Structural Conservation of Neurotropism-associated VspA within the Variable Borrelia Vsp-OspC Lipoprotein Family*

Vsp surface lipoproteins are serotype-defining antigens of relapsing fever spirochetes that undergo multiphasic antigenic variation to avoid the immune response. One of these proteins, VspA of Borrelia turicatae, is also associated with neurotropism in infected mice. Vsp proteins are highly polymorphic in sequence, which may relate to their specific antibody reactivities and host cell interactions. To determine whether sequence variations affect protein structure, we compared B. turicatae VspA with three related proteins: VspB of B. turicatae, Vsp26 of the relapsing fever agent Borrelia hermsii, and OspC of the Lyme disease spirochete Borrelia burgdorferi. Recombinant non-lipidated proteins were purified by affinity or ion exchange chromatography. Circular dichroism spectra revealed similar, highly α-helical secondary structures for all four proteins. In vitro assays demonstrated protease-resistant, thermostable Vsp cores starting at a conserved serine at position 34 (Ser34). All proteins aggregate as dimers in solution.In situ trypsin treatment and surface protein cross-linking showed that the native lipoproteins also form protease-resistant dimers. These findings indicate that Vsp proteins have a common compact fold and that their established functions are based on localized polymorphisms. Two forms of VspA crystals suitable for structure determination by x-ray diffraction methods have been obtained.

Relapsing fever and Lyme disease are bacterial infectious diseases caused by spirochetes of the genus Borrelia. In contrast to other spirochetes that affect humans, such as the syphilis agent Treponema pallidum, their transmission to vertebrate hosts depends on arthropod vectors (1). Throughout their vector-host life cycle, Borrelia cells display abundant lipoproteins on their surfaces. The genomic sequence of the Lyme disease spirochete Borrelia burgdorferi revealed over 130 putative lipoprotein genes on its chromosome and plasmids (2,3). With the exception of 3 matrix-binding proteins (4,5), their specific functions have not been determined.
Included in this class of lipoproteins with unknown function is the tick-associated Outer Surface Protein A (OspA) 1 (6), which is in current use as a Lyme disease vaccine for humans (7). OspA is anchored to the outer membrane via an N-terminal triacyl-modified cysteine (8,9) and is intrinsically resistant to proteases such as trypsin, despite its surface accessibility and high lysine content (10 -12). The OspA crystal structure revealed 21 antiparallel ␤-strands and one C-terminal ␣-helix, which fold into globular N-and C-terminal domains connected by an unusual, freestanding nonglobular ␤-sheet (13)(14)(15).
Less is known about the structure of the major outer membrane lipoproteins of relapsing fever Borreliae, which are immunodominant and determine serotype (16). Collectively called variable membrane proteins, they have been divided into two families: the variable large proteins (Vlps) of 36 -40 kDa and the variable small proteins (Vsps) of 20 -23 kDa (17). Vlp and Vsp proteins appear unique to the genus Borrelia. They have been described in the relapsing fever species Borrelia turicatae (17)(18)(19)(20), Borrelia hermsii (21), Borrelia recurrentis (22), and Borrelia crocidurae (23). B. burgdorferi OspC, a major outer surface protein in early Lyme disease (24), is phylogenetically related to the Vsps (17), which has led to the term Vsp-OspC family.
While their signal peptides necessary for proper translocation and processing are conserved, the mature Vlp and Vsp-OspC lipoproteins are highly polymorphic. OspC amino acid sequences of different Lyme disease Borreliae can vary as much as 25% (25,26). There is evidence that this variation is maintained, even within a local population, by frequency-dependent balancing selection (27). Vlp and Vsp proteins diverge even more, with 40 to 80% amino acid identities among them (17,28). In contrast to the single plasmid-encoded ospC of B. burgdorferi (29), several 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 (21,30,31). The resulting multiphasic antigenic variation of Vsps and Vlps allows the spirochete to repeatedly evade the host's immune response, which leads to recurrent spirochetemia and the characteristic febrile episodes (16). This strategy is analogous to that for variant surface glycoproteins (VSG) of African trypanosomes, the PfEMP1 proteins of the malarial parasite Plasmodium falciparum, and the PilE/PilS and Opa surface proteins of Neisseria species (32,33).
Experimental relapsing fever in mice has demonstrated a role for Vsp proteins in differential tissue localization as well as for avoidance of the immune response. In a clonal population of B. turicatae, expression of VspB was associated with high densities of spirochetes in the blood, while expression of VspA led to early invasion and persistent infection of the central nervous system (17-19, 34, 35). VspA may help to evacuate the bacterium to an immunopriviledged niche, while VspB may facilitate efficient transmission to the next feeding tick. Tissue culture assays yielded clues about pathogenesis mechanisms. B. turi-catae expressing VspA penetrated human umbilical vein epithelial cell monolayers more readily than those expressing VspB (18). On the other hand, VspB increased binding of the spirochete to mammalian endothelial and glial cells (36), predominantly by the direct interaction of VspB with host cell surface glycosaminoglycans (37).
Structural data for Vsp-OspC proteins are limited. NMR studies of B. burgdorferi strain B31 OspC have suggested that the core of OspC consists of four ␣-helices, while the N-and C-terminal sequences are highly flexible (38). OspC dimers were observed both with recombinant protein in solution (38) as well as with cell surface-exposed native lipoprotein (11). However, the effect of sequence variation on Vsp-OspC protein structure, and thus the structural basis of their established functions, remained unknown. To address this, we determined structural features of VspA and three other members of the Vsp-OspC protein family by biochemical and biophysical methods. A comparison indicates that Vsp-OspC proteins share a highly ␣-helical, compact fold, which includes a dimerization domain and confers protease resistance to a large central portion of the Vsps. For high resolution x-ray structure determination, we obtained two forms of VspA crystals.

Expression and Purification of Recombinant Histidine-tagged Fusion
Proteins-Non-lipidated, soluble fusion proteins with N-terminal histidine tags (hisVspA, hisVspB, hisVsp26, and hisOspC) were obtained as described for hisVspA (17). Briefly, gene segments lacking the N-terminal signal sequence and lipidation site were amplified by polymerase chain reaction using Taq thermostable DNA polymerase (Roche Molecular Biochemicals). Forward and reverse primers for hisVspB, his-Vsp26, and hisOspC are listed in Table I. The polymerase chain reaction products were ligated into pET15b (Novagen) and cloned in Escherichia coli BL21(DE3) (Novagen). Recombinant proteins were expressed after induction with isopropyl-␤-D-thiogalactopyranoside and subsequently purified by nickel affinity chromatography as described (17,19). All four proteins eluted Ͼ95% pure at an imidazole concentration of 300 mM. Recombinant, non-lipidated OspA was provided by R. Huebner (Aventis, Swiftwater, PA).
Expression and Purification of N-terminal Truncated Core Proteins-To obtain core proteins S34VspA and S34VspB, vspA and vspB sequences starting with the Ser 34 codon and ending with the translational stop codon were amplified from plasmid clones (see above and Refs. 17 and 19) by polymerase chain reaction using Pwo polymerase (Roche Molecular Biochemicals). The forward primer for both sequences was S34vspAB-fwd. Reverse primers were S34vspA-rev and S34vspBrev, respectively (see Table I). The polymerase chain reaction products were digested with NdeI and BamHI and ligated into pET29b (Novagen). Recombinant plasmids pET29::S34vspA and pET29b::S34vspB were obtained after transformation of E. coli Inv␣FЈ (Invitrogen). E. coli BL21(DE3) (Novagen) transformants containing the plasmids were grown overnight from single colonies at 30°C in LB containing 30 g of kanamycin/ml. A 1:50 dilution of the overnight cultures was grown at 37°C in Terrific broth (39) with kanamycin. Expression was induced with isopropyl-␤-D-thiogalactopyranoside at a final concentration of 1 mM when the culture OD 595 reached 0.9. Incubation continued overnight at 30°C. Cells were harvested by centrifugation at 8,000 ϫ g for 10 min at 4°C and resuspended in 0.2 ϫ volume of anion exchange column buffers: (i) 20 mM Tris hydrochloride, 0.02% NaN 3 , pH 9.0 (AIEX-A), for S34VspA and (ii) 20 mM ethanolamine hydrochloride, 0.02% NaN 3 , pH 10.0 (AIEX-B), for S34VspB. Cleared lysates were obtained by ultrasonication, centrifugation at 15,000 ϫ g for 10 min and passing through a 0.45-m pore size filter (Nalgene).
Cleared lysates were passed over a HiPrep 16/10 Q XL anion exchange column (Pharmacia) in AIEX-A or -B column buffers. Under the respective conditions both S34VspA and S34VspB bound quantitatively to the matrix. The proteins were eluted in a linear gradient of AIEX-A or -B column buffers containing 0 to 200 mM NaCl. S34VspA and S34VspB eluted at salt concentrations of 80 and 130 mM, respectively. Peak fractions containing the recombinant proteins were pooled, concentrated using Centricon Plus-20 centrifugal concentrators (Amicon), and dialyzed overnight against deionized H 2 O at 4°C. Protein aliquots were flash-frozen on dry ice and stored at Ϫ80°C.
Proteolytic and Cross-linking Assays-Recombinant purified proteins were treated separately with four proteases. The two serine proteases, trypsin and plasmin (Roche Molecular Biochemicals), were used at 10 to 400 g/ml or 1 mg/ml, respectively, in 100 mM Tris-HCl, pH 8.5, while lysosomal cysteine protease cathepsin B and aspartyl protease cathepsin D (Calbiochem) were used at concentrations of 1.5 units/ml in 250 mM Na citrate, 1 mM EDTA, 2 mM dithiothreitol, pH 5.5. Reactions were incubated at 37°C for 1 h and then stopped with Roche Complete protease inhibitor mixture supplemented with 1 M pepstatin.
For in situ protease treatment, B. turicatae Oz1 serotype A cells were grown in BSK II medium and harvested as described (40). Intact Borrelia cells were treated in situ with trypsin (Roche Molecular Biochemicals) as described (41). Cross-linking of surface-exposed proteins was performed using formaldehyde as described for B. burgdorferi OspC (11).
Protein Gel Electrophoresis and Immunoblot Analysis-Proteins were separated on 12% polyacrylamide-SDS gels and visualized by Coomassie Blue staining. For immunoblots, proteins were electrophoretically transferred to nitrocellulose membranes (Immobilon-NC, Millipore) as described (42). Membranes were rinsed in 20 mM Tris, 500 mM NaCl, pH 7.5 (TBS), and either air-dried or processed directly. 5% Dry milk in TBS with 0.05% Tween 20 was used for membrane blocking and subsequent incubations for 1 h each. TBS with 0.05% Tween 20 alone was used for intervening washes. Rabbit antisera were used at 1:500 dilutions, and mouse monoclonal antibody hybridoma supernatants at a 1:10 dilution. Alkaline phosphatase-conjugated Protein A/G (ImmunoPure Protein A/G CIP conjugate, Pierce) at a 1:5000 dilution was used as the second ligand, and a stabilized alkaline phosphatase substrate solution (1-Step NBT/BCIP, Pierce) was used for colorimetric detection.
N-terminal Protein Sequence Analysis and Mass Spectrometry-Nterminal protein sequences of recombinant proteins were determined on a Model 477A protein sequencer (Applied Biosystems) from protein samples previously purified by gel electrophoresis and transfer to a polyvinylidene difluoride membrane (43). Molecular masses of purified recombinant proteins were determined by matrix-assisted laser desorption ionization-time of flight on a Voyager biospectrometry workstation (Perseptive Biosystems) using sinapinic acid as matrix.
Circular Dichroism Spectroscopy and Secondary Structure Prediction-Circular dichroism (CD) spectra were obtained using a Jasco J-720 spectropolarimeter. Proteins were diluted to a concentration of 20 M in deionized H 2 O. Samples were placed in 0.1-mm path length cells at 25°C and spectra were acquired using a scan speed of 20 nm/min, response time of 1 s, bandwidth of 1.0 nm, and step resolution of 0.5 nm. 20 Acquisitions between 260 and 180 nm wavelength were co-added. Spectral data were deconvoluted using the DICROPROT program suite (44) and secondary structure ratios were determined by the least squares fit method using the Fasman algorithm. Predictions of second- The underlined sequences were added to introduce restriction sites and are noncomplimentary to the Borrelia sequences.
ary structure were obtained using software available on the protein sequence analysis server at the Biomolecular Engineering Research Center of Boston University (45). Gel Filtration Liquid Chromatography-For size determination, proteins were separated on a HiPrep 16/60 Sephacryl S-200 high-resolution column (Amersham Pharmacia Biotech). The column was first equilibrated with column running buffer (20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 0.2% NaN 3 ) and then calibrated separately with blue dextran 2000 (for void volume determination), albumin (67 kDa), chymotrypsinogen A (25 kDa), ovalbumin (43 kDa), and ribonuclease A (13.7 kDa) (Amersham Pharmacia Biotech). Approximately 10 mg of concentrated recombinant proteins previously purified by anion exchange chromatography were loaded in 1 ml of column running buffer. For each protein, the partition constant K av was calculated from the peak elution volume using the following formula: where V e ϭ elution volume, V o ϭ column void volume, and V t ϭ total bed volume.
Crystallization of S34VspA and preliminary diffraction analysis-Recombinant purified S34VspA was dialyzed against deionized H 2 O, then diluted to 8 mg/ml. Crystals suitable for structure determination by x-ray diffraction methods were obtained using standard vapor diffusion techniques. Aggregated thin plates were produced in conditions numbers 6 and 22 during initial screening with Hampton Research sparse matrix kit I. These conditions share precipitant and buffer conditions of 30% PEG 4000 and 0.1 M Tris/HCl, pH 8.5, but differ in salt additive of 0.2 M MgCl 2 or 0.2 M NaCH 2 COOH, respectively. The precipitant, salt, and pH were varied systematically about these two conditions to optimize crystal size and quality (see "Results"). S34VspA crystals were characterized using a conventional rotating anode x-ray source and native diffraction data were collected at beamline X12C of the Brookhaven National Laboratory National Synchrotron Light Source on a CCD detector. Diffraction images were processed with DENZO/SCALEPACK (46).

Vsp-OspC Protein
Purification-While pathogenesis-associated functions of Borrelia Vsp proteins, e.g. in antigenic variation and tissue tropism, have been established, the structural basis for these functions remain unknown. One question is how the observed primary protein sequence polymorphisms affect secondary, tertiary, and quarternary protein structure. Beginning to address this by biochemical and biophysical methods, we first expressed VspA of B. turicatae in E. coli. We chose this protein to focus on because of its association with nervous system invasion. Two related proteins, VspB of the same strain of B. turicatae and Vsp26 of B. hermsii, as well as the more distantly related OspC of B. burgdorferi, were also expressed. Their conserved signal peptides and overall amino acid identi-ties of 40 to 70% are representative of the entire Vsp-OspC protein family (Fig. 1). The proteins were purified in an Nterminal histidine-tagged form by standard nickel affinity chromatography. In these constructs (hisVspA, hisVspB, his-Vsp26, and hisOspC), the Vsp-OspC sequences start with residue Asn 20 (Fig. 1), i.e. they lack the signal peptide and acylation site at Cys 19 . The recombinant proteins therefore were soluble, but otherwise mimic the full-length processed form.
Protease Resistance of Vsp Core Proteins-Previous studies showed that B. burgdorferi OspA and OspC are highly resistant to trypsin in situ, i.e. when expressed in a native, lipidated form on the surface of the bacteria (10,11) and that protease resistance is an intrinsic property of recombinant OspA (12). To determine whether VspA and the related proteins are also protease resistant, we incubated recombinant hisVspA, his-VspB, hisVsp26, and hisOspC with trypsin at concentrations ranging from 12.5 g to 400 g/ml for 1 h at 37°C. Recombinant, non-lipidated OspA and casein (Sigma) were used as controls (Fig. 2).
As expected, casein was digested to completion, while recombinant OspA was unaffected by trypsin. In contrast to the study of in situ OspC (11), several independently purified recombinant hisOspC preparations did not exhibit comparable trypsin resistance in vitro. This difference in results suggests that OspC is either stabilized by the lipid-mediated attachment to the cell surface, or that the recombinant hisOspC obtained from our construct is destabilized by the additional, non-native sequence at its N terminus.
Although the processed VspA, VspB, and Vsp26 proteins have 27 to 30 predicted trypsin sites distributed over the whole length of their protein sequences, the three proteins were merely truncated 5 to 8 kDa by trypsin digestion (Fig. 2). Similar results were obtained with plasmin, which has trypsinlike activity. Yet, to reach an equivalent cleavage efficiency, plasmin concentrations had to be increased to 1 mg/ml, which corresponds to about 10 times the physiological concentration (47). This indicated that most of the sites were not accessible to the protease, thus yielding a protease-resistant core. No further truncations were observed with trypsin concentrations of up to 10 mg/ml, and VspA maintained its protease-resistant core even after being boiled for 10 min (not shown). Protein sequencing of the trypsin-treated VspA, VspB, and Vsp26 protein cores revealed the N-terminal amino acids in all three FIG. 1. Amino acid sequence alignment of VspA, VspB, Vsp26, and OspC. The protein sequence alignment was performed using CLUSTAL X (71). Color coding of conserved residues is according to Gonnet Pam250 matrix groups. Asterisks (*) in the alignment header indicate fully conserved amino acid residues, colons (:) indicate conservation of "strong" Gonnet Pam250 matrix group residues, and periods (.) conservation of "weaker" group residues. The shaded area of the alignment header indicates the protease resistant core determined for VspA, VspB, and Vsp26 (see also Fig. 2). An L denotes the lipid-modified cysteine (C19). Shaded bars and black frames in the alignment footer indicate predicted Vsp and NMR-determined OspC ␣-helices (␣1 to ␣4) (38), respectively (see also Fig. 5B). Amino acid residues are numbered to the right of the alignments according to the full-length, unprocessed protein sequences. cases to be SDGTV. This sequence can be found starting at a conserved serine at position 34 (Ser 34 ) (Fig. 1).
To purify the protease-resistant cores and for further structural analysis of VspA (see below), we constructed N-terminal truncated versions of VspA and VspB starting at Ser 34 (S34VspA and S34VspB). A methionine codon was added to initiate translation. S34VspA and S34VspB proteins were purified to Ͼ95% purity from crude E. coli lysates in a single-step anion-exchange chromatography step. Trypsin truncated S34VspA and S34VspB by approximately 1 kDa, revealing a succeptible site at the C terminus as well (Fig. 2). The location of this site, i.e. the C terminus of the core protein, was derived from matrix-assisted laser desorption ionization-time of flight mass spectrometry data. The mass of undigested S34VspA was determined as a control, yielding a major peak of 18,507 Da. It corresponds to a protein starting with Ser 34 and ending with Asn 214 with a predicted size of 18,475 Da. As confirmed by N-terminal sequencing, the f-Met was removed, probably by the E. coli methionyl aminopeptidase (48). For trypsin-treated S34VspA, the mass of the major peak was 17,739 Da, which corresponds to a predicted 17,700 Da protein starting with Ser 34 and ending with Lys 206 . The mass difference of 768 Da would also match the removal of 8 C-terminal amino acids, a predicted 792 Da.
To assess the resistance to other physiologically relevant proteases, we treated the proteins with two lysosomal proteases. Cathepsin B is a broad specificity cysteine endopeptidase with preference for R-R bonds, but also has peptidyl dipeptidase activity, liberating C-terminal dipeptides. Cathepsin D is an aspartic endopeptidase with pepsin A-like activity, cleaving F-V, Q-A, A-L, L-Y, Y-L, G-F, F-F, and F-Y bonds (49). Incubation conditions were derived from the ones used to study antigen presentation of ovalbumin (50). Casein was digested by both proteases, while OspA appeared unaffected. Neither VspA, VspB, Vsp26, nor OspC was cleaved by cathepsin D, although they contain 1 to 4 predicted endopeptidase sites (Fig. 1).
Cathepsin B treatment resulted in several distinct protein bands for all Vsp proteins. Interestingly, the smallest bands for VspA and VspB, presumably the proteolytic end products, were similar if not identical in size to the ones obtained with trypsin. We therefore performed mass spectrometry on cathepsin Bdigested S34VspA. The mass of the final proteolytic product, 17,733 Da, was almost identical to trypsin-digested VspA. This suggests that cathepsin B removes 8 amino acids from the C terminus of S34VspA, which would be consistent with its peptidyl dipeptidase activity. Together, these data indicate that Vsp proteins, although differing in primary amino acid sequence, have a conserved core that is resistant to proteases with different specificities.
Vsp-OspC Protein Dimerization In Solution-Huang et al. (38) have recently shown that a truncated recombinant form of B. burgdorferi OspC forms a 37-kDa dimer in solution. To see if this is a common structural feature of the Vsp-OspC protein family, we determined the size of the recombinant proteins by gel filtration. Purified recombinant proteins were loaded on a high-resolution Sephacryl column, which had been calibrated with protein size standards. The apparent molecular masses were calculated from the partition coefficients to be as follows: hisVspA, 43.5 kDa; hisVspB, 42.5 kDa; hisVsp26, 49.2 kDa; hisOspC, 49.2 kDa; S34VspA, 38.5 kDa; S34VspB, 40.5 kDa (see Fig. 3). The predicted monomer sizes for the recombinant proteins range from 18.6 kDa (S34VspA) to 23 kDa (hisVsp26). This shows that VspA, like the other relapsing fever Vsp proteins and the related OspC, forms dimers in solution.
VspA Dimerization and Protease Resistance In Situ-In a next step, we investigated whether the dimerization and protease resistance of Vsp proteins could also be observed in situ, i.e. on the surface of intact Borrelia cells. Dimerization has recently been shown for in situ B. burgdorferi OspC (11). We applied the same approach to VspA expressed by B. turicatae Oz1 serotype A. Cross-linking of closely associated surface proteins by formaldehyde revealed that VspA dimers form also in situ (Fig. 4A).
We next assessed whether the native VspA protein dimers also exhibited the protease-resistant properties observed with the recombinant protein. Accessibility of the C-terminal trypsin sites would lead to a truncated, yet cell-associated protein.
If the N-terminal sites in both dimer subunits were accessible as well, the protein parts containing the two lipid anchors would be cleaved and the cores would be released into the supernatant. We therefore incubated intact B. turicatae cells with trypsin and then analyzed the pelleted cell-associated proteins and solubilized proteins in the supernantant. VspA was detected using an anti-VspA polyclonal antibody (Fig. 4B). In untreated B. turicatae cell pellets, VspA was found in its full-length lipidated form running at about 22 kDa. No protein was detectable in the supernatant. In the cell-associated fraction of trypsin-treated B. turicatae cells however, two additional forms of VspA approximately 20 and 18 kDa in size were detected, with the 20-kDa band being the most prominent. In the supernatant, only the 18-kDa band was detected. The 20-kDa band likely represents the C-terminal truncated lipidated protein, while the 18-kDa band corresponds in size to the VspA core. This indicates that both the N-and C-terminal domains were accessible to proteases such as trypsin. It also suggests that the VspA core can still be anchored to the surface via association with a lipidated dimer subunit, and that release into the surrounding environment occurs only after cleavage of both membrane anchor-containing peptides.
Conserved Vsp-OspC Secondary Structure-We used CD spectroscopy as a first approach to compare the secondary structures of the recombinant Vsps and OspC (Fig. 5A). The obtained mean residue ellipticity values were plotted against the wavelength. Peaks at 192 nm and two troughs at 208 and 222 nm are indicative of predominantly ␣-helical secondary structures. The CD spectra of VspA, VspB, Vsp26, and OspC overlap extensively, thus the high helix content appears conserved among the Vsp-OspC protein family. Deconvolution of spectral data lead to estimates of secondary structure ratios. For histidine-tagged proteins, the mean values for ␣-helical, ␤-sheet, and coiled secondary structure were 71, 2, and 27%, respectively. For the N-terminal truncated proteins, the coiled structure content decreased to 16%, with ␣-helix and ␤-sheet ratios of 80 and 4%.
As we currently lack further secondary or tertiary structure information on the Vsps, we resorted to computer algorithms for further comparisons. While a comparison of predicted secondary structures confirmed the experimental data of highly ␣Ϫhelical proteins, they also revealed further conservation. All studied proteins are calculated to fold into 4 helices separated by turns (Fig. 5B). Furthermore, the N and C termini likely form flexible, unordered tails of approximately 20 and 10 amino acids, respectively. Both these structural features are in accordance with the NMR-derived data for OspC (38).
Crystallization of S34VspA-To date, B. burgdorferi OspA is the only Borrelia lipoprotein whose crystal structure has been determined (13)(14)(15). Initial attempts to crystallize the histidine-tagged forms of VspA and VspB were unsuccessful. We surmised that the predicted flexible, unordered N-and C-terminal sequences could inhibit crystallization. In our next set of experiments, we therefore used S34VspA, which lacks most of the flexible N-terminal tail.
We have collected two complete diffraction data sets at 100 K. Dataset I (S34VspA/nickel chloride) has 38,032 reflections merged from 130,123 individual measurements and is 98.9% complete to a resolution of 2.7 Å, with a linear R-merge of 4.2% (14.7% for the highest resolution shell, 2.8 -2.7 Å). Dataset II (S34VspA/magnesium formate) has 27,615 reflections merged from 84,782 individual measurements and is 96.2% complete to a resolution of 3.0 Å, with a linear R-merge of 6.7% (20.9% for the highest resolution shell, 3.11 -3.0 Å). In both cases the mosaic spread of the crystal was determined to be 0.95°. We have recently succeeded in obtaining selenomethionine-substituted crystals of S34VspA for structure determination using the multiple anomalous dispersion method (51). DISCUSSION This study has shown that the neurotropism-associated VspA of B. turicatae shares secondary, tertiary, and quarternary structural characteristics with other Borrelia Vsp-OspC family proteins, despite as little as 40% identity in primary sequence. CD spectroscopy demonstrated conserved, highly ␣-helical secondary structures that are predicted to fold into a four-helix bundle. Supporting this model were the findings that postswitch mutations in expressed vsp alleles are mainly conservative (30,52), and that non-conservative amino acid changes, insertions, and deletions cluster in regions outside the predicted helices (Fig. 1). We conclude that localized polymorphisms, e.g. variable loops, rather than differences in overall domain structures determine specific Vsp-OspC functions such as antigenic variation and niche selection. Further validating our model of structural conservation were experiments showing that the proteins had same size proteaseresistant cores and dimerized in a physiologic solution as well as on the bacterial surface. These shared characteristics are likely a consequence of a common protein fold and conserved residues. Supporting this, the Vsp cores coincide with the predicted helical domains. The predicted flexible unordered N-and C-terminal tails appear susceptible to proteolysis. In the case of trypsin, the enzyme active site must obtain access to the target peptide bond (K-X) to accomplish cleavage. This can only be achieved in regions where the main chain atoms are not involved in formation of stable secondary structural elements (53). In OspA, the lysines are in antiparallel ␤-strand sheets or short loops and therefore protected (15). Similarly, the participation of Vsp-OspC core lysines in ␣-helix hydrogen bonding would make these trypsin sites inaccessible. On the other hand, lysines in the loops may be shielded from proteolytic attack by dimer subunit interactions. A comparable mechanism has been observed with two B. burgdorferi surface proteins: OspA closely interacts with an integral membrane protein, P66, thereby protecting a surface-exposed loop of P66 from proteolysis (11).
The common, intrinsic property of Vsp-OspC proteins to dimerize appears mediated by sequences present in the protein cores. This conservation despite significant divergence between the Vsp and OspC proteins suggests that some of the invariant core residues participate in this process. In the absence of further structural information, a stretch of constant amino acids found at the C-terminal end of the predicted first ␣-helix ( Fig. 1) is currently our strongest candidate for a dimerization domain.
The effect of Vsp-OspC core features on the proteins' biological functions could be manifold. Various tick organs have proteases (54,55), thus protease resistance would be of benefit during the migration of the spirochetes from the mitgut through the hemolymph to the salivary glands (1). The resist-ance to trypsin and plasmin might also serve the spirochetes in their vertebrate hosts. After host plasminogen acquisition and activation, surface bound plasmin facilitates the dissemination of bacteria in the tick and fosters heart and brain invasion in infected mice (56,57).
While the lipid moiety of Borrelia lipoproteins such as B. burgdorferi OspA is a determinant of mitogenicity (58), the common structural features of Vsp-OspC proteins might modulate host immune responses as well. An efficient peptide antigen presentation to T cells via the major histocompatibility complex requires the limited proteolytic processing of exogenous proteins by endosomal and lysosomal proteases (59). Experiments with specific protease inhibitors and protease gene knockout mice or the elimination of protease sensitive sites have shown this (60). Borrelia spirochetes may have evolved the latter strategy, rendering their major antigens resistant to proteolytic attack and thus leading to the persistence of Vsp-OspC core proteins in infected mammals. Furthermore, the multimerization of the highly abundant Vsp-OspC proteins might produce an array that is analogous to viral capsid particles (61). Both antigen persistence and repetitive structure have been associated with a T-independent response (62,63). Intriguingly, this type of immune response has been observed in the clearance of B. turicatae infection (64) and was important for protective immunity and resolution of Lyme disease in an animal model (65). The current study has focused on the activity of a limited selection of proteolytic enzymes in vitro, and further immunological studies will be needed to comprehensively address the possible effects of Vsp protease resistance on antigen presentation.
The mechanisms of antigenic variation in African trypanosomes and relapsing fever Borrelia are strikingly similar (66). A comparison of the known structural features of these proteins reveals further convergences. VSGs also share common, highly ␣-helical folds and form stable dimer units in solution (67). Both VSGs and Vsp-Vlps are anchored to the cell's surface via post-translational modifications, the VSGs via a C-terminal glycosylphosphatidylinositol anchor (68), and the Vsp-Vlps, like other Borrelia lipoproteins (8,9), via an N-terminal triacylmodified cysteine. Interestingly, VSG glycosylphosphatidylinositol anchors are hydrolyzed by an endogenous trypanosomal phospholipase under stress conditions, which leads to VSG shedding (69). This may expedite the exchange of coat proteins during an antigenic switch, and at the same time divert the host immune system from live trypanosomes expressing the same VSG. A similar mechanism has been recently observed with the oral pathogen Streptococcus mutans, which expresses an enzyme that releases the adhesin P1 from its cell wall (70). This activity might allow the bacterium to shed P1-antibody complexes or to recolonize after detaching from an adherent surface.
The structure-function analysis of the Vsp and Vlp proteins will likely yield important clues in our understanding of antigenic variation and tissue tropism. By integrating the structural information gained from the VspA crystals obtained here with data from in vitro immunological, biochemical, and cell biological studies, we will be able to define the domains interacting with antibodies and other host factors.