Mutations in a Bacterial Mechanosensitive Channel Change the Cellular Response to Osmotic Stress*

MscL is a channel found in bacterial plasma membranes that opens a large pore in response to mechanical stress. Here we demonstrate that some mutations within this channel protein (K31D and K31E) evoke a cellular phenotype in which the growth rate is severely depressed. Increasing the osmolarity of the growth medium partially rescues this “slowed growth” phenotype and decreases an abnormal cytosolic potassium loss observed in cells expressing the mutants. In addition, upon sudden decrease in osmolarity (osmotic downshock) more cytoplasmic potassium is released from cells expressing the mutants than cells expressing wild-type MscL. After osmotic downshock, all cells remained viable; hence, the differences in potassium efflux observed are not due to cell lysis but instead appear to be an exaggeration of the normal response to this sudden change in environmental osmolarity. Patch clamp studies in native bacterial membranes substantiate the hypothesis that these mutant channels are more sensitive to mechanical stresses, especially at voltages approaching those estimated for bacterial membrane potentials. These data are consistent with a crucial role for MscL in the adaptation to large osmotic downshock and suggest that if the normally tight regulation of MscL gating is disrupted, cell growth can be severely inhibited.

The ability to detect mechanical force, mechanosensation, is an inherent property of essentially all living organisms. Many animals, including humans, need mechanosensation for the senses of touch, hearing, and balance, as well as for regulating cardiovascular function; plants have the ability to detect wind and gravity; even many microorganisms have a primitive sense of touch or require mechanosensation for adapting to changes in osmolarity. Many studies in a variety of organisms have suggested that mechanosensitive (MS) 1 channels play an important role in mechanosensation (1)(2)(3)(4). Despite their obvious importance to life, little is known about how MS channels function. Although ligand-and voltage-gated channels have had many model systems for a number of years, for mechanically gated channels, the first and currently the only gene shown to encode an activity is mscL, originally isolated from Escherichia coli in 1994 (5) (see Refs. 6 and 7 for recent reviews).
Bacteria may provide a system for determining general principles, applicable to many systems, of the structure and function of ion channels. Several bacterial genes predicted to encode voltage-gated ion channels apparently share molecular mechanisms and structural elements with channels in several other systems (8). MS channel activities have also been found in many bacterial species (9 -14). In E. coli, MS channel activities have been studied in native membranes (9) and in reconstituted proteoliposome systems (15). Three classes of MS channel activity have been reported in this organism (16) representing at least two discrete molecular entities (17). One of these activities is MscL (for mechanosensitive channel of large conductance), which has fast kinetics and a conductance of greater than 2.5 nS; an activity with smaller conductance (0.8 nS), MscS; and an MS channel activity with the smallest (mini) conductance (0.4 nS), called MscM. The observation that MscL could be solubilized, fractionated through a chromatographic column, reconstituted into artificial liposomes, and yet retain functional channel activities as measured by patch clamp (15,17) permitted the isolation of the protein responsible for this activity and ultimately, cloning of the corresponding gene (5). This and subsequent studies (18,19) demonstrated that mscL expression in a lipid bilayer is both necessary and sufficient for MscL channel activity; no other proteins are necessary. Structural studies have suggested that MscL is a pre-formed homohexamer (18) (but see Ref. 20) composed of subunits each containing two transmembrane domains (18).
Bacterial MS channels are thought to play a role in sensing and/or adapting to changing osmotic environments. Several studies have demonstrated that when bacteria are challenged with a rapid osmotic downshock, many of the smaller cytoplasmic components are jettisoned into the medium, yet the bacteria remain viable (21)(22)(23)(24)(25). Because of their large conductance, and gating by mechanical stress, bacterial MS channels have been thought to be a major pathway for this cytoplasmic efflux. It was, therefore, disconcerting to find that the mscL-null mutant had no obvious osmotic phenotype (7); a redundancy of function with the other MS channels (MscS and MscM), whose activities are still present in the mscL-null, seemed a likely explanation for the apparent lack of phenotype.
Recently, we generated a number of site-directed mutants with altered channel properties as assayed by patch clamp of native membranes (26). Here we demonstrate that one of these mutants, K31E, inhibits the growth of the bacterial cell. Analysis of this phenomenon by whole cell and electrophysiology has provided evidence that is consistent with MscL being one of the pathways for the normal release of cytoplasmic components upon osmotic downshock, and that Lys-31, a charge in or near the first transmembrane domain (18), is an important residue for normal gating of MscL in the bacterial cell.

EXPERIMENTAL PROCEDURES
Constructs, Strains, and Cell Growth-The K31D MscL mutation was generated from an M13 construct containing the mscL open reading frame (18) using the Sculpture IVM kit (Amersham) employing the following mutating oligonucleotide: GGC ATT CGG GGA TAT TGT CTC T. The mutation was confirmed by sequencing, then subcloned using the pB10a expression construct (5). Generation of other mutants has been described (26). All mutants were expressed in the pB10a construct (5). The mscL-null PB104 bacterial strain (18,26) was used to host these expression constructs. In all experiments shown, the wildtype control is a PB104 strain containing the p5-2-2 plasmid (pB10a containing the wild-type mscL open reading frame) (5). For plate phenotype studies (Fig. 1) and growth for electrophysiological studies (Table I and Fig. 7), cells were grown in Luria-Bertani medium (LB). For growth and whole cell physiological studies (Figs. [2][3][4], cells were grown in minimal medium, K10 (27): 46 mM Na 2 HPO 4 , 23 mM NaH 2 PO 4 , 8 mM (NH 4 ) 2 SO 4 , 0.4 mM MgSO 4 , 6 M FeSO 4 , 1 g/ml thiamine, 10 mg/ml histidine, leucine, threonine, valine, and isoleucine, 0.2% glucose, 10 mM KCl, and 100 M ampicillin. K10 was assumed to be 208 mosM. Isopropyl-␤-D-thiogalactopyranoside (IPTG) was added to 1 mM for the induction of wild-type and mutant mscL transcription; normal induction time was 1 h, 2 h for electrophysiological studies. By estimating the number of channels seen by patch clamp in an area of membrane equivalent to that of a single cell, we estimate that there are normally about 10 to 100 functional MscL channels in a normal bacterial cell, absolutely zero in the mscL-null, zero to six in a mscL-null that contains the p5-2-2 expression plasmid but has not been induced by IPTG, and 50 to 200 in a mscL-null containing the p5-2-2 plasmid and has been induced. PB104 without plasmid, with p5-2-2, or with empty plasmid, with or without IPTG, all had similar growth rates, demonstrating that the modest increase in MscL expression in this expression system (about 5 times endogenous levels) had little physiological consequences in normal laboratory conditions. No obvious differences in the number of functional channels per patch were observed between cells expressing the wild-type and mutant MscL; Western analysis has shown previously that the amount of MscL protein in induced cells expressing the wild-type was indistinguishable from the K31E mutant (26). All growth and warming of buffers for the potassium experiments was at 37°C.
Potassium Content and Efflux Experiments-To measure the potassium loss over 15 min, cells were grown to midlog phase in K10 with or without 325 mM NaCl. After induction of MscL expression for 1 h, 10 ml of culture was harvested by filtration onto a 47-mm diameter 0.45-m nitrocellulose filter (Millipore). The filter was placed on the wall of a 50-ml conical tube (Falcon) that contained 6 -10 ml of prewarmed (37°C) K0 (46 mM Na 2 HPO 4 and 23 mM NaH 2 PO 4 ). This tube was then mixed vigorously for 3 to 5 s, and a 1-ml sample was filtered onto a 25-mm diameter 0.45-m nitrocellulose filter (Millipore) and washed with 3 ml of 37°C K0. For the cells grown in the high osmolarity medium containing NaCl, the NaCl concentration was maintained in the K0 resuspension and wash buffer; no significant osmotic downshock was given. The mixing, filtering, and washing took approximately 20 to 25 s to complete, and therefore the first time point is labeled as 0.4 min. The second time point was taken at 15 min and was again 1 ml of sample filtered and washed. A 15-min time point was chosen because preliminary experiments suggested that a significant percentage of potassium loss that took place over the course of 30 min occurred within the first 10 to 15 min.
The potassium efflux assays were designed to measure the decrease of internal potassium due to osmotic downshock and were performed as described previously (22) with the following modifications. All cultures were grown in K10 with 325 mM NaCl, to midlog phase, and induced for 1 h. At harvest, cultures were kept at 37°C, 1 ml was filtered onto a 25-mm diameter 0.45-m nitrocellulose filter and washed with 3 ml of 37°C wash solution. The wash solutions contained K0, plus varying concentrations of NaCl. Note that as the concentration of NaCl decreases, the extent of osmotic downshock increases. Under these conditions, increasing the wash solution up to 20 ml had little effect on the values obtained suggesting that there is a phase of very rapid potassium release upon downshock that is essentially complete within the first 3 ml of wash. A normal 3-ml wash took approximately 3 to 5 s.
For both the potassium loss and the efflux studies, the filters were placed in a 50-ml plastic beaker and dried at 80°C to 95°C overnight, and then the solutes were resuspended in 3 to 4 ml of double-distilled water. The solution was then assayed for potassium by flame photometry (Buck Scientific, PFP7).
Cell Viability Studies-Cells grown in K10 to midlogarithmic growth were incubated for an additional 1 h induced or uninduced. The OD 650 of each culture was determined at harvest, and the cells were serially diluted into LB medium and plated in triplicate onto LB plates containing 100 M ampicillin. Colony forming units (CFU) were determined per OD 650 unit. To determine the viability after downshock, the same procedure was used except that the cells were grown in K10 with 325 mM NaCl, and dilutions were made in K0 with or without 325 mM NaCl; the first dilution of this series was 10 l of culture into 10 ml of buffer, thus providing the initial osmotic downshock for the cells diluted into buffer not containing NaCl.
Single-channel Analysis-Giant cells (28) were generated and used in patch clamp experiments (29) as described previously (9); recordings were from excised inside-out patches at room temperature and were performed at negative voltages (positive pipette voltages), but channel currents are presented as upward. The pipette solution was: 200 mM KCl, 90 mM MgCl 2 , 10 mM CaCl 2 , and 5 mM HEPES adjusted to pH 6.0; the bath solution was the same plus 0.3 M sucrose. Pressure was assessed using a pressure transducer (Micro Switch; Omega, Stamford, CT) calibrated by a pneumatic transducer tester (Bio-Tec, Winooski, VT) and is presented normalized to MscS (applied pressure/threshold pressure required to open MscS in the same patch) as previously established (26). Data were acquired with 10 kHz filtration and a sampling rate of 30 kHz and analyzed with pCLAMP6 software.

Expression of Mutant MscL with a Charge Change at the Lys-31 Position Leads to a Slowed-growth Plate Phenotype-
Cells expressing one of two site-directed MscL mutants, K31D or K31E, were assayed for a plate phenotype. Because the wild-type and mutated mscL genes were under the transcriptional control of an inducible promoter, expression could be chemically induced by IPTG. As expected, on plates that did not contain IPTG, all cells grew equally well ( Fig. 1, left). This growth was comparable to that of the wild-type strains or mscL-null mutants (not shown). However, on plates containing IPTG, the growth of cells expressing the K31D or K31E mutant was severely depressed (Fig. 1, right). No growth retardation was observed on IPTG plates for cells expressing K31R, K31C, or K31I (not shown), suggesting that the plate phenotype observed was due to the charge reversal at the Lys-31 position.

Cells Expressing the K31D or K31E Mutant MscL Show a Slowed-growth Phenotype in Liquid Medium That Is Partially
Reversed by Increasing the Osmolarity of the Medium-Growth rates of cells expressing one of these mutants or wild-type MscL were measured in a liquid medium ( Fig. 2A). Induction of expression led to a severe slowed-growth phenotype for cells expressing the mutants but not the wild-type MscL. Interestingly, the slowed-growth phenotype of the mutants in these liquid cultures did not appear as pronounced as on plates, presumably because of differences in the assays. Viability studies demonstrated that the slowed increase in OD 650 of liquid cultures expressing the mutants cannot be explained by death of a fraction of the population. In a typical experiment, the number of viable K31E-containing cells prior to induction (1.5 ϫ 10 8 CFU/OD 650 unit) remained the same after 1 h of Addition of 325 mM NaCl increased the growth rate of the cells expressing the mutant MscL (Fig. 2B). Although the growth rate of cells expressing the MscL wild-type was decreased with increasing NaCl concentrations (Fig. 3, top), as little as 100 mM NaCl increased the growth rate of the cells expressing the K31D and K31E mutants, and maximum growth rates were observed at 200 -300 mM NaCl (Fig. 3, bottom). As the concentration of NaCl increased to 500 mM, the growth rate of the cells expressing these mutants approached that of cells expressing the wild-type MscL.
To determine whether the remediation of the slow-growth phenotype was specific for NaCl, or a function of increased medium osmolarity, we tried two other osmolites: KCl and sorbitol. As seen in Fig. 4, all three osmolites inhibited growth of all uninduced cells. As expected, the growth rates of induced wild-type MscL cultures were similar to that of uninduced cells in all media tested. However, the poor growth of the cells expressing the K31D or K31E mutant was remediated by each of the osmolites. These data are consistent with the hypothesis that medium osmolarity is the stimulus for remediation of the poor growth phenotype.
IPTG-induced cultures that were placed in medium containing 325 mM NaCl grew as fast in the first 0.5 h as in the next 1.5 h (wild-type: 0.41 versus 0.38 h Ϫ1 ; K31D: 0.23 versus 0.23 h Ϫ1 ; K31E: 0.25 versus 0.21 h Ϫ1 for the two growth rates); there was no significant lag to the remediation of the slowed-growth phenotype. Therefore, it does not seem likely that time-depend-ent transcriptional and translational events are responsible for the osmotic remediation.

Cells Expressing K31D or K31E Mutant MscL Show an Abnormally Large Potassium Loss That Is Partially Reversed by
Increasing the Osmolarity-To determine if the cells expressing the mutants lost cytoplasmic solutes at an increased rate, we measured the intracellular potassium of cells after being placed in K0, a potassium-free buffer (Fig. 5). As expected, if uninduced cells were grown in K10, then placed in K0, they lost only a small proportion of their potassium over the course of 15 min (about 30%). In contrast, if the cells were induced with IPTG for 1 h prior to being placed in buffer, the cells expressing mutant MscL not only started out with less internal potassium (60% that of cells expressing wild-type MscL), but lost substantially more potassium in 15 min (cells expressing mutants lost 70 to 80%, cells expressing wild-type lost 30%). Uninduced cells grown at high osmolarity accumulated about 1.5 times as much internal potassium as those grown in lower osmolarity medium and lost only a small amount of potassium upon being placed in K0 for 15 min (10 -15%). Of induced cells grown at high osmolarity, those expressing the mutants started at internal potassium levels 85% of that of cells expressing wild-type MscL. Note that although the cells expressing mutants still lost more potassium than wild-type expressing cells (30% versus 12%), this loss was significantly less than the loss of potassium in low osmotic medium (30% versus 70 -80%). These data demonstrate that cells expressing the mutants lose more potassium than cells expressing wild-type MscL and that this loss is exacerbated by low osmolarity.
Upon Osmotic Downshock, Cells Expressing K31D or K31E Mutants Retain Less Cytoplasmic Potassium than Cells Expressing Wild-type MscL-When bacteria are challenged with a rapid osmotic downshock, many of the smaller cytoplasmic components are jettisoned into the medium, yet the bacteria remain viable (21)(22)(23)(24)(25). One of the characteristics of this phenomenon is that the solute efflux is very rapid and transient in nature (24). We therefore tested whether the cells expressing mutants lost more potassium than cells expressing wild-type MscL when subjected to such an osmotic downshock. In contrast to the experiments above where changes in potassium levels were measured over a long time course (15 min) after essentially no changes in osmolarity, in this experiment the cells were osmotically downshocked using the wash solution; hence, rapid changes in potassium concentration occurring in less than 5 s were being observed (see "Experimental Procedures"). As seen above in Fig. 5, when induced for 1 h after growth in high osmolarity medium, no significant differences were observed between the total amount of internal potassium in the cells expressing wild-type versus the K31D or K31E mutants (all contained about 300 M/OD 650 unit). If, however, the cells were exposed to an osmotic downshock, from the starting 858 mosM to less than 550 mosM, less internal potassium was measured. As shown in Fig. 6, less potassium remained in the cells expressing K31D and K31E mutants relative to the wild-type when the shocking buffer was between 600 and 200 mosM. These data cannot be attributed to differences in MscL expression because it has been previously demonstrated by Western blot of membrane proteins that the K31E mutant is expressed at levels indistinguishable from that of the expressed wild-type MscL (18). Nor can the results of these experiments be attributed to lysis of the mutant expressing cells because even at the largest downshock, the cells retained 100% viability

FIG. 5. Abnormal potassium loss from cells expressing the K31D or K31E MscL mutants is partially inhibited by high osmolarity.
Cultures were grown to midlog phase in K10 (top row) or K10 with 325 mM NaCl (bottom row) with (Induced; right column) or without (Uninduced; left column) IPTG. Cells were harvested and placed in buffer not containing potassium, and intracellular potassium levels were measured within 20 to 25 s (labeled as 0.4 min; f) and again at 15 min (u). Shown is the average Ϯ S.E. of 5 to 8 independent experiments; note that the S.E. for K31D induced, 15 min, is Ϯ3 and is too small to be seen on this graph.
(all values for wild-type and mutants before and after downshock were between 3.9 and 5.4 ϫ 10 8 CFU/OD 650 unit). Therefore, these data demonstrate that, upon osmotic downshock, the cells expressing K31D and K31E mutants exhibit a greater efflux of potassium than those expressing the wild-type MscL channel.

The K31D and K31E Mutant MscL Channels in Native Membranes Open at Lower Membrane Tension than Wild-type MscL in Patch Clamp
Studies-Single-channel analysis by patch clamp in native membranes (giant cells) was used to analyze the K31D, and K31E, MscL mutant channel activity relative to wild-type. Previously, we characterized the open-state dwell time and threshold sensitivity (relative to MscS opening) of several mutants and the wild-type MscL channel (26). These studies revealed that the K31E mutant had a shorter open dwell time and was more sensitive to membrane tension relative to the wild-type. Here, we have repeated these observations with consistent results (Table I). In addition, we characterized the K31D channel activities. As expected from the phenotypic studies above, we found that the single-channel properties of the K31D mutant were significantly different from the wild-type MscL and indistinguishable from the K31E mutant (Table I).
Because we are dealing with charges that are apparently in a transmembrane domain (18), the voltage potential across the membrane may have profound effects. All single channel recordings in our previous report (26) and in Table I were made at a Ϫ20 mV potential; however, several studies suggest that the resting membrane potential of E. coli is approximately Ϫ150 to Ϫ200 mV (30 -33). Unfortunately, native bacterial membranes are not stable in the patch electrode at very negative membrane potentials, and the addition of the pressure required to open the MS channels increases the stress and makes it difficult to acquire data at these voltages. However, once a pressure at which MscL activity is observed at Ϫ20 mV is clamped, the voltage potential can sometimes be changed to Ϫ100 mV for short periods of time, then returned to Ϫ20 mV without breaking the seal of the patch. When successfully performed with the wild-type channel, an increase in the probability of opening (P o ) is observed (Fig. 7A, left). Similar results were obtained with the K31E mutant (Fig. 7A, right); however, the increase in P o consistently seemed greater for the mutant relative to wild-type MscL. Quantitation of the fold increase in P o between Ϫ20 mV and either Ϫ80 or Ϫ100 mV demonstrated that as the voltage became more negative, approaching that of the normal resting potential of E. coli, the increase in P o was greater for the K31E mutant than wild-type MscL (Fig. 7B). These data are consistent with the hypothesis that a charge change at the Lys-31 position leads to more severe changes in single channel properties as voltages approach the bacterial resting potential; therefore, the data presented in Table I on the pressure sensitivity of the Lys-31 mutants may be a gross underestimate of the increase in pressure sensitivity of these mutants in vivo. DISCUSSION Previously, we expressed wild-type and mutant MscL channels in a mscL-null E. coli strain and identified several sitedirected mutations that led to changes in MS channel activities as assayed by patch clamp (26). We found that mutations at two sites, Lys-31 and Gln-56, could lead to changes in channel kinetics and/or shifts in the pressure sensitivity curve. We have assayed several of these mutants, including all that led to dramatic changes in channel dwell times or pressure sensitivity, for their ability to evoke a plate phenotype when expressed (not all are shown in this study). We previously found two mutations that led to channels that opened with less pressure, K31E and Q56P. These two mutants were the only ones that evoked an obvious plate phenotype. The Q56P mutation led to a slow growth plate phenotype that is less severe than the K31E mutation and is only seen at a lower temperature (22°C) (not shown). Patch clamp studies, however, initially suggested that Q56P was a more severe mutation than K31E because it had larger shift in the pressure sensitivity curve at moderate membrane potentials (26). Furthermore, under the patch clamp conditions used, the Lys-31 MscL mutant channels still opened at higher pressures than MscS. Therefore, the observed FIG. 6. Upon osmotic downshock, cells expressing K31D or K31E mutant retain less cytoplasmic potassium than cells expressing wild-type MscL. Cells were grown in K10 with 325 mM NaCl (858 mosM). All cells were induced with IPTG for 1 h. As presented under "Experimental Procedures," for each point an aliquot of the culture was filtered onto a nitrocellulose filter and then washed with K0 plus varying concentrations of NaCl to yield the desired osmotic downshock. The difference between the concentration of NaCl in the growth medium and the NaCl in the wash buffer constituted the "downshock." The filters were dried, and the potassium remaining in the cell was measured by flame photometry. Shown is the percent of maximum potassium concentration (100% ϭ no downshock) plotted against the osmolarity of the wash solution. Points and error bars are mean Ϯ S.E. of four independent experiments. From these four experiments, five spurious points that deviated greater than 30 times the standard deviation derived from the other 3 experiments were interpreted to arise from contamination and therefore were not used. All four experiments showed a leftward shift in this curve for cells expressing the mutants relative to those expressing wild-type MscL. Data from either wild-type or mutant mscL genes expressed in a mscL-null mutant strain are presented. K31D and K31E are substitutions of lysine at position 31 to aspartate or glutamate, respectively.
b The analysis was performed as previously described (26). All kinetic data were derived from at least 6 independent experiments (obtained from at least 2 independent spheroplast preparations) each containing at least 200 events that were fitted to a 3 open state model; 2 and 3 are the two longer time constants. The shortest for each group, which was always Ͻ0.3 ms, could not be measured accurately and is not shown. Mean Ϯ S.D. for all other values are shown. All groups had a minimum of 2000 total events for analysis. A previous study (26) of individual patches at different pressure stimulation demonstrated that channel mean open time did not vary significantly over the range of P 0 used for these experiments. c p Ͻ 0.001 from wild-type as determined by Student's t test. No significant difference, in any parameter measured, was observed between the K31D and K31E mutants.
increase in sensitivity to pressure does not seem sufficient to explain the phenotype if poor growth is due to excessive loss of cytoplasmic solutes; the critical pressure for solute release from MS channels ought to be set by the lower threshold of MscS and not MscL. Presumably, differences in environment between the intact cell and the membrane patch account for this discrepancy. One of these differences is the membrane potential. Here we present evidence that suggests that the pressure sensitivity of wild-type MscL is voltage-sensitive; the channel becomes more sensitive to membrane tension at more negative membrane potentials. This voltage dependence of the pressure sensitivity is apparently exaggerated for the K31E mutant; the pressure sensitivity of the K31E mutant increases disproportionately as the membrane potential approaches the estimated potential across the E. coli envelope. Thus, the discrepancy between electro-and whole cell physiology is probably at least partially due to the differences in membrane potential across the patched membrane versus the bacterial cell (although other differences, such as channel modulation in the whole cell, could also play a role in increasing the severity of the K31D and K31E phenotype). Presumably, the membrane potential influences the charge at position 31 because the charge is in the membrane (18) and therefore senses the electric field. Consistent with this interpretation is the observation that the reversal in charge is apparently the relevant feature because K31R, K31C, and K31I mutants do not show the slow-growth phenotype.
Several studies have demonstrated the rapid release of some of the cytoplasmic contents into the medium when a bacterial cell is subjected to an osmotic downshock (21)(22)(23)(24)(25). The components released include potassium, proline, glutamate, ATP, lactose, and trehalose. Many of these molecules are osmoprotectants that are either synthesized in, or transported into, the cytoplasm to a high concentration when growth is in high osmolarity medium. This rapid efflux is thought to be a means by which bacteria rapidly adjust to hypoosmotic stress. The initial discovery of channels in the membrane of E. coli with very large conductances and gated by tension (9) suggested pathways for this release of components. A few studies have provided evidence consistent with this hypothesis. MS channel activities in E. coli, with conductances consistent with those of MscM and MscS, have been observed in response to osmotic changes when assayed in the excised (34) as well as the wholecell (protoplast) (35) patch clamp configuration (MscL apparently was not observed because the greater stimulation required to open this channel is difficult to achieve in the wholecell configuration). Also, the MS channel activities (35,36) and the MscL protein (18,20) have both been localized to the inner membrane, the principal cellular barrier of the E. coli cell. Finally, it is known that gadolinium (Gd 3ϩ ) blocks MS channels in a wide range of systems including bacteria (4), and a direct correlation has been demonstrated between the concentration of Gd 3ϩ required to block bacterial MS channels and that required to block efflux of lactose and ATP from E. coli and ATP from Streptococcus faecalis in response to osmotic downshock (37). The Gd 3ϩ treatment did not stop the efflux of glutamate or potassium (although the rate of potassium efflux was shown to be slower in the presence of Gd 3ϩ ), suggesting that the major pathway for these molecules may normally be independent of the MS channels blocked in this study. None of these studies, however, is without its caveats, nor do the studies directly demonstrate that the MS channel activities observed in patch clamp play a role in the efflux response upon osmotic downshock.
One crucial piece of evidence to support the assertion that bacterial MS channels are the pathway through which bacteria jettison cytoplasmic components upon osmotic downshock was missing: a cellular phenotype. Given the previous evidence, one might expect to find a bacterial cell that is lacking MS channels to show differential growth with changing osmolarity or to exhibit decreased survival upon osmotic downshock. Probably because there is redundancy of function with other MS channels, no phenotype has yet been reported for the E. coli mscLnull mutant. However, several of the findings in this study, derived from gain-of-function mutants, now provide phenotypic evidence that MscL normally plays a role in osmotic adaptation. The "slowed growth" phenotype observed for the K31D and K31E MscL mutants is exacerbated by low osmolarity. Because of the rapid kinetics of growth recovery when cells expressing the mutants are placed in high osmotic medium, it seems unlikely that the partial rescue is an indirect effect such as the conditioning of cells to the higher osmotic medium. Previous studies of osmotic dependent changes in protein expression, including Kdp induction (38) and shifts in porin expression (39), have suggested that in healthy cells these changes are not detected until at least 10 min after cells are show the same patch after the membrane voltage was changed to Ϫ100 mV; note there is an increase in P o even though the pressure is the same as in the upper traces and that this increase appears larger for the mutant than the wild-type. The baseline (no MscL channels open) is marked in these traces with a dashed line; note that for simplicity, openings in all cases are shown as upward. The bottom traces are from the same patch at the same pressure several seconds after the membrane potential was returned to Ϫ20 mV. B, the fold increase in P o at different voltages as compared with that observed at Ϫ20 mV. The points and error bars are the mean Ϯ S.E. for 3 or more independent measurements; a total of 99 measurements are represented. placed in hyperosmotic medium, and maximal changes are not seen until 30 to 35 min. One might expect that these changes may even occur more slowly in the cells expressing the K31D or K31E mutants, which have a doubling time of about one-fourth of their normal rate. On the other hand, potassium pump activity has been reported to occur very rapidly when the osmolarity of the medium is increased (40). Perhaps this accounts for the observation that when KCl is used as an osmolite, the cells have a significantly faster growth rate than when NaCl or sorbitol is used ( Fig. 4; p Ͻ 0.05 but Ͼ 0.01). However, it seems unlikely that this increased potassium pumping activity can account for all of the data because the potassium loss observed over time in potassium-free buffer is partially inhibited by increasing the buffer osmolarity (Fig. 5); in this experiment there should be no potassium to pump into the cell. Hence, the simplest interpretation is that the partial growth rescue is at least partially due to a direct effect, perhaps a slight decrease in turgor force across the cell membrane in the higher osmotic medium, that allows for better growth of the mutants by eliminating improper gating of the mutant channel. In this scenario, the increase in turgor at lower osmolarity would cause the channel to open in vivo with some low frequency, transiently decreasing the turgor by causing the release of solutes and the breakdown of the proton gradient that serves as a major energy source (see Refs. 41-43 for reviews). Thus, the cells expressing these mutant MscLs would be in a state of catch-up in their metabolic energy, as well as having to resynthesize or recover any lost osmolites, metabolites, or amino acids. Preliminary results suggest that the average internal pH of a population of cells expressing the K31D and K31E mutants is similar to that of cells expressing wild-type MscL (not shown). This implies that the mutant MscL channels open infrequently in vivo and that the proton gradient, and presumably also the turgor and metabolic state, quickly recovers. This is perhaps not surprising, given the observation that cells expressing the K31D and K31E MscL still grow, albeit at a slower rate.
The increased efflux of potassium upon osmotic downshock strongly supports the model that, in vivo, the mutant MscL channels open more easily upon membrane tension caused by low or rapidly decreasing osmolarity. These studies are complemented by the in vitro patch clamp studies that demonstrate that the mutant channels gate more easily by suction in the pipette, especially at voltages approaching those predicted for bacterial membrane potentials. Hence, the data suggest not only that MscL is one of the pathways for cytoplasmic components to efflux upon osmotic downshock, but that if the normally tight regulation of MscL gating is disrupted, the rate of cell growth can be severely inhibited. Genes predicted to encode MscL homologues have been found in several bacterial species, including some pathogens (7). 2 Therefore, if misgating of MscL channels could be accomplished pharmacologically rather than genetically, it could be the genesis of a new class of antimicrobial agents.
A plate phenotype, such as that observed for the K31D and K31E mutants (Fig. 1), demonstrates that screens may be developed to select for functional mutations within MscL and other MS channels. The use of an expression plasmid that allows one to turn on or off expression within the cell and the ability to selectively mutagenize only the target gene are some of the advantages this system has for such studies. A simple screen for colonies that show a "slowed or no growth" phenotype when a randomly mutagenized MscL is expressed has already been employed, leading to the isolation of 19 single-site mutations. 3 Interestingly, one of the randomly generated mutations is at Lys-55, only one amino acid away from the Gln-56 mutants presented previously (26). More impressive, however, is the observation that 14 of the mutants isolated are clustered between amino acid residues 13 and 30, very close to the Lys-31 mutants described here. Seven of these fourteen not only are in this region but, like K31D and K31E, cause the MscL protein to become more electrically negative. Given the proximity of this critical region to the membrane (18), an obvious question is whether this protein domain contributes to the pore or gate of the channel. In the future, the use of genetic screens to identify new functional regions within MscL as well as to identify new proteins involved with osmotic adaptation, in combination with site-directed mutagenesis, whole cell physiology, and patch clamp techniques, promises to give insight into the physiological and functional significance of domains of the MscL channel protein and other MS channels in bacteria.