Proc. Nati. Acad. Sci. USAVol. 82, pp. 1692-1696, March 1985Biophysics

Channels formed by botulinum, tetanus, and diphtheria toxins inplanar lipid bilayers: Relevance to translocation of proteinsacross membranes

(voltage-dependent channels/pH-gating/channel-sizing)

DAVID H. HOCH*, MIRYAM ROMERO-MIRA*, BARBARA E. EHRLICH*, ALAN FINKELSTEIN*,BIBHUTI R. DASGUPTAt, AND LANCE L. SIMPSONt*Departments of Physiology and Biophysics and of Neuroscience, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, NY 10461; tFoodResearch Institute, University of Wisconsin, 1925 Willow Drive, Madison, WI 53706; and tDepartment of Pharmacology, College of Physicians and Surgeons,Columbia University, 630 West 168 Street, New York, NY 10032

Communicated by A. M. Pappenheimer, Jr., November 5, 1984

ABSTRACT The heavy chains of both botulinum neuro-toxin type B and tetanus toxin form channels in planar bilayermembranes. These channels have pH-dependent and voltage-dependent properties that are remarkably similar to those pre-viously described for diphtheria toxin. Selectivity experimentswith anions and cations show that the channels formed by theheavy chains of all three toxins are large; thus, these channelscould serve as "tunnel proteins" for translocation of activepeptide fragments. These findings support the hypothesis thatthe active fragments of botulinum neurotoxin and tetanus tox-in, like that of diphtheria toxin, are translocated across themembranes of acidic vesicles.

Diphtheria toxin (1), botulinum neurotoxin (2), and tetanustoxin (3) are proteins that are similar in origin and macro-structure. All three toxins are synthesized by bacteria (Cory-nebacterium diphtheriae, Clostridium botulinum, and Clos-tridium tetani) as single polypeptide chains (diphtheria tox-in, "'60 kDa; clostridial neurotoxins, ==150 kDa). Whenexposed to trypsin or trypsin-like enzymes, they are cleavedto yield two-chain molecules in which a heavy-chain poly-peptide is linked by a disulfide bond to a light-chain polypep-tide. The two-chain structure is the active form of the threetoxins.

Various techniques have been used to generate polypep-tide fragments from the toxins. The most straightforward ofthese is disulfide-bond reduction, which releases the heavychain from the light chain. Alternatively, the clostridial neu-rotoxins have been exposed to limited proteolysis (e.g., withpapain) to generate a fragment B and fragment C. Finally,traditional techniques have been used to select mutant orga-nisms that synthesize incomplete toxins, such as the CRM45fragment of diphtheria toxin, from which the B45 fragmentcan be formed. The various toxins and their fragments areillustrated in Fig. 1.

Diphtheria toxin is the best characterized of the three tox-ins. The entire molecule has been sequenced (4, 5), and threeseparate domains mediating binding activity (6, 7), channel-forming activity (8), and enzymatic activity (9) have beencharacterized. The carboxyl terminus of the heavy chain me-diates binding to cell surface receptors, and the light chain isan enzyme that catalyzes ADP-ribosylation of elongationfactor 2. The heavy chains of the clostridial neurotoxins alsomediate binding (10-12), but the nature of the putative enzy-matic activity of their light chains has not been established.An important question that remains unanswered for all

three toxins is the mechanism by which they reach the cyto-sol. There is evidence that diphtheria toxin is internalized by

receptor-mediated endocytosis (13); there is also evidencethat the endocytic vesicles become acidified and that the ac-tive fragment of diphtheria toxin requires this acidic environ-ment for its translocation across the vesicle membrane (14).There are findings that suggest that botulinum and tetanusneurotoxins follow a similar pathway into the cytosol (15-17). These observations leave open the question of whetherthe active fragments from the three toxins use the samemechanism for crossing the vesicle membrane.

Previous studies have shown that the amino terminus ofthe heavy chain from diphtheria toxin (18) and whole diph-theria toxin (19) form channels in lipid bilayers, and it hasbeen proposed (18) that these channels provide the pathwayfor the light chain to cross membranes. Here we report thatthe heavy chains of both botulinum neurotoxin type B andtetanus toxin also form channels in lipid bilayers. Further-more, for all three toxins, channel formation is maximalwhen the protein-containing (cis) side of the artificial mem-brane is at low pH (-4.0) and the opposite (trans) side is atpH -7.0, a pH gradient comparable to that across the mem-branes of acidic vesicles in cells. The channels for all threetoxins are very large, as determined by selectivity experi-ments with large anions and cations, and this finding is com-patible with the idea that the channels function as "tunnelproteins" for translocation of fully extended active frag-ments. In addition, tetanus toxin channels display a voltagedependence similar to that of diphtheria toxin channels,opening when positive voltages are applied to the cis side ofartificial membranes and closing when negative voltages areapplied. We discuss these findings in the context of currentviews on protein translocation across membranes.

MATERIALS AND METHODS

Planar lipid bilayer membranes separating two salt solutionswere formed at room temperature from the union of twomonolayers (20) across a hole (100 ,Am diameter; pretreatedwith squalene) in a Teflon partition. The composition of thesalt solutions is described below and in the figure captions;the membranes were made of asolectin, a crude mixture ofsoybean phospholipids (lecithin type II; Sigma) from whichneutral lipids were extracted (21). Various proteins were ap-plied to the asolectin membranes, including diphtheria toxin(R. J. Collier, Department of Bacteriology, Univ. of Califor-nia, Los Angeles), the B45 fragment of diphtheria toxin(A. M. Pappenheimer, Jr., Department of Biology, HarvardUniv.), tetanus toxin and its heavy and light chains (P. Bo-quet, Department of Microbiology, Pasteur Institute), andfragment B of tetanus toxin (0. Zwisler and K. D. Hungerer,Behring Institute). Botulinum neurotoxin type B and itsheavy and light chains were prepared by one of us (B.R.D.);

1692

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Proc. Natl. Acad. Sci. USA 82 (1985) 1693

LIGHT HEAVYs-s

a H2Ni - lICOOHCRM45 ----------------------

B45 -----------

DIPHTHERIA TOXINand FRAGMENTS

LIGHT

bHEAVY

s-s..

I-------------------------II-------------

B C

TETANUS TOXINand FRAGMENTS

HEAVY

100

---COOH BOT ULI NUMTOXIN

150

FIG. 1. Schematic diagram of diphtheria, tetanus, and botulinum toxins. The dashed lines below the toxins indicate the portions of themolecules found in various toxin fragments. The scale is in kDa. All toxin molecules are shown in the "nicked" form, with their heavy and lightchains connected by a disulfide bond. Disulfide bonds are not drawn to scale. (a) Diphtheria toxin and fragments. CRM45 lacks the 17-kDacarboxyl-terminal portion of the toxin which is involved in receptor binding. Disulfide bond reduction of "nicked" CRM45 yields the B45fragment. (b) Tetanus toxin and fragments. Fragments B and C are produced by proteolytic cleavage (papain) of the heavy chain. Fragment C isinvolved in receptor binding. Note that fragment B of tetanus toxin corresponds to CRM45 of diphtheria toxin. (c) Botulinum toxin.

the details of the isolation will be reported elsewhere.After addition of toxins or fragments to the cis compart-

ment, known voltages were applied across the membraneand the resulting current responses were measured. Themembrane conductance (g) in symmetric salt solutions is de-fined as current (I) divided by voltage (g = I/V), where V isthe potential of the cis compartment. In the absence of tox-ins or fragments, g was -5 picosiemens (pS).

120 pA 50 pA 11004secK

t t .I -L-A

b

l=

RESULTSpH Gradient and Channel Formation. The addition of frag-

ments possessing the amino terminus of the heavy chains ofclostridial neurotoxins (i.e., fragment B of tetanus toxin orheavy chains of botulinum and tetanus toxin) to one side ofan asolectin membrane separating solutions at a symmetricpH of 7.0 resulted in modest single-channel activity (Figs. 2aand 3a). There was considerable dispersion in single-channel

50

120 pA4 sec

12 pA_ 0 4 sec

2 pAi 4 sec +50 r) mV

FIG. 2. Effects of cis and trans pH on the rate of conductance rise produced by tetanus toxin and its fragment B. The figure illustrates thatthese molecules are more active when membranes separate symmetric low pH solutions than when they separate symmetric neutral pHsolutions and are dramatically more active in the presence of a pH gradient (cis, low; trans, neutral). The membranes separated identical saltsolutions (1 M KCl/5 mM CaCl2/5 mM 3,3-dimethylglutaric acid/0.1 mM EDTA; unless otherwise indicated, this solution was used in allexperiments described). (a) Both cis and trans solutions were at pH 7.0, and the potential across the membrane was clamped at 50 mV (cispositive). At the time indicated by the first arrow, fragment B was added to the cis compartment to a concentration of 4 tkg/ml. There was

virtually no change in membrane conductance (current). About 10 min later, the cis pH was lowered to 4.5 with dimethylglutaric acid (secondarrow), resulting in a tremendous increase in conductance. The vertical jumps in the record reflect the indicated changes in current scale.Similar records were obtained with both whole tetanus toxin and its heavy chain. (Inset) Single-channel activity at symmetric pH 7.0 producedby whole tetanus toxin (10 /Lg/ml). The record shows channels turning on when the voltage was switched to +50 mV. Note the higher sensitivityin the Inset. Fragment B produced channels with similar single-channel conductances. (b) Both cis and trans solutions were at pH 4.0, and thepotential across the membrane was clamped at +40 mV. About 10 min prior to the start of the record, tetanus toxin was added to the ciscompartment to a concentration of 5 ;Lg/ml. A slow rise in macroscopic conductance is seen in the record (in contrast to the virtual absence ofresponse at pH 7.0). When the trans pH was raised to pH 6.5 with Hepes (marked by the horizontal bar), the rate of conductance risedramatically increased. The vertical jumps in the record reflect the indicated changes in current scale. Similar records were obtained with bothheavy chain and fragment B. (Inset) Single-channel activity at symmetric pH 4.0 produced by tetanus toxin (5 ;Lg/ml). Single-channel activitycould be observed with the voltage of the cis compartment at -50 mV. When the voltage was switched to +50 mV, there was a slow rise inmacroscopic conductance (seen at a higher current-gain here than in the main part of the figure). Note the much higher activity at pH 4.0 than atpH 7.0 (see Inset to a). Also note that single-channel conductances were about one-third as great here at pH 4.0 than at pH 7.0 (see Inset to a

and text).

s-sI ..

LIGHT

C H2N

0- 50

Biophysics: Hoch et aL

A T

pA

Proc. NatL. Acad2 Sci. USA 82 (1985)

I20 pA

20 sec120 pAj204sNec

a

t | 4 Secl10 pA1 ~~~~~~~~4sec

FIG. 3. Effects of cis and trans pH on the rate of conductance rise produced by the heavy chain of botulinum neurotoxin. The figureillustrates that the molecule is most active in the presence of a pH gradient (cis, low; trans, neutral). (a) Both cis and trans solutions started atpH 7.0 and the potential across the membrane was clamped at 50 mV (cis positive). The salt solution in this experiment was 0.2 M NaCl/5 mMCaCl2/5 mM dimethylglutaric acid/O.1 mM EDTA/5 mM dithiothreitol. (The presence of dithiothreitol was not required.) Prior to the start ofthe record, heavy chain of type B botulinum neurotoxin was added to the cis compartment to a concentration of 2 ,ug/ml. There was virtually nochange in membrane conductance (current). About 10 min later, the cis pH was lowered to 4.6 with dimethylglutaric acid (arrow); within 30 secthere began to occur a large increase in conductance. The intact type B botulinum toxin (2 ,ug/ml) was ineffective under the same conditions.(Inset) Single-channel activity induced by the heavy chain of type B botulinum neurotoxin (2 ,ug/ml) at symmetric pH 7.0. The voltage of the ciscompartment was +80 mV. The salt solutions were the same as in Fig. 2. (b) Same conditions as in a, except that the solutions were atsymmetric pH 4.0 instead of 7.0. Again there was no activity for >15 min after protein addition. However, about 1 min after a pH gradient wasestablished across the membrane, in this case by raising the trans pH to 6.5 with Hepes (arrow), the conductance began to increase. The effectseen in this record was not as large as in a. In general, we obtained larger effects with the heavy chain of botulinum toxin when starting atsymmetric high pH and lowering the pH of the cis side, than when starting at symmetric low pH and raising the pH of the trans side. (Inset)Single channel activity at symmetric pH 4.0. The voltage of the cis compartment was +70 mV. All other conditions were the same as in the Insetto a.

conductances; in 1 M KCl, the predominant conductancelevels were 45 pS for heavy chain of tetanus toxin and forfragment B and 15 pS for heavy chain from botulinum neuro-toxin. Within seconds after lowering the pH of the cis com-partment to 4.5, there was a tremendous increase in channelactivity, resulting in a rapid rise in macroscopic membraneconductance (Figs. 2a and 3a).At a symmetric pH of 4.0, there was more single-channel

activity than at a symmetric pH of 7.0; this was especiallytrue for fragments obtained from tetanus toxin (Figs. 2b and3b). In 1 M KCl, the predominant single-channel conduc-tances were 15 pS for heavy chain of tetanus toxin and forfragment B and 20 pS for heavy chain of botulinum neuro-toxin. This channel activity, which led to a slow rise in mac-roscopic conductance, was dramatically enhanced when thepH of the trans compartment was raised to 6.5, leading to a

several hundred-fold increase in the rate of conductance rise(Figs. 2b and 3b).These results show that fragments containing the amino

terminus of heavy chains from clostridial neurotoxins formchannels in lipid bilayer membranes, confirming in part anearlier report (16). These fragments are more active on mem-branes separating solutions at low symmetric pH (4.0 to 5.0)than at neutral symmetric pH (7.0), and they are dramatical-ly more active on membranes separating a pH gradient (cis,pH 4.0-5.0; trans, pH 6.0-7.0). The intact tetanus toxin mol-ecule behaved similarly to its fragments, particularly withrespect to formation of pH-dependent channels (Fig. 2). Un-der the conditions used in this study, the intact botulinumneurotoxin type B molecule did not form pH-dependentchannels.The light chains of botulinum and tetanus toxins, which

are presumed to be the pharmacologically active fragments,were tested for channel-forming activity. Whether tested atsymmetric low pH (4.0), symmetric neutral pH (7.0), or inthe presence of a pH gradient (cis, pH 4.0; trans, pH 7.0),the light chains were devoid of channel-forming activity.These findings indicate that the channel-forming proper-

ties of clostridial neurotoxins and their fragments are analo-gous to those previously described for diphtheria toxin andits fragments (18, 19). In addition to functional similarities,there are also structural similarities. For both diphtheria tox-in and tetanus toxin, the entire heavy chain is not essentialfor channel activity. (We have not yet investigated this pos-sibility for botulinum neurotoxin.) The carboxyl terminus ofthe heavy chain, which is presumably involved in receptor

recognition, could be removed, and the remaining amino-ter-minal end of the heavy chain still retained full channel-form-ing activity. Thus, fragment B, an incomplete form of teta-nus toxin, and CRM45, an incomplete form of diphtheriatoxin, both form channels in artificial membranes (Fig. 2;ref. 18).There is a recent report (22) that tetanus toxin forms chan-

nels in planar lipid bilayers at a symmetric pH of 7.0 and thatthese channels are formed only when mixed gangliosides are

present in the membrane. As described above, we have ob-served similar low levels of toxin-induced channel activity inasolectin membranes at a symmetric pH of 7.0. This activitywas not affected by the presence of ganglioside GTlb (aso-lectin to ganglioside ratio 9:1), a putative receptor for tetanustoxin (23).

Voltage-Gating of Channels. Channels formed by tetanustoxin and its fragment B showed voltage-dependent behaviorremarkably similar to that seen with channels formed bydiphtheria toxin (18, 19). At symmetric low pH, positivevoltages applied to the cis compartment opened channelsand negative voltages closed them; the rate of opening andclosing increased with the absolute magnitude of the voltage.In the presence of a pH gradient (cis, 4.0; trans, 7.0), therewas significant opening of channels at zero voltage, and larg-er negative voltages were required to close them.The voltage dependence of channels formed by the heavy

chain of botulinum neurotoxin was investigated mainly in thepresence of a pH gradient. Under these conditions the chan-nels opened with both positive and negative voltages. How-ever, in the presence of concentration gradients (10:1)formed with glucosamine chloride or potassium glutathio-nate, the channels could be closed with large negative volt-ages applied to the cis compartment.

Size of Toxin-Induced Channels. An important questionthat must be addressed before postulating that channels pro-vide a tunnel for their respective active fragments is whetherthe channels are large enough to accommodate movement ofpeptide chains. Earlier work (18) with multilamellar vesiclesindicated that the channels formed by the B45 fragment ofdiphtheria toxin allowed the nonelectrolytes cyclodextrinand polyethylene glycol 1500 to pass through, which suggest-ed a diameter of at least 18 A. A channel of this size is largeenough to accommodate the extended light chain of the diph-theria toxin molecule (18).From reversal-potential measurements made in the pres-

ence of gradients formed with salts of large anions or cat-

t 2 pA

bR.VW-

1694 Biophysics: Hoch et al.

Proc. Natl. Acad. Sci. USA 82 (1985) 1695

Table 1. Reversal potentials for 10:1 concentration gradients under various pH conditions

mV

K+Cl Glucosamine'Cl K+NAD-

pH 3.5 cis = pH 4.0; pH 5.5 pH 3.5 cis = pH 4.0; pH 5.5 pH 3.9(symmetric) trans = pH 7.0 (symmetric) (symmetric) trans = pH 7.0 (symmetric) (symmetric)

Diphtheria toxin -22 +40 -42 -20 +39Ideal -51 (Cl- +51 (K+) -45 (Cl-) -45 (Cl) +46 (K+)

Tetanus toxin fragment B +25 +8Ideal +51 (K+) -45 (Cl)

Botulinum toxin heavy chain +30 -12Ideal +51 (K+) - 45 (Cl)

Open hole (diffusion potential) 0 0 0 -25 -25 -25 +32

K+Cl- and glucosamine+Cl- (Sigma) gradients were 1 M vs. 0.1 M; K+NAD- (Sigma, Grade III-S) gradients were 0.7 M vs. 0.07 M.Diphtheria toxin measurements were done at the symmetric pHs indicated. Tetanus toxin fragment B and botulinum toxin heavy chainexperiments were done in the presence of a pH gradient [cis = 4.0; trans = 7.0]. Positive potentials indicate cation selectivity; negativepotentials indicate anion selectivity. (All measurements were within ± 1 mV of the recorded entries.) The "ideal" rows indicate the potentialsfor a membrane exclusively permeable to K+ or Cl-. Thus, for example, if the diphtheria toxin entry under K+NAD- had read +46 (the idealvalue) instead of +39, this would have indicated that NAD- did not go through the toxin-induced channel. For K+NAD- and K+Cl-, the idealK+ value was determined on membranes treated with nonactin (1 ,ug/ml), a cyclic antibiotic that is a carrier for univalent cations. Forglucosamine+Cl- and K+Cl-, the ideal Cl- value was determined with a Ag/AgCl electrode. The "open hole" row represents the diffusionpotential measured across the hole in the partition with no membrane present. The diffusion potential is independent of pH. Thus the entriesunder K+NAD- show that NAD- is restricted somewhat by the diphtheria toxin-induced channel (+39) as compared to an open hole (+32)but is certainly not completely excluded (+46). The channels formed by tetanus toxin and the heavy chain of botulinum toxin are morepermeable to glucosamine than is the channel formed by diphtheria toxin, suggesting that the former channels may be larger.

ions, we confirmed with planar lipid bilayers that the diph-theria toxin channel is indeed large (Table 1). The ion selec-tivity of the diphtheria toxin channel is a sensitive functionof pH, with the ratio of K+ to Cl- permeability switchingfrom 1:3 at pH 3.5 to 15:1 at pH 5.5. These values were ob-tained by applying the Goldman-Hodgkin-Katz equation tothe data in Table 1. The permeability of a channel dependson both the size and charge of the permeant ion. The possi-ble complications that charge effects can have on studies todeduce channel size were emphasized in experiments withglucosamine, a cation whose diameter is 7-8 A. At pH 3.5,where the diphtheria toxin channel prefers anions, glucos-amine was almost totally excluded. At a pH of 5.5, where thechannel prefers cations, the ratio of glucosamine to Cl- per-meability was comparable to the value in free solution, indi-cating that the channel did not discriminate between the twoions (Table 1). These findings show that charge is a criticaldeterminant in the selectivity properties of the diphtheriatoxin channel. The data (Table 1) also show that the diphthe-ria toxin channel is sufficiently large to allow penetration byNAD (12-16 A), a result that is compatible with the initialsizing experiments done on vesicles and nonelectrolytes(18).Over the pH range 3.5-7.0, channels formed by tetanus

toxin, fragment B, and the heavy chains from tetanus toxinand botulinum neurotoxin preferred cations, though onlyweakly (Table 1). Their permeability to glucosamine, thelargest cation tested, was even greater than that shown bythe diphtheria toxin channel (Table 1). Thus, these channels,too, are very large. The size of the channels formed by thethree toxins are compatible with their proposed function astunnel proteins to allow the passage of fully extended activefragments.

DISCUSSIONThere is strong evidence that the active fragment of diphthe-ria toxin enters the cytosol from acidic vesicles (13, 14), andthere is suggestive evidence that the same is true for botuli-num neurotoxin (15) and tetanus toxin (16). The results re-ported in the present study are in complete accord with thisview and, furthermore, suggest an explanation for the acidic-vesicle requirement. The data indicate that a portion of theheavy chains from all three toxins interacts with lipid bilayer

membranes to form channels and that this activity is particu-larly marked in the presence of a pH gradient comparable tothat found across intracellular acidic-vesicle membranes. Inaddition, channel formation is enhanced in the presence of avoltage gradient that may be equivalent to that which isfound across acidic-vesicle membranes.There are two possible conclusions that could be drawn

based on the data. The first is that the pH-dependent andvoltage-dependent channels formed by the toxins are epi-phenomena that accompany the interaction of toxin frag-ments with the membrane. Indeed, recent experiments inwhich the light chain of diphtheria toxin was labeled by anintramembranous photoreactive probe appear to indicatethat the light chain does not traverse the membrane throughthe channel formed by the heavy chain (24). This conclusionis predicated on the assumption that the photoreactive probeused in these experiments did not have access to the interiorof the channel. However, the channel wall of the toxin maynot present an impenetrable barrier to the reactive group ofthe probe, just as the wall of the monazomycin channel doesnot prevent alkyl chains from passing through it (25). Dis-missing these channels forces one to assume that there issome other as yet unexplained mechanism that accounts fortoxin translocation. Alternatively, the data can be interpret-ed to mean that the pH-dependent channels are tunnel pro-teins through which the active fragments pass. This conclu-sion does not require one to invoke additional membrane-related mechanisms; to the contrary, it is entirely compatiblewith our ion-permeability experiments which suggest thatthe pH-dependent channels are large enough to allow trans-location of toxin fragments. As previously suggested (18),the pH gradient across acidic vesicles can also act as a driv-ing force for the translocation of the light chain through thechannel into the cytosol, by causing an unfolding of thatchain at the lower vesicular pH and its subsequent refoldingat cytosolic pH.

It is important to stress the presumed role of the pH-de-pendent channels in the mechanisms of action of the toxins.The data in this paper do not suggest that these channels areformed in the plasma membrane or that pH-dependent chan-nels cause depolarization of excitable membranes. The factthat channel activity is pronounced only when the toxins (orrelevant fragments) are in a low pH environment indicatesthat channel formation cannot occur under physiological

Biophysics: Hoch et aL

Proc. Natl. Acad Sci. USA 82 (1985)

conditions at the cell membrane. A more plausible conclu-sion is that the pH-dependent channels are formed in acidi-fied vesicles and the role of these channels is to permit trans-location of the pharmacologically active fragments into thecytosol.The findings and conclusions on transmembrane move-

ment of toxins may be important to the general issue of pro-tein translocation across membranes. There are two basicmodels that have been proposed for this process, one ofwhich can be called the channel model and the other ofwhich can be called a non-channel model. A well known ex-ample of the former is Blobel and Dobberstein's signal hy-pothesis (26). In this model the hydrophobic leader peptide,or signal sequence, recruits several membrane proteins that,together with the signal sequence, form a transmnembranechannel through which a protein can penetrate (or in the caseof integral membrane proteins insert into) the endoplasmicreticulum membrane.Prominent examples of non-channel models are the di-

rect-transfer model of Von Heijne and Blomberg (27) andWickner (28) and the helical hairpin hypothesis of Engelmanand Steitz (29). In these models, the hydrophobic contribu-tions of the leader sequence and subsequent amino acids canovercome the energy barrier associated with insertingcharged and polar residues into a low dielectric constant me-dium, and this in turn leads to direct extrusion of secretedproteins through (and the insertion of integral proteins into)the lipid bilayer.

Proponents of the various models point out that the exis-tence of channels that allow for the passage of proteins hasnever been directly demonstrated. In this context, we sug-gest that the toxin-induced channels we have described mayprovide a good model for endogenous proteins that form tun-nels for the passage of other proteins or polypeptides. Wepoint out, however, that tunnel-forming proteins may not becharacteristic of all multicomponent toxins that have activefragments. For example, we have not found conditions un-der which abrin or ricin form channels in lipid bilayer mem-branes. This suggests that tunnel proteins such as thoseformed by botulinum, tetanus, and diphtheria toxins mayrepresent only one of several mechanisms by which proteinsachieve translocations.

This work was supported in part by National Institutes of HealthGrants T32GM7288, GM29210, NS15409, and NS17742 and by U.S.Army Medical Research and Development Command Contracts 17-83-C-3034 and 17-82-C-2005.

1. Gill, D. M. (1978) in Bacterial Toxins and Cell Membranes, ed.Jeljaszewicz, J. T. & Wadstrom, T. (Academic, London), pp.291-332.

2. Simpson, L. L. (1981) Pharmacol. Rev. 33, 155-188.3. Robinson, J. P. & Nash, J. H. (1982) Mol. Cell. Biochem. 48,

33-44.4. Greenfield, L., Bjorn, M. J., Horn, G., Fong, D., Buck,

G. A., Collier, R. J. & Kaplan, D. A. (1983) Proc. Natl. Acad.Sci. USA 80, 6853-6857.

5. Kaczorek, M., Delpeyroux, F., Chenciner, N., Streeck, R. E.,Murphy, J. R., Boquet, P. & Tiollais, P. (1983) Science 221,855-858.

6. Ittelson, T. R. & Gill, D. M. (1973) Nature (London) 242, 330-332.

7. Zanen, J., Muyldermans, G. & Beugnier, N. (1976) FEBS Lett.66, 261-263.

8. Falmagne, P., Capiau, C., Zanen, J., Kayser, G. & Ruys-schaert, J.-M. (1982) Toxicon 20, 243-246.

9. Chung, D. W. & Collier, R. J. (1977) Biochim. Biophys. Acta483, 248-257.

10. Morris, N. P., Consiglio, E., Kohn, L. D., Habig, W. H., Har-degree, M. C. & Helting, T. B. (1980) J. Biol. Chem. 255,6071-6076.

11. Goldberg, R. L., Costa, T., Habig, W. H., Kohn, L. D. &Hardegree, M. C. (1981) Mol. Pharmacol. 20, 565-570.

12. Simpson, L. L. (1984) J. Pharmacol. Exp. Ther. 229, 182-187.13. Leppla, S. H., Dorland, R. B. & Middlebrook, J. L. (1980) J.

Biol. Chem. 255, 2247-2250.14. Sandvig, K. & Olsnes, S. (1980) J. Cell Biol. 87, 828-832.15. Simpson, L. L. (1983) J. Pharmacol. Exp. Ther. 5, 546-552.16. Boquet, P. & Duflot, E. (1982) Proc. Natl. Acad. Sci. USA 79,

7614-7618.17. Dolly, J. O., Black, J., Williams, R. & Melling, J. (1984) Na-

ture (London) 307, 457-460.18. Kagan, B. L., Finkelstein, A. & Colombini, M. (1981) Proc.

Natl. Acad. Sci. USA 78, 4950-4954.19. Donovan, J. J., Simon, M., Draper, R. & Montal, M. (1981)

Proc. Natl. Acad. Sci. USA 78, 172-176.20. Montal, M. (1974) Methods Enzymol. 32, 545-554.21. Kagawa, Y. & Racker, E. (1971) J. Biol. Chem. 246, 5477-

5487.22. Borochov-Neori, H., Yavin, E. & Montal, M. (1984) Biophys.

J. 45, 83-85.23. Van Heyningen, W. E. (1963) J. Gen. Microbiol. 31, 375-387.24. Zalman, L. S. & Wisnieski, B. J. (1984) Proc. Natl. Acad. Sci.

USA 81, 3341-3345.25. Heyer, E. J., Muller, R. U. & Finkelstein, A. (1976) J. Gen.

Physiol. 67, 703-729.26. Blobel, G. & Dobberstein, B. (1975) J. Cell Biol. 67, 835-851.27. Von Heijne, G. & Blomberg, C. (1979) Eur. J. Biochem. 97,

175-181.28. Wickner, W. (1979) Annu. Rev. Biochem. 48, 23-45.29. Engelman, D. M. & Steitz, T. A. (1981) Cell 23, 411-422.

1696 Biophysics: Hoch et aL