Insertion of Anthrax Protective Antigen into Liposomal Membranes

Protective antigen (PA), the receptor-binding component of anthrax toxin, heptamerizes and inserts into the endosomal membrane at acidic pH, forming a pore that mediates translocation of the enzymic components of the toxin to the cytosol. When the heptameric pre-insertion form of PA (the prepore) is acidified in solution, it rapidly loses the ability to insert into membranes. To maximize insertion into model membranes, we examined two ways to bind the protein to large unilamellar vesicles (LUV). One involved attaching a His tag to the von Willebrand factor A domain of one of the PA receptors, ANTXR2, and using this protein as a bridge to bind PA to LUV containing a nickel-chelating lipid. The other involved using a His tag fused to the C terminus of PA to bind the protein directly to LUV containing the same lipid. Both ways enhanced pore formation at pH 5.0 strongly and about equally, as measured by the release of K+. Controls showed that pore formation in this system faithfully reproduced that in vivo. We also showed that binding unmodified ANTXR2 von Willebrand factor A to the prepore in solution enhanced its pore forming activity by slowing its inactivation at acidic pH. These findings indicate that an important role of PA receptors is to promote partitioning of PA into the bilayer by maintaining the prepore close to the target membrane and presumably in the optimal orientation as it undergoes the acidic pH-dependent conformational transition to the pore.

Many toxins produced by pathogenic bacteria act by enzymically modifying intracellular target molecules. Most of these toxins are bipartite entities, so-called A-B toxins in which one part, the B (binding) moiety, binds to a cell surface receptor and facilitates delivery of the other part, the A (catalytic) moiety, to the cytosol (1). In some of these toxins the B moiety, besides binding to a receptor, also has the ability to undergo membrane insertion and form a pore (channel). The structures and functions of such pores are of great interest in understanding how toxin A moieties translocate across membranes.
Anthrax toxin is a tripartite A-B toxin system, composed of two catalytic moieties, edema factor (EF) 2 and lethal factor (LF), and a single receptor-binding/pore-forming moiety, protective antigen (PA). PA (83 kDa) binds to cell-surface receptors and is cleaved by furin or a furin-like protease to an active, 63-kDa form (PA 63 ) (2). PA 63 oligomerizes into a heptameric, receptorbound prepore, which contains high-affinity binding sites for EF and/or LF (3). The entire toxin-receptor complex is internalized by receptor-mediated endocytosis, and within the endosome the prepore undergoes an acidic pH-dependent conformational rearrangement to form a cation-selective, transmembrane pore (4). The PA pore mediates translocation of EF and LF across the endosomal membrane into the cytosol, where EF, an 89-kDa calmodulin-dependent adenylate cyclase, elevates levels of cAMP (5), and LF, a 90-kDa zinc protease, inactivates mitogen-activated proteins kinase kinases (6).
Two cellular receptors for PA have been identified: anthrax toxin receptor/tumor endothelial marker 8 (called ANTXR1) and capillary morphogenesis protein 2 (called ANTXR2) (7,8). Both are type-I transmembrane proteins, containing a signal sequence, an extracellular von Willebrand factor A (VWA) domain, and a single-pass transmembrane region. The VWA domains of ANTXR1 and ANTXR2 contain a conserved metal ion-dependent adhesion site motif. The ANTXR2 VWA domain (abbreviated as R2 below) binds PA with high affinity (K d Ϸ 170 pM when liganded by Mg 2ϩ ) (9). X-ray structural studies of complexes consisting of R2 bound to PA or to the heptameric prepore have revealed that R2 binds not only to PA domain 4 (receptor binding domain) but also to domain 2 (10,11). Domain 2 contains a mobile loop, the 2␤2-2␤3 loop (residues 303-322), and the seven 2␤2-2␤3 loops of PA heptamer are believed to move to the base of the structure during the low pH-induced conformational rearrangement to form a 14-strand transmembrane ␤-barrel (12). Binding of R2 to the PA prepore shifts the threshold of prepore-to-pore conversion to a more acidic pH range (10). Whole-cell voltage-clamp measurements of PA pore current have shown that ANTXR2 mediates PA * This work was supported by National Institutes of Health Grants RO1 AI 22021, AI 48489, and PO1 AI 56013. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Fig. 1 pore formation on the plasma membrane of cultured cells when they are exposed to acidic conditions (13).
Liposomes have been widely used as model systems to study protein-membrane interactions. In an early study, the PA prepore was found to permeabilize unilamellar asolectin vesicles (lacking receptors) to K ϩ under acidic conditions (14). Later studies 3 suggested that PA did not partition efficiently into liposomal membranes. In an attempt to make liposomes a more useful model system for studying this toxin, we have now devised ways to bind PA to the surface of large unilamellar vesicles (LUV). Synthetic Ni 2ϩ -chelating lipids provide a convenient way to attach His-tagged proteins to liposomal membranes (15)(16)(17), and we generated a His-tagged form of R2 (His-R2) to enable it to bind to LUV containing Ni 2ϩ -chelating lipids (DOGs-NTA-Ni) and serve as a bridge between PA and the membrane. We also prepared PA containing a His tag at its C terminus, enabling it to bind directly to the surface of LUV containing the Ni 2ϩ -chelating lipid. The effects of binding PA to LUV on toxin assembly and pore formation were examined and are reported below.

MATERIALS AND METHODS
Plasmid Construction-ANTXR2 cDNA encoding residues Ser 38 -Cys 218 was inserted into pET15b at NdeI and BamHI sites, which resulted in a construct pET15b-ANTXR2(Ser 38 -Cys 218 ) that expressed a N-terminally His 6 -tagged ANTXR2 VWA domain, His-R2. ANTXR2 cDNA encoding residues Ser 38 -Cys 218 was amplified by PCR (Forward primer: ggt ggt gaa ttc tcc tgc aga cga gcc ttt g; Reverse primer: ggt ggt ctc gag tca atg gtg atg gtg atg atg aca tga ctg agc tag tat aga att) to add a His 6 sequence after Cys 218 . The amplified DNA was inserted into pGEX-4T at EcoRI and XhoI sites to generate a construct, pGEX4T-ANTXR2(Ser 38 -Cys 218 )-His 6 , that expresses a GST-R2-His fusion protein. The construct of pET22b-PA-His was generated by introducing a His 6 DNA sequence to the PA C terminus, using QuikChange mutagenesis (Forward primer: ggc tat gag ata gga ctc gag cac cac cac cac cac cac tga; Reverse primer: tca gtg gtg gtg gtg gtg gtg ctc gag tcc tat ctc ata gcc). All the constructs were confirmed by DNA sequencing.
Protein Purification-The expression and purification of recombinant PA and R2 were performed as described (9). LF N -DTA, a fusion of the PA-binding domain of LF (LF N ) with the catalytic domain of diphtheria toxin (DTA) was expressed and purified as described (18). His-R2 was expressed in BL21 (DE3) and purified by metal (Ni 2ϩ ) affinity chromatography followed by dialysis in buffer (20 mM Tris-Cl, 150 mM NaCl, pH 8.5) to remove imidazole. GST-R2-His was expressed in BL21 (DE3) and purified by metal (Ni 2ϩ ) affinity chromatography. The GST tag was cleaved by thrombin, and free GST and thrombin were removed by passing through a glutathione-Sepharose column and a benzamidine column. PA-His was expressed in BL21 (DE3) and purified by metal affinity chromatography followed by a size-exclusion chromatography. Protein concentrations were determined from the A 280 using calculated extinction coefficients. Nicked PA (nPA) and PA heptamer prepore were prepared as described (19).
Liposome Preparation-1,2-Dioleoyl-sn-glycero-3-phosphocholine (DOPC) and 1,2-dioleoyl-sn-glycero-3-{[N(5-amino-1carboxypentyl) iminodiacetic acid]succinyl} (nickel salt) (DOGs-NTA-Ni) were purchased from Avanti Polar Lipids, Inc. DOPC (20 mg/ml) and DOGs-NTA-Ni (10 mg/ml) in chloroform were mixed at the indicated molar ratio, the lipid mixture was dried under a stream of nitrogen gas to form a lipid film, and the lipid film was placed under high vacuum for 3 h to remove all residual solvent. The dried lipid film was rehydrated with proper buffers to form multilamellar vesicles, which were subjected to three freeze-thaw cycles and extrusion through a 200-nm pore size polycarbonate filter (Nucleopore Inc.) using a mini extruder (Avanti Polar Lipids). Size-exclusion chromatography and cryo-electron microscopy were used to check the quality of the prepared LUV. The majority of liposomes were unilamellar, with diameter ϳ150 -200 nm. Variation in the mol % of DOGs-NTA-Ni did not change the diameter of liposomes.
K ϩ Release Assay-The K ϩ release assay was performed as described (14), with modifications. A K ϩ buffer (10 mM HEPES, 100 mM KCl, pH 7.4) was used to rehydrate the dry lipid film followed by three freeze-thaw cycles and extrusion. Right before K ϩ release assay, the K ϩ buffer-rehydrated liposomes were subjected to buffer exchange by passing them through a G-50 column equilibrated with a Na ϩ -containing buffer (10 mM HEPES, 100 mM NaCl, pH 7.4). The resulting liposomes, with K ϩ inside and Na ϩ outside, were mixed with indicated proteins at pH 8.5 (the pH of the liposome-protein mixture was adjusted to pH 8.5 by adding 1/20 volume of 1 M Tris-HCl, pH 8.5). After 30 min at room temperature, the protein-liposome mixture (total 200 l) was added to 5 ml of 10 mM buffers (NaAc, pH 4.5-5.0; Bistris, pH 5.5-6.5; HEPES, pH 7.0 -7.5; Tris-Cl, pH 8 -9) containing 100 mM NaCl. The mixtures were stirred continuously using a magnetic stirrer, and release of K ϩ from liposomes was monitored using a K ϩ -selective electrode (Orion Research) connected to a microcomputer pH/mV/ Temp meter 6171. The signal was transmitted by a data transmitter (DATAQ) and displayed with DATAQ software. The channel-forming peptide, gramicidin A (5 g/ml), was added to determine the total or remaining K ϩ in the liposomes. Note: Because of intrinsic variables and limitations in preparation, different pools of liposomes differed in terms of the liposome concentration, the amounts of residual K ϩ outside liposomes, and entrapped K ϩ inside. Therefore, a control, buffer alone, was always included in the experiments, enabling the results obtained from the same pool of liposome preparation to be compared.
His-tagged Protein-Liposome Binding Assay-LUV (200 nm) prepared in present study can be easily collected by centrifugation at 100,000 ϫ g for 30 min (17,20). Purified His-tagged proteins were incubated with liposomes at indicated conditions. Membranes were pelleted at 100,000 ϫ g for at least 30 min in an airfuge, and the pellet was analyzed by SDS-PAGE.
Time-lapse Intensity Measurements of NBD Emission-PA mutants G305C and N306C were labeled with NBD (N,NЈ-dimethyl-N(iodoacetyl)-NЈ-(7-nitrobenz-2-oxa-1,3diazol)ethylenediamine) as described (21). The liposomes (DOPC:DOGs-NTA-Ni ϭ 100:8) were incubated with indicated proteins at pH 8.5 for 30 min at room temperature, and the incubated protein-liposome mixture (total volume, 1.8 ml) was transferred to a cuvette with a stirring bar in an ISS K2 fluorometer. The NBD was excited at 488 nm, and emission was recorded at 544 nm. Crossed polarizers on excitation, emission beams, and a 520-nm filter were used to reduce the background because of scatter. After addition of 0.2 ml of 1 M pH 8.5 buffer or pH 5.0 buffer to the cuvette, a sharp decrease of the background scatter signal was observed because of dilution. This was followed by a specific signal resulting from shifting NBD from a polar (solution) to a non-polar environment (liposomal membrane).

Acidic Conditions Cause Rapid Inactivation of the PA Prepore-
It is well documented that purified PA prepore and nPA can form ion-conductive pores in planar phospholipid bilayers and unilamellar asolectin vesicles under acidic conditions (14,(22)(23)(24). The heptameric prepore rapidly loses the ability to form ion-conductive pores when exposed to acidic pH, however, limiting the partitioning into membranes. As shown in Fig. 1, in the absence of membranes the prepore loses a large fraction of its ability to induce K ϩ leakage from KCl-loaded DOPC LUV within 15 s after being exposed to pH 5.0 buffer, and this activity is almost completely lost within 1 min. We observed a similar inactivation when prepore was acidified before being added to planar bilayers 4 . Under acidic conditions, the prepore converts to an SDS-resistant state and forms aggregates, suggesting mechanisms for this inactivation.
His-R2 Mediates Binding of PA to Liposomes Containing Ni 2ϩ -chelating Lipids-In principle it should be possible to increase the partitioning of PA into liposomes by binding the protein to the membrane surface before it is exposed to low pH. To examine the use of a receptor domain to mediate binding of PA to membranes, we first prepared an N-terminally Histagged form of the ANTXR2 VWA domain (His-R2) and measured its binding to DOPC LUV doped with various levels of the Ni 2ϩ -chelating lipid (DOGs-NTA-Ni). The binding showed a clear dependence on DOGs-NTA-Ni up through 16 mol % ( Fig.  2A). The binding reached equilibrium by 15 min, and there was no nonspecific binding of His-R2 with extended incubation times. Imidazole effectively inhibited the binding, consistent with a Ni 2ϩ -mediated interaction (supplemental Fig. 1). The behavior of a C-terminally His-tagged ANTXR2 VWA domain (R2-His) was identical to that of the N-terminally tagged pro-4 B. Krantz, R. Melnyk, and R. Collier, unpublished data.  tein (data not shown), consistent with the locations of the N and C termini of the ANTXR2 VWA domain; adjacent to each other and diametrically opposite the conserved metal ion-dependent adhesion site motif, the focal point of PA binding (10).
We incubated PA with His-R2 and LUV containing 8% DOGs-NTA-Ni and then isolated the membranes by centrifugation and analyzed their associated proteins by SDS-PAGE. As shown in Fig. 2B, binding of PA to liposomes was at background levels in the absence of His-R2 and increased dramatically in its presence. Next, we tested the requirements for binding of LF N -DTA to liposomes (Fig. 2C). LF N -DTA, a fusion protein consisting of the enzymatic domain of DTA fused to the C terminus of the LF N , enters cells via the same pathway as LF and EF and blocks protein synthesis. Binding of LF N -DTA to nickel-containing liposomes was observed in the presence of His-R2 and trypsin-activated PA (nPA) (Fig. 2C). Native PA did not functionally replace the activated form, consistent with the fact that LF and EF bind to oligomers of PA 63 , which form only after proteolytic activation of PA. The results suggest that nPA associated with liposomes via His-R2, and after dissociation of PA 20 , PA 63 oligomerized on the liposomal membrane, allowing LF N -DTA to bind.
His-R2-dependent, PA-mediated Pore Formation-As shown in Fig. 3A, binding to His-R2 greatly increased the activity of nPA in releasing entrapped K ϩ from DOPC liposomes containing 8% DOGs-NTA-Ni (pH 5.0.) Within 60 s K ϩ release in the presence of His-R2 reached a level close to that seen following addition of gramicidin. In the absence of His-R2, the level of release by nPA was much lower. His-R2-dependent K ϩ release was strongly dependent on acidic pH and showed a sharp threshold at pH ϳ5.5 (Fig. 3B). In the absence of His-R2, the release was slower and more limited, and the threshold was about one pH unit higher. These results are consistent with the observation that binding of ANTXR2 VWA domain to PA prepore shifts the pH threshold of prepore-pore conversion to a more acidic pH range (10). Free and His-R2-complexed prepore gave results similar to those with free and His-R2-complexed nPA, respectively, in releasing K ϩ as a function of pH.
We characterized two PA mutants as controls in the K ϩ release assay. D512K, an oligomerization-deficient PA mutant FIGURE 3. His-R2 facilitates PA-dependent K ؉ release from LUV charged with a Ni ؉ -chelating lipid. Pore formation was monitored by K ϩ release from LUV containing 8% DOGs-NTA-Ni, as described under "Materials and Methods." A, representative K ϩ release traces generated by buffer alone, nPA (40 nM), and nPA (40 nM) ϩ His-R2 (80 nM) at pH 5.0. B, PA pore formation as a function of pH in the presence of buffer alone, nPA (40 nM), or nPA (40 nM) ϩ His-R2 (80 nM). Percentage K ϩ release was calculated as K ϩ release before gramicidin addition/ total K ϩ release ϫ 100. The signal from the buffer alone reflected extravesicular K ϩ in the liposome preparation and spontaneous K ϩ leakage. The traces shown are representative ones from three independent experiments. (3), showed low K ϩ release compared with wild-type nPA when complexed with His-R2 (Fig. 4A), indicating that PA 63 oligomerization was required. The other mutant, PA (K397D, D425K), oligomerizes to form the prepore but is unable to undergo the conformational change to the pore (25,26). This mutant is also a dominant-negative inhibitor (DNI) of toxin action; when the mutant co-assembles with wild-type PA, it inhibits pore formation and translocation. When complexed with His-R2, the DNI homoheptamer gave almost no K ϩ release, and the population of heteroheptamers assembled from a 1:1 mixture of wild-type PA and DNI (WT:DNI) showed greatly reduced K ϩ release relative to wild-type heptamer (Fig. 4B).
To demonstrate the insertion of PA 63 into liposomal membranes directly, we attached an environment-sensitive dye (NBD) to either of two sites within the protein and monitored its fluorescence emission intensity during pore formation. The intensity of NBD fluorescence at 544 nm increased upon a shift from a polar to a non-polar environment, and earlier we monitored insertion of various residues of the 14-strand ␤-barrel into membranes of Chinese hamster ovary-K1 cells after replacing those residues with Cys and reacting the protein with a thiol-reactive form of NBD (21). In that study, NBD attached to an introduced Cys at position 305 (facing the lipid bilayer) showed a strong increase in fluorescence, whereas at position 306 (facing the aqueous lumen) the fluorophore showed little change. In the current study we observed a strong increase of fluorescence intensity at pH 5.0 with NBD-PA(G305C) during pore formation but not with NBD-PA(N306C) (Fig. 4C). The increase seen with the labeled PA (G305C) was dependent on low pH, His-R2, and liposomes. These data correlate results in the LUV system with those in cells and support the ␤-barrel model of PA insertion.
PA Bound to Liposomes via a His tag on Its C Terminus Mediates Efficient Pore Formation-To probe the role of the ANTXR2 VWA domain in this system, we compared attaching PA to liposomes via a ANTXR2 VWA bridge protein with attaching it directly via a His tag on the C terminus of PA. nPA-His and His-R2-complexed nPA showed approximately equal levels of binding to DOPC liposomes containing 8% DOGs-NTA-Ni (Fig. 5A) and similarly high levels of activity in promoting K ϩ release (Fig. 5B). Also, when attached to mem-branes by either mode, nPA gave greater activity than unbound nPA lacking His tag (Fig. 3A). nPA-His behaved similarly to PA lacking the His tag in terms of oligomerization and ANTXR2 binding (data not shown), and both had low activity in releasing K ϩ from DOPC liposomes lacking the Ni 2ϩ -chelating lipid (Fig. 5C). Because PA that was bound to liposomes directly via its C terminus was as active in pore formation as PA bound via ANTXR2 VWA bridge, His-R2 apparently functioned primarily as a membrane anchor in this system.

R2 Lacking a His Tag Inhibits Aggregation of Heptameric PA in Solution and Facilitates Pore
Formation-When we performed controls to test the possibility that R2 lacking a His tag might affect pore formation by PA prepore, we found that indeed it increased K ϩ release from liposomes (Fig. 6A). However, R2 did not bind to liposomes at pH 4 -9 (data not shown), suggesting that, unlike His-R2, it did not mediate binding of PA to LUV. We therefore examined an alternative possibility; binding of R2 might slow the inactivation of heptameric PA 63 . We incubated the heptamer with excess R2 at pH 5.0 for various brief periods and then added the mixture to DOPC liposomes and monitored the kinetics of K ϩ release. As shown in Fig. 6B, in comparison with the data in Fig.  1, R2 significantly retarded the inactivation process. Also, we examined the effect of R2 on aggregation of the prepore by monitoring turbidity at 340 nm after lowering the pH from pH 8.5 to 5.0 (Fig. 6C). Although R2 did not significantly change the final level of aggregation, it caused a delay in the formation of aggregrates, on the order of minutes. Consistent with the known high affinity of R2 for PA, the maximal shift in kinetics was observed with a relatively small molar excess of the receptor domain. Bovine serum albumin at molar ratio of the protein to PA (8:1) did not affect the aggregation kinetics (data not shown).

DISCUSSION
Many intracellularly acting toxins initially bind to a receptor on the surface of host cells and later translocate across the endosomal membrane under the influence of acidic pH. In some of these proteins (e.g. diphtheria toxin) the pH-dependent membrane insertion process has been studied extensively in model membranes devoid of receptors (27)(28)(29)(30)(31)(32). In the case of anthrax toxin, the PA prepore has been shown to form ionconductive pores (channels) in both planar phospholipid bilayers and liposomes in the absence of receptors, and much information about the properties of these pores has been obtained recently from studies in planar bilayers (24,33). Although liposomes offer advantages over planar bilayers in addressing some aspects of pore formation and function, they have not been as useful in studying PA prepore insertion, in part because of the apparent inefficiency of insertion revealed in earlier studies (14). 3 In the current study we have addressed this problem and developed ways to overcome it by binding PA to vesicles in advance of pore formation. Besides making liposomes a more useful system to study anthrax toxin, the results shed light on the role of receptors in toxin action.
Within seconds after the pH was lowered to 5.0, the prepore converted in solution to an inactive form, lacking the ability to permeabilize DOPC LUV membranes to K ϩ . Concurrently the heptameric PA 63 complex became resistant to dissociation by SDS and fell out of solution, generating visible aggregates. Therefore, acidic pH triggered a conformational rearrange-ment in the prepore, yielding a short-lived intermediate that was competent to insert into membranes and then converted to an inactive form. The structures of the active and inactive forms are unknown. Conceivably the active form corresponds to the pore, containing a fully formed 14-strand ␤-barrel, as a transiently soluble intermediate, and the inactive form corresponds to aggregates of this form. However, the pathway and the end product(s) may be more complex. Regardless, the pathway leading to inactivation in solution clearly competes with the pathway leading to pore formation, and this can cause partitioning into membranes to be inefficient.
In an attempt to rectify this problem, we sought ways to bind PA to the surface of LUV in advance of acidification. Availability of a full-length functional receptor was still a limiting factor, but the use of metal-chelating lipids in combination with a poly His-tagged receptor ecto-ligand-binding domain appeared to be an alternative. Incorporating a His-tagged poliovirus receptor into liposomes via nickel-chelating lipids was recently used to study non-enveloped virus cell entry (34,35). We used a similar approach, based on the availability of the purified PAbinding domain (VWA) of ANTXR2 and the ability to produce a His-tagged form of this protein efficiently in Escherichia coli. LUV doped with a Ni 2ϩ -chelating lipid could bind His-R2, and this protein in turn could serve as a receptor for monomeric PA or prepore. The interaction between monomeric PA and R2 shows a K D ϳ 0.2 nM, and because the prepore can bind seven copies of R2, its linkage to R2 on liposomes would be expected to be effectively irreversible. Although the K D for the interaction between hexa-His tag and Ni 2ϩ -NTA was about 1 M (36), in the present study the His-R2 binding to DOGs-NTA-Ni was optimized by using excess DOGs-NTA-Ni relative to His-R2, which allowed at least 90% His-R2 bind to liposomes (data not shown). In addition, when prepore was saturated with R2, the interaction of the entire complex with DOGs-NTA-Ni-containing membranes would be expected to be effectively irreversible because of the multi-point contact of the seven His-R2 molecules with the nickel-containing lipid.
We showed that the efficiency of pore formation in doped LUV was strongly enhanced when PA was bound to vesicles via the His-R2 bridge protein, as judged by the kinetics of K ϩ release. The insertion half-time was on the order of several seconds, as measured by K ϩ release or enhancement of fluorescence of NBD at position 305 and was much more rapid than that observed with unbound PA. Thus the insertion was largely complete before the complex could proceed down the pathway to inactivation (if such a pathway is in fact operative with membrane-bound PA.) Consistent with observations in other systems, the binding to His-R2 shifted the pH threshold of conversion to the pore to a lower range (9,13,37); mutations in PA known to affect heptamerization or prepore-to-pore conversion inhibited pore formation in the predicted way (3,25); and membrane insertion by a fluorescently tagged form of PA gave results consistent with prior findings in cells and with the ␤-barrel model of pore formation (21). Also, binding of LF N -DTA was shown to depend on proteolytic activation of PA (3). As judged by these various criteria, doped DOPC LUV containing His-R2 as receptor represent a valid model of the membrane insertion process in vivo. FIGURE 6. R2 facilitates pore formation in liposomal membranes by slowing inactivation of heptameric PA at pH 5. 0. A, representative K ϩ release curves from liposomes generated at pH 5.0 by buffer alone, prepore (3 nM), and prepore (3 nM) complexed with R2 (40 nM). B, prepore (4 M) complexed with R2 (56 M) was treated with 1/10 volume of 1 M pH 5.0 buffer (final pH is 5.0) and after 15, 30, 60, and 120 s of treatment, the low pH-treated (PA 63 ) 7 ⅐R2 complexes were assayed for ability to release K ϩ from LUV. Prepore⅐R2 complexes that had not been treated at low pH treatment (0 s) were included as a control. C, prepore (64 nM) was incubated in buffer (1 mM MgCl 2 , 20 mM Tris-HCl, 150 mM NaCl, pH 8.5) with or without R2 at R2/PA molar ratios of 0, 1, 2, 4, and 8. After 60 min at room temperature, 1 M pH 8.5 or 5.0 buffers were added to the solutions with continuous stirring. The intensity (I) of scattered light at 340 nm was then monitored with time in a K2 fluorometer and was expressed as I-I 0 at 340 nm. Inset, plot of t1 ⁄2 values of PA aggregation versus R2/PA molar ratio.
An almost identical enhancement of pore formation was observed when PA was bound to doped LUV via a His tag appended to its C terminus. This implies that an important function of the PA receptor is to concentrate PA close to the surface of the membrane during prepore-to-pore conversion. Orientating the prepore with respect to the membrane is likely also to be important, but in our experiments binding either directly or via His-R2 occurred via the PA C-terminal domain 4, so the membrane-penetrating ␤-barrel would form on the surface of the prepore facing the membrane. Thus, our results do not discriminate between this and alternative orientations. Binding via the His-R2 bridge protein would place the prepore farther away (ϳ40 Å) from the membrane surface than binding via the C-terminal His tag on PA, but this did not diminish the efficiency of insertion. Furthermore, in the native ANTXR2 receptor there is an additional domain between the VWA domain and the transmembrane domain, so that ANTXR2bound PA prepore is at an even greater distance from the membrane surface in vivo. The 14-strand ␤-barrel of the pore is predicted to be ϳ100 Å long (38), with the transmembrane segment adjacent to the insertion tip, allowing a significant "reach" of prepore not immediately adjacent to the membrane surface. There is evidence in vivo that the receptor dissociates from the PA heptamer after conversion to the pore (37), and this would be consistent with the apparent insensitivity to the distance from the membrane at which the pH-dependent conformational transition to the pore occurs.
The finding that the R2 retards conversion of the PA 63 heptamer to an inactive form in solution is consistent with earlier results showing that the domain binds tightly to the prepore and lowers the pH threshold of the conformational transition to the pore by about one pH unit (10). By binding to both domain 4 and domain 2 of PA, the ANTXR2 receptor serves as a molecular clamp, which inhibits the conformational transition of domain 2 to form the 14-strand ␤-barrel. Conversion of the unliganded PA prepore to the SDS-resistant pore conformation can occur even at neutral pH, and binding to the ANTXR1 receptor affects the pH threshold of this conversion to a lesser extent than ANTXR2 (37). Our results do not speak to the question of whether partitioning into the membrane of membranebound nPA varies significantly with small changes in the pH at which prepore-to-pore conversion occurs. The ability to bind PA to liposomes doped with metal-chelating lipids, as shown here, makes such vesicles useful tools to study this and other questions pertaining to the action of anthrax toxin.