hermsii was originally reported to acquire complement regulator FHL-1 from serum, as well as to bind a cloned fragment of FH containing domains 1-7 [31]. Domains 6-7 are identical in FH and FHL-1 and contain a binding site for several microbial proteins, like fHbp of Neisseria, CspA of Lyme disease borrelia and streptococcal M-protein [41]. We therefore decided to test whether BhFhbA binds this region by using an FH fragment of domains 5-7 (FH5-7).

Borrelia immune evasion
BhFhbA was immobilized to ELISA-plates, and purified FH5-7 or FH19-20 fragments were added. After washing steps, the binding of FH5-7 and FH19-20 was detected by a polyclonal anti-FH antibody. When compared to FH19-20, the FH5-7 fragment displayed very modest binding to BhFhbA (Fig 3A-3D). Moreover, FH5-7 did not affect the binding of FH19-20 to BhFhbA in a competition assay (Fig 3B). To analyze the binding interaction in another setup, we used MST to test if FH5-7 or FH19-20 affects the binding of full-length FH to BhFhbA in a competition assay. FH5-7 slightly decreased the binding of full-length FH to BhFhbA, but FH19-20 abolished it completely (S6B Fig).
These results suggest that BhFhbA binds FH predominantly via domain 20. We cannot exclude that under some circumstances, BhFhbA also interacts with FHL-1 or FH via domains other than 20, but the major binding site for BhFhbA is clearly FH domain 20.

PLOS PATHOGENS
Borrelia immune evasion

FhbA decreases complement activation and enhances serum survival
After entering the body, RF Borrelia can survive and multiply in the blood, and cause massive spirochetemia, where bacteria are present in the blood at high densities (10 5 −10 6 bacteria/ml). We mimicked natural conditions by incubating 50,000 live B. hermsii bacteria in 100 μl of whole blood treated with hirudin to prevent coagulation, and measured the amount of terminal complement complexes (TCC) as indicators of complement activation [42]. We first confirmed that, as expected, the presence of bacteria in blood increases complement activation, which is seen as an increase in the amount of TCC in the sample. Adding purified BhFhbA to the reaction to inhibit binding of FH to B. hermsii led to even higher levels of TCC compared to bacteria alone. Conversely, adding of the FH-binding defective mutant BhFhbA/Phe154Ala had no effect on the levels of TCC, which were similar to bacteria alone ( Fig 3E). Some enhancement of complement activation was also detected when BhFhbA was incubated in the absence of bacteria, whereas BhFhbA/Phe154Ala had no effect. We hypothesize that highaffinity binding of BhFhbA to FH may affect its ability to regulate complement in the fluid phase and/or can lead to formation of complement activating immune complexes.
We then expressed BhFhbA wild type and BhFhbA/Phe154Ala mutant on the outer membrane of E. coli to study how binding of FH affects bacterial survival. We chose E. coli as it is a Gram-negative bacterium, and laboratory strains lacking any evasion mechanisms are efficiently killed by complement [43]. We utilized the autotransporter adhesin involved in diffuse adherence-I (AIDA-I) system [44] (Fig 3F) to deliver the BhFhbA protein to the outer surface of E. coli and showed by immunoblot analysis that His-tagged BhFhbA was present in the outer membrane fraction (Fig 3G). We then performed serum sensitivity assays by utilizing BhFhbA wild type and mutant proteins (see Materials and Methods). Survival of the strain expressing wild-type BhFhbA was significantly (p<0.05) higher when compared to the control strain and to the strain expressing BhFhbA/Phe154Ala mutant ( Fig 3H). However, in the absence of complement, all three strains showed similar survival ( Fig 3I). These results show that binding of functional BhFhbA to FH is necessary and sufficient for increasing the survival of the bacteria.

A new family of immune evasion proteins revealed by bioinformatic searches
The crystal structure of BhFhbA inspired us to study the distribution of FhbA-like and other FH-binding proteins within the whole Borreliaceae family. First, we performed a thorough search for homologous proteins within all available whole genomes (in total 154) and separately deposited sequences from the family Borreliaceae (Table 3). Analysing sequence data from borrelia is demanding, as Borreliaceae have very complex genomes with both linear and circular plasmids [45] where length, diversity and composition vary between different species. It is thus possible that some FH-binding proteins might be absent from the databases due to plasmid loss, which has been reported in the Borreliella clade.
Our structure-guided sequence database analysis approach allowed us to reliably identify 10 species with sequences homologous to BhFhbA in RF borreliae clade, and 3 species in the LD group (Figs 4A and S8). A phylogenetic tree of these homologous proteins ( Fig 4B) shows that they form three different clusters. Nonetheless, in all 10 species of RF borreliae, the residue corresponding to Phe154, which is essential for FH-binding in BhFhbA, is conserved. In B. crocidurae, both sequenced strains (Achema and DOU) have stop codons in the signal sequence that alter the amino terminal regions of the proteins. It would be interesting to see if these truncated genes translate into functional proteins and provide protection like FhbA protein from B. hermsii.
The translated genomic regions of three species from the LD borreliae clade aligned well with other BhFhbA proteins, although with some changes (Fig 4A). In Bo. valaisiana, the protein is two helices shorter, but the hinge region and the key Phe154 residue mediating FHbinding are conserved. Bo. bissettiae (NC_015916.1) has a Leu instead of conserved key Phe, several stop codons, and a single nucleotide deletion that leads to a frame shift (marked by '?' in Fig 4A). The only sequence that seems not to be altered at the DNA level is from Bo. afzelii. Potentially it can be translated into functional protein, though it has Leu instead of conserved Phe, as in Bo. bissettiae. Further studies are required to test whether any of these genes are expressed and provide similar protection as FhbA.
Overall, it appears that FhbA-like genes are present in all RF clade species, except for B. anserina, and that the key binding loop containing Phe154 is conserved. However, although FhbA-like genes are identifiable at the DNA level also in the LD clade as well, it is not known if they encode functional proteins. Table 3. fhbA-related, cspA (CRASP-1/BBA68), cspZ (CRASP-2) and ospE genes reported in borreliae. '+' means that the gene has been identified at least in one source.'-'means that the sequence was not found in any of the sources used. Data were acquired from the sequenced genomes, published reports and individual sequences deposited to the databases. Relevant genes were searched as described in Materials and Methods. Accession numbers for FhbA-related proteins are listed in the legend to

Clade
Species Genes Borreliella valaisiana fhbA-related gene is detectable at the DNA level, but appears truncated if translated in silico 2 fhbA-related protein is not conserved in the FH-binding position Phe 154 3 cspZ described in [46] 4 ospE has 30% identity to the N-terminal half of another hypothetical protein BGP333 (GenBankID:AAU86184): the FhbA-related protein is truncated 5 Strain isolated from ticks [47] 6 FhbA-related protein is not found in the genome available, but protein has been detected 32 7 Causative agent of avian spirochetosis [48] https://doi.org/10.1371/journal.ppat.1010338.t003

PLOS PATHOGENS
Borrelia immune evasion

Identification of a conserved binding mechanism for FH binding
To examine if the BhFhbA binding mechanism we identified occurs in other members of this protein family, we cloned, expressed and purified a novel FhbA homologue from B. duttonii (BdFhbA). We chose this protein, because B. duttonii is known to bind FH [28]. BdFhbA protein has high conservation (41.8% identity) to BhFhbA (Fig 5A). We modelled the structure of BdFhbA with Phyre2 using our crystallized BhFhbA as a template, and observed an almost identical structure, except for shorter helix 3 and its connection to helix 4, as expected from the sequence alignment (Fig 5A and 5B). The location of the helices, the positioning of the hinge area, and the orientation of the critical residue Phe130 (corresponding to BhFhbA Phe154 in B. hermsii) were conserved. We next measured binding of FH19-20 to BdFhbA. Gel filtration experiment showed that the proteins form a stable complex (Figs 5C and S9). Moreover, MST experiments revealed that FH19-20 bound BdFhbA (Fig 5D), although with lower affinity (1.06 ± 0.38 μM) than to BhFhbA (Fig 2 and Table 2). The lower affinity might be due to small differences in the hydrophobic binding pocket. For example, BhFhbA Met137, which sits in the hydrophobic cavity, is The signal sequence is marked in magenta and secondary structure elements derived from the crystal structure of BhFhbA:FH19-20 are shown in blue above the sequence. The conserved hinge region is marked with a red rectangle, and an arrowhead points at the key Phe residue important for tight binding to FH20. The sequence alignment consists of protein sequences, predicted protein sequences and translated genomic regions that match the BhFhbA used as the search sequence (see Materials and Methods for details). Asterisk ' � ' represents a stop codon and question mark '?' stands for an incomplete codon, where a frameshift appears to have occurred. In generating the protein alignment, the frameshift was ignored, and the translation frame was preserved to allow further protein alignment after that problematic codon. Sequence parts with grey background are translated genomic sequences, which most probably are not present in derived proteins due to stop codons or frameshifts.

PLOS PATHOGENS
Borrelia immune evasion replaced by Thr in BdFhbA. We also mutated the key phenylalanine (corresponding to Phe154 in BhFhbA) in BdFhbA to alanine (Phe130Ala). As expected, both the gel filtration (Figs 5C and S9) and MST assays (Fig 5D) showed a drastic decrease in the binding of FH19-20 to BdFhbA/Phe130Ala. We also showed that full length FH binds wild type BdFhbA but not the BdFhbA/Phe130Ala mutant (S10 Fig) and retains its cofactor activity in FI-mediated cleavage of C3b (Fig 5E). Finally, BdFhbA binds FH predominantly via domain 20 and possesses only weak affinity towards FH5-7 (Fig 5F), like BhFhbA. CD spectrometry showed that both wildtype BdFhbA and BdFhbA/Phe130Ala were correctly folded (S7 Fig).
Together, these results clearly show that two homologous proteins, BhFhbA and BdFhbA, bind FH19-20 via a novel, conserved mechanism. Based on structure-assisted multiple-

PLOS PATHOGENS
Borrelia immune evasion sequence alignments, we predict that other members of the family utilize the same mechanism.
We delineated the full spectrum of potential FH-binding proteins in LD and RF clades using the same set of genomes as for the FhbA searches to identify homologues for FH-binding proteins from Bo. burgdorferi. We used sequences from the known structures of CspA/ CRASP-1 [49], CspZ/CRASP-2 [50] and OspE [22] in the search. Ten of the eleven genomes analyzed in the LD clade have proteins homologues to CspA/CRASP-1, six of eleven to CspZ/ CRASP-2 and ten of eleven to OspE/CRASP-5 (in Bo. bavariensis the protein has an N-terminal region displaying 30% identity to OspE (Tables 3 and S3). In contrast, the members of the RF clade have neither CspA/CRASP-1 nor CspZ/CRASP-2 homologues.
There were two species that lacked all four classes of FH-binding proteins ( Table 3). The first is Bo. chilensis, which was originally isolated from ticks [47], and currently there is no data about its vertebrate host. The second is B. anserina, a bird isolate. Although another bird infecting species, Bo. garinii, was shown to bind avian FH [51], it is not known whether Bo. anserina binds FH at all.
Taken together, FH binding proteins clearly fall into four groups, which also coincide with the phylogenetic classification of the borreliae species. All LD clade species with FH-binding proteins have CspA/CRASP-1 proteins, and the majority also possess CspZ/CRASP-2 and OspE/CRASP-5 -proteins. RF clade species exclusively have FhbA proteins and lack the other three, suggesting a significant role for FhbA proteins in complement evasion of RF borreliae.

Discussion
By determining a 2.2 Å resolution crystal structure of B. hermsii surface protein BhFhbA in complex with FH19-20, we revealed the molecular mechanism by these two proteins interact with each other. The structure, combined with mutagenesis and binding studies, led to the identification of a conserved aromatic residue, Phe154, that is central in binding FH. Using structure-guided sequence analysis with the structure of BhFhbA as a search model, we identified several putative homologous proteins in relapsing fever and Lyme disease borreliae. To confirm the common binding mechanism between the FhbA-related proteins and FH, we also expressed a novel FH19-20 binding protein (BdFhbA) from B. duttonii, which causes relapsing fever, and mutated the key residue, Phe130. MST and gel filtration demonstrated that these two proteins share the same FH-binding mechanism. We also delineated four different FHbinding proteins families of Borrelia/Borreliella and show that the LD clade of borreliae has four different groups of FH-binding proteins, but the RF borreliae clade appears to possess only one.
BhFhbA was originally identified as an FH binding protein from B. hermsii, which causes relapsing fever [29]. The protein was predicted to be composed of four α-helices flanked by three loops. When compared to our crystal structure, the locations of predicted coiled coils and loop regions match poorly. Thus, earlier random [31] and site-directed mutagenesis [52] studies aimed at the predicted loop regions of the protein also targeted α-helical and core

PLOS PATHOGENS
Borrelia immune evasion regions of the protein, affecting secondary structure elements and protein folding. We mapped six previously published mutants with reduced or no FH-binding activity [31] to our structure (S11 Fig). Detailed inspection of the environment of each mutated position explains decreased binding. For example, the Asn172Thr mutation disrupts two hydrogen bonds that keep helices 2 and 3 together. Though Asn172 is located far from the active hinge region, such a mutation is likely to affect overall folding or stability of the protein. Nevertheless, these earlier studies support the importance of the hinge region in FH binding.
BhFhbA binds to domain 20 of FH (Figs 1A and S1). Structural analysis of FH19-20 complexes with two microbial proteins, FhbA and OspE from Bo. Burgdorferi [22] suggests that the general microbial binding site on FH20 mediates interaction (S1 Table). However, the interactions are different: a hydrophobic binding pocket is formed between FH20 and BhFhbA, whereas the FH20:OspE interaction is mainly electrostatic in nature and mediated by hydrogen bonds. Binding of FH to sialic acid on erythrocytes, endothelial cells and platelets has been shown to protect host cells from complement [53]. Interestingly, the structure of the FH19-20: sialic acid:C3d [40] complex revealed that the same general microbial binding site in FH20 is involved in binding to sialic acid. This is a rare and interesting example of convergent evolution of a binding site utilization; the binding ligands and mechanisms are different, but the binding patch site on FH largely overlaps.
The other important interaction site in FH is in domains 6-7, part of which is present in full-length FH as well as in FHL-1. Typically, microbes bind FH via domain 20, or FH and FHL-1 via domain 7. BhFhbA is a rare example, as it was first reported to bind FHL-1 and FH fragment 1-7 [54] and later FH via domain 20 [30]. Here, we examined if BhFhbA has two binding sites on FH by comparing the interactions of FH19-20 and FH5-7 to FhbA (Figs 3A-3D and S6). Our results suggest that FhbA has only weak affinity for FH5-7 and it cannot compete with FH19-20 in binding to BhFhbA (Figs 3B and S6). When all FhbA-related proteins analyzed so far are considered, this result is not surprising. No binding of FH1-7 to FhbArelated protein HcpA from B. recurrentis [33], to BpcA of B. parkeri [32] or to CbiA of B. miyamotoi [40] was observed. It cannot, however, be excluded that under certain circumstances, e.g., in specific tissue locations, RF borrelia acquire FHL-1 as well, but the most important binding site for this protein is on FH20.
Microbes bind domains other than N-terminal domains FH1-4 so that this region can bind to and downregulate C3b. It has been previously shown that FH bound to BhFhbA retains its cofactor-activity in cleaving C3b both using purified proteins [30] and when BhFhbA is expressed on the bacterial cell surface [55]. We confirmed that FH bound to BdFhbA retains its cofactor-activity ( Fig 5E). Furthermore, we previously showed that FH20 bound to BhFhbA or other microbial proteins enhances the cofactor-activity of FH in cleaving C3b [30]. The mechanism of this enhanced regulatory function is not clear, but we speculate that simultaneous binding to target via domain 20 and to C3b via domain 19 facilitates enhanced activity. Indeed, we previously solved the structure of a tripartite complex between microbial protein OspE, C3d and FH19-20 [56]. Our model of FhbA on the surface of Borrelia is based on the hypothesis that BhFhbA acts similarly to OspE and binds simultaneously to FH20 and C3d (Fig 6).
Using BhFhbA as a search model in structure-guided sequence database analysis enabled us to identify 10 homologous DNA loci from the RF and three from the LD clades (Fig 3). All ten identified proteins in the RF borreliae group are very similar within the FH20 binding region, as both the key phenylalanine residue and the surrounding hinge region are highly conserved. Moreover, for BhFhbA and BdFhbA, mutations of the key phenylalanine dramatically affected binding to both FH19-20 and full-length FH (Figs 2, 4, 5, S4, S6 and S10). The only exception is B. turicatae, which has an asparagine instead of aspartic acid in the +2 position after Phe154. Consistent with our predicted binding mechanism, B. turicatae BtcA is the only FhbA-related protein of RF borreliaea that does not bind FH [32] (Table 4). Interestingly, there are three FhbA-related sequences in the borrelia from the LD clade, which cluster separately from the RF group (Fig 4B). Sequence data show that there are many deleterious alternations at the DNA level. It is not yet known, if these FhbA-related sequences are functional on a protein level.
Analysis of all available sequence data from borreliae demonstrates that the LD clade has evolved to have three to four different classes of FH-binding proteins, whereas the RF clade has just one, which we name the FhbA-related protein family (Table 4). FhbA-related proteins may be able to inhibit complement more efficiently than the other FH-binding proteins, thus compensating for the lack of other FH-binding proteins. The interactions of FhbA-related proteins with other complement proteins (Table 4) might also affect overall regulation of complement. It cannot be excluded that other, yet unknown, FH-binding proteins exist in relapsing fever spirochetes, or that some other complement evasion mechanisms, like binding of C4BP [57] or C1-inhibitor [58] provide enhanced protection.

Borrelia immune evasion
Five FH-binding FhbA-related proteins have been shown to mediate serum resistance. B. hermsii strain YOR, which expresses BhFhbA, is more resistant to serum and causes more persistent infections in mice when compared with strain REN, which naturally lacks FhbA [52]. Expression of three FhbA-related proteins (BhCRASP-1 [59], HcpA [33], BpcA [32]) in the serum-sensitive strain Bo. burgdorferi B313 led to increased serum survival of the mutant strain. In addition, B. miyamotoi CbiA established serum resistance when expressed in serum sensitive Bo. garinii strain G1 [34]. Surprisingly, an fhbA knockout strain created from B. hermsii strain YOR retained resistance to complement in vitro and in mice, even though the strain did not express FhbA nor bound FH [55].
To analyze the effect of BhFhbA on serum mediated killing, we expressed BhFhbA and BhFhbA mutant Phe154Ala on the surface of a serum-sensitive laboratory strain of E. coli. In that environment wild-type BhFhbA, but not the mutant, protected bacteria from complement killing (Fig 3E). The binding mechanism suggested by our structure thus appears to be important also in a more physiological context. Similar results were obtained from the assay, where we incubated live B. hermsii borrelia in whole blood and measured complement activation ( Fig 3E). Wild-type BhFhbA competed with B. hermsii for FH whereas the Phe154Ala mutant did not.
Our results thus demonstrate that BhFhbA is also functional on the surface of E. coli and can provide protection from complement in a natural environment. Previous results from the fhbA knockout strain suggest that other yet unidentified mechanism(s) to prevent formation of membrane-attack complexes may exist in B. hermsii strain YOR. This is not, perhaps, surprising because pathogens typically have several mechanisms that act alone or in tandem to help the bacteria evade innate immunity.
In summary, we present here a high-resolution structure of BhFhbA, an outer-surface complement evasion mediating protein from Borrelia hermsii, in complex with FH19-20. We found a dozen highly homologous proteins from Lyme disease and relapsing fever spirochetes, thus identifying a new family of immune evasion proteins, which we name the FhbA-related protein family. We propose that FhbA-related proteins are important complement evasion molecules in RF borreliae, and thus represent important targets to develop tools to prevent infections caused by borreliae.

Ethics statement
Blood samples were drawn from healthy human volunteers by trained professionals after donor review of information fact sheet and written and signed consent as approved by the Ethical Committee (decision HUS/135/2020) of Hospital district of Helsinki and Uusimaa.
Bacteria and sera. Borrelia hermsii strain HS1 was a kind gift from prof. Bergström, University of Umeå, Sweden. Bacteria were cultured in BSK-H media (Sigma-Aldrich, Darmstadt, Germany) at +33˚C in 5% CO 2 and 100% humidity and number was determined by calculation under dark-field microscopy using 40x magnification. Prior to usage bacteria were pelleted (8,000 g 15 min at RT) and washed 3 times with PBS (phosphate-buffered saline; 120 mM NaCl, 30 mM phosphate, pH 7.4). Blood was drawn into hirudin (Roche Diagnostics, Mannheim, Germany) tubes from healthy human volunteers after informed written and signed consent (Ethical Committee decision HUS/135/2020, Hospital district of Helsinki and Uusimaa). The plasma was isolated by centrifugation. To obtain serum the blood was drawn into serum tubes, blood was allowed to clot, after which it was centrifugated and serum collected. Serum was kept at -80˚C and heat-inactivated (+56˚C, 30 min) to remove complement activity.
Plasmid construction and mutagenesis. Wild type BhFhbA without its leader peptide concentration was 8 mg/ml of BhFhbA and the crystallization trials used sitting drop vapour diffusion in 200 nl drops (100 nl of protein complex solution and 100 nl of well solution). Plate-like crystals first appeared in the Helsinki Random Screen 1 (HR1) as well as in the Helsinki Complex screen. Hit conditions were manually optimised by preparing hanging drops with 2 μl of protein and 2μl reservoir. Harvestable 3D crystals appeared within two weeks at 20˚C from the following conditions: 0.1 M MES pH 6.7, 0.2 M Ammonium Sulfate, 20% (v/v) PEG 4000. The crystals were picked, cryoprotected with 20% (v/v) ethylene glycol in the mother liquid, and flash frozen in liquid nitrogen for storage and transportation to the synchrotrons. Diffraction data were collected at European Synchrotron Radiation Facility (ESRF, Grenoble, France) on beamline ID23-2, at 100 K on a Pilatus3 detector (Dectris). A full 360d ataset (3600 images) was collected at an oscillation angle of 0.1˚, transmission energy of 18.3%, with 0.2 s exposure time per frame. Data were merged and scaled using X-ray Detector Software (XDS) and autoPROC. Molecular replacement was done using Phaser from the Phenix package with the published structure of FH19-20 (PDB 4J38) as a search model. An initial structure of BhFhbA was built manually into clear electron density. Several cycles of manual building using Coot [36] and refinement with BUSTER [37] resulted in a final structure with R work /R free = 0.19/0.24. We used ACHESYM software for reindexing and redefining the origin. Coordinates and structure factors were deposited to the PDB with accession code 6ZH1.
Binding assays. Binding affinities between FH19-20 or FH (from Complement Technologies, US) and wild type and mutants of FhbA proteins was determined using Microscale Thermophoresis (MST) with a Monolith NT.115 instrument (Nanotemper Technologies, Germany). FH19-20 and FH were labelled with RED-tris-NTA dye (Nanotemper Technologies, US) in PBS according to the manufacturer's instructions. 10 μl of 300 nM labelled protein was mixed with 10 μl of ligand in PBS/0.025% Tween-20, the mixture loaded into standard treated capillaries (Nanotemper Technologies, US) and thermophoresis was measured at 22C for 22-30 s with 20% LED power and 20%/60% infrared laser power. Three independent measurements were made, and results were analysed using the MO. Affinity Analysis software version 2.1 (Nanotemper Technologies, US). For gel filtration 100 μl of proteins (20 nmole) were eluted individually and in combination with FH19-20 on a Superdex 200 Increase 10/300 GL column attached to an Ä KTA (GE-healthcare) with PBS buffer at +4˚C. 1 ml fractions from each run were collected and subjected to SDS-PAGE analyses on TGX gradient (4-20%) precast mini-gels (Biorad, CA, US), which were fixed, stained with QC Colloidal Coomassie Stain (Biorad), and proteins were visualized using Image Lab (Biorad). For high salt experiments, PBS buffer supplemented with 500 mM NaCl was used.
Western blotting. 500 ng of purified BhFhbA or mutant proteins were subjected to nonreducing SDS-PAGE and transferred onto nitrocellulose membranes. Nonspecific binding was blocked with 3% fat-free milk in PBS for 1 hour at 22˚C. The membranes were incubated with purified FH19-20 (10 μg/ml in 3% fat-free milk in PBS) for 12 hours at +4˚C. After three washes (PBS, 0.05%Tween) membranes were incubated with goat polyclonal anti-FH antibody (Quidel, CA, US) at a 1:2000 dilution in 3% fat-free milk in PBS for 3 hours at RT. After three washes with PBS/0.05% Tween a HRP-conjugated rabbit donkey anti-goat antibody (Merck, NJ, US) was added at a dilution of 1:2000, and the membranes were incubated at 22˚C for 1 hour. After three washes the bound antibodies were detected by enhanced chemiluminescence. Mutant proteins were also detected using mouse anti-His-antibody (Merck, NJ, US) at a 1:2000 dilution and HRP-conjugated secondary rabbit-anti-mouse antibody (1:5000 dilution) to detect his-tagged mutant proteins.
Cofactor-activity assay. 1 μg/ml of BdFhbA was coated on 96-well ELISA-plates (Thermo Scientific, MA, US) in 0.5 M NaHCO₃ pH 9.6 for 12 hours +4˚C. Wells were washed 3 times with 300 μl PBS, after which heat-inactivated NHS (10% in PBS) or 16 nM of FH was added into the wells, which were then incubated at +37˚C for 60 min and washed 3 times with PBS. 50 nM of C3b and 40 nM of Factor I were added, after which wells were incubated for 60 min at +37˚C. As a control, C3b was incubated with Factor I or with factor I and Factor H (16 nM). Samples from wells were collected and run on SDS-PAGE under reducing conditions. and transferred onto nitrocellulose membranes. Nonspecific binding was blocked with 3% fatfree milk in PBS for 1 hour at 22˚C, after which monoclonal anti-C3c antibody (Quidel, San Diego, CA, USA) (1:2000 in 3% fat-free milk in PBS) was added and membrane was incubated for 12 hours at +4˚C. After three washes with PBS/Tween, 0.5%, bound primary antibody was detected with HRP-conjugated rabbit anti-mouse antibody (Jackson ImmunoResearch, Cambridgeshire, UK) (1:500 dilution in 3% fat-free milk in PBS) for 60 min. After three washes, the bound antibodies were detected by enhanced chemiluminescence.
Complement activation assays. 80 μl blood with 5 x 10 5 B. hermsii bacteria were incubated at 37˚C 5% CO 2 atmosphere for 30 min in the presence of BhFhbA wild type or BhFhbA Phe154Ala mutant (9 μM concentration). Complement activity was stopped by adding 50 mM EDTA (ethylenediaminetetraacetic acid). Plasma was separated from the blood cells by centrifugation at 600 × g for 10 minutes. diluted 1:30 and analyzed by SC5b 9 Enzyme Immunoassay according to the manufacturer's instructions (MicroVue SC5b 9 Plus Enzyme Immunoassay, Quidel, CA, US).
Serum sensitivity assay of E. coli. Transformed E. coli strains were grown o/n at +37˚C on a shaker at 200 rpm in LB media containing 5 μg/ml chloramphenicol, diluted to 0.1 OD600, and expression of proteins was induced with 1mM IPTG. For western blot analysis of the outer membrane localization of the His-tagged proteins cells were collected and the fractionations were done as published previously [45]. After 3 hours growth, bacteria were pelleted (3,000 g 10 min at RT) and diluted into veronal buffered saline (VBS) so that each reaction contained 10 5 bacteria with 10% NHS with 5 mM MgCl 2 and 10 mM EGTA. In the control, serum was inactivated with 10 mM EDTA in VBS. The samples were diluted 1:1 into ice cold PBS at time point 0 and after 15 min incubation at +37˚C. To obtain suitable amounts of bacteria, 50 μl from the assays and serial dilutions (1:10, 1:100) were plated onto Luria-Bertani plates containing 5 μg/ml chloramphenicol. After growth at +37˚C, colonies were counted and the percentage survival (number of bacteria after exposure to serum divided by bacteria at time point zero x 100).
ELISA assays. 1 μg/ml of BhFhbA or BdFhbA were coated onto 96-well ELISA-plates (Thermo Scientific, MA, US) in 0.5 M NaHCO₃ pH 9.6 for 12 hours +4˚C. Wells were washed 5 times with 300 μl PBS and blocked (60 min RT). Buffers used in all dilutions and washes were 0.5% BSA in PBS for monoclonal antibodies and 0.5% Tween in PBS for polyclonal antibodies. Serial dilutions of FH19-20 or FH5-7 were prepared on non-adherent plastic plates. 100 μl pf primary monoclonal antibody VIG8 [60] against domain 20 of FH (1 μg/ml) or polyclonal goat-anti FH at a 1:2000 dilution was added and plates were incubated 60 min at +37C . After 5 washes, secondary HRP-conjugated anti-mouse or anti-goat antibody (both from Jackson ImmunoResearch, Cambridgeshire, UK) were added at a 1:5000 dilution and incubated for 60 min at +37˚C. After five washes the substrate, o-phenyl-diamine diluted in H 2 O and supplemented with 0.04% H 2 O 2 , was added. After a 15 min incubation at +22˚C the reaction was stopped by adding 50 μl of 2M H 2 SO 4 per well. The absorbances were read with an ELISA-reader using a 492 nm filter.
Sequence analysis and protein family identification. The full sequence of the FhbA protein from B. hermsii YOR (UniprotID: W5SB08) was submitted to Position-Specific Iterated (PSI)-BLASTp search using the NCBI portal against the non-redundant protein sequences database. After each iteration, the sequences for the next round were manually selected in order to have a representative group of unique sequences from different borreliae families. It took 5-6 rounds to obtain a comprehensive list of family representatives. Increasing repetitions provided no new members. All available borreliae genomes (in total 158 genomes) were downloaded from NCBI and searched for FhbA-related proteins and other FH-binding proteins starting from the following proteins: B. hermsii FhbA (UniprotID: W5SB08), Bo. burgdorferi CcpA (CRASP-1) (UniprotID:Q66ZA0,), Bo. burgdorferi CspZ (CRASP-2) (UniprotID: O5066,), and Bo. burgdorferi OspE (UniproID: Q45001) with exonerate program [61], using parameters to match the translated genomic DNA sequence to that of the bait protein sequence. This allowed us to identify sequences at the DNA level even when the protein prediction failed, or when protein annotation was incorrect or missing in the databases. For CspZ (CRASP-2) protein, the reported genomes were missing the full-length sequence, but Rogers et al. reported [46] two isolated sequences, which were not deposited in any well-known database, but are nevertheless included in Table 3. Four genomes (B. chilensis, B. turicatae, B. anserina, and B. recurrentis) did not have the plasmids containing the fhbA gene, but the sequences had been isolated by different research groups and annotated accordingly. We combined the sequences from both searches and selected the unique representatives from each of the species for the multiple sequence alignment. In case of discrepancy or multiple matches, we selected the closest homologue to the target protein used in the search.
In this study we refer to FhbA-related sequencies from different species as homologs [62]. The accession codes for sequences used in the alignment are in the legend to Fig 4A. 3D protein homology model was obtained for each representative member using Phyre2 [63] and SWISS model [64] servers freely available on-line, with our X-ray structure of BhFhbA as a template. Secondary structure elements were independently predicted using Jperd4 (PMID: 25883141) software and compared to those obtained from Phyre2 and SWISS model. Homology models were used to correct the sequence alignment by introducing gaps at relevant places, to obtain a structure-based or structure-assisted sequence alignment. This step was crucial to preserve the correct alignment of secondary-structure elements (α-helices).
Circular dichroism (CD) spectrometry of purified proteins. The CD spectra for the wild type and mutant FhbA-proteins were collected in 20% PBS buffer diluted with water, at 20˚C, on a J-720 spectropolarimeter (Jasco MA, US), in 300 μl quartz cuvette of 0.1 cm light path length with the following parameters: continuous scanning mode with scanning speed of 20 nm/min, band width 0.5 nm, wave range 190-260 nm, data pitch 0.5 nm. All the proteins were thawed and diluted to 12.5 or 15 μM. The accumulation of 5 scans for each protein has been plotted as a single curve.
Statistical analysis and data fitting. K d -values were calculated from each individual experiment separately (and shown as an average in In the serum survival test statistics was calculated using SPSS software, IBM statistics from at least 4 replicates using one-way Anova supplemented with Dunnet's post-hoc test for unequal variances.  Table. Hydrogen bonds between FH20 and OspE and FH20 and BhFhbA. Distances (Å) are from the PISA-server (Krissinel and Henrick, 2007 [39]) and data from binding inhibition assays from (Meri et al. 2013 [30]). In biochemical assays effect of mutant protein to binding of radiolabeled FH19-20 to BhFhbA/OspE was measured, thus in regard to those results the table has no distances. For example (italics), mutation of Trp 1183 to alanine decreased binding of FH19-20 to FhbA and OspE (grey marking) and the PISA server found a hydrogen bond between OspE Asn 77 (Nδ2) and FH Trp 1183 (O) with a length of 3.08 Å.