Evolutionary specialization of MscCG, an MscS-like mechanosensitive channel, in amino acid transport in Corynebacterium glutamicum

MscCG, a mechanosensitive channel of Corynebacterium glutamicum provides a major export mechanism for glutamate in this Gram-positive bacterium, which has for many years been used for industrial production of glutamate and other amino acids. The functional characterization of MscCG is therefore, of great significance to understand its conductive properties for different amino acids. Here we report the first successful giant spheroplast preparation of C. glutamicum amenable to the patch clamp technique, which enabled us to investigate mechanosensitive channel activities of MscCG in the native membrane of this bacterium. Single channel recordings from these spheroplasts revealed the presence of three types of mechanosensitive channels, MscCG, MscCG2, and CgMscL, which differ largely from each other in their conductance and mechanosensitivity. MscCG has a relatively small conductance of ~340 pS followed by an intermediate MscCG2 conductance of ~1.0 nS and comparably very large conductance of 3.7 nS exhibited by CgMscL. By applying Laplace’s law, we determined that very moderate membrane tension of ~5.5 mN/m was required for half activation of MscCG compared to ~12 mN/m required for half activation of both MscCG2 and CgMscL. Furthermore, by combining the micropipette aspiration technique with molecular dynamics simulations we measured mechanical properties of the C. glutamicum membrane, whose area elasticity module of KA ≈ 15 mN/m is characteristic of a very soft membrane compared to the three times larger area expansion modulus of KA ≈ 44 mN/m of the more elastic E. coli membrane. Moreover, we demonstrate that the “soft” properties of the C. glutamicum membrane have a significant impact on the MscCG gating characterized by a strong voltage-dependent hysteresis in the membrane of C. glutamicum compared to a complete absence of the hysteresis in the E. coli cell membrane. We thus propose that MscCG has evolved and adapted as an MscS-like channel to the mechanical properties of the C. glutamicum membrane enabling the channel to specialize in transport of amino acids such as glutamate, which are major osmolytes helping the bacterial cells survive extreme osmotic stress.


Results
Structural features and molecular evolution of MscS-like mechanosensitive channels in Corynebacterium species. Corynebacterium glutamicum possesses two MscS-like channels (MscCG and MscCG2) 19,25 and one MscL-like channel CgMscL whereas E. coli has six MscS-like channels (MscS, KefA, YbdG, YbiO, YnaI, and YjeP) and one MscL channel. Note that the crystal structures of the E. coli MscS and Mycobacterium tuberculosis MscL channels were solved as a functional heptamer and pentamer, respectively (Fig. 1A). The primary amino acid sequence of the C. glutamicum MscCG (CgMscCG) has a large C-terminal extension (287-533 aa) not present in E. coli MscS. This C-terminal extension is characteristic of the MscCG-type channels found in Corynebacteria species (Supplementary Table S1). The third pore-lining transmembrane helix and the adjacent region known as the "MscS family conserved region" (shown in red in Fig. 1A) is also highly conserved in the MscCG channels, although several residues in this region important for the channel gating differ from the residues in MscS (Figs 1B, and S1A,B). In the MscCG-type channels the G113 residue important for the channel inactivation, is replaced by Ser 31 . Also, N117 residue, which is involved in the MscS inactivation 32 , is replaced by negatively charged Asp in Corynebacteria species. The characteristic fourth transmembrane helix at the C-terminal end is conserved as well in the MscCG channels across the species (Supplementary Fig. S1A,B).
Given that the MscCG channel functions as a glutamate exporter it is possible that MscCG-type channels have evolved differently from the canonical MscS channel. Note that the canonical MscS is the smallest protein in the MscS channel family, which does not necessarily mean that this channel is the ancestral channel. It is also possible that the ancestral channel was larger and that it had a more specialized function, whereas the current canonical MscS might be a structurally stripped-down version of a structurally larger MscS-like channel. To test this hypothesis, we performed phylogenetic analysis using the MscS conserved region. MscCG-type channels showed the best homology in this region among Corynebacteria species. Multiple sequence alignment revealed however, that MscCG (CgMscCG) and MscCG2 (CgMscCG2) are more similar to KefA (MscK), YjeP (MscM), YbiO, and YnaI, than to the canonical MscS (Figs 1B, S1A and Table S1). Consistently, a phylogenetic tree based on the maximum likelihood method supports the view that MscCG-type channels were grouped into a separate branch distant from the canonical MscS (Fig. 1C). Although 50 may not be a very good bootstrap score for most of the shown branches, MscCG2 is located in the same branch with YbiO with the bootstrap score of 78, which suggests that MscCG-type channels might have co-evolved with the YbiO channel, rather than evolved from the canonical MscS. Note that this result does not mean YbiO is the ancestral channel of MscCG and MscCG2. These  51 are shown on the left side. The predicted secondary structures of MscS homologs in C. glutamicum and E. coli, and the predicted secondary structures of MscL homologs in C. glutamicum, M. tuberculosis, and E. coli are shown on the right side, respectively. Orange bars indicate predicted transmembrane helices by TMHMM (http://www. cbs.dtu.dk/services/TMHMM-2.0/) and red bars indicate the MscS family conserved region including the poreforming transmembrane helix. (B) Sequence alignment of the MscS family conserved region of C. glutamicum CgMscCG, CgMscCG2, and six MscS paralogs of E. coli. The black bar indicates the pore-lining transmembrane helix, and the two glycine residues G113 and G121 known as kink and the asparagine residue N117 important for the inactivation shown in MscS are highlighted. (C) A phylogenetic tree built by the maximum likelihood method using the MscS conserved domain. Orange box shows clustered CgMscCG-type channels. The number at the branches indicates bootstrap value larger than 50, out of 100 bootstraps. Note that the bootstrap values are only an indication of the node quality in the phylogenetic tree, but not for the phylogenetic tree shape. Also note that MscCG2 is located in the same branch with YbiO with the bootstrap score of 78. results indicate that MscCG channels may have evolved as paralogs after the gene duplication of the MscS gene and acquired more specialized functions different from the common ancestral gene.
Preparation of C. glutamicum giant spheroplasts. Similar to the preparations of giant spheroplasts in E. coli 8 , Vibrio cholerae 33 , and Bacillus subtilis 34 , lysozyme was also used in this study to digest the peptidoglycan layer during the preparation of giant spheroplasts in C. glutamicum. However, since C. glutamicum has an extremely thick cell-wall with an outer lipid-rich layer containing mycolic acids and arabinogalactan on top of the peptidoglycan layer, lysozyme alone was insufficient for digestion of the cell-wall. In addition to lysozyme we had to use ethambutol to remove the mycolic acids and arabinogalactan layer from C. glutamicum cells before treating the cells with lysozyme. C. glutamicum cells were grown in the peptone-yeast-glucose (PYG) medium. The cells became round after 21 h in the culture medium containing 2.5 mg/ml ethambutol. The peptidoglycan layer of the round cells was digested after 1 h upon addition of 10 mg/ml lysozyme, which also resulted in increased number of round cells demonstrating that the cell-wall could be digested by the combination of ethambutol and lysozyme ( Fig. 2A).
The size of the round cell wall-less cells of ~1 μm in diameter was not large enough for the patch-clamp recording. To enlarge the cells further to ≥5 μm in diameter we let them grow for several days in the presence of ethambutol and ampicillin to inhibit the formation of the peptidoglycan layer and the synthesis of the arabinogalactan layer. Since the cells were osmotically very fragile, we cultured the cells in the presence of high concentration (0.8 M) of NaCl to prevent cell lysis. The cells grew gradually and their size plateaued at ~5 μm in diameter after 3 days ( Supplementary Fig. S2A). We confirmed that these enlarged cells were C. glutamicum by 16S rRNA gene sequencing analysis (data not shown). When looking at the enlarged cells under a phase contrast microscope, we observed two types of cells having either black or white appearance, respectively ( Supplementary Fig. S2B). To investigate whether the surface structure of the enlarged cells was clean enough to apply the patch-clamp technique, we examined them by transmission electron microscopy. As previously reported, we observed the thick cell-wall of C. glutamicum in the normal cell culture. Even though the enlarged cells had a different amount of the cell wall left, we were able to observe some enlarged cells with a relatively thin cell wall (Fig. 2B). Since some cell wall was still present in the enlarged cells, we named them "giant spheroplasts" in analogy to the giant spheroplasts from E. coli cells 8 . Mechanosensitive channels recorded in giant spheroplasts of C. glutamicum. To record mechanosensitive channel currents in the giant spheroplasts of C. glutamicum we first tested whether a tight gigaohm seal between the cell membrane and the glass wall of a patch pipette could be established. We attempted sealing on the giant spheroplasts of black appearance identified in the recording chamber with a phase-contrast microscope and were able to obtain a seal of ~1 GΩ (Fig. 2D). Moreover, we could excise the spheroplast membrane patch in the inside-out recording mode by snapping the pipette (Supplementary Video S1). In contrast, the spheroplasts of white appearance were difficult to seal on because their membrane was quite fluid and not amenable to form a gigaohm seal due to frequent vesiculation of the spheroplast membrane upon application of pipette suction (Supplementary Video S2). Consequently, we used only black giant spheroplasts for mechanosensitive channel recording in this study.
In order to detect mechanosensitive channel activities in the spheroplast membrane, we applied ramp pressure protocol that has been used in previous studies of E. coli MscS and MscL 1,35 . E. coli MscS and MscL were named as mechanosensitive channel of small conductance and mechanosensitive channel of large conductance, respectively, since these channels have significantly different conductance with 1 nS and 3 nS. Similarly, we were able to observe three different channel conductances in the C. glutamicum membrane upon the application of the ramp pressure. When the ramp pressure applied to the membrane gradually reached ~−100 mmHg at the pipette voltage of +30 mV, small currents of ~10 pA were observed first (Fig. 2D) and then after the saturation of these currents, larger currents of ~30 pA and ~100 pA were elicited at ~−210 mmHg (Fig. 2D). These three types of mechanosensitive channel activity ceased completely when the negative pressure was released (Fig. 2D). The smallest currents of ~10 pA are consistent with the conductance of MscCG reported previously 23,26,27 . To confirm the existence of MscCG in the spheroplast membrane, we recorded mechanosensitive currents in the giant spheroplasts of the MscCG-knockout strain. As expected, only two types of mechanosensitive channel activities of ~30 pA and ~100 pA corresponding to MscCG2 and CgMscL were observed at the pipette voltage of +30 mV (Fig. 2E), suggesting that the 10 pA small currents corresponded to MscCG. The 30 pA and 100 pA currents of MscCG2 and CgMscL, respectively, are comparable to that of E. coli MscS and MscL reported previously, and there is no other mechanosensitive gene found in the C. glutamicum genome. Note that MscCG currents were always elicited at the lowest pressure among the three types of mechanosensitive channels and did not show any sub-conductance states whereas the MscCG2 currents were elicited at almost the same level of pressure to activate CgMscL and exhibited frequent sub-conductance levels (Fig. 2D,E). From a number of experiments, we determined the average single channel conductance by calculating current amplitude divided by applied voltage. The calculated conductances of MscCG (n = 3), MscCG2 (n = 3), and CgMscL (n = 3), were 0.34 ± 0.01, 1.00 ± 0.11, and 3.70 ± 0.27 nS, respectively (Fig. 2F). This result shows that MscCG2 and CgMscL are comparable to the conductance of E. coli MscS (~1 nS) and MscL (~3 nS), and that these three channels can be identified by their +30 mV. The channel conductance was calculated as the single channel current divided by the voltage. Bars show standard errors (n = 3). (*Student's t-test, P < 0.05, **Student's t-test, P < 0.01). (G) Activation thresholds of CgMscCG and CgMscCG2. The pressure required for the first activation of CgMscCG (P CG ), CgMscCG2 (P CG2 ), and CgMscL (P L ) was measured as shown in D and E, and the activation threshold ratio against P L was calculated. Bars show standard errors (n = 5) (*student's t-test, P < 0.05).
SCIeNTIFIC REPORTS | (2018) 8:12893 | DOI:10.1038/s41598-018-31219-6 conductances. To evaluate mechanosensitivity of MscCG, MscCG2, and CgMscL, we measured the threshold pressure required for the first channel opening observed in a recording (P CG , P CG2, and P L ) from which we calculated the activation threshold ratio of each channel relative to the CgMscL threshold. P L /P CG was 2.19 ± 0.35 (n = 3), whereas P L /P CG2 was 1.06 ± 0.02 (n = 5) (Fig. 2G), indicating that MscCG2 requires much higher pressure for activation than MscCG given that its activation threshold is almost identical to MscL.
We further determined the membrane tension required to activate MscCG and CgMscL in the C. glutamicum native membrane since bacterial mechanosensitive channels have been shown to be activated by membrane tension, but not pressure 36 . To determine membrane tension in spheroplast patches we measured the radius of curvature (r) of the giant spheroplast membrane patch using a confocal microscope equipped with the patch-clamp system. To calculate membrane tension (T), we used the Laplace's law (T = Pr/2) 37 , which relates membrane tension T to the applied negative pressure P and the radius of the patch curvature r (Fig. 3A). The amount of applied suction was measured as negative pressure (P) by a piezoelectric pressure transducer (Omega Engineering). Given the minimum recording interval time of 0.7 s in our confocal microscope system, which is much longer compared to the time resolution of the patch-clamp recording, we applied a slow pressure ramp to activate MS channels and recorded the images of the membrane patch at the exact time when channels started to gate first in order to determine the corresponding radius of the membrane curvature. MscCG was activated at ~−85 to ~−105 mmHg within the time interval b (Fig. 3A,B). At that time, the radius of the patch curvature was 0.9 μm, which corresponded to membrane tension between ~5.1 and ~6.3 mN/m. The CgMscL channels were activated at higher negative pressure of ~−226 to ~−240 mmHg within the time interval c when the radius of the patch curvature was 0.8 μm (Fig. 3A,B) corresponding to the threshold of activation by membrane tension between ~12.1 and ~12.8 mN/m. Note that although the radius of the membrane curvature did not change much at the activation points for MscCG and CgMscL in our experiments, the pressures required to activate each channel were very different. These values are comparable to the activation thresholds of MscS and MscL in azolectin liposome membrane, which were previously determined as ~6 mN/m and ~12 mN/m, respectively 37 . However, the activation threshold determined for MscCG in the C. glutamicum membrane (~5.1 to ~6.3 mN/m) is much smaller compared to the threshold of 9-20 mN/m required for MscS as previously determined in the E. coli giant spheroplast membrane 38 . MS channels sense the force from membrane lipids 7,28 , and thus, it is expected that the lipid composition and physical properties of the membrane bilayer would affect the mechanosensitivity of these channels. For example, activation thresholds of E. coli MscS and MscL differ significantly between azolectin liposomes and E. coli spheroplasts 33,34 . One of the explanations for this difference is that the areal elasticity modulus K A of the E. coli spheroplast membrane is three times larger than the elasticity modulus of liposome membranes. The C. glutamicum membrane consists mainly of cardiolipin and phosphatidyl-glycerol, which is very different from the E. coli membrane consisting mainly of phosphatidyl-ethanolamine and phosphatidyl-glycerol 39 . We measured the mechanical properties of the C. glutamicum membrane using the micropipette aspiration (MA) method as previously done for the liposome membranes 40 . In the MA method, membranes are pulled into a pipette and the length of the pulled membrane inside the pipette (displacement) is measured after applying a constant negative pressure. Using a recently developed framework of patch-fluorometry for azolectin liposome membranes 40 , we were able to determine the areal elasticity modulus K A of both the C. glutamicum and E. coli membranes. Both membranes were pulled into pipettes by applying −10, −20, −30, and −40 mmHg pressure steps (Fig. 4A, and Supplementary Videos S3 and S4). The K A was calculated by plotting the relationship between the membrane tension and the areal strain (Fig. 4B). The K A of the C. glutamicum membrane was calculated as 14.7 ± 2.2 mN/m (n = 3), whereas the K A of the E. coli membrane was 43.7 ± 2.3 mN/m (n = 3) (Fig. 4C), suggesting that the C. glutamicum membrane is much softer than the E. coli membrane.

MD simulations of mechanical properties of the C. glutamicum and E. coli membranes.
Given that the C. glutamicum membrane consists of only negatively charged lipids, we employed MD simulations to examine whether the difference observed between the C. glutamicum and E. coli membranes stem from their membrane lipid composition. We modelled the C. glutamicum membrane based on its lipid composition previously reported 39 according to which the lipid composition was 50.2% 1-palmitoyl-2-oleoyl-sn-glycero-3 -phosphoglycerol (POPG), 30.2% 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoinositol (POPI), and 19.6%, cardiolipin (CL) molar ratio (Fig. 4D). The E. coli membrane model was created in parallel based on a previous study 41 , according to which the molar ratio of phosphatidyl-ethanolamine to phosphatidyl-glycerol was ~4:1. Both membrane bilayer models were stretched in MD simulations by surface tensions comparable to those applied in the MA experiments, i.e. 5, 10, 15, and 20 mN/m for the E. coli membrane and 5, 7.5, 10, and 12.5 mN/m for the C. glutamicum membrane. The results showed that the response of both membrane models to membrane tension was quasi-linear (Fig. 4E). The K A was 42.1 ± 1.5 mN/m for C. glutamicum membrane and 87.5 ± 9.5 mN/m for E. coli membrane, respectively (Fig. 4F). Although individually larger than the experimentally determined values of K A both values are comparable and of the same order of magnitude. These results suggest that the mechanical properties of the C. glutamicum membrane are comparable to those of liposome membranes rather than to the E. coli membrane. They also indicate that the membrane softness of C. glutamicum membrane originates from its lipid components.

MscCG channels exhibit voltage-dependent gating hysteresis in C. glutamicum but not in E. coli.
MscCG channels have been reported to function as "major glutamate exporters" in C. glutamicum, although interestingly, not as efficiently when expressed heterogeneously in E. coli 26 . The gating hysteresis defined as different gating thresholds between opening and closing has been suggested to be important for continuous glutamate export in C. glutamicum 26,27 . Thus, we examined to which extent the gating hysteresis of MscCG differs in the C. glutamicum and E. coli membranes. Since the MscCG currents can be distinguished from other currents of similar conductance, and MscCG is always activated at the lowest pressure among the three types of mechanosensitive channels, we were able to measure the single level of MscCG currents in the WT strain. We applied a slow ramp pressure that includes the same rate of loading and unloading to both membranes. MscCG opened at −34 mmHg and then closed at −14 mmHg in the C. glutamicum membrane whereas in the E. coli membrane, the opening and closing thresholds were −51 mmHg and −64 mmHg, respectively (Fig. 5A,B, black traces). Importantly, this gating hysteresis of MscCG was enhanced significantly only in the C. glutamicum membrane when a high pipette voltage was applied. At the pipette voltage of +100 mV (corresponding to the membrane potential of −100 mV in the inside-out patch recording mode) most of the MscCG channels remained open even when the pressure was completely released (Fig. 5A,B, red traces). To quantify the voltage effects on the gating hysteresis, we plotted the closing threshold/opening threshold ratio determined at pipette voltages from −100 mV to +100 mV (Fig. 5A,B). In the C. glutamicum membrane, the ratio became significantly decreased as the pipette voltage was increased. On the other hand, the gating hysteresis of MscCG did not change significantly and was even occasionally reversed in the E. coli membrane (Fig. 5B). These results suggest that the gating hysteresis observed with MscCG is voltage-dependent only in the C. glutamicum membrane. Note that the opening threshold of MscCG is unaffected by voltage, while the closing of MscCG channels slowed down dramatically at high positive pipette voltages resulting in the channels remaining open even upon release of suction applied to the patch membrane.
The current rectification of MscCG is also present in the C. glutamicum membrane as previously reported for MscCG recorded in E. coli membrane 23 . MscCG currents were significantly smaller (~70%) at negative pipette voltages compared to positive pipette voltages (Fig. 5C). This ~70% reduction was also observed in E. coli membrane (Fig. 5D). The current-voltage plots show the slope conductance of 286 pS and 95 pS in the C. glutamicum membrane at positive and negative pipette voltages, respectively, whereas the slope conductance in the E. coli membrane was larger at both voltages being 309 pS at positive and 107 pS at negative voltages (Fig, 5C,D).

Discussion
We developed a C. glutamicum giant spheroplasts preparation amenable to the patch clamp technique and characterized for the first time the gating of endogenous mechanosensitive channels, MscCG, MscCG2, and CgMscL. Surprisingly, the electrophysiological characteristics of MscCG and MscCG2 were remarkably different, supporting the view that these two channels might have a different evolutionary origin as members of the MscS-like channel family. Given that most of the MscS-like channels have complex structure by having more than three transmembrane helices characteristic of the canonical MscS of E. coli and were shown to be involved in more specialized processes such as calcium signaling in fission yeast cellular signaling (Msy1 and Msy2) 9 and apoptosis signaling in plants (MSL10) 42 , MscCG channels have four transmembrane helices and thus seem to have acquired more complex structure and gating characteristics for their more specialized function in C. glutamicum. Considering the differences in the gating and physiological function between MscCG and MscS it is likely that the MscCG channels have evolved after gene duplication of the mscS gene and "evolutionary tinkering" adding the fourth transmembrane helix and an extended C-terminal domain to enable its function as a glutamate exporter in C. glutamicum.
The electrophysiological characterization of MscCG functional properties in this study strongly supports the view for specialized function of MscCG in C. glutamicum cells. In addition to its low activation threshold of ~6 mN/m that is roughly half the activation threshold of CgMscL and MscCG2, MscCG lacks desensitization, inactivation and frequent sub-conducting states observed in the MscCG2 channels. All these characteristics are expected to contribute to a continuous export of glutamate as long as membrane tension keeps the channels open. Furthermore, MscCG exhibited a pronounced gating hysteresis in the C. glutamicum membrane characterized by a significant delay in the channel closing upon release of negative pressure applied to a spheroplast patch at low membrane potentials around −100 mV (corresponding to a pipette voltage of +100 mV in the inside-out patch recording mode). Since the resting membrane potential in a C. glutamicum cell is about −170 mV (corresponding to V p = +170 mV in the inside-out excised patch configuration), MscCG may stay longer open at such high membrane potentials enabling more efficient export of glutamate facilitated by the absence of inactivation in MscCG due to mutations in residues shown to be important for the MscS channel inactivation 6 . Significantly, our data demonstrate the presence of strongly voltage-dependent gating hysteresis in MscCG, which was only observed with channels recorded in C. glutamicum giant spheroplasts and not with channels heterologously expressed in E. coli giant spheroplasts. This result indicates that although the hysteresis may be an inherent property of the channel protein itself as recently shown for the MscS channels 27,43 , in the MscCG case the hysteresis appears to be largely due to the properties of the negatively charged lipid membrane of C. glutamicum. How voltage may influence the closing of MscCG channels is not known. It is however, possible that the evolutionary modification through addition of a negatively charged loop connecting the TM3 and TM4 helices in the C-terminal domain may be the structural component of MscCG interacting electrostatically with the membrane. Given the importance of the C-terminal domain for the MscCG gating 26 , it is likely that the electrostatic interaction between the negative charges of the loop and the negatively charged membrane lipids is modulated by the cell membrane potential, which would affect the open probability and gating hysteresis of MscCG in a way promoting the glutamate export at high membrane potentials.
The C. glutamicum cytoplasmic membrane is made of negatively charged phosphatidyl-glycerol (PG), phosphatidyl-inositol (PI), and cardiolipin (CL). Our MA experiments and molecular dynamics simulations both indicate a mechanically very soft membrane. It has previously been shown that a lipid bilayer made of phosphatidyl-glycerol (PG) alone is much softer than a bilayer made of phosphatidyl-ethanolamine alone (PE) 41,44 , which suggests that the softness of the C. glutamicum membrane results from the negatively charged lipids. An interesting fact is that all mechanosensitive channels, MscCG, MscCG2, and CgMscL, are activated within a physiological range of membrane tension (≤20 mN/m) in this soft membrane. In Mycobacterium tuberculosis, the mechanosensitivity of MtMscL requires the presence of a small amount of PI to be activated within this physiological range of membrane tension, since it was shown that it is difficult to activate MtMscL when expressed in E. coli giant spheroplasts 45 . In support of this finding MtMscL was shown in mass-spectrometry experiments to associate with PI lipids 46 . The effects of the negatively charged lipids on the gating of MscCG thus require further study.
In summary, the gating of endogenous mechanosensitive channels, MscCG, MscCG2 and CgMscL found in C. glutamicum cytoplasmic membrane was characterized using the patch clamp recording of the channel activity from giant spheroplasts developed in this study. Our data strongly suggest that not only the properties of the MscCG channel protein but also its interactions with the lipid membrane environment have been evolutionarily fine-tuned towards the specialized function of this MscS-like channel as a glutamate exporter. Since C. glutamicum bacteria have been utilized for industrial glutamate production exceeding 2 billion tons per year 21 , the electrophysiological characterization of MscCG described here may in addition to contributing to a better understanding of the amino acid export mechanism facilitated by these channels also help to further improve the biotechnological process used industrially for amino acid production.

Materials and Methods
Strains, media and growth. The WT strain of Corynebacterium glutamicum ATCC13869 was used for all the experiments. The ΔMscCG strain of Corynebacterium glutamicum ATCC13869 generated for a different study 19 was used in our study. Vegetable peptone (Oxoid, VEGETABLE PEPTONE No 1, Code: VG0100) was used for CM2B-V, PYG-V, and GP-V medium. After autoclaving, precipitation of the Vegetable peptone was observed, and all media were resuspended by shaking gently before the use. Cells were incubated at 31.5 °C, 135 rpm for aerobic culture with the incubator INNOVA 44 (Eppendorf, Germany) and collected with the centrifuge Sorvall Legend RT (Thermo scientific).
Phylogenetic analysis. MscCG homologs were collected with the BLAST program (https://blast.ncbi.nlm. nih.gov/Blast.cgi) and candidates were screened based on the annotation. Multiple sequences were aligned with the CLUSTALW program (http://www.genome.jp/tools-bin/clustalw) as well as manually, and the sequences of the MscS conserved region were extracted to construct the phylogenetic tree with the TREE program (http:// www.genome.jp/tools-bin/ete).  (Axon Instruments), and data were acquired at a sampling rate of 5 kHz with 2 kHz filtration. Pressure ramp activating the channels was applied to patch pipettes using a high-pressure clamp apparatus (ALA Scientific Instruments, USA; HSPC-1) and was monitored using a piezoelectric pressure transducer (Omega Engineering). The channel currents were analysed using pClamp 10 analysis software (Molecular Devices, Sunnyvale, CA). To obtain a better seal by improving salt bridge between the patch membrane and the pipette glass wall, magnesium and calcium concentration was increased. The open probability from all channels was obtained by the single-channel analysis using Clampfit software.
Micropipette aspiration method. The images of the spheroplast patch membrane of C. glutamicum and E. coli were observed with Zeiss LSM 700 confocal microscope using a long working distance water immersion objective (63×; NA 1.15; Carl Zeiss) in a Faraday cage. To visualize the membrane, the pipette tip was bent ~30 °C with a microforge (Narishige; MF-900) to make it parallel to the bottom of the chamber. The areal elasticity modulus K A was determined from the slope of the membrane tension, vs. areal strain graph using the following equation: SCIeNTIFIC REPORTS | (2018) 8:12893 | DOI:10.1038/s41598-018-31219-6 Membrane tension, T, was calculated based on the principle of surface chemistry (Laplace's law) and is expressed as follows, where R d is the radius of curvature of the membrane patch. The areal strain (fractional area change) is conventionally calculated by measuring the relative change in the apparent area of the expanded membrane inside the pipette, A (after the application of suction), with respect to its initial area, A 0 (zero pressure). The areal strain is thus defined by Using geometric relations, the area after the application of suction (A) was calculated depending on the shape of the pipette. If the pipette shape is conical like in our experiments, then the membrane surface area is calculated as where L is the protrusion length, L; r and R are the pipette radii at the pipette tip and at the dome, respectively. h is the height of the dome of the patch.

Molecular dynamics simulations.
All MD simulations were performed with NAMD2.10 47 . Visual Molecular Dynamics (VMD) 48 and Pymol (Molecular Graphics System, Version 1.8 Schrödinger, LLC) were used for all visualizations. CHARMM-GUI was used to generate the C. glutamicum membrane lipid system 49 . We used a model of top 6 lipid types found in the E. coli cell membrane as previously modelled 41 . The C. glutamicum membrane was modelled based on its main membrane lipid composition 39 . The lipid was solvated by adding 20 Å water boxes on the top and bottom of the membrane along the thickness direction. The TIP3P water molecule model was used along with the SOLVATE program (https://www.mpibpc.mpg.de/grubmueller/solvate). The lipid heads and tails were, in turn, randomized for ~1 ns at 298 K, while the rest of the system was fixed in an NVT system. Then the whole system was equilibrated for 30 ns with a time step of 1 fs with no restraints in an NPT ensemble. To gain the areal elasticity modulus K A of the C. glutamicum and E. coli membrane lipid bilayers, they were both stretched with the surface tensions close to their experimental range in different sets of simulations for 25 ns with a time step of 1 fs and in NPT ensembles. The E. coli bilayer was simulated under the surface tensions of 5, 10, 15, and 20 mN/m. The C. glutamicum bilayer was, however, simulated under the surface tensions of 5, 7.5, 10, and 12.5 mN/m due to its lower lytic tension observed in the MA experiments. Each simulation was repeated three times for both C. glutamicum and E. coli membranes. In all the simulations, the particle-mesh Ewald (PME) method was used for computing electrostatic interactions beyond a real-space cut-off of 1.2 nm with a Fourier grid spacing of 0.1 nm. A modified Nosé-Hoover Langevin piston pressure control provided in NAMD was utilized to control fluctuations in the barostat (at 1 atm). This method is combined with a method of temperature control (at 298 K) (Langevin dynamics) to simulate the NPT ensemble. The CHARMM36 (C36) Force field was used for all MD calculations. For data analysis, a TCL code was developed and used to calculate the average surface area of the lipid through different trajectories in different simulations. This code was employed to average the area over N (e.g., 10) slabs along the bilayer thickness direction (between the phosphate groups) as the lipid area is not uniform across the thickness direction. Then an average of areas over different trajectories is calculated as follows, M is the number of steady trajectories. A i (i = 1, 2, …, N) is the cross-sectional area of the lipid along the bilayer thickness (Fig. 4D). When the bilayer is stretched, the first few trajectories can be excluded from the deformed area analysis, to make sure the steady surface area is measured. The K A , was calculated using Eqns 1 and 2 for analyzing the micropipette aspiration data. To be able to obtain a statistically significant result the K A values of the two models were compared and each four data points (e.g. at surface tensions of 5, 10, 15, 20 mN/m) were assumed as one data set.