Binding stoichiometry and kinetics of the interaction of a human anthrax toxin receptor, CMG2, with protective antigen

The protective antigen (PA) moiety of anthrax toxin binds to cellular receptors and mediates entry of the two enzymatic moieties of the toxin into the cytosol. Two PA receptors, ATR/TEM8 and CMG2 (capillary morphogenesis protein 2), have been identified. We expressed and purified the von Willebrand A (VWA) domain of CMG2 and examined its interactions with monomeric and heptameric forms of PA. Monomeric PA bound a stoichiometric equivalent of CMG2, while the heptameric prepore form bound seven equivalents. The K d of the VWA-domain:PA interaction is 170 pM when liganded by Mg 2+ , reflecting a 1000-fold tighter interaction than most VWA-domains with their endogenous ligands. The dissociation rate constant is extremely slow, indicating a 30 hour lifetime for the CMG2:PA monomer complex. CMG2’s metal ion dependent adhesion site (MIDAS) was studied kinetically and thermodynamically. The association rate constant (~10 5 M -1 s -1 ) is virtually identical in the presence or absence of Mg 2+ or Ca 2+ , but the dissociation rate of metal ion liganded complex is up to four orders of magnitude slower than metal ion free complex. Residual affinity ( K d ~960 nM) in the absence of divalent metal ions allowed the free energy for the contribution of the metal ion to be calculated as 5 kcal mol -1 , demonstrating that the MIDAS metal ion is directly coordinated by CMG2 and PA in the binding interface. The high affinity of the VWA-domain for PA supports its potency in neutralizing anthrax toxin, demonstrating its potential utility as a novel therapeutic for anthrax.


Introduction
The pathology of the anthrax bacillus, B. anthracis, is due, in part, to the production of anthrax toxin, an ensemble of three nontoxic monomeric proteins, which combine at the surface of host cells to form toxic noncovalent complexes (Fig. 1A). Two of these proteins are enzymes that modify cytosolic substrates. Lethal Factor (LF; 90 kD) is a Zn 2+ protease that cleaves several mitogen-activated protein kinase kinases (1,2), and Edema Factor (EF; 89 kD) is a Ca 2+ -and calmodulin-dependent adenylate cyclase (3). The third protein, Protective Antigen (PA 83 ; 83 kD), binds to cellular receptors and transports LF and EF to the cytosol.
The initial step in the action of the toxin is the binding of PA to a cell surface receptor. Receptor-bound PA is cleaved into two fragments by a furin-family protease (4). Dissociation of the smaller fragment allows the larger fragment, which remains receptor-bound, to self-associate into ring shaped heptamers [(PA 63 ) 7 , also referred to as prepore (5).] Prepore may then bind up to 3 molecules of LF and/or EF with nanomolar affinity (6,7). The resulting complexes are endocytosed to an acidic compartment (8,9), where the heptamers are converted from the prepore state to an integral membrane, ionconductive pore (10). The process of translocating LF and EF into the cytosol is linked to the formation of pore, but the nature of this linkage is poorly understood. Within the cytosol these enzymatically active moieties may then disrupt normal cellular physiology.
Two anthrax toxin receptors, CMG2 (11) and ATR/TEM8 (anthrax toxin receptor/tumor endothelial marker 8) (12) are known. Each is a single peptide chain, consisting of an extracellular domain, a membrane spanning region, and a cytoplasmic tail. In their extracellular domains, there is an ~200 amino acid von Willebrand-type A (VWA) domain, which shows 60% amino acid identity between the two proteins (11). This domain adopts a dinucleotide binding, or Rossmann, fold, which is composed of a sandwich of six to eight amphipathic α-helices that surround a hydrophobic β-sheet (Lacy et al., submitted; Fig. 1B). The VWA-domain fold is found in many cell adhesion proteins and generally promotes protein-protein interactions (13). Many VWA-domains contain a highly conserved metal ion-dependent adhesion site (MIDAS) that is often involved in ligand interactions (14). The metal ion adopts an octahedral geometry and is coordinated by residues from three of its loops as well as two to three ordered water molecules. Usually, a glutamic or aspartic acid side chain from the ligand completes this metal ion's coordination sphere; therefore, the metal ion acts as a bridge between the ligand and VWA-domain. Consistent with a metal ion mediated interaction, both CMG2 and ATR have been shown to bind PA more tightly in the presence of divalent cations (11,12).
Here, we quantify the binding interaction of soluble CMG2 VWA-domain with PA 83 monomer and PA 63 heptamer. Monomeric PA bound a stoichiometric equivalent of CMG2, while the heptameric prepore form bound seven equivalents. The equilibrium dissociation constant for CMG2 VWA-domain's interaction with monomeric PA 83 is very tight (170 pM), and the dissociation rate constant is extremely slow (~10 -5 s -1 ). We show that the tight binding affinity relies on the presence and identity of the divalent MIDAS metal ion. Knowledge of the affinity and slow dissociation rate of CMG2:PA complexes supports the notion that CMG2 VWA-domain may be used clinically as an inhibitor of anthrax toxin.

Materials and Methods
Plasmid construction. A DNA sequence encoding residues 35 to 225 of CMG2's VWA-domain (referred to as CMG2 35-225 ) was cloned into pGEX4T-1 (Amersham Pharmacia), using 5' BamH I and 3' Not I restriction sites. pGEX4T-1 includes a thrombin cleavable glutathione-S-transferase tag onto the N-terminus of the expressed protein. Two truncated versions of CMG2 were then generated using PCR and the same 5' BamH I and 3' Not I sites: (i) residues 38 to 218 (CMG2 S38 ) and (ii) residues 40 to 217 (CMG2 R40 ). The latter version eliminated the natural disulfide bond. To generate a version of CMG2 with a single, unique cysteine point mutation on the amino-terminus, two successive rounds of site directed mutagenesis were performed on CMG2 R40 . The C175A mutation was introduced to eliminate the remaining buried cysteine, and the R40C mutation created a unique cysteine residue on the more accessible amino-terminus, making CMG2 C40 .
Preparation of Proteins. Recombinant lethal factor amino-terminal domain (residues 1 to 263; LF N ) was purified as previously described (15). Recombinant PA was expressed in BL21(DE3) using pET22b-PA (Novagen), which directs expression to the periplasm. Growth and expression of PA was carried out in a 10 L Bioflo 110 fermenter (New Brunswick Scientific). Using ECPM1 growth media (16), cells were grown at 37 o C to ~5 OD 600nm , sparged by a 60% air/O 2 mixture, and induced at 30 o C using 1 mM IPTG. PA was extracted from the periplasm and further purified using Q sepharose anion-exchange chromatography (Pharmacia), in 20 mM Tris, pH 8.0 (Buffer A) and Buffer A + 1 M NaCl (Buffer B). The protein was further purified using an S-200 superdex gel filtration column (Pharmacia) in Buffer A. For fluorescence studies, PA 83 6 with either the K563C or E733C point mutation was purified as previously described (15) and stored at -80 o C in 10 mg/ml aliquots in 10 mM dithiothreitol (DTT).
CMG2 VWA-domain variants were grown in a similar way except induction was carried out at 37 o C. Harvested bacteria were lysed by French press and sonication. GST-CMG2 was loaded onto a ~50 mL Glutathione Sepharose 4B column (Pharmacia), the column was washed in Buffer A, and bound fusion was cleaved from the immobilized GST using bovine α-thrombin (Sigma) overnight at room temperature. Thrombin was removed from the eluate with benzamidine sepharose, and CMG2 was further purified using an S-200 superdex column (Pharmacia). The purified stock protein solution was concentrated to 10 mg/ml, 10 mM DTT was added to protect the free thiol, and the solution was frozen at -80 o C.
Fluorescence labeling and modification of PA and CMG2. Buffers A and B were purged by sparging with N 2 for 5 minutes, and 0.5 mM TCEP was added to each buffer.
An S-200 gel filtration column was pre-equilibrated with 15% Buffer B. Approximately 20 mg of either PA 83 K563C, PA 83 E733C or CMG2 C40 was purified using gel filtration to remove excess DTT, which would react with the maleimide activated fluorophores.
One mg of either maleimide thiol reactive fluorophore, Alexa Fluor 488 (AF488) or Alexa Fluor 546 (AF546), was added to each protein elutate in approximately a 10 ml volume (2 mg/ml final protein concentration) for 2 hours at room temperature. Two separate preparations of PA 83 K563C were labeled with AF488 donor and AF546 acceptor; PA 83 E733C was labeled with the donor AF488, making PA 83 E733C-AF488; and CMG2 C40 was labeled with the acceptor AF546, making CMG2 C40 -AF546.
Following the completion of the reaction, unreacted dye reagent was blocked with 10 by guest on July 8, 2020 http://www.jbc.org/ Downloaded from mM 2-mercaptoethanol; and labeled protein was subsequently purified from free dye by a second round of S-200 gel filtration, using 15% Buffer B in the absence of TCEP. Protein solutions were assessed for labeling efficiency using pre-determined extinction coefficients for the fluorophores at their respective absorbance maxima and 280 nm.
Labeling efficiency was 95% or greater. Uniformly labeled PA heptamer, (PA 63 ) 7 E733C-AF488, was generated as previously described (17). Using an identical procedure,  Each 100 µL aliquot was separately incubated for about 1-2 hours at room temperature to ensure complete binding.
Similarly, (PA 63 ) 7 E733C-AF488 was diluted to 16 nM; however, 0.5 µM of LF N was added to the reaction mix to minimize non-specific interactions with the PA:LF binding face. The molar ratio was varied as described for PA 83 except the range spanned 0.3 to 25.
Each aliquot was then analyzed in an ISS fluorimeter interfaced to an Ar + laser.
The 488 nm line was used for excitation of the donor fluorophore (AF488). Donor and acceptor emission were acquired at 520 and 570 (±10) nm, respectively. The apparent FRET signal was defined by the ratio of the acceptor to donor fluorescence emission.
Aliquots were diluted to 2 mL in universal buffer. Fluorescence counts were recorded for 10 seconds, and all aliquots in a given titration were measured in triplicate. Kinetic and equilibrium data analysis. Binding association and dissociation rate constants, k a and k d , respectively, were obtained by fitting to a mono-exponential model (Eq. 1a), whereas PA heptamerization kinetics were fit to a second order model (Eq. 1b),

Multi-angle
where the amplitude at time zero, A, decays with respect to time, t. The offset, C, is the final value reached when the system achieves equilibrium. For binding association kinetics, pseudo-first-order conditions are maintained, i.e. CMG2 concentrations are >5fold above the concentration of PA 83 . Observed kinetic rate constants (k obs ) were separately fit or averaged to obtain association and dissociation rate constants.
Equilibrium dissociation constants are calculated from kinetic measurements of the association and dissociation rate constants according to After an additional size-exclusion chromatography step, the CMG2 VWA-domain was obtained in high purity as judged by SDS-PAGE (Fig. 1C).  5C). Reduction of the disulfide bond in CMG2 S38 similarly reduced the association rate by a factor of two with respect to the oxidized form ( Table 1).

Expression and purification of
Because of the inherent difficulties in measuring extremely slow dissociation rates (lifetimes >10 4 s) using SPR, we developed a FRET-based binding system. Association rates were initially estimated using stopped-flow to confirm the fidelity of the novel VWA-domain binding assay. Blue laser light was used to excite the donor fluorophore (AF488) on PA, which transferred energy to the acceptor fluorophore (AF546) on CMG2 C40 , allowing binding kinetics to be observed (Fig. 5B). Association rate constants (k a ) measured for this FRET based system (1.1×10 5 ± 5×10 3 M -1 s -1 ) were similar to those measured by SPR using CMG2 R40 in the presence of Ca 2+ (Fig. 5C).
Taking advantage of this FRET system, we also measured the association rates of Mg 2+ -bound CMG2. Here, protein solutions were pre-incubated with 2 mM MgCl 2 and 1 mM EGTA, reducing the effective concentration of even 1 µM contaminating Ca 2+ to less than 1 pM. Surprisingly, the association rate constant in the presence of Mg 2+ , as compared with Ca 2+ , is two times slower, at 5.3×10 4 ±9×10 2 M -1 s -1 . Nonetheless, the dissociation rate was 10-fold slower (9.2×10 -6 ±1×10 -7 s -1 ; Fig. 5D Correspondingly, FRET studies on the heptamerization of PA 63 (Fig. 4)  The coordination of the MIDAS metal often involves a glutamic or aspartic acid residue from the ligand (20). Consistent with this model, several studies have indicated that mutations at Asp683 in domain 4, the receptor binding domain of PA, compromise receptor binding (21)(22)(23), implicating this residue for that role. Also, if CMG2's MIDAS metal ion is geometrically modeled to bind PA via Asp683, then the predicted distance between E733 on PA and the amino-terminus of CMG2 is ~60 Å-consistent with the distance calculated from the observed FRET efficiency (Fig. 3A). (CD11b) with ~1 nM affinity due to a slow dissociation rate, 10 -5 s -1 (28,29). The divalent metal ion dependency for NIF:CD11b was less significant, i.e. ~2-to 5-fold reduced in the absence of metal ions (28,29), whereas we observed a 6000-fold reduction in affinity for PA:CMG2. NIF and PA may bind their receptor VWA-domain in a metal ion independent manner, but to very different degrees, consistent with the model that a second MIDAS-independent binding surface is utilized by CMG2. The tight binding strategy adopted by NIF clearly aims to block and inhibit adhesion-dependent responses of the neutrophil by competitively binding host integrin VWA-domains (28).