|Home | About | Journals | Submit | Contact Us | Français|
The virulence of Bacillus anthracis is critically dependent on the cytotoxic components of the anthrax toxin, lethal factor (LF) and edema factor (EF). LF and EF gain entry into host cells through interactions with the protective antigen (PA), which binds to host cellular receptors such as CMG2. Antibodies that neutralize PA have been shown to confer protection in animal models and are undergoing intense clinical development. A murine monoclonal antibody, 14B7, has been reported to interact with domain 4 of PA (PAD4) and block its binding to CMG2. More recently, the 14B7 antibody was used as the platform for the selection of very high affinity single chain antibodies that have tremendous potential as a combination anthrax prophylactic and treatment. Here we report the high resolution X-ray structures of three high affinity single chain antibodies in the 14B7 family; 14B7 and two high affinity variants 1H and M18. In addition, we present the first neutralizing antibody-PA structure, M18 in complex with PAD4 at 3.8 Å resolution. These structures provide insights into the mechanism of neutralization and on the effect of various mutations on antibody affinity and enable a comparison between the binding of the M18 antibody and CMG2 with PAD4.
Anthrax remains a significant threat as a biological weapon due in large part to its ease of both large-scale manufacture and weaponization in the spore state. Following spore inhalation, anthrax is lethal in humans due to the combined actions of secreted toxins.1, 2 An effective countermeasure strategy requires an effective anti-toxin therapy 3–19 to be used in combination with antibiotics, or as a stand alone treatment of an antibiotic resistant strain of anthrax.20
We, and others, have been developing a combination prophylactic-post exposure therapeutic for anthrax based on an engineered antibody against the anthrax protective antigen (PA) toxin.7–25 Briefly, the PA toxin facilitates host cellular targeting and transport of the lethal factor (LF) and edema factor (EF) into the cytoplasm. LF is a protease that targets mitogen-activated protein kinase kinases (MAPKKs) and EF functions as an adenylate cyclase. The action of LF and EF in the cytoplasm of target cells triggers a series of biochemical events that lead to cell death.1, 2
The intoxication process is initiated when monomeric full-length protective antigen (PA83) is processed by host proteases to form the PA63 fragment, which binds as a heptamer with high affinity to the TEM8 and CMG2 cellular receptors on host cells such as macrophages. Post-exposure administration of high affinity antibodies that block the PA-receptor interaction has been shown to be effective in reducing mortality in animal models.21–25 Anti-PA antibodies can also serve as prophylactics to prevent infection from spore inhalation, although the mechanism of prophylaxis is not well understood.20, 26–29
The 14B7 murine monoclonal antibody (KD = 4.3 nM),11 originally developed at USAMRIID,12 was shown to delay time-to-death following exposure to anthrax spores in a guinea pig model.24 14B7 is known to recognize the receptor-binding region of PA and thereby block PA-host cell interactions.30 Originally, we used phage display to isolate an affinity enhanced version of the 14B7 variant called 1H, exhibiting a KD of 250 pM.13 A humanized version of this antibody is currently in advanced clinical development.20 The approximately 20-fold affinity enhancement of 1H compared to 14B7 is achieved with two mutations, Q55L and S56P, in CDR L2. In subsequent studies, an even higher affinity variant of 14B7 called M18 was isolated from a library of random mutants screened by bacterial display and flow cytometry.11 M18 has 10 mutations (light chain I21V, L46F, S56P, S76N, Q78L, and L94P; heavy chain S30N, T57S, K64E, and T68I) and exhibits a KD of 35 pM.
Crystallographic studies of antibody fragments in complex with a protein antigen have been ongoing for more than 25 years.31–40 Generally, antibodies to protein antigens target a discontinuous epitope on the antigen.32 It is also common for all 6 complementarity determining regions (CDRs) of the antibody to interact with the antigen32,40–42 and, on occasion, for framework residues to make contact as well.32 Shape complementarity along the interaction surface appears to be important,35,40,43 and a nonpolar “hotspot” is generally found to contribute the majority of the binding energy. A study of the affinity maturation of antibodies to lysozyme revealed that improved shape complementarity and burial of nonpolar surface at the expense of polar surface were generally correlated with increased affinity.35 In addition, structural studies with small molecule haptens have indicated that affinity maturation via somatic mutation might involve freezing out complementary conformations of CDR loops, involving mutations in residues that can be up to 15 Å away from the antigen.44
Here we report the crystal structure of M18 in complex with domain 4 of PA and the crystal structures of antibodies 14B7, 1H, and M18. The PA-M18 complex offers a detailed explanation for the neutralizing activity of the 14B7 family of antibodies, and also provides some insight into the affinity enhancements seen with engineered versions of 14B7. In the future, the PA-M18 complex structure will serve as a useful reference for targeting specific regions of 14B7 family antibodies in order to make even higher affinity variants, and for mutagenesis in the event a non-neutralizable variant of PA is encountered through either deliberate or natural variation.
The scFv of 14B7*, an N-terminal variant of parent monoclonal antibody 14B7 (Figure 1), crystallized in space group P41212 with cell constants a = b = 80.18, c = 67.83 Å and with one molecule per asymmetric unit, giving a Vm of 2.3 Å3/dalton. The structure was solved by molecular replacement using the κ chain of the blood group A Fv (PDB accession code 1JV5)45 and the heavy chain of Fab17-IA (PDB code 1FOR)46 as search models for the light and heavy chains, respectively. The final 1.3 Å structure consists of residues 1–107 of the light chain and 4–112 of the heavy chain. The 20 residue linker between the C-terminus of the light chain and the N-terminus of the heavy chain is not visible in electron density maps nor were residues 1–3 and the C-terminal residue (113) of the heavy chain. We assume the missing residues are all disordered. A representative section of electron density is shown in Figure 2a.
Crystals of the high affinity variant 1H, crystallized under the same conditions as 14B7* but at 4 °C, grew in the same crystal form as 14B7* but diffracted to only 2.8 Å. The structure was solved by molecular replacement using the 14B7* structure as the search model. The final refined structure consists of residues 1–108 of the light chain and 3–113 of the heavy chain and has an Rworking of 21.7% and an Rfree of 25.6%. In addition, there is clear electron density for one leader residue preceding the N-terminus of the light chain, and this residue is included in the final model. As with 14B7*, the 20 residue linker between the light and heavy chains was not visible in electron density maps. The refined model includes 16 solvent molecules.
Crystals of the ultrahigh affinity variant M18 scFv were grown under the same conditions as 1H and diffracted to 2.0 Å. Unlike 14B7* and 1H, the M18 scFv crystallized in space group P1 with cell constants a = 36.07, b = 54.21, c = 61.95 Å, α = 71.8°, β = 75.6°, and γ = 71.3° and with two molecules per asymmetric unit, giving a Vm of 2.2 Å3/dalton. The structure was solved by molecular replacement using the 14B7* structure as the search model. The final refined structure consists of residues of 1–108 of the light chain and 2–113 of the heavy chain and has an Rworking of 19.4% and an Rfree of 22.9%. In addition, there is density for two leader residues preceding the N-terminus of the light chain. As with 14B7* and 1H, the 20 residue linker between the light and heavy chains was not visible in electron density maps. The refined model includes 286 solvent molecules.
A Ramachandran plot for the 14B7* structure shows 98.6% of residues in the most favorable region and 0.9% in additional allowed space. A Ramachandran plot for the 1H structure shows 96.4% of residues in the most favorable region and 3.1% in additional allowed space. For the M18 scFv structure, there are 98.2% in the most favorable region and 1.8% in the additional allowed space. Heavy chain Gly 100, located on the CDR H3 loop, is found in an energetically disfavored conformation in the 14B7* and 1H structures.
In an effort to crystallize M18 in complex with PA, a 1.1:1 molar ratio mixture of M18 scFv and PA83 was used to screen for crystallization conditions. After eleven months of incubation, a crystal was observed in one condition. Molecular replacement analysis using PHASER47 revealed that the crystal contained M18 in complex with the PA domain 4 (PAD4), indicating that proteolysis of PA83 yielded sufficient PAD4 such that a crystal of the complex could form. The complex crystallized in space group P21 with cell constants a = 48.9, b = 299.7, c = 69.0 Å, and β = 94.5°. Four antigen-scFv complexes were found in the asymmetric unit, giving a Vm of 3.1 Å3/Dalton. The crystal diffracted to only 3.8 Å, but four-fold non-crystallographic restraints permitted the refinement of individual atomic positions and of group temperature factors for each residue. The final refined structure has an Rwork of 23.2% and an Rfree of 27.6%. Figure 2b shows electron density from a 2Fo-Fc map at the interface between PAD4 and both the H and L chains of M18. Crystallographic data for the three scFv structures and the PAD4-M18 complex are shown in Table 1.
As mentioned above, the asymmetric unit of the complex contains four PAD4-M18 heterodimers. In particular, each heterodimer is related to a second heterodimer by a non-crystallographic two-fold rotation with the dimer interface of this greater dimer involving principally β-strands of the light chain 62–67 and 70–74. The two greater dimer particles in the asymmetric unit are, in turn, related by a rotation of 178° about an axis that is around 91° from the greater dimerization axis. Contact between the greater dimer particles of the asymmetric unit involves a single PAD4 monomer in each.
The four M18 molecules within the asymmetric unit are virtually identical. As was observed in the structures of unbound 14B7*, 1H, and M18, electron density is observed for residues of the light chain and heavy chain but not for the 20 amino acids linking the C-terminus of the light chain to the N-terminus of the heavy chain in the scFv. In two of the M18 monomers, residues 1–113 of the heavy chain are observed while only heavy chain residues 2–113 can be seen in the other two monomers.
The four PAD4 molecules within the asymmetric unit are also virtually identical with the largest differences seen in the N-terminal loop, residues 592–605. This loop extends away from the rest of the molecule, and as a result, the crystal packing environment differs for each PAD4 monomer in the asymmetric unit producing subtle conformational differences. There is also some variation in the PAD4 sequence length observed. For three of the PAD4 monomers, density for residues 592–735 is observed, but for the fourth monomer, only residues 593–735 are seen. Based on the crystal structure of PA, domain 4 was defined as residues 596–735.48 Residues 593–595 form a slightly exposed loop at the boundary of domains 3 and 4, but some conformational change likely preceded proteolysis at residue 591.
The structures of the 14B7* scFv and the engineered variants 1H and M18 all show the expected typical immunoglobulin fold presenting a large binding surface formed by the CDRs. The folds of the three scFv antibodies are virtually identical. The rms distance between equivalent Cα positions of 14B7* and 1H is 0.24 Å, and similarly, the rmsd between Cα’s 14B7* and M18 is 0.54 Å. The only significant difference occurs in the CDR H3 loop of M18 with a Cα rmsd of 2.0 A.
However in the PAD4-M18 complex, the H3 loop assumes a conformation more like that of 14B7* and 1H. In the 14B7* and 1H scFv crystals, the H3 loop packing involves hydrophobic contacts between heavy chain L97 and L98 interacting between two adjacent molecules related by crystallographic symmetry. This differs from the packing in the P1 crystal of M18 scFv, where H:L97 makes hydrophobic contacts with heavy chain W33 and Y52, as well as light chain T93 and a leader Lys in an adjacent molecule. H:L98 is contacted by light chain P95 and heavy chain F47 in an adjacent molecule. Although the difference in H3 conformation appears to be due to crystal packing forces, it does demonstrate the flexibility of the H3 loop. A superposition of the Cα traces of the 14B7*, 1H, and M18 scFv’s, along with that of M18 from the antibody-antigen complex Cα traces is shown in Figure 3.
The structure of PA has been solved previously at high resolution, both as the apo-protein48 and in complex with the anthrax toxin receptor.49 The majority of contacts between PA and the CMG2 cellular receptor are made with PAD4, although some contacts with PA domain 2 are also seen. Residues 593–735 are observed for the four PAD4 monomers in the asymmetric unit of the PAD4-M18 crystal. A cartoon drawing of the PAD4-M18 complex is shown in Figure 4. The PAD4 domain complexed with antibody is virtually identical to that from the unbound antigen (Cα rmsd of 0.7 Å), indicating that antibody binding does not cause significant conformational change in the antigen.
The PAD4-M18 complex has a buried surface area of ~1700 Å2, with the light and heavy chains contributing virtually equally. As expected, the CDRs of both the light and heavy chains of M18 face PAD4 in the antibody-antigen complex. All of the CDRs except L3 make some direct contact with PAD4, but the strongest interactions are made with H3 and L2. The major epitope of PAD4 is at an end of one of its β sheets (strands 5-6-9-10). There are three main contact loops: (1) β5–β6 (residues 646–658), comprised of the loop between strands 5 and 6, strand 6, and the loop immediately following strand 6; (2) α3-β9 (residues 683–694), the loop between helix 3 and strand 9; and (3) pre-β10 (residues 716–719), the loop immediately preceding strand 10.
The M18 complex interface involves several hydrophobic contacts. L685 is a surface leucine, which upon binding with M18, is inserted into a pocket formed by CDRs H1 and H3 and the side chains of heavy chain residues W33 and L97. Additional hydrophobic interactions are formed between Y688 and L2 residue Y49 and H3 residues L97 and L98, as shown in Figure 5. Other hydrophobic interactions include L652 contacting light chain residue Y50, and P686 contacting heavy chain residue L97.
The complex also involves several polar interactions (Figure 5). The hydroxyl of Y688 forms a hydrogen bond with the OE1 atom of L2 residue Q55. Several ion pairs are formed between PAD4 loops and CDRs L1, L2, and H2. These include D658 with L1 residue R30, D648 with L2 residue R53, D683 with H2 residue R50, and K684 with H2 residue D56. Several more hydrogen bonds are formed between all three PAD4 loops and CDR loops L1, L2, H1, and H3.
Previously, alanine-scanning experiments of residues 657 and 679–693 of PAD4 implicated N682, K684, L685, P686, L687, and Y688 as being significant for interactions with the 14B7 antibody30 (Figure 5). By far, the largest influence on 14B7 binding was found for the L685A substitution, which showed a 10,000-fold decrease in binding affinity. It was also found that substitutions for N682, L687, and Y688 each reduced binding to 14B7 ~100-fold, while substitutions for K684 and P686 reduced binding affinity ~20-fold.
In light of the PAD4-M18 structure, it is easy to rationalize the results found with PAD4 substitutions like K684A, L685A, P686A, and Y688A, each of which shows clear interactions with M18 and a significant buried surface area. These results are supported by computational alanine-scanning mutagenesis of the PAD4-M18 complex,50,51 which uses a simple free energy function comprised of a Lennard-Jones potential, an implicit solvation model and other terms to calculate the effects of an alanine mutation on the free energy of a protein-protein complex. PA residues L652, K684, E654, L685 and Y688 are predicted to be primarily responsible for the binding energy of the complex, each contributing >1 kcal/mol free energy and the latter three residues each making similar contributions of ~3 kcal/mol. The reasons for drastic reduction in binding to 14B7 by the L687A or N682A PA variants are less obvious. L687 does not interact directly with the antibody, but it does form hydrophobic interactions with other PAD4 residues such as I646, I656, F678, and V696. N682 does not form any favorable interactions with PAD4 in the complex but is pointed toward the backbone of heavy chain residues 97 and 98. The predicted free energy contributions of PA residues L687 and N682 to complex formation are negligible (<0.05 kcal/mol).
Previous results of the effect of alanine substitutions in 14B7 52 are similarly easy to rationalize. Light chain residues Y50 and R53, as well as heavy chain residues W33, R50, Y52, and Y100 were found to be very important for binding, and these are all on the binding interface with PA (Figure 5). Light chain residues W35, N92, and W96 were also found to be important in the alanine-scan,52 but each of these are located within the interior of the light chain fold so they are likely key to establishing a stably folded structure.
Previously, the structure of the complex of PA with the I domain of its host cell receptor, CMG2, was solved at 2.5 Å resolution.49 The PA-CMG2 receptor complex revealed a total buried surface area of 1920 Å2, of which ~1360 Å2 is buried by the PAD4 interface. The remainder is buried by the interface of the receptor with PA domain 2. In contrast, there is no evidence that the 14B7 family of antibodies, including M18, contacts PA outside of domain 4. Consistent with the biochemical data, a superposition of the entire PA molecule (with domain 4 superimposed on that of our complex) indicates that M18 can contact only domain 4. That is, the complex represents the complete interaction of M18 with the PA protein.
Overall, the binding interfaces found in the PAD4-CMG2 and PAD4-M18 structures overlap well (Figure 6a). The CMG2 receptor interacts with the same three binding loops of PAD4 observed to make antibody contacts in the PAD4-M18 complex, and several of the PA residues are important in both cases (L682, P686, L687) based on alanine-scanning results.30 The binding interface between PAD4 and CMG2 centers on a key interaction between D683 of PAD4 and a Mg+2 atom bound in a so-called metal ion dependent adhesion site (MIDAS) motif of CMG2. In this way, the PA is mimicking the ligand recognition of integrins. Not surprisingly, when D683 was replaced by Ala, all PA toxicity was lost, presumably because the D683A variant cannot bind CMG2 or other cellular receptors such as TEM8.30 However, when D683 was converted to alanine, there was no measurable effect on the PAD4-14B7 interaction.30 In the PAD4-M18 structure, D683 is located in the middle of the protein-protein interface, and makes a salt bridge with H2 residue R50 of M18 (Figure 5b). It is unclear why the loss of this ionic interaction should have so little effect on PAD4 antibody interaction.
There are some other clear differences between the PAD4-CMG2 and PAD4-M18 structural interactions. For example, the PAD4-CMG2 binding interface shows the insertion of the CMG2 V115 side chain into a pronounced hydrophobic “socket” in PAD4 composed of I646, I656, and L687 (Figure 6b). In contrast, in the PAD4-M18 complex, this same PAD4 hydrophobic socket is not occupied by any antibody residue, hydrophobic or otherwise.
In an effort to create a therapeutic anti-toxin antibody for use as a prophylactic as well as late stage treatment for anthrax, 14B7 was used as the platform for antibody engineering. Initially, sequence randomization followed by phage display was used to isolate 1H, which exhibited a KD of 250 pM.13 This approximately 20-fold affinity enhancement compared to 14B7 means that 1H may have a higher affinity for PA than do the cellular receptors.13 Presumably for this reason, a humanized, whole IgG form of 1H (Anthim™) was shown to be an effective prophylactic and post-infection treatment for inhalation anthrax in animal models.20 Anthim™ successfully completed a phase I clinical human safety trial and is currently in later stages of development.
The two key mutations in 1H relative to 14B7 are light chain residues Q55L and S56P. Both changes increase hydrophobicity on the periphery of the binding interface with PA and, by analogy to the PAD4-M18 structure, will contact Y688 and S690 of PA. For example, the Cδ atom of L:P56 contacts Cα and Cβ of S690. It is therefore tempting to ascribe the 1H affinity enhancement to an increase in hydrophobic contacts.
The ultrahigh affinity variant M18 shares the L:S56P mutation with 1H and presumably makes similar interactions with the antigen. There was no loss in binding energy predicted by computational alanine-scanning analysis for an L:P56A substitution in the M18 complex, indicating that effects contributed by P56 may be indirect. For instance, it may stabilize the CDR L2 loop, as suggested by Sivasubramanian et al.;52 or it may subtly reorient nearby contact residues, such as L:R53, to optimize their energetic contributions.
Interestingly, substitution of L:Q55 with alanine independently increases 14B7 affinity for PA from 4.8 to 1.7 nM, while substitution with leucine further increases 14B7 affinity to 0.52 nM, as observed experimentally13 and predicted computationally. This hierarchy of substitutions is computationally predicted to stabilize the scFv binding partner in the order L > A > Q as well as contribute some binding free energy (~0.45 kcal/mol) in the M18 complex. However, M18 does not contain the L:Q55L mutation. In the PAD4-M18 structure, only the polar –OH group of Y688 makes contact with the L:55 and L:56 side chains. In fact, in the case of L:Q55 of M18, there is a predicted hydrogen bond that will be lost upon replacement with leucine in 1H. Comparison of the 14B7 backbone conformation to that of 1H (and M18) in this region reveals no significant change in folded structure (Figure 3). Because the PAD4-M18 structure does not have the L:Q55L mutation, we can only speculate that there is some contribution from a hydrophobic contact with Y688 by L:L55 in 1H. Finally, considering the constrained conformational dynamics of a proline moiety, it is worth considering that a major contribution of the L:S56P and possibly L:Q55L mutations could be to alter the dynamics of the L2 loop to be more favorable for binding.
Besides L:S56P, only the L:L46F mutation in M18 is located near the binding interface with PA (Figure 7). The role of L:L46 in affinity enhancement is not obvious, and the energetic contributions to binding are predicted to be negligible. It may form a slight hydrophobic contact with Y688 of PA. The L:F46 Cε2-Y688 Cε2 distance is 5.3 Å. F46 forms a stronger hydrophobic interaction with the side chain of H:L98 in the H3 loop; H:L98, in turn, is within 4 Å of the Y688 side chain so L:L46 may be acting to stabilize, indirectly, an important hydrophobic contact. In addition, the proximity of L:L46F to L:S56P might mean that they are working in concert to control subtle features of the conformation or dynamics of the L2 loop. Of course, we must not rule out the possibility that some of the other mutations in M18 contribute to an increase in affinity through subtle, indirect effects.
Interestingly, the contributions of L:L46F, L:Q55L, and L:S56P appear to be additive. For example, adding the L:Q55L mutation to M18, which already contains L:L46F and L:S56P, leads to a new variant called M18.1 that exhibits a dissociation constant of 21 pM, representing the highest affinity observed to date in the 14B7 family of engineered antibodies (Table 2). Unfortunately, this enhanced affinity comes at the expense of expression level as M18.1 expresses significantly more poorly than M18 in E. coli. Adding L:L46F to 14B7 enhances binding affinity by an order of magnitude, and removing L:L46F from M18 causes an analogous decrease in affinity.
The results of this study appear to be quite similar to those reported in the case of anti-fluorescein scFv’s binding to their hapten antigen.53 Complex structures reported for the parental and affinity-matured scFv’s showed no significant structural changes, even with the large improvement in affinity. Characterization of the scFv:fluorescein interactions by isothermal titration calorimetry revealed a sizable enthalpic contribution to the change in binding free energy. They concluded that the interaction of many subtle changes resulted in a large net effect on affinity.
In contrast, comparison of the structures of anti-hen egg-white lysozyme (HEL) Fabs and their resultant complexes with HEL revealed significant movement of CDRs H1 and H2 induced upon binding, which help improve shape complementarity of the antibody fragment to its protein antigen.35 They observed that the highest affinity Fab variants required the least amount of movement to transit between unbound and bound states, while the low affinity parental Fab had the largest distance between unbound and bound conformations; thus, the least amount of distortion from the unbound state upon complex formation appeared to lead to increased affinity. However, their observation that increasing affinity correlated with the increase in hydrophobic interactions adjacent to a central “hot spot” appears consistent with our observations in the PA:scFv system.
The PAD4-M18 structure described here represents the first structural characterization of a neutralizing antibody interacting with PA. Overall, the structure can be used to nicely rationalize the extensive biochemical data available. The one exception is the PA D683 position that would appear to make an important contribution to M18 binding, even though this is not reflected in the previously reported D683A variant of PA binding to 14B7.30 In addition, the results of affinity enhancements obtained by either phage panning (1H) or APEx (M18) can be rationalized as changes in key interactions in this same region of the interface, contacting Y688 and/or S690 of PA. However, a detailed explanation for the energetic contributions from other mutations in M18 that contribute to affinity enhancement, for example L:L46F, is not straightforward although it does appear to be operating in an additive fashion.
The PAD4-M18 structure can serve as the basis for engineering a new generation of ultrahigh affinity anti-PA antibodies. For example, using targeted mutagenesis, it may be possible to fill the hydrophobic “socket” in PAD4 composed of I646, I656, and L687 and enhance the already very strong PA-antibody interaction. In addition, if a variant of PA is derived either from natural or deliberate variation, the PAD4-M18 structure can serve as an invaluable reference for designing targeted libraries that should be able to yield a high affinity therapeutic derivative of 14B7 in a timely fashion.
M18, 1H, and 14B7* scFv’s were cloned into the periplasmic expression vector pAK400. Protein was prepared as described previously.54 See Figure 1 for their sequences, as well as that of the parent monoclonal antibody 14B7. Briefly, a pelB leader sequence was utilized to express protein into the periplasm of E. coli strain Jude-1, and osmotic shock followed by immobilized metal affinity chromatography was used for initial purification. Size exclusion chromatography (AKTA FPLC and Superdex 200 column, GE Healthcare) allowed isolation of monomeric scFv. Monomeric scFv was then digested overnight at 4° C with Carboxypeptidase A (Calbiochem, San Diego, CA, 1:100 mass ratio, protease:scFv). A second round of FPLC was followed by concentration to 10 mg/ml (Amicon Ultra 15, 10000 MWCO) into 1xHBS-EP (10 mM Hepes, pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% Polysorbate-20). Full-length anthrax protective antigen (PA83) was purchased from List Biological Labs (Campbell, CA), resuspended at 1 mg/ml, and subjected to size exclusion FPLC. Protein was concentrated to 10 mg/ml into 1xHBS-EP.
Mutant scFv’s 14B7-L46F and M18-F46L were constructed using the QuikChange protocol (Stratagene, La Jolla, CA). After protein expression, kinetic studies were carried out using a Biacore 3000 (Biacore, Inc., Piscataway, NJ). Briefly, amine coupling chemistry was used to immobilize 750 RU of either PA83 or BSA (control) to a CM5 chip. Samples were injected for 60 seconds at 100 μl/min in 1 xHBS-EP and allowed to dissociate for 15 minutes at the same flow rate. Regeneration of the flow cells was accomplished using a 30 second injection of 4 M MgCl2 at 100 μl/min.
The 14B7* ScFv protein was crystallized at room temperature in sitting drops from mixtures containing a one-to-one ratio of antibody solution (15 mg/mL) and reservoir solution [20% (w/v) PEG 4000, 0.1 M HEPES, pH 7.5, 10% iso-propanol]. The higher affinity variants 1H and M18 were crystallized under the same conditions at 4 °C. Prior to data collection, crystals were transferred to a cryoprotectant solution (the reservoir solution containing 25% glycerol) for 1–5 seconds. Each crystal, mounted in a cryoloop (Hampton Research, Laguna Niguel, CA), was frozen by dipping in liquid nitrogen and placed in the cold stream on the goniostat.
A mixture of PA83 and M18 with a 1:1.1 molar ratio was used to screen for crystallization conditions by the sitting drop method. Eleven months after the crystallization conditions were set up, a crystal was found in a condition at 4 °C with the reservoir solution 10% PEG 20,000, 0.05 M Tris-HCl, pH 7.5. Prior to data collection, crystals were transferred to a cryoprotectant solution (6% PEG 20,000, 0.05 M Tris-HCl, pH 7.5, 25% glycerol) for 1–5 seconds. The complex crystal, mounted in a cryoloop, was frozen by dipping in liquid nitrogen and placed in the cold stream on the goniostat.
X-ray diffraction data from the 14B7* crystal were collected at 100 K on beamline 8.2.1 of the Advanced Light Source (Lawrence Berkeley National Laboratory). Diffraction data of the 1H, M18, and complex crystals were collected at 100 K on an RAXIS IV++ image plate detector (Rigaku, The Woodlands, TX) with X-rays generated by a Rigaku RU-H3R rotating anode generator operated at 50 mV, 100 mA. Diffraction images were processed and data reduced using HKL2000.55
The structure of the 14B7* scFv was solved by molecular replacement using the light chain of blood group A Fv (PDB accession code 1JV5)45 and the heavy chain of Fab17-IA (accession code 1FOR)46 as search models; the structures were aligned with the Evolutionary Protein Molecular Replacement (EPMR) program.56 An initial chain tracing was generated from the molecular replacement solution using ARP/warp.57 The structures of the 1H and M18 scFv’s were solved by molecular replacement using the 14B7* structure as the search model with the program MOLREP.58
Crystal cell parameters of the complex crystal suggested that the asymmetric unit might contain two PA83-M18 complexes. Molecular replacement searches using several programs and with various search models, including PA83 (PDB accession code 1ACC),48 PA63, 14B7*, and individual PA domains, failed to find a solution. Finally, a search using the program PHASER47 with the single chain antibody 14B7* as a model gave a result that had a rotation function Z-score of 5.7 and a translation function Z-score of 5.9; the program authors have stated that scores of this magnitude may indicate a correct solution. Subsequent searches with 14B7* found solutions for three additional molecules, with increasingly improved Z-scores. Following that, solutions for the positions of four copies of PA domain 4 (PAD4) were determined. Examination of the packing of the four M18 scFv and four PAD4 molecules in the asymmetric unit showed that there were four essentially identical M18 scFv-PAD4 heterodimers; this initial model had an R-factor of 32.6% suggesting that this was the correct molecular replacement solution.
Model building was carried out using O.59 Refinement of models was performed with the Crystallography and NMR System (CNS) (Version 1.1).60 The structures of the scFv’s were refined using the slow-cooling protocol. For the 3.8 Å resolution complex structure, atomic positions were restrained by non-crystallographic symmetry and refined using conjugate gradient minimization. Two group temperature factors for each residue in the complex were also refined. There were several rounds of refinement followed by manual rebuilding of the model. To facilitate manual rebuilding, a difference map and a 2Fo-Fc map, σA-weighted to eliminate bias from the model,61 were prepared. 5% of the diffraction data were set aside throughout refinement for cross-validation.62 PROCHECK63 and MolProbity64 were used to determine areas of poor geometry and to make Ramachandran plots. Computations and model building were carried out on Silicon Graphics Indy (Mountain View, CA), Gateway SB Select (Poway, CA), and HP Pavilion a1610n (Hewlett-Packard Co., Palo Alto, CA) computers.
Superpositions of protein molecules were done with O. Model pictures were made using MOLSCRIPT,65 BOBSCRIPT,66 Raster3D,67 and PYMOL (DeLano Scientific, San Carlos, CA). Computational alanine-scanning mutagenesis of the PAD4-M18 complex was done using ROBETTA.50, 51
Coordinates of the refined model of the PAD4-M18 complex have been deposited in the Protein Data Bank with entry code 3ETB. Coordinates of the refined models of the 14B7*, 1H, and M18 molecules have been deposited in the Protein Data Bank with entry codes 3ESU, 3ET9, and 3ESV, respectively.
This work was supported by NIH grants GM 63593, AI 75509, by the Robert A. Welch Foundation, and by the College of Natural Sciences support to the Center for Structural Biology.
This research was performed under an appointment to the Department of Homeland Security (DHS) Scholarship and Fellowship Program, administered by the Oak Ridge Institute for Science and Education (ORISE) through an interagency agreement between the U.S. Department of Energy (DOE) and DHS. ORISE is managed by Oak Ridge Associated Universities (ORAU) under DOE contract number DE-AC05-06OR23100. All opinions expressed in this paper are the author’s and do not necessarily reflect the policies and views of DHS, DOE, or ORAU/ORISE.
The Advanced Light Source is supported by the Director, Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.