The interaction at low pH (≤6) of the 177-residue T domain (residues 202–378) of diphtheria toxin with endosomal or plasma membranes results in translocation of the toxin's catalytic domain (residues 1–185) across the target membrane into the cytosol. Under these same low pH conditions, the T domain also forms channels in planar lipid bilayers (and plasma membranes). In association with this channel formation, the catalytic domain is also translocated across planar bilayers (Oh et al. 1999
), although the channel's role, if any, in this process is unclear. Thus, the T domain contains the entire translocation machinery; no other proteins, or even the receptor-binding portion of the toxin (residues 386–535), are required for translocation. How does the T domain accomplish this? This paper begins to address this question by outlining the T domain's topography in the channel's open state. That is, we have determined which of its regions lie on the cis and trans sides of the membrane, and which, by inference, reside within the membrane.
We used two strategies to map the topography. Both of them used substituted cysteine mutagenesis, followed by linking prosthetic groups at the selected sites. In the first, we chemically linked a synthetic hexahistidine (H6) tag to each site. In so doing, we exploited the phenomenon of a voltage-dependent channel block by the H6 tag. Added to the cis solution, the free synthetic H6 peptide blocks channel conductance at positive voltages; added to the trans solution, it blocks at negative voltages. Hence, if the chemically attached H6 tag blocked the channel at positive voltages, this constituted evidence that the H6 tag, and the residue to which it was linked, was located on the cis side; channel block at negative voltages suggested the opposite location. Interference of H6 tag channel block by micromolar concentrations of trans nickel further confirmed a residue's trans location.
In the second approach, we employed a biotinylation–streptavidin assay. Its rationale was that if a biotinylated residue is located on either the cis or trans side, streptavidin addition to the appropriate solution would likely affect some aspect of channel behavior. We found that streptavidin binding after channel formation generally resulted in an increase in current magnitude and noise; for residue 347, it resulted in channel closure. Furthermore, if a residue ultimately resides on the trans side, its binding to streptavidin before membrane insertion should interfere with channel formation by preventing it from reaching the trans side, which was indeed what we found. The results of both approaches lead us to propose the model for T domain topography shown in , which we now proceed to explicate.
Our model of the distribution of T domain residues in the open channel state. Residues whose positions were determined in this study are shown in black with white numbers.
Evidence for the Model
Both the H6 tag and the biotinylation–streptavidin assay place residues 235, 261, and 267 on the trans side of the membrane ( and ); the inhibition of normal channel formation by preincubating biotinylated 235 mutant with streptavidin supports this conclusion. Given that the T domain's amino terminus is also translocated to the trans solution (Senzel et al. 1998
), we conclude that the entire very polar amino-terminal third of the T domain (residues 202–270) is translocated across the membrane to the trans side. In fact, since the catalytic domain of whole toxin is also translocated (Oh et al. 1999
), all of the toxin's first 270 residues move across the membrane in association with channel formation!
The H6 assay places residues 291, 293, and 294 on the cis side ( and ), implying that the polar 10-residue acidic segment, residues 290–299 (), lies there. (The biotinylation–streptavidin assay confirmed this for residue 291; see the next section for residues 293 and 294.) Connecting the dots, this means that the hydrophobic segment consisting of residues 271–289 traverses the membrane. This segment is slightly longer than the TH5 helix seen in the crystal structure of the water soluble form of the toxin (Choe et al. 1992
; Bennett et al. 1994
). It is tempting to suppose that this hydrophobic segment remains α-helical in the membrane-associated form of the T domain, as it is long enough to traverse the bilayer.
The H6 tag assay places residue 320 on the cis side (), implying that the charged 10-residue segment, residues 318–327 (), lies there. Between the two charged segments (residues 290–299 and 318–327) on the cis side of the membrane is an 18-residue uncharged segment (residues 300–317) corresponding to helices TH6 and TH7 (essentially one helix, TH6-7) in the crystal structure (Choe et al. 1992
; Bennett et al. 1994
). A helical wheel representation of TH6-7 depicts it as amphipathic. We suggest that this segment retains its helical character, with its nonpolar surface dipping into the hydrophobic portion of the bilayer's cis leaflet, while its polar surface contacts this leaflet's headgroups and the nearby water layer. We have no direct experimental evidence to support this suggestion, but it is the most parsimonious model consistent with its connecting two charged segments lying on the cis surface.
It is interesting to compare the topography presented so far with that inferred from the crystal structure of the toxin's water soluble form; in particular, the portion from TH5 to TH6-7. In the crystal structure, this forms an α-helical hairpin. The obvious inference is that this inserts as such into the membrane, with its acidic connecting loop (residues 290–299) facing the trans solution, and with the amino terminus of TH5 (residue 274) and the carboxy terminus of TH6-7 (residue 315) close to the cis surface (). Our data directly contradict this disposition of the TH5 to TH6-7 region. The acidic connecting loop (residues 290–299) cannot be facing the trans solution, since residues 291, 293, and 294 face the cis solution. (Similarly, our placement of residue 267 on the trans side makes it unlikely that residue 274 is close to the cis surface.) Our results are consistent with only TH5 (not both TH5 and TH6-7) spanning the membrane (), with an orientation opposite that in the double dagger model ().
Figure 11 Double dagger model of Choe et al. 1992. See text for discussion of incompatibilities between this model and ours ().
We turn now to the carboxy terminal end (residues 322–378), which actually forms the channel (Silverman et al. 1994a
; Huynh et al. 1997
). In the crystal structure, most of this region (residues 318–376) forms the α-helical hairpin TH8-9 (Choe et al. 1992
; Bennett et al. 1994
). In planar bilayers, pH mutagenesis–titration experiments (Mindell et al. 1994a
,Mindell et al. 1994b
), as well as studies of accessibility of cysteine mutants to sulfhydryl-specific reagents (Huynh et al. 1997
), clearly place residues 326–336 closer to the cis than to the trans side, and residues 348–352 close to the trans side. Our experiments placing residue 320 on the cis side () and residue 347 on the trans side () confirm these earlier findings that most of helix TH8 spans the membrane (though not necessarily as an α-helix; Huynh et al. 1997
). The disposition of helix TH9 is less clear, but residue 359 appears not as close to the trans side as residue 351 (Huynh et al. 1997
). Combining this fact with our present finding (by both assays) that residue 376 is on the cis side ( and ) justifies our having TH9 span the membrane (), although not necessarily as an α-helix.
Validity of the H6 Tag and Biotinylation–Streptavidin Assays
A basic assumption in using these assays to determine T domain topography is that the H6-tagged and biotinylated sites end up on the same side of the membrane as where they normally reside when unencumbered by these labels. There are two reasons for believing this. First, the assays' results are consistent with previous findings. Their placement of residues 235, 261, and 267 on the trans side agrees with there being a site (or sites) between residues 209 and 265 susceptible to trans trypsin (Senzel et al. 1998
). Similarly, the assignments of residues 320 and 376 to the cis side and residue 347 to the trans side are consistent with the previously determined membrane orientation of the TH8-9 region (Mindell et al. 1994a
,Mindell et al. 1994b
; Huynh et al. 1997
Second, the assays were largely in agreement, even though the H6 tag is very hydrophilic and the biotin moiety (plus its linker) rather nonpolar. Only in the region of residues 291–299 is there a possible discrepancy. While residues 291, 293, and 294 appear cis by the H6 assay, the latter two, when biotinylated, do not induce normal channel formation. Furthermore, the cis streptavidin-induced increase in current noise for biotinylated 291 mutant was not reversible. Biotinylated 291 mutant produced a much greater conductance (when cis pH was not raised) after cis TCEP addition, which suggests that biotinylation may decrease channel-forming efficiency. Thus, attaching biotin or H6 peptide may have held these residues on the cis side when they would otherwise have entered or traversed the membrane. However, biotin and H6 peptide did not prevent residues 235, 261, and 267 from traversing the membrane. Moreover, even if residues 291, 293, or 294 are indeed artificially held on the cis side by the H6 peptide, the amino terminus of the T domain is still translocated to the trans side. We showed this by attaching the H6 peptide at residue 293 with the amino-terminal H6 tag also present; the amino-terminal H6 tag closed the channels at negative voltages, while the attached H6 tag at residue 293 closed them at positive voltages ().
Should it turn out that residues 291, 293, and 294 were artifactually held on the cis side by the attached moieties and that they are normally on the trans side in the open channel state, the double dagger model in is still precluded, because of our placement of residue 267 on the trans side. What we would have, instead of the disposition of residues shown in , is that, instead of TH5 being a membrane-spanning segment, TH6-7 would be membrane spanning, with everything upstream of it lying on the trans side.
The failure of the H6 tag at residue 347 to block the channel may result from its being in the channel-forming region; perhaps the H6 tag tethered there lacks the flexibility to reach its binding site with the correct orientation. Alternatively, the H6 tag attached at residue 347 may prevent channel formation, so that the observed conductance actually resulted from a minority of mutant 347 molecules that lacked the attached H6 tag because they failed to react with the maleimidyl-H6 peptide.
Certain aspects of the biotinylation–streptavidin results are puzzling. (a) As noted in results
, the increased current noise produced by streptavidin's binding to biotinylated 235, 261, 267, 291, and 376 mutants is not reflected in increased single-channel current noise. It may result from a streptavidin-induced cooperative interaction among channels, but its nature and mechanism are unknown. (b) It is surprising that the biotin–streptavidin complex formed by trans streptavidin's binding to biotinylated 235, 261, and 267 mutants does not impede the amino-terminal H6 tag's access to the channel's binding site. (c) That trans streptavidin's binding to the biotinylated 347 mutant closes or blocks the channels is not surprising since that residue lies in the channel's ion-conducting pathway (Huynh et al. 1997
), but why do ~20% of the channels remain unaffected? Conceivably, in a subpopulation of the channels, the tethered biotin is locked into a configuration inaccessible to trans streptavidin. (d) Occasionally, cis addition of biotinylated 235 mutant preincubated with streptavidin produced a large burst of conductance corresponding to very rapid insertion of hundreds of channels. Surprisingly, these channels displayed gating indicating that the amino terminus was translocated across the membrane, despite residue 235 being bound by streptavidin. Possibly, this rapid insertion of channels occurred through a transient membrane defect induced by their TH8-9 regions, which allowed the normally translocated region to cross the membrane with its attached streptavidin.
Comparison of Planar Bilayer Results with Results in Other Systems
The T domain topography presented in comes from the results described in this paper and is consistent with earlier cited results on planar lipid bilayers; it pertains exclusively to the open channel state. The topography of all, or parts, of the T domain has also been derived from fluorescence, spin-label, and protease–protection assays on lipid vesicles and cells. In the following brief summary of these results and their comparison to our planar bilayer findings, the reader should keep in mind that the number of open channels (a few thousand at most) in a given planar bilayer represents a minute fraction (<10−10) of the total number of T domain molecules added to the cis compartment, and we know neither how many of these molecules are adsorbed onto the membrane surface, nor their membrane topography. In contrast, experiments on lipid vesicles and cell membranes generally measure the average disposition of all of the T domain molecules, of which an unknown and perhaps a very small fraction are in the open channel state.
The lipid vesicle conclusions agree with ours for the TH8-9 region; namely, it is inserted into the bilayer like a hairpin, with the residues near 349 close to or on the trans side, and residues near 320 and 376 close to or on the cis side (Oh et al. 1996
; Quertenmont et al. 1996
; Wang et al. 1997
; Kachel et al. 1998
). The cell data do not make as strong a statement about the orientation of the TH8-9 region, but they are consistent with the lipid vesicle and planar bilayer results in that this region is protected from cis protease digestion (Moskaug et al. 1991
), and the mutation of either Glu349 or Asp352 to Lys inhibits toxicity and channel formation (Silverman et al. 1994b
; Lanzrein et al. 1997
Protease experiments on lipid vesicles do not agree with our proposed orientation of the TH5-7 region in . In vesicles, the entire TH5-7 region is completely digested by cis proteinase K (Quertenmont et al. 1996
), whereas our model predicts that at least TH5 should be protected from digestion. Protease data on cells, however, are consistent with our orientation of the TH5-7 region, in that residue 299 is exposed on the cis side, whereas helices TH5 and TH6-7 are protected from digestion (Moskaug et al. 1991
). The cell data are also consistent with our placing the loop connecting TH5 and TH6-7 on the cis side (rather than on the trans side as diagrammed in ), in that mutating either Asp290 or Glu292 to Lys does not reduce toxicity or channel formation (Silverman et al. 1994b
). [The mutation of Asp295 to Lys reduces toxicity and channel formation in cells (Falnes et al. 1992
; Silverman et al. 1994b
; Lanzrein et al. 1997
), but given that the mutations to Lys of the nearby Asp290 or Glu292 do not, the effect of the Asp295-to-Lys mutation cannot be interpreted as simply resulting from placing a positive charge in the loop.]
Protease experiments on lipid vesicles agree marginally with our results for the region upstream of TH5. The amino terminal end of the T domain down to approximately residue 249 was protected from digestion by cis proteases, but this represented only 12% of the amount of TH8-9 protected (Quertenmont et al. 1996
). We, on the other hand, place the amino-terminal end down to at least residue 267 on the trans side. Moreover, in the experiments by Quertenmont et al. 1996
, the entire catalytic domain was digested by cis proteases, whereas in planar bilayers this is translocated to the trans side (Oh et al. 1999
) and should therefore have been protected from digestion. In addition to the caveats mentioned at the beginning of this section, we also note that in these experiments the proteases were allowed to act for hours, making their correspondence to the planar bilayer experiments even more problematic. In cells, the amino-terminal end of the T domain was digested by pronase E (Moskaug et al. 1991
), thereby placing this region on the cis side, in disagreement with our findings.
In sum, by picking and choosing, one can select data from vesicle and cell membrane experiments supporting most of the T domain topography proposed in . As noted at the start of this section, however, we feel it is a priori very difficult to relate the results of those experiments to ours of the open-channel state.