Identification of Shallow and Deep Membrane-penetrating Forms of Diphtheria Toxin T Domain That Are Regulated by Protein Concentration and Bilayer Width*

The α-helix-rich, hydrophobic transmembrane (T) domain of diphtheria toxin is believed to play a central role in membrane insertion by the toxin and in the translocation of its catalytic domain across membranes. In this report, T domain structure was studied using site-directed single-Cys mutants. The residues chosen, 322 (near the amino-terminal end of helix TH8), 333 (within helix TH8), and 356 (within helix TH9) were substituted with Cys and labeled with the fluorescent probe bimane. (Residues 333 and 356 should be located within the bilayer in the transmembrane state, and residue 322 should not penetrate the bilayer.) After insertion of T domain into model membrane vesicles, the location of bimane label relative to the lipid bilayer was characterized by its fluorescence emission and by its quenching with nitroxide-labeled phospholipids. It was found that when the T domain is added to dioleoylphosphatidylcholine-containing vesicles, all three residues reside close to the outer surface. However, at high T domain concentration or in thinner dimyristoleoylphosphatidylcholine-containing vesicles, a large fraction of residues 333 and 356 penetrate deeply into the membrane. In contrast, residue 322 remains exposed to aqueous solution under these conditions. These conclusions were confirmed by a novel antibody binding method. Antibodies that quench the fluorescence of 4,4-difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-3-indacene (BODIPY) groups were used to evaluate the exposure of BODIPY-labeled 322, 333, and 356. Maximum exposure of residues 333 and 356 to externally added antibody was only observed under conditions in which bimane fluorescence showed that these residues do not penetrate the bilayer. In contrast, residue 322 remained exposed under all conditions. We propose that the deeply penetrating T domain conformation represents a transmembrane or near-transmembrane state. The regulation of the transmembrane/nontransmembrane equilibrium should be a key to understanding diphtheria toxin membrane insertion and translocation. Our results suggest that toxin-toxin interactions may play an important role in regulating this behavior.

Diphtheria toxin is a protein secreted by Corynebacterium diphtheriae. It can be split into two chains, A (21 kDa) and B (37 kDa), joined by a disulfide bond (1). The crystal structure of the toxin shows that it consists of three domains (2)(3)(4)(5). The A chain is the catalytic (C) domain. The B chain contains the transmembrane (T) and receptor binding (R) domains. Membrane penetration is believed to occur after the toxin reaches endosomes (6). The low pH within the endosomal lumen induces a partial unfolding of the toxin, resulting in exposure of the hydrophobic regions and translocation of the A chain of the toxin into the cytoplasm (6). Once in the cytoplasm, the A chain catalyzes the transfer of the ADP-ribosyl group of NAD ϩ to elongation factor 2, inactivating protein synthesis.
The T domain is made up of nine ␣-helices, several of which contain hydrophobic sequences that play a critical role in membrane insertion and translocation (2)(3)(4)(5)(6)(7). Several recent studies have concentrated on two of the most hydrophobic helices, TH8 and TH9, since they are the most likely to form a transmembrane structure upon insertion into a bilayer (8 -15). Proteolysis studies indicate that in membrane-inserted T domain TH8 and TH9 are Pronase-insensitive (8). Studies on pore formation by T domain mutants are consistent with a model in which the loop between TH8 and TH9 reaches the trans side of the membrane in the pore-forming state (9 -13). (The trans side is the membrane surface opposite that (cis) into which the toxin initially inserts.) Studies using electron spin resonance (ESR) 1 have 1) shown that TH9 maintains helical structure upon insertion; 2) identified the aqueous and lipid facing sides of the helix; and 3) suggested a transmembrane structure (see "Discussion") (14,15).
Isolation of T domain from Escherichia coli strains expressing the T domain mutants H322C (322C), I333C (333C), and A356C (356C) was performed as described previously (9). (These T domain constructions contain the His 6 tag followed by a tetrapeptide linker (GSHM) attached to T domain residues 202-378 (9). The protein without any mutations in the T domain sequence is referred to as wild type.) Final purity appeared to be Ͼ95% as judged by SDS gel electrophoresis. Protein was stored at 4°C in 20 mM Tris-Cl or 10 mM Tris-Cl, 250 mM NaCl, pH 8.0. T domain concentration was determined from the absorbance at 280 nm using ⑀ ϭ 18,200 M Ϫ1 cm Ϫ1 , and converted to g using an approximate molecular weight of 20,000.
Fluorescence Measurements-Fluorescence was measured at room temperature with a Spex 212 Fluorolog spectrofluorometer operating in the ratio mode. Unless otherwise noted, measurements were made in a semimicrocuvette (excitation path length 10 mm, emission path length 4 mm). The excitation and emission slit widths used were 2.5 and 5 mm, respectively. Trp fluorescence was measured with excitation at 280 nm. Emission intensity was generally measured at 330 and 350 nm for 4 -8 s, or emission spectra were measured (at a rate of 1 nm/s). Bimane fluorescence was measured with excitation at 380 nm. Emission intensity was generally measured at 440 and 480 nm for 1-5 s, or emission spectra were measured. BODIPY fluorescence was measured for 20 -30 s with excitation at 488 nm and emission wavelength at 516 nm. For Trp and bimane, background intensities and spectra from samples lacking protein were subtracted. Background values for BODIPY were negligible (Ͻ1%). Spectra were generally smoothed after subtraction.
Bimane Labeling of T Domain-For the 322C and 356C mutants, about 50 g (2.5 nmol) of mutant protein diluted to 1 ml with a Tris-Cl/NaCl, pH 8, buffer was dialyzed overnight against 1 liter of 10 mM Tris-Cl, 150 mM NaCl with 2 mM dithiothreitol and then dialyzed for 6 -8 h against the same buffer without dithiothreitol. The protein was then mixed with 2 l of 15 mM monochlorobimane (probe:protein ratio 12:1 mol/mol) dissolved in ethanol. After shaking at room temperature for 15-20 min, the sample was dialyzed at 4°C overnight against 4 liters of 10 mM Tris-Cl, 140 mM NaCl, pH 8, and then against another 4 liters of 10 mM Tris-Cl, pH 8, for an additional 6 h and stored at 4°C. In all cases, it was assumed that there was full recovery of protein (see below).
Unless the 333C mutant was unfolded during the labeling reaction, the level of labeling was very low. Therefore, to label 333C, 45 l of 8 M urea was added to 45 l of 333C (1.1 mg/ml). Then 2 l of 15 mM monochlorobimane in ethanol was added. After shaking for 15-20 min, 2.79 ml of 10 mM Tris, 140 mM NaCl, pH 8, was added to dilute the urea and induce refolding. Both residual urea and excess bimane were removed by dialysis, as described above.
We confirmed that this protocol allowed renaturation by examination of Trp fluorescence. There was a red shift in Trp emission at high urea concentrations indicative of unfolding. This was totally reversed upon dilution of urea. In addition, the renatured protein underwent the normal pH-induced conformational change (see "Results"). Interestingly, only partial reversal of the fluorescence red shift was obtained after guanidinium chloride denaturation.
In all labeling experiments wild type T domain was labeled (using the same protocol as for the mutants) in parallel to the mutants. The degree of labeling of wild type T domain as judged by fluorescence intensity was found to be less than 1% for 322C and 356C and about 10% for 333C.
BODIPY Labeling of T Domain-To BODIPY-label the T domain, 5-6 l of 2 mM BODIPY-iodoacetamide dissolved in Me 2 SO was added to 50 g of wild type T domain, 322C, or 356C pretreated as described above dissolved in 994 -995 l of 10 mM Tris-Cl, 140 mM NaCl, pH 8. The solution was mixed well on a gyrotory shaker for 1 h at about 4°C and then dialyzed for about 20 h against 12 liters of 10 mM Tris-Cl, 140 mM NaCl, 0.7% (v/v) Me 2 SO, pH 8, with three buffer changes. 333C was labeled similarly except that labeling was performed after unfolding, as described for bimane labeling (see above). Labeling specificity was analyzed by comparing BODIPY-iodoacetamide fluorescence with that of wild type T domain labeled with the same probe. Fluorescence from the labeled mutants was 20 -300 times that of wild type. In a number of labeling experiments, it was determined that the final concentration of the labeled protein was estimated to be 70 -100% of the initial concentration. For simplicity, in most experiments it was assumed protein concentration was not altered by labeling.
Effect of pH on T Domain Fluorescence-Samples containing 5 g/ml T domain were transferred into 1.4 ml of 10 mM Tris-Cl, pH 8, by dialysis. Then 42 l of 5 M NaCl was added to bring NaCl concentration to 150 mM. The samples were then gradually titrated with 2-5-l aliquots of 0.8 M acetic acid to pH 4.6 and then with aliquots of 0.9 M HCl to achieve lower pH. For unlabeled protein, Trp fluorescence was measured after each addition; for bimane-labeled T domain mutants, bimane fluorescence intensity was measured after each addition. Readings were taken about 1 min apart. For these experiments, a 1-cm path length cuvette was used with 5-mm excitation and emission slits.
Effect of Acyl Chain Length on Bimane Fluorescence-Chloroform solutions of DOPG were mixed with DMoPC (14 carbon acyl chains), dipalmitoleoyl-PC (16 carbon acyl chains), DOPC (18 carbon acyl chains), dieicosenoyl-PC (20 carbon acyl chains), or dierucoyl-PC (22 carbon acyl chains) in order to obtain a 30% DOPG, 70% PC (mol/mol) mixture (120 nmol total lipid). The mixed lipids were dried under an N 2 stream and then redissolved with 7.5 l of ethanol. Vesicles were prepared by rapidly diluting this solution with 300 l of 500 mM sodium acetate, 150 mM NaCl, pH 4.4 (17). 2 Then 5 g of labeled T domain mutant in 300 l of Tris-Cl, pH 8, was added, yielding a final pH 4.5. Final lipid concentration was 200 M. Fluorescence intensity was then measured as described above.
Effect of Decane Addition on Bimane Fluorescence-To determine the effect of decane on the fluorescence of bimane-labeled T domain, samples containing DOPG/DMoPC or DOPG/DOPC were prepared as described above. Aliquots of 1:9 (v/v) decane/ethanol were added. After a 1-min incubation, fluorescence intensity was measured. There was no effect on fluorescence seen in control experiments in which ethanol alone was added.
Effect of Increasing T Domain Concentration on Bimane Fluorescence-Bimane-labeled T domain mutants were incorporated into 30% DOPG, 70% DOPC ethanol dilution vesicles prepared as described above, except using half the lipid and half the labeled T domain concentration to conserve materials. 2-6-l aliquots of purified wild type T domain were then added. After a 30-s incubation, during which fluorescence restabilized, fluorescence was remeasured.
Fluorescence Quenching of Bimane-labeled T Domain Mutants-For fluorescence quenching experiments, lipid vesicles were prepared in a similar manner to that described above. Organic solvent solutions of DMoPC or DOPC, DOPG, and (when desired) spin-labeled PC were mixed. Samples contained 30 mol% DOPG and 70 mol% total PC. Samples containing spin-labeled PC, had 15 mol% of 12SLPC, 5SLPC, or TempoPC. 3 The mixtures were dried under a N 2 stream, redissolved in a few drops of CHCl 3 , and redried with N 2 . Then they were dissolved in 7.5 l of ethanol and diluted with 300 l of 333 mM sodium acetate, 150 mM NaCl, 3.5 mM Tris-Cl, pH 4.4, to form vesicles. T domain dissolved in 300 l of 10 mM Tris-Cl, pH 8, was then added, to give a final pH of 4.5. Final lipid concentration was either 100 or 200 M, and the amount of protein in the samples was 2.5 or 5 g, respectively, to maintain a constant protein:lipid ratio. Emission spectra were scanned from 430 to 500 nm, and the fluorescence of samples with and without spin-labeled PC was recorded at the wavelength of maximum emission in the sample without quencher (i.e. 30% DOPG, 70% DMoPC or DOPC). For experiments in which the effect of protein concentration on quenching was measured, a small aliquot of a stock solution containing 10 g of wild type T domain was then added to the samples. Fluorescence spectra were remeasured after a 15-min incubation.
Quenching by Anti-BODIPY Antibodies-Anti-BODIPY experiments were done using sonicated small unilamellar vesicles prepared as de-scribed previously (17). Mixtures containing 1 mol of total lipid composed of 70% DMoPC and 30% DOPG or composed of 70% DOPC and 30% DOPG (mol/mol) were prepared. These were dried under a stream of N 2 at about 30°C, dissolved in CHCl 3 , redried with N 2 , hydrated with 250 l of 6.7 mM Tris-Cl, 150 mM NaCl, 167 mM sodium acetate, pH 4.5, and then sonicated in a bath sonicator (model G112SP1T, Lab Supplies Co., Hicksville, NY) for 25 min or until optically clear. Typically, a 40-l aliquot of the sonicated vesicles was mixed with 740 l of 6.7 mM Tris-Cl, 150 mM NaCl, 167 mM sodium acetate, pH 4.5, and then 0.5-2.5 g (10 -60 l) of BODIPY-labeled mutant was added, mixed with a micro-stir bar, and allowed to incubate for about 5 min. (The reason for the variation in protein concentration is that the BODIPY labeling efficiency varied. The experiments were performed with about equal amounts of BODIPY probe as judged by fluorescence.) The final volume of each sample was 800 l, and the final lipid concentration was 200 M. Fluorescence measurements were made, and then 40 l of anti-BODIPY was added to the samples. (It was found that this amount of antibody gives maximum quenching (not shown).) Every 15 min after antibody was added, the samples were remixed, and fluorescence was remeasured until the readings stabilized (up to 1 h). For samples with excess unlabeled protein, a small aliquot (5-20 l) containing 10 g of wild type T domain was added immediately after the labeled mutant. The neutral pH buffer in the wild type protein did not affect the final pH.
Effect of Removal of His 6 Tag-The N-terminal hexahistidine tag was removed in control experiments. Bimane-labeled 356C with the His 6 extension on the N terminus was treated with thrombin as described previously (9). The removal of the His 6 tag on the labeled protein was confirmed by SDS gel electrophoresis and by the chromatography on a nickel-nitrolotriacetic acid-agarose column (Qiagen, Chatsworth, CA). In the latter experiment, thrombin-treated T domain washed through, whereas untreated T domain bound, as detected by bimane fluorescence. Bilayer-inserted bimane-labeled 356C T domain lacking the His 6 tag showed the same blue shift as the protein with the His 6 tag in DMoPC-containing model membranes relative to that in DOPC-containing model membranes; the same red shift when decane was added to the DMoPC-containing model membranes; and the same blue shift when excess unlabeled T domain was added to labeled T domain incorporated into the DOPC-containing model membranes. We have also seen a lack of effect of the His 6 tag on T domain membrane insertion measuring the Trp fluorescence of wild type T domain and T domain Trp mutants. 4 Since the His 6 tag had no effect on T domain behavior, it was not removed in subsequent experiments.

Fluorescence Properties of Mutant and Wild Type T Domains in Solution-
The behavior of mutants with a single Cys at residues 322, 333, and 356 were studied to examine the structure of the T domain in model membranes. These mutants were chosen because they are in the region containing TH8 and TH9, two highly hydrophobic helices that have the capacity to form a transmembrane structure (9 -15).
We first examined whether the T domain mutants would undergo the low pH-induced conformational change observed in both whole toxin (20) and isolated T domain (9). This conformational change is the critical step inducing membrane penetration in cells (6,7). The conformational change was probed by measuring wavelength shifts in the Trp fluorescence of the T domain. (T domain has two Trp residues, Trp 206 in helix 1 (TH1) and Trp 281 in helix 5 (TH5)). Wavelength shifts were detected by the ratio of Trp fluorescence emission intensity at 330 nm relative to that at 350 nm (21). Fig. 1 shows the effect of pH on the 330:350 ratio for T domain in solution. There is a conformational change resulting in a lower ratio (i.e. a red shift) below pH 5, indicative of exposure of Trp to a more polar environment. This suggests increased contact of Trp to solution at low pH, in agreement with results showing partial unfolding of T domain at low pH (9,22). Just as importantly, Fig. 1 shows that the 330:350 ratio was similar for the wild type and mutant T domains, with the low pH transition occurring at about pH 5 in all cases. This shows that the introduction of a Cys has no significant effect on the low pH-induced transition.
Fluorescence Properties of Bimane-labeled Mutant T Domains in Solution-To examine the effect of pH on the behavior of the TH8/TH9 region, the experiment above was repeated with bimane-labeled Cys mutants. Bimane was chosen because it is a relatively small, uncharged, and fluorescent Cys-specific reagent.
The emission properties of the T domain-attached bimane in aqueous solution depended on its attachment site. Shifts in bimane fluorescence were measured by the ratio of bimane emission intensity at 440 nm relative to that at 480 nm. At neutral pH, bimane-labeled 322C and 356C mutants had a relatively low 440:480 ratio (i.e. were red-shifted), and labeled 333C had a much higher ratio (Fig. 2). This suggests that bimane attached to 333C is in a relatively nonpolar, buried location, consistent with the inability to label 333C without first unfolding the T domain (see "Experimental Procedures") and with both ESR studies (14) and the crystal structure of the toxin.
The effect of pH on probe fluorescence was then examined. As shown in Fig. 2, a conformational change was observed at low pH with bimane-labeled 356C undergoing a blue shift and labeled 333C undergoing a large red shift. These results show that the TH8/TH9 region participates in the low pH-induced change, in agreement with previous studies (9,14). The pH transition for labeled 333C occurs at a somewhat higher pH (by 1-1. buried bimane label. Bimane-labeled 322C remained redshifted at all pH values. This suggests it remains exposed to aqueous solution both at neutral and low pH. Fluorescent Properties of Model Membrane-inserted T Domain: Conformation Is Affected by Bilayer Structure-T domain readily inserts into the bilayer of model membrane vesicles at low pH (9,14,15). The binding of the labeled T domain mutants to 30% PG, 70% PC (mol/mol) vesicles was tight, with concentrations of 100 M lipid sufficient for complete binding as judged by fluorescence intensity changes upon lipid binding and centrifugation experiments (data not shown).
Our first experiments on T domain inserted into DOPG/ DOPC vesicles indicated that the bimane-labeled 322C, 333C, and 356C all occupied shallow depths in the bilayer (see below). This was a surprise, because of the hydrophobic nature of TH8 and TH9. Therefore, we searched for conditions that would regulate the depth of T domain insertion. We recently found that bilayer width can control the transmembrane orientation of hydrophobic helices (23). Therefore, we examined whether bilayer width would affect the insertion of the T domain. This was first done by comparing fluorescence in DMoPC-containing vesicles relative to that in DOPC-containing vesicles. DMoPC has the same double bond position as DOPC but has 14 carbon acyl chains in place of the 18 carbon fatty acyl chains of DOPC. This decrease of 4 carbons results in about 7 Å thinner bilayers for the pure lipids (24,25). As shown in Fig. 3 and Table I, all three bimane-labeled mutants gave relatively red-shifted fluorescence in DOPG/ DOPC vesicles. This suggests that in these vesicles all three residues are located in a relatively polar environment. However, bimane-labeled 333C and 356C underwent a distinct blue shift in DOPG/DMoPC vesicles relative to that in DOPG/ DOPC, as judged both by max and the 440:480 ratio, although bimane-labeled 322C did not. This indicated that some conformational change occurred in the T domain in which residues 333 and 356, which are within the hydrophobic regions of helices TH8 and TH9, move to a more nonpolar location in DMoPC-containing vesicles. In other words, these results suggest there is deeper insertion of the T domain in DMoPCcontaining vesicles than in DOPC-containing vesicles. The lack of a response in bimane-labeled 322C fluorescence to the change in lipid composition does not contradict this model, because residue 322, which is at the polar end of TH8, would not be expected to become buried in the membrane in any case.
Fluorescent Properties of Model Membrane-inserted T Domain: Dependence on Bilayer Width-To examine the correlation between bilayer width and conformation in more detail, the 440:480 and max ratio was examined for T domain inserted into vesicles containing PC with various acyl chain lengths (Fig. 4). This experiment showed that bimane-labeled 333C and 356C occupy more nonpolar locations (i.e. give blue-shifted fluorescence) when acyl chain length is less than 18 carbons. Similar effects were observed in the absence of DOPG (data not shown).
The effect of bilayer width was also examined by altering width in situ with decane. The addition of decane has been shown to increase bilayer width in several studies (26,27), and we found that it alters the transmembrane insertion of hydrophobic helices in the same manner as increasing acyl chain length (23). Fig. 5 shows that the addition of decane to DMoPCcontaining vesicles partially or totally reverses the blue shift seen with bimane-labeled 333C and 356C without affecting the fluorescence of bimane-labeled 322C. This is consistent with the effect of bilayer width on T domain conformation described above. In addition, it demonstrates these conformations can be interconverted in situ. In other words, T domain inserted in DMoPC-containing vesicles was not permanently trapped in the "blue-shifted" (i.e. deeply inserted) conformation. This suggests that there is a dynamic equilibrium between the two conformations (also see below).
Fluorescent Properties of Model Membrane-inserted T Domain: Effect of Protein Concentration-Altering the concentration of T domain within the bilayer had an effect similar to that of changing bilayer width. As shown in Fig. 3 and Table I, when wild type T domain was added to the bimane-labeled protein incorporated into DOPG/DOPC vesicles, a blue shift was observed for bimane-labeled 333C and 356C, but not for labeled 322C. Fig. 6 shows that the addition of 5-10 g of unlabeled T domain to 2.5 g of bimane-labeled T domain was sufficient to obtain a maximal blue shift in fluorescence. These results suggest that there is an interaction between T domains that affects their conformation.
The different T domain conformations predominating at different protein concentrations are also in a dynamic equilibrium. This is shown by the fact that the subsequent addition of unlabeled T domain to membrane-inserted labeled T domain affected the conformation of the labeled protein. This would not have been observed if the labeled T domain had been permanently trapped in the "red-shifted" (i.e. shallow) conformation upon insertion.
The interaction between T domain molecules did not depend on the absolute protein concentration but rather on the concentration of T domain in the bilayer. This was shown by studies on bimane-labeled 356C in which lipid concentration was varied instead of protein concentration. Blue-shifted fluorescence was obtained at low lipid concentrations (data not shown). Thus, it is the protein:lipid ratio (i.e. the concentration of T domain in the bilayer) that determines the shift in emission wavelength.
There was no further blue shift of bimane-labeled 356C or 333C fluorescence seen when protein concentration was increased in DOPG/DMoPC vesicles (data not shown). This suggests that increasing protein concentration has no additional  effect on the deeply inserted conformation present in DOPG/ DMoPC vesicles. Fluorescence Quenching by Spin-labeled Lipids: Assessing the Depth of Labeled Cys Residues-The experiments above show that residues 333 and 356 move into a more nonpolar environment in thin bilayers and at high T domain concentration in the bilayer. Fluorescence quenching by spin(nitroxide)labeled lipid was measured to determine more directly whether these emission shifts actually reflect a change in membrane depth of these residues. To do this, the quenching of the labeled T domain mutants by lipids carrying a shallow (TempoPC), medium (5SLPC), or deep (12SLPC) nitroxide was compared. A deep or shallow location of a fluorescent group is indicated when the quenching is strongest by the deep or shallow nitroxide, respectively. When there are two populations, one deep and one shallow, the quenching by the deep and shallow nitroxide is stronger than by the medium depth nitroxide.
Fluorescence Quenching by Spin-labeled Lipids: Bimane Attached to 322C, 333C, and 356C Locates Shallowly in DOPCcontaining Vesicles at Low T Domain Concentration- Table II illustrates the quenching of bimane-labeled 322C, 333C, and 356C. At low protein concentration in DOPC-containing bilayers, all three residues are at a shallow location. This is shown by the observation that the quenching is strongest by the shallow nitroxide located on the polar head group of TempoPC (F 12SLPC /F TempoPC Ͼ 1). These data are consistent with the wavelength shift data and a conformation of TH8 and TH9 in which these helices are close to the bilayer surface in DOPCcontaining vesicles (Fig. 7).
Fluorescence Quenching by Spin-labeled Lipids: Deep Location of Bimane Attached to 333C and 356C Both in DMoPC-containing Vesicles and in DOPC-containing Vesicles at High T Domain Concentration- Table II also provides information on the location of residues 333 and 356 in DMoPC-containing bilayers and in DOPC-containing bilayers at high protein concentration. Both residues 333 and 356 are located more deeply under these conditions. This is shown by the increase in quenching by the deep 12SLPC nitroxide relative to that both by the medium 5SLPC and shallow TempoPC nitroxide (decrease in F 12SLPC /F TempoPC ; ⌬ Ͻ 0). Since DMoPC and high protein concentration are also the conditions under which bimane-labeled 333C and 356C exhibit a blue shift, quenching confirms that the emission wavelength changes reflect deeper penetration of the T domain into the bilayer. 5 Quenching also shows bimane attached to 333C and 356C must be present at more than one depth in DMoPC-containing vesicles and at high T domain concentration, since the quenching by the shallow and deep nitroxides is greater than that by the intermediate nitroxide. This indicates there must be coexisting shallow and deeply penetrating populations (Fig. 7). 6 5 One concern we had in these experiments is that in the thinner DMoPC-containing bilayers quenching by 12SLPC could arise from 12SLPC molecules in both leaflets of the bilayer (18). This could increase 12SLPC quenching and make even a shallow residue appear deep. This possibility can be ruled out because 1) bimane-labeled 322C does not appear to be at a deep location in DMoPC-containing vesicles; 2) deep locations of the bimane-labeled 333C and 356C are also observed at high T domain concentrations, where this problem is not an issue; and 3) antibody binding confirms the deep location of labeled 333C and 356C. 6 Because of the presence of two populations at different depths, it is not useful to use the parallax analysis (18,19) to determine average bimane depth.

Fluorescence Quenching By Spin-labeled Phospholipids: Bimane Attached to Residue 322 Remains Shallow under All
Conditions-Table II also shows that, in contrast to the behavior of bimane-labeled 333C and 356C, bimane-labeled 322C remains at a shallow location in DMoPC-containing bilayers and at high protein concentration. This is shown by the observation that the shallow TempoPC nitroxide gives the strongest quenching under all conditions. This result is in agreement with the lack of a blue shift for bimane-labeled 322C under any conditions. Assessing T Domain Conformation by a Novel Antibody Binding Assay-As an additional approach to confirming the ability of the T domain to take on two different conformations, an antibody binding assay that takes advantage of the quenching of BODIPY fluorescence by anti-BODIPY antibodies was developed. Anti-BODIPY binding to BODIPY-labeled T domain mutants was used to assess the exposure of the labeled residues to external solution. Table III shows the quenching of BODIPY-labeled T domain mutants by anti-BODIPY. At lower protein concentrations in 30% DOPG, 70% DOPC, all three residues are almost equally exposed to the antibody. There is a moderate decrease in binding relative to that of free probe, which probably reflects steric blockage of antibody binding by BODIPY attachment to the protein. 7 The pattern of antibody binding is quite different in 30% DOPG, 70% DMoPC. There is a significant reduction in antibody binding to BODIPY-labeled 333C and 356C but not to BODIPY-labeled 322C. This suggests that residues 333 and 356 become less exposed to external solution but that residue 322 does not. A similar result is obtained at high T domain concentrations, although the loss of exposure of residue 333 is significantly less than that of residue 356. These results are consistent with the bimane fluorescence and fluorescence quenching experiments. In addition, the observation that there is still antibody binding to BODIPY-labeled 333C and 356C in the conditions yielding deeper insertion is in agreement with the observation from quenching that a population of shallowly penetrating molecules is present even when some molecules insert deeply (see above).

DISCUSSION
The Two Different Conformations of Membrane-inserted T Domain-This study demonstrates that membrane-inserted T domain occurs in at least two conformations. In one conformation, residues in TH8 and TH9 have a shallow location, suggesting a form in which these helices are close to the surface. In the other form, the helices must be more deeply inserted into the bilayer (Fig. 7). One question to be resolved in future studies is whether the deeply inserted form is fully transmembraneous.
Comparison with Previous Studies of T Domain Conformation-The existence of a conformation of the T domain in which TH8 and TH9 are close to the bilayer surface was not detected previously by ion conductivity or ESR studies (9 -15). In the former case, this is not unexpected because it is only the poreforming transmembrane population that is detected. On the other hand, using spin-labeled T domain, ESR, which has the potential to detect both shallow and deeply inserted species, showed that the structure of TH8 and TH9 was more consistent with a transmembrane structure. Based on the lipid and protein concentrations used in the ESR studies, the fluorescence results suggest that a mixture of the shallow and deeply penetrating conformations could have been present. It is possible that subtle differences in experimental conditions, differences between labeling groups, or different sensitivity to the presence  a Values shown are the percentage of quenching relative to that obtained for free BODIPY-iodoacetamide probe (74%) in the same pH 4.5 buffer. The results shown are the average of at least two experiments.  b This is equal to (F 12SLPC /F 0 )/(F TempoPC /F 0 ). The values shown are the average of ratios determined from individual experiments performed on the same day rather than the average of the entries in Table II. c ⌬ ϭ (F 12SLPC /F TempoPC ) Ϫ (F 12SLPC /F TempoPC ) DOPC .
of mixtures explains the differences between the ESR and fluorescence results. T Domain Protein-Protein Interactions-How does high T domain concentration promote deep insertion? The most likely possibility is that at high T domain concentration the T domain forms oligomers within the bilayer. Oligomers could allow for deep insertion by burying polar residues within their interior, i.e. away from the lipid bilayer. Oligomerization has already been identified as a key step in membrane insertion by anthrax toxin and ␣-hemolysin (28,29). In the case of ␣-hemolysin, there is a shallowly inserted form, which is believed to be converted to the transmembrane state upon formation of an oligomeric structure in the membrane (30). Therefore, it would not be surprising if oligomeric interactions were important for deep diphtheria toxin insertion as well. Some evidence for oligomeric interactions between diphtheria toxin molecules has been previously obtained from studies of pore formation and other properties (31)(32)(33). In preliminary studies, we have more directly detected oligomer formation in membrane-inserted whole toxin. 8 However, the nature of the oligomeric species is not yet clear. For example, whether there is a specific oligomeric stoichiometry is unknown. The presence of co-existing deep and shallow populations at high T domain concentrations brings up the unusual possibility of the formation of an oligomer in which different T domains have different positions in the bilayer. Alternately, there may simply be a mixture of separate shallow and deep populations.
Effect of Bilayer Width on Insertion-This study also demonstrated bilayer width could influence the degree of membrane penetration. Specifically, deep insertion was observed in thin DMoPC-containing bilayers. One possible explanation is that oligomerization of the T domain is promoted in thin bilayers, so that oligomers are present even at dilute concentrations in such bilayers. This is supported by the observation that the deep insertion inferred from bimane fluorescence in DMoPCcontaining bilayers is concentration-independent. Alternately, a deeply inserted monomeric structure may form. Transmembrane insertion may simply be promoted by the fact that in a thinner bilayer fewer polar groups must be buried within a hydrophobic milieu. In any case, these results suggest that altering bilayer thickness may be a useful tool for manipulating membrane protein insertion.