| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Biophys J, October 1999, p. 1960-1972, Vol. 77, No. 4
Laboratory of Molecular Biology, University of Wisconsin, Madison, Wisconsin 53706, USA
 |
ABSTRACT |
|---|
 TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Mechanosensitive channel large (MscL) encodes the large
conductance mechanosensitive channel of the Escherichia
coli inner membrane that protects bacteria from lysis upon
osmotic shock. To elucidate the molecular mechanism of MscL gating, we
have comprehensively substituted Gly22 with all other
common amino acids. Gly22 was highlighted in random
mutagenesis screens of E. coli MscL (Ou et al., 1998
,
Proc. Nat. Acad. Sci. USA. 95:11471-11475). By analogy
to the recently published MscL structure from Mycobacterium tuberculosis (Chang et al., 1998, Science.
282:2220-2226), Gly22 is buried within the constriction
that closes the pore. Substituting Gly22 with hydrophilic
residues decreased the threshold pressure at which channels opened and
uncovered an intermediate subconducting state. In contrast, hydrophobic
substitutions increased the threshold pressure. Although hydrophobic
substitutions had no effect on growth, similar to the effect of an
MscL deletion, channel hyperactivity caused by
hydrophilic substitutions corrrelated with decreased proliferation.
These results suggest a model for gating in which Gly22
moves from a hydrophobic, and through a hydrophilic, environmment upon
transition from the closed to open conformation.
| |
INTRODUCTION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Organisms must respond specifically to a variety
of pressure stimuli such as touch, gravity, barometric pressure,
turgor, and osmotic changes. It is becoming apparent that pressure
stimuli often cause ion channel opening, effectively transducing a
particular mechanical stimulus into an electrical or chemical one that
the cell can interpret (for reviews, see: French, 1992; Bargmann, 1994;
Sackin, 1995; Hamill and McBride, 1996; Kernan, 1997; Sukharev et al.,
1997; Sachs and Morris, 1998).
Mechanically gated ion conductances are found in over 30 cell types,
including animal tissues exposed to osmotic or cardiovascular deformation (Lansman et al., 1987; Christensen, 1987; Ubl et al., 1988). Such mechanosensitive conductances are also detected in plant,
yeast, and bacterial cells (Sukharev et al., 1997); in Escherichia coli, such conductances are called MscM, MscS,
and MscL
respectively, Mechanosensitive
channel Mini, Small, and
Largeconductances (Sukharev et al., 1993; Berrier et al.,
1996). MscL, which was the first mechanosensitive channel cloned
(Sukharev et al., 1994), forms a mutimeric channel (now thought to be a
pentamer) in the inner membrane that is both necessary and sufficient
to allow gating by membrane stretch (Berrier et al., 1989; Sukharev et al., 1994; Blount et al., 1996c; Häse et al., 1997; Saint et al.,
1998; Sukharev et al., 1999a). Each subunit has only 136 amino acids
that generate two 
-helical transmembrane domains, a connecting
periplasmic loop, and two cytoplasmic domains (Sukharev et al., 1994;
Blount et al., 1996b; Arkin et al., 1997), a topology recently
supported by the crystallographic structure of the Mycobacterium tuberculosis homologue, Tb-MscL (Chang et al., 1998). Sieving and
conductance studies show that the channel opens to a large, ~30-40-Å pore (Cruickshank et al., 1997; Sukharev et al., 1999b), which passes a large nanoSiemens current that is relatively nonspecific (Martinac et al., 1987, Sukharev et al., 1993, 1994). Several studies
support this channel's role in osmoregulation: 1) mscL gain
of function (GOF) mutants at K31 that lose potassium upon hypotonic
shock can be rescued by osmotically supported medium (Blount et al.,
1997); 2) hypotonic shock causes MscL channels to pass solutes,
including small proteins such as the 12 kD thioredoxin (Ajouz et al.,
1998); and 3) a marine bacterium can be rescued from osmotic lysis by
heterologously expressed E. coli MscL (Nakamaru et al.,
1999). The most compelling evidence shows that cells deprived of both
MscL and YggB, which contributes to MscS activity, lyse upon osmotic
shock. Loss of either gene alone has no effect (Levina et al., 1999).
Because MscL serves as a model for how a protein changes conformation
in response to membrane tension, much research has centered upon
determining the molecular basis for its gating. Two studies implicate
the lower half of the first transmembrane domain as important in
gating. The pentameric Tb-MscL crystallographic structure shows a
constriction between Ile14 and Val21 that is
formed by the lower half of all five TM1 helices (Chang et al., 1998).
The biological significance of this highly conserved region (Moe et
al., 1998) was underscored by random mutagenesis of E. coli
(Ou et al., 1998). In this genetic screen, mutations that resulted in
slow or no growth phenotypes were isolated. Although the entire
E. coli mscL gene was mutagenized, 14 of 18 mscL
mutants isolated had substitutions in one of the amino acids ranging
from positions 13 to 30, which correspond to residues 11 to 28 of
Tb-MscL. The most severe growth defects correlated with channels that
open with little or no stretch force (Ou et al., 1998). Among this group were numerous mutations in Gly22 (Ou et al., 1998), a
residue highly conserved among bacteria (Moe et al., 1998). By analogy,
the Tb-MscL crystal structure gives us a static snapshot of the closed
E. coli channel. Given that the open pore of the E. coli channel must be huge to account for its ability to conduct
bulky solutes (Cruickshank et al., 1997; Ajouz et al., 1998; Sukharev
et al., 1999b), and given that the solved structure appears to
represent a closed state with only a 3-Å opening at the cytoplasmic
end (determined by examination of the structure from Chang et al.,
1998), there must be an enormous change in protein conformation upon
channel gating. Here, we have tested the contribution to channel gating
of one residue, glycine 22 (G22), which, by analogy to Tb-MscL, would
be buried within the wall of the constricted pore (Chang et al., 1998).
We have replaced this residue with all 19 other common amino acids and tested the effects on cellular growth and channel gating. Our results
suggest a model for the environmental changes that occur in this region
of the channel upon transitioning from the closed to the fully-open state.
| |
MATERIALS AND METHODS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Mutagenesis of mscL
We have generated all possible 19 amino acid changes at MscL's
amino acid 22 using a megaprimer polymerase chain reaction (PCR)
strategy (Barik, 1993). This strategy required two rounds of PCR, both
using as template the wild-type mscL in pB10b (Moe et al.,
1998; Ou et al., 1998). The first PCR was primed with either a precise
or degenerate oligonucleotide encompassing the codon we wished to
change. Synthesis of the complementary strand was primed from 17 nucleotides outside the 3' end of the open readin frame (ORF) using an
Xho1-tagged primer sequence that extended into the pB10b vector: 5'-GTT
CGC GGA CTT TCG TCg agc tcg agc tc-3' (Sukharev et al., 1994). We then
used this product as a megaprimer to generate the entire ORF using a
BglII-tagged 5' primer that included the initial ATG (5'-AGA TCT AGA
TCT CAT AGG GAG AAT AAC ATG-3').
The PCR product containing the mutant mscL ORF was then gel purified
(QIAGEN Inc., Valencia, CA), digested with XhoI and BglII, and ligated
into an antisense plasmid pB10c (Moe et al., 1998). This construct was
then transformed by electroporation (Biorad, Hercules, CA) into the
mscL knockout strain PB104 (Blount et al., 1996c) and the
amplified DNA extracted (QIAGEN) for sequencing by the University of
Wisconsin Biotechnology Center (Madison, WI), using Amplitaq or BigDye
(Perkin-Elmer Biosystems, Foster City, CA). After the engineered amino
acid change was detected, the ORF was excised with XhoI and BglII and
ligated into the plasmid PB10b, which can express the ORF under the
control of a lac-inducible promoter (Moe et al., 1998). All cloning
steps and expression were performed within the mscL knockout E. coli PB104. Each entire ORF was sequenced through the flanking
BglII and XhoI sites described above to ensure that there were no
unintended changes.
Bacterial growth assays
Plate assay
Single colony isolates from Luria-Bertani (LB) + amp (ampicillin 100 µg/µL in all cases) (Lech and Brent, 1995) plates of the
mscL-null E. coli PB104 harboring the G22X mutant
plasmids were grown overnight in liquid LB + amp. Typically, cells were at an OD600 of 1.4 to 1.5 at this point although the slow
growing G22D and G22E were at ~1.2. The cells were diluted 1:10 into
fresh LB + amp and grown on the shaker at 37°C for an additional
hour. Then the cells were serially diluted and plated in the presence or absence of 1 mM isopropyl-
-D-thiogalactoside (IPTG)
(Research Products International, Mt. Propsect, IL). The plates were
incubated for 19 h at 37°C when growth was scored. The assay was
performed at least three times using either one or two single colony
isolates during each assay.
Liquid growth
Single colony isolates from LB + amp plates of PB104 harboring the G22X mutant plasmids were grown for 17-18 h in liquid LB + amp at 37°C. Then they were diluted into prewarmed (37°C) medium (100 µL in 2 mL) and grown for an additional 2 h. They were then further diluted to an OD650 of 0.02 in 25 mL of prewarmed (37°C) LB + amp, each in a 125-mL flask. The OD650 was determined every half hour during the subsequent growth. After 2 h 15 min, IPTG was added to the flask to a final concentration of 1 mM to induce expression of the G22X plasmid. The next OD650 reading was taken 20 min later and, thereafter, every 30 min. The growth data from six different experiments using half of the G22X mutants plus controls in each experiment were subjected to statistical analysis and expressed as the mean and the standard deviation of the population (Microsoft Excel, Microsoft Corp., Redmond, WA) at each data point. n = 3 or 4 for each G22X mutant and n = 4 and n = 12 for the knockout strain with the empty vector and wild type, respectively. The growth rate after induction was determined using the LINEST function (Microsoft Excel) on the data generated between time 3:05 h (50 min after induction) and 5:05 h.Electrophysiological techniques
Spheroplast preparation
E. coli spheroplasts were prepared and wild-type and mutant MscL activity was recorded essentially as previously described (Blount et al., 1999). Cephalexin (final concentration 0.06 mg/mL) was
added to log-phase cells growing in modified LB medium that contained
0.5% NaCl instead of 1% NaCl (Martinac et al., 1987). After
incubation for 1.5-3 h, IPTG was added (final concentration 1.3 mM) to
induce mscL expression. The induction time varied from 5 min
to 1 h. Longer preincubation and induction were necessary when a
mutant with a slow-growth phenotype was used. The cells were then
harvested and lysed with lysozyme (0.2 mg/mL), and spheroplasts were
collected by centrifugation.
Patch clamping
Electrodes with resistances of 3.5-4.5 M
(bubble number = 4.2-4.8) were used to clamp inside-out patches. Pipette solutions contained 200 mM KCl, 90 mM MgCl2, 10 mM CaCl2,
and 5 mM Hepes (pH 6.0), whereas the bath solution additionally
contained 0.3 M sucrose to stabilize the spheroplasts. Currents were
measured with a List EPC 7 amplifier (List Medial, Darmstadt, Germany) and filtered with an 8-pole Bessel filter at 3 kHz. The potential of
the pipette was held +20 mV higher than that of the bath. Current recordings were digitized at 10 kHz with Digidata 1200 interface using
Clampex ver. 7.0.0.86 software (Axon Inst., Foster City, CA) and stored
in a PC. Pressure was applied by syringe-generated suction through the
patch-clamp pipette and measured with a pressure guage (143PC05D,
Honeywell, Minneapolis, MN).
Data were analyzed using FETCHAN ver. 6.0.5, pSTAT ver. 6.0.5, and
AxoGraph ver. 3.5.5 software (Axon Inst.). The kinetics of the channels
that had a high threshold were not examined in all MscL mutants because
the pressure needed to activate the MscLs was close to the lysis
pressure of the membrane.
Determination of the gating threshold
The threshold of MscL (or mutant MscL) gating was expressed as the ratio of the pressure required to gate MscL (or a mutant MscL channel) relative to that of MscS (Blount et al., 1996a). For mutant
MscLs showing an open substate, gating into the substate was defined as
the threshold. The absolute value of this threshold varied most likely
due to variation in the geometry of the patch (Sukharev et al., 1999b).
| |
RESULTS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
The hydrophobicity of amino acid 22 systematically affects the growth of E. coli harboring the mutant MscL protein
Plate growth
Bacteria were grown and then plated in serial dilutions on LB + amp plates with or without the inducer, IPTG. When the mutant construct was induced, there was a gradient of growth inhibition that correlated with the hydrophobicity of residue 22 (Fig. 1). In the presence of IPTG, growth was completely absent for bacteria harboring mscLs containing the most hydrophilic substitutions at amino acid 22 (Fig. 1 A). Substituting an amino acid having an intermediate hydrophobicity, such as proline, tyrosine, or serine, slowed growth. In these cases, growth was evident, but colonies were noticeably smaller after only 19 hr of growth (Fig. 1 A) and, after 42 h of growth, colony size was still restricted (data not shown). Hydrophobic substitutions showed no visible effect on growth or viability after 19 h at 37°C (Fig. 1 A). Therefore, the hydrophobicity of the substitution at G22 of MscL systematically affected the plate growth of the mscL-null strain harboring this construct.
|
). This extreme growth inhibition in the
absence of induction was not seen with basic substitutions at G22,
showing that an acidic substitution at residue 22 has an added
detrimental effect not yet understood.
Growth in liquid media
The G22X series of strains was also grown in liquid LB + amp as described in Materials and Methods. Cells that were in exponential steady-state growth were induced, and their subsequent growth was followed with optical density measurements at 650 nm. Figure 1 B reveals the same three groups highlighted by the growth on plates. Those inhibited by expression of the G22X plasmid, including all basic and some polar substitutions, hug the bottom of the graph. Those significantly slowed by expression of the G22X plasmid, i.e., G22P, G22S, and G22Y, comprise the middle group. Those unaffected by expression of the G22X plasmid, i.e., the remaining hydrophobic substitutions, along with wild type and the knockout plus empty vector, comprise the upper group. The growth rates clearly correlate with the hydrophobicity of the residue 22 substitution (Fig. 1 C). Hydrophobicities given here and below are from Kyte and Doolittle (1982). There is a sharp threshold of allowed hydrophobicity at which
growth rates vary greatly (for example, 0.37 OD650 units/hr for wild type of hydrophobicity 
0.40 versus 0.007 OD650
units/hr for G22T of hydrophobicity 0.70). On either side of this
narrow hydrophobicity threshold between 1.3 and 0.40, growth is
either vigorous (up to a hydrophobicity of 4.5) or completely inhibited (down to a hydrophobicity of 4.5) (Fig. 1 C).
It was not possible to examine G22D and G22E using the liquid growth
assay, because these strains were consistently slow to come out of
stationary phase resulting from the overnight growth and were not in
steady-state growth at the time of induction.
The threshold for mechanosensitive gating corresponds to the hydrophobicity of residue 22
Channel activity was recorded from excised patches from
spheroplasts expressing the various plasmid-borne mscL
alleles in an mscL deletion strain. When negative pressure
(suction) was applied to the membrane, two types of channel activities
were readily observed. Figure
2 A,ii shows that
MscS (
), native to all strains, was activated at a lower suction and
passed about 25 pA of unitary current (at 20 mV). Wild-type MscL
(
) activated next with greater suction, at a pressure 1.6 times that
required to open MscS, and the MscL channel's unitary current was
about 80 pA (Fig. 2 A,ii). Because MscL and MscS
gate in response to membrane tension, not pressure, individual patch
geometries affect the pressure dependence of activation. To account for
variable patch geometries, the gating threshold of MscL (or the mutant G22X channels) is given as the ratio of the suction at which MscL (or
the mutant MscL) is found to gate relative to that at which MscS opens
(see Materials and Methods). For wild-type MscL, that threshold was
1.64 ± 0.08 (mean ± SD, n = 9) over a range
of 110-190 mmHg (Table 1).
|
|
We determined the gating threshold ratio of each of the G22X MscL channels expressed in an mscL knockout background (Table 1). When a hydrophobic residue such as alanine was substituted for glycine 22, higher pressure was required to open this channel (compare G22A and wild type traces, Fig. 2 A, i and ii). The gating threshold pressure ratio of 2.47 ± 0.20 (G22A/MscS, n = 4) was significantly higher than the wild-type MscL/MscS ratio (1.64) (Table 1). In contrast, when glycine 22 was substituted with a polar or charged amino acid, the mutant MscL channel was activated at a pressure lower than that required to gate wild-type MscL. For example, the negative pressure required to open the polar substitution mutant G22T was near that required to open MscS, whereas the hydrophilic substitution mutant G22K opened before MscS (Fig. 2 A, iii and iv).
The effect of the hydrophobicity of residue 22 on the gating threshold of all 20 possible G22X substitutions is summarized in Fig. 2 B and Table 1. Hydrophilic substitutions eased gating, whereas more hydrophobic substitutions hampered gating. The MscL gating threshold rose from 0 (i.e., the channel was active in the absence of intentionally applied suction) to about 2 in parallel with the increase in hydrophobicity of the amino acid at position 22. The threshold of MscLs with a very hydrophobic substitution may be an underestimate because we sometimes did not observe MscL activity up to the lytic pressure (about 2.5 times the MscS threshold) of the patch membrane. The sign of the charge that leads to hydrophilicity did not affect the threshold at our resolution (compare G22R or G22K with G22D or G22E, Fig. 2 B and Table 1).
Combining the results presented in Fig. 2, A and B, one can see that the growth rates of the G22X mutants largely correlate with the gating thresholds of their MscL channels (Fig. 2 C). Although the relationship is clear for cells harboring channels whose residue 22 is clearly more or less hydrophobic than glycine, the relationship breaks down for G22X mutants for which the hydrophobicity of residue 22 is close to that of glycine.
The hydrophobicity of amino acid 22 affects dwell times in the fully open state and in an open substate
The wild-type MscL channel gates through several transient
subconductance states to the fully open state of about 3 nS upon the
application of a membrane tension (Fig.
3 A,ii and
Sukharev et al., 1994). Its full-open time distribution (Fig.
3 B,ii) can be fitted with three Gaussian
components dominated by the one with a time constant of about 20 ms. In
this study, we registered the three time constants (T) and their
proportions (in parentheses) to be 19.4 ± 2.5 ms (0.67), 2.5 ± 1.1 ms (0.09), and 0.1 ± 0.1 ms (0.24, n = 3),
not significantly different from those previously reported (Fig.
3 B,ii and Blount et al., 1996b).
|
Replacing G22 with the more hydrophobic alanine caused a large increase in open dwell. Here, the open time was largely accounted for by a single Gaussian distribution with a T of 145 ± 33 ms (Fig. 3 B,i). Continuous openings for over 0.5 s were often encountered (Fig. 3 B,i). Substitutions with other hydrophobic residues gave openings similar to wild type (Fig. 3 C).
In contrast, replacing G22 with threonine, a mildly hydrophilic residue, shortened the open dwell, now dominated by events at or below 2 ms (T = 0.91 ± 0.28 ms) (Fig. 3 B,iii). This gave the flickery impression of the mutant channel activities (Fig. 3 A,iii). Substitutions with two other mildly hydrophobic residues, P and S, gave results more similar to wild type (Fig. 3 C).
Replacing G22 with a charged or strongly polar residue yielded channel
activities that were very flickery, e.g., G22K MscL activity had a
fully open dwell dominated by events shorter than 0.5 ms (T = 0.24 ± 0.08 ms, Fig. 3 A,iv and
B,iv). In addition, these channels also lingered
in a subconductance state. Fig. 3 A,iv and its
inset show dwells in such a substate for tens of milliseconds (T = 11.0 and 1.1 ms) in the case of G22K. Here the substate conductance is
about 0.5 nS in conductance, about 
of the full unitary
conductance, and most similar to the conductance of the lowest substate
in the wild type (Sukharev et al., 1999b). Wild-type MscL rarely stays
in this subconductance state for more than 1 ms. Substitution of G22
with the positively charged R, K, or H, the negatively charged D, or
their amine equivalents, N or Q, all resulted in this cluster of
biophysical phenotypes: low gating threshold, flickery activity, and a
stable substate at or near the lowest subconductance level (Fig.
3 C). G22E showed a stable substate but had a channel
opening (9.9 ms) longer than these mutant MscLs. The results indicate
that these toxic channels (Fig. 1) disfavor the closed states, but,
among the open states, they favor the lowest subconductance state over
the fully-open state at intermediate open probability.
Hypersensitive mutant MscLs are pressure sensitive at all states
When an increasing suction ramp was applied to the membrane containing hypersensitive mutant MscL channels, the prominent lowest open substate appeared first and the full-open state occurred at higher pressure. Thus, such mutant MscLs first go through a low-threshold C-to-S transition and then, with increasing pressure, through a high-threshold S-to-O transition, where C is the closed, S is the first substate, and O is the fully-open state. We tested the mechanosensitivity of different states by evaluating the transitions at different pressures for G22N MscL, a channel that is active even in the absence of applied suction (Fig. 4).
|
Without suction, G22N fluctuated mainly between the closed and the open substate without full opening (Fig. 4 A, top trace). When a moderate suction was applied to the membrane (105 mmHg), the channel largely remained in the substate but flickered to the fully open state (Fig. 4 A, middle trace). The closed state was infrequent, indicating that the C-to-S transition was almost saturated toward S. When a higher pressure was applied (210 mmHg), the channel stayed mostly in the fully-open state and flickered back to the substate, as its S-to-O transition approached saturation (Fig. 4 A, bottom trace). These recordings at static pressures confirmed the observation in the pressure-ramp experiment. We next used such data for the quantitative evaluation of the mechanosensitivity of each transition.
If we assume a serial three-state model, in which the channel is in one
of the three states, C, S, or O, the equilibrium can be expressed as
|
(1) |
|
(2) |
|
(3) |
![P=<FR><NU><UP>exp</UP>[(p−p<SUB>1/2</SUB>)/s<SUB><UP>p</UP></SUB>]</NU><DE>1+<UP>exp</UP>[(p−p<SUB>1/2</SUB>)/s<SUB><UP>p</UP></SUB>]</DE></FR>](/content/vol77/issue4/fulltext/1960/img005.gif)
|
(4) |
P)] versus
pressure (Martinac et al., 1987). The p1/2 was
56 ± 37 mmHg for the C-to-S transition and 132 ± 18 mmHg
for the S-to-O transition (n = 3). This indicates that
the fully open state is brought about by at least two steps that are
distinct with respect to conductance and pressure sensitivity. All
mutant MscLs of the lower group in Fig. 3 C behaved in a
manner similar to G22N. Quantitative analysis on another member of this
group, G22D, gave results similar to those of G22N shown here.
pH Affects the mechanosensitivity of G22H MscL
Because the hydrophobicity of residue 22 governs mechanical gating and channel kinetics, one might expect that these parameters would change with pH, if H+ is accessible to histidine 22 of G22H MscL. Because histidine has a pKa of 6.5, we examined the activities of G22H MscL at pH 6.0 and at pH 7.5. Preliminary experiments at a lower pH (pH 5.0) in the bath and/or pipette solution showed little difference from results obtained at pH 6.
When the pH of the bath solution (i.e., the cytoplasmic side of an
inside-out patch) was 6.0, which allows the histidine (if exposed) to
become more positively charged from the cytoplasmic side of the
channel, the G22H channel acted like a hydrophilic substitution mutant.
It opened with very little suction: it gated at a suction that was less
than half that required to open MscS (Fig.
5 A, left side).
With increasing suction, G22H showed two conductance states: an open
substate with an amplitude of about 20 pA and a full-open state with an
amplitude of about 80 pA (all at 20 mV). When the pH of the bath
solution was increased from pH 6.0 to 7.5 by perfusion, the G22H
channel acted like a channel with an uncharged residue 22, and opening
G22H required greater suction than that required to gate MscS (Fig.
5 A, right side. Note the reordering of the two
thresholds marked by the arrow for G22H and MscL, and the arrowhead for
MscS). This change was reversible. A paired t-test of the
change in the pressure needed to activate G22H MscL (48 ± 10 mmHg
at pH 6.0 and 156 ± 63 mmHg at pH 7.5, p < 0.05, n = 3) indicated a significant decrease in sensitivity
with this increase in pH on the cytoplasmic side. In contrast, the
sensitivity of wild-type MscL was not affected by the pH of the bath
solution (Fig. 5 B and Blount et al., 1996c).
|
These results showed that histidine 22 was accessible to cytoplasmic
H+. To examine the effect of periplasmic pH, the tip of the
pipette was filled with pH 6.0 solution (with 0.3 M sucrose, to prevent immediate mixing) and back-filled with pH 7.5 solution (without sucrose, to allow a gradual change in pH at the extracellular side)
(Blount et al., 1996c). The suction needed to gate G22H MscL was half
that needed to open MscS within 10 min after filling the pipette and
before significant mixing of the pH 6.0 and pH 7.5 solutions (Fig.
5 C, left side). When the same patch was
examined after 1 h, the G22H MscL opened only when a suction
greater than that needed to open MscS was applied (Fig.
5 C, right side). A paired t-test
(52 ± 13 mmHg at pH 6.0 and 152 ± 57 mmHg at pH 7.5, p < 0.05, n = 3) indicated that this
increase was also statistically significant. In the converse experiment
of tip filling at pH 7.5 and back filling at pH 6.0, we observed the
reverse change, i.e., the G22H channel first had a high and then a low
pressure threshold (data not shown). The sensitivity of wild-type MscL
did not change with the periplasmic pH (Fig. 5 D). The
change in the gating threshold of G22H MscL was not simply dependent on
time because the sensitivity did not change when the pipette and bath
solution were both kept at pH 6.0 for up to two hours (data not shown).
When both the pipette and bath were at pH 7.5, the threshold also did
not change with time; again, more suction was required to gate G22H
than MscS as in the pH 6.0/pH 7.5 experiments shown above (data not shown). These data suggest that 22H is accessible to protons from both sides of the membrane.
| |
DISCUSSION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Besides being the smallest amino acid, glycine is a residue of
intermediate hydrophobicity (Kyte and Doolittle, 1982 and Table 1). G22
is located in the first transmembrane helix of MscL and is largely
conserved, although alanine is an allowed substitute in more distant
bacterial homologues (Sukharev et al., 1997; Chang et al., 1998; Moe et
al., 1998). By analogy with the M. tuberculosis-MscL structure, this glycine is virtually buried just under the outermost constriction of the closed channel and faces an adjacent subunit helix,
which also contributes to the closed pore (Chang et al., 1998, Batiza
et al., 1999). Changing glycine 22 of E. coli's MscL to a
more hydrophobic residue (Table 1) results in a channel that requires
greater tension to open (Fig. 2) and that has a comparable (or longer)
dwell time in the fully open state (Fig. 3). However, the bacteria
tolerate these changes well (Fig. 1). In contrast, hydrophilic
substitutions at G22 reduce cell viability (Fig. 1). They result in
easy gating of the mutant mscL channel (Fig. 2) and channel flickering
in the fully open state (Fig. 3) with the exception of G22E (see
Results). The most hydrophilic substitutions also reveal the presence
of a stable open substate (Figs. 3 and 4). No other parameter of the
side chains, such as the sign of charge and the size of side chain,
seems to correlate with the gating defects systematically. For example,
the G22A and G22I substitutions have almost identical effects on
channel kinetics (Fig. 3) and cell growth (Fig. 1) although the size
differs significantly (the van der Waals volumes of alanine and
isoleucine are 67 Å and 124 Å, respectively).
Our data indicate the relative hydrophobicity of a residue's
environment inside the gate before, during, and possibly after opening
(Fig. 6 A). A change to a
more hydrophobic amino acid at residue 22 makes the channel harder to
open, whereas a more hydrophilic amino acid at residue 22 makes the
channel open more easily to a substate (Fig. 2). Thus, the
hydrophobicity of residue 22 affects the energy barrier between the
closed state and the first open substate. [Note that the transition
between the closed state and the first open substate is the major
pressure-dependent step in wild-type gating (Sukharev et al., 1999b)].
More hydrophobic substitutions increase this barrier, whereas more
hydrophilic substitutions decrease this barrier. Every logical
explanation that accounts for these specific changes (i.e., changing
the depths of the energy wells and/or the height of the transition
barrier peak) requires residue 22 to be in a relatively more
hydrophobic environment in the closed state than in the open substate
(Fig. 6 A, closed state and open substate).
|
The dwell-time kinetics (Fig. 3) may reveal more details about the
environment of residue 22. Based upon the closing rate constant data
presented by Sukharev and coworkers, WT exhibits a 9-fold increase in
the open dwell-time over the pressure range of opening MscL in
reconstituted liposomes (Sukharev et al., 1999b). Because the dwell
time determinations for the G22X mutants were made at
PO < 0.1 (Fig. 3 B), by
necessity, the pressures at which these measurements were made varied
considerably depending upon the particular G22X gating threshold. In
addition, the measurements presented here were made of channels in
spheroplasts. Given these caveats, the trends presented here suggest
that the hydrophilicity of this substitution decreases the open dwell
time about 2-3 fold greater than that expected in wild type. Thus, the
rates of the backward reactions (O
S) increase with the
hydrophilicity of the G22X substitution; therefore, the hydrophilicity
of residue 22 decreases the energy barrier between the fully open state
and the open substate, suggesting that residue 22 is in a more
hydrophobic environment in the fully open state than it is in the open
substate (Fig. 6 A, open substate and fully open state).
Given the strong sequence conservation, the closed conformation of the
M. tuberculosis homologue (Chang et al., 1998) presumably describes that of the E. coli MscL channel as well (Batiza
et al., 1999). The crystallographic structure of Tb-MscL shows that, in
the closed state, residues I14 through V21 define the closed gate
presumably restricting water passage within. The corresponding residues
in E. coli MscL are V16 through V23. Tb-MscL's A20 (which corresponds to E. coli's G22) is buried within the walls of
this closed fist, because it is in van der Waals contact with Tb-A18 (which corresponds to A20 in E. coli) in an adjacent TM1
helix. Therefore, E. coli's G22 would likely be within a
hydrophobic environment in the closed state (Fig. 6 A,
closed). This is consistent with our conclusion that G22 is in a more
hydrophobic environment when the channel is closed than when it is
partially open.
One would expect large movements of the transmembrane helices during
opening to account for the increase from 0 to 3 nS in conductance. M1,
in particular, is expected to swing toward the periphery, to straighten
up, and, perhaps, to rotate around its own long axis in the process
(Chang et al., 1998). The cytoplasmic end of M1 is expected to traverse
a distance of some 15 Å, because sieving (Cruickshank et al., 1997)
and conductance analyses (Cruickshank et al., 1997; Sukharev et al.,
1999b) estimate the fully-open pore to be 30-40 Å in diameter. This
requires all transmembrane helices to line the fully open pore
(Cruickshank et al., 1997; Chang et al., 1998; Sukharev et al., 1999b).
Our data suggest that, during this movement, residue 22 is in a
hydrophilic environment (possibly facing the aqueous lumen) in a
partially open state that corresponds to the subconductance state
favored when the residue is charged (Fig. 6 A, open
substate). In the fully open state at the end of this movement, our
dwell-time data are best explained by placing residue 22 in a more
hydrophobic environment relative to the open substate, such as its
being in contact with hydrophobic residues of a neighboring M2 (Fig.
6 A, fully open state) or possibly in contact with the
lipid bilayer. To achieve this, M1 may rotate clockwise or
counterclockwise (Fig. 6 A, fully open state) to minimize
previously buried hydrophobic surfaces now exposed to the lumen (Chang
et al., 1998; Batiza et al., 1999). However, one cannot rule out
residue 22's being exposed to the lumen in the fully open state.
G26 and L19, among other M1 residues including G22, were also
highlighted by the GOF screen by Ou and colleagues after random mutagenesis (Ou et al., 1998) (Fig. 6 B). By analogy with
the Tb-MscL structure, L19, G22, and G26 would be near corresponding residues V17, A20, and I24 in an adjacent M1 domain when the channel is
closed (Chang et al., 1998, 1msl). The recovered mutants of G26S, L19Y,
and G22D, N, and S, were all changes to more hydrophilic residues, and
these mutations all resulted in "severe" or "very severe"
growth defects, an increase in the sensitivity to stretch, and
flickering kinetics (Ou et al., 1998). These results corroborate the
present interpretation of the G22 results: a hydrophilic substitution at these residues disrupts the hydrophobic interactions that keep the
channel closed. The model for MscL opening defined by our results (Fig.
6 A) suggests that this face of M1 highlighted by the Ou et
al. (1998) GOF mutants (Fig. 6 B) might experience
environmental changes during channel opening similar to those proposed
for G22. These results also underscore the rationale for conservation
of glycine, a residue of intermediate hydrophobicity, at amino acid 22. Glycine 22 is presumably one of several residues defining the
constricted region of the pore that oppose the tension required for
gating. Any deviation changes the gating properties of the MscL
channel. Five MscL homologues having a glycine at this position that
were tested by Moe and coworkers opened at a tension similar to that
required to open wild type (Moe et al., 1998). However, the MscL
homologues of Staphylococcus aureus,
Synechocystis, and M. tuberculosis have an alanine at
this relative position. Although S. aureus opens at a
tension similar to that of wild type, Synechocystis requires
three times the suction of MscS to open (Moe et al., 1998), a value
similar to the 2.47 gating threshold ratio of G22A (Table 1). It will
be interesting to see whether or not the Tb-MscL channel is relatively stiff.
Interestingly, none of the residues in an adjacent M1 helix presumably
close to L19, G22, and G26 in the closed channel (i.e., V17, A20, or
I24 on the opposite face of the M1 helix) was recovered in the Ou and
coworkers' screen (1998). This suggests that V17, A20, and I24 do not
undergo environmental changes similar to those of L19, G22, and G26.
They may remain in a hydrophobic environment throughout the gating
process, shielded by the adjacent helices and the lipid bilayer.
Arkin et al. (1998) have shown by amide
H+/D+ exchange that two-thirds of the whole
MscL protein is water accessible and suggested that MscL may have a
wide aqueous vestibule. However, the length of the closed gate revealed
by the Tb-MscL structure (Chang et al., 1998) makes it surprising that
the G22H MscL changed its sensitivity when either the bath or pipette
solution was changed to pH 7.5 (Fig. 5). This was not due to proton
leakage because 1) the high pH (i.e., low proton concentration)
determines the sensitivity, 2) the seal resistance is as high as about
4 G, and 3) this occurs even when the pH of the pipette (held +20 mV relative to potential in the bath) was higher than that of the bath.
Therefore, this histidine is accessible to protons from either the top
or bottom of the channel. The accessibility from both sides may be due
to the presence of two protonation sites in histidine and a cleft
induced by the increased size of residue 22, and/or multiple
conformations in the closed state.
The increase in open probability of MscL with membrane tension has been
attributed largely to a decrease in channel closed time. Sukharev et
al. (1999b) examined the tension dependence of the transitions among
the closed state, several open substates, and the fully open state in
wild-type MscL. They identified at least three substates in the
wild-type MscL, which occur much less frequently than the substate
revealed by the present G22X MscLs. Only one of the transitions, namely
from the closed to the first substate (C
S1), was shown
to be dramatically tension dependent. Their results can also be
interpreted to mean that other transitions occur at tensions lower than
that for the CS1 transition in the wild-type. We used
the opportunity presented by the G22N MscL with its prominent
S1 and found both CS and SO to be mechanosensitive.
We found that the tension needed for the S-to-O transition (100-150
mmHg) is lower than the suction required for the C-to-S transition
(160-300 mmHg). Thus, in the G22N MscL, and probably in wild-type MscL
as well, membrane tension pulls on M1 through all stages of gating.
An electromechanical model by Gu et al. (1998) indicates that a
domain of the N-terminus region of MscL swings as a gate. The gating is
proposed to be brought about by tilting the coulombic forces between
the charged residues of the N-terminal, C-terminal, and membrane
spanning domains. Our finding that the sign of the charge at position
22 is not important in determining the tension sensitivity is not
consistent with this model. We suspect that the hydrophobic
interaction, rather than electrostatic interaction, will play a
significant role in gating because the latter force is relatively low
across the aqueous phase of the lumen.
We are pleasantly surprised by the way in which the hydrophobicity of residue 22 can predict both a biophysical parameter, the gating threshold (Fig. 2), and a complex physiological one, the growth rate (Fig. 1). Figure 2 C shows that the gating threshold correlates well with the growth rate. However, the parallel is not complete. Several factors may contribute to inconsistencies for channels having a substituted amino acid whose hydrophobicity is close to that of glycine. Perhaps some parameter for which glycine has been precisely chosen, in addition to its effect on the gating threshold, contributes to MscL functioning during cell growth. Also, variation in the amount of channel protein expressed in each type of transformant is possible, and the growth rates may be affected by these differences. Nonetheless, the correlation between the mutant MscL's gating behavior and the growth rate of the mutant population seems remarkable, given that one describes a physical parameter of a single protein in vitro, whereas the other describes a characteristic of a population of living cells. Additionally, most of the growth-sensitive channels described here are more sensitive to stretch, but not all have flickery openings (Figs. 1 and 3), suggesting that the former but not necessarily the latter is detrimental to growth. Both of the mutants having acidic substitutions are unusually poor growers, perhaps due to charge-specific filtering when the channel is in the partially-open substate.
Plants, fungi, bacteria, and other walled cells maintain a large turgor
that is used to disrupt the bonds cementing wall material so that new
wall can be added during growth (Koch and Woeste, 1992). MscLs
that activate at lower tensions interfere with growth, presumably
because the turgor needed for growth cannot be attained due to channel
opening (Fig. 2 C). MscLs with a threshold higher than
normal do not affect growth possibly because the cells can maintain
normal turgor, and MscL is not activated at normal membrane tension. In
contrast, the huge turgor imposed by a severe hypo-osmotic shock
requires pressure valves to jettison solutes from the swollen cell.
Although it has been postulated that mechanosensitive channels serve
this function (Berrier et al., 1992; Ajouz et al., 1998), two recent
reports substantiate this idea: 1) Heterologously expressing MscL
rescues the marine bacterium Vibrio alginolyticus from lysis upon osmotic downshock (Nakamaru et al., 1999); and 2) although YggB is
necessary for MscS activity, the double mutant
mscL
yggB lyses after
a severe hypotonic shock (Levina et al., 1999). It would be interesting
to find out which of the G22X channels can protect these organisms from
downshock lysis.
In summary, the intermediate hydrophobicity of G22 is a key element of
MscL channel's pressure sensitivity and possibly, the dwell time. When
this residue is altered, the hydrophobicity of the substitution
determines the channel gating characteristics and cell viability. In
addition, changing G22 to a hydrophilic residue reveals a
pressure-dependent open substate, and further analysis shows that both
this substate and the fully open state are pressure sensitive in
hypersensitive mutants. A model for moving G22 into different
environments during gating has been proposed, which is consistent with
the recently published structure of Tb-MscL (Chang et al., 1998). This
comprehensive analysis can be applied to other highly conserved channel
residues to further dissect mechanosensitive gating.
| |
ACKNOWLEDGMENTS |
|---|
The authors thank Yoshiro Saimi and Steve Loukin for many helpful discussions regarding the conclusions of this study. We also are indebted to Jean Yves Sgro of the Institute for Molecular Biology and Doug Davies of the Enzyme Institute, both at the University of Wisconsin, Madison, for assistance in analyzing the Tb-MscL structure. We also thank Leanne Olds for assistance in preparing figures. This study was supported by National Institutes of Health grant GM 47856 and the visit of K.Y. was supported by the Ministry of Education, Science, Sport, and Culture of Japan.
| |
FOOTNOTES |
|---|
Received for publication 12 April 1999 and in final form 28 June 1999.
Address reprint requests to Ching Kung, Laboratory of Molecular Biology, University of Wisconsin, Madison, WI 53706. Tel.: 608-262-9472; Fax: 608-262-4570; E-mail: ckung{at}facstaff.wisc.edu.
* The first two authors contributed equally.
Dr. Yoshimura's present address is Department of Biological Science, Graduate School of Science, University of Tokyo, Hongo, Tokyo 113-0033, Japan.
Mr. Schroeder's present address is Enzyme Institute, University of Wisconsin, Madison, WI 53706.
Dr. Blount's present address is Department of Physiology, University of Texas, Southwestern Medical Center, Dallas, TX 75235.
| |
REFERENCES |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
the MscL gene, protein, and activities.
Ann. Rev. Physiol.
59:633-657[Abstract/Full Text].
Biophys J, October 1999, p. 1960-1972, Vol. 77, No. 4
© 1999 by the Biophysical Society 0006-3495/99/10/1960/13 $2.00
This article has been cited by other articles:
 |
|
 |
|
  K. Yoshimura, J. Usukura, and M. Sokabe Gating-associated conformational changes in the mechanosensitive channel MscL PNAS, March 11, 2008; 105(10): 4033 - 4038. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Nomura, M. Sokabe, and K. Yoshimura Interaction between the Cytoplasmic and Transmembrane Domains of the Mechanosensitive Channel MscS Biophys. J., March 1, 2008; 94(5): 1638 - 1645. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Carney, J. M. East, and A. G. Lee Penetration of Lipid Chains into Transmembrane Surfaces of Membrane Proteins: Studies with MscL Biophys. J., May 15, 2007; 92(10): 3556 - 3563. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I. Iscla, G. Levin, R. Wray, and P. Blount Disulfide Trapping the Mechanosensitive Channel MscL into a Gating-Transition State Biophys. J., February 15, 2007; 92(4): 1224 - 1232. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
|
), hydrophobic (
), polar (
), and
basic (
).
< 2 ms versus normal or long: 
