The closed structure of the MscS mechanosensitive channel. Cross-linking of single cysteine mutants.

Mechanosensitive channels must make a large conformational change during the transition from the closed to the open state. The crystal structure of the open form of the Escherichia coli MscS channel was recently solved and depicts a homoheptamer (1). In this study, cross-linking of site-specific cysteine substitutions demonstrates that residues up to 10-33 A apart in the crystal structure readily form disulfide bridges in the closed form and can also be cross-linked by a 10-A linker. Cross-linking between adjacent subunits stabilizes the heptameric form of the channel providing biochemical evidence to support the crystal structure. The data are consistent with the published model (1) in that the membrane domain is highly flexible and that the closed to open transition may involve a significant displacement of transmembrane helices 1 and 2, possibly by as much as 30 A. The data are also consistent with significant flexibility of the cytoplasmic domain.

Mechanosensitive (MS) 1 channels underpin vital sensory processes in higher organisms and are responsible for structural integrity of bacterial cells during the transition from high to low osmolarity (downshock) (2)(3)(4). In the bacterial membrane closed MS channels must maintain the cation impermeability of the membrane, which is an essential requirement for energy transduction. Open MS channels in Escherichia coli exhibit conductances of 0.3-3 nanosiemens (5), consistent with the formation of large pores in the membrane of at least an 11-Å diameter (1,6,7). Clearly MS channels must undergo large conformational changes in the protein structure during the transition from the closed to the open state. The recent crystal structures of the bacterial MS channels, MscL and MscS, have been achieved for the closed and open states, respectively (1,8). Model building, coupled with biochemical analysis of single cysteine mutants, has supported molecular dynamics-based analyses of the structural transitions that may open MscL (6, 7, 9 -11). However, the two proteins are so individual that there is little from the one structure that informs the possible mechanism of the other.
The MscS protein has been much less intensively studied as a result of the relatively recent discovery of the structural gene (12). Genetic and biochemical analyses established that the protein has three transmembrane (TM) helices and a large carboxyl-terminal cytoplasmic domain (13). This was confirmed by resolution to 3.8 Å of a crystal structure of the MscS channel that also showed the protein to be a homoheptamer (1). The large cytoplasmic domain was demonstrated to form a 40-Å diameter chamber perforated by eight holes, seven of which arise at the boundary between pairs of monomers and the eighth from the formation of a ␤ barrel by the carboxyl-terminal 15 residues. In the crystal structure the TM3 helix bends at Gly 113 , such that the carboxyl-terminal part of this helix lies parallel to the membrane surface. The loop between TM2 and TM3 (residues 91-95) forms an extended chain that is part of the wall of the open pore. Helices TM1 and TM2 are close packed to each other but are displaced away from TM3. It has been suggested that channel closure might be effected by closer packing of TM1 and TM2 with TM3 with propagation of this change through the TM2-TM3 loop into TM3 and the cytoplasmic domain (1). Here we present evidence that in the closed channel the transmembrane helices and residue 267 at the base of the carboxyl-terminal domain can be in close proximity to each other. These data support recent electrophysiological data suggesting that a rearrangement of the carboxyl-terminal domain is essential for channel gating (14).

Growth and Downshock Protocols
Strains were grown routinely at 37°C in Luria-Bertani medium (LB) containing per liter: 10 g of tryptone, 5 g of NaCl, and 5 g of yeast extract. Medium was supplemented with ampicillin (25 g/ml) if required. Agar plates contained 1.4 g/liter agar. Analysis of survival after downshock was essentially as described previously (13). Survival assays were performed in MJF465, and both induced and uninduced cell preparations were tested.

Mutant Creation
The single Cys mutants were created using the Stratagene QuikChange site-directed mutagenesis protocol using the primers listed in Table I. Six single Ser to Cys mutants at residues 9, 26, 49, 58, 95, and 267 were created in pMscSH 6 . The mutant plasmids were verified by sequencing on both strands and were transformed into MJF465. Mutant S267C was also constructed in pMscS to eliminate potential interference in the analysis by the His tag.

Membrane Preparation and Western Blotting
Membranes harvested from transformants that had been induced for 30 min using 0.3 mM IPTG as described previously (13) were resus-* This work was supported by The Wellcome Trust (040174). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed.  (15). Protein concentration in membrane preparations was assayed by the Folin-Ciocalteau method (16). Western blots were performed as described previously using either anti-His 6 antibodies (mouse IgG2a isotype) (Sigma) or peptide antibodies-specific for MscS (13). Preformed SDS-polyacrylamide gels (Novex) were used to separate the proteins prior to transfer onto nitrocellulose.

Purification of MscS
Membrane preparations derived from cells expressing the S267C mutant (1.5 mg/ml membrane protein) were incubated with 31.25 M oPDM for 15 min at 25°C and the reaction stopped by the addition of DTT (5 mM final concentration). A sample was retained for Western blotting to verify that reaction with oPDM was as observed in earlier experiments. The treated membranes were harvested by centrifugation (50,000 rpm, Beckman TLA100.4 rotor, for 1 h at 4°C) and the pellets weighed to calculate the total protein present. The pellets (ϳ0.2 g) were suspended in 5 ml of 50 mM sodium phosphate buffer, pH 7.4, containing 300 mM NaCl, 50 mM imidazole, and 2% Triton X-100 (membrane grade, Roche Diagnostics). The membranes were incubated at 37°C for 2 h and were shaken at intervals. The solubilized protein was centrifuged (50,000 rpm, Beckman TLA100.4 rotor, for 1 h at 4°C) and the clear supernatant retained. In previous experiments it had been established that 2% Triton X-100 gave efficient solubilization (Ͼ95%) of MscS (data not shown). The supernatant was mixed with 0.25 ml nickelnitrilotriacetic acid-agarose (Qiagen), placed in Qiagen 1-ml polypropylene columns, and the columns were capped and placed on a rocking platform at room temperature for ϳ16 h. The columns were then allowed to settle and the unbound protein collected. The column was then washed four times with 2.5 ml of 50 mM sodium phosphate buffer, pH 7.4, containing 300 mM NaCl, 50 mM imidazole, 10% glycerol, 10% ethanol, 10 mM ␤-mercaptoethanol, and 0.2% membrane-grade Triton X-100. The buffer also contained 1 Complete™ EDTA-free protease inhibitor tablet (Roche Diagnostics) per 50 ml of solution. The membrane protein was eluted by addition of four 0.25 ml volumes of the above wash buffer amended to contain 0.3 M imidazole. Eluted protein fractions were treated to remove excess detergent prior to gel electrophoresis using the Pierce PAGEprep TM product following the manufacturer's instructions. Samples were separated on Novex NuPAGE 4 -12% BisTris gels run in MES buffer.
Protein bands corresponding to the monomer through to the heptamer were excised from the SDS-PAGE gels and were reduced, Salkylated, and in-gel-digested with trypsin (Promega, Madison, WI). The tryptic peptides were extracted from the gel piece, and an aliquot of the mixture was desalted using a GELoader tip containing POROS R2 sorbent (PerSeptive Biosystems). The peptides were then eluted onto a sample plate with 0.5 l of a saturated solution of ␣-cyano-4-hydroxycinnamic acid (Sigma, Poole, UK) in 70% acetonitrile, 5% formic acid. Mass spectra were acquired with a PerSeptive Biosystems Voyager DE-STR MALDI-TOF mass spectrometer operated in reflectron-delayed extraction mode. Spectra were calibrated with trypsin auto-digestion products and proteins identified by interrogation of protein databases using MASCOT and MS-Fit data base searching programs.

Cross-linking Reactions
All cross-linking protocols were based on published methods (17,18). Copper/Phenanthroline-Prewarmed (37°C) membranes, containing 1.5 mg of membrane protein/ml, were treated with freshly prepared copper/phenanthroline reagent at 37°C for 10 min. The final concen-trations of Cu 2ϩ and o-phenanthroline were 0.167 and 0.5 mM, respectively. The reagent was made by mixing equal volumes of freshly prepared 30 mM phenanthroline (dissolved in ethanol) and aqueous 10 mM CuSO 4 . This stock was diluted to give the appropriate concentrations of Cu 2ϩ and o-phenanthroline (see above). Control incubations were performed with ethanol. Reactions were stopped by addition of NEM (10 mM final concentration).
MTS-1-MTS-This reagent was dissolved in Me 2 SO to provide a 1 mM stock solution and was used at a final concentration of 100 M. Reactions were incubated at 25°C for 1 h and quenched with NEM (as above). Controls contained an equivalent volume of Me 2 SO.
o-Phenylenedimaleimide-oPDM was dissolved in Me 2 SO to provide a 250 mM stock solution, which was kept at room temperature in a foil-wrapped tube and was discarded after each experiment. The final concentration of oPDM was between 15 and 250 M. Similar data were obtained at all concentrations of oPDM and 31.25 M oPDM was adopted for routine analysis of cross-linking. Membranes (1.5 mg of protein/ml) were incubated at either 4 or 25°C for up to 1 h. The reaction was terminated by addition of DTT (5 mM final concentration).

Creation and Physiology of Single Cysteine Mutants of
MscS-Single cysteine mutants were created in the 286-residue E. coli MscS protein, which is Cys-free, by replacing selected serine residues (Ser 9 , Ser 26 , Ser 49 , Ser 58 , Ser 95 , and Ser 267 ). Residues Ser 9 and Ser 26 are not visible in the crystal structure, which starts at residue Tyr 27 . However, we can estimate the separation of S26C residues from the position of Tyr 27 . In adjacent subunits of the crystal structure, which is proposed to depict the open channel, the mutated residues are 10 -33 Å apart (Fig. 1a). Both basal and induced expression of the mutant proteins protected the triple MS channel-deficient mutant (MJF465; mscL Ϫ mscK Ϫ mscS Ϫ ) (12) against downshock ( Fig. 1b; data not shown for induced channels as this was found to be 100% in all cases). Protection by the basal expression of the MscS channel is a strong indicator of normal organization and function in the channel (13). The resultant proteins were expressed at similar levels to the Cys-free parent after induction with 0.3 mM IPTG (Fig. 2, a and b). These data suggest that the cysteine mutations do not significantly disturb the MscS channel organization. In non-reducing gels S9C mutants were detected as dimers, even prior to incubation with an oxidising agent (Fig. 2a), but were converted to monomers by DTT (Fig. 2b). Neither S26C nor S95C, which are also located on the periplasmic face of the channel, were observed as dimers in the absence of oxidising agents. In the crystal structure, S9C is not visible (1) but is predicted to be periplasmic (13,20). It alone has sufficient conformational flexibility to form disulfide bridges without addition of any oxidizing agent.
Cross-linking by Oxidizing Agents-Treatment of membrane preparations, in which the channel is in the closed state (21), with copper/phenanthroline reagent (18) generated dimers for all of the proteins except the wild type, which is a Cys-free protein, and S49C (Fig. 2a). The monomer and dimer of MscS have slightly faster mobility than expected from the theoretical mass, as is frequently observed with membrane proteins. Higher oligomers (usually proteins with mobility consistent with a tetramer) were observed for control incubations of S9C and for the copper/phenanthroline-treated S26C and S58C. The effect was very marked in the latter where the tetramer was observed more often than the dimer and the dimer was only visible on longer exposures of the Western blot. All of the disulfide bridges could be reduced by incubation with DTT except that for S95C, which proved to be quite resistant to the reducing agent (Fig. 2b). Incubation with the disulfide bridgegenerating reagent MTS-1-MTS (17) also resulted in dimer formation for all the Cys mutants, except S49C, and incubation with DTT reduced the dimers to the monomer (data not shown).
Cross-linking with oPDM-oPDM cross-links cysteine residues that are capable of adopting positions within ϳ10 Å of each other (range 9.4 Ϯ 0.4 Å) (22). Exposure of the MscS single Cys mutants led to S26C, S58C, S95C, and S267C forming higher order oligomers that were stable to SDS (Fig. 3a). Only S95C formed solely the dimer, whereas tetramers and hexamers were evident for S26C on extended exposure of the Western blot and the dominant form obtained with S58C had the mobility of the tetramer (Fig. 3a). The most startling result was seen with S267C, which formed multiple oligomeric forms from dimer through to the heptamer (Fig. 3a). The multimeric forms were seen to proceed via formation of the dimer followed by the appearance of trimers, then tetramers, pentamers, and finally the hexamer and heptamer (Fig. 3b). This observation was only made with the S267C mutant after incubation with oPDM. All of the cross-linked forms observed with S267C after oPDM treatment exhibited unusually fast mobilities suggesting that the protein might exhibit a more compact conformation. Incubation at high temperature or with urea-containing loading buffer, immediately prior to loading the gel, did not breakdown the multimers. These data give the clearest support to the heptameric form seen in the crystal structure (1) and suggests that this organization is maintained in the closed state.
The observation of the heptamer was made prior to the crystal structure being available, at a time when the protein was thought to be hexameric (21). This led us to perform a number of controls to eliminate the possibility of artifacts. The cloned MscS protein has a carboxyl-terminal His 6 tag to facilitate detection and purification. To ensure that this extra sequence was not responsible for the formation of the higher oligomers, the cross-linking was repeated in a MscS S267C construct lacking the His 6 tag and was detected using antibodies to the native MscS protein. The same result was obtained, separate and distinct bands up to the heptamer (Fig. 4a). We expect the MscS channel to be closed in membrane preparations (21). To confirm that the cross-linking effected was be-  (1). S9C cannot be placed as these residues are not visible in the crystal structure but are known to lie on the same face of the membrane as S26C and S95C (13). The images were created and inter-residue distances measured using MDL Chime via Protein Explorer (28). b, survival of MJF465 subject to downshock from minimal medium containing 0.3 M NaCl into distilled water. Data are shown for uninduced plasmids. Induction led to 100% protection in all cases (data not shown).

FIG. 2. Disulfide bond formation in MscS
Cys mutants using copper/phenanthroline. Membranes derived from MJF465 expressing MscS (either wild type or site-specific cysteine mutants) were harvested after induction for 30 min using 0.3 mM IPTG as described previously (13). Membranes (1.5 mg of membrane protein/ml) were preincubated at 37°C for 10 min prior to addition of freshly prepared copper/phenanthroline reagent (0.167 mM copper/0.5 mM phenanthroline) and the incubation continued for 10 min. Control tubes were incubated with ethanol (10%), the solvent used for the copper/phenanthroline. Reactions were stopped by addition of 10 mM NEM. Samples of the completed reactions were subjected to SDS-PAGE under non-reducing (a) or reducing conditions (b) (Novex NuPAGE 4 -12% BisTris gels run in MES buffer), transferred onto nitrocellulose membranes, and MscS proteins detected by Western blotting using anti-His 6 tag primary antibodies (Sigma) and anti-mouse HRP conjugate secondary antibodies (Sigma). Perbio Supersignal West Dura Substrate was used for a standard ECL method of detection. Blots were exposed to Amersham Biosciences film and recorded using a Kodak Image Station 440CF. tween subunits of the closed channel, reactions were performed in growing cells where the channel must be closed. Bands up to the heptamer were visible in Western blots of whole cells and in membranes purified from the cross-linked cells (Fig. 4b).
The wild type protein, which lacks cysteine residues, was not affected by oPDM, which suggests that the observed pattern of oligomers was not caused by reaction of the cross-linker with other residues. E. coli cells express a MscS-related protein MscK that contains a number of cysteine residues. However, the results shown here were obtained in MJF465, which lacks MscK, and identical patterns of cross-linking were observed in strains that possessed MscK. To determine whether the higher molecular mass oligomers arose by cross-linking to other proteins, we performed tryptic peptide fingerprinting on the purified protein. The MscS S267C protein was purified by Ni 2ϩchelate chromatography after treatment of the membranes with oPDM and the purified material resolved by SDS-PAGE. A clear pattern of seven bands was seen (Fig. 4c). Peptide fingerprinting by MALDI-TOF mass spectrometry demonstrated that no other protein was linked to MscS in any of the seven bands. Mowse scores of Ն1e ϩ 005 were observed for MscS in all seven protein bands, with no other significant hits. All seven proteins gave similar tryptic digest patterns, suggesting that no further cross-links were formed in the higher order oligomers. Furthermore, no evidence was found for other intra-MscS cross-links that could account for the higher order oligomers. Eleven of the 16 potential tryptic peptides were observed. Two of the missing peptides correspond to large (4.2 and 4.9 kDa) molecular mass hydrophobic peptides (TM1 and TM3) that are not readily detected by mass spectrometry. One of the missing peptides contained the cross-linked residue, S267C, and could be detected as an oPDM-modified peptide, but the cross-linked peptide itself was not observed in the MS. These data support a model in which the oligomeric state is probably formed by stabilization of the multimeric state rather than by additional cross-linking. This suggestion is supported by the observation that oxidation of single cysteine residues in several mutants also stabilized higher order oligomers (Figs. 2a  and 3a). The native MscS protein, when expressed to very high levels in E. coli, also forms multiple molecular forms up to the heptamer, 2 suggesting that this property is intrinsic to MscS and that cross-linking merely stabilizes this tendency. DISCUSSION The analysis reported here is of cross-linking of specific residues in the closed state of the MscS channel. Previous workers have established that MS channels in isolated membranes are in the closed state (11). Most of our studies were conducted on isolated membranes and represent analysis of the closed state. Additional studies showed that identical patterns of cross-linking of S267C were observed in isolated membranes and in whole cells, where for physiological reasons the channel must be closed (Fig. 4). MscL represents the most in-depth analysis of the closed and open configurations of an MS channel, with the former observed in the crystal state and the latter by molecular dynamics simulations (6,8,10,23). All the models envisage the transition from a compact protein to an expanded state as the channel moves from the closed to the open state. Cross-linking data has been used extensively to support these models. In the MscS crystal structure, which is proposed to represent the open state of the channel (1), the distance between the residues selected for mutagenesis range from 10 to 33 Å (Fig. 1a). None should easily form disulfide bonds, yet the majority do so in the closed state of the channel (Fig. 2a). Our data suggest that spontaneous cross-linking, which could trap a rare conformation of a flexible MscS protein, does not take place to any significant extent except for S9C (Fig. 2). The 2 R. Bass, personal communication. with oPDM at 25°C is not influenced by the presence of the His tag. pMscS S267C was created as described under "Experimental Procedures" (13). Membrane preparation and cross-linking with oPDM was as described in the legend to Fig. 3a and the Western blotting performed using antibodies specific to MscS (13). b, in vivo cross-linking was carried out by addition of 100 M oPDM to exponentially growing MJF465/pMscSH 6 (S267C) cells, which had been induced for expression by 0.3 mM IPTG 30 min prior to oPDM addition. Control cells were treated with Me 2 SO, the solvent used for oPDM. Whole cells and membranes prepared from these cells were analyzed for crosslinking as above. c, purified MscS after cross-linking with oPDM. Membranes derived from MJF465/pMscSH 6 (S267C) were incubated with 31.25 M oPDM at 25°C for 15 min and the MscS protein purified as described under "Experimental Procedures." The proteins were separated on Novex NuPAGE 4 -12% BisTris gels run in MES buffer. The gel was stained with GelCode Blue dye. Samples were also used to develop Western blots and for peptide fingerprinting (see "Experimental Procedures"). latter may be a substrate for the periplasmic disulfide oxidoreductase system (24). Conformational flexibility is known to influence the apparent distance between residues, as revealed by cross-linking experiments (25). Generally it is considered that proximity and cross-linking efficiency are correlated, but if a polypeptide chain exhibits significant flexibility then residues can be cross-linked, although far apart in the native structure.
Most of the mutants were cross-linked by oPDM, suggesting that these residues can readily adopt positions within the distance required for efficient cross-linking. oPDM has been estimated to bridge 9.4 Ϯ 0.4 Å (22) and, with the possible exception of S95C, would be expected to react poorly with the Cys mutants in the open channel configuration. The only residue that did not form either disulfide bridges or react with oPDM, S49C, is predicted to be the least favorably arranged for crosslinking. The crystal structure suggests that the S49C side chains are not positioned in a way that would easily facilitate reaction. The lack of reactivity of S49C provides an internal control for nonspecific cross-linking, since the absence of crosslinks involving this residue under any of the conditions tested suggests that those that are formed by other residues are specific and a realistic indication of proximity or dynamics in the closed structure. S58C readily formed disulfides with copper/phenanthroline and cross-linked oligomers with oPDM, suggesting that in the closed structure these residues can be located as few as ϳ3 Å apart compared with up to ϳ33 Å in the crystal structure.
Rees and colleagues (1) proposed a model for the gating of MscS that envisaged a significant lateral movement within the plane of the membrane during the transition from the closed to the open state. The data presented here are consistent with such a model, since residues that are distant in the crystal are cross-linked in a manner predicted from the model. In this case, the cross-linking data would predict that transmembrane helices 1 and 2 undergo a displacement of up to 30 Å during the transition from the closed to the open state. The most COOHterminal mutant studied, S267C, lies at the base of the large carboxyl-terminal domain of the channel, at the boundary between the ␣␤ domain and the final ␤ barrel that completes the protein. Adjacent S267C residues are separated by ϳ20 Å in the crystal structure but readily form disulfide bridges and are cross-linked by oPDM (Fig. 3, a-c). These data suggest that the cytoplasmic domain may undergo a significant conformational change during the closed to open transition. Support for this suggestion comes from the observed inhibition of gating of MscS by cross-linking of carboxy-terminal His tags by Ni 2ϩ (14). The inhibition required high concentrations of Ni 2ϩ , suggesting that multiple cross-links were needed to block the re-organization of the carboxyl-terminal domain. Second, small deletions affecting the carboxyl terminus of MscS prevent the stable insertion of the protein into the membrane, consistent with the importance of this domain for the structure of the channel (13).
The data presented here strongly support the published model for MscS, which is derived from the crystal structure (1) and supporting genetic work (13). The cross-linking of S267C provides the first biochemical evidence for the heptameric structure of MscS. The progressive formation of higher oligomers, most evident with S267C treated with oPDM, suggests that the MscS protein can form SDS-stable complexes similar to those seen for K ϩ channels (26,27). The cross-linking data, particularly for S267C, suggest that when two subunits are linked by an inflexible bond, the interactions with other subunits are strengthened, i.e. a conformational change consequent upon cross-linking of pairs of subunits is transmitted to adjacent subunits such that higher oligomers become stable. The rate of dimer formation is rapid relative to the formation of non-covalently linked oligomers (0.5-3 min compared with 9.5-12 min, although the trimer is evident after 3-min incubation), suggesting that a significant conformational change is required for the stabilization of the contact between the subunits. In conclusion, we believe our data support the heptameric model for MscS and suggests significant flexibility in the protein that may be essential for the transition between the closed and open states.