Molecular Microbioiogy (1992) 6(22), 3277-3282
MicroReview How bacterial protein toxins enter celis: the role of partiai unfoiding in membrane transiocation Erwin London Department of Biochemistry and Cell Biology, State University of New York at Stony Brook, Stony Brook, New York 11794-5215, USA. Summary Bacterial protein toxins translocate across membranes by processes that are stili mysterious. Studies on diphtheria toxin have shown that partial unfolding processes play a major role in toxin membrane insertion and translocation. Similar unfolding behaviour is seen with other bacterial toxins. The lessons gained from this behaviour allow us to propose novel mechanisms for toxin translocation.
Introduction Toxin proteins produced by pathogenic bacteria have remained a subject of intense research for over a century. In the biomedical field they are being developed as therapeutic agents, directed against tumour or HIV-infected cells in the form of conjugates attached to targeting groups such as antibodies. In cell biology, their ability to modulate signal transduction has made them valuable tools for probing the regulation of cell behaviour. In biochemistry, their ability to translocate across membranes is being exploited as a simple system providing clues to the translocation of ordinary cellular proteins. In this review the process whereby protein toxins enter cells is analysed, with an emphasis on diphtheria toxin. For a more detailed discussion the reader is referred to London (1992).
Diphtheria Toxin — Structure and Function Diphtheria toxin (M, 58348) is secreted by Corynebacterium dlphtheriae. It oan be readily cleaved by proteolysis into A and B chains which remain joined by a disulphide bond. Its native structure has just been solved by X-ray crystallography (Choe et ai, 1992). The A ohain Received 2 June, 1992; revised and accepted 5 August, 1992. *For correspondence. Tel. (516) 632 8564; Fax (516) 632 8575.
(=C or catalytic domain) forms the A/-terminal third of the protein. It has the ability to inhibit protein synthesis by inactivation of elongation factor 2. This inactivation is due to ADP-ribosylation of diphthamide, a modified His residue. It has also been proposed that the toxin has a nuclease activity, but this idea is very controversial. The crystal structure shows the A chain to contain both ahelices and p-sheets. The B chain comprises the remaining two thirds of the protein. The C-terminal R portion of the B chain controls toxin-receptor interaction. It is a p-sheet domain (Choe et ai. 1992) that binds the toxin to its cellular receptor, which appears to be a specific small plasma-membrane protein complexed with a second membrane protein (Iwamoto et ai. 1991). The receptor cDNA has very recently been cloned. It encodes a protein of 185 residues with a single transmembrane domain and an extracellular domain corresponding to a heparin-binding epidermal growth factor (Naglich et ai, 1992). The B chain also plays a critical role in the membrane insertion of the toxin. Its W-terminal T domain is an a-helical region (Choe et ai, 1992) containing several long hydrophobic sequences. These sequences tend to be localized within hydrophobic helices which remain largely buried until the conversion of the B chain to a hydrophobic state (see below). The possibility that the B chain has additional biological activities has not been ruled out (Kagan, 1991).
The critical role of endocytosis and low pH After the toxin binds to its receptor, it undergoes receptormediated endocytosis and enters endosomal vacuoles, A large number of genetic, cell biological and biochemical studies have demonstrated that the exposure of the toxin to the acidic lumen of the endosome triggers its membrane penetration. Given the pH at which the toxin inserts into membranes (pH 5-5.5), late endosomes appear the most likely location for membrane translocation. On the other hand, one group has proposed that toxin entry might only be completed after recycling of endocytic vesicles to the cell surface (Hudson ef ai, 1988). It is presumed that the receptor functions to carry the toxin to the endosomes. It is not yet clear whether it plays any additional role in translocation, or whether there are
3278
E London
any endosomal proteins that might aid in the translocation process.
The membrane penetration step Exposure to low pH renders diphtheria toxin hydrophobic, as shown by its binding to micelles of mild non-ionic detergents, and its insertion into model and biological membranes. This hydrophobic behaviour is linked to major low-pH-induced changes in its folding, which include exposure of previously buried proteolysis sites (in the W-terminal half of the B chain) and Trp residues (see London, 1992 for details). The conformational change at low pH can be best summarized as a partial unfolding event in which tertiary structure is disrupted, but in which secondary structure remains largely intact (Zhao and London, 1986), This corresponds to the behaviour of proteins in the partly unfolded molten globule conformation which has now been recognized to exist for many proteins. It appears that the low-pH-induced exposure of the hydrophobic regions within the B chain, which results from this partial unfolding, is the central event triggenng insertion. Low-pH-induced exposure of hydrophobic regions in the B chain is supported by the pH dependence of the behaviour of truncated molecules composed solely of the B chain (McGill etai. 1989), Low-pH-induced partial unfolding probably largely involves alterations of electrostatic forces due to protonation of His. Asp and Glu residues. Protonation could result in formation of buried charges, breakdown of salt bridges, and the strengthening of repulsions owing to increased local or global positive charge, all of which favour unfolding. Local increases in hydrophobicity owing to the loss of charge on anionic residues at low pH is likely to be another important factor in toxin hydrophobicity. In this context, it had been noted that there is a clustering of acidic residues adjacent to the hydrophobic stretches within the B domain (Kieleczawa et ai, 1990). More specifically, the crystal structure of the toxin suggests that the protonation of the acidic residues at the tips of pairs of hydrophobic helices In the T domain renders these tips sufficiently hydrophobic to initiate membrane insertion (Choe efa/., 1992), Low pH can also induce changes within the isolated A chain that are remarkably similar to those within whole toxin, as they involve unfolding linked to increased hydrophobicity and membrane insertion (Zhao and London, 1988). The ability of the isolated A chain to insert is surprising since unlike the B chain it has no sequences that appear to be hydrophobic. Furthermore, A chain unfolding may be quite different from that of the B chain, involving opening of a hinge that bisects its structural lobes (Choe et ai, 1992), Another important difference
between A chain and whole-toxin behaviour is that the changes within the A ohain are readily reversible. This suggests that its transient unfolding and insertion could have an important role in translocation. It should be noted that diphtheria toxin is not the only bacterial toxin in which low-pH-induced partial unfolding is closely linked to membrane insertion. Similar behaviour has recently been observed for Pseudomonas exotoxin A (Jiang and London, 1990) and coiicins (Merrill ef ai. 1990; van der Goot et ai. 1991). In addition, such behaviour has now been observed for the SecA protein, a component of the translocation machinery in E. coli (Ulbrandtefa/,, 1992).
The structure of membrane-inserted toxin — an unsolved mystery We know relatively little about the structure of toxins in their membrane-inserted state. Analysis of the conformation of membrane-inserted diphtheria toxin is complicated by the fact that it can be in one of several conformations. As illustrated in Fig. 1, at lower temperatures membraneinserted diphtheria toxin takes on conformations in which the A chain is folded (L' and R'), and at higher temperatures conformations in which the A chain is at least partially unfolded (L" and R") (Jiang et ai. 1991a), In addition, there is a change in toxin conformation after toxin that has inserted into model membranes at low pH {i,e., in the L state) is returned to neutral pH. After pH neutralization (i.e., in the R state) the toxin remains membrane bound, but is inserted into the membrane to a much lesser degree. Determination of the detailed structures of these different conformations would go a long way towards explaining the translocation process. Ordinarily, crystallography would provide the details of tertiary structure. However, the crystal structures of toxins in their soluble native state seem to provide only limited clues as to their structure within membranes. This is presumably because the structural rearrangements accompanying membrane insertion are substantial. Nevertheless, from the crystal structure of diphtheria toxin it has been possible to identify several hydrophobic helices in the T domain that are likely to take on a transmembranous orientation upon insertion (Choe etai, 1992). On the other hand, it appears that hydrophobic transmembrane a-helices cannot represent the whole story of toxin membrane insertion. The diphtheria toxin A chain and Pseudomonas exotoxin sequences do not show the presence of hydrophobic helices and yet exhibit hydrophobic behaviour. What structure could such molecules have when membrane-inserted? One possibility is that other helices are more hydrophobic than believed. This is possible because hydropathy scales that predict transmembrane helices are generally
Toxin membrane translocation
Fig. 1. Illustration of the conformation of membrane-inserted diphtheria toxin under different conditions. N=native state, L=low-pH conformation, R=contofmation of low-pH-treated toxin after pH is neutralized (reversed). Under conditions in which a domain is partly unfolded it is shown wilh an irregular boundary. This diagram is not meant to illustrate the exact relationship of the membrane-inserted A and B domains. Reprinted with permission from Jiang etal. (1991a), copyright American Chemical Society.
not adjusted for pH, At low pH, protonation of Asp and Glu should render sequences containing them more hydrophobic. In agreement with this idea, the unfolding of diphtheria toxin at tow pH does result in more hydrophobic behaviour than unfolding at high temperature or at high pH (Kieleczawa ef ai, 1990), Furthermore, as noted above, the anionic residues adjacent to the most hydrophobic sequences in diphtheria toxin have suggested that Asp and Glu protonation could strongly modulate hydrophobicity. Nevertheless, adjustment of hydropathy scales for pH does not suggest that additional transmembrane helices exist at low pH (unpublished observations). Alternatively, a toxin could insert with some transmembrane helices that are less hydrophobic than those found in ordinary membrane proteins. Since the maintenance of bilayer integrity upon toxin insertion may not be important, it is possible that the interactions of toxin and lipid involve some contact of polar amino acid residues with hydrophobic acyl chains. A related possibility is that the toxin forms oligomers in membranes. In oligomers many transmembrane helices could be involved in internal proteiri-protein interactions and so would not have to be very hydrophobic (Papini ef ai, 1987). In the same way, contacts" of the toxin helices with a cellular transmembrane protein could substitute for contact with lipid. It should be noted that the self-association of toxin molecules within the membrane has been poorly characterized. Preliminary studies show that membraneinserted toxin has a strong tendency to oligomerize (London, 1992). If the toxin can form oligomers of different sizes and pore properties, as occurs in the case of complement proteins (Malinski and Nelsestuen, 1989), its behaviour could be quite dependent on experimental conditions, and this might explain some of the variations in the behaviour of membrane-inserted toxin seen from study to study. Another possibility for the structure of membrane-
3279
inserted toxin is that some sequences form transmembrane p-sheets. The crystal structure of porins shows that transmembrane p-sheet barrels exist (Weiss etai, 1991). Furthermore, there is one toxin, staphylococcal a-toxin, that does appear to be p-sheet-rich when membraneinserted (Tobkes etai. 1985), Transmembrane p-sheets may have often eluded detection up to now because they are very hard to detect from sequence data, being shorter than transmembrane helices (10 versus 20 residues) and having a pattern of alternating hydrophobic and hydrophilic residues that is easily obscured. In the case of diphtheria toxin transmembrane p-sheets might play a role in translocation of the A chain. In this regard, it is interesting that the opening of the A domain hinge mentioned above could result in the formation of an extended sheet structure. Furthermore, inspection of the sequences encompassed by the p-sheets of the A chain show that several contain hydrophobic residues in alternating positions. In particular, this pattern overlaps p-sheets at positions 75-85, 88-96, 132-140, and 159-168. Therefore, diphtheria toxin could form a hybrid structure with both transmembrane T domain a-helices and A chain p-sheets. A final possibility is that some new type of structure might be present. For example, there could be a hydrophobic surface formed by the spatial clustering of residues that are not contiguous in the primary sequence, analogous to the hydrophobic core of soluble proteins. Under conditions of partial unfolding the hydrophobic core of a toxin might remain largely intact and thus form a lipid interacting site at the tertiary structure level.
The membrane translocation step Information on how diphtheria toxin may translocate across the lipid bilayer has come from several types of studies. The analysis of the effects of low pH on model membrane-inserted toxin has led to the proposal of various translocation mechanisms. One model came from observations that show that whole toxin and isolated B chain both can induce pore formation in membranes at low pH. This raised the possibility that the A chain passes into the cytoplasm via this pore, or that the pore is formed by a discarded hydrophobic B chain wrapper which acts to shield the membrane-inserted A chain from contact with the bilayer (Kagan etai. 1981; Misler, 1984). However, it is still not clear whether the pore formed by the toxin is large enough to allow translocation of the A chain. Another possibility comes from studies demonstrating that the A chain will insert into lipid bilayers. This has led to the proposal that there is transmembrane insertion of the A chain with a partly surrounding B chain serving to limit the degree of A chain contact with the bilayer (Papini et ai, 1987; Zhao and London, 1988). The B chain may also serve to properly orient the inserted A chain. Anothe'
HLA- like cleft Fig. 2. The HLA-like iris model for cholera toxin and heat-labile toxin. A. Schematic illustration of the 'Side' view of the crystal structure of heatlabile toxin from Sixma e( a/. 1991. B. 'Top' view of the structure of heat labile toxin B subunits and the conformational change that would occur upon iris opening.
possible mechanism for translocation is that the toxin, which can destabilize and fuse bilayers, may escape endosomes by inducing membrane lysis. This seems unlikely in view of the limited size of the lesions formed by toxin in model membrane vesicles (Jiang ef ai, 1991 b). Several studies have examined toxin-induced cytotoxicity to infer information about translocation. In some cases it has been shown that specific deletions greatly reduce cytotoxicity, implicating the altered regions in translocation. Environmental effects on translocation, such as transmembrane gradients of pH and membrane potential have also been examined via cytotoxicity. However, such experiments are sometimes too indirect to determine whether the translocation step itself is involved. A more direct approach has been to use low-pH treatment to artificially induce membrane translocation. Studies by Olsnes, Sandvig and colleagues on artificially induced translocation across plasma membranes have revealed that: (i) a sub-population of receptor bound toxin molecules can translocate to the cytoplasm, (ii) A chains with polypeptide extensions at their W-terminus can be translocated, and (iii) cleavage of the peptide bond linking the A and B chains is necessary for translocation and an additional cleavage of a few amino acids from the C-terminal of the A chain may further promote translocation (Moskaug et ai, 1989), They have also found that upon insertion into plasma membranes the disulphide bond linking A and B chains reaches the cytoplasm, and that
the C-terminal region of the B chain is protected from protease digestion. Analysis of the protected regions has been used to formulate a preliminary model of B chain topography (Moskaug ef ai, 1991) which differs somewhat from that predicted by the crystal structure, Translocation has also been studied directly in model membrane vesicles. Exposure of toxin trapped within the lumen of such vesicles to low pH induces the insertion of the trapped toxin into the membrane and its exposure to the external solution (Gonzalez and Wisnieski, 1988; Jiang etai. 1991b). Interestingly, the exposure appears to be greater for the high-temperature state in which the A chain is unfolded. Reduction of the disulphide linking the A and B chains completes translocation, resulting in release of the A chain into the external solution (Jiang et al., 1991b). Unlike exposure to the external face of the vesicles, the degree of release is similar for toxin in its low- and high-temperature states. The explanation of the different dependencies of exposure and release upon conformation probably arises from the fact that some of the 100-250 toxin molecules trapped in each vesicle in these experiments form non-specific, protein-sized pores. A single large pore in a vesicle could be sufficient to allow non-specific release of many A chains that would not othenAfise penetrate the membrane. An unresolved question is whether a single toxin molecule trapped within a vesicle would form such nonspecific pores or translocate the A chain by a more specific process. Until this and the number of toxin molecules within endosomes under physiological conditions are more directly determined, it will be difficult to specify the exact roles of pore formation and A chain unfolding in translocation.
A molecular mechanism for cholera toxin translocation: the HLA-like iris model Two key ideas coming from diphtheria toxin behaviour are that the translocating polypeptide will be at least partly unfolded during the translocation process, and that there will be major rearrangements of the native tertiary structure during translocation. Using these ideas, I would like to speculate about the mechanism for translocation for a case where the crystal structure provides some tantalizing clues, namely for cholera toxin and its close relative, the heat-labile toxin of E coli. The crystal structure of heat-labile toxin shows that these toxins contain five B subunits forming a ring that surrounds a single extended A2 chain which is attached to a globular Al domain (Fig. 2A) (Sixma et ai. 1991). It has been found that these toxins bind to plasma membranes via interaction with ganglioside GM, and act by ADP-ribosylation of Gg proteins. Thus, it had been thought that these toxins directly penetrate the plasma membrane to reach their targets. Some
Toxin membrane translccation studies of cholera toxin interaction with ganglioside-containing model membranes have suggested that the B subunits remain at the surface at the membrane, such that the A subunit penetrates the membrane by itself (Ribi ef ai. 1988), The opposing suggestion that the unfolded A subunit would pass through a pore formed by membraneinserted B chains (Gill, 1976) is not supported by such studies. However, there is growing evidence that the translocation process is more complex than simple penetration of the A subunit directly into the plasma membrane. Cholera toxin action shows a lag-time often associated with a dependence of translocation upon endocytosis and may well require endocytosis in order to act (Lencer et ai. 1992), The cytoplasmic ARF protein which aids cholera toxin-catalysed ADP-ribosylation has been found to be concentrated at the Golgi apparatus (Serafini ef ai. 1991), not the plasma membrane, and is believed to play a role in endoplasmic reticulum-Golgi trafficking (Balch ef ai.. 1992). Along this line it is fascinating that the A subunits of these toxins have at their C-terminus the signal for transport of tumenal proteins to the endoplasmic reticulum, the KDEL sequence (Chaudhary ef ai. 1990), A KDEL-like sequence has already been shown to play an important role in Pseudomonas exotoxin action (Chaudhary et ai. 1990). In addition, the cholera and heat-labile toxin A subunits do not seem to possess classical transmembrane hydrophobic a-helices. Therefore, it is quite possible that cholera and heat-labile toxin translocation is more complicated than just a simple plasma membrane insertion of the A subunit in its native conformation. There are several features in heat-labile toxin crystal structure that are suggestive of how A subunit translocation might occur. The ring of B subunits, thought to sit on the membrane surface upon GMi binding, has a smooth cylindrical shape with an outer wall largely formed by a series of |i-sheets, thus resembling the structure of membrane-inserted ponn (Weiss ef ai, 1991). The length of the B subunit cylinder is sufficient to span at least most of the hydrophobic core of a membrane if it were to insert. In addition, the spacial relationship of the B subunits to one another shows that they form a structure that looks like an iris (Fig, 2B), A sliding action between B subunits could result in the opening of a large pore (Fig. 2B). The strange orientation of the side chains of Arg and Lys residues that line the interior wall formed by the central a-helices of the B subunits is also suggestive of a structure that might open up. In the native state they are all pressed against the helices, lying parallel to the helix axis, and are reminiscent of the ribs of a closed umbrella. Therefore, it is tempting to propose that these toxins undergo a major conformational change, linked to B subunit insertion, which results in the opening of the iris and passage of the A subunit through the pore thus formed.
3281
There is one startling feature of this model that increases its potential significance. The interior of the open iris would have five clefts, each bounded by a psheet outer wall and by two (x-helical lateral walls (Fig. 2B). This corresponds to a structure that is known to bind peptide chains In an unfolded conformation, namely the deft of the human lymphocyte antigen (HLA) protein (Bjorkman ef ai, 1987). The HLA protein presents antigens in the form of peptide chains held within its cleft. The fact that in the HLA protein antigen chains are held in an extended conformation (Madden etai, 1991) and that the HLA deft structure is about the length that would span the hydrophobic core of a membrane suggests that a modification of this type of deft could be adapted to protein translocation. I would speculate that in cholera and heatlabile toxin much of the A subunit polypeptide 'moves' across the membrane along the analogous 8 subunit clefts in an unfolded form. The binding and release of the A subunit from the clefts may be regulated by the exact distance between B subunits. It is interesting that a similar B subunit structure is seen with verotoxin-1, although it is only large enough to span half a membrane by itself (Stein ef ai, 1992). Could other cases of protein translocation also involve a similar mechanism? It is certainly premature to consider this model as anything other than a spur to further experimentation. Perhaps its most important implication is that exploring the structure and function of protein toxins will be an even richer source of information than previously realized. We are in an exciting period of research on these fascinating proteins.
References Balch, W.E. ef ai {1992) J Biol Chem 267:13053-13061. Bjorkman, P.J. etai (1987) Wafure329: 512-518. Chaudhary, V.K. ef ai (1990) Proc NatI Acad Sci USA 87:
308-312. Choe, S. etai (1992) Nature357: 216-221. Gill, D.M. (1976) Biochemistry 15: 1242-1248. Gonzalez, J.E., and Wisnieski, B.J, (1988) J Biol Chem 263:
15257-15259, Hudson, T.H. et ai (1988) J Biot Chem 263: 4773-4781. Iwamoto, R. ef ai (1991) J Biol Chem 266: 20463-20469. Jiang. J.X., and London, E. (1990) J Biol Chem 265: 8636-8641, Jiang. J. X, etai (1991a) Biochemistry30: 3857-3864. Jiang, J.X, et ai (1991 b) J Biol Chem 266: 24003-24010. Kagan, B.L. (1991) Biochim Biophys Acta ^069•. 145-150. Kagan, B.L. ef ai (1981) Proc Natt Acad Sci USA 78: 4950-4954. Kieleczawa, J. ef ai (1990) Arch Biochem Biophys 282:
214-220. Lencer, W.I. ef ai (1992) J Cell Biol 117:1197-1209. London, E. (1992) Biochim Biophys Acta ^•\ 13: 25-5). Madden, D.R. etai {^99^) Nature353: 321-325. Malinski, J.A., and Nelsestuen, G.L. (1989) Biochemistry 28:
61-70.
3282
E. London
Merrill, A.R, efa/. (1990) Biochemistry 29: 5829-5836. McGill, S. etai (1989) EMBOJ 8: 2843-2848. Misler, S. (1984) Biophys J 45:107-109. Moskaug, J.O. etai {^ 989) J Biol Chem 264:15709-15713. Moskaug, J,0. etai (^ 99:) J Biol Chem 266: 2652-2659. Naglich, J.G. efa/.(1992) Ce//69:1051-1061. Papini, E, e/a/. (1987) Eur J e/ochem 169: 637-644 . Ribi, H.O. etai (1988) Sc/ence239:1272-1276. Serafini, T. etal. (1991) Ce//57: 239-253. Sixma, T.K. etai (1991) Wafure351: 371-377,
Stein, P.E. etal. (1992) /Vafure 355: 748-760. Tobkes, N. etai (1985) Biochemistry 24: 1915-1920 , Ulbrandt, N.D. etai (1992) JS/o/Cfiem 267:15184-15192. van der Goot, F.G, efa/. (1991) /Vafure 354: 4 0 8 ^ 1 0 . Weiss, M.S. etai (1991) Sc/ence 254: 1627-1630. Zhao, J.-M., and London. E. (1988) J Biol Chem 263: 15369-15377. Zhao, J.-M., and London, E. (1986) Proc NatI Acad Sci USA 83: 2002-2006.
MicroReview How bacterial protein toxins enter celis: the role of partiai unfoiding in membrane transiocation Erwin London Department of Biochemistry and Cell Biology, State University of New York at Stony Brook, Stony Brook, New York 11794-5215, USA. Summary Bacterial protein toxins translocate across membranes by processes that are stili mysterious. Studies on diphtheria toxin have shown that partial unfolding processes play a major role in toxin membrane insertion and translocation. Similar unfolding behaviour is seen with other bacterial toxins. The lessons gained from this behaviour allow us to propose novel mechanisms for toxin translocation.
Introduction Toxin proteins produced by pathogenic bacteria have remained a subject of intense research for over a century. In the biomedical field they are being developed as therapeutic agents, directed against tumour or HIV-infected cells in the form of conjugates attached to targeting groups such as antibodies. In cell biology, their ability to modulate signal transduction has made them valuable tools for probing the regulation of cell behaviour. In biochemistry, their ability to translocate across membranes is being exploited as a simple system providing clues to the translocation of ordinary cellular proteins. In this review the process whereby protein toxins enter cells is analysed, with an emphasis on diphtheria toxin. For a more detailed discussion the reader is referred to London (1992).
Diphtheria Toxin — Structure and Function Diphtheria toxin (M, 58348) is secreted by Corynebacterium dlphtheriae. It oan be readily cleaved by proteolysis into A and B chains which remain joined by a disulphide bond. Its native structure has just been solved by X-ray crystallography (Choe et ai, 1992). The A ohain Received 2 June, 1992; revised and accepted 5 August, 1992. *For correspondence. Tel. (516) 632 8564; Fax (516) 632 8575.
(=C or catalytic domain) forms the A/-terminal third of the protein. It has the ability to inhibit protein synthesis by inactivation of elongation factor 2. This inactivation is due to ADP-ribosylation of diphthamide, a modified His residue. It has also been proposed that the toxin has a nuclease activity, but this idea is very controversial. The crystal structure shows the A chain to contain both ahelices and p-sheets. The B chain comprises the remaining two thirds of the protein. The C-terminal R portion of the B chain controls toxin-receptor interaction. It is a p-sheet domain (Choe et ai. 1992) that binds the toxin to its cellular receptor, which appears to be a specific small plasma-membrane protein complexed with a second membrane protein (Iwamoto et ai. 1991). The receptor cDNA has very recently been cloned. It encodes a protein of 185 residues with a single transmembrane domain and an extracellular domain corresponding to a heparin-binding epidermal growth factor (Naglich et ai, 1992). The B chain also plays a critical role in the membrane insertion of the toxin. Its W-terminal T domain is an a-helical region (Choe et ai, 1992) containing several long hydrophobic sequences. These sequences tend to be localized within hydrophobic helices which remain largely buried until the conversion of the B chain to a hydrophobic state (see below). The possibility that the B chain has additional biological activities has not been ruled out (Kagan, 1991).
The critical role of endocytosis and low pH After the toxin binds to its receptor, it undergoes receptormediated endocytosis and enters endosomal vacuoles, A large number of genetic, cell biological and biochemical studies have demonstrated that the exposure of the toxin to the acidic lumen of the endosome triggers its membrane penetration. Given the pH at which the toxin inserts into membranes (pH 5-5.5), late endosomes appear the most likely location for membrane translocation. On the other hand, one group has proposed that toxin entry might only be completed after recycling of endocytic vesicles to the cell surface (Hudson ef ai, 1988). It is presumed that the receptor functions to carry the toxin to the endosomes. It is not yet clear whether it plays any additional role in translocation, or whether there are
3278
E London
any endosomal proteins that might aid in the translocation process.
The membrane penetration step Exposure to low pH renders diphtheria toxin hydrophobic, as shown by its binding to micelles of mild non-ionic detergents, and its insertion into model and biological membranes. This hydrophobic behaviour is linked to major low-pH-induced changes in its folding, which include exposure of previously buried proteolysis sites (in the W-terminal half of the B chain) and Trp residues (see London, 1992 for details). The conformational change at low pH can be best summarized as a partial unfolding event in which tertiary structure is disrupted, but in which secondary structure remains largely intact (Zhao and London, 1986), This corresponds to the behaviour of proteins in the partly unfolded molten globule conformation which has now been recognized to exist for many proteins. It appears that the low-pH-induced exposure of the hydrophobic regions within the B chain, which results from this partial unfolding, is the central event triggenng insertion. Low-pH-induced exposure of hydrophobic regions in the B chain is supported by the pH dependence of the behaviour of truncated molecules composed solely of the B chain (McGill etai. 1989), Low-pH-induced partial unfolding probably largely involves alterations of electrostatic forces due to protonation of His. Asp and Glu residues. Protonation could result in formation of buried charges, breakdown of salt bridges, and the strengthening of repulsions owing to increased local or global positive charge, all of which favour unfolding. Local increases in hydrophobicity owing to the loss of charge on anionic residues at low pH is likely to be another important factor in toxin hydrophobicity. In this context, it had been noted that there is a clustering of acidic residues adjacent to the hydrophobic stretches within the B domain (Kieleczawa et ai, 1990). More specifically, the crystal structure of the toxin suggests that the protonation of the acidic residues at the tips of pairs of hydrophobic helices In the T domain renders these tips sufficiently hydrophobic to initiate membrane insertion (Choe efa/., 1992), Low pH can also induce changes within the isolated A chain that are remarkably similar to those within whole toxin, as they involve unfolding linked to increased hydrophobicity and membrane insertion (Zhao and London, 1988). The ability of the isolated A chain to insert is surprising since unlike the B chain it has no sequences that appear to be hydrophobic. Furthermore, A chain unfolding may be quite different from that of the B chain, involving opening of a hinge that bisects its structural lobes (Choe et ai, 1992), Another important difference
between A chain and whole-toxin behaviour is that the changes within the A ohain are readily reversible. This suggests that its transient unfolding and insertion could have an important role in translocation. It should be noted that diphtheria toxin is not the only bacterial toxin in which low-pH-induced partial unfolding is closely linked to membrane insertion. Similar behaviour has recently been observed for Pseudomonas exotoxin A (Jiang and London, 1990) and coiicins (Merrill ef ai. 1990; van der Goot et ai. 1991). In addition, such behaviour has now been observed for the SecA protein, a component of the translocation machinery in E. coli (Ulbrandtefa/,, 1992).
The structure of membrane-inserted toxin — an unsolved mystery We know relatively little about the structure of toxins in their membrane-inserted state. Analysis of the conformation of membrane-inserted diphtheria toxin is complicated by the fact that it can be in one of several conformations. As illustrated in Fig. 1, at lower temperatures membraneinserted diphtheria toxin takes on conformations in which the A chain is folded (L' and R'), and at higher temperatures conformations in which the A chain is at least partially unfolded (L" and R") (Jiang et ai. 1991a), In addition, there is a change in toxin conformation after toxin that has inserted into model membranes at low pH {i,e., in the L state) is returned to neutral pH. After pH neutralization (i.e., in the R state) the toxin remains membrane bound, but is inserted into the membrane to a much lesser degree. Determination of the detailed structures of these different conformations would go a long way towards explaining the translocation process. Ordinarily, crystallography would provide the details of tertiary structure. However, the crystal structures of toxins in their soluble native state seem to provide only limited clues as to their structure within membranes. This is presumably because the structural rearrangements accompanying membrane insertion are substantial. Nevertheless, from the crystal structure of diphtheria toxin it has been possible to identify several hydrophobic helices in the T domain that are likely to take on a transmembranous orientation upon insertion (Choe etai, 1992). On the other hand, it appears that hydrophobic transmembrane a-helices cannot represent the whole story of toxin membrane insertion. The diphtheria toxin A chain and Pseudomonas exotoxin sequences do not show the presence of hydrophobic helices and yet exhibit hydrophobic behaviour. What structure could such molecules have when membrane-inserted? One possibility is that other helices are more hydrophobic than believed. This is possible because hydropathy scales that predict transmembrane helices are generally
Toxin membrane translocation
Fig. 1. Illustration of the conformation of membrane-inserted diphtheria toxin under different conditions. N=native state, L=low-pH conformation, R=contofmation of low-pH-treated toxin after pH is neutralized (reversed). Under conditions in which a domain is partly unfolded it is shown wilh an irregular boundary. This diagram is not meant to illustrate the exact relationship of the membrane-inserted A and B domains. Reprinted with permission from Jiang etal. (1991a), copyright American Chemical Society.
not adjusted for pH, At low pH, protonation of Asp and Glu should render sequences containing them more hydrophobic. In agreement with this idea, the unfolding of diphtheria toxin at tow pH does result in more hydrophobic behaviour than unfolding at high temperature or at high pH (Kieleczawa ef ai, 1990), Furthermore, as noted above, the anionic residues adjacent to the most hydrophobic sequences in diphtheria toxin have suggested that Asp and Glu protonation could strongly modulate hydrophobicity. Nevertheless, adjustment of hydropathy scales for pH does not suggest that additional transmembrane helices exist at low pH (unpublished observations). Alternatively, a toxin could insert with some transmembrane helices that are less hydrophobic than those found in ordinary membrane proteins. Since the maintenance of bilayer integrity upon toxin insertion may not be important, it is possible that the interactions of toxin and lipid involve some contact of polar amino acid residues with hydrophobic acyl chains. A related possibility is that the toxin forms oligomers in membranes. In oligomers many transmembrane helices could be involved in internal proteiri-protein interactions and so would not have to be very hydrophobic (Papini ef ai, 1987). In the same way, contacts" of the toxin helices with a cellular transmembrane protein could substitute for contact with lipid. It should be noted that the self-association of toxin molecules within the membrane has been poorly characterized. Preliminary studies show that membraneinserted toxin has a strong tendency to oligomerize (London, 1992). If the toxin can form oligomers of different sizes and pore properties, as occurs in the case of complement proteins (Malinski and Nelsestuen, 1989), its behaviour could be quite dependent on experimental conditions, and this might explain some of the variations in the behaviour of membrane-inserted toxin seen from study to study. Another possibility for the structure of membrane-
3279
inserted toxin is that some sequences form transmembrane p-sheets. The crystal structure of porins shows that transmembrane p-sheet barrels exist (Weiss etai, 1991). Furthermore, there is one toxin, staphylococcal a-toxin, that does appear to be p-sheet-rich when membraneinserted (Tobkes etai. 1985), Transmembrane p-sheets may have often eluded detection up to now because they are very hard to detect from sequence data, being shorter than transmembrane helices (10 versus 20 residues) and having a pattern of alternating hydrophobic and hydrophilic residues that is easily obscured. In the case of diphtheria toxin transmembrane p-sheets might play a role in translocation of the A chain. In this regard, it is interesting that the opening of the A domain hinge mentioned above could result in the formation of an extended sheet structure. Furthermore, inspection of the sequences encompassed by the p-sheets of the A chain show that several contain hydrophobic residues in alternating positions. In particular, this pattern overlaps p-sheets at positions 75-85, 88-96, 132-140, and 159-168. Therefore, diphtheria toxin could form a hybrid structure with both transmembrane T domain a-helices and A chain p-sheets. A final possibility is that some new type of structure might be present. For example, there could be a hydrophobic surface formed by the spatial clustering of residues that are not contiguous in the primary sequence, analogous to the hydrophobic core of soluble proteins. Under conditions of partial unfolding the hydrophobic core of a toxin might remain largely intact and thus form a lipid interacting site at the tertiary structure level.
The membrane translocation step Information on how diphtheria toxin may translocate across the lipid bilayer has come from several types of studies. The analysis of the effects of low pH on model membrane-inserted toxin has led to the proposal of various translocation mechanisms. One model came from observations that show that whole toxin and isolated B chain both can induce pore formation in membranes at low pH. This raised the possibility that the A chain passes into the cytoplasm via this pore, or that the pore is formed by a discarded hydrophobic B chain wrapper which acts to shield the membrane-inserted A chain from contact with the bilayer (Kagan etai. 1981; Misler, 1984). However, it is still not clear whether the pore formed by the toxin is large enough to allow translocation of the A chain. Another possibility comes from studies demonstrating that the A chain will insert into lipid bilayers. This has led to the proposal that there is transmembrane insertion of the A chain with a partly surrounding B chain serving to limit the degree of A chain contact with the bilayer (Papini et ai, 1987; Zhao and London, 1988). The B chain may also serve to properly orient the inserted A chain. Anothe'
HLA- like cleft Fig. 2. The HLA-like iris model for cholera toxin and heat-labile toxin. A. Schematic illustration of the 'Side' view of the crystal structure of heatlabile toxin from Sixma e( a/. 1991. B. 'Top' view of the structure of heat labile toxin B subunits and the conformational change that would occur upon iris opening.
possible mechanism for translocation is that the toxin, which can destabilize and fuse bilayers, may escape endosomes by inducing membrane lysis. This seems unlikely in view of the limited size of the lesions formed by toxin in model membrane vesicles (Jiang ef ai, 1991 b). Several studies have examined toxin-induced cytotoxicity to infer information about translocation. In some cases it has been shown that specific deletions greatly reduce cytotoxicity, implicating the altered regions in translocation. Environmental effects on translocation, such as transmembrane gradients of pH and membrane potential have also been examined via cytotoxicity. However, such experiments are sometimes too indirect to determine whether the translocation step itself is involved. A more direct approach has been to use low-pH treatment to artificially induce membrane translocation. Studies by Olsnes, Sandvig and colleagues on artificially induced translocation across plasma membranes have revealed that: (i) a sub-population of receptor bound toxin molecules can translocate to the cytoplasm, (ii) A chains with polypeptide extensions at their W-terminus can be translocated, and (iii) cleavage of the peptide bond linking the A and B chains is necessary for translocation and an additional cleavage of a few amino acids from the C-terminal of the A chain may further promote translocation (Moskaug et ai, 1989), They have also found that upon insertion into plasma membranes the disulphide bond linking A and B chains reaches the cytoplasm, and that
the C-terminal region of the B chain is protected from protease digestion. Analysis of the protected regions has been used to formulate a preliminary model of B chain topography (Moskaug ef ai, 1991) which differs somewhat from that predicted by the crystal structure, Translocation has also been studied directly in model membrane vesicles. Exposure of toxin trapped within the lumen of such vesicles to low pH induces the insertion of the trapped toxin into the membrane and its exposure to the external solution (Gonzalez and Wisnieski, 1988; Jiang etai. 1991b). Interestingly, the exposure appears to be greater for the high-temperature state in which the A chain is unfolded. Reduction of the disulphide linking the A and B chains completes translocation, resulting in release of the A chain into the external solution (Jiang et al., 1991b). Unlike exposure to the external face of the vesicles, the degree of release is similar for toxin in its low- and high-temperature states. The explanation of the different dependencies of exposure and release upon conformation probably arises from the fact that some of the 100-250 toxin molecules trapped in each vesicle in these experiments form non-specific, protein-sized pores. A single large pore in a vesicle could be sufficient to allow non-specific release of many A chains that would not othenAfise penetrate the membrane. An unresolved question is whether a single toxin molecule trapped within a vesicle would form such nonspecific pores or translocate the A chain by a more specific process. Until this and the number of toxin molecules within endosomes under physiological conditions are more directly determined, it will be difficult to specify the exact roles of pore formation and A chain unfolding in translocation.
A molecular mechanism for cholera toxin translocation: the HLA-like iris model Two key ideas coming from diphtheria toxin behaviour are that the translocating polypeptide will be at least partly unfolded during the translocation process, and that there will be major rearrangements of the native tertiary structure during translocation. Using these ideas, I would like to speculate about the mechanism for translocation for a case where the crystal structure provides some tantalizing clues, namely for cholera toxin and its close relative, the heat-labile toxin of E coli. The crystal structure of heat-labile toxin shows that these toxins contain five B subunits forming a ring that surrounds a single extended A2 chain which is attached to a globular Al domain (Fig. 2A) (Sixma et ai. 1991). It has been found that these toxins bind to plasma membranes via interaction with ganglioside GM, and act by ADP-ribosylation of Gg proteins. Thus, it had been thought that these toxins directly penetrate the plasma membrane to reach their targets. Some
Toxin membrane translccation studies of cholera toxin interaction with ganglioside-containing model membranes have suggested that the B subunits remain at the surface at the membrane, such that the A subunit penetrates the membrane by itself (Ribi ef ai. 1988), The opposing suggestion that the unfolded A subunit would pass through a pore formed by membraneinserted B chains (Gill, 1976) is not supported by such studies. However, there is growing evidence that the translocation process is more complex than simple penetration of the A subunit directly into the plasma membrane. Cholera toxin action shows a lag-time often associated with a dependence of translocation upon endocytosis and may well require endocytosis in order to act (Lencer et ai. 1992), The cytoplasmic ARF protein which aids cholera toxin-catalysed ADP-ribosylation has been found to be concentrated at the Golgi apparatus (Serafini ef ai. 1991), not the plasma membrane, and is believed to play a role in endoplasmic reticulum-Golgi trafficking (Balch ef ai.. 1992). Along this line it is fascinating that the A subunits of these toxins have at their C-terminus the signal for transport of tumenal proteins to the endoplasmic reticulum, the KDEL sequence (Chaudhary ef ai. 1990), A KDEL-like sequence has already been shown to play an important role in Pseudomonas exotoxin action (Chaudhary et ai. 1990). In addition, the cholera and heat-labile toxin A subunits do not seem to possess classical transmembrane hydrophobic a-helices. Therefore, it is quite possible that cholera and heat-labile toxin translocation is more complicated than just a simple plasma membrane insertion of the A subunit in its native conformation. There are several features in heat-labile toxin crystal structure that are suggestive of how A subunit translocation might occur. The ring of B subunits, thought to sit on the membrane surface upon GMi binding, has a smooth cylindrical shape with an outer wall largely formed by a series of |i-sheets, thus resembling the structure of membrane-inserted ponn (Weiss ef ai, 1991). The length of the B subunit cylinder is sufficient to span at least most of the hydrophobic core of a membrane if it were to insert. In addition, the spacial relationship of the B subunits to one another shows that they form a structure that looks like an iris (Fig, 2B), A sliding action between B subunits could result in the opening of a large pore (Fig. 2B). The strange orientation of the side chains of Arg and Lys residues that line the interior wall formed by the central a-helices of the B subunits is also suggestive of a structure that might open up. In the native state they are all pressed against the helices, lying parallel to the helix axis, and are reminiscent of the ribs of a closed umbrella. Therefore, it is tempting to propose that these toxins undergo a major conformational change, linked to B subunit insertion, which results in the opening of the iris and passage of the A subunit through the pore thus formed.
3281
There is one startling feature of this model that increases its potential significance. The interior of the open iris would have five clefts, each bounded by a psheet outer wall and by two (x-helical lateral walls (Fig. 2B). This corresponds to a structure that is known to bind peptide chains In an unfolded conformation, namely the deft of the human lymphocyte antigen (HLA) protein (Bjorkman ef ai, 1987). The HLA protein presents antigens in the form of peptide chains held within its cleft. The fact that in the HLA protein antigen chains are held in an extended conformation (Madden etai, 1991) and that the HLA deft structure is about the length that would span the hydrophobic core of a membrane suggests that a modification of this type of deft could be adapted to protein translocation. I would speculate that in cholera and heatlabile toxin much of the A subunit polypeptide 'moves' across the membrane along the analogous 8 subunit clefts in an unfolded form. The binding and release of the A subunit from the clefts may be regulated by the exact distance between B subunits. It is interesting that a similar B subunit structure is seen with verotoxin-1, although it is only large enough to span half a membrane by itself (Stein ef ai, 1992). Could other cases of protein translocation also involve a similar mechanism? It is certainly premature to consider this model as anything other than a spur to further experimentation. Perhaps its most important implication is that exploring the structure and function of protein toxins will be an even richer source of information than previously realized. We are in an exciting period of research on these fascinating proteins.
References Balch, W.E. ef ai {1992) J Biol Chem 267:13053-13061. Bjorkman, P.J. etai (1987) Wafure329: 512-518. Chaudhary, V.K. ef ai (1990) Proc NatI Acad Sci USA 87:
308-312. Choe, S. etai (1992) Nature357: 216-221. Gill, D.M. (1976) Biochemistry 15: 1242-1248. Gonzalez, J.E., and Wisnieski, B.J, (1988) J Biol Chem 263:
15257-15259, Hudson, T.H. et ai (1988) J Biot Chem 263: 4773-4781. Iwamoto, R. ef ai (1991) J Biol Chem 266: 20463-20469. Jiang. J.X., and London, E. (1990) J Biol Chem 265: 8636-8641, Jiang. J. X, etai (1991a) Biochemistry30: 3857-3864. Jiang, J.X, et ai (1991 b) J Biol Chem 266: 24003-24010. Kagan, B.L. (1991) Biochim Biophys Acta ^069•. 145-150. Kagan, B.L. ef ai (1981) Proc Natt Acad Sci USA 78: 4950-4954. Kieleczawa, J. ef ai (1990) Arch Biochem Biophys 282:
214-220. Lencer, W.I. ef ai (1992) J Cell Biol 117:1197-1209. London, E. (1992) Biochim Biophys Acta ^•\ 13: 25-5). Madden, D.R. etai {^99^) Nature353: 321-325. Malinski, J.A., and Nelsestuen, G.L. (1989) Biochemistry 28:
61-70.
3282
E. London
Merrill, A.R, efa/. (1990) Biochemistry 29: 5829-5836. McGill, S. etai (1989) EMBOJ 8: 2843-2848. Misler, S. (1984) Biophys J 45:107-109. Moskaug, J.O. etai {^ 989) J Biol Chem 264:15709-15713. Moskaug, J,0. etai (^ 99:) J Biol Chem 266: 2652-2659. Naglich, J.G. efa/.(1992) Ce//69:1051-1061. Papini, E, e/a/. (1987) Eur J e/ochem 169: 637-644 . Ribi, H.O. etai (1988) Sc/ence239:1272-1276. Serafini, T. etal. (1991) Ce//57: 239-253. Sixma, T.K. etai (1991) Wafure351: 371-377,
Stein, P.E. etal. (1992) /Vafure 355: 748-760. Tobkes, N. etai (1985) Biochemistry 24: 1915-1920 , Ulbrandt, N.D. etai (1992) JS/o/Cfiem 267:15184-15192. van der Goot, F.G, efa/. (1991) /Vafure 354: 4 0 8 ^ 1 0 . Weiss, M.S. etai (1991) Sc/ence 254: 1627-1630. Zhao, J.-M., and London. E. (1988) J Biol Chem 263: 15369-15377. Zhao, J.-M., and London, E. (1986) Proc NatI Acad Sci USA 83: 2002-2006.
