Crystal Structure of Neurotropism-Associated Variable Surface Protein 1 (Vsp1) of Borrelia turicatae

Abstract

Vsp surface lipoproteins are serotype-defining antigens of relapsing fever spirochetes that undergo multiphasic antigenic variation to allow bacterial persistence in spite of an immune response. Two isogenic serotypes of Borrelia turicatae strain Oz1 differ in their Vsp sequences and in disease manifestations in infected mice: Vsp1 is associated with the selection of a neurological niche, while Vsp2 is associated with blood and skin infection. We report here crystal structures of the Vsp1 dimer at 2.7 and 2.2 Å. The structures confirm that relapsing fever Vsp proteins share a common helical fold with OspCs of Lyme disease-causing Borrelia. The fold features an inner stem formed by highly conserved N and C termini and an outer “dome” formed by the variable central residues. Both Vsp1 and OspC structures possess small water-filled cavities, or pockets, that are lined largely by variable residues and are thus highly variable in shape. These features appear to signify tolerance of the Vsp-OspC fold for imperfect packing of residues at its antigenic surface. Structural comparison of Vsp1 with a homology model for Vsp2 suggests that observed differences in disease manifestation may arise in part from distinct differences in electrostatic surface properties; additional predicted positively charged surface patches on Vsp2 compared to Vsp1 may be sufficient to explain the relative propensity of Vsp2 to bind to acidic glycosaminoglycans.

Relapsing fever (RF) and Lyme disease (LD) are infectious bacterial diseases caused by spirochetes of the genus Borrelia. In contrast to other spirochetes that affect humans, such as Treponema pallidum or Leptospira interrogans, the transmission of borreliae to vertebrate hosts depends on arthropod vectors (6). LD is the most common vector-borne disease in North America (4), while RF is endemic in palearctic, afrotropical, neotropical, and nearctic regions, with sporadic cases occurring in the Southwestern to Northwestern parts of the United States (8). Throughout their vector-host life cycle, Borrelia cells display surface lipoproteins that are anchored to the bacterial membrane lipid bilayer via an N-terminal triacyl-modified cysteine (10, 13). Because of their abundance at the host-pathogen interface, these surface lipoproteins have received considerable attention as potential virulence factors and vaccine targets. Indeed, development of Borrelia burgdorferi outer surface protein A (OspA) (37) as a first-generation LD vaccine for humans (66) was initiated before the determination of its structure (25, 40, 41) and before the emergence of clues to its potential biological function within the tick vector (50, 51).

A well-established mouse model of infection has revealed the overall functions of the major outer membrane lipoproteins of RF borreliae. Collectively called variable membrane proteins (VMPs), they are immunodominant and determine serotype (3). The VMPs fall into two phylogenetically distinct families: the variable large proteins (Vlps) of 36 to 40 kDa and the variable small proteins (Vsps) of 20 to 23 kDa (17). Vlp and Vsp proteins appear to be unique to the genus Borrelia. They have been described in the RF species Borrelia turicatae (17, 19, 53, 59), Borrelia hermsii (7), Borrelia recurrentis (68), and Borrelia crocidurae (65). B. burgdorferi OspC, a major outer surface protein involved in establishing mammalian LD infection (32, 57), is phylogenetically related to the Vsps (17), which has led to the term Vsp-OspC family. By analogy, B. burgdorferi VlsE (for VMP-like sequence, expressed) was named for its similarity to the Vlps (72).

While signal peptides necessary for proper translocation and processing are conserved, the mature Vlp-VlsE and Vsp-OspC lipoproteins are highly polymorphic. OspC amino acid sequences from LD borreliae can vary in identity by as much as 25% (38, 42). There is evidence that this variation is maintained by frequency-dependent balancing selection and that OspC sequence diversity is a consequence of the multiplicity of vertebrate species that serve to maintain a local population (14, 69). Vlp and Vsp proteins of RF borreliae diverge even more than OspC proteins of LD borreliae, with up to 60% variation in amino acid identity (17, 35). In contrast to the single plasmid-encoded ospC gene of B. burgdorferi (62), about 60 archival copies of B. hermsii vsp and vlp genes are maintained on linear plasmids and sequentially expressed from a promoter site after gene conversions or DNA rearrangements (7, 23, 53, 60). Reminiscent of similar pathogenic strategies used by African trypansomes, malarial parasites, or Neisseria species, the resulting multiphasic antigenic variation of Vsps and Vlps allows the spirochete to repeatedly evade the host's immune response. This leads to recurrent spirochetemia and characteristic febrile episodes (3).

Experimental RF in mice has demonstrated a role for Vsp proteins in differential tissue localization, as well as for avoidance of the immune response. B. turicatae serotypes expressing Vsp1 (previously called VspA) and Vsp2 (VspB; 60% amino-acid sequence identity to Vsp1) were derived from a single infected mouse (19). Both serotypes carry silent copies of the vsp1 and vsp2 alleles and were found to differ genetically only around the expression locus (54). Expression of Vsp2 is associated with high densities of spirochetes in the blood, while expression of Vsp1 leads to early invasion and persistent infection of the central nervous system (16-19, 43, 52, 53). Vsp1 may help to target the bacterium to an immunoprivileged niche, while Vsp2 may facilitate efficient transmission to the next feeding tick. Tissue culture assays have yielded clues about pathogenesis mechanisms. B. turicatae spirochetes expressing Vsp1 penetrate human umbilical vein epithelial cell monolayers more readily than those expressing Vsp2 (19). On the other hand, Vsp2 increases binding of the spirochete to mammalian endothelial and glial cells (67), predominantly by the direct interaction of Vsp2 with host cell surface glycosaminoglycans (GAGs) (44, 74).

The clear correlation between Vsp amino acid sequence and phenotype for B. turicatae serotypes 1 and 2 presents a unique opportunity to probe the molecular basis of niche selection in this experimental system. As a starting point, we report here crystal structures of two different core fragments of Vsp1, Vsp1^45-201 and Vsp1^34-214, against X-ray diffraction data to resolutions of 2.2 Å and 2.7 Å, respectively. Comparison of the Vsp1 structure to a homology model for Vsp2, as well as to previously determined crystal structures of OspCs from LD borreliae, has revealed distinctive physicochemical features that could be responsible for tissue tropism and disease manifestation. These data will serve as a keystone for further structure-function studies of antigenically variant surface lipoproteins of RF Borrelia spirochetes.

MATERIALS AND METHODS

Gene constructs and expression.

Vsp1^34-214 expression in Escherichia coli BL21(DE3) cells (Invitrogen) harboring the expression plasmid pET29b::S34vsp1 was reported previously (73). A selenomethionine derivative (SeMet[44,95]-Vsp1^34-214) was prepared by introducing two Ile-to-Met mutations into pET29b::S34vsp1 at Vsp1 amino acid positions 44 and 95 by sequence overlap extension PCR (36). To introduce the I44M mutation, two primary DNA fragments were amplified using primer pairs S34vsp1/2-fwd and vsp1I44M-rev or vsp1I44M-fwd and S34vsp1-rev, respectively (Table 1) (73). The two primary amplicons were then used as templates and spliced in a secondary PCR with the flanking S34vsp1/2-fwd and S34vsp1-rev primers. The resulting vsp1I44M DNA fragment was digested with NdeI and BamHI and ligated into pET29b, resulting in pET29b::S34vsp1I44M. To introduce the I95M mutation, we used the primers vsp1I95M-fwd and vsp1I95M-rev with the above flanking primers on a pET29b::S34vsp1I44M template. The resulting pET29b::S34vsp1I44M/I95M expression plasmid was used to transform E. coli BL21(DE3) cells. Growth in l-selenomethionine-containing minimal medium followed the procedure of Doublié (27).

TABLE 1.

Primers

Namea	Target	Sequence (5′ to 3′)b
vsp1I44M-fwd	Introduces I44-to-M mutation	CTAGCTACAATGACTAAAAACATT
vsp1I44M-rev	Introduces I44-to M-mutation; complement to vsp1I44M-fwd	AATGTTTTTAGTCATTGTAGCTAG
vsp1I95M-fwd	Introduces I95-to-M mutation	GCAAAATTAATGTCAGGAGCATAT
vsp1I95M-rev	Introduces I95-to-M mutation; complement to vsp1I95M-fwd	ATATGCTCCTGACATTAATTTTGC
T45vsp1-fwd	5′ end of truncated vsp1 with NdeI site	GGAATTCCATATGACTAAAAACATTACC
S202vsp1-rev	3′ end of truncated vsp1 with BamHI site	CGGGATCCTTATGAAAGCGCATTTATTG

^aPrimers S34vsp1/2-fwd and S34vsp1-rev are defined in reference (73), where they are referred to as S34vspAB-fwd and S34vspA-rev, respectively.

^bUnderlined nucleotides are introduced restriction sites. Boldface nucleotides are introduced Ile-to-Met point mutations, start codons, or stop codons.

To prepare Vsp1^45-202, a DNA fragment starting with the Thr-45 codon and ending with the Ser-202 codon was amplified by PCR with the oligonucleotide primer pair T45vsp1-fwd and S202vsp1-rev (Table 1) using pET29b::S34vsp1 (73) as the template. The resulting amplicon was digested with NdeI and BamHI and ligated into pET29b, yielding pET29b::T45vsp1S202. An overnight culture of E. coli BL21(DE3) cells transformed with pET29b::T45vsp1S202 was transferred at 1:50 dilution to two 1-liter volumes of LB broth containing 50 μg/ml kanamycin. After growth for 3 h at 37°C, expression was induced by addition of isopropyl-β-d-thiogalactopyranoside to a final concentration of 1 mM. The cells were incubated at 37°C for 4 additional hours and then harvested and stored for several days at −80°C.

Purification.

Vsp1^34-214 and SeMet[44,95]-Vsp1^34-214 were purified by ion-exchange and gel filtration chromatography as described by Zückert et al. (73). Purification of Vsp1^45-202 was performed as follows, with all steps at 4°C. Cells were resuspended in 50 ml of 20 mM Tris-HCl and 200 mM NaCl, pH 8.0, and lysed by ultrasonication. The soluble fraction was obtained by centrifugation at 4,000 × g for 30 min. Finely ground ammonium sulfate powder was added to the supernatant to 50% saturation, and the precipitate was removed by centrifugation at 9,000 × g for 30 min. Finely ground ammonium sulfate powder was then added to the resulting supernatant to 75% saturation, and the precipitate, which contained Vsp1^45-202, was collected by centrifugation at 16,000 × g for 30 min. The precipitate was resuspended in 20 ml of 20 mM Tris-HCl, pH 8.0, and dialyzed overnight against two changes of the same buffer. The resulting material was passed through a HiLoad 26/60 Superdex 75 prep grade column (Amersham). Peak fractions corresponding to Vsp1^45-202 were pooled and loaded onto a HiPrep 16/10 Q-Sepharose column (Amersham). The protein was eluted using a linear gradient of 0 to 200 M NaCl in 20 mM Tris-HCl, pH 8.0. Peak fractions corresponding to Vsp1^45-202 were pooled and concentrated to 53.38 mg/ml using Centricon YM-10 centrifugal filter units (Millipore) and stored at −20°C in 1-ml aliquots. The final yield was approximately 100 mg.

Crystallization.

Vsp1^34-214 and SeMet[44,95]-Vsp1^34-214 were crystallized against 28% (wt/vol) PEG 4000, 64 mM Tris-HCl, pH 8.5, 18% (vol/vol) glycerol, and 100 to 160 mM NiCl₂, as reported previously (73). Crystallization conditions for Vsp1^45-202 were identified using Crystal Screen (Hampton Research) and optimized via grid screening of pH, precipitant concentration, temperature, and additives (Additive Screen; Hampton Research). Bar-shaped crystals were produced by hanging-drop vapor diffusion at 4°C against a reservoir composed of 0.1 M sodium acetate, pH 4.8, 25% (wt/vol) PEG 4000, 0.1 M LiCl, and 0.1 mM taurine.

Data collection.

All X-ray diffraction data were collected on charge-coupled device detectors at the Protein Crystallography Research Resource of the National Synchrotron Light Source at 100 K. Native diffraction data for the monoclinic Vsp1^34-214 crystal to 2.7 Å were reported previously (73). Diffraction data were measured to 3.0 Å for a triclinic SeMet-Vsp1^34-214 crystal at a single wavelength corresponding to the selenium K edge (λ = 0.9795 Å). Native diffraction data were measured to 2.2 Å for Vsp1^45-202 after “crystal annealing” (34) as follows. A single crystal was soaked for 2 minutes at 25°C in a cryoprotectant solution composed of reservoir plus 20% (vol/vol) ethylene glycol and then flash-cooled to 100 K. Extensive streaking was observed in initial diffraction images. The soaking/flash-cooling procedure was repeated twice, each time yielding remarkable improvement in the diffraction pattern as assessed by reduction in streaking of individual reflections. All data reduction was performed using the HKL package (49). Summary statistics for each data set are given in Table 2.

TABLE 2.

X-ray diffraction and model statistics

Parameter	Value
	SeMet-Vsp134-214	Vsp145-202	Vsp134-214
Crystal
Space group	P1	C2	C2
Unit cell lengths (Å)	a = 34.6, b = 88.2, c = 120.2	a = 175.8, b = 34.2, c = 99.4	a = 241.3, b = 69.1, c = 87.6
Unit cell angles (degrees)	α = 74.1, β = 85.2, γ = 84.0	α = γ = 90.0, β = 112.7	α = γ = 90.0, β = 104.8
Asymmetric unit content	4 dimers	2 dimers	4 dimers
Solvent fraction	0.492	0.433	0.489
Diffraction statisticsa
Beamline	NSLS X12C	NSLS X25	NSLS X12C
Wavelength (Å)	0.9795	1.0	1.0
Resolution range (Å)	50.0-2.9 (3.0-2.9)	50.0-2.2 (2.3-2.2)	50.0-2.7 (2.8-2.7)
No. of merged reflections	25,727	26,408	38,032
Completeness	0.847 (0.808)	0.957 (0.845)	0.989 (0.920)
Average redundancy	3.89	2.93	3.42
Rmergeb	0.134 (0.567)	0.050 (0.166)	0.042 (0.147)
Model statisticsc
PDB identifier		1YJG	2GA0
Resolution range (Å)		50.0-2.2 (2.28-2.20)	50.0-2.7 (2.78-2.70)
R		0.206 (0.247)	0.216 (0.286)
Rfree		0.265 (0.289)	0.272 (0.353)
Rfree test size (%)		5.0	5.0
No. of protein atoms		4,508	9,456
No. of solvent atoms		216	135
RMS deviations
Bond lengths (Å)		0.009	0.015
Bond angles (degrees)		1.258	1.544
Ramachandran statistics (% residues)
Most-favored regions		92.4	92.0
Additional allowed		7.2	7.4
Generously allowed		0.3	0.7
Disallowed regions		0.0	0.0

^aSeMet-Vsp^34-214 statistics are given for merged Bijvoet pairs.

^bR_merge = (∑_i∑_j < I_j − <I>)/(∑_i∑_j<I>), where ∑_i is taken over all reflections, ∑_j is taken over all observations of each reflection, and <I> is the mean intensity of all observations of reflection i.

^cR = ∑ ‖F_o| - |F_c‖ / ∑ |F_o|, where ∑ is taken over all reflections included in model refinement, |F_o| is the amplitude of the observed structure factor, and |F_c| is the amplitude of the calculated structure factor. R_free is taken over the test set of reflections not included in model refinement.

Structure determination.

Molecular-replacement solutions were obtained initially for the monoclinic Vsp1^34-214 crystal using the program Amore (48). The search model was the B. burgdorferi strain B31 OspC dimer (Protein Data Bank [PDB] identifier 1GGQ, chains A and B) (39). Four dimers were found within the asymmetric unit. Each chain was replaced with a homology model for Vsp1 based on aligned OspC structures from B. burgdorferi strains B31 (1GGQ [39]), HB19 (1F1M [39]), and N40 (1G5Z [29]) and the Vsp1 sequence (UniProt entry O34000_BORTU) using the program Modeler (63). This model was improved by alternating refinement against the crystal diffraction data using CNS v. 1.1 (15), employing strong noncrystallographic symmetry (NCS) restraints, and adjustment to fit NCS-averaged and/or model-phased maps, but R and R_free remained above 35%.

Next, following the same molecular-replacement and refinement strategy, we obtained a partially refined model for the triclinic SeMet-Vsp1^34-214 crystal that included residues 39 to 201 of each chain (R = 0.304; R_free = 0.331). The eight strongest peaks (13.7 to 6.8 σ) of an anomalous difference Fourier map calculated with model phases corresponded precisely to the eight expected positions for residue 95 SeMet substitutions. Weaker peaks (5.3 to 4.4σ) were also found, corresponding to two of eight expected positions for residue 44 SeMet substitutions and five fortuitous nickel(II) cation sites. While this result confirmed the validity of the provisional model, attempts to further improve fit to the diffraction data were unsuccessful.

We subsequently acquired 2.2-Å X-ray diffraction data for the monoclinic Vsp1^45-202 crystal and obtained a molecular-replacement solution with Amore using the SeMet provisional model. Two dimers were found in the asymmetric unit. The model was improved by alternating refinement against the diffraction data using the maximum likelihood target of Refmac5 (47), employing strong NCS restraints, and adjustment to fit NCS-averaged and/or (2fo-fc) model-phased maps. The higher-resolution maps revealed that numerous side chains that had appeared to fit the Vsp1^34-214 density maps had incorrect conformations, particularly branched residues Ile, Leu, Val, and Thr. The Molprobity web server was used to identify additional residues with incorrect side chain conformations (24). The solvent structure was built automatically using the ARP/wARP option (55) of Refmac5 as implemented in CCP4i v. 5.0 (56). Individual isotropic temperature factors were calculated for each atom, and translation-libration-screw parameters were calculated for each chain (70). The final Vsp1^45-202 model includes residues 45 to 201 for chains A, B, and E; residues 47 to 201 for chain D; and 216 solvent atoms.

We next used the Vsp1^45-202 structure as a molecular-replacement model to redetermine the structure of the monoclinic Vsp1^34-214 crystal. The model was improved by following the procedure described above; terminal residues were added if observed in electron density. The final Vsp1^45-202 model included residues 38 to 201 for all eight chains, 14 nickel(II) cations, and 135 solvent atoms.

Additional model statistics are given in Table 2.

Homology modeling.

A structural-homology model for the B. turicatae Vsp2 dimer (uniprot sequence entry Q9Z6I0_BORTU) was calculated using DeepView (33). Four crystal structures were used as templates: Vsp1 (1YJG; 60.3% identity) and B. burgdorferi OspC structures from strains B31 (1GGQ; 45.8% identity), HB19 (1F1M; 41.8% identity), and N40 (1G5Z; 47.3% identity) (29, 39). The sequence alignment was hand edited to move two insertions in Vsp2 relative to Vsp1 from buried to exposed positions. A homology model of Vsp2 was then calculated as a full dimer with the Swissmodel server using structurally aligned dimer templates. The resulting structure was asymmetric with respect to the VR1 region (defined in Results and Discussion). In particular, the conformation of residues 90 to 92 near the molecular twofold-symmetry axis differed. From each chain, twofold-symmetric homology models of Vsp2 were constructed. One model was considered improbable because of significant stereochemical clashes and exposure of Leu 91 to the solvent. The second model had reasonable stereochemical complementarity and a buried position for Leu 91. This second symmetric model was used in the comparison with Vsp1.

Structure analysis.

Eighteen Vsp sequences from B. turicatae and B. hermsii (excluding translations of silent genes) were aligned using ClustalW v. 1.82 (21). The resulting alignment is presented in the supplemental material. The Shannon entropy was calculated for each residue position (http://bio.dfci.harvard.edu/Tools/). The calculated values were transferred to the B-factor column of the Vsp1 dimer PDB file and plotted on the Vsp1 dimer surface.

Electrostatic-potential maps were calculated with APBS (2) via a PyMol plug-in (http://www-personal.umich.edu/∼mlerner/PyMOL/). Charges were assigned using the PDB2PQR server (26). Cavities and pockets were identified using CASTP (12).

Graphics.

Molecular images were generated using PyMol (http://pymol.sourceforge.net/). Coordinates for tripalmitoyl-S-glycerylcysteine (Pam₃Cys) were obtained using CACTVS (http://www2.ccc.uni-erlangen.de/software/cactvs/) and the CORINA server (31). Coordinates for heparin were obtained from the crystal structure of basic fibroblast growth factor-heparin complex (PDB entry 1BFC) (30).

Protein structure accession numbers.

The coordinates are deposited in the Protein Data Bank (11), entries 1YJG (Vsp1^45-202) and 2GAO (Vsp1^34-214).

RESULTS AND DISCUSSION

Vsp1 core structure.

In posttranslational processing of neurotropism-associated B. turicatae surface lipoprotein Vsp1, the 18-residue N-terminal signal sequence is removed and the mature N-terminal cysteine residue is lipidated to become a tripalmitoyl-S-glycerylcysteine (Pam₃Cys). To obtain a crystal structure of Vsp1, the strategy of preparing recombinant material lacking the lipid anchor was used (28).

Vsp1^34-214 is a protease-resistant fragment that forms a stable homodimer and is missing only the first 16 N-terminal residues of the mature protein (73). We were able to obtain preliminary crystallographic models using X-ray diffraction data from crystals of native Vsp1^34-214 (resolution limit, 2.7 Å) and from a selenomethionine-substituted derivative (resolution limit, 2.9 Å), using known structures of B. burgdorferi OspC as molecular-replacement search models (∼45% sequence identity to Vsp1). However, efforts to complete refinement or to obtain higher-resolution diffraction data were not successful. We hypothesized that the presence of ambiguous crystal lattice interactions between neighboring Vsp molecules formed by largely disordered N and C termini was limiting the crystal order. We therefore redesigned the expression construct to produce a smaller core fragment lacking the disordered terminal residues, Vsp1^45-202. Diffraction data subsequently obtained to a resolution limit of 2.2 Å enabled completion of crystallographic-model refinement of both of the native fragments, Vsp1^34-214 (final model, R = 21.6%, R_free = 27.2%), and Vsp1^45-202 (final model, R = 20.6%, R_free = 26.5%) (for more details, see Materials and Methods and Table 2). The Vsp1^34-214 crystal coordinates encompass residues 38 to 201 of each chain; the unmodeled N- and C-terminal residues are completely disordered. The Vsp1^45-202 crystal coordinates encompass residues 45 to 201; only the C-terminal residue is disordered. The two structures are otherwise essentially identical. A representative section of electron density from the Vsp1^45-202 structure is shown in Fig. 1.

FIG. 1.

Representative electron density for the B. turicatae Vsp1^45-202 dimer, in stereo. A solvent-exposed region of the Vsp1 dimer is shown (gray bonds) with solvent atoms (gray spheres). Labels indicate residues involved in solvent-mediated salt bridges. Predicted protein-solvent and solvent-solvent hydrogen bonds are indicated by black dashed lines. The sigma-weighted model map (gray mesh) is contoured at 1.5 σ.

Vsp1 is a mainly α-helical, twofold-symmetric dimer (Fig. 2). Each subunit is composed of four long α-helices that form an antiparallel bundle (α1, residues 38 to 73; α2, residues 91 to 114; α3, residues 119 to 143; α5, residues 172 to 200) and two additional short structural elements (α4, residues 143 to 148, and a two-stranded β-sheet: β1, residues 77 to 80; β2, residues 83 to 86). The dimer interface is largely hydrophobic and buries 1,900 Ų on each subunit. The interface consists primarily of a twisted parallel association of the two α1 helices, with additional contributions by each subunit from the N terminus of α2, the C terminus of α5, and the short β-sheet.

FIG. 2.

Schematic representation of the B. turicatae Vsp1^45-202 dimer, in stereo. In each panel, the subunit at left is shaded light gray, and the subunit at right is shaded dark gray. (Top) Side view. The protein region distal to the spirochetal membrane, i.e., the region most readily exposed to the host environment, is at the top; the two lipid-anchored N-termini proximal to the spirochetal membrane (residues 20 to 44), as well as the two free C termini (residues 203 to 214), would extend from the ends shown at the bottom. The five α-helices of the left subunit (α1 to α5), as well as N and C termini, are labeled. (Bottom) Top view. Selected α-helices and both β-strands of the left subunit are labeled.

The Vsp1 crystal structures confirm the close architectural identity of Vsps from RF borreliae and OspC from LD borreliae (29, 39). Pairwise least-squares superimpositions of Cα atoms between the Vsp1^45-202 dimer and any of the OspC dimer models from B. burgdorferi strains B31, N40, and HB19 (29, 39) yield root-mean-square deviations in the range 1.4 to 1.7 Å (302 to 304 common atom pairs). The largest overall structural difference is a slight lengthening of helices α2 and α3 in Vsp1 relative to OspC, the result of a five-residue insertion in the region of the α2-α3 turn (not shown). The close structural identity confirms that Vsp and OspC proteins share a conserved, compact fold.

Variation on a conserved scaffold.

The Vsp/OspC family lipoproteins are characterized by highly conserved N and C termini and a highly variable central span of ∼100 residues (see Pfam [9] entry PF01441, lipoprotein_6, at http://www.sanger.ac.uk). The conserved and variable parts partition into two anatomically distinct regions of the Vsp/OspC dimer fold: the four conserved termini together form an inner “stem,” while the variable central span forms a “dome” that surrounds the stem. These two regions are readily distinguished when viewing a map of sequence variation of RF borreliae on the Vsp1^45-202 structure (Fig. 3a; the inner stem is largely blue/green, and the outer dome is largely orange/red).

FIG. 3.

Variability of the Vsp1 dimer. The views are identical to Fig. 2. (a) Atoms (solid spheres) are colored according to their per-residue variability in 18 RF (B. turicatae and B. hermsii) Vsp sequences (see Materials and Methods). The most conserved residues are colored blue, the least conserved residues are colored red, and intermediate levels are green through orange. (b) Schematic cylinder representation of the Vsp dimer, with four variable regions identified by structure-sequence alignment highlighted as summarized in the key in red, green, yellow, and magenta, respectively. Conserved regions are blue.

Comparison of sequence alignments and Vsp/OspC structures reveals that the variable central span can be logically divided into four distinct variable regions (VR) based on the presence of three short intervening regions with highly conserved motifs (Fig. 3b; see the sequence alignment in the supplemental material). The intervening regions (residues 94 to 98, “LLAGA”; residues 138 to 140, “KLK”; residues 160 to 161, “IA”), together with the N and C termini, make direct contacts with each other and form a conserved scaffold. VR1 extends from the C terminus of α1 to the N terminus of α2 (Vsp1 residues 80 to 93). The two symmetry-related copies of VR1 are in direct contact with each other and form the central part of the dome. Because of variable sequence lengths (RF Vsps have up to four additional residues inserted in VR1 relative to Vsp1) and the requirement for self-association, VR1 is also the most troublesome region for homology modeling. VR2 spans most of α2 and α3 (residues 99 to 137) and forms the dome outer flanks. VR3, extending from the C terminus of α3 into α4 (residues 141 to 159), and VR4, comprising the “superhelical turn” between α4 and α5 (residues 162 to 174), together form the dome inner flanks. Definition of these four VRs provides a useful starting point for designing chimeras to investigate the structural basis of known functional differences between Vsps.

Variation in RF Vsps has also been examined at the genomic level (61). While the mechanism of recombination that produces the variation is unknown, the presence of repetitive nucleotide repeats within B. hermsii Vsp genes is highly suggestive of intracellular recombination between the multiple gene cassettes that exist within the genome of a single organism. Interestingly, the genomic level analysis identified one Vsp gene region in particular that is commonly repeated elsewhere at seemingly random gene positions. This recombination “hot spot” (nucleotides 200 to 250, residues 67 to 83) encompasses a highly conserved stretch of amino acid residues that begins near the C terminus of helix α1 and ends at the terminus of strand β1. Both of these secondary-structural elements are intimately involved in formation of the Vsp dimer interface. We suggest that the observed genomic repetition pattern may have arisen in part because of selection against recombinant Vsp genes with mutations in this region.

Vsp-OspC cavities and pockets.

A distinctive feature of the crystal structure of Vsp1 is a twofold-symmetric bilobed solvent-filled pocket 8 Šdeep and 15 Šin length that sits directly beneath the surface of the dome center. Between two and four ordered solvent molecules (up to two per lobe) are observed in the six independent observations of this pocket in the Vsp1^45-202 and Vsp1^34-214 crystal structures; in principle, the pocket is large enough to fit a small ligand (the volume based on the molecular surface is 266 ų). Similarly distinctive solvent-filled pockets or cavities are present at equivalent positions in each of the OspC structures. Based on two OspC structures, it was hypothesized that cavities beneath the dome might play an important functional role, e.g., perhaps forming part of a binding site for an unknown ligand (39). With four high-resolution examples of Vsp-OspC family members now available, we decided to look further into this possibility.

We found that the cavities and/or pockets of Vsp1 and the three OspC structures near the dome center vary substantially in size or volume, shape, and connectivity to the surface (Fig. 4). Two of the four structures (Vsp1 and OspC from strain N40) possess bilobed pockets with surface-accessible openings, while the other two structures (OspCs from strains B31 and HB19) possess enclosed cavities. The main reason why the cavities/pockets do not share overall shape is that they are lined largely with variable residues from VR1 and VR3. Given the high variability in the structures of the four Vsp-OspC members, it appears unlikely that the cavities or pockets are binding sites for a common ligand. A possible alternative explanation for their presence is that in the evolution of this family of proteins, variation of the dome to evade host immunity has taken precedence over optimal “packing” of the dimer interface.

FIG. 4.

Cavities and pockets of B. turicatae Vsp1 and B. burgdorferi OspC proteins. In each panel, the molecular surface of the indicated dimer is viewed from the inside, looking down the twofold-symmetry axis toward the top (membrane-distal end). OspC structures are from PDB entries 1GGQ (strain B31), 1F1M (strain HB19), and 1G5Z (strain N40) (29, 39). Fully enclosed cavities are labeled “c”; deep pockets of the molecular surface are labeled “p.”

Model for native, lipidated Vsp1.

The crystal structures described here constitute ∼80% by mass of the complete Vsp1 lipoprotein. The missing 20% encompasses the most highly conserved regions of the Vsp/OspC lipoprotein amino acid sequences: the N and C termini. We present here a rough theoretical model of the overall structure of Vsp1 anchored into a spirochete membrane by a pair of N-terminal Pam₃Cys lipid moieties (Fig. 5). Missing residues are represented as continuous α-helices in order to represent a minimum distance between lipid anchor and dimer core (∼40 Å), but their true conformations are unknown. In a fully extended conformation (not shown), the distance would be approximately doubled. The actual distance is likely to lie between these two extremes.

FIG. 5.

Model for the complete Vsp1 dimer. The Vsp1^45-202 dimer core structure is shown with its N and C termini extended to represent two complete lipidated Vsp1 chains (residues 19 to 214). Both termini are arbitrarily displayed in α-helical conformation. For the N terminus, the α-helical conformation yields the likely minimum distance between the Pam₃Cys lipid anchors and the dimer core (40 Å). The distance would approximately double (80 Å) with full extension of the polypeptide chains.

Several lines of evidence suggest that native Vsp1 dimer N and C termini either adopt a mostly “random-coil” conformation or are disordered. First, in crystals of Vsp^34-214, terminal residues 34 to 38 and 202 to 214 are completely disordered. Second, there are single conserved proline residues within each native terminus (Pro 26 and Pro 204) that would interrupt helical structure. Third, on B. turicatae spirochetes, both N and C termini are labile to trypsin (73). The evidence for disorder of Vsp1 termini essentially parallels similar evidence obtained for OspC (29, 39). It therefore appears likely that Vsp (and OspC) cores are attached to the spirochete membrane through a flexible double tether, with two flexible C-terminal “tail” extensions. It is curious that these termini, which constitute the most highly conserved amino acid sequences for the Vsp-OspC family, are outside of folded regions. The termini may be conserved for other reasons, e.g., to facilitate lipoprotein secretion (64) or to mediate interactions with other bacterial surface proteins or host factors.

Our model indicates a proximal position for the conserved stem and a distal position for the variable dome of Vsp1 relative to the spirochete membrane. This would make the variable dome regions readily accessible to host molecules, such as serotype-specific neutralizing immunoglobulin M antibodies (1, 5, 22, 46, 71) or extracellular-matrix components (see below). However, if the double tether is highly flexible, the position and orientation of the core could vary with respect to the spirochete membrane, possibly sitting sideways or even upside down. Electron microscope images of antibody-labeled B. burgdorferi indicate that the C-terminal region of OspC is exposed on the surface of the spirochete (45).

Vsp electrostatic potential and tissue tropism.

Homology modeling was employed by Kumaran et al. (39) in the comparison of representative B. burgdorferi OspCs from each of 21 major ospC allele groups. Electrostatic-potential maps of OspC models had revealed a surprising correlation between surface protein electrostatic potential and phenotypic behavior of B. burgdorferi. In each of the four OspC groups known to be invasive in humans, but not in any other OspC groups, a strong negatively charged patch was observed at the dome center of the Vsp/OspC fold. This suggested that surface protein electrostatic potential affects the mobility of spirochetes through mammalian tissues. One caveat regarding this analysis, however, is that OspC proteins were compared between nonisogenic B. burgdorferi strains. Thus, the phenotypes attributed to OspC might instead be attributable to other genetically linked determinants.

In contrast, the phenotypic differences between two isogenic B. turicatae serotypes appear to be entirely the consequence of differences in expressed Vsp genes (54). One goal of our studies is to identify structural and/or physicochemical differences between Vsp1 and Vsp2 that could be responsible for the phenotypic differences, particularly with regard to tissue localization. Since we have yet to succeed in crystallizing Vsp2 (the construct Vsp2^34-214 tends to aggregate in solution), we have performed an initial structural comparison between the crystal structure of Vsp1 and a homology model for Vsp2.

Vsp2 binds significantly better to extracellular-matrix GAGs, such as heparin and chondroitin sulfate, than Vsp1 (44, 74). The specificity of GAG-protein interactions is governed by the ionic interactions of the sulfate and carboxylate groups of GAGs with the basic amino acids on the protein, as well as the optimal structural fit of a GAG chain into the binding site of the protein (58). Heparin-binding sites are commonly observed on the external surfaces of proteins and correspond to shallow pockets of positive charge (20).

Electrostatic-potential maps for the Vsp1 structure and Vsp2 homology model (Fig. 6) show that Vsp1 and Vsp2 share a negative-charge feature at their dome centers, similar to that seen for OspCs from invasive groups. However, Vsp2 also has more predicted regions of positive charge, including areas within the stem, the dome arms, and to either side of the dome center. This pronounced increase in positive-electrostatic-potential patches may be sufficient to explain the relative propensity of Vsp2 to bind acidic GAGs. Consistent with this idea, sulfation of GAGs, which establishes a net negative charge, is required for spirochete binding to mammalian cells (44). Electrostatic interaction with extracellular-matrix molecules may therefore be an important factor in defining phenotypic behavior: while B. turicatae serotype 2 cells expressing Vsp2 prefer to interact with blood vessel, heart, and joint tissues, neurotropic serotype 1 cells displaying Vsp1 may simply avoid getting stuck before reaching their niche.

FIG. 6.

Predicted electrostatic potential, Vsp1 versus Vsp2. Electrostatic potential is plotted at the molecular surface of the Vsp1 dimer crystal structure (left) and the Vsp2 dimer homology model (right). Deep red corresponds to an electrostatic potential of −8 kT/e; deep blue corresponds to +8 kT/e. The view is identical to Fig. 2, top. The structure of a six-residue fragment of the glycosaminoglycan heparin is shown at center (sulfur, oxygen, nitrogen, and carbon are orange, red, blue, and gray, respectively). Vsp2 has been shown to bind heparin, but Vsp1 does not. Electrostatic-potential and homology model calculations are described in Materials and Methods.