Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Salmonella species translocate virulence effector proteins from the bacterial cytoplasm into mammalian host cells by means of a type III secretion apparatus, encoded by the pathogenicity island-1 (SPI-1). Little is known about the assembly and structure of this secretion apparatus, but the InvG protein is essential and could be an outer membrane secretion channel for the effector proteins. We observed that in recombinant Escherichia coli, the yield of InvG was enhanced by co-expression of InvH, and showed that mutation of invH decreased the level of InvG in wild-type Salmonella typhimurium. In E. coli, InvG alone was able to form an SDS-resistant multimer, but InvG localization to the outer membrane was dependent upon InvH, a lipoprotein itself located in the outer membrane, and no other SPI-1 specific protein. InvG targeted to the outer membrane by InvH became accessible to extracellular protease. InvG and InvH did not, however, appear to form a stable complex. Electron microscopy of InvG membrane protein purified from E. coli revealed that it forms an oligomeric ring-like structure with inner and outer diameters, 7 nm and 15 nm respectively.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Salmonella typhimurium invades the intestinal epithelium, causing severe gastroenteritis in humans. When bacteria contact the host cell, they synthesize transient extracellular organelles called invasomes, subvert host signal transduction pathways and induce cytoskeletal changes and membrane ruffles that surround and internalize the pathogen (Finlay and Falkow, 1990; Francis et al., 1992; Ginocchio et al., 1994; Takeuchi, 1967). The invasion process, encompassing cell entry and subversion, is dependent on a type III secretion apparatus encoded by the Salmonella pathogenicity island-1 (SPI-1) (reviewed in Galán, 1996). This machinery secretes several virulence effector proteins including InvJ, SpaO, SipA, SipB, SipC, SipD, SptP and AvrA and may also translocate a subset of these into the host cell cytosol (Hueck et al., 1995; Collazo and Galán, 1997a; Hardt and Galán, 1997; Fu and Galán, 1998). These effector proteins may stimulate host signal transduction pathways or regulate the secretion and translocation of other effector proteins (Collazo and Galán, 1997b). Similar type III protein secretion systems are conserved among human and animal pathogens, e.g Yersinia enterocolitica, Shigella flexneri and Chlamydia psittaci and plant pathogens such as Pseudomonas syringae and Xanthomonas campestris (Van Gijsegem et al., 1993; Hsia et al., 1997; Lee, 1997), and the components are homologous to proteins of the Salmonella flagellar secretion apparatus and the flagellar hook basal body (Dreyfus et al., 1993; Lee, 1997).
Although the genetics of these systems is established, there is little understanding of the type III secretion mechanism by which Salmonella effector proteins cross the bacterial cell membranes and are translocated into the eukaryotic cell. Based on primary sequence analyses and secondary structure predictions, a subset of the SPI-1 proteins, InvA, InvC, SpaQ, SpaR, SpaS, SpaP, PrgH and PrgK, could together mediate transport across the cytoplasmic membrane (reviewed in Groisman and Ochman, 1993; Collazo and Galán, 1997b), and, similarly, the amino acid sequences of InvG and InvH suggest that these proteins could be localized to the outer membrane and may effect exit from the bacterial cell (Kaniga et al., 1994; Pegues et al., 1995). Periplasmic intermediates have not been identified in type III secretion systems, but as yet there are no obvious candidates within the putative type III machinery that could mediate transport across the periplasmic space or determine interaction between the inner and outer membrane components of the secretion machinery. Furthermore, there is little information on how the complete secretion apparatus is assembled. To examine the secretion process in Salmonella, we expressed InvG in laboratory Escherichia coli and established that InvG is an oligomeric, ring-like, putative outer membrane channel, comparable to recently analysed multimers involved in the assembly of f1 filamentous phage protein and the Pseudomonas aeruginosa pilus, as well as the type II secretion of virulence-associated proteases and lipases by P. aeruginosa, and the type III secretion of Yersinia Yops (Drake et al., 1997; Koster et al., 1997; Linderoth et al., 1997). In so doing, we identified a specific role for a lipoprotein, InvH, in localizing InvG to the outer membrane and demonstrated that InvH is the only SPI-1 encoded protein required for localization.
Yield of InvG is increased in recombinant E. coli by co-expression of InvH
The invG gene was amplified by PCR from S. typhimurium SJW1103 and cloned downstream of the T7 promoter of plasmid pAR3040 (Studier and Moffat, 1986). The resulting recombinant plasmid, pAMCG1, was used to express InvG in E. coli C41(DE3) (Miroux and Walker, 1996), but IPTG induction of mid-exponential-phase cells gave only low levels of the expected 62 kDa protein on in vivo [35S]-Met radiolabelling (Fig. 1A). To determine whether the yield of InvG protein was affected by the presence of other SPI-1 gene products, the InvG expression plasmid was co-transformed with SPI-1 genes from the inv and spa regions that encode putative structural components or substrates of the type III translocator. These genes were placed under T7 control on a compatible plasmid vector so that they could be co-induced with InvG. The presence of InvH (expressed from the recombinant plasmid pAMCH2) increased between two- and fivefold the amount of [35S]-InvG seen using autoradiography (Fig. 1A). Purified InvG from both cells expressing the protein alone or with InvH was N-terminal sequenced, confirming the identity of the protein. Identical sequences of SEKIP were obtained, suggesting processing by leader peptidase between Ser-24 and Ser-25.
The level of S. typhimurium InvG is reduced by invH mutation
To assess whether the effect of InvH on InvG was evident in the native S. typhimurium, InvG was assayed in isogenic wild-type S. typhimurium and a transposon mutant lacking the invH gene product, using antisera prepared against the purified recombinant E. coli InvG. The InvG antibodies recognized a 62 kDa protein in the strain S. typhimurium TNP-5, but not in the invG transposon mutant, and InvG levels were reduced two- to threefold in the invH ::TnphoA mutant (Fig. 1B). In contrast, mutation of the unrelated gene pagC did not cause any decrease in InvG levels.
InvH is an outer membrane lipoprotein
The amino acid sequence of the InvH protein shows at residues 13–16 a consensus sequence, L–X–G/A–C (Braun and Wu, 1994), which suggested that the protein might be cleaved by lipoprotein signal peptidase and modified by fatty acids, a type of lipid modification that serves to anchor proteins in the inner or outer bacterial membrane (reviewed in Pugsley, 1993). When [3H]-palmitate was supplied to E. coli cells during induction of T7 expression, the InvH protein incorporated the isotope (Fig. 2A), as was the case with the known and unrelated flagellar lipoprotein FlgH (Schoenhals and Macnab, 1996), which was cloned and expressed under the control of the T7 promoter to create the plasmid pAMCH1.
In agreement with Altmeyer et al. (1993), more than 90% of InvH was found in the membrane fraction of recombinant E. coli BL21(DE3) cells after cell fractionation. To determine whether the protein was localized in the inner or outer membrane, membrane fractions of cells containing radiolabelled InvH were subjected to sucrose density centrifugation. After equilibration of the gradient, the InvH protein localized to fractions containing 1.5–2.0 M sucrose, characteristic of an outer membrane protein (as demonstrated by co-migration of the OmpA outer membrane marker) (Fig. 2B). The minor peak of InvH at the very bottom of the gradient represents membranes of increased density due to overexpression or poor resuspension of membranes prepared for fractionation. In contrast, NADH oxidase activity localized between 1.1–1.3 M sucrose at the top of the gradient, as would be expected of this inner membrane marker.
To determine whether InvH formed a multimer in the membrane, the protein was solubilized from the E. coli cell membranes with non-ionic detergents and subjected to gel filtration chromatography on a Superdex 200 HR10/30 column. InvH elution from the column indicated a mass of less than 60 kDa, which is consistent with the molecular weight of a single 15 kDa InvH molecule plus the average mass of an octyl-POE (C8En) detergent micelle of 30 kDa (Fig. 2C).
Protease treatment was used to gain information regarding the membrane topology of InvH. Proteinase K treatment of E. coli BL21(DE3)recA cells expressing InvH did not alter cellular levels of full-length InvH, but when whole cells were made permeable by resuspension in isotonic sucrose/EDTA solutions the InvH protein was degraded (Fig. 2D). This would be compatible with exposure of a periplasmic domain to the protease.
InvH, but no other SPI-1 gene product, is necessary for localization of InvG to the outer membrane
The InvG primary sequence suggests that it might be an OM protein (Kaniga et al., 1994). As InvH increased InvG yield, we examined InvG localization in recombinant E. coli with the InvH protein, which we showed above (Fig. 2B) to be itself in the outer membrane. Membrane preparations of E. coli C41(DE3) cell cultures were centrifuged onto 1.0–2.0 M sucrose density gradients and fractionated. When expressed in the presence of InvH, the InvG protein co-migrated with OmpA, an outer membrane marker, detected in fractions containing 1.5–2.0 M sucrose (Fig. 3A).
Examination of the fractions by SDS–PAGE and immunoblotting revealed that InvG expressed in the absence of InvH was located predominantly in fractions containing 1.3–1.5 M sucrose (Fig. 3B), co-localizing with the inner membrane marker protein. Once again, in the presence of InvH, InvG was in fractions of 1.5–2.0 M sucrose, co-migrating with the outer membrane protein marker (Fig. 3B). A significant amount of InvG localized to the bottom of the sucrose gradient, presumably due to the increased density of packed InvG oligomers. It should also be noted that small quantities of InvG were detected in inner membrane fractions, a result that we interpret to represent an intermediate step in processing.
This InvH-dependent localization of InvG to the bacterial cell surface was confirmed by protease accessibility. When whole cells of the E. coli C41(DE3) strain expressing InvG alone were treated with proteinase K, and the proteins analysed by SDS–PAGE and immunoblotting, the InvG protein was seen to be unaffected, but in the cells co-expressing InvH, the InvG protein was degraded by proteinase K and a specific 35 kDa breakdown product was detected (Fig. 3C). Proteolytic treatment of cell lysates of E. coli C41(DE3) expressing only InvG resulted in complete digestion of this protein (not shown). This would be expected if the domains protected in the mature InvG of the outer membrane were accessible in the inner membrane intermediate.
As InvG is dependent upon InvH for its correct localization, we sought to determine whether the two proteins could be readily isolated as a stable complex. Stabilized spheroplasts of E. coli cells expressing InvG in the presence or the absence of InvH were treated with the non-ionic detergent Triton X-100. Whereas InvH could be selectively extracted from the membranes by this treatment, InvG remained in the membrane fraction of both strains (Fig. 4), suggesting that the two proteins do not form a stable complex in solution. Attempts at using disuccinimidyl gluterate (DSG) to chemically cross-link the two proteins in the E. coli cell membrane were also unsuccessful.
Outer membrane InvG forms a ring-like oligomeric structure
Many outer membrane proteins (e.g. porins, PulD) have been shown to form SDS-resistant complexes (Hardie et al., 1996a). Only a minority of the InvG protein was detected in the monomeric (62 kDa) form after SDS–PAGE analysis of wild-type S. typhimurium SJW1103 cell lysates resuspended in SDS loading buffer (Fig. 5A); most of the protein was retained in the stacking gel, suggesting formation of an SDS-stable InvG complex in vivo. Boiling cell lysates in the SDS buffer did, however, dissociate InvG into its monomeric form.
To determine whether InvH interaction is necessary for this potential oligomerization, InvG was examined in wild-type TNP-5 S. typhimurium and the invH ::TnphoA mutant (Fig. 5B). Cell lysates were resuspended in SDS loading buffer with or without boiling and analysed by SDS–PAGE and immunoblotting. The same experiment was performed on the E. coli C41(DE3) expressing InvG with or without recombinant InvH (Fig. 5C). In both S. typhimurium and E. coli, monomeric InvG was only detected after boiling the cell lysates, and co-expression of InvH had no effect on the molecular weight shift. The appearance of monomeric InvG after boiling with SDS could be due to protein-lipopolysaccharide (LPS) linkage rather than oligomerization of InvG. To eliminate this possibility, we confirmed the effect when bacterial LPS structure was severely shortened. Molecular weight shift was equally apparent in wild-type S. typhimurium SL1030 and its rfaH− derivative, a deep rough mutant (Fig. 5D), indicating that LPS structure did not affect the result.
InvG outer membrane protein purified from recombinant E. coli was examined by electron microscopy to establish the nature of the apparent oligomeric InvG structure. Initial images of uranyl-acetate-stained protein indicated that InvG formed ring-like structures, but these clumped together and were difficult to interpret (Fig. 6A). Further images showed small ordered arrays of rosette-like structures containing six or seven rings (Fig. 6B). Only after treatment with alkali (100 mM NaOH), followed by neutralization, were small numbers of individual rings dispersed from the clumps. The negative staining with uranyl acetate revealed that the individual InvG ring-like structure has an approximate diameter of 15 nm with a potential pore inside the ring of 7 nm diameter (n = 55) (Fig. 6C).
Little detail is known of the assembly of the Salmonella SPI-1-encoded type III secretion apparatus, or the mechanism it affords for the translocation of effector proteins out of the pathogen into the host target cell. It appears, however, that the putative outer membrane protein InvG is a central component, as mutations in invG prevent secretion of effector proteins (Collazo et al., 1995; Kaniga et al., 1995a,b; Hardt and Galán, 1997). We noted that the yield of the 62 kDa InvG protein in recombinant E. coli cells was increased by co-expression of the SPI-1-encoded 15 kDa InvH protein, and that invH mutation in wild-type S. typhimurium caused a decrease in the level of InvG. This led us to the finding that InvH is an essential helper protein for InvG, determining its localization to the outer membrane. We suggest that mislocalized InvG could be degraded, so causing the observed decrease in InvG yield in cells lacking InvH. Mutations in invH severely diminish the ability of S. typhimurium to invade epithelial cells in tissue culture and in bovine ileal-loop models (Stone et al., 1992; Altmeyer et al., 1993; Lodge et al., 1995; Watson et al., 1995), and recently an invH mutant was shown to lose the ability to secrete a potential effector protein (Watson et al., 1998). Our findings suggest that these effects could be indirect and result from improperly localized InvG, producing a phenotype similar to an invG mutation.
The localization of InvG to the outer membrane by InvH, which we showed to be a lipoprotein also localized to the outer membrane, appears analogous to the stabilization of Klebsiella oxytoca type II secretion protein PulD (22% identical to InvG) by the outer membrane-bound lipoprotein PulS. PulD levels are reduced in pulS mutants, and heterologous expression of PulD in E. coli was only productive when PulS was co-induced. As was found with InvG–InvH, PulS is required for localization of PulD to the outer membrane (Hardie et al., 1996a), but PulS shows no primary sequence identity to InvH. This localization function shared by PulS and InvH is apparently not conserved in the PulS homologue OutS of Erwinia chrysanthemi, which although required for stabilization of the PulD homologue OutD, does not appear to localize the protein to the outer membrane (Shevchik et al., 1997). Our data suggest that the pathways governing localization of these outer membrane secretion proteins are mechanistically related, albeit directed by distinct small proteins, primarily lipoproteins. The common elements at the core of these pathways remain to be elucidated. As with InvG–InvH, no stable complex was observed between PulD and PulS, but in vivo interactions could be transitory, as indicated by the interaction between PulD and a MalE–PulS chimera (Hardie et al., 1996b). We demonstrate that in whole cells, the InvH protein is partially digested by proteinase K only when the periplasmic face is exposed. This suggests that the protein possesses a periplasmic domain, and that this domain could be responsible for interaction with InvG in the inner membrane. Recent evidence shows that cellular levels of the Yersinia YscC protein (31% identical to InvG) are enhanced by the VirG lipoprotein, and differential solubilization of inner and outer membranes by the ionic detergent sarcosyl indicates that VirG may possibly influence outer membrane targeting (Koster et al., 1997). Again, VirG has no primary sequence identity to InvH or PulS.
Analysis of amino acid similarities suggest that InvG could be a component of an outer membrane channel (Collazo and Galán, 1997b), and we showed that InvG oligomerizes to form an SDS-resistant complex. This oligomerization was not dependent on the InvH protein, and since without InvH, InvG is localized to the inner, not to the outer membrane, it seems that the InvG complex could form in the inner membrane and be localized from there as a preformed multimer. This view is also compatible with the observation that the InvG signal sequence is cleaved when InvG is expressed in the absence of InvH, suggesting that InvH is required only for a late step in the localization of the InvG complex.
Our analysis of the InvG oligomer purified from E. coli co-expressing InvG and InvH showed that InvG forms an oligomeric ring-like structure, compatible with the view that it could be an outer membrane secretion channel. YscC and PulD form oligomeric structures that are SDS stable (Hardie et al., 1996a; Koster et al., 1997), but, unlike these oligomers, InvG multimers can be heat dissociated, as may be the case with the OutD oligomers (Shevchik et al., 1997). In Neisseria gonorrhoeae, formation of a PilQ multimeric structure requires the lipoprotein PilP, which, like the PulS and VirG lipoproteins discussed above, shows no identity to InvH (Drake et al., 1997). Both the filamentous phage protein pIV and X. campestris XpsD can be isolated in oligomeric form (Chen et al., 1996; Kazmierczak et al., 1994), and YscC of Y. enterocolitica, pIV, and P. aeruginosa XcpQ and PilQ have been shown to form membrane-bound ring-like oligomers, with, respectively, outer diameters of 20 nm, 14 nm, 20 nm and 18 nm and inner diameters of 5 nm, 6.5 nm, 9.5 nm and 5.3 nm. These oligomers have been estimated to contain ≈12–14 monomeric subunits (Koster et al., 1997; Linderoth et al., 1997; Bitter et al., 1998), but tetramers of OutD have been identified (Shevchik et al., 1997). InvG ring-like structures were sometimes seen in small rosette-shaped arrays appearing to contain six oligomers around an electron-dense central multimer. Such hexagonal symmetry could indicate an InvG ring composed of six monomers, or, alternatively, six dimers associating to form a dodecamer.
The InvG ring-like multimers were formed in E. coli co-expressing only InvH; other SPI-1 components are therefore not required. This suggests that, unlike the related assembly of flagellar outer membrane rings (Jones and Macnab, 1990; Kubori et al., 1992), InvG does not require other factors to nucleate multimerization. Recent work suggests that membrane-associated components of the type III secretion apparatus may form a needle-like structure, with an outer membrane-associated ring-like structure connected to an inner membrane platform via an analogue of the flagellar rod (Kubori et al., 1998). The InvG structure observed in this work could form the outer membrane ring central to this complex machinery.
The invG, invH and flgH genes were isolated from the wild-type S. typhimurium SJW1103 (Yamaguchi et al., 1984); InvG expression was assayed in 14–16 h shaken cultures of S. typhimurium TNP-5, and in isogenic pagC ::TnphoA, invG ::TnphoA and invH ::TnphoA mutants of TNP-5 (Lodge et al., 1995; E. Morgan, personal communication). The S. typhimurium rfaH mutant SL1060 was derived from SL1030 (Farewell et al., 1991). Strain E. coli BL21(DE3)recA and the BL21(DE3) derivative C41(DE3) (Studier and Moffat, 1986; Miroux and Walker, 1996) were used for expression of proteins under control of the T7 bacteriophage promoter. Bacteria were maintained on Luria broth agar or Bacto Terrific broth agar or in Bacto Terrific broth (Difco) or 2 × TY medium. Chloramphenicol (20 μg ml−1) and/or ampicillin (100 μg ml−1) were used to select for plasmids.
The pACT7 expression plasmid was constructed by subcloning a fragment spanning the T7 promoter and cloning sites from vector pAR3040 (Studier and Moffat, 1986) into the unique ClaI and Sal I sites of pACYC184 (Chang and Cohen, 1978). Synthetic oligonucleotides (Table 1) were used to amplify invG, invH and flgH loci from the S. typhimurium SJW1103 chromosome by PCR (Perkin Elmer GeneAmp apparatus). Amplified fragments were digested with NdeI and BamHI or Bgl II, and ligated into NdeI/BamHI-digested pAR3040 (invG and flgH ). The invH fragment was cloned into pACT7 via NdeI and Bgl II restriction sites.
Table 1. . Oligonucleotides used for PCR (5′–3′).
Solubilization of InvG/InvH
[35S]-methionine-labelled cells were converted to spheroplasts as described (Koronakis et al., 1991). Spheroplasts were stabilized with 20 mM MgCl2 and resuspended in 5% Triton X-100, 20 mM Tris, pH 7.4, 10% glycerol and allowed to incubate at room temperature for 2.5 h. Membrane and soluble fractions were separated at 16 000 × g for 5 min.
Cell fractionation and sucrose density gradients
Whole cells from 1 l cultures were resuspended in 20 mM Tris, pH 7.4, 1 mM Pefabloc SC Protease inhibitor (Pentafarm), and passed through a French pressure cell three times at 150 kPa before removing whole cells by 15 min of centrifugation at 6000 × g, adding 20 mM EDTA to the suspensions and layering the resulting cell lysates on a 5 ml sucrose gradient prepared in a stepwise fashion with layers of 1.0 M, 1.2 M, 1.4 M, 1.6 M, 1.8 M and 2.0 M sucrose (30–55% w/v sucrose) containing 5 mM EDTA. Alternatively, spheroplasts made from cells expressing [35S]-InvH were lysed by freeze–thaw, and membranes were isolated, resuspended in 20 mM EDTA and layered on the sucrose gradient. The gradients were centrifuged for 16 h at 75 000 × g after which 250 μl fractions were removed with a positive displacement pipette and precipitated with 10% TCA.
Protease treatment of whole cells
E. coli C41(DE3) cells were suspended in 20 mM Tris, pH 7.4, and incubated with Proteinase K (50–200 μg ml−1) for 30 min at room temperature before addition of 1.5 mM Pefabloc and 10% TCA. For membrane permeabilization, cells were resuspended in 0.25 M sucrose, 20 mM Tris, pH 7.5, 10 mM EDTA before treatment with proteinase K.
E. coli C41(DE3) carrying pAMCG1 (InvG) and pAMCH2 (InvH) was grown to mid-exponential phase, and T7-controlled expression was induced with 1 mM IPTG. After 3 h induction, cells were spun down and resuspended in 20 mM Tris, pH 7.4, and lysed in a French pressure cell. Pefabloc (1 mM), and 20 mM MgCl2 were added to the cell suspension and membranes were spun down at 48 000 × g for 30 min, resuspended in 20 mM Tris, pH 7.4 in the presence of lysozyme and incubated 16 h at 4°C to dissociate proteins from peptidoglycan. Membranes were then isolated in the presence of 20 mM MgCl2 and washed in 8 M urea before centrifugation at 200 000 × g for 2 h. The preparation was then washed in 20 mM Tris, pH 7.4, 1% Triton X-100, 10 mM EDTA and centrifuged at 48 000 × g for 30 min. This was performed eight times (Fig. 7). Tris/Triton/EDTA buffer was replaced with 0.5% Triton X-100, 10 mM EDTA, 10 mM CAPS buffer, pH 10.0 [TEC buffer (Akiba et al., 1991)] in some preparations.
Molecular weight chromatography
[35S]-InvH was extracted from the membranes prepared from lysed spheroplasts with 0.1% Triton X-100, 20 mM MgCl2, 20 mM Tris, pH 7.4. The Triton X-100 buffer was exchanged with 20 mM Tris-HCl, pH 7.4, 0.3% C8En on an ion exchange column. Samples were loaded on a calibrated Superdex 200 HR10/30 column (Pharmacia) equilibrated with 20 mM Tris, pH 7.4, 0.5% C8En. The protein was eluted at 1 ml min−1 using the same buffer and 1 ml fractions were collected and precipitated with 10% TCA.
SDS–PAGE, protein transfer and immunodetection
Protein samples were suspended in SDS loading buffer with 8 M urea and analysed by SDS 10% or 15% PAGE (Laemmli, 1970). Proteins were visualized using Coomassie staining or electroblotted onto PVDF membrane (ProBlot, Applied Biosystems), blocked with 5% skimmed milk and probed with rabbit polyclonal anti-serum raised against purified InvG raised commercially (Scottish Antibody Production Unit, Law Hospital, Carluke, UK) and used at a 1:8000 dilution. Goat anti-rabbit IgG–HRP conjugate, 1:5000 dilution (Pierce), was used to probe the blot before visualization with SuperSignal substrate (Pierce).
N-terminal sequencing was performed on electroblotted, Coomassie-visualized InvG subjected to Edman degradation (Protein Chemistry Facility, University of Cambridge Department of Biochemistry).
Protein samples were adsorbed onto freshly glow-discharged, carbon-coated copper grids for 2–4 min. Deposition was found to be especially difficult with the InvG protein and was facilitated with alkali treatment. Samples were washed with water alone or with 1% Triton X-100, followed by water before uranyl acetate staining. Grids were examined with a Philips CM100 transmission electron microscope operated at 80 kV and magnification up to 64 000 ×.
We thank Eva Koronakis and Colin Hughes for critical review of this manuscript, and Jeremy Skepper for help with electron microscopy. We also thank R. Macnab, E. Morgan, and J. Stephen for strains. This work was supported by a Wellcome Trust grant (V.K.) and British Marshall and Cambridge Overseas Trust scholarships (A.M.C.).