Dynamics of Protein-Protein Interactions at the MscL Periplasmic-Lipid Interface

Abstract

MscL, the highly conserved bacterial mechanosensitive channel of large conductance, is one of the best studied mechanosensors. It is a homopentameric channel that serves as a biological emergency release valve that prevents cell lysis from acute osmotic stress. We previously showed that the periplasmic region of the protein, particularly a single residue located at the TM1/periplasmic loop interface, F47 of Staphylococcus aureus and I49 of Escherichia coli MscL, plays a major role in both the open dwell time and mechanosensitivity of the channel. Here, we introduced cysteine mutations at these sites and found they formed disulfide bridges that decreased the channel open dwell time. By scanning a likely interacting domain, we also found that these sites could be disulfide trapped by addition of cysteine mutations in other locations within the periplasmic loop of MscL, and this also led to rapid channel kinetics. Together, the data suggest structural rearrangements and protein-protein interactions that occur within this region upon normal gating, and further suggest that locking portions of the channel into a transition state decreases the stability of the open state.

Introduction

The bacterial mechanosensitive channel of large conductance, MscL, is located in the cytoplasmic membrane and gated by membrane tension (1–6). It is one of the best studied mechanosensitive channels and serves as a paradigm to address how a mechanosensor is able to sense tension in the membrane (6–8). The crystal structure of Mycobacterium tuberculosis MscL (Mt-MscL) revealed a homopentamer in a closed or nearly closed conformation (9), in which each subunit has two transmembrane domains (TM1 and TM2) connected by a periplasmic loop, an N-terminal S1-helix that runs along the membrane, and a C-terminal helical bundle (Fig. 1 A). Within the complex, the five TM1 domains tightly pack to form the inner helix bundle or pore of the channel, and the five TM2 helices form the outer part of the channel that faces the lipids. Opening of the channel is achieved by rotating and tilting of TM1 helices in the membrane plane, and then the channel opens like the iris of an old-fashioned camera, thus allowing ion permeation (7,10–12). MscL has an open pore of 30 Å in diameter (13), suggesting that some domains, especially TM1, undergo large movements as the channel opens. The periplasmic loop is predicted to also show large conformational changes upon channel gating, as determined by molecular-dynamics simulations (14–16). Previous studies suggested that the loop acts as a spring that resists the opening of the channel and sets the energy level needed to gate the channel (17–19). For example, one previous study showed that one residue, Q56 at the loop of Escherichia coli MscL (Eco-MscL), plays a role in both channel mechanosensitivity and kinetics (20), and another study reported that interactions between the Eco-MscL Q65 residue and lipid headgroups during channel opening are important for channel gating (15,19)

Figure 1.

Crystal structure of Mt-MscL and alignment of MscL homologs. (A) Crystal structure of Mt-MscL, showing a homopentamer with each subunit having two transmembrane domains (TM1 and TM2) connected by periplasmic loop (peri-loop) marked in blue, an N-terminal cytoplasmic helix (S1) connected to TM1, and a C-terminal cytoplasmic helical bundle connected to TM2 by a short cytoplasmic loop. The amino acid shown in red is G47 in Mt- MscL, which is equivalent to F47 in S. aureus and I49 in E. coli MscL. (B) Sequence alignment of MscL homologs from M. tuberculosis (Mt-MscL), E. coli (Eco-MscL), and S. aureus (Sa-MscL). Included in blue boxes are identical and similar residues in the M. tuberculosis, E. coli, and S. aureus MscL protein.

Many studies have defined structural changes and protein-protein and protein-lipid interactions at the cytoplasmic regions of the protein that occur upon channel gating (21–23). However, much less is known about such interactions on the periplasmic side of the protein. One exception is a recent study that showed that one single residue at the periplasmic lipid-aqueous interface, F47 of S. aureus and I49 of E. coli MscL (highlighted in Fig. 1, A and B), controls both channel kinetics and mechanosensitivity (24). Mutagenesis studies of the S. aureus MscL (Sa-MscL) F47 demonstrated that hydrophobic amino acid substitutions led to long-open-dwell-time channels, whereas bulky aromatic amino acids confer short openings. Analogous studies with E. coli MscL gave consistent results and thus confirmed the conserved function of this residue. Together, the data suggested that this residue changes its local environment, including potential protein-protein interactions, as the channel gates.

In this study, to identify possible interactions involved in controlling channel kinetics, we made single- and double-cysteine mutations, and assayed potential interactions by cysteine trapping and patch-clamp recordings. Our results suggest that two different interactions with Sa-MscL F47 and Eco-MscL I49 occur upon gating, and trapping the channels when these interactions occur leads to an intermediate gating state that effects a destabilized open state and thus shorter open dwell times.

Materials and Methods

Strains and cell growth

All mutants were generated using the Mega primer method as described previously (25) and inserted into either the pB10d expression construct (26,27) or the pET21a expression vector. The E. coli strain PB104 (ΔmscL:Cm) was used as host for the expression of C-terminal his-tagged Sa-MscL F47C, W59C, F47C/W59C, F47K, W59K, and F47K/59K, and Eco-MscL I49C, A58C, I49C/A58C, I49K, A58K, and I49K/58K in expression vector pB10d for electrophysiological experiments (20,26). PB116 (11) was used as host for the expression of C-terminal his-tagged Sa-MscL F47C, W59C, F58C, G60C, I61C, F47C/W59C, F47C/F58C, F47C/G60C, F47C/I61C, and Eco-MscL I49C in expression vector pET21a. PB104 was also used as host for expression of C-terminal his-tagged Eco-MscL A58C, I49C/A58C, I49C/Q56C, I49C/F57C, and I49C/V59C MscL in expression vector pB10d for the disulfide-trapping experiments. Cultures were grown routinely in Lennox broth (LB) medium (Fisher Scientific, Pittsburgh, PA) containing 100 μg/ml ampicillin. Expression was induced by addition of 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG; Anatrace, Maumee, OH).

In vivo disulfide trapping

Disulfide-trapping experiments were performed as described previously (11,21,28,29). Briefly, cells were grown in LB medium containing 0.5 M NaCl until OD reached ∼0.2, and induced with 1 mM IPTG for 10–60 min at 37°C. They were then shocked with copper phenanthroline (CuPe; final concentration 150 μM, except for Sa-F47C and Eco-I49C, which were so efficient that we decreased the concentration to 1.5 μM) at a 1:20 dilution in water at 37°C for 15 min. The samples were immediately spun down and resuspended in nonreducing sample buffer and run on a 4–20% gel (Bio-Rad, Hercules, CA) for western blot analysis. MscL-specific primary antibody (1:15,000) and secondary rabbit anti-mouse horseradish peroxidase (HRP; 1:50,000; Bio-Rad) were used for mutant Eco-MscL A58C as previously described (20). For all other mutants, we used anti-Penta His (Qiagen, Valencia, CA) at 1:4000 and secondary goat anti-mouse HRP (Bio-Rad) at 1:100,000. Blots were developed with the use of HRP substrate (Millipore, Billerica, MA) and exposed to film.

Electrophysiology

Excised, inside-out patches from E. coli giant spheroplasts were examined at room temperature as described previously (30,31). Patch buffer contained 200 mM KCl, 90 mM MgCl₂, 10 mM CaCl₂, and 5 mM HEPES (Sigma-Aldrich, St. Louis, MO), pH 6.0. Data were acquired at −20 mV at a sampling rate of 20 kHz with a 5 kHz filter, using an AxoPatch 200B amplifier in conjunction with Axoscope software (Axon Instruments, Union City, CA). For simplicity, the channel openings are shown upward in the figures. A piezoelectric pressure transducer (World Precision Instruments, Sarasota, FL) was used to monitor the pressure introduced to the patch membrane by suction throughout the experiments. Measurements were performed using Clampfit9 from pCLAMP9 (Axon Instruments). For experiments requiring oxidative conditions, 150 μM Cu-phenanthroline (Sigma-Aldrich) was backfilled into patch pipettes as described previously (26).

For the kinetic experiments, the data were fit to a three-open-state model as previously described (20,26,27). The two longer time constants (τ2 and τ3) could be easily determined; the shortest (τ1) is beyond the resolution of the equipment used. All kinetic data were obtained from at least 300 opening events per mutant from three independent experiments performed in at least two independent spheroplast preparations.

Mechanosensitivity was determined as previously described (26,27,31). Briefly, endogenous MscS, the bacterial mechanosensitive channel of small conductance that is activated at lower membrane tension, was used as an internal standard within the patch, and the ratio of gating pressure for MscL divided by the threshold pressure of MscS was determined (here we will refer to this as the pL/pS ratio). All mechanosensitivity data were obtained from at least three independent recordings performed in at least two independent spheroplast preparations.

Results

Intersubunit cross-linking of Sa-F47C as well as Eco-I49C MscL decreases channel open dwell times

A large iris-like expansion of the MscL pore occurs upon channel gating (10–12), suggesting that the previously identified residues at the TM1/periplasmic loop interface undergo large movements as the channel opens. Given these large changes and the results from our previous study (24), these residues presumably have dynamic interactions with multiple environments. To identify potential interactions with the residue at the TM1/periplasmic loop interface, we performed an in vivo disulfide trapping assay (11,21). Briefly, mutant-expressing cells were osmotically downshocked in the presence of the oxidative reagent CuPe, followed immediately by nonreducing SDS-PAGE and western blot analysis. Note that neither Eco- nor Sa-MscL contains endogenous cysteines. We first tested and found that the single-cysteine mutants at the TM1/periplasmic loop interface formed disulfide bridges. As shown in Fig. 2 A, some of the Sa-F47C MscL and Eco-I49C MscL channel proteins migrated as a dimer upon addition of as little as 1.5 uM CuPe, indicating that an interaction between cysteine residues from two subunits within the pentameric complex occurred within both Sa- and Eco-MscL. Dimerization was osmotic-shock dependent, presumably because the channel is induced to gate, thus changing its conformation and the proximity of these residues, which are not predicted to be close in the closed structure (9). Increasing the CuPe treatment of Sa-F47C and Eco-I49C to 150 μM upon osmotic shock resulted in a similar dimerization pattern.

Figure 2.

Cross-linking Sa-F47C and Eco-I49C cysteines with the analogous residue in another subunit in the pentamer decreases the channel open dwell time of Sa-MscL and Eco-MscL. (A) Western blot analysis of disulfide trapping experiments. All samples are either nonshocked (NS) or shocked (S⁺) in the presence of the oxidizing reagent CuPe. (B) Patch-clamp recordings of Sa-F47C and Eco-I49C mutants. O and C refer to channel open and closed states, respectively. The mechanosensitivity of channels measured as (pressure threshold for MscL)/(pressure threshold for MscS) is shown as mean ± SEM at the top left of each trace. The scale bar is shown at the top right. (C) Open-dwell-time analysis showing the τ2 and τ3 of each mutant channel in solid and open boxes, respectively. ^∗∗p < 0.01 compared with corresponding non-CuPe groups (two-tailed Student’s t-test). The data were fit to a three-open-state model; τ2 and τ3 are the longer time constants. The shortest (τ1) is below resolution and is not shown; n, which is the same for τ2 and τ3 of each mutant, is shown on the top of the box.

To determine whether these intersubunit cross-links in Sa-MscL F47C and Eco-MscL I49C modulate channel function, we performed single-channel patch-clamp analysis under varying redox conditions. In determining the channel kinetics, we found, as observed in previous studies (20,26,27), that the data fit a three-open-state model well. Only the longer time constants, τ2 and τ3, were easily resolved, because the shortest (τ1) was <1 ms and thus below the resolution of our recordings. As shown in Fig. 2, B and C, the τ2 and τ3 open dwell times of Sa-F47C MscL dramatically decreased when CuPe was backfilled within the patch pipette, as previously described (26). To determine mechanosensitivity, MscS is routinely used as an internal standard, and the ratio of the threshold pressure required to gate MscL relative to MscS (pL/pS ratio) is determined (20,26). Although the absolute values of the amount of pressure may vary from patch to patch because of variations in the radius of curvature of the patch, because MscL and MscS both sense membrane tension, we find the threshold ratios are invariant. Using this approach, we found that wild-type Sa-MscL had a pL/pS ratio of 1.96 ± 0.07 (24), and the Sa-F47C MscL showed a slightly increased sensitivity (1.75 ± 0.05) compared with wild-type Sa-MscL. However, the mechanosensitivity of the Sa-F47C channel was not significantly altered upon changes in the redox conditions (pL/pS ratio: control 1.75 ± 0.05 versus CuPe 1.74 ± 0.06, p > 0.5, Student’s t-test). Similarly, as shown in Fig. 2, B and C, the open dwell times of Eco-I49C MscL were significantly decreased with the presence of CuPe in the pipette as compared with the control. Again, the channel mechanosensitivity was not significantly changed (pL/pS ratio: control 1.31 ± 0.05 versus CuPe 1.23 ± 0.1, p > 0.1, Student’s t-test). Note that although the wild-type Eco-MscL had a pL/pS ratio of 1.47 ± 0.03, which is consistent with that previously observed (25), the mechanosensitivity of Eco-I49C (pL/pS ratio: 1.31 ± 0.05) was slightly changed but still within a relatively normal physiological range. There was no apparent change in conductance upon CuPe treatment for the Sa-F47C and Eco-I49C mutants. Hence, the data strongly suggest that the cysteines mutated at the TM1/periplasmic loop interface can interact with the analogous site on another subunit, and when they are trapped in this structure, the major difference observed is the decreased channel open dwell times.

A double-cysteine scan of the periplasmic loop of Sa-MscL reveals an interaction that, when disulfide trapped, decreases the channel open dwell times

We next aimed to determine whether specific interactions between the residue at the TM1/periplasmic loop interface and residues in the periplasmic loop could modulate channel function. Given that F47 is at the lipid-periplasmic interface, and that the Sa-MscL periplasmic loop is quite small, it seemed likely that an interacting residue might be toward the periplasmic loop near the TM2 interface of the channel. We therefore performed a cysteine scan of the region, within the Sa-MscL 47C MscL background, sequentially mutating residues F58 to I61 to cysteines. We utilized our in vivo cysteine trapping approach and relied on the possibility that such an interaction may be more efficient than the F47C-F47C disulfide bridging described in the previous section, thus leading to higher-order multimers in the western analysis. As illustrated in Fig. 3 A, our data indeed show that Sa-MscL W59C in the periplasmic loop efficiently interacts with Sa-MscL F47C to produce a disulfide bridged pentamer, whereas each single-cysteine mutant (Sa-W59C or Sa-F47C) produces some dimers. Whereas all of the single mutants and some of the double mutants produced some dimers, no significant amount of pentamers was observed after the double mutants F47C/F58C, F47C/G60C, or F47C/I61C were cysteine trapped (Fig. S1 in the Supporting Material). Thus, it appears that F47C at the interface of the TM1/periplasmic loop can specifically interact with W59C in the periplasmic loop.

Figure 3.

Disulfide trapping of Sa-F47C/W59C and Eco-I49C/A58C in the periplasmic loop of MscL decrease the open dwell time. (A) Western blot analysis of disulfide trapping experiments. All samples are either nonshocked (NS) or shocked (S⁺) in the presence of oxidizing reagent CuPe. (B) Patch-clamp recordings of Sa-MscL and Eco-MscL double and single mutants. O and C refer to channel open and closed states, respectively. The mechanosensitivity of channels measured as (pressure threshold for MscL)/(pressure threshold for MscS) is shown as mean ± SEM at the top left of each trace. The scale bar is shown at the top right. ^∗∗∗p < 0.0001 compared with corresponding non-CuPe groups (Student’s two-tailed t-test). (C) Open-dwell-time analysis showing the τ2 and τ3 of each mutant channel in solid and open boxes, respectively. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001 compared with corresponding non-CuPe groups (two tailed Student’s t-test); n, which is the same for τ2 and τ3 of each mutant, is shown on the top of the box.

To determine whether the double-cysteine interactions in the Sa-MscL F47C/W59C mutant changed channel function, we performed patch-clamp recordings. As shown in Fig. 3, B and C, single-channel recordings and open-dwell-time analysis show that channel open dwell times were drastically decreased when CuPe was applied in the patch pipette. The mechanosensitivity and channel conductance were not significantly changed. In contrast, after similar CuPe treatment, the single-mutant Sa-MscL W59C MscL did not have any measurable change in either open dwell times or mechanosensitivity. Although we cannot totally rule out the possibility that the decreased open dwell times for the Sa-MscL F47C/W59C mutant are due to F47C-F47C dimerization, the efficiency of the pentamerization of this mutant (Fig. 3 A) makes it unlikely. Hence, it appears that in Sa-MscL, F47C could efficiently cross-link with W59C; the appearance of pentamers suggests it is from different subunits. Locking the channel in this intermediate protein-protein interactive state leads to decreased channel open dwell times. The results suggest that in wild-type Sa-MscL, F47 and W59 from different channel subunits interact with each other, and locking this interaction leads to a decrease in channel open dwell times.

A double-cysteine scan of the periplasmic loop of Eco-MscL also reveals an interaction that, when disulfide trapped, decreases the channel open dwell times and mechanosensitivity

Having found that the residue at the TM1/periplasmic loop in Sa-MscL specifically interacts with another residue in the periplasmic loop, and that locking this interaction leads to a decrease in open dwell times, we next wanted to determine whether this is also the case for Eco-MscL. Due to the nonconserved nature of periplasmic loop, we scanned much of the region Q56–H74 (except for R62 and D63, because of their charges). Again, by using our in vivo cysteine trapping method, we found that Eco-MscL I49C/A58C specifically formed higher-order multimers. As shown in Fig. 3 A, western blot shows that nonshocked Eco-MscL I49C/A58C appeared mostly as a monomer with a small portion of dimer formation. When mutant-expressing cells were shocked and treated with CuPe, dimers and trimers were formed. In contrast, the single-cysteine mutants, Eco-MscL I49C and A58C, formed dimers, but no trimers, when shocked and treated with CuPe. Fig. S2 shows that after CuPe treatment, additional double mutants, including Eco-MscL I49C/Q56C and I49C/V59C, formed only a small portion of dimers, and I49C/F57C did not form any visible dimers after CuPe treatment. Therefore, Eco-MscL A58C appeared to be the best candidate for interacting with Eco-I49C.

To determine the effect of disulfide trapping on the channel function of Eco-MscL I49C/A58C, we performed patch-clamp analysis. As shown in Fig. 3, B and C, single-channel recordings and open-dwell-time analysis showed that Eco-MscL I49C/A58C had decreased open dwell times after CuPe treatment, whereas A58C did not (Fig. 3, B and C). Interestingly, the measure for channel sensitivity, the channel pL/pS ratio, was substantially increased after CuPe treatment for Eco-I49C/A58C, but not the single mutants Eco-I49C or Eco-A58C, although slight tension sensitivity changes were observed. No obvious change in channel conductance was observed for the above mutants. These results suggest that the decrease in channel open dwell times is due to cross-linking between Eco-MscL I49C and A58C, and not due to I49C-I49C dimers, the latter of which do not appear to affect mechanosensitivity. Although the interaction between I49C and A58C seems to be less efficient than that observed for Sa-MscL, the cross-linking is still efficient enough to cause changes in both channel kinetics and mechanosensitivity.

An electrostatic repulsion test supports the interpretation that Sa-MscL residues F47/W59 interact upon normal gating

One of the caveats of disulfide trapping is that it is often unclear whether one has trapped a normal state of the protein or a state that is rarely obtained. Potential misinterpretations can be partially resolved by scanning an entire region, as we did here. However, we also recently developed a method, which we termed the electrostatic repulsion test (ESReT) (11), that can add additional support to the hypothesis that two residues interact during the normal gating process. Briefly, if interactions are required for the gating process, and both residues are mutated to like charges, the electrostatic repulsion should inhibit the gating process. Unfortunately, when we attempted this with the Eco-MscL I49 and A58 residues, which were both mutated to lysines, we found that although the amount of channel activity was severely depressed, the channels showed heterogeneous behavior in both mechanosensitivity and kinetics, suggesting that there may indeed be other, perhaps less used, pathways to channel opening. On the other hand, for Sa-MscL, we found that Sa-W59K showed a very similar channel behavior as wild-type Sa-MscL (pL/pS ratio: 2.2 ± 0.1, very short open dwell time < 1 ms). In contrast, the Sa-F47K channel displayed a gain-of-function (GOF) phenotype (pL/pS ratio: 0.95 ± 0.02) with a medium-long open dwell time (≤ 0.5 s) (24). Interestingly, combining the functionally wild-type W59K with the GOF F47K significantly decreased the GOF phenotype (pL/pS ratio: 1.60 ± 0.1). In addition, whereas the single mutants had short or measurable kinetics, the double mutant showed an extremely stabilized open state that routinely lasted for >1 s (Fig. S3) and was even observed for 240 s, suggesting that once the channel is opened, the electrostatic repulsion inhibits channel closure. These data support the hypothesis that these residues interact during the normal gating process.

Discussion

Our previous study demonstrated the importance of Sa-MscL F47C and Eco-MscL I49C residues at the TM1/periplasmic interface in defining channel kinetics and mechanosensitivity, and suggested that these residues change their local environment upon channel gating (24). Here, by employing a cysteine-trapping approach for both Sa- and Eco-MscL, we found that these functional residues at the TM1/periplasmic loop interface can interact with a matching residue from a different subunit. Disulfide trapping of cysteine mutants at this site led to pronounced cross-linking to form dimers, and this trapping led to decreases in channel open dwell times (Fig. 2). This cross-linking is not likely to occur in a fully closed conformation, since the equivalent sites in the Mt-MscL crystal structure are at least 26 Å apart in the neighboring subunits (Fig. 1 A). As previously reported, asymmetric movements occur during normal Eco- and Mt-MscL channel gating that allow cysteines mutated in sites that are predicted to be distal in the Mt-MscL crystal structure to cross-link (32,33). Thus, cross-linking of channels that contain single-cysteine mutations at the TM1/periplasmic loop interface, where some dimers are observed in western blots, likely locks the channels into a gating-transition state in which they attain a fully open state that subsequently is destabilized, leading to shorter open dwell times.

The second interaction we identified was between the residue at the TM1/periplasmic loop interface and another residue in the nonconserved periplasmic loop. Similar interactions between the TM1 and loop from an adjacent subunit in the counterclockwise direction have been proposed, based on molecular-dynamics simulations, to stabilize the opening of the Mt-MscL channel (14). In Sa-MscL, W59C in the periplasmic loop can cross-link with F47C at the TM1/periplasmic loop interface, which leads to a decrease in open dwell times (Fig. 2). Intersubunit interactions between these two cysteines are so efficient that all subunits in the channel complex are cross-linked to form almost exclusively pentamers, as visualized by western blot. Without changing channel mechanosensitivity, these interactions may trap the channel into a gating transition state in which the open state is destabilized.

Similarly, Eco-MscL A58C from the periplasmic loop can also cross-link with I49C at the TM1/periplasmic loop interface (Fig. 2). Although these interactions are not as efficient as those of Sa-MscL mentioned above, and thus show only monomers, dimers, and trimers on western blots, they are efficient enough to again trap the channel into a state in which opening is destabilized. Locking this interaction also causes a decrease in mechanosensitivity, suggesting that the channel is trapped into a more closed transition state. A previous study found that both MscL open dwell times and mechanosensitivity increased after trypsin, chymotrypsin, or pronase treatment of the periplasmic side of MscL channels in giant proteoliposomes, thus suggesting that this region serves as a spring element for gating (17). Interestingly, most of these cleavage sites are located near Eco-MscL A58, so it seems likely that these cleavages alter the local conformation and inhibit A58 from interacting with I49 of the other subunits.

In summary, we have demonstrated two different interactions that occur with residues previously reported to define MscL channel kinetics and mechanosensitivity: Sa-MscL F47 and Eco-MscL I49C at the TM1/periplasmic lipid interface. The data suggest that this site is very dynamic, changing environments and interacting with other subunits in the complex in various ways upon normal channel gating.