Nature Structural Biology9, 729 - 733 (2002)
Published online: 9 September 2002; | doi:10.1038/nsb839
Conversion of a transmembrane to a water-soluble protein complex by a single point mutation
Yulia Tsitrin1, Craig J. Morton2, Catherine El Bez3, Patrick Paumard1, Marie-Claire Velluz1, Marc Adrian3, Jacques Dubochet3, Michael W. Parker2, 5, Salvatore Lanzavecchia4
& F.G. van der Goot1, 5
1 Department of Genetics and Microbiology, University of Geneva, 1 rue Michel Servet, 1211 Geneva, Switzerland.
2 The Biota Structural Biology Laboratory, St. Vincent's Institute of Medical Research, 41 Victoria Parade, Fitzroy, Victoria 3065, Australia.
Proteins exist in one of two generally incompatible states: either membrane associated or soluble. Pore-forming proteins are exceptional because they are synthesized as a water-soluble molecule but end up being located in the membrane — that is, they are nonconstitutive membrane proteins. Here we report the pronounced effect of the single point mutation Y221G of the pore-forming toxin aerolysin. This mutation blocks the hemolytic activity of the toxin but does not affect its initial structure, its ability to bind to cell-surface receptors or its capacity to form heptamers, which constitute the channel-forming unit. The overall structure of the Y221G protein as analyzed by cryo-negative staining EM and three-dimensional reconstruction is remarkably similar to that of the wild type heptamer. The mutant protein forms a mushroom-shaped complex whose stem domain is thought to be within the membrane in the wild type toxin. In contrast to the wild type heptamer, which is a hydrophobic complex, the Y221G heptamer is fully hydrophilic. This point mutation has, therefore, converted a normally membrane-embedded toxin into a soluble complex.
Pore-forming proteins include perforin and pro- or anti-apoptotic proteins of the Bcl-2 family, as well as several bacterial toxins, such as aerolysin. Aerolysin is secreted by Aeromonas hydrophila as a water-soluble precursor proaerolysin1. The mature protein is obtained after proteolytic cleavage at the C-terminus, which occurs outside the bacterial cell. The toxin can then oligomerize into a ring-like heptameric structure, which is amphipathic and inserts into a lipid membrane to form a pore.
Aerolysin is divided into four domains, 1−4 (Fig. 1a). On the basis of a low-resolution (25 Å), three-dimensional reconstruction of the aerolysin transmembrane channel and the crystal structure of proaerolysin2, domain 4 has been proposed to form the transmembrane pore. Interestingly, the upper boundary of domain 4 is delineated by a row of three aromatic residues (repeated seven times in the heptamer), Tyr 298, Phe 410 and Tyr 221, which are reminiscent of the 'aromatic belts' found in other membrane proteins that are thought to anchor and stabilize those proteins in the bilayer3,
4. In addition, these aromatic residues are flanked by two hydrophobic residues, Leu 277 and Leu 196 (Fig. 1a), suggesting that this region could be involved in membrane interactions. Here we analyzed the importance of these aromatic residues by mutating each to a glycine.
Figure 1. Structural effects of the Y221G point mutation.
a, Ribbon diagram of the wild type proaerolysin monomer2. Residues Leu 196, Tyr 298, Phe 410, Tyr 221 and Leu 277 are shown as space-filled models. This figure was drawn using MolScript30. b, Human erythrocytes (0.4% (v/v) packed cells:PBS) were incubated with 1.6 g ml-1 trypsin-activated wild type or mutant aerolysin at 37 °C. The optical density (OD) at 600 nm wavelength was measured as a function of time (1 cm path length). c, Near UV CD spectra of wild type (filled square) and Y221G (empty circles) proaerolysin. Each spectrum is the average of 25 scans. d, Aerolysin overlays31 were performed on baby hamster kidney cell extracts (40 g) using wild type and Y221G proaerolysin (0.14 g ml-1). NCAM (N) and semaphorin 7 (S) were specifically identified6. e, Wild type and mutant proaerolysins (0.2 mg ml-1) in 150 mM NaCl and 20 mM HEPES, pH 7.4, were activated for 10 min with trypsin (5 g ml-1) and analyzed by SDS-PAGE.
Structural effects of the Y221G mutation The Y298G mutation had no effect on the hemolytic activity of the toxin (Fig. 1b). In contrast, mutation F410G delayed hemolysis and the Y221G mutation completely blocked hemolysis. This latter effect may be due to the absence of an aromatic side chain, because the Y221F mutation did not have an inhibitory effect (Fig. 1b). Alternatively, the effect of the Y221G mutation may be due to the special properties of glycine flexibility and adoption of dihedral angles not normally observed for other residues. To investigate whether the inhibitory effect of the Y221G mutation was due to a change in the overall structure, we measured the near UV CD spectrum of Y221G proaerolysin. The spectrum was identical to that of wild type (Fig. 1c), suggesting that the tertiary structure of the protoxin was not affected by the mutation. In addition, the Y221G mutant was still able to bind to target cells5 and specifically recognize glycosylphosphatidyl inositol (GPI)-anchored proteins such as NCAM and semaphorin 7 (previously identified in baby hamster kidney (BHK) cell extracts6) (Fig. 1d), again suggesting correct folding. Even more important, the ability to form heptamers was not affected by these mutations because the oligomerization of F410G and of Y221G was indistinguishable from that of wild type as analyzed by SDS-PAGE (aerolysin heptamers are SDS resistant and can therefore be analyzed by SDS-PAGE; Fig. 1e), in agreement with a previous observation5.
We next analyzed whether the most pronounced of the above-mentioned mutations, Y221G, affected the structure of the heptamers. The near UV CD spectrum of Y221G heptamers differed significantly from that of wild type heptamers (Fig. 2a), indicating that their tertiary structure must differ despite the similarity of their secondary structures (Fig. 2b). The Y221G heptamer showed the same unfolding curve in guanidinium-Cl (data not shown) as the wild type heptamer, indicating that they are equally stable7.
Figure 2. CD analysis of wild type and Y221G heptamers using a, near and b, far UV CD spectroscopy.
Analysis of the Y221G heptamer To further confirm the similarity in structure between the wild type and the Y221G heptamers, the two complexes were analyzed by cryo-negative staining EM, a technique that offers good contrast while maintaining native structure8. The wild type heptamers formed irregular, mostly aggregated structures (data not shown) and micelle-like structures (Fig. 3a) that are not amenable to single particle EM analysis. In contrast, regular structures were observed for the Y221G mutant (Fig. 3a). Single particle analysis and class averaging (Fig. 3b) were performed, followed by a three-dimensional reconstruction (Fig. 3c). The resolution estimated by Fourier shell correlation9 corresponds to 13.5 Å using the sqr(5) / N criterion10,
11 and to 17 Å using the 50% criterion; the reconstruction has been low pass filtered at 13.5 Å. Y221G was found to form a complex of two identical mushroom- or funnel-shaped heptamers, joined generally by their larger bases (Fig. 3c). The funnel axis has a seven-fold symmetry, and a set of 14 two-fold axes are found in the plane where the two heptamers contact each other. Each funnel is composed of seven elongated subunits interwoven as in a wicker basket. The assembly has a maximal diameter of 15 nm and a height of 17.5 nm. The inner diameter of the channel is 2 nm at the tips. The present cryo-negative staining EM analysis and three-dimensional reconstruction give improved data compared with previous studies12. Not only has the resolution significantly increased, but the stem region of the mushroom-shaped complex can be seen and analyzed for the first time. This region was not seen in the previous study, because it was supposedly obscured by the membrane.
Figure 3. Three-dimensional reconstruction of the Y221G heptamer and analysis of its hydrophobicity.
a, Wild type and Y221G mutant heptamers were visualized by cryo-negative staining EM8. b, Class averages of the Y221G complexes were made showing head-to-head dimers of heptamers. c, Three-dimensional reconstruction of the Y221G heptamer. The image shows a dimer of heptamers joined together by their large base. d, Model of the aerolysin heptamer. Orthogonal views of the fit of domains into the channel density are shown. Note that only the C chain is shown. The left hand view is looking from the top of the mushroom-shaped head. Each subunit is indicated by a different color. The head of the mushroom-shaped density is formed by domains 1 and 2 and the stalk by domains 3 and 4. The slightly bulbous base of the image is caused by domain 4 projecting away from the channel lumen. Tyr 221 is highlighted as green space filling models.
Three-dimensional model of the Y221G heptamer We constructed a high-resolution model of the heptamer using docking methods and found a convincing fit into the three-dimensional reconstruction (Fig. 3d). The model predicts that conversion of proaerolysin dimers (as found in the crystal)2 to aerolysin heptamers involves a simple rotation of domain 1 about the long linker connecting domains 1 and 2 and an alteration to the helical twist of the -sheet that runs through the major lobe of the molecule. The domain 1−2 dock seems highly plausible because (i) a large surface area of 832 Å2 is buried in the domain interface, (ii) a three-dimensional profile plot13 shows that all residues are in a favorable environment (in particular, there were 13 potential hydrogen bonds, 1 salt bridge and 50 van der Waals interactions in the domain interface) and (iii) residues that were implicated in oligomerization by mutagenesis (His 107, His 132 and Asp 139)14,
15 are located at this interface. Detailed analysis of interactions involving domains 3 and 4 has not been performed because loss of the propeptide in the conversion of proaerolysin to heptameric aerolysin would cause significant and unpredictable conformational changes in these domains. Altogether, these observations show that the heptamer of the Y221G mutant has a mushroom-like shape very reminiscent of that proposed for the wild type heptamer12 and that of the staphylococcal -hemolysin pore16. According to the current model, residue 221 is located in a subunit interface (Fig. 3d); although, structural rearrangements caused by activation could just as readily make it point into the lipid bilayer.
The observation that heptamers associate into dimers (Fig. 3c) can be readily rationalized. If the two heptamers are docked together using the electron microscopy data as a guide, the region Asp 334-Asn 335-Arg-336 is brought close to the equivalent region in a symmetry-related mate of the other heptamer, causing a double salt bridge to form between the Asp and Arg in each subunit. In addition, hydrogen bonds between the two subunits involving the Asn residue are possible.
The Y221G heptamer is hydrophilic In addition to providing a 13.5 Å resolution three-dimensional model of the Y221G heptamer, the EM analysis showed that these heptamers do not aggregate. This observation is critical because it suggests that these heptamers, in contrast to the wild type heptamers, do not expose hydrophobic surfaces. Hydrophobicity of certain exposed surfaces on the wild type heptameric complex is thought to be the driving force for membrane insertion of aerolysin17, and its absence could explain the inability of Y221G to form channels.
To test this hypothesis, we measured binding of the hydrophobic dye ANS (8-anilino-1-naphthalenesulfonate) upon the triggering of oligomerization by trypsin treatment of proaerolysin. Although binding of ANS was readily observed upon addition of trypsin to the wild type protoxin, ANS fluorescence remained low upon processing of Y221G proaerolysin (Fig. 4a) despite the observation that oligomerization did occur (Fig. 1e, this gel corresponds to the sample at the end of the experiment in Fig. 4a). These observations suggest that the Y221G heptamer was indeed hydrophilic. Next, we analyzed the partitioning properties of the Y221G heptamers in Triton X114 (ref. 18) in comparison with those of wild type monomers and heptamers. As expected, wild type monomers were found in the aqueous phase and wild type heptamers in the detergent phase (Fig. 4b). In marked contrast, both the Y221G monomers and the heptamers were found in the aqueous phase, confirming that Y221G heptamers do behave as soluble proteins.
a, Binding of the hydrophobic dye ANS upon activation and oligomerization of wild type (filled square) and Y221G mutant (open square) was performed as described17. Proaerolysin (0.2 mg ml-1) in 150 mM NaCl and 20 mM HEPES, pH 7.4, was incubated with 100 M of ANS. At the time indicated by an arrow, trypsin was added at a final concentration of 5 g ml-1. At the end of the experiments, samples were analyzed by SDS-PAGE to control that heptamer formation had indeed occurred (Fig. 1e). 'a.u.' represents arbitrary units. b, Samples of trypsin-activated wild type and Y221G mutant aerolysin containing monomers and heptamers were submitted to partitioning in Triton X114 (ref. 18). The protein profiles of the initial sample (Tot.), the aqueous phase (Aq.) and the detergent phase (Det.) were analyzed by SDS-PAGE to detect the monomers and the heptamers.
Conclusions A mutant pore-forming toxin able to heptamerize but unable to form channels has been found in the past, namely the H35R mutant of staphylococcal -hemolysin19. Although the hydrophobicity of the H35R -hemolysin heptamer has not been analyzed, we presume that the complex would be soluble similar to the Y221G aerolysin heptamer. Unfolding of the central loop of -hemolysin into a membrane inserting -barrel was shown not to occur in the H35R heptamer, which would lead to a structure resembling a mushroom without a stem domain. This is in marked contrast with what is observed here for the Y221G heptamer. Similar to the wild type toxin, Y221G forms a mushroom-shaped structure with a stem domain. This observation leads to the remarkable conclusion that a single point mutation (leading to seven mutations in the heptamers) has converted a transmembrane structure — that is, the aerolysin channel-forming complex — into a soluble complex without grossly affecting its structure. The model of the heptamer (Fig. 3d) provides some clues as to why mutating Tyr 221 results in a soluble channel. Previous work has demonstrated that several conformational changes must occur in the conversion to the membrane state: a long loop in domain 3 (indicated by an arrow in Fig. 1a) must move20 and cleavage of the propeptide propagates conformational changes throughout the molecule21. Tyr 221 is located on a strand that leads into the domain 3 loop and likely acts as a pivot point for conformational changes caused by the loop movement. Tyr 221 also directly interacts with the propeptide and removal of its side chain would affect the conformational changes caused by activation of the toxin. Thus, the mutation of Tyr 221 likely inhibits one or more conformational changes that are required to expose the hydrophobic regions that form the membrane-inserted channel.
Methods Circular dichroism. CD spectra of proaerolysin, wild type and the Y221G mutant were measured using 1 cm path length cuvettes at a protein concentration of 0.45 mg ml-1. The spectrum of the buffer was subtracted from that of the protein. Heptamers were prepared from trypsin-activated aerolysin as described22 and then dialyzed against 2 M urea, 150 mM NaCl and 20 mM HEPES, pH 7.4. Urea was added to ensure solubility of wild type heptamers, which tend to aggregate. However urea in the 0−6 M range had no effect on the CD spectra of either wild type or Y221G heptamers, in agreement with the previous observation that urea has no effect on the Trp fluorescence of the heptamer7. Spectra in the near and far UV regions of the heptamers were taken using path lengths of 1 cm and 0.02 cm, respectively, and a protein concentration of 0.36 mg ml-1.
Cryo-negative stain and three-dimensional reconstruction. Wild type and Y221G mutant heptamers in 20 mM HEPES, pH 7.3, were obtained as described22 and analyzed by cryo-negative staining8. Images were collected at 0° tilt of the grid, a condition in which the molecules show strongly preferred orientations. To perform a three-dimensional reconstruction, 9,400 particles were recorded from tilted and untilted micrographs. The out-of-focus range of the images was 700−1,200 nm. A preliminary model was obtained from 638 pairs of views extracted from 0−20° tilt pairs as described23 without imposing any symmetry. All images were then aligned by iterated projection matching24 on the basis of the refining model. D7 symmetry was observed and subsequently imposed during the refinement. Three-dimensional reconstruction was performed by the Radon approach25. From untitled views, 2,500 molecules were selected for the final reconstruction. CTF (Contrast Transfer Function) correction was applied on phases only (phase flipping). The surface shown was chosen so as to enclose a volume corresponding to 680 kDa (14 48.5 kDa).
Generation of a three-dimensional model of the Y221G heptamer. A model of the heptamer was constructed using the crystal structure of proaerolysin2 and the protein-docking program GRAMM26. Previous work had indicated that only small conformational changes, particularly in tertiary interactions, occur in the conversion of proaerolysin to the aerolysin heptamer21,
27; hence, possible structural changes within domains were ignored in the modeling process. Strictly speaking aerolysin consists of two domains: the small lobe (domain 1) and the large lobe (domains 2−4). Hence, the large lobe should be treated as a single domain for docking purposes because it has a core of -sheets that runs through the large lobe, from tip to tip, so that domains 2−4 cannot move independently of each other. The highest ranked solutions from the GRAMM runs were inspected by computer graphics. No sensible docking solutions were found involving either domains 3 or 4. This was not surprising given the likelihood that these domains undergo significant conformational changes on activation of the protoxin and disassembly of the dimer28. A heptamer of domains 1 and 2 was constructed based on the dimensions of the mushroom-shaped head seen in published images of the channel and energy minimized using Discover (Accelrys). The assembly revealed that the C-terminus of domain 1 was near enough to join the N-terminus of domain 2 in the adjacent monomer, providing further confidence in the model. The domain 1−2 heptamer was manually fit into the mushroom-shaped head of the EM images presented here (Fig. 3c). However, superposition of domains 2−4 of the proaerolysin monomer onto domain 2 of the constructed heptamer revealed slight clashes of domains 3 and 4 with each other. This problem was readily resolved by assuming the helical twist of the -sheet running through domains 2−4 could be altered; such flexibility has been observed in the crystal structure of the dimer where crystal contacts cause a partial unwinding in one monomer with respect to the other2. The rebuilt heptamer was automatically fitted into the EM images using Situs29. The close fit between model and map (Fig. 3d) is reflected by a correlation coefficient of 0.78.
Acknowledgments We would like to thank J. Lakey for analysis of the far UV CD spectroscopy curves, and J. Gruenberg and G. von Heijne for critical reading of the manuscript. This work was supported by grants of the Swiss National Science Foundation to J.D. and F.G.v.d.G., and by the National Health and Medical Research Council of Australia to M.W.P. M.W.P. is a NHMRC Senior Principal Research Fellow.
Competing interests statement:
The authors declare that they have no competing financial interests.
g ml-1 trypsin-activated wild type or mutant aerolysin at 37 °C. The optical density (OD) at 600 nm wavelength was measured as a function of time (1 cm path length). c, Near UV CD spectra of wild type (filled square) and Y221G (empty circles) proaerolysin. Each spectrum is the average of 25 scans. d, Aerolysin overlays
2 nm at the tips. The present cryo-negative staining EM analysis and three-dimensional reconstruction give improved data compared with previous studies
chain is shown. The left hand view is looking from the top of the mushroom-shaped head. Each subunit is indicated by a different color. The head of the mushroom-shaped density is formed by domains 1 and 2 and the stalk by domains 3 and 4. The slightly bulbous base of the image is caused by domain 4 projecting away from the channel lumen. Tyr 221 is highlighted as green space filling models.
-sheet that runs through the major lobe of the molecule. The domain 1−2 dock seems highly plausible because (i) a large surface area of 832 Å2 is buried in the domain interface, (ii) a three-dimensional profile plot
48.5 kDa).
