One face of a transmembrane helix is crucial in mechanosensitive channel gating

Abstract

MscL is a mechanosensitive channel in bacteria that responds directly to membrane tension by opening a large conductance pore. To determine functionally important residues within this molecule, we have randomly mutagenized mscL, expressed the genes in living bacteria, and screened for gain-of-function mutants with hampered growth. Expression of these genes caused leakage of cytoplasmic solutes on little or no hypo-osmotic stress. In excised patches, the mutant channels gated at membrane tensions that are less than that required for the gating of the wild-type MscL. Hence, the data suggest that the slowed or no-growth phenotype is caused by solute loss because of inappropriate gating of the channel. Most of the mutations mapped to the first transmembrane domain. When this domain is modeled as an α-helix, the most severe mutations are substitutions of smaller amino acids (three glycines and one valine) on one facet, suggesting an important role for this structure in MS channel gating.

Mechanosensitive (MS) channels have been implicated in touch, hearing, balance, cardiovascular regulation, and the sensing of gravity and osmotic stress (1, 2). However, as molecular entities, these sensors largely have remained elusive, especially when compared with ligand- or voltage-gated channels. One notable exception is the Escherchia coli mechanosensitive channel of large conductance, MscL (3–5), which resides in the cytoplasmic membrane of E. coli (6–9) where it apparently detects and responds to osmotic downshock (10–13). By tracing its activity in reconstituted membrane subfractions, a corresponding subunit protein was identified, and its gene was cloned (4). This homohexameric channel (6), whether expressed in native (14, 15) or reconstituted membranes (16) or as reconstituted purified protein (6, 17), responds directly to stretch forces delivered through the lipid bilayer by opening a large solute-nonspecific pore of very large conductance (2.5 nS) (3, 18). This small protein (136 residues) and its homologues (19) are currently the only isolated MS channel molecules available for direct experimentation.

Here, we have screened for and identified mscL genes with single substitutions that, when expressed in E. coli, can cause a slowed or no-growth phenotype (Fig. 1). Measurements of potassium within cells expressing these mutant MscL channels suggest that the cells lose this solute with low or no osmotic downshock. In addition, patch–clamp studies of the mutant MscL channels expressed in native bacterial membranes confirm that many of these channels open at inappropriately low membrane tensions. Hence, the slowed or no-growth phenotype appears to be caused by solute loss because of inappropriate gating of the mutant channels. Although mutations were generated randomly, presumably throughout the entire protein, most of the mutants isolated by this genetic screen mapped to one portion of the molecule: residues that are predicted to be on one surface of an α-helix within the first hydrophobic domain.

Figure 1.

(a) The experimental scheme. A plasmid with an inducible Lac promoter and the mscL coding region was mutagenized in vitro and used to transform mscL-null E. coli. Transformant colonies whose replicas failed to appear after induction (+IPTG) were isolated. We then determined their growth rates, K⁺ contents, and MS conductances as well as the mutated DNA sequence so as to correlate functional with structural changes. (b) Growth on Luria–Bertani nutrient plates. Each row shows the growth pattern of three 10-μl drops of inoculate from cultures of 0.3 OD₆₅₀ diluted 10⁴, 10⁵, and 10⁶ fold. (Left) Plates without IPTG where all strains grew normally. (Center) With 1 mM IPTG, only the wild type (WT) grew normally. (Right) Same plate as center but photographed much later and at a higher contrast. The last two panels together show that induced K55T colonies grew slowly, G30E even more slowly, and N15D or V23A not at all on plates. (Induced N15D grew very slowly in liquid media, but V23A not at all; see Table 1) (c) Loss of cell K⁺ on osmotic downshock. One hour after induction, bacteria were downshocked from 858 milliosmol to different hypo-osmolarity (abscissa) before cell K⁺ was determined by flame photometry and plotted as percentage of the induced but no-shock control (mean ± SEM, n = 3–8). N15D (severe) and G30E (milder mutant) but not K55T (mildest) mutant consistently lose more K⁺ at milder downshocks. V23A (very severe) has a low viability and was not tested.

MATERIALS AND METHODS

Genetics.

All mutant and wild-type mscL were expressed in the vector pB10b (19). Mutagenesis was performed in two ways: First, treatment of the wild-type mscL containing plasmid p5–2-2b with hydroxylamine (20): 20 μl of fresh 2M hydroxylamine (in 100 mM Na pyrophosphate/2 mM NaCl, pH 6.0) was mixed with 30 μl of plasmid (1 μg/μl TE buffer) and incubated at 75°C for 1–4 h. At each time point, a 10-μl aliquot of the mix was added to 1 ml of stop buffer and the DNA was precipitated, washed, and dissolved in 20 μl of TE. Mutant mscLs were subcloned into and sequenced in pBluescript (Stratagene), and those ORFs with single substitutions were subcloned back into fresh pB10b. Second, error-prone PCRs using Taq polymerase and an established protocol (21), with varied amount of MnCl₂. Reaction products were gel purified and then subcloned into pB10b. The mscL null E. coli strain PB104 (6) was used to host the wild-type or mutant expression plasmids. Most mutant mscL in the selected plasmids were sequenced directly, but at least one isolate of each allele was subcloned into a fresh pB10b, resequenced, and used for further study. Both strands were sequenced by using an ABI Prism Automated Sequencing Kit and ABI 337 automated sequencer (Perkin–Elmer).

Growth Phenotypes.

Growth rates in the defined K10 medium (22), with or without 1 mM isopropyl β-d-thiogalactoside (IPTG) (induced or uninduced, in Table 1), were determined as described and expressed as percentage wild-type rate of OD₆₅₀ increase (10). Colony forming units per OD₆₅₀ were counted on IPTG-free Luria–Bertani (LB) agar. Viability was gauged by the changes in colony forming units after 1 h of induction with 1 mM IPTG (10) and was photographed at times given or scored as in Table 1. Growth in Fig. 1b was on LB plates ± 1 mM IPTG. All media contained 100 μM ampicillin.

Table 1.

Growth phenotypes and K⁺ content of the gain-of-function MscL mutants

Strain	Growth in liquid media		Viability (plate assay)	[K+]in
	Uninduced	Induced		Uninduced	Induced	After downshock
Wild type	100	106  ±  8	+++	100	99  ±  7	56  ±  3
Group 1, very severe
G22D	92  ±  4*	Nil***	−	ND	ND	ND
G22N	96  ±  5	Nil***	+	ND	ND	ND
G22S	90  ±  8	Nil***	+	ND	ND	ND
V23A	23  ±  4***	Nil***	−	ND	ND	ND
V23D	104  ±  9	Nil***	−	ND	ND	ND
V23G	69  ±  4***	Nil***	+	ND	ND	ND
G26S	93  ±  5	Nil***	+	ND	ND	ND
G30R	99  ±  3	Nil***	−	ND	ND	ND
Group 2, severe
N15D	85  ±  5*	18  ±  3***	++	92  ±  9	59  ±  10*	32  ±  7**
L19Y	103  ±  4	Nil***	++	102  ±  4	45  ±  25*	51  ±  11
G46D	96  ±  1	60  ±  4***	+	94  ±  10	62  ±  27*	59  ±  16
N100D	92  ±  3*	47  ±  7***	++	100  ±  12	62  ±  10*	62  ±  5
Group 3, milder
G30E	101  ±  4	60  ±  4***	+++	106  ±  6	94  ±  9	22  ±  6***
Group 4, mildest
R13C	96  ±  6	44  ±  5***	+++	95  ±  7	65  ±  16*	43  ±  6*
G14E	91  ±  4*	52  ±  6***	+++	92  ±  10	73  ±  17*	49  ±  6
V21E	104  ±  6	48  ±  2***	+++	103  ±  4	78  ±  22	55  ±  13
K55T	102  ±  5	80  ±  9*	+++	105  ±  7	106  ±  10	46  ±  7*
I92F	100  ±  6	70  ±  9**	+++	103  ±  8	92  ±  13	50  ±  6

All values mean ± SD (not SEM); n = 3 or 4 for growth rates, and n = 5 for [K⁺]_in; ∗, P ≤ 0.05; ∗∗, P ≤ 0.001; ∗∗∗, P ≤ 0.0001, by Student’s t test against control. Liquid growth compared with wild type uninduced. Un- or induced [K⁺]_in compared with wild type uninduced. After downshock is percentage of induced (but no-shock) level of each of the mutant strain, not wild type. ND, not determined because of no or low viability. Viability by plate assay refers to colony-forming units among cells after 1-h induction. 

K⁺ Content and K⁺ Efflux on Downshocks.

Potassium content, [K⁺]_in, was determined by flame photometry (10, 23). Cells were grown in K10 with 325 mM NaCl (858 milliosmol) and induced with IPTG for 1 h, and aliquots were collected and challenged with media of osmolarities (K10 with varying NaCl) plotted in Fig. 1c, as described. Table 1 shows percentage of [K⁺]_in of uninduced (no ITPG), induced before, and induced after a downshock from 858 to 458 milliosmol.

Patch–Clamp Studies.

Inside-out patches were excised from giant cells generated as described (3–5, 14, 15, 23). IPTG (1 mM) was added only during the last 5 min of cephalexin treatment. The pipet solution was: 200 mM KCl, 90 mM MgCl₂, 10 mM CaCl₂, and 5 mM Hepes adjusted to pH 6.0; the bath solution was the same plus 0.3 M sucrose. Recordings were at room temperature and at −20 mV. Recordings were cytoplasmic negative and were filtered at 10 kHz and were sampled at 30.3 kHz and then at 3 kHz filtration by Fetchan. The P_o vs. pressure plots and the Boltzmann fits were as described (24). Pressure was assessed by using a pressure transducer and is presented as the applied pressure divided by the MscS pressure threshold in the same patch. Threshold is the pressure needed, within 7 s, to activate two or more MscS units (24). For kinetic analyses (24), currents were recorded as above but bypassing the Fetchan filter and then were analyzed with pstat. Note that each strain expresses its normal mscS from its mscL-null chromosome and its mutated mscL from the mutated plasmid p5–2-2b.

RESULTS

The Strategy for the Isolation of Gain-of-Function MscL Mutants.

Fig. 1a diagrams our approach to find domain(s) important in MS channel activity. Loss of mscL has little effect under the various laboratory conditions tested (3), presumably because of proteins with redundant functions, e.g., MscS (below). Our searches here are for “gain-of-function” mutants whose MscL channels may be abnormally active. Note that we avoided studying other genes by mutating only the mscL in an inducible plasmid and not the E. coli chromosome. Note also that the entire mscL was mutagenized randomly. From ≈50,000 transformants of these plasmids, we isolated 58 mutants by screening for IPTG-induced growth impediment. They manifest a range of difficulties in colonizing nutrient plates with IPTG (Fig. 1b), in populating IPTG liquid media (Table 1, Growth), or in simply surviving for 1 h in such a medium (Table 1, Viability). These phenotypic gradations were used together to sort broadly the mutants into very severe, severe, milder, or mildest groups (Table 1, Groups 1–4).

Cells Expressing MscL Mutants Leak Solutes.

MscL is apparently one of the safety valves in vivo for the defensive jettison of solutes on sudden hypo-osmotic stress (10–13). Hence, one might expect the mutant channels to release more solutes than wild type on a mild osmotic downshock (10) or even without any shock. We measured [K⁺]_in, the cellular content of the most significant osmolite, potassium (25, 26). This was not done with the very severe mutants (Group 1) because they begin to die on induction. Without induction, other mutants have 100% the wild-type [K⁺]_in (Table 1, “uninduced”). Induction alone (no downshock) lowers the [K⁺]_in in the mutants but not the wild type; some have only 50% that of the wild type (Table 1, “induced,” as percentage of wild-type [K⁺]_in). The defect in K⁺ retention roughly parallels the severity of growth retardation in IPTG (Fig. 1b; Table 1). A 400-milliosmol downshock left the wild type with ≈60% of its K⁺. Mutants retained some 60% or less of their already diminished pre-shock [K⁺]_in (note that Table 1, last column, lists percentage of induced level of that strain, not of wild type). We found the wild type to begin ejecting K⁺ on a 200-milliosmol downshock (from 858 to 650 milliosmol) and lose some 90% of [K⁺]_in at the extreme downshocks (to ≤200 milliosmol) (Fig. 1c). Consistent with the single-point values (Table 1, last column) G30E and N15D start and finish ejecting their K⁺ at downshocks by ≈100 milliosmol milder than the wild-type levels (Fig. 1c).

Patch–Clamp Studies Show that Mutant MscLs Gate at Lower Membrane Tensions than Wild Type.

MscL channels in membrane patches excised from giant cells were examined directly by patch–clamp (23). All gain-of-function mutant MscLs were found to encode MS channel activity (Fig. 2, not all shown). Because stretch is a function of suction and geometry, and the latter varies from patch to patch, we used the ubiquitous second MS channel of a smaller conductance, MscS, as an internal control. Wild-type MscL conductance invariably appears at suction 1.4–1.5 times of that which triggers the MscS conductance (23, 24), e.g., the first panel of Fig. 2 shows that two wild-type MscS units activated (asterisk) at a smaller suction (lower trace), and MscLs (triangle) did not activate until the suction was higher. Half activation (P_1/2) occurs at ≈1.7 to 1.8 times (24) (Fig. 2). All mutants appeared to be hypersensitive to membrane tension. As an example of its class, the “very severe” V23A mutation reverses the activation order such that the MscL activates before MscS, which appears as negative images buried in the activities of MscL (Fig. 2, second panel). This hypersensitivity to pressure is evident in the less severe mutants as well, though often to lesser degrees (Fig. 2, lower rows). The MscL open probability plotted against pressure can be fitted with a Boltzmann distribution (14, 24) (Fig. 2, middle column) and the V23A curve is clearly shifted far to the left. The slopes of the wild-type and mutant curves showed no statistically significant difference within the resolution of the experiments. Thus, these mutations change the gating set point with little or no change in the resistive or elastic properties of the patch or the channel protein or in the spatial parameter of the channel [e.g., ΔA (27)].

Figure 2.

MscL single-channel conductance analyses (20). Each row is from a single patch. (Left) Each panel is a representative episode of applying suction (lower trace) onto, and recording the current (upper trace) through a patch of membrane excised from a giant cell of the strain marked. Typical of >30 episodes from 10 to 50 cells of each strain. ∗, unit conductances of MscS, endogenous in all strains. ∇, of wild-type (WT) or mutant MscL conductances from plasmids. Here, pressures ranged from nil to just above the MscL thresholds, corresponding to the feet of the activation curves (Center). (Center) MscL open probability vs. suction normalized to that needed to activate MscS, i.e., MscS-threshold sets unity of the abscissa. Smooth curves are Boltzmann fits. The wild-type curve is repeated as the broken lines to ease mutant comparisons. P_1/2: WT 1.9 ± 0.2 (mean ± SD, n = 4); V23A 1.4 ± 0.2* (n = 4); N15D 1.6, 1.4 (n = 2), G30E 1.5 ± 0.1** (n = 4), K55: 1.6, 1.6 (n = 2). Slope: WT: 0.08 ± 0.02; V23A 0.1 ± 0.04; N15D: 0.06, 0.04; G30E 0.09 ± 0.01; K55T 0.07, 0.08. ∗, P = 0.017; ∗∗, P = 0.007, t test against WT; all others P > 0.05; not tested where n is only 2. (Right) Open-time distribution histograms. In wild type, open dwell can be fitted with three components (24) with time constants of 38, 7, and <0.3 ms. Group 1–3 mutants showed shortened open dwell: V23A, all < 0.3 ms; N15D, 1.0 ms, and <0.3 ms; G30E, 10, 3, and <0.3 ms. K55T of Group 4 was like the wild type: 33, 8, and <0.3 ms.

The patch–clamp survey of all mutant channels and the detailed characterization of one from each class (Fig. 2) demonstrated that the extent of the mutant MscL hypersensitivity to stretch correlates well with the severity of the whole-cell phenotypes (growth and [K⁺]_in, in Table 1). MscLs of Group 1 mutants (very severe) are all activated at stretch forces far below the MscS threshold. Group 2 and 3 channels activate around that threshold. Group 4 (mildest) channels clearly activate above the MscS threshold but appeared more sensitive than the wild-type MscL [gauged as established (24); data not shown). Patch-clamp resolved differences among individual mutants within the same general phenotypic group (Table 1). For example, we found the Group 1 G22D and G22N channel to be even more sensitive than the V23A channel (Fig. 2), activating under patch–clamp without any applied suction (data not shown). Most mutant conductances flicker rapidly, especially the more severe ones. The open dwell of the wild-type conductances has two major components >1 ms that are missing in the V23A and N15D, for example (Fig. 2, right column). Thus, the more severe mutations also lower the transition barrier between the closed and the open state.

DISCUSSION

Most mutations were found to be between residues 13 and 30 (Fig. 3a). Unlike site-directed mutagenesis, forward genetics scans the entire target. That few sites lying outside this region caused very harmful channel misfiring indicates the importance of the region. Our screen apparently has approached saturation (especially for Group 1) because we found repeated hits at individual residues, often of the same base changes (see legend to Fig. 3). However, the span from residues 13–30 cannot be the sole determinant of all channel properties because gain-of-function mutations, especially the milder ones, exist elsewhere. Sixteen of the 20 directed mutations we made previously had little effect or make the MscL less active (24). Only K31D, K31E, Q56P, and Q56R are like some of the Group 3 or 4 mutants (10, 24). Although they were not encountered here, their immediate neighborhoods were (G30, K55). Bacteria, after 1-h induction, showed some 50–200 MscL conductance units per ≈6-μm² membrane patch. Leakage of the lacUV5 promoter was detected as 0–6 units per patch from uninduced cells (10, 23). This explains why some mutants grow poorly even in IPTG-free media (Table 1) but also raises the caveat that very extreme mutants would have vanished from all plates and therefore eluded our screen. However, even if one substitution has this effect, it seems likely that one of the 18 other possible substitutions at that site would yield a retrievable mutant.

Figure 3.

(a) Amino acid substitutions in gain-of-function mutants displayed on the wild-type MscL sequence. (b) Locations of mutated sites diagrammed on a membrane topology model of a MscL subunit. (c) A helical wheel of the inner half of M1, showing the clustering to one side the sites at which mutations have severe effects. In b and c, the dark gray and light gray symbols signify sites that can be mutated to give very severe (Group 1 in Table 1), only the less severe (Group 2 and 3), or only very mild (Group 4) mutants, respectively. The mutations are R13C (3;1+1*, i.e., three independent isolates from hydroxylamine mutageneses; and two from error-prone PCRs, one of which, marked with ∗, is a multiple mutant with one or more additional substitutions outside residue 13 through 30), G14E (1;0), N15D (0;3+1*), L19Y (1;2*), V21E (0, 1+2*), G22S (1;0), G22N (1;0), G22D (1*, 2), V23G (0, 1), V23A (0, 4+2*) V23D (0, 1+3*), G26S (2+3*;3+2*), G30E (5;1), G30R (3;0), G46D (3, 1), K55T (0;1), I92F (0;1), and N100D (0;1).

CD, Fourier transform infrared spectroscopy (28), and computer modeling (3, 5, 29) indicated MscL to be highly helical, and MscL-PhoA fusion tests (6) showed that the MscL peptide crosses the membrane twice, with both termini in the cytoplasm. The random mutagenesis study described here directs attention to one small region of the MscL protein, which appears to be one face of a transmembrane α-helix, as having important functional significance in the gating of the pore. One interpretation of our results is that this domain changes environment on channel gating. Because a large pore is generated on gating (18), suggesting large conformational changes, we suspect that these residues move between a more hydrophobic environment (lipids or hydrophobic protein interior) and a more hydrophilic one (hydrophilic residue interactions or aqueous phase such as protein exterior or channel lumen).

Some insight into the structural requirements for the implicated domain may be gained by assessing specific residue changes. For example, we find that a valine at position 23 is among the most important residues. Valine is one of the smaller hydrophobic amino acids. However, when this residue is substituted with an even smaller hydrophobic residue, alanine, a severe phenotype, is induced. Alanine replacement usually is thought to be inert and not change local protein structure. Hence, these data strongly suggest that the size of the residue at this location is important and that substituting a smaller residue causes mis-gating. This is supported further with the V23G mutation also found in this study. Three glycines also are implicated as being important. Glycines have several unique properties. Being the smallest residue, they are known to allow local flexibility and close protein packing. They are also mildly polar. Which of these properties are crucial cannot be clearly deduced from the mutations we found. On the other hand, a survey of all of the mutations in this region shows several substitutions that add an electric charge, suggesting that polarity is important. Perhaps the residues on this helical face move into an aqueous environment on gating, thus stabilizing the open conformation relative to the closed. Alternatively, these added charges might be within the membrane potential field, thus straining certain conformations (see ref. 10 for a more complete discussion). Although the random mutagenesis method described in this paper has defined a facet of a transmembrane α-helix as being important and gives us a first glimpse into structural requirements, further investigation will be needed to understand fully the molecular mechanisms behind the conformational changes associated with MS channel gating.

The importance of the transmembrane and relative location of at least some of the residues implicated in this may extend beyond the E. coli MscL protein. Twelve MscL homologues now have been found in a large variety of bacteria (19). Sequence comparison shows that the 1 Val and 3 Gly residues are among the most conserved, independently indicating their importance. (GXXX)n predicting side-chain-free facets in transmembrane helices of several proteins can be found in eukaryotic databases. Although MscL as a whole has no clear homologues in eukaryotic databases, searches using residues 13–30 identify a putative membrane protein in the fission yeast, Schizosaccharomyces pombe, which in turn has homologues in the budding yeast, the nematode, the mouse, and the human. We are currently testing whether they also are used in mechanosensation.

ABBREVIATIONS

MS

mechanosensitive

IPTG

isopropyl β-d-thiogalactoside