Immunochemical Analysis Shows All Three Domains of Diphtheria Toxin Penetrate across Model Membranes*

Diphtheria toxin undergoes membrane insertion and translocation across membranes when exposed to low pH. In this study, the translocation of the toxin has been investigated by the binding of antibodies to two preparations of model membrane-inserted toxin. In one prep- aration, toxin was added externally to model membrane vesicles and then inserted by exposure to low pH. In the other preparation, toxin was entrapped in the vesicles at neutral pH, and then inserted by decreasing pH. At neutral pH, externally added antibodies could not bind to entrapped toxin, although they could bind to exter- nally added native toxin. However, after low pH exposure, antibodies against all three toxin domains (cata- lytic (C), transmembrane (T), and receptor-binding (R)) could bind to entrapped toxin, and also to externally added membrane-inserted toxin. The binding to the entrapped toxin shows that all three domains of the toxin translocate to the trans face of the membrane after exposure to low pH. The observation that antibodies bind to both external and entrapped preparations of toxin after low pH exposure shows that toxin inserts in a mixed orientation. sn -glycero-3-phosphoethanolamine; sn sn

domain, and a B chain, which comprises transmembrane (T) and receptor-binding (R) domains (1). After endocytosis, the acid environment of the endosome lumen triggers a conformational change which allows the toxin to insert and then translocate the C domain across the membrane bilayer (2). The C domain is then released into the cytoplasm. Hydrophobic helices in the T domain are the most obvious sequences that may be involved in translocation (1). However, several studies have indicated that the C domain may also participate actively in membrane interaction and translocation (3)(4)(5)(6)(7)(8)(9). Furthermore, the R domain, which has been thought to act simply to bind to the receptor for the toxin, becomes hydrophobic at low pH (10), raising the possibility it has an important role in translocation.
In order to clarify the role each domain plays in translocation, monoclonal antibodies to the C, T, and R domains and polyclonal anti-peptide antibodies were used to examine the interaction of all three toxin domains with the membrane bilayer. A system was used in which the toxin is added externally to vesicles or entrapped within the vesicle lumen. This latter system is analogous to endosomes, allowing the detection of translocation (5,11). The results show that all three domains traverse the bilayer and become exposed to the trans side of the membrane after exposure to low pH. This is a strong argument for the proposal that all three domains participate directly in translocation of the C domain in vitro. The observations that the unfolding of the C domain further promotes translocation and that the toxin inserts in a mixture of orientations are additional clues to the translocation mechanism.

EXPERIMENTAL PROCEDURES
Materials-ImmunoPlate MaxiSorp, flat-bottom 96-well microtiter plates were purchased from Nunc. Goat-anti rabbit Ig labeled with alkaline phosphatase was obtained by Southern Biotechnology Associates (Birmingham, AL). Goat anti-mouse Ig labeled with alkaline phosphatase was obtained by Jackson Laboratories (West Grove, PA) and Fisher Biotech. Sigma 104 phosphatase substrate (5-mg tablets) and octyl glucoside (OG) were purchased from Sigma. DOPC and DOPG were obtained from Avanti Polar Lipids (Pelham, AL). Anti-fluorescein whole antibody and F ab fragments (1-4 mg/ml in 0.1 M potassium phosphate, pH 8, with 2 mM sodium azide), rhodamine-DHPE, biotin-DHPE, biotin-X-DHPE, and streptavidin were obtained from Molecular Probes (Eugene, OR). Phast SDS-polyacrylamide gel electrophoresis gels and PhastSystem supplies were obtained from Pharmacia Biotech Inc. Nitrocellulose membranes were purchased from Schleicher & Schuell. Toxin was prepared as described previously (4,13). Antibodies were prepared as described in accompanying paper except that they were diluted with Tris buffer (15 mM Tris-Cl, 150 mM NaCl, pH 7.2.) prior to use. Antibody concentrations were identical to the antibody concentration reported in the accompanying article (12), except where noted.
Lipid Preparation and Entrapping Toxin-LUV with and without entrapped toxin were prepared by detergent dialysis as described pre-* This work was supported by National Institutes of Health Grant GM 31986.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.
Demonstrating Entrapped Toxin in the Native Conformation Does Not React with Externally Added Antibodies-A competition ELISA protocol was used as in the accompanying article (12), with the exception that free dimer toxin was used as coating antigen. For ␣-T1 (see accompanying article for antibody nomenclature), 115 l of toxin-containing vesicles were diluted to 200 l with Tris buffer and then different aliquots were diluted with vesicles without toxin to maintain 1 mg/ml lipid concentration. For ␣-C2 and ␣-R1, 88 l of toxin-containing vesicles were diluted to 109 l with Tris buffer and then different aliquots were diluted with vesicles without toxin to maintain 1.7 mg/ml lipid concentration. Tris-diluted antibodies were added and samples incubated for about 2 h at room temperature. Samples were then diluted to 320 l with Tris buffer, and 100 l aliquoted into the ELISA plate as described previously (12). Final antibody concentrations were 0.20 g/ml for ␣-T1, 0.98 g/ml for ␣-C2, and 0.56 g/ml for ␣-R1.
Demonstrating Entrapped Toxin Does React with Externally Added Antibodies after the Addition of Detergent-For experiments with ␣-T1, 9.4 l of 240 mg/ml OG was added to a 115-l aliquot of LUV-entrapped toxin or toxin added externally to LUV having the same protein and lipid concentrations to give an OG concentration of 15 mg/ml (for ␣-C2 and ␣-R1, 6 l of OG was added to 88-l toxin-LUV aliquots). For ␣-T1, 35 l of Tris buffer was added and the samples incubated for 30 min at room temperature (this step was skipped for the other antibodies). Then for ␣-T1, 40 l of additional Tris buffer was added (15 l for ␣-C2 and ␣-R1) and incubated for 15 min. Tris-diluted antibodies were added and samples incubated for about 2 h at room temperature. Samples were then diluted to 320 l with Tris buffer, and 100 l aliquoted into the ELISA plate as described previously (12). Final antibody concentrations were as in the preceding section.
Assay for Antibody Permeation into the Lumen of Model Membrane Vesicles-66-kDa FITC-dextran (predialyzed in 8-kDa dialysis tubing to remove small dextrans) was entrapped in LUVs with or without adenosine-3Ј-phosphate 5Ј-uridine phosphate-bound monomer (79.7% DOPC, 20.2% DOPG, and 0.12% pyrene-DHPE) as described above. Aliquots of coentrapped samples (lipid concentration of 1.8 mg/ml, toxin concentration of 7.2 g/ml, and a FITC-dextran concentration of 0.39 g/ml) were exposed to low pH by the addition of 200 mM sodium acetate, pH 4.4, for 30 min at room temperature (final pH 4.6). The samples were neutralized with 200 mM Na 2 HPO 4 , diluted with Tris buffer to give a volume of 800 l, incubated for 15 min at room temperature, and then fluorescein fluorescence was measured. 12 l of an anti-fluorescein isothiocyanate IgG (␣-FITC) and, where desired, aliquots of Tris-diluted ␣-R1 (10 l), ␣-C1 (30 l), or ␣-T2 (30 l) were added. In the case of ␣-R1, the ␣-FITC was added just after antitoxin. In the other cases, ␣-FITC was added 15 min before antitoxin. Fluorescein fluorescence was then monitored versus time. The final toxin and anti-toxin antibody concentrations were 1.07 and 0.53 g/ml for ␣-R1, 3.0 and 0.74 g/ml for ␣-C1, and 2.6 and 0.62 g/ml for ␣-T2. At the end of the incubation time, OG was added (50 l of 200 mg/ml for ␣-R1 or 40 l of 240 mg/ml for ␣-C1 and ␣-T2) and fluorescence was monitored for about 15 min.
A similar experiment using ␣-T2 was performed as for ␣-T2 above except toxin was added externally to LUV with only FITC-dextran entrapped (lipid concentration of 1.8 mg/ml and FITC-dextran concentration of 0.38 g/ml) and at the end of the experiment 150 l of 53 mg/ml OG was added. Similar experiments as above were also performed with the LrЉ conformation toxin. The coentrapped LUV had an initial lipid concentration of 1.4 mg/ml, toxin concentration of 6.7 g/ml, and a FITCdextran concentration of 0.27 g/ml. For the experiments with toxin added externally to LUV, the vesicles had an initial lipid concentration of 1.3 mg/ml and FITC-dextran concentration of 0.27 g/ml. The toxin concentrations were adjusted to equal those in the LrЈ samples before incubation at low pH for 30 min as above but at 37°C. pH was neutralized as above at 37°C, fluorescein fluorescence was measured, and then ␣-FITC was added as above and fluorescence remeasured. 15 min after pH neutralization, sample temperature was decreased to 24°C over 10 min, and then the desired ␣-toxin was added as above, and the fluorescein fluorescence monitored for 2 h.
Fluorescein fluorescence was measured with a Spex 212 Fluorolog spectrofluorometer. The samples were measured in quartz cuvettes having a 1-cm excitation and a 0.4-cm emission path lengths. The excitation and emission wavelengths were 503 nm excitation and 523 nm emission.
Topography Assay-ELISA-An aliquot of vesicle-entrapped toxin, or externally added toxin with the same vesicle and toxin concentration, was brought to about pH 4.6 by the addition of 200 mM sodium acetate pH 4.4 (about 1 volume of acetate to 5 volumes of vesicles), and then incubated for 30 min at either room temperature (to prepare LrЈ conformation toxin) or 37°C (to prepare LrЉ conformation toxin). The sample was then neutralized at the same temperature with 200 mM Na 2 HPO 4 (1.4 -2 volumes/volume of pH 4.6 sample) and enough Tris buffer to equalize toxin concentration between preparations. Different aliquots were removed, and (to all but the sample with the most toxin) empty LUV in Tris buffer was then added to bring samples to the same volume without changing lipid concentration. Tris buffer and Trisdiluted antibody was added to bring volume to 320 l. The final antibody concentrations were 0.65 g/ml for ␣-C1, 0.98 g/ml for ␣-C2, 0.20 g/ml for ␣-T1, 0.77 g/ml for ␣-T2, 0.56 g/ml for ␣-R1, 0.25 g/ml for ␣-R2, 6.3 g/ml for ␣-C 141-157 , and 4.7 g/ml for ␣-T 224 -237 . (These are the same concentrations as used in the protocol for the competition ELISA in the absence of lipid; see accompanying article (12).) The ELISA results were generally very reproducible when duplicates were compared. More variability was observed when comparing experiments performed on different days.
To examine translocation at lower toxin to vesicle ratios, a similar protocol was used except that the concentration of toxin used for entrapment was decreased by 10-, 20-, or 40-fold.
Topography Assay-Western Blotting-To detect antibody bound to vesicle-inserted toxin by Western blotting, biotin-labeled LUV with entrapped or externally added toxin, were prepared essentially as described above, with a composition of (w/w) 78.5% DOPC, 19.9% DOPG, 0.16% rhodamine-PE, 0.74% biotin-X-PE, and 0.74% biotin-PE. (Biotinylated lipid concentrations were based on the manufacturer's specifications.) Vesicles with toxin were incubated with 200 mM acetate, pH 4.4, and neutralized with 200 mM Na 2 HPO 4 as above to prepare toxin in the LrЈ and LrЉ conformations. To prepare vesicles with native toxin, premixed sodium acetate and 200 mM Na 2 HPO 4 were added. 10 l of Tris-diluted antibodies were then added to bring volume to 120 -124 l and incubated for 1 h at room temperature. The final toxin and antibody concentrations were about 7.0 and 1.7 g/ml, respectively, for ␣-C1 experiments, 6.5 and 0.54 g/ml for ␣-T1 experiments, and 6.5 and 0.67 g/ml for ␣-R2 experiments.
To pellet the vesicles 10 l of 1.3 mg/ml streptavidin (concentration determined spectroscopically using ⑀ 280 nm ϭ 40,542 M Ϫ1 cm Ϫ1 ) was added and samples incubated 30 min at room temperature (13). 120 l of each sample was centrifuged using a Beckman airfuge at 26 p.s.i. (95,000 rpm, 120,000 ϫ g) for 30 min using an A-100 rotor. 100 l was removed and washed three times, using 100 l of Tris buffer followed by a 20-min spin for each wash. After the final spin, 100 l was removed and 20 l of Tris buffer and 10 l of 110 mg/ml OG was added. After the pellet was completely dissolved by vortexing, 15 l of the sample was added to a 5-l solution of 50 mM Tris-Cl, 1 mM Na 2 EDTA, 2.5% (w/v) SDS, 0.01% (w/v) bromphenol blue, pH 8.0, and 7.5% (v/v) 2-mercaptoethanol) and boiled for 10 min. Then a 1-l aliquot of each sample was run on a 4 -15% PhastGel using the PhastSystem apparatus, transferred to a nitrocellulose membrane and detected as in the accompanying paper (12).

Preparations Used to Examine the Topography of Membrane-
inserted Toxin-Antibody binding was used to evaluate the conformation of diphtheria toxin inserted in model membranes. Fig. 1 schematically illustrates the preparation of the two types of model membrane vesicles containing toxin that were examined. In one preparation, the toxin was added externally to (outside of) model membrane vesicles, and then inserted by exposure to low pH. In the other preparation, the toxin was trapped in the aqueous lumen (inside) of the vesicle and then inserted by exposure to low pH. In the latter (entrapped) preparation, sites on the toxin exposed to external solution after its insertion are those facing the trans surface of the membrane (the side of the membrane opposite to that in which the toxin inserts) whereas in the former (external toxin) preparation externally exposed toxin sites face the cis side of the membrane (the side of the membrane into which the toxin inserts). The ability to study toxin in both of these preparations allows determination of the orientation of specific sites on the toxin. Notice that, as shown in Fig. 1, if toxin orients in a single direction, a monoclonal antibody will only react with one preparation.
Antibodies Do Not Bind to Entrapped Native Toxin-Competition ELISA was used to demonstrate that toxin in the entrapped toxin preparation was inaccessible to antibody prior to membrane insertion. In this assay the amount of binding of antibody to toxin is measured by the toxin's ability to inhibit antibody binding to ELISA wells (see Fig. 1 and "Experimental Procedures").
Antibody binding to entrapped native toxin was measured with antibodies ␣-C2, ␣-T1, and ␣-R1. Fig. 2 shows there is only very weak inhibition of antibody binding assayed by an ELISA in the presence of entrapped toxin, indicating that the entrapped toxin is inaccessible to antibody binding and only small amounts of toxin may be outside of the vesicles. 2 (This was true even in samples tested after 4 -6d storage of the vesicles at 4°C.) In contrast, when samples were solubilized with the detergent OG, a strong toxin-induced inhibition of antibody binding to ELISA plates, identical to that observed with toxin not entrapped in vesicles, was observed (Fig. 2).
Antibodies Do Not Permeate into the Vesicle Lumen Even after Toxin Insertion into the Bilayer-Another requirement of the antibody binding experiments is that antibodies do not permeate into the vesicle lumen. This is demonstrated by the results above for native toxin, but since the toxin forms pores upon low pH-induced insertion (14 -16) it is necessary to demonstrate that antibodies do not permeate the vesicle after toxin insertion. To do this the ability of an anti-fluorescein IgG (␣-FITC) to bind to entrapped FITC-dextran was monitored. The interaction of ␣-FITC with fluorescein is easily measured by the 90% decrease in fluorescein fluorescence upon antibody binding (Table I).
In these experiments fluorescein conjugated to a 66-kDa dextran was entrapped with or without toxin in the lumen of model membrane vesicles. After exposure to low pH to insert toxin, and the addition of anti-toxin antibodies where desired, the ␣-FITC was added and the fluorescein fluorescence was monitored. Table I shows samples with entrapped toxin and FITC-dextran do not form pores large enough to release 66-kDa FITC-dextran or allow ␣-FITC to enter into the vesicle lumen even after exposure to low pH. 3 This was true for toxin in both the LrЈ and LrЉ conformations (see below). Binding of anti-toxin antibodies (either ␣-C1, ␣-T2, or ␣-R1) did not affect this result, showing that anti-toxin interactions with toxin do not affect FITC-dextran and IgG permeation. These results imply that antibody binding to toxin inserted into vesicles will reflect the exposure of sites on the toxin to the aqueous environment on the outside of the vesicles to antibody.
Antibody Binding to Membrane-inserted Toxin: All Domains Penetrate to the Trans Face of the Membrane-In order to determine what sites of toxin are exposed to the vesicle exterior upon its insertion, antibody binding to membrane-inserted toxin was measured using the competition ELISA (see above). Membrane-inserted toxin was prepared by the low pH treatment of either toxin added externally to vesicles or entrapped toxin. Experiments were performed both with low pH-treated toxin in the LrЈ (23°C) conformation, 4 in which the C domain is folded, and LrЉ (37°C) conformation, in which the C domain is partially unfolded (4).
The relative strength of binding for different antibodies is summarized in Table II. One observation is that after mem- 2 The possible weak binding seen in the entrapped preparation might be due to a small amount of toxin that had leaked out of the vesicles. Extrapolation of the inhibition curves to determine the amount of toxin that inhibits ELISA by 50% (IC 50 ) for entrapped toxin and comparison to the IC 50 after detergent release can be used to estimate the amount of toxin outside of the vesicles. This gives an upper limit of 5-10% of the toxin in the entrapped preparation, after correction for the effect of OG on antibody binding. (The correction varied from no effect for ␣-R1 to a 2-fold difference for ␣-T1.) This percent is much too small to account for the amount of reaction seen with the entrapped toxin after low pH treatment. 3 Similar experiments were performed with an ␣-FITC F ab , showing that after low pH-induced toxin insertion and pH neutralization, the F ab , which are slightly smaller than toxin (about 50 kDa), was also unable to permeate the vesicles (data not shown). 4 We have changed our previously assigned names for the conformations the toxin takes after low pH treatment and pH neutralization from RЈ and RЉ to LrЈ and LrЉ, respectively. This was done to avoid confusion between the R domain and R conformation.
FIG. 1. Schematic illustration of the preparation of membraneinserted diphtheria toxin and the assay for antibody binding to membrane-inserted toxin. Entrapped toxin was prepared by detergent dialysis. Membrane-inserted toxin was prepared by the low pH treatment of toxin added externally to LUV and toxin entrapped within the lumen of an LUV. The side of membrane insertion is considered the cis side, and the opposite side of membrane insertion is considered the trans side. Notice that a given site (illustrated as a black dot at the apex of toxin molecule) is exposed on the external surface in only one preparation when orientation is unidirectional. brane insertion of toxin by exposure to low pH, antibodies can bind to the entrapped toxin preparation. Since toxin remains membrane-bound (5), this reflects the exposure of sites in membrane-inserted toxin to the trans face of the membrane. The IC 50 values obtained are generally comparable to those obtained for toxin in solution (see accompanying article (12)), indicating that the translocation of a large fraction of toxin molecules occurs. Furthermore, since interaction of ␣-C, ␣-T, and ␣-R antibodies with the toxin is observed, there must be translocation of all three domains to the trans surface of the vesicle. 5 To see if the toxin concentration affects this translo-cation, toxin was entrapped at lower toxin-to-vesicle ratios than in the protocol used above (65 toxin molecules/vesicle), and the binding of antibodies ␣-C2, ␣-T2, and ␣-R2 was measured. After low pH treatment of toxin entrapped at 6.5, 3.2, or 1.6 toxin molecules/vesicle, binding of all three antibodies was observed (data not shown).) 6 In at least the experiments with ␣-T2 and ␣-R2, the efficiency of binding appeared to be independent of toxin/vesicle ratio. Since there are so few toxins trapped per vesicle, these results suggest that translocation to the trans side of membrane does not involve the formation of large toxin oligomers.
Antibody Binding to Membrane-inserted Toxin: Toxin Inserts in a Mixture of Orientations-Another observation evident from the antibody binding data is that all of the anti-toxin antibodies are able to bind to both membrane-inserted toxin in both the externally added and entrapped toxin preparations. The only way this could occur is if the toxin inserts in a mixture of orientations, such that some toxin molecules have sites facing one side of the membrane, and others in the same sample have the same sites facing the opposite side of the membrane (see schematic figures). The observation that the IC 50 for antibody binding to entrapped and externally added toxin preparations are generally similar implies that there is a large fraction of toxin molecules facing in each orientation. This result could be interpreted in terms of a mixture of transmembraneous toxin molecules facing in different directions or a mixture of transmembraneous toxin and toxin that only penetrates the membrane incompletely.
The IC 50 is on the average somewhat lower for externally added toxin than for entrapped toxin in the LrЈ conformation, and averaging the difference in IC 50 values for externally added and entrapped toxin suggests that about 70% of the molecules have sites oriented facing the cis side of the membrane and 30% are oriented facing the trans side in this conformation. However, the difference in IC 50 could potentially be influenced by a difference in reactivity as well as the number of sites facing cis and trans sides.
Antibody Binding to Membrane-inserted Toxin: Effect of Toxin Conformation-Monoclonal antibody binding to mem- 5 An unlikely alternate interpretation of both the FITC-dextran experiments described earlier and the observation that entrapped toxin binds antibodies is that FITC-dextran binding to toxin blocks pore formation, and binding of antibody to entrapped toxin is due to pore formation and penetration of antibodies to the lumen of the vesicles. This was ruled out by experiments in which both toxin and FITC-dextran were entrapped in vesicles and then antibody binding to the entrapped toxin was measured. It was found that FITC-dextran did not significantly alter IC 50 , whereas a large increase in IC 50 would be expected if FITC-dextran was preventing antibody entry into the vesicles (data not shown). 6 These average toxin/vesicle values can be calculated because previous studies have shown the trapping of toxin in vesicles is about that expected for a random trapping process (5).

FIG. 2. Demonstration of toxin entrapment in vesicles by the inability of entrapped toxin to bind to externally added antibodies.
Binding of antibody to toxin was assayed by competition ELISA, in which the percentage of absorbance is inversely related to the amount of toxin binding (see text for details). Samples contained: entrapped toxin before the detergent octyl glucoside was added (q), after detergent added (E), or toxin added externally to vesicles after detergent added (ϩ).   (12) a The concentrations of antibodies and membrane-inserted toxin used in these samples are close to typical amounts used to evaluate antibody binding to toxin (see below). The experiments were performed with toxin in the LrЈ and LrЉ conformation (value in parentheses).
b The relative fluorescence was calculated by the fluorescence intensity at the indicated time relative to time zero, defined to be 100 (see footnote c). The fluorescence values were corrected for dilution.
c In control experiments about 10% release of FITC-dextran was found during the period of exposure to low pH, as judged by a 10% decrease in intensity of fluorescein dextran fluorescence by ␣-FITC. This agrees with previous observations (5). No additional intensity decrease was found after pH was neutralized. The release at low pH probably reflects a somewhat larger pore size as compared to after pH neutralization.
d Anti-toxin antibodies were generally added 15 min after pH neutralization; 15 min after neutralization was defined as time zero. e A 90% decrease of fluorescein fluorescence upon ␣-FITC addition is also found for FITC-dextran in solution in the absence of OG.
brane-inserted toxin in the LrЈ and LrЉ conformations was examined. In general, binding is slightly tighter when the toxin is in the LrЈ conformation, as judged by a lower IC 50 . ␣-C2 behavior is somewhat different (see Table II), showing closer to equal binding to LrЈ and LrЉ toxin. Binding of the anti-peptide antibodies ␣-C 141-157 and ␣-T 224 -237 to toxin differs from that of the monoclonals in that binding to the LrЉ toxin is tighter than binding to the LrЈ toxin. It is possible that the increased degree of unfolding in the LrЉ toxin promotes the binding of antipeptide antibodies by increasing the exposure of peptide epitopes, whereas the non-contiguous surface epitopes seen for monoclonal antibodies (17,18), are somewhat disrupted by the increased unfolding in the LrЉ toxin.
Antibody Binding to Membrane-inserted Toxin: Evidence That Increased Translocation of C and T, but Not R Domain, Occurs upon C Chain Unfolding-The IC 50 values in Table II show that for externally added toxin most antibodies bind the LrЈ conformation toxin more tightly than the LrЉ conformation toxin. The weaker binding of an antibody to its epitope on externally added toxin in the LrЉ conformation may be explained in one of two ways; 1) there is a structural change of an epitope to a weaker binding form in the LrЉ conformation, or 2) in the LrЉ conformation the epitope has translocated to the trans side of the membrane bilayer (and so faces the antibodyinaccessible vesicle lumen for most of the toxin molecules). It is possible to distinguish these possibilities by comparing antibody binding to entrapped toxin in the LrЈ and LrЉ conformations. If the decrease in antibody binding to toxin in LrЉ conformation were due to a change in epitope structure, antibody binding would show a decrease in binding to LrЉ toxin relative to LrЈ toxin similar to that seen with externally added toxin. However, if the weaker reactivity of the LrЉ toxin is due to translocation of the epitope across the bilayer, the ratio of reactivity of the entrapped LrЉ toxin relative to LrЈ toxin should be higher than for externally added toxin.
The second explanation is seen to be correct for several epitopes when IC 50 values are examined. Table III shows the IC 50 LrЉ/IC 50 LrЈ ratio is lower for entrapped toxin than for externally added toxin in several cases, i.e. the reactivity of several antibodies with LrЉ toxin normalized to the reactivity of the LrЈ toxin is higher in the entrapped preparation than in the externally added preparation. We have defined the change in the LrЉ/LrЈ IC 50 ratio as the relative translocation index. This is a measure of how much a site moves to the trans face upon unfolding of the C domain, the major change in the toxin conformation on going from LrЈ to LrЉ conformation (5). Inter-estingly, a translocation index close to 2, indicating increased translocation for toxin in the LrЉ conformation, is seen for the anti-C and some anti-T antibodies, but not for the anti-R antibodies, where the index is close to 1. This suggests that additional C and T domain sites move to the trans face in the LrЉ toxin but the location of at least some R domain sites do not change location. The increased movement of sites to the trans face in the LrЉ toxin confirms and amplifies previous observations of increased toxin movement to the trans surface of the membrane in the LrЉ toxin (5) and studies showing a role of C chain unfolding in translocation (5, 19) (see "Discussion").
Direct Demonstration of Antibody Binding to Membrane-inserted Toxin-Monoclonal antibody binding to membrane-inserted diphtheria toxin was more directly detected than in the ELISA by using centrifugation to isolate anti-toxin-bound vesicles containing membrane-inserted toxin. After solubilization, pelleted vesicles were analyzed for bound antibodies by probing with anti-mouse antibodies, and for toxin by probing with polyclonal anti-toxin. Fig. 3 shows the results of such experiments using ␣-C1, ␣-T1, and ␣-R2. When the toxin preparations were not exposed to low pH (native toxin), none of the antibodies bound to the vesicles (lanes 1 and 4 in each panel, upper half). This rules out b In samples containing externally added toxin membrane inserted toxin in the LrЈ conformation a significant amount of toxin, about 25% as judged by Western blotting, was found not to insert. Since the antibody binding of LrЈ toxin in solution is generally slightly weaker than in the membrane, this should have little effect on our results, perhaps shifting IC 50 values upwards by 10 -15% relative to what they would be if insertion were complete. The exceptions are ␣-C2 and ␣-R2, which bind antibodies stronger in solution, and therefore probably have significantly higher IC 50 than shown in the Table for the LrЈ toxin in the membrane. In the LrЉ conformation, almost all of the externally added toxin remained membrane bound. In addition, entrapped toxin remains totally membrane associated for both the LrЈ and LrЉ conformations ((5) and data not shown).  In contrast, when toxin was inserted by exposure to low pH, the antibodies bound to the vesicles both with externally added (lanes 2 and 3 in each panel) and entrapped toxin (lanes 5 and 6 in each panel), in agreement with the ELISA results. This confirms the conclusion that there is translocation of sites on all three domains of the toxin to the trans face upon insertion. 7

Structural Model For Membrane-inserted Diphtheria Tox-
in-This study shows that antibodies can be used to examine the structure of membrane-inserted diphtheria toxin. The observation that all three domains of the toxin undergo translocation to the trans face of the membrane requires a model of translocation in which all three domains participate in the insertion and translocation process in vitro (but perhaps not in vivo, see below). Despite the prejudice that the T domain, with its potentially transmembrane ␣-helices, should be the main part of the toxin undergoing insertion, the observation that isolated C (3, 6) and R domains (Ref. 10 and data not shown) become hydrophobic at low pH, isolated C domain can membrane insert (3) and proteolytic evidence for R domain insertion (23) support a role for these domains in the insertion and translocation process. Since the C and R domains lack long hydrophobic stretches, this brings up the question of the structure of membrane-inserted C and R domains. We have proposed that transmembrane ␤-sheets, similar to those formed by porins (20), may be formed by these domains (21). ␤-Strand 4 in the R domain is rich in hydrophobic residues, while ␤-strands 8 and 10 in the R domain and ␤-strands 4, 5, 6, and 8 in the C domain have the striking regions of alternating hydrophilic and hydrophobic residues found in transmembraneous ␤-strands. An alternate possibility is that these strands insert in the membrane in a non-transmembraneous fashion.
It has been observed that isolated catalytic, transmembrane, and receptor binding domains of Pseudomonas exotoxin A become hydrophobic at low pH (22). This may mean that the translocation mechanism for diphtheria toxin and exotoxin A is similar. This is interesting in terms of the similar secondary structure, but relative lack of sequence similarity, between the transmembrane and receptor binding domains of diphtheria toxin and Pseudomonas exotoxin A.
Unfolding of the C Domain and Translocation of Sites to the Trans Surface-There is an increasing body of evidence that the partial unfolding of the C domain promotes its translocation (5,23,24). Jiang et al. (5) showed that upon the unfolding of the C domain at 37°C, additional regions of the A and B chain become exposed to the trans side of membrane-inserted toxin, suggesting translocation of both chains across a membrane bilayer. The observations in this report showing increased movement to the trans surface for toxin in the LrЉ conformation support this proposal, and indicate that the transmembrane insertion of some part of the T domain is also promoted by C domain unfolding. A model of toxin translocation is shown in Fig. 4. It shows that the C domain and certain regions of the T domain translocate across the bilayer upon unfolding of the C domain by heating to 37°C, while the R domain does not translocate across the bilayer. Since our anti-T antibodies are against the amino-terminal end of the T domain, it suggests that some of the amino-terminal end of the T domain translocates across the bilayer upon C domain unfolding. This proposal is supported by protease and mutagenesis studies, which show that the amino-terminal end of the T domain is important for translocation of the toxin across a membrane (25)(26)(27). The observation that this change does not occur for the R domain suggests the R domain may not change its degree of insertion during the translocation process. In future studies it will be interesting to see how the exposure of T and R domains to the trans surface changes upon the release of the C domain from the membrane.
Orientation of Membrane-inserted Toxin in vivo-The observation that diphtheria toxin inserts in a mixed orientation brings up the question of whether this can occur in vivo. This cannot be answered at present, but it should be noted that several studies suggest this possibility. Two studies of toxin interaction with cells have been interpreted as indicating the presence of two populations of toxin, only one of which translocates (28,29). In entry via endosomes it has been reported that only a third of the toxins translocated (29), which agrees with the percent toxin penetration to the trans surface found in this study. Another study has shown that not all toxin molecules insert such that they are accessible to a specific chemical modification (30). This could be interpreted in terms of toxin oligomers with mixed orientations. Nevertheless, it should be remembered the mixture of orientations (specifically the minor (30%) orientation) may not occur in vivo. 7 The sensitivity of the Western blotting technique is insufficient to examine the difference between the level of binding to toxin in LrЈ and LrЉ conformations. Vesicles with antibody bound to membrane-inserted toxin were isolated by centrifugation. Western blotting followed by reaction with alkaline phosphate conjugated to anti-mouse antibodies was used to directly detect the amount of antibody bound to the vesicles (top half). Western blotting followed by reaction with horse polyclonal anti-toxin antibodies and rabbit anti-horse antibodies conjugated to alkaline phosphate was used to detect the amount of membrane-inserted toxin in each sample (bottom half). ␣-C1, ␣-T1, and ␣-R2 bound to vesicles with membraneinserted toxin were run on a non-reducing 4 -15% Phast System gel.