Lactococcus lactis Uses MscL as Its Principal Mechanosensitive Channel

The functions of the mechanosensitive channels from Lactococcus lactis were determined by biochemical, physiological, and electrophysiological methods. Patch-clamp studies showed that the genes yncB and mscL encode MscS and MscL-like channels, respectively, when expressed in Escherichia coli or if the gene products were purified and reconstituted in proteoliposomes. However, unless yncB was expressed in trans , wild type membranes of L. lactis displayed only MscL activity. Membranes prepared from an mscL disruption mutant did not show any mechanosensitive channel activity, irrespective of whether the cells had been grown on low or high osmolarity medium. In osmotic downshift assays, wild type cells survived and retained 20% of the glycine betaine internalized under external high salt conditions. On the other hand, the mscL disruption mutant retained 40% of internalized glycine betaine and was significantly compromised in its survival upon osmotic downshifts. The data strongly suggest that L. lactis uses MscL as the main mechanosensitive solute release system to protect the cells under conditions of osmotic downshift.

The functions of the mechanosensitive channels from Lactococcus lactis were determined by biochemical, physiological, and electrophysiological methods. Patchclamp studies showed that the genes yncB and mscL encode MscS and MscL-like channels, respectively, when expressed in Escherichia coli or if the gene products were purified and reconstituted in proteoliposomes. However, unless yncB was expressed in trans, wild type membranes of L. lactis displayed only MscL activity. Membranes prepared from an mscL disruption mutant did not show any mechanosensitive channel activity, irrespective of whether the cells had been grown on low or high osmolarity medium. In osmotic downshift assays, wild type cells survived and retained 20% of the glycine betaine internalized under external high salt conditions. On the other hand, the mscL disruption mutant retained 40% of internalized glycine betaine and was significantly compromised in its survival upon osmotic downshifts. The data strongly suggest that L. lactis uses MscL as the main mechanosensitive solute release system to protect the cells under conditions of osmotic downshift.
Mechanosensitive channels play an important role in prokaryotic cell volume regulation (1). In small, single-cell organisms, this regulation can mean the difference between life and death under extreme osmotic downshift conditions. By diffusion over the semipermeable cell membrane and/or aquaporins embedded in the membrane, water can enter and leave the cell until equilibrium is established between internal and external osmolality. This allows microorganisms to adapt to changes in external osmolyte concentrations. When the external osmolyte concentrations increase (hyperosmotic stress), water will leak out of the cell, causing loss of turgor, and ultimately the cell may plasmolyse. Bacteria respond to this hyperosmotic stress by rapid uptake of ions (K ϩ ) and/or compatible solutes or increasing the intracellular osmolyte concentration through synthesis of compatible solutes. The increase in internal compatible solute concentration compensates for the high external osmolality, allowing water to diffuse back and the cell to regain its original volume and turgor (2)(3)(4).
Conversely, when the external osmolyte concentration suddenly decreases, water will diffuse into the cell, causing it to swell and, in extreme conditions, lyse. This is where the mechanosensitive channels are thought to play a role by opening in response to the increased membrane tension effected by the rapid increase in cell volume. The best known example of a channel in this role is the mechanosensitive channel of large conductance from Escherichia coli (MscL Ec ), but homologues are present in most eubacteria (5)(6)(7). MscL opens near the lytic tension limit of the bacterial membrane. A second mechanosensitive channel, that of small conductance, MscS, has been characterized in only a few organisms (8,9). MscS opens at lower membrane tensions and has a smaller conductance than MscL, making it useful for fine regulation of internal compatible solute concentration. Crystal structures of MscL and MscS are available (10,11). The electrophysiological characteristics of MscS from E. coli (12) and of a number of MscL homologues (5) have been described. So far, their physiological roles in osmoprotection are not fully understood.
Previously, we have established the mechanism of the osmotic regulation of the upshift-activated glycine betaine transporter OpuA from Lactococcus lactis (13,14). Although the physicochemical properties of the membrane and lipid-protein interactions also play a critical role in the osmotic activation of OpuA, the mechanism is entirely different from that underlying the gating of MS 1 channels. We have shown that increasing internal ionic strength, a consequence of osmotic upshift, activates OpuA by altering the electrostatic interactions between anionic lipids and charged residues at the cytoplasmatic face of the protein (15). For a better understanding of the total osmoregulatory response of L. lactis, we present here the electrophysiological and biochemical characterization of L. lactis MscL (MscL Ll ; 45.4% identity with MscL Ec ; Fig. 1A) and YncB (hereafter referred to as MscS Ll ; 24.0% identity with MscS from E. coli; Fig. 1B) and show that MscL Ll is critical for protection of L. lactis against osmotic downshifts. In fact, the majority of glycine betaine, accumulated upon upshift activation of OpuA, seems to exit the cell via MscL Ll upon subsequent osmotic downshift. We also examined the expression of the opuA, mscL, and yncB genes during cell growth under low and high salt conditions.

Strains, Plasmids, and Growth Conditions
Experiments were performed using the strains listed in Table I and  plasmids listed in Table II. E. coli strains were grown in Luria broth, with 100 g/ml ampicillin when required. L. lactis cells were grown in M17 broth (Difco) or chemically defined medium (27) supplemented with 25 mM glucose and 365 mM KCl for high salt conditions (1050 mosmol/kg). Chloramphenicol (5 g/ml) or erythromycin (5 g/ml) was added when the cells were transformed with plasmids. For growth on solid medium, 1.5% (w/v) agar was added to the broth.

Cloning of mscL and yncB
The primers that were used for the amplification of mscL and yncB are listed in Table III. Since the yncB gene product is homologous to MscS from E. coli and functions similarly, the YcnB protein is referred to as MscS Ll . The superscripts Ll and Ec will be used to denote genes or proteins from L. lactis and E. coli, respectively. For PCR amplifications, Expand High-Fidelity DNA polymerase (Roche Applied Science) was used, and reactions were performed according to the manufacturer's instructions. Chromosomal DNA of L. lactis IL1403 was used as template, the annealing temperature was 50°C for mscL Ll and 53°C for yncB, and the elongation times were 30 and 60 s, respectively. For the histidine-tagged version of MscL Ll , a two-step PCR was performed using primers LL.SD.5Ј and LL.SD.4H3Ј (see Table III) to introduce a 4-histidine tag, and then a second step was used to engineer another two histidines, a stop codon and a restriction site for cloning.
The mscL Ll and mscL Ll 6H genes were first subcloned in pBluescript, isolated from E. coli JM110, using the XhoI-XbaI restriction sites. This construct was used to transform JM110, and after plasmid isolation, the genes were excised from the pBluescript vector and ligated into pNZ8020 and pB10b again using the XhoI-XbaI restriction sites. The resulting plasmids pNZ8020mscL Ll 6H and pB10bmscL Ll were then used to transform L. lactis JIM7049 and E. coli PB104, respectively. YncB10H was ligated directly after PCR into the vectors pNZ8048, pET324, and pBAD and used to transform L. lactis IL1403 and E. coli JM465, respectively. Finally, all constructs were midi-prepped (Qiagen), and the DNA sequence was analyzed to confirm fidelity.

Construction of Single Cysteine Mutants and a L. lactis mscL Disruption Strain
The primers used for the construction of the mutants of MscL Ll are listed in Table III. The cysteine mutants were constructed using pB10bmscL Ll 6H as template. First, LL.SD.6H3Ј was combined with the primer for the specific mutant to create a megaprimer, which in a second amplification step was combined with LL.SD.5Ј to obtain the complete mscL sequence with the desired cysteine modification. All other conditions were as described above for the cloning of MscL6H.
For the construction of the L. lactis JIM7049 mscL disruption strain, an internal gene fragment, from codon 12 to codon 86, was amplified. From earlier studies with truncation mutants of E. coli MscL, it could be predicted that a disruption with this gene fragment should lead to complete inactivation of MscL (28). The amplification was performed using the same conditions as described above for amplification of mscL6H. The internal gene fragment was ligated in pOri280, using the XbaI and BamHI restriction sites. The vector was hosted in L. lactis Ll108 for propagation of the plasmid, and, after confirmation of the DNA sequence, the plasmid was introduced into L. lactis JIM7049. The cells were grown on M17-glucose solid medium, supplemented with erythromycin (5 g/ml) to select for colonies that harbored the plasmid DNA integrated into the chromosome (22). The mscL disruption strain is designated JIM7049⌬MscL.

Transcriptional Analysis
Cells were grown to the midexponential phase of growth on CDM and diluted 1:1 to CDM without NaCl or CDM plus 0.5 M NaCl (final concentration) and harvested 10 min after dilution. RNA was extracted as described (29) at concentrations of 4 -5 g/l. The RNA was used as template for the avian myeloblastosis virus reverse transcriptase, using the first strand cDNA synthesis kit for RT-PCR (Roche Applied Science), according to the manufacturer's instructions and in the presence of the supplied random hexanucleotide primers. The obtained cDNA was subsequently used as template in a PCR with Taq polymerase using the RT-MscL, RT-MscS, and RT-OpuA forward (fw) and reverse (rev) primers listed in Table III For PCR amplification, the annealing temperature was 50°C, and the elongation time was 60 s. The products were analyzed on a 1.5% agarose gel after 15 and 22 cycles. As a control, the PCR was also performed on chromosomal DNA of L. lactis; as a negative control, the RNA samples were treated in exactly the same manner, except that the avian myeloblastosis virus reverse transcriptase was omitted in the reaction with RNA as template.

Preparation of MS Channel-containing Membranes for Patch Clamp Analysis
Hybrid Proteo-GUVs-L. lactis NZ9000 containing pNZ8020mscL Ll 6H was grown in M17-glucose medium to an A 600 of 0.8, after which mscL expression was induced for 3 h with a 1:1000 dilution of the supernatant of a culture from the nisinA-producing L. lactis NZ9700 strain (20). Inside-out membrane vesicles were prepared by lysing the bacteria (20 mg/ml protein) with a high pressure homogenizer (Kindler type NN2002; single passage at 10,000 p.s.i.), following (partial) digestion of the cell wall with 10 mg/ml lysozyme for 30 min at 30°C. After removing unlysed cells and cell wall debris by centrifugation at 20,000 ϫ g, the membrane vesicles were washed once by centrifugation at 150,000 ϫ g and then resuspended in 50 mM KP i , pH 6.5. Aliquots of 0.5 ml were frozen in liquid nitrogen and stored at Ϫ80°C. The membrane vesicles were mixed with liposomes (2:25 (w/w) total membrane protein/ lipid ratio), centrifuged at 270,000 ϫ g, and resuspended in 5% ethylene glycol and dehydrated on a glass slide. To obtain fused proteo-GUVs, rehydration was performed by placing rehydration buffer (200 mM KCl, 0.1 mM EDTA, 0.01 mM CaCl 2 plus 5 mM HEPES, pH 7.2) on top of the dried membranes to obtain a final lipid concentration of 100 mg/ml as described (30). Alternatively, after drying on glass slides coated with indium tin oxide, hybrid proteo-GUVs were obtained by rehydration in a flow chamber with 300 l of 10 mM KCl, 2 mM MgCl 2 , 0.25 mM HEPES, pH 7.2, plus 320 mM sucrose (the sucrose was added to make the rehydration medium equiosmolar to the buffer for patch clamp analysis). The flow chamber was closed with a second indium tin oxidecoated glass slide. A voltage of 1.2 V at 10 Hz was applied for at least 3 h through electrodes sealed on the glass plates. The resulting giant unilamellar vesicles (GUVs), 5-50 m in diameter, were used in patch clamp experiments (31).
Proteo-GUVs-Membrane vesicles, containing overexpressed channel protein (MscL Ll or MscS Ll ), were solubilized in 50 mM KP i , 35 mM imidazole, 300 mM NaCl, pH 7.0, plus 3% octyl-␤-D-glucoside (Anatrace) at 4°C for 20 min and under continuous stirring of the suspension. The solubilized proteins and remaining membrane material were separated by ultracentrifugation at 270,000 ϫ g for 20 min at 4°C. The supernatant was then loaded onto nickel-nitrilotriacetic acid affinity resin preequilibrated with solubilization buffer. The column was washed with 20 -30 volumes of solubilization buffer containing 0.5% Triton X-100, after which the proteins were eluted with 50 mM KP i , 300 mM NaCl, pH 7.0, plus 0.2% Triton X-100 with a step gradient of imidazole. The purified proteins were mixed with Triton X-100-destabilized preformed liposomes (10 mg/ml lipid); the lipid mixtures were composed of dioleoyl 18:1 (⌬9 cis) phospholipids (dioleoylphosphatidylcholine, dioleoylphosphatidylethanolamine, and/or dioleoylphosphatidylserine), typically at protein/lipid ratios of 1:2000 -20,000 (mol of monomeric channel protein/mol of lipid), and reconstitutions were performed as described (32). The proteoliposomes were converted to proteo-GUVs by dehydration and rehydration in the presence or absence of an electrical field, as described under "Hybrid Proteo-GUVs." Spheroplasts-For determination of pressure ratios, relative to Msc-S Ec , of MscL Ll and the individual cysteine mutants, giant spheroplasts were prepared from E. coli PB104 containing pB10bmscL Ll 6H or its derivatives. For other channel characteristics, giant spheroplasts were prepared from E. coli MJF465 containing pB10bmscL Ll , pB10bmscL Ll 6H, or pET324yncB10H. All spheroplasts were prepared as described (30).

Patch Clamp Experiments
Experiments were performed as described previously (30). After preparation, an aliquot of proteo-GUVs or 1-5 l of a spheroplast sample was transferred to a sample chamber containing a ground electrode and 300 l of patch clamp buffer: 5 mM HEPES, pH 7.2, 200 mM KCl, 40 mM MgCl 2 for proteo-GUVs; 5 mM HEPES, pH 7.2, 200 mM KCl, 90 mM MgCl 2 , 10 mM CaCl 2 for spheroplasts. Channel activity was recorded using an Axopatch 200A amplifier together with a digital converter and Axoscope software (Axon Instruments, Foster City, CA). Data were acquired at a sampling rate of 33 kHz and filtered at 10 kHz. The presented traces were additionally filtered to decrease electronic noise, using Clampfit 8.0 software (Axon Instruments) with the low pass Boxcar filter at smoothing point 7. Offline analysis was performed using PClamp 6.0 software (Axon Instruments).

Glycine Betaine Efflux
Glycine betaine efflux was performed as described (13) with some modifications. Cells were grown overnight to late log phase in high osmolality medium (1050 mosmol/kg) and washed three times in their original volume with an equimosmolar buffer (50 mM KP i , 500 mM KCl, pH 6.5). The cells were then resuspended to a protein concentration of ϳ10 mg/ml and stored on ice. For uptake of glycine betaine, the cells were diluted 10-fold in 50 mM KP i , pH 6.5, 500 mM KCl, supplemented with 10 mM glucose and 1.88 mM of [ 14 C]glycine betaine. After 40 min of uptake at 30°C, aliquots of the cells were subjected to no dilution or a 3-, 5-, 10-, 20-, 50-, and 100-fold dilution into 50 mM KP i , pH 6.5. This resulted in final osmolalities of 1050, 420, 295, 200, 155, 125, and 115 mosmol/kg, respectively. Samples were taken in 5-fold to determine the steady state internal glycine betaine concentrations prior to the osmotic downshift (final internal glycine betaine content) and at different time intervals (at 1 (in duplicate), 5, 10, and 30 min) after each of the downshifts. At least three independent uptake and downshift experiments were performed on three independent cultures of L. lactis IL1403 and JIM7049⌬MscL. The steady state levels of internalized glycine betaine after uptake were ϳ900 nmol/mg protein for both strains.
Release is plotted as a percentage of retained glycine betaine.

Analysis of Cell Viability
Survival of E. coli MJF465, carrying pB10bmscL Ll 6H, pET324yncB10H, or empty plasmid controls, under osmotic downshift conditions, was analyzed as described (12). Because E. coli MJF465 is not able to use arabinose as an energy source and glucose represses the expression of the arabinose-inducible system, the isopropyl 1-thio-␤-D-galactopyranoside-inducible construct pET324yncB10H was used here. The cysteine mutants were screened for survival with, and without, MTSET as described (33). For survival under osmotic downshift conditions of L. lactis IL1403 or JIM7049⌬MscL, cells were grown overnight in chemically defined medium (27), supplemented with 25 mM glucose and 5 g/ml erythromycin (where applicable), and diluted 1:100 to the same medium supplemented with 365 mM KCl (1050 mosmol/kg). Cells were allowed to grow to an A 600 of ϳ0.8 and then diluted 100-fold into 50 mM KP i , pH 6.5, plus 500 mM KCl (1050 mosmol/kg) or into 50 mM KP i , pH 6.5 (final osmolality of 115 mosmol/kg). Prior to plating onto agarcontaining media, the cells were diluted serially with sterile

Metabolic Activity after Osmotic Downshift
Because L. lactis grows in chains, the number of CFU is not necessarily a quantitative indicator of the survival of cells after downshift. Therefore, metabolic activity based on the production of acid by L. lactis during fermentation of glucose was also determined. The method was adapted from Ref. 34. In brief, cells were cultured, washed, and osmot-ically stressed as described above for the glycine betaine efflux assay, but with 5 mM KP i , pH 6.5, and 65 mM KCl (105 mosmol/kg) instead of 50 mM KP i buffer (105 mosmol/kg) to reduce the buffering capacity. Acidification rates per mg of total protein in the presence of 10 mM glucose were determined for all samples and compared with the acidification rates of the unshocked sample (100% value).

Other Procedures
Protein Biochemistry-Purified proteins were analyzed on 15% acrylamide SDS-PAGE (35). Protein concentration was determined by Amido Black 10B staining as described (36). MscL was also identified and analyzed for potential modifications or proteolysis products by MALDI-TOF mass spectrometry as described (37). The MscL Ϫ phenotype of L. lactis JIM7049⌬MscL was checked by preparation of cell membranes and subsequent SDS-PAGE analysis. The presence or absence of MscL Ll was determined by immunodetection using specific rabbit serum with polyclonal antibodies raised against MscL Ll (at a titer of 1:30,000) and the Western light chemiluminescence detection kit (Tropix Inc., Bedford, MA). Determination of Osmolality-Osmolalities of media and buffers were measured by freezing point depression with an Osmostat 030 (Gonotec, Berlin).

Cloning, Expression, and Purification
Cloning and expression of mscL Ll and yncB Ll was achieved in both E. coli (not shown) and L. lactis ( Fig. 2A). For amplification of MscL Ll and MscS Ll in L. lactis, the nisin-inducible expression system was used (38). In E. coli, the lacUV5 promoter system was used for amplification of MscL Ll , and the arabinose-inducible system was used for amplification of MscS (24). MscL Ll -6H and MscS Ll -10H were purified by nickel-nitrilotriacetic acid chromatography. Based on the yield of purified protein, both channels were found to be amplified in L. lactis to levels between 5 and 10% and in E. coli between 2 and 5% of total membrane protein.
As can be seen in Fig. 2, A and B, MscL Ll protein amplified in and purified from L. lactis NZ9000 migrated over a wide range in SDS-PAGE and gave rise to a "fuzzy" appearance. To establish the nature of the differently migrating species, various fractions from the gel were analyzed by MALDI-TOF mass spectrometry. Four distinct slices of the gel were digested with CNBr and trypsin, and the resulting peptide fragments were identified. The fragments of the protein, after cleavage, accounted for 70% of the total protein (data not shown), including both the N and C termini, indicating that the heterogeneity observed on the gel was not due to proteolytic degradation of the amplified protein. We suggest that the broad and fuzzy band reflects incomplete unfolding of the protein in the presence of SDS or a very strong association between individual subunits, which is only dissociated during SDS-PAGE analysis.

In Vivo Complementation of E. coli MJF465 (MscL Ϫ /MscS Ϫ )
For initial characterization of MscL Ll and MscS Ll , the corresponding genes were expressed in E. coli MJF465 (MscL Ϫ / MscS Ϫ ), using plasmids pB10bmscL Ll 6H and pET324yncB10H. This allowed a comparison of the ability of the L. lactis and E. coli channels to rescue the osmotic-fragile phenotype of E. coli MJF465 (12). Clearly, both MscL Ll and MscS Ll were able to rescue MJF465 in osmotic downshift experiments (Table IV), albeit not as well as the endogenous E. coli channels.
It has been demonstrated that MscL Ec can be activated by MTSET modification of a cysteine introduced at position 22 (33,39,40). The equivalent residue, Gly 20 , in MscL Ll and its flanking residues (Val 21 , Ile 22 , and Ile 23 ; boxed region in Fig. 1) were replaced by cysteines, and the phenotype of the mutants, upon labeling with MTSET, was determined (Table IV). The cysteine positions that showed a clear gain-of-function phenotype as a result of increased channel activity after reaction with MTSET under iso-osmotic conditions, G20C and V21C (equivalent to positions G22C and V23C in MscL Ec ), were further characterized in patch clamp experiments.

Electrophysiological Characterization
MscS Ll in E. coli-Channel activity of an excised patch from E. coli MJF465 expressing MscS Ll 10H (from pBADyncB10H) is shown in Fig. 3A. The slope of the current-voltage relationship in Fig. 4A indicates that MscS Ll allows a conductance of 0.87 Ϯ 0.07 nanosiemens. The fact that the data in Fig. 4A could be fit to a straight line reveals that the channel shows no rectification between Ϫ50 and 50 mV. Open dwell times of one to several hundred ms were observed, and the channel appeared to desensitize when pressure was maintained for longer periods of time (n ϭ 5). Overall, the channel characteristics of MscS Ll are analogous to those described previously for MscS from E. coli (12).
MscL Ll in E. coli-Channel activity of an excised patch from E. coli MJF465, expressing MscL Ll 6H (from pB10bmscL6H) is shown in Fig. 3B. The current-voltage relationship for MscL Ll (Fig. 4A) shows that MscL Ll has a conductance of 2.00 Ϯ 0.25 nanosiemens. Also, MscL Ll did not show rectification between Ϫ50 and 50 mV. Two short but distinct dwell times were observed in 34 independent recordings (both from lipid fusions and spheroplasts; Fig. 4B). One dwell time could be determined at 3.6 Ϯ 0.29 ms (mean and S.E.); the second was shorter than 1 ms and could not be determined more accurately given the experimental parameters. These dwell times are comparable with the dwell times of MscL from Staphylococcus aureus and are shorter than the dwell times observed for MscL from E. coli (5). The pressure at which MscL Ll fully opened for the first time was determined and compared with the pressure at which MscS Ec  which was found to have a MscL Ec /MscS Ec pressure ratio of 1.64 Ϯ 0.08 (41). The open probability of MscL Ll was determined by the average current (measured at a given pressure) divided by the current if all channels in the patch were open. The open probability was plotted against the pressure ratio MscL/MscS Ec and fitted to a Boltzmann distribution (Fig. 4C). Both MscL Ec and MscL Ll show initial openings around the same relative pressure. However, further opening of MscL Ll required a higher energy input than opening of MscL Ec as reflected by the different slopes of the sigmoidals (42). Finally, MscL Ll appeared to have distinct substates at 0.26, 0.53, and 0.75 of full conductance (Fig. 4D), which compares with 0.22, 0.45, 0.70, and 0.93 of full opening for MscL Ec (43).
MscS Ll and MscL Ll in L. lactis-Channel activity in a patch excised from a hybrid proteo-GUV, prepared from L. lactis IL1403 is shown in Fig. 3C. MscL Ll activity, but no MscS Ll activity, was observed in 30 independent patches. MscS Ll activity was, however, observed in patches excised from hybrid proteo-GUVs from L. lactis JIM7049, overexpressing MscS Ll (from pNZ8048yncB10H) as shown in Fig. 3D, indicating that the yncB gene product can be functional but is possibly not expressed in wild type L. lactis. The observation that MscS Ll is apparently not functional in wild type L. lactis is underscored by the observation that no MS channel activity whatsoever was observed in patches of JIM7049⌬MscL membranes fused with liposomes (n ϭ 10 with cells grown under low salt conditions; n ϭ 11 with cells grown under high salt conditions) (Fig. 3E). Even if the native L. lactis JIM7049⌬MscL membrane/liposome ratio was increased 10-fold from 2:25 to 20:25 (w/w) (n ϭ 5, data not shown), MS channel activity was not observed. Recordings of channel activity of patches excised from proteo-GUVs from L. lactis JIM7049 pNZ8048yncB10H membranes expressing MscS Ll were used to determine the pressure ratio for full openings of MscL Ll relative to MscS Ll . This value was 1.48 Ϯ 0.05 (n ϭ 14), but substate channel activity of MscL Ll could be observed before the first opening of MscS Ll (e.g. see Fig. 3D, #).
Cysteine Mutants of MscL Ll in E. coli-The cysteine mutants that had a gain-of-function phenotype in response to MTSET labeling were analyzed with respect to the pressure ratio MscL Ll /MscS Ec , using E. coli PB104 expressing the channel from pB10bmscL Ll 6H (G20C) or pB10bmscL Ll 6H (V21C). Both G20C and G21C had a high pressure ratio of 2.6 Ϯ 0.3 (n ϭ 8) and 3.7 Ϯ 0.2 (n ϭ 4), respectively, in the absence of MTSET, but their activity became pressure-independent upon the addition of MTSET (data not shown).

Expression of mscL, yncB, and OpuAA
Since electrophysiological characterization of L. lactis IL1403 and JIM7049⌬MscL membranes, fused with liposomes, showed no MscS Ll activity (Fig. 3C), it seemed possible that yncB was not transcribed in these cells. RT-PCR was performed on RNA extracted from L. lactis IL1403 cells grown in CDM or CDM plus 0.5 M NaCl, and the resulting amplified DNA fragments were analyzed on gel (Fig. 5). After 15 PCR cycles, it is clearly visible  that the transcription of opuAA (the first gene of the opuA operon) was increased in the cells grown at high salt (Fig. 5A, compare lane A3 with lane B3), which is in agreement with a higher glycine betaine uptake capacity under these conditions (13). RT-PCR DNA corresponding to mscL and yncB fragments were visible after 22 PCR cycles. Both mscL and yncB are transcribed in wild type L. lactis IL1403 cells, albeit to a low level compared with the opuAA gene of the opuA operon; the expression of the mscL and yncB was not significantly influenced by the osmolality of the medium (Fig. 5B, compare lanes A1 and B1 and  compare lanes A2 and B2). The negative controls showed that the extracted RNA was not contaminated with significant amounts of DNA, since no product was obtained when reverse transcriptase was omitted from the reaction mixture (Fig. 5B, lanes  A Ϫ and B Ϫ ); the lanes marked C1, C2, and C3 correspond to PCR products of mscL Ll , yncB, and OpuAA, respectively, using chromosomal DNA as template.

Physiological Characterization of JIM7049⌬MscL
Plasmid pOri280⌬MscL Ll is not replicated in L. lactis JIM7049 unless it integrates into the chromosome, which preferentially takes place at the homologous position of the chromosomal mscL gene. By selecting for clones that grew in the presence of erythromycin, a disruption strain of mscL was obtained. MscL Ll could not be detected by Western blot analy-sis in membranes of these cells, using antibodies raised against MscL Ll (Fig. 2B).
To determine the role of MscL Ll under conditions of hypoosmotic stress, L. lactis IL1403 and JIM7049⌬MscL Ll cells were grown on CDM supplemented with 365 mM KCl. After washing in 50 mM KP i plus 500 mM KCl, pH 7.0, the cells were incubated with [ 14 C]glycine betaine for 40 min to allow internalization of the compatible solute and thereby compensate for the high external osmolality. A steady state level of uptake was attained, corresponding to an internal concentration of about 900 nmol of glycine betaine/mg of total protein. The cells were then diluted into various hypotonic media, and the fate of the internal glycine betaine content was determined. In Fig. 6A the internal concentrations are expressed as a percentage of the internal glycine betaine before osmotic downshift. The L. lactis IL1403 cells released ϳ80% of their glycine betaine when diluted 5-fold or more. The release of glycine betaine from L. lactis JIM7049⌬MscL was significantly lower; around 40% of the glycine betaine was retained by the cells. Taken as a whole, the data suggest that at least a fraction of the internalized glycine betaine is released via MscL Ll .
Next, cell survival was examined using protocols developed for E. coli. Under osmotic downshift conditions, 98 Ϯ 6% (n ϭ 3) of wild type CFU survived and 80 Ϯ 5% (n ϭ 3) of the disruption were collected from three independent spheroplast patches with 1-3 channels per patch, in which MscL Ll activity was saturated. D, relative conductive substates of MscL Ll (compared with full opening) at Ϫ20 mV determined by using an all-points histogram based on data from five independent patches. Data were normalized, and the number of events per conductance substate was determined, from which the most frequently visited and therefore most stable substates were obtained. strain. Because L. lactis grows in chains of mostly two or four cells, this could mean that in reality between 33 and 55% of the cells survive; to obtain a CFU from a chain of cocci, it is sufficient if only a single cell is viable (also see "Discussion").
To obtain a quantitative measure of the consequences of an osmotic downshift on the metabolic activity of L. lactis IL1403 and JIM7049⌬MscL, the capacity to ferment glucose was determined before and after osmotic downshift. Cells were grown in high salt medium and preloaded with glycine betaine and subjected to osmotic downshifts as described above. For L. lactis IL1403, only the metabolic activity after 100-fold dilution was determined, since even under this extreme downshift condition 100% of glucose metabolism was retained. In contrast, the rate of acidification of the buffer around L. lactis JIM7049⌬MscL decreased with the -fold dilution (Fig. 6B). The decrease in metabolic activity is quantitatively similar to the release of glycine betaine from JIM7049⌬MscL, suggesting that the release of glycine betaine in the disruption mutant might be caused by loss of cell integrity rather than efflux through the MscL Ll channel. DISCUSSION MS Channels of L. lactis-MscL and MscS from L. lactis form functional mechanosensitive channels when expressed in E. coli or L. lactis. Both channels provided E. coli MJF465 (MscS Ϫ /MscL Ϫ ) with protection against osmotic downshift as demonstrated in plate assays, and both channels were functional in patch clamp experiments using membrane patches from E. coli spheroplasts, membrane patches from hybrid proteo-GUVs prepared from L. lactis membrane vesicles, or membrane patches from proteo-GUVs prepared from purified and membrane-reconstituted protein. Importantly, in wild type L. lactis only MscL Ll seems to be functional, and this MS channel appears critical for survival of the organism under conditions of hypo-osmotic stress.
The unitary conductances of MscS Ll and MscL Ll were 0.87 and 2.0 nanosiemens, respectively, which are within the range previously reported for MscS and MscL from other organisms (8,9,39). Open dwell times for MscL Ll 6H were short compared with MscL Ec and most other MscL homologues studied to date, giving rise to a "flickery" appearance in channel recordings. The activation thresholds of both L. lactis channels were comparable with those of the E. coli homologues, as were the channel substates observed for MscL Ll (43). We also showed that MscL Ll could be activated by modifying cysteine residues with MTSET (G20C or V21C) at the constriction site of the pore, demonstrating that MscL from L. lactis functions very much like MscL from E. coli.
Survival and Metabolic Activity after Osmotic Downshift-Cell survival of E. coli MJF465 after osmotic downshift conditions was determined after heterologous expression of MscS Ll and MscL Ll . The experiments showed that both channels were able to partially rescue the MscL Ϫ /MscS Ϫ phenotype of the MJF465 strain. It can therefore be concluded that, in E. coli, both proteins form functional MS channels. In L. lactis, it was harder to determine the effect of an osmotic downshift on cell survival because the cells grow in chains. The number of CFU of the disruption strain JIM7049⌬MscL decreased by 20% when the cells were subjected to an osmotic downshift. If a binomial distribution of cell death in a chain of four cells is assumed, then the distribution of live and dead cells would be L 4 ϩ 4L 3 D ϩ 6L 2 D 2 ϩ 4LD 3 ϩ D 4 (where L represents the fraction of live cells and D is the fraction of dead cells). Since D 4 (i.e. all cells in one chain are dead) is the only case that gives rise to a decrease in the number of CFU, the fraction of dead cells can be calculated as the fourth root of the fractional decrease in CFU. In this case, the fraction of dead cells is (0.20) 0.25 ϭ 0.67, implying that only 33% of the cells survived the osmotic downshift. Assuming a chain length of two cells and a normal distribution, the survival would be 55%. With most cells growing in chains of 2 and 4, the observed decrease in cell viability of 20% thus corresponds to an actual decrease in viability of 45-67%.
The glucose fermentation capacity of L. lactis JIM7049⌬MscL decreased by 60% after osmotic downshift. Since the IL1403 wild type strain (which normally expressed MscL Ll ) was not affected by the osmotic downshift, a fraction of the JIM7049⌬MscL may have lysed, as a consequence of the inability to rapidly release osmolytes upon osmotic downshift. Strikingly, the decrease in glucose fermentation capacity of 60% compares well with the actual decrease in cell viability of 45-67%.
MscS Activity Is Not Detectable in Wild Type L. lactis-Based on the following observations, L. lactis IL1403 does not seem to possess detectable levels of functional MscS, although the yncB gene is transcribed. First, MscS Ll activity was not observed in patches excised from hybrid proteo-GUVs of IL1403, which displayed MscL activity. Second, MscS Ll activity was not observed in patches excised from hybrid proteo-GUVs of JIM7049⌬MscL, which lacked any MS activity even when prepared from cells grown under different osmotic conditions. Finally, efflux of glycine betaine upon osmotic downshift was significantly decreased in JIM7049⌬MscL. The residual 60% of efflux can be explained by cell lysis, as could be inferred from the decrease in CFU in the drop plate assay and the decrease in metabolic activity upon osmotic downshift. In contrast to L. lactis, an E. coli knock-out strain released the same amount of glycine betaine as the wild type E. coli upon osmotic downshift (44), suggesting that in E. coli another MS channel compensates for the absence of MscL or that in the wild type strain MscL is not activated under the conditions used.
However, when yncB was expressed in E. coli MJF465 (using pBADyncB10H) or in L. lactis JIM7049 (using pNZ8048yncB10H), MS channel activity was observed. RT-PCR showed that the yncB gene is transcribed to levels comparable with mscL but much lower than opuAA. Moreover, and contrary to the opuA genes, the expression of mscL and yncB genes in L. lactis IL1403 does not seem to be influenced by the osmolality of the medium. In E. coli, the expression of mscL and mscS is growth phase-dependent, and the transcription level of both genes increases when the osmolality of the medium increases (45). In analogy with MscS from E. coli, one could argue that MscS from L. lactis is post-translationally inactivated; for instance, the open probability of MscS from E. coli is decreased by lowering of the pH and long exposures to membrane tension (12). Evidence for inactivation by prolonged exposure to membrane tension of MscS Ll can be observed in Fig. 3A (yncB expressed in E. coli), and it is possible that MscS in L. lactis IL1403 is kept in an inactive state, for which we have not been able to find the activation trigger. In contrast to wild type L. lactis, overexpression of yncB in L. lactis yielded channel activity, suggesting that, when excess MscS channel is produced, a population of the molecules escapes the inhibition mechanism.
What are the consequences of inactive MscS for L. lactis? It is possible that MscS is less critical for L. lactis than for E. coli. Substate activity of MscL Ll can be observed at pressures at which MscS Ll activity is not yet observed. The presence of substates, together with the short dwell times of MscL Ll , may be sufficient for L. lactis to fine regulate the internal osmolyte concentration without requiring a MscS-like activity.
In conclusion, MscS and MscL of L. lactis are genuine mechanosensitive channel proteins with properties similar to those described for MscS and MscL from E. coli. We show, however, that MscL Ll is the primary, if not the only, mechanosensitive channel used by L. lactis under the conditions studied, through which a major portion of compatible solutes such as glycine betaine is released upon osmotic downshift.