Interaction of Diphtheria Toxin with Mammalian Cell Membranes*

Uptake of ‘*51-labeled diphtheria toxin and serologically related proteins by a sensitive human HeLa cell line and by a resistant mouse L929 cell line has been studied. The evidence suggests that there is an initial rapid reaction between a recognition site present on the toxin Fragment B and specific plasma membrane receptors on the sensitive cell (there are approximately 4000/HeLa cell). This initial interaction is followed by a slow irreversible process during which there is a major conformational alteration of the toxin molecule causing the enzymically active 22,000.dalton Fragment A to become exposed to the cytosol. We suggest that it is at this point that cleavage of the NHz-terminal disulfide bond occurs lead-ing to release of Fragment A into the cytoplasm. The toxin Fragment B remains attached to the membrane, probably formed in a complex with receptor, and blocks entry of additional toxin molecules through the same site. Specific membrane receptors are lacking from mouse cells. Both HeLa cells and L929 cells internalize toxin, related nontoxic proteins, and inert molecules such as inulin nonspecifically into endocytotic vesicles. At 30” the bulk internalization of extracellular fluid is about 1.2% of their cell volume per h for both cell lines. Fragment A does not traverse the plasma membrane by a mechanism that depends on endocytosis. The interaction of diphtheria toxin with the sensitive cell membrane is discussed in relation to other protein toxins and certain glycopeptide tropic hormones in which relatively large, hydrophilic polypeptide fragments or subunits are

One of the many functions of the eukaryotic cell membrane is to serve as a barrier to prevent macromolecules from reaching the cytoplasm in an intact biologically active form. Nevertheless, there are certain types of large hydrophilic molecules that possess special properties enabling them to penetrate the lipid bilayer and reach the cytosol. In viral infections, for example, it is obvious that mechanisms must exist to enable viral nucleic acid to traverse the plasma membrane. Several protein toxins, or at least large polypeptides derived from them, have been shown to reach the cytoplasm without loss of their lethal enzymic activity. Thus diphtheria toxin (1,2), cholera toxin (3), and the toxic seed proteins, abrin and ricin (4), have each been shown to consist of an enzymically active polypeptide A, which must reach the cytoplasm to exert its lethal effect, and polypeptide B, joined to A through a disulfide bridge. In each case polypeptide B recognizes receptors on the sensitive cell, and by an as yet undetermined mechanism, allows polypeptide A to get across the plasma membrane. It seems likely that the passage of the (Y * This work was supported by Grant GB35579X from the National Science Foundation. $ Postdoctoral Fellow supported in part by French government and in part by the Institut National de la Sante et de la Recherche MBdicale (INSERM). subunits of certain peptide hormones across the plasma membrane of their target cell may be facilitated by their B subunits by a similar mechanism (5,6).
The diphtheria toxin molecule (62,000 daltons) is well suited for a study of the mechanism by which an enzymically active polypeptide is able to cross the plasma membrane to reach the cytoplasm. Its 22,000-dalton Fragment A is a potent enzyme (7) which, in uiuo as well as in uitro, catalyzes cleavage of NAD+ with the transfer of its ADPribose moiety to inactivate the eukaryotic polypeptide chain elongation factor, EF-2. ' Only a few molecules of Fragment A (perhaps only a single molecule) (8) need reach the cytoplasm in order to block protein synthesis and kill a eukaryotic cell. Nevertheless, many animal species such as rats and mice and cell lines cultured from them, are relatively resistant to the action of diphtheria toxin because they appear to lack specific membrane receptors recognized by the 38,000-dalton toxin Fragment B. Several serologically related nontoxic proteins (CRMs) are available (9). Thus CRM197, which cannot be distinguished immunologically from toxin itself, is the product of a missense mutation  Pappenheimer and Brown (16). For each 100 pg of toxin, 400 yCi of carrier-free iodine-125 (New England Nuclear) was used to yield about 2 x 10' cpm/pg of protein .  The labeled proteins  were  filtered  through  a small Sephadex  G-50 column  equilibrated  with  MEM supplemented with 2% fetal calf serum. As shown in Fig. 1, the radioactivity of '*"I-labeled toxin was almost equally distributed between Fragments A and B. Uptake of Labeled Toxin by Cell Suspensions-The procedure described below is rather long and tedious, but must be followed in order to reduce the background counts to a minimum and obtain quantitatively significant results. Very small amounts of aggregated or denatured labeled toxin, if bound to cells, will give counts that are too high. Some toxin may be nonspecifically adsorbed to cells. Washing with excess antitoxin is thus of great importance. Cells from spinner cultures were harvested by centrifugation and resuspended in MEM containing 2% fetal calf serum so as to contain about 5 x lo* cells/ml. Suspensions were kept at 0" prior to use. Immediately before each experiment, the '2sI-labeled toxin was passed through a double layer of 0.45 p Millipore filters. This step is essential to remove large toxin aggregates.
One-milliliter aliquots of cell suspension were then distributed into sterile glass tubes (16 x 150 mm) and equilibrated by stirring at given temperatures for 5 min before addition of labeled toxin.* Each toxin concentration was tested in duplicate.
After stirring at a given temperature for a given length of time, the cells were quickly transferred with Pasteur pipettes to tubes containing 1 ml of ice-cold PBS, mixed, and thereafter maintained at low temperature during the entire washing procedure that follows. Cells were collected in an International centrifuge at 800 rpm for 5 min. They were washed three times with l-ml portions of cold PBS. The washed cells were then resuspended in cold PBS containing 20 units of horses diphtheria antitoxic globulin previously passed through a 0.45 cc Millipore filter. The cells were transferred to glass tubes, stirred for 1 h in the cold, and then given three more l-ml washings with cold PBS without antitoxin. Finally, the cells were collected on glass fiber filters (Whatman GF/C) previously dipped in fetal calf serum. Cells were washed three times on the filter with cold PBS and then with 5% trichloroacetic acid and counted in a Picker y counter. In order to distinguish entry by specific binding from nonspecific internalization of molecules in solution through vesicle formation, suitable controls are essential.
We have found that a 15-to 2O.fold excess of unlabeled toxin or unlabeled CRM197 added 30 min before labeled toxin provided the most suitable control.
Since endocytotic vesicle formation is negligible in the cold, uptake of label at low temperature could not be used as a control. Control cells treated with excess unlabeled toxin before addition of labeled toxin were put through the same washing procedure as described above. For estimation of the number of toxin molecules specifically bound per cell, control counts were subtracted from counts of cells treated with labeled toxin only. Uptake of Znulin by Cell Suspensions-The same procedure was used to measure ["Clinulin uptake as described for labeled toxin except that the step in which cells were washed with antitoxin was unnecessary and ice-coid PBS was used instead.
[carboxyl-"C]Inulin (New England Nuclear, 1.9 &i/g) was dissolved by heating in distilled water. Twenty microliters of 0.1% solution was added to each milliliter of cell suspension.
Cells were harvested, washed as described for labeled toxin uptake, and counted at hourly intervals. Protein Synthesis-The rate of protein synthesis was estimated by measuring ["Clleucine incorporation into trichloroacetic acid-insoluble HeLa cell protein during a l-h pulse at 35" (10 after treatment with 0.6% trypsin at 37" in the presence of 1% /3-mercaptoethanol.
After 10 min the reaction was stopped by addition of 0.25% soy bean trypsin inhibitor.
thoroughly washed in the usual manner, except that the washing with antitoxin was omitted. Instead, the cells were finally stirred for 1 h at 10' with PBS. The washed cells were then suspended in 8 ml of 10 mM TrisiHCl containing 15 mM iodoacetate at pH 8.0 and allowed to swell for 5 min after which they were ruptured in a Dounce tissue homogenizer.
Plasma membrane ghosts were then purified by a single zonal centrifugation in a sucrose gradient as described by Atkinson and Summers (17). The membrane ghost fraction was collected from the interphase and after a 4-fold dilution in Tris buffer containing 3 mM MgCl, and 10 mM NaCl, was centrifuged for 15 min at 7000 x g. The pellet, which contained about 60% of all the L-['Hlfucose incorporated, consisted mainly of plasma membranes.
As pointed out by Atkinson and Summers (17), almost all of the fucose present in cells is associated with external membrane glycoproteins.

Nonspecific Internalization of Extracellular
Fluid-To estimate cellular uptake of molecules in solution, [carboxyl-'Clinulin uptake was measured as a function of time. As pointed out by Berger and Karnofsky (18), uptake of this inert molecule serves as a reliable measure of bulk movement of extracellular fluid into mammalian cells in culture. Fig. 2 shows that at 30", labeled inulin uptake by HeLa cells was linear over a 5-h period. The rate of fluid taken in by endocytosis into vesicles amounted to about 1.2% of the cell volume per h. The figure shows that the same rate of extracellular fluid internalization could be calculated from uptake of 1z61-labeled toxin at two different concentrations by cells pretreated for 30 min with a large excess of unlabeled toxin. In other experiments, it was found that the nontoxic related protein, CRM45, was internalized at the same rate. Fig.  2 shows that there was relatively little uptake of extracellular fluid by cells maintained at low temperature. The rate of inulin uptake by mouse L929 cells was also linear when followed over a 5-h period at 30". Within experimental error, the volume of extracellular fluid taken up per hour by mouse cells was the same as by HeLa cells whether measured by inulin uptake or by labeled toxin.
Specific Binding of Toxin by HeLa Cells-It is clear that one , , cannot use total uptake of label to measure specific binding of toxin to membrane receptors without correcting for nonspecific internalization by endocytosis. If the number of toxin-specific entry sites per cell is small, this correction could amount to an important fraction of the total uptake of radioactive label. Moreover, if specific membrane sites become saturated relatively quickly, as appears to be the case, we would expect that the magnitude of the correction for nonspecific uptake would continue to increase even when specific entry sites were no longer available.
Curve A of Fig. 3 shows the uptake of '261-labeled toxin by sensitive HeLa cells following l-h exposure at 30' to increasing concentrations of labeled toxin. Curve B of Fig. 2 plots the number of labeled toxin molecules internalized nonspecifically into vesicles by cells pretreated for 30 min with a 20-fold excess of unlabeled toxin. By subtracting B from A we obtain the corrected Curve C for specific binding. The curve for specific binding as a function of toxin concentration is replotted in Fig.  4 which summarizes data from several experiments. The curve is hyperbolic and approaches a limiting value of close to 4200 molecules bound/cell. Approximately the same limiting value for number of toxin molecules bound per cell is approached, even at low concentrations, if binding is followed as a function of time. However, after a few hours exposure to labeled toxin, the calculation of specific binding becomes imprecise. Thus following 5-h treatment of HeLa cells with lo-' M toxin, the Diphtheria Toxin and Membranes 5773 correction for nonspecific endocytosis approaches 90% of the total uptake of label. Even at relatively high toxin concentration, therefore, toxin does not interfere with endocytosis. A similar conclusion was drawn by Bonventre et al. (19).
CRM45 is present. The same is true for L-cells (Fig. 6) which show no specific binding of toxin at all. This confirms earlier studies suggesting that mouse cells are resistant to toxin because they lack specific receptors (16,20,21  Previous studies (10) have shown that the nontoxic protein, CRM197, can compete with toxin for entry into HeLa cells. Fig. 7 shows that unlabeled toxin competes similarly with lz51labeled toxin for specific binding to HeLa cell membranes.
From competition experiments such as these, it was concluded that there must exist specific receptors on the sensitive cell membrane. It was suggested that these receptors interact reversibly with intact Fragment B (10). However, we have thus far been unable to obtain unequivocal evidence for reversibility, either by direct measurements at low temperature or by chase experiments using excess unlabeled toxin added 2 to 20 min after addition of labeled toxin. Although other explanations are possible, it seems reasonable to suppose that there is an initial, rapid, reversible interaction between toxin and receptor followed by a slow irreversible process. In any case it seemed worthwhile to reinvestigate the relationships between CRM197, toxin, and the cell. rate of protein synthesis was followed. The figure shows that when the cycloheximide was removed, ["Clleucine incorporation into protein resumed in all flasks, and after a brief lag, continued at its initial rate in the control Flask D to which no toxin had been added. Cells treated for 2 h with toxin alone (Flask C) or treated simultaneously for 2 h with toxin and CRM197 in a ratio of 8:l (Flask B), incorporated ["Clleucine at about 80% the initial rate during the 1st h, but the rate fell rapidly thereafter and was only 40% of the initial rate 3 h after removing cycloheximide. On the other hand, the rate of protein synthesis in Flask A, in which cells were exposed to 5 x lo-@ M CRM197 for 3 h before adding 4 x lo-' M toxin, declined at a significantly slower rate. As would have been expected from Figs. 2 and 3 and from our previously published studies on the kinetics of intoxication (lo), the cells in Flask A behaved as if fewer than 20% of the entry sites were functional at the time toxin was added.
Blocking of Toxin Entry Sites by CRMl97-If the interaction of the toxin Fragment B with membrane receptor sites is in fact irreversible, or nearly so, then if CRM197 is allowed to interact with cells before addition of toxin, entry of toxin molecules through these sites should be blocked. Under these conditions there should no longer be any competition between CRM197 and toxin. Several lines of evidence suggest that this is indeed the case.
1. Fig. 8 plots the results of an experiment in which HeLa cells, in a series of spinner flasks, were treated with a relatively low concentration of CRM197 for 3 h in the presence of cycloheximide in order to inhibit synthesis of protein and of possible new receptor sites. They were then exposed, in the experimental Flask A, to a high saturating dose of toxin for 2 h. After washing free of excess cycloheximide, CRM197, and toxin, the cells were resuspended in fresh medium and their 2. In order to obtain a rough estimate of the time required for "irreversible" blocking of entry sites by CRM197, we have carried out several experiments in which HeLa cells were first exposed at 35" to a relatively low CRM197 concentration (4 x lo-' M) and then, after increasing intervals of time, toxin was added to give a concentration of lo-' M or greater. After 2 h treatment with toxin, ["Clleucine was added and its incorporation into protein measured during a l-h pulse. The results of a typical experiment are shown in Table I. Approximately 1 h exposure to CRM197 at 35" was required to obtain maximal blocking of toxin entry. At this time the rate of leucine incorporation was almost as great as the control without toxin and was 35 to 40% faster than when both CRM197 and toxin P ct" I RATIO: UNLABELED/LABELED TOXIN FIG. 7. Competition between labeled and unlabeled toxin for HeLa receptor sites. One-milliliter HeLa suspensions (1.2 x IO6 cells) were added to a series of tubes. To successive tubes were added Y-toxin to IO-' M together with increasing concentrations of unlabeled toxin. Tubes were incubated for 1 hat 30" after which the cells were collected, processed, and counted in the usual manner. The value obtained at the highest ratio of unlabeled to labeled toxin was used as the correction for nonspecific uptake. The curve is drawn to show uptake of iodine-125 to be expected for free competition between labeled and unlabeled toxin. HeLa cell s&pensions (2.5 x lo6 cells/ml in MEM containina 2% fetal calf serum) were nlaced in each of four small spinner flasks aid 5 x lOma M CRM197 was added to Flasks A and D. Fifteen minutes later, 2.4 x lo-" M cycloheximide, which is known not to interfere with toxin binding (22), was added to each flask. After 3 h at 35", 4 x 10e7 M diphtheria toxin was added to Flasks A and C. To Flask B was added both CRM197 and toxin to concentrations of 5 x 1Om8 M and 4 x 10m7 M, respectively. Flask D served as a control. After two more hours at 35", cells from each flask were collected, washed twice with medium in the centrifuge to remove cycloheximide, excess CRM197, and toxin, and were finally resuspended in medium containing 1 unit/ml of diphtheria antitoxin. At intervals, samples were withdrawn from each flask and pulse-labeled with [U-"Clleucine for 1 h. 0, CRM197 for 3 h, then toxin for 2 h (Flask A); 0, CRM197 and toxin together at 3 h (Flask B); X, toxin only at 3 h (Flask C); A, Control; no toxin (Flask D). Note: after addition of cycloheximide, protein synthesis fell to 15 to 20% or less in all flasks.

Diphtheria
Toxin and Membranes 5775  Aliquots of membrane suspension were mixed with an equal volume of a 10% suspension of antitoxin (anti-B) covalently bound to Sepharose 4B beads and centrifuged for 20 min at 7000 x g. The pellets were incubated 2 h at room temperature. The beads were then suspended in cold PBS, washed twice with PBS at low speed, and finally collected on GF/C filters. After further washing on the filter, they were dried and counted for tritium. As