Protein Localization in Escherichia coli Cells: Comparison of the Cytoplasmic Membrane Proteins ProP, LacY, ProW, AqpZ, MscS, and MscL

Abstract

Fluorescence microscopy has revealed that the phospholipid cardiolipin (CL) and FlAsH-labeled transporters ProP and LacY are concentrated at the poles of Escherichia coli cells. The proportion of CL among E. coli phospholipids can be varied in vivo as it is decreased by cls mutations and it increases with the osmolality of the growth medium. In this report we compare the localization of CL, ProP, and LacY with that of other cytoplasmic membrane proteins. The proportion of cells in which FlAsH-labeled membrane proteins were concentrated at the cell poles was determined as a function of protein expression level and CL content. Each tagged protein was expressed from a pBAD24-derived plasmid; tagged ProP was also expressed from the chromosome. The osmosensory transporter ProP and the mechanosensitive channel MscS concentrated at the poles at frequencies correlated with the cellular CL content. The lactose transporter LacY was found at the poles at a high and CL-independent frequency. ProW (a component of the osmoregulatory transporter ProU), AqpZ (an aquaporin), and MscL (a mechanosensitive channel) were concentrated at the poles in a minority of cells, and this polar localization was CL independent. The frequency of polar localization was independent of induction (at arabinose concentrations up to 1 mM) for proteins encoded by pBAD24-derived plasmids. Complementation studies showed that ProW, AqpZ, MscS, and MscL remained functional after introduction of the FlAsH tag (CCPGCC). These data suggest that CL-dependent polar localization in E. coli cells is not a general characteristic of transporters, channels, or osmoregulatory proteins. Polar localization can be frequent and CL independent (as observed for LacY), frequent and CL dependent (as observed for ProP and MscS), or infrequent (as observed for AqpZ, ProW, and MscL).

Modern developments in fluorescence microscopy have led to a new understanding of the organization of bacterial cells, particularly protein and lipid localization (21, 56). Analysis of the subcellular localization of diverse proteins and lipids has shown that they are not uniformly distributed. The phospholipid cardiolipin (CL) localizes at the poles and septal regions (36), and there is evidence for segregation of phosphatidylethanolamine (PE) from phosphatidylglycerol (PG) in the membranes of living Escherichia coli cells (69). Localization of many proteins that are integral or peripheral to the cytoplasmic membrane has been studied by fusing them to green fluorescent protein (GFP) (or its derivatives), and it is possible to classify the fusion proteins according to their subcellular localization. The first group, comprised of proteins that are concentrated at the cell poles, includes chemoreceptors (31, 62), the lactose permease LacY (43), and the metabolic sensor kinases DcuS and CitA (55). Members of the second group form helices that extend from pole to pole and include MreB (25), MinD (57), the Sec protein export system (58), and RNase E, which is the main component of the RNA degradosome in E. coli (67). Other proteins may appear to be similarly distributed due to their association with the Sec system (58). Members of the third group are uniformly distributed and include the mechanosensitive channel MscL (45) and the sensor kinase KdpD (32).

The polar localization of proteins appears to be a critical feature of the complicated internal localization of bacteria. For example, it is important for temporally and spatially accurate placement of the septum during cell division (15). However, the mechanism of protein organization at bacterial cell poles is still unclear, and in many cases its functional role has not been determined. Do the poles merely serve as a receptacle for proteins, superstructures, or membrane domains with no functional effects, or is this location functionally important for membrane proteins and lipids?

Recent evidence indicates that the subcellular localization of the transporter ProP in E. coli is related to membrane phospholipid composition, cardiolipin localization, and ProP function (51, 52). E. coli cells from cultures grown to exponential phase contain mostly the zwitterionic phospholipid PE (approximately 75 mol%) and the anionic phospholipids PG (approximately 20 mol%) and CL (approximately 5 mol%) (8). (Note that cardiolipin is diphosphatidylglycerol.) However, the phospholipid composition depends on the bacterial growth conditions. We found that the proportion of CL among E. coli lipids varies directly with growth medium osmolality (68), and increased CL synthesis was at least partially attributed to regulation of the cls locus encoding cardiolipin synthase (52). There is residual CL in cls bacteria, indicating that there is an alternative pathway for CL synthesis (51). The CL-specific fluorescent dye 10-N-nonyl-acridine orange (NAO) was used to show that CL clusters at the poles and septa in growing E. coli cells (36, 52). This result was corroborated by analyzing the phospholipid composition of E. coli minicells (DNA-free cells resulting from asymmetric cell division) (24, 51).

ProP is an osmosensory transporter that senses increasing osmolality and responds by mediating the cytoplasmic accumulation of organic osmolytes (e.g., proline, glycine betaine, and ectoine). Biochemical regulation of the ProP protein ensures that ProP activity increases with increasing assay medium osmolality (49). We showed that ProP and CL colocalize at the poles and near the septa of dividing E. coli cells and that the polar concentration of ProP correlates with the polar concentration of CL (52). Moreover, we showed that the osmolality required to activate ProP increased in parallel to the CL content when E. coli was cultivated in media with increasing osmolality (51, 52, 68). The osmolality required to activate ProP was also a direct function of CL content in proteoliposomes reconstituted with purified ProP (51). We concluded that concentration at the cell poles controlled the osmoregulatory function of ProP by placing the transporter in a cardiolipin-rich environment.

To determine whether CL-dependent membrane protein localization is a general phenomenon in E. coli, we compared the subcellular localization of ProP with that of its paralogue LacY, a well-characterized lactose transporter (16). LacY and ProP are both members of the major facilitator superfamily and H⁺ symporters. LacY transports the nutrient lactose, and LacY activity decreases while ProP activity increases with increasing osmolality (9). Nagamori et al. reported polar localization of a LacY-GFP fusion protein in E. coli (43). We confirmed this observation and demonstrated that, in contrast to the behavior of ProP, the polar concentration of LacY did not correlate with the polar concentration of CL (51).

In this work we further explored the relationship between CL and protein localization in E. coli. We compared ProP with other proteins related to cellular osmoregulation. Bacteria use arrays of osmoregulatory mechanisms to survive and function when the osmotic pressure of their environment changes. In E. coli, the aquaporin AqpZ mediates transmembrane water flux, the transporters ProP, ProU, BetT, and BetU mediate organic osmolyte accumulation at high osmotic pressure, and the mechanosensitive (MS) channels MscL and MscS mediate solute efflux in response to osmotic downshock (71). Localization of these proteins might be expected since AqpZ might influence cell morphology changes by accelerating water flux at particular positions on the cell surface and the pressure sensitivities of MscL and MscS are known to depend on membrane curvature in vitro (18).

For ProP and LacY, we labeled the inserted peptide tag CCPGCC with the biarsenical fluorescein reagent FlAsH-EDT₂ (fluorescein arsenical helix binder, bis-EDT adduct) (1, 2) to examine the subcellular localization of AqpZ, the integral membrane component ProW of the osmoregulatory ATP-binding cassette (ABC) transporter ProU, and the MS channel proteins MscS and MscL in cls⁺ and cls bacteria. Fluorescence microscopy was used to determine the proportion of cells with labeled protein concentrated at the poles as a function of bacterial CL content and protein expression level. For ProP, the frequency with which MscS was concentrated at cell poles was proportional to the level and polar concentration of CL. LacY concentrated at the cell poles at a high and CL-independent frequency. The frequencies with which AqpZ, MscL, and ProW concentrated at the cell poles and septa were low (up to 12%) and CL independent.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

Strains and plasmids used in this work are listed in Table 1. Strain NCM3306 (MG1655 ΔaqpZ::Cam [60]) was obtained from Sydney Kustu (Berkeley, CA); strains MM294-Strept^R (spontaneous streptomycin-resistant mutant of MM294 [E. coli Genetic Stock Center no. 6315; F⁻ glnV44 λ⁻ rfbC1? endA1 spoT1? thi-1 hsdR17 creC510]) and MM1211 (MM294-Strept^R aqpZ::lacZ-kan) (7) were obtained from Giuseppe Calamita (University of Bari, Bari, Italy).

TABLE 1.

E. coli strains and plasmids used in this study

Strain or plasmid	Description or genotype	Source or reference
Strains
MG1655	Wild type	E. coli Genetic Stock Center
MJF465	Frag1 ΔmscL::Cm ΔyggB ΔkefA::kan	27
MM294-StrepR	Streptomycin-resistant derivative of MM294 (F−glnV44 λ−rfbC1? endA1 spoT1? thi-1 hsdR17 creC510)	7
MM1211	MM294-StrepRaqpZ::lacZ-kan	7
NCM3306	MG1655 ΔaqpZ::cam	60
WG170	F−lacZ trp rpsL thi Δ(putPA)101 proP219	63
WG350	F−lacZ trp rpsL thi Δ(putPA)101 Δ(proU)600 Δ(proP-melAB)212	11
WG269	F−lacZ trp rpsL thi Δ(putPA)101 proP219 Δ(proU)1 zjd-351::Tn10	38
WG980	WG350 cls::Tn10dTet3	52
WG1127	WG350 ΔproW859::FRT	This study
WG1128	WG980 ΔproW859::FRT	This study
WG1224	WG350 ΔproP837::FRT	This study
WG1259	WG350 proP678::Lumio	This study
WG1260	WG1259 cls::Tn10dTet3	This study
Plasmids
pTR10	E. coli ORF mscS inserted into expression vector pBAD24	This study
pTR11	pTR10 derivative that encodes MscS-CCPGCC	This study
pTR12	E. coli ORF mscL inserted into expression vector pBAD24	This study
pTR13	pTR12 derivative that encodes MscL-CCPGCC	This study
pTR14	E. coli ORF aqpZ inserted into expression vector pBAD24	This study
pTR15	pTR14 derivative that encodes MVCCPGCC-AqpZ	This study
pTR17	pDC10 derivative that encodes MLCCPGCC-ProP	This study
pDC288	pKO3 derivative with proP678::Lumio inserted at the BamHI site	This study

Unless otherwise stated, the bacteria were cultivated in Luria-Bertani (LB) medium (37) or in NaCl-free morpholinepropanesulfonic acid (MOPS) medium (44) with glycerol (0.4%, vol/vol) as the carbon source, NH₄Cl (9.5 mM) as the nitrogen source, tryptophan (245 μM), and thiamine (1 mg/ml) to meet auxotrophic requirements. Ampicillin (100 μg/ml) was added to maintain plasmids, and arabinose was added as indicated below to adjust gene expression from the P_BAD promoter of vector pBAD24 (17). NaCl was added to adjust the osmolality, and osmolalities were measured with a Wescor vapor pressure osmometer (Wescor, Logan, UT). To monitor growth in liquid medium, cultures (24 ml) were prepared in 125-ml sidearm flasks and incubated at 37°C with rotary shaking at 200 rpm. Optical densities were monitored with a Bausch and Lomb Spectronic 88 spectrophotometer.

Plasmids and molecular biological manipulations.

The basic molecular biological techniques used were performed as described by Sambrook and Russell (54). PCR were carried out as described by Brown and Wood (6). Oligonucleotides were purchased from Cortec DNA Services (Kingston, ON, Canada). Construction of plasmid pDC232 encoding MVCCPGCC-ProP and plasmid pDC245 encoding MCCPGCC-LacY was described previously (51, 52). The genes encoding ProW, MscS, MscL, and AqpZ were each amplified using PCR primers that introduced restriction endonuclease cleavage sites appropriate for insertion downstream from the AraC-controlled P_BAD promoter in the vector pBAD24 (enzymes NcoI and HindIII) (17). Each amplicon was cloned into this vector, and the desired recombinant plasmids were recovered from the ligation mixtures by standard chemical transformation of E. coli DH5α (19). Site-directed mutagenesis was performed as described by Culham et al. (12) to introduce the oligonucleotide sequence encoding the desired FlAsH tag (Table 1). The entire sequence of the gene encoding each protein was confirmed (GenAlyTiC, Guelph, ON, Canada). Plasmids pTR3 and pTR4 (encoding ProW and its FlAsH-tagged derivative, respectively) were introduced into E. coli strains WG1127 (ΔproW cls⁺) and WG1128 (ΔproW cls), plasmids pTR10 to pTR15 (encoding MscL, MscS, and AqpZ and their FlAsH-tagged derivatives [Table 1]) were introduced into strains WG350 (mscL⁺ yggB⁺ aqpZ⁺ cls⁺) and WG980 (mscL⁺ yggB⁺ aqpZ⁺ cls), and plasmids pTR14 and pTR15 (encoding AqpZ and its FlAsH-tagged derivative, respectively) were also introduced into strain NCM3306 (aqpZ::Cm cls⁺).

Strains WG1127 and WG1128 were constructed by introducing the Keio collection (3) proW::Km insertion into WG350 and WG980, respectively, by P1 transduction and then deleting the kanamycin resistance cassette as described by Datsenko and Wanner (13). Strain WG1224 was created in an analogous manner by introducing the Keio collection (3) proW::Km insertion with WG350 as the target strain and then deleting it.

The gene encoding MLCCPGCC-ProP was inserted into the E. coli chromosome as follows. Plasmid pDC10 consists of the proP open reading frame (ORF) plus flanking DNA inserted into the vector pGEM4Z (11). An oligonucleotide encoding the FlAsH tag CCPGCC was inserted after the initiation codon of proP in pDC10. Site-directed mutagenesis using primer 5′-GGCCAACTCTGCGAGGAAAGCTATGCTGTGCTGCCCGGGCTGCTGCAAAAGGAAAAAAGTAAAACCGAT-3′ and its complement, with pDC10 as the template, created plasmid pTR17, which includes proP678::Lumio. The proP678::Lumio ORF plus 500 bp 5′ of the initiation codon and 500 bp 3′ of the termination codon was amplified by PCR performed with primers 5′-CGCGGGATCCCACCGACCACGCGTCAC-3′ and 5′-CGCGGGATCCCAACCCTGCTGCGGATG-3′, which were designed to add a BamHI site at each end of the amplicon. This amplicon was inserted at the BamHI site of the allelic exchange vector pKO3 (29), creating plasmid pDC288. The proP678::Lumio ORF was then introduced into the chromosome of E. coli strain WG1224 by allelic exchange as described by Link et al. (29). Chloramphenicol- and kanamycin-sensitive isolates that survived the Sac selection were screened on MOPS medium supplemented with 0.6 M NaCl and 1 mM proline to identify the isolates that were ProP⁺ (11). Chromosomal insertion of proP678::Lumio into one such isolate, WG1259, was confirmed by PCR amplification and cleavage of the resulting amplicon to verify introduction of an NciI site during creation of proP678::Lumio. The allele cls::Tn10dTet3 was introduced into WG1259 by P1 transduction.

Labeling, fluorescence microscopy, SDS-PAGE, and Western blot analysis.

Bacteria containing the tagged protein MVCCPGCC-ProP, MLCCPGCC-ProP, MCCPGCC-LacY, MACCPGCC-ProW, MscS-CCPGCC, MscL-CCPGCC, or MVCCPGCC-AqpZ or no tagged protein were cultivated as described above and labeled with the fluorescent probe FlAsH-EDT₂ in the presence of 2,3-dimercapto-1-propanol (BAL) or with 10-N-nonyl-3,6-bis(dimethylamino acridine) (NAO) (Invitrogen, Carlsbad, CA) as described previously (52). Unless otherwise stated, bacteria were cultivated without arabinose. FlAsH-EDT₂ and BAL were purchased from Invitrogen (Carlsbad, CA), and FlAsH-EDT₂ was synthesized as described by Adams and Tsien (1). The identity and purity of the reagent were confirmed by nuclear magnetic resonance spectroscopy (see Fig. S1 in the supplemental material). Cells were viewed with an Imaging RetigaEX charge-coupled device (CCD) camera mounted on an Axiovert 200M inverted fluorescence microscope (Carl Zeiss Microimaging Inc.) equipped with a Zeiss Plan Neofluor 100× oil NA1.3 objective as described previously to detect green FlAsH fluorescence or CL-specific, red NAO fluorescence (52) and with phase-contrast light microscopy.

Labeled cells were analyzed by SDS-PAGE to assess the labeling specificity and integrity of the FlAsH-tagged proteins. Bacteria were cultivated as described previously for transport assays (12) in low-osmolality MOPS medium supplemented with arabinose (1 mM), and SDS-PAGE was performed as described by Laemmli (26) using Bio-Rad MiniProtean II cells (Bio-Rad Laboratories Ltd., Hercules, CA) (12% T, 2.67% C). The SDS loading buffer did not contain β-mercaptoethanol. FlAsH fluorescence was detected with a Typhoon 9410 variable-mode imager (GE Healthcare Life Sciences, Piscataway, NJ) equipped with a 526-nm short-pass filter and a CCD camera attachment.

For Western blotting, bacteria were cultivated as described previously for transport assays (12) in low-osmolality MOPS medium without arabinose. Cell extracts were separated by SDS-PAGE, and Western blotting was performed using anti-ProP antibodies as described previously (49).

Transport assays.

Bacteria were cultivated and transport assays were performed with them as described by Culham et al. (10). Initial uptake rates were measured using [1-¹⁴C]glycine betaine at a final concentration of 10 μM or l-[U-¹⁴C]proline at a final concentration of 200 μM. All assays were conducted in triplicate, and all experiments were performed at least twice. Representative rates obtained in single experiments are expressed below as means ± standard deviations. Nonlinear regression was used to fit initial rates of proline uptake (a₀) at various osmolalities (Π/RT) to the following equation:

(1)

where A_max is the rate that would be attained at infinite osmolality, Π_1/2/RT is the osmolality at which a₀ equals A_max/2, and B is related to the slope of the activation curve (12).

Stopped-flow spectroscopy.

Samples for stopped-flow experiments were prepared as described by Mallo and Ashby (33), with the following modifications. Using frozen stocks, E. coli strains were inoculated into 5 ml LB medium adjusted to pH 7.5 with NaOH and incubated overnight at 37°C with shaking at 200 rpm. They were then subcultured into 24 ml of the same medium in 125-ml flasks to obtain an optical density at 600 nm (OD₆₀₀) of 0.1 and incubated with shaking until the OD₆₀₀ was 0.8 (exponential phase, 2 h) or about 6.5 (stationary phase, 10 h) (optical densities were determined using a Bausch and Lomb Spectronic 88 spectrophotometer). The cells were harvested by centrifugation for 10 min at 10,000 rpm at 4°C. Each pellet was resuspended with cold phosphate-buffered saline (PBS) (14 mM NaCl, 2.7 mM KCl, 4 mM Na₂HPO₄, 1.8 mM KH₂PO₄; pH 7.4) to obtain an OD₆₀₀ of 0.8.

The absorbance of each cell suspension was measured with an SX20 stopped-flow spectrometer (Applied Photophysics Limited, Leatherhead, England) after 1 volume of cells in PBS was mixed with 1 volume of PBS-0.5 M NaCl, PBS-1 M proline, or PBS-1 M glycerol (for upshock) in a 100-μl optical cell at 4°C. The distance between the optical cell and the side window nine-stage photomultiplier was 3 cm, and the path length of the optical cell was 2 mm. The absorbance of the suspension was recorded over a 5-s period after mixing, and 10 replicate measurements were obtained. The optical density of the suspension after mixing was about 0.4, as measured with the Bausch and Lomb Spectronic 88 spectrophotometer, for cells from both exponential- and stationary-phase cultures. The average of the last five replicate measurements was analyzed. The scale of the absorbance (A) values was adjusted as follows to make the units of the values obtained by regression more convenient: (A·1,000) + 10. To determine rate constants (k), the absorbance was plotted versus time (t) for the period from 0.1 s to 5.0 s after mixing, and the data were fitted to the foll owing equation by nonlinear regression:

(2)

where A₀ and A_max are the absorbance at zero time and the absorbance at infinite time, respectively.

Lysis of E. coli upon osmotic downshock.

Cultures were grown, cells were treated, and lysis of the cells was measured exactly as described by Levina et al. (27).

Preparation of giant spheroplasts.

Spheroplasts were prepared from E. coli strain MJF465 lacking MscL, MscS, and MscK essentially as described previously (34, 53). Plasmids pTR10 to pTR13 encoding MscL and MscS with and without the FlAsH tag were introduced into this strain. A culture was grown in LB medium until the OD₆₀₀ was 0.4 to 0.5 and then diluted 1:10 into Luria-Bertani medium, and cephalexin was added to a concentration of 60 μg/ml. The culture was then incubated at 42°C for 2 to 2.5 h until single-cell filaments reached a length (50 to 150 μm) sufficient for formation of giant spheroplasts (diameter, 5 to 10 μm). After this, arabinose (final concentration, 0.0015%) and glycerol (0.4%) were added, and the cells were incubated at 25°C and 180 rpm for about 1 h. Then filaments were harvested by centrifugation, and the pellet was rinsed without resuspension by gentle addition of 1 ml of 0.8 M sucrose (0.4% glycerol). After a second centrifugation, the supernatant was removed, and the pellet was resuspended in 2.5 ml of 0.8 M sucrose (0.4% glycerol). Then reagents were added in the following order: 150 μl of 1 M Tris (pH 7.2), 120 μl of lysozyme (5 mg/ml), 50 μl of DNase I (5 mg/ml), and 150 μl of 0.125 M EDTA. The mixture was incubated at room temperature for 3 to 5 min to hydrolyze the peptidoglycan layer, and the progress of spheroplast formation was followed under a microscope. At the end of this incubation, 1 ml of a solution containing 20 mM MgCl₂ (to remove the EDTA and activate the DNase), 0.7 M sucrose, 0.4% glycerol, and 10 mM Tris (pH 7.2) was added. The mixture was then diluted with 5 ml of a solution containing 10 mM MgCl₂, 0.8 M sucrose, 0.4% glycerol, and 10 mM Tris (pH 7.2). Aliquots were prepared and stored in a freezer at −20°C.

Electrophysiological recordings.

Single-channel analysis was performed with giant spheroplasts (64). Borosilicate glass pipettes (Drummond Scientific Co., Broomall, PA) were pulled using a Flaming/Brown pipette puller (P-87; Sutter Instrument Co., Novato, CA) to obtain a diameter which corresponded to a pipette resistance between 3.0 and 6.0 MΩ. Ion currents arising from activation of the proteins using negative pipette pressure (suction) were recorded using an Axon 1D patch-clamp amplifier (Axon Instruments), filtered at 2 kHz, and digitized at 5 kHz. Single-channel analysis was done using pCLAMP10 software (Axon Instruments). The bath and pipette solutions were asymmetrical and consisted of 250 mM KCl, 90 mM MgCl₂, and 5 mM HEPES (pH 7.2) and 200 mM KCl, 40 mM MgCl₂, and 5 mM HEPES (pH 7.2), respectively.

Protein assays.

The protein concentrations of cell suspensions were determined by the bicinchoninic acid assay (59), using a BCA kit from Pierce (Rockford, IL) with bovine serum albumin as the standard.

RESULTS

Analysis of lipid and protein localization: CL and ProP as prototypes.

We previously reported that the level and polar concentration of CL depend on the bacterial genotype (cls bacteria retain only low levels of CL) and growth medium osmolality (the cellular CL content increases with increasing growth medium osmolality) (51, 52, 68). The subcellular localization of bacterial components like CL can be illustrated by reproducing representative fluorescence micrographs (Fig. 1), but selected images do not reveal how frequently a feature is observed in a bacterial population. Such images can be supplemented by reporting the proportion of cells in which membrane components are concentrated at cell poles and septa. For example, we observed concentration of CL-specific, red NAO fluorescence at the poles in very different proportions of cls⁺ and cls E. coli cells (Fig. 2G). CL-specific red fluorescence was evident at the poles in 73% of cls⁺ bacteria and in less than 3% of cls bacteria from liquid cultures with a variety of osmolalities. Within each population (cls⁺ or cls), the frequency of polar localization did not vary significantly with growth osmolality (Fig. 2G) or CL content (Fig. 2H). Systematic variation might have occurred for CL contents in the range from 0.6 to 4 mol% that is not accessible to us in vivo. As reported previously, the intensities of the polar CL “patches” varied (51, 52). Concentration of CL at the cell poles in some bacteria (particularly cls bacteria) may have eluded detection by fluorescence microscopy, or CL may concentrate at the cell poles less frequently when it is present as a smaller mole fraction of the membrane phospholipid (22, 41).

FIG. 1.

Localization of ProW, MscS, MscL, AqpZ, LacY, ProP, and CL in cls⁺ and cls E. coli strains. Bacteria were grown in high-osmolality MOPS media without arabinose to exponential phase and prepared for microscopy by staining with FlAsH-EDT₂ or NAO as described previously by Romantsov et al. (52). Tagged ProP was expressed from a pBAD24-derived plasmid (MVCCPGCC-ProP in WG350/pDC232 [cls⁺] and WG980/pDC232 [cls]) (ProP^P) or from the chromosome (MLCCPGCC-ProP in WG1259 [cls⁺] and WG1260 [cls]) (ProP^C). The other proteins were expressed from pBAD24-derived plasmids, as follows: MACCPGCC-ProW in WG350/pTR4 (cls⁺) and WG980/pTR4 (cls), MscS-CCPGCC in WG350/pTR11 (cls⁺) and WG980/pTR11 (cls), MscL-CCPGCC in WG350/pTR13 (cls⁺) and WG980/pTR13 (cls), MVCCPGCC-AqpZ in WG350/pTR15 (cls⁺) and WG980/pTR15 (cls), and MCCPGCC-LacY in WG350/pDC245 (cls⁺) and WG980/pDC245 (cls). CL localization was examined in WG350/pDC232 (cls⁺) and WG980/pDC232 (cls). Selected images illustrate proteins or CL concentrated at the cell poles and randomly distributed in the cytoplasmic membrane. The frequencies with which the proteins were concentrated at the poles are shown in Fig. 2.

FIG. 2.

Frequencies with which CL, ProP, LacY, ProW, MscS, MscL, and AqpZ were concentrated at cell poles. Bacteria expressing the proteins were cultivated and imaged as described in the legend to Fig. 1, employing MOPS media adjusted with NaCl to the indicated range of osmolalities and lacking arabinose. CL was detected by staining with NAO as described in Materials and Methods. The frequency with which CL or each protein was concentrated at a cell pole (one or both poles) was determined by visual inspection for three groups of 100 randomly selected cells. Adjustment of the depth of focus was required to fully examine each cell. Each protein was found to concentrate at the septa of some dividing cells. This was always accompanied by concentration of the protein at the other pole of the same cell and was therefore counted as polar concentration. The data are expressed means ± standard deviations for the three determinations. The frequency data are plotted versus the osmolality of the bacterial growth medium (left panels) and the proportions of CL among the phospholipids of bacteria cultivated under the conditions described above (right panels). The CL contents were determined previously and reported by Tsatskis et al. (68) and Romantsov et al. (51). ProP^P indicates plasmid-encoded ProP, and ProP^C indicates chromosome-encoded ProP. All other proteins were plasmid encoded (P).

We previously exploited the well-controlled, low-level protein expression provided by the pBAD24 vector (17) and the propensity of FlAsH-EDT₂ to specifically label small hexapeptide tags in vivo (2) to show that the transporters ProP and LacY concentrate at the poles of E. coli cells (51, 52). Here we extended these results by comparing the behavior of plasmid- and chromosome-encoded ProP and by further quantifying the relationship between CL content and the proportion of cells with ProP or LacY concentrated at the poles.

We previously reported that insertion of the oligonucleotide encoding the FlAsH tag at the 5′ end of proP impaired the expression of this gene from the pBAD24 vector (as indicated by Western blotting) and that the activities of FlAsH-tagged ProP and untagged ProP are similar (52). Here we show that the cellular levels of FlAsH-tagged ProP expressed from a pBAD24-derived plasmid (P_BAD promoter, no arabinose induction) or from the chromosome (native proP promoters, no osmotic induction) are lower than the level of native ProP expressed from the chromosome (native proP promoters, no osmotic induction) (Fig. 3A). This conclusion is based on measurements of both protein levels (Western blotting) and transport activity. Tagged ProP was expressed from the plasmid without arabinose induction or from the chromosome, and then the bacteria were labeled with FlAsH-EDT₂ and examined by fluorescence microscopy. The proportions of cells in which FlAsH fluorescence was concentrated at the poles were slightly lower for the chromosome-encoded protein than for the plasmid-encoded protein (Fig. 2A and B). For both expression systems, the proportion of cells in which FlAsH fluorescence was concentrated at the poles varied with the growth medium osmolality (Fig. 2A) and hence with the mole fraction of CL in the bacterial membranes (Fig. 2B). Further, systematic variation might have been observed if CL contents in the range from 0.6 to 4 mol% were experimentally accessible.

FIG. 3.

Expression level of ProP and integrity of tagged proteins. (A) Activities of tagged (circles) or untagged (triangles) ProP variants expressed from the E. coli chromosome under control of the wild-type proP promoters (filled triangles and open circles) or from a pBAD24-derived plasmid under control of the P_BAD promoter (filled circles). E. coli (cls⁺) was cultivated in low-osmolality MOPS medium (0.15 mol/kg) without arabinose, proline uptake assays were conducted, data were analyzed to obtain regression lines, and Western blotting was performed as described in Materials and Methods. (Inset) Western blot showing the level of ProP in each preparation detected with anti-ProP antibodies. Lane 1, plasmid-encoded MVCCPGCC-ProP; lane 2, chromosome-encoded MLCCPGCC-ProP; lane 3, chromosome-encoded native ProP. (B) Fluorescence image of an SDS-PAGE gel displaying the arrays of proteins present after FlAsH labeling of bacteria that expressed each protein considered in this study. Bacteria were cultivated in low-osmolality medium with 1 mM arabinose, electrophoresis was performed, and the image was obtained as described in Materials and Methods.

The proportion of cells in which FlAsH fluorescence was concentrated at the poles did not vary when expression of plasmid-encoded, tagged ProP was induced with up to 100 μM arabinose (see Table S1 in the supplemental material). The cellular level of tagged ProP varies from well below to approximately 2-fold above that of the native, chromosome-encoded transporter with arabinose induction in this range (Fig. 3A) (52). SDS-PAGE analysis and fluorescence imaging of the proteins in FlAsH-labeled bacteria revealed good labeling specificity and no significant degradation of FlAsH-labeled ProP (Fig. 3B). We therefore concluded that neither the polar localization of ProP nor its correlation with CL localization is an artifact of ProP expression from a heterologous promoter, of its under- or overexpression, or of its degradation. For analogous studies performed with this and other tags, we cannot rule out effects of the FlAsH tag or FlAsH labeling on ProP localization.

Analysis of protein localization: other membrane proteins.

We extended the approach described above to compare the subcellular location of ProP with that of the transporter LacY, the aquaporin AqpZ, and three osmoregulatory proteins (ProW, MscL, and MscS). Each protein was expressed from a pBAD24-derived plasmid, and the optimal FlAsH tag was introduced at the N or C terminus of a protein subunit. Each tagged protein complemented the corresponding genetic defect (see reference 51 for LacY and data below for the other proteins). For each protein, FlAsH-EDT₂ labeling of cells cultivated without arabinose and fluorescence microscopy revealed some cells in which fluorescence was concentrated at the poles and/or septa (Fig. 1). As discussed further below, both the proportion of cells with fluorescence concentrated at the poles and the correlation of this behavior with CL content were protein specific. For each protein, the frequency of polar localization was independent of the protein expression level after induction with up to 1 mM arabinose (see Table S1 in the supplemental material). The ranges of protein expression obtained here with the pBAD system may be higher or lower than those obtained when each native protein is expressed from the chromosome via its native promoter.

To assess the integrity of the tagged and FlAsH-labeled proteins, we imaged SDS-PAGE gels displaying the arrays of proteins present in each preparation after FlAsH labeling (Fig. 3B). Induction (1 mM arabinose) was used to obtain protein levels detectable with this technique. The background labeling was weak (Fig. 3B, control lane). Each protein migrated with an electrophoretic mobility close to that reported previously (35, 42, 49, 50, 65, 66). More slowly migrating species may represent dimers of the target proteins, and more rapidly migrating species, perhaps representing degraded protein, were observed at extremely low levels.

Data reported below indicate the subcellular localization of proteins in which FlAsH-EDT₂ labeled the introduced 6-amino-acid tag CCPGCC, not the localization of the native protein. This shortcoming is shared with many other studies of protein localization.

Transporters LacY and ProW.

We previously reported that FlAsH fluorescence was concentrated at the poles in approximately one-half the cells expressing MCCPGCC-LacY from a pBAD24-derived plasmid without arabinose induction (low- or high-osmolality cultures of cls⁺ or cls E. coli) (51). Here we quantified this result with bacteria cultivated at a range of osmolalities (Fig. 1 and 2E and F). The data further indicate that concentration of LacY at the cell poles is not correlated with cellular CL content.

ProU is an osmosensory transporter and a member of the ATP-binding cassette (ABC) superfamily with substrate specificity similar to that of ProP (30). ProU consists of the membrane-associated ATPase ProV, the integral membrane protein ProW, and the periplasmic ligand-binding protein ProX. ProW is encoded by the second gene in the proU operon, and proW overlaps proV by seven codons. For this study, chromosomal proW was replaced by a 15-codon open reading frame without eliminating the proV stop codon (see Materials and Methods). CCPGCC was introduced at the N terminus of the ProW protein for FlAsH labeling. Staining with FlAsH-EDT₂ and fluorescence microscopy revealed that polar localization of ProW was infrequent (Fig. 2A) and that it did not vary with the CL content (Fig. 2B). The mean frequencies of polar localization were 11% ± 2% for cls⁺ bacteria and 13% ± 2% for cls bacteria.

Glycine betaine uptake activity was used as a measure of the integrity of the ProU system incorporating FlAsH-tagged ProW. Measurements were obtained using ProP-deficient host bacteria to avoid interference by this system. The proW deletion was complemented by plasmid-borne proW, indicating that proV and proX were expressed from the chromosome and that the transporter was assembled in the membrane despite expression of proW in trans from proVX (Table 2). Similar glycine betaine uptake activities were conferred by plasmids encoding ProW (pTR3) and MACCPGCC-ProW (pTR4). This similarity suggested that the FlAsH tag had not disrupted ProW function. However, even after arabinose induction, each activity was three times lower than the activity of ProW encoded by the intact, chromosomal proU operon (proVWX) (Table 2). The difference may have arisen from assembly of plasmid-encoded ProW with chromosome-encoded ProV and ProX.

TABLE 2.

Activities of ProU systems incorporating wild-type and FlAsH-tagged ProW

Relevant genotypea	ProW variant	Glycine betaine uptake rate (nmol min−1 mg protein−1)b
proU+ (proVWX+)	ProW	28.2 ± 0.5
ΔproW	None	0.08 ± 0.08
ΔproW pBAD24	None	0.05 ± 0.09c
ΔproW pTR3	ProW	10.8 ± 0.1c
ΔproW pTR4	MACCPGCC-ProW	9.1 ± 0.8c

^aThe host bacteria either were proU⁺ (WG170) or harbored an in-frame proW deletion (WG350 ΔproW859::FRT).

^bBacteria were cultivated and transport assays were performed as described in Materials and Methods. Cells were grown under osmotic stress conditions (0.3 M NaCl) to induce proU transcription.

^cCells were induced with 1 mM arabinose for one generation (∼2.5 h).

Mechanosensitive channels.

MscL and MscS (also known as YggB) are mechanosensitive channels with large and small conductance, respectively, that open to release cytoplasmic solutes after an osmotic downshock. Each of these proteins is an oligomer consisting of identical subunits (five subunits for MscL and seven subunits for MscS). We introduced the FlAsH tag at the C terminus of each channel protein because the pressure sensitivities of the proteins were not affected when other C-terminal tags were attached and overall the channels behaved like the wild-type channels (4, 34, 64). Staining with FlAsH-EDT₂ and fluorescence microscopy revealed that polar localization of MscS was evident in a larger proportion of cells than polar localization of MscL (Fig. 2C) and that polar localization of MscS varied with the cellular CL content, whereas the polar localization of MscL did not vary with the cellular CL content (Fig. 2D). The frequency with which MscS was concentrated at the cell poles ranged from 15% ± 3% in cls bacteria cultivated at low osmolality to 29% ± 3% in cls⁺ bacteria cultivated at high osmolality, whereas the mean frequencies with which MscL was concentrated at the cell poles were 3% ± 2% in cls bacteria and 6% ± 2% in cls⁺ bacteria. Thus, the MscS channel protein behaved like the transporter ProP, whereas the MscL channel protein behaved like the transport protein ProW in terms of subcellular localization.

The osmotic lysis method of Levina et al. (27) was used to determine whether introduction of the tag affected channel function in vivo. Strain MJF465 is a mutant that lacks MscS, MscL, and MscK. Plasmids encoding wild-type and FlAsH-tagged MscS or MscL were introduced into this strain to test the abilities of these channels to protect cells during sudden downshock. Cell lysis, indicated by release of material with absorbance at 260 nm, occurred after downshock of strain MJF465 with or without the pBAD24 vector (Fig. 4A) but not after downshock of this strain expressing the wild-type or FlAsH-tagged MscS or MscL protein (Fig. 4B). Thus, we reproduced the observation of Levina et al. that low-level, plasmid-based expression of either MscS or MscL is sufficient to protect MJF465 from osmotic lysis (27, 34, 64) and showed that this was also true for the FlAsH-tagged variants of these proteins.

FIG. 4.

Analysis of MscS and MscL function using osmotic lysis of E. coli. Bacteria were grown in the presence of 0.5 M NaCl, filtered, and resuspended in the presence (filled symbols) or absence (open symbols) of 0.5 M NaCl as described by Levina et al. (27). At intervals, samples were centrifuged, and the OD₂₆₀ of the supernatant was recorded. (A) Circles, MJF465 (ΔmscL::Cm ΔyggB ΔkefA::kan); triangles, MJF465/pBAD24. (B) Circles, MJF465/pTR10 (MscS⁺); triangles, MJF465/pTR11 (MscS-CCPGCC); squares, MJF465/pTR12 (MscL⁺); diamonds, MJF465/pTR13 (MscL-CCPGCC).

Patch-clamp analysis was also used to check channel activity. Giant spheroplasts were made by using standard methods (see Materials and Methods) and were analyzed by cell-attached recording. The slope conductances were not significantly affected by tagging the channels (Table 3) and were close to those recorded previously (34, 64). The channel opening threshold was not affected by the tag; however, subtle differences were observed when negative voltage was applied to the tagged and wild-type MscL-containing spheroplasts. The tagged MscL protein showed an unusual propensity to open at subconducting levels (Fig. 5). This prevented us from obtaining a slope conductance for this tagged protein, especially at applied voltages less than −40 mV. In contrast, no significant differences were seen at negative voltages in the channel gating behavior of tagged and wild-type MscS protein. Indeed, the FlAsH-MscS protein exhibited normal rectification (Fig. 6A) between the positive and negative voltages (Table 3), as well as desensitization or inactivation (Fig. 6B), which is typical of the behavior of wild-type MscS in giant spheroplasts.

TABLE 3.

Functional analyses of mechanosensitive channels incorporating the FlAsH tag

Channel	Slope conductance (nS)a
	Positive voltages	Negative voltages
MscS	1.30 ± 0.04	0.79 ± 0.05
MscS-CCPGCC	1.29 ± 0.04	0.85 ± 0.03
MscL	3.94 ± 0.05	NDb
MscL-CCPGCC	4.04 ± 0.03	ND

^aData for recordings were obtained from inside-out excised patches of giant spheroplasts of E. coli strain MJF465 (mscL mscS mscK) grown in LB media and expressing MscL or MscS channel proteins with and without the FlAsH tag (CCPGCC).

^bND, not determined (see the text).

FIG. 5.

Patch-clamp recordings of FlAsH-labeled MscL. The recordings are of inside-out excised patches from giant spheroplasts of E. coli strain MJF465 (mscL mscS mscK) expressing FlAsH-labeled MscL. (A) Recording at 30 mV; (B) recording at −30 mV. Flickery channel behavior is seen at positive voltages (expansion), and subconducting states are apparent at negative voltages. C and O1 to O9 indicate closed and open channel levels, respectively.

FIG. 6.

Patch-clamp recordings of FlAsH-labeled MscS. The recordings are of inside-out excised patches from giant spheroplasts of E. coli strain MJF465 (mscL mscS mscK) expressing FlAsH-labeled MscS at (A) −30 and 30 mV and (B) channel open behavior at 30 mV. In panel B, analogous to wild-type MscS behavior, the channels rapidly inactivate or desensitize. C and O1 to O10 indicate closed and open channel levels, respectively.

Aquaporin AqpZ.

We introduced the FlAsH tag at the N terminus of AqpZ (a homotetramer) because an AqpZ derivative with a 23-residue N-terminal extension containing 10 consecutive His residues was functional in proteoliposome studies (5). Staining with FlAsH-EDT₂ and fluorescence microscopy revealed polar localization at a low frequency (Fig. 2E) that did not vary with the cellular CL content (Fig. 2F). The frequencies with which AqpZ was concentrated at the cell poles were 12% ± 2% in cls bacteria and 10% ± 1% in cls⁺ bacteria. Thus, the aquaporin AqpZ behaved like the transport protein ProW and the channel protein MscL in terms of subcellular localization.

Multiple methods were used to assess the function of AqpZ and its FlAsH-tagged derivative. Calamita et al. reported differences in colony phenotype and growth in batch culture between E. coli strains MM294-Strept^R (aqpZ⁺) and MM1211(aqpZ) (7), but Soupene et al. found no such differences between a wild-type E. coli strain (MG1655) and strain NCM3306 (MG1655 ΔaqpZ::cam) (60). Examining all four strains, we observed no difference in growth characteristic of the aqpZ defect (data not shown). Spectroscopy (33) and electron microscopy (14) have revealed differences between wild-type and AqpZ-deficient bacteria. Mallo and Ashby used stopped-flow light-scattering spectroscopy to show that the response of E. coli MG1655 to an osmotic upshift imposed with proline was much faster than that of E. coli NCM3306, particularly for bacteria from stationary-phase cultures (33).

Stopped-flow spectroscopy was used to compare the functions of wild-type and FlAsH-tagged AqpZ. Strains MG1655 and NCM3306 and their plasmid-bearing derivatives were used for these experiments as phase-contrast microscopy showed that strains MM294-Strept^R and MM1211 are filamentous (data not shown). The absorbance increased when osmotic upshocks were imposed on cells from exponential-phase (2 h) and stationary-phase (10 h) cultures of the aqpZ⁺ and aqpZ strains. For each preparation, the rates of increase in absorbance, which were attributed to increased light scattering due to water efflux, were similar for upshocks imposed with NaCl, glycerol, and proline. As expected, the absorbance decreased over time when upshocks were imposed with glycerol but not when upshocks were imposed with NaCl or proline (data not shown). The changes in scattering were fitted to a single exponential function to obtain the rate constants (k_w) shown in Table 4. The k_w for aqpZ⁺ bacteria (MG1655) was much larger than that for aqpZ bacteria (NCM3306) from stationary-phase cultures, and the differences in k_w between aqpZ⁺ and aqpZ bacteria from exponential-phase cultures were much smaller. Our k_w values are lower than those reported by Mallo and Ashby, but we observed qualitatively similar effects of culture growth phase and the aqpZ defect. The osmotic shock response of AqpZ-deficient bacteria became slower in stationary phase as the response of the wild-type bacteria accelerated.

TABLE 4.

Osmotic upshock responses of bacteria expressing AqpZ and its FlAsH-tagged variant

Strain	AqpZ variant	kw (s−1)a
		Upshock imposed with NaCl on culture from:		Upshock imposed with proline on culture from:
		Exponential phase	Stationary phase	Exponential phase	Stationary phase
MG1655	AqpZ	6.82 ± 0.05	19.24 ± 0.37	5.56 ± 0.04	27.38 ± 0.63
NCM3306	None	6.02 ± 0.04	0.480 ± 0.004	5.19 ± 0.03	0.350 ± 0.007
NCM3306/pBAD24	None	6.00 ± 0.07	0.450 ± 0.004	5.09 ± 0.07	0.304 ± 0.008
NCM3306/pTR14	AqpZ	7.31 ± 0.10	27.64 ± 0.40	7.18 ± 0.06	17.89 ± 0.17
NCM3306/pTR15	MVCCPGCC-AqpZ	6.94 ± 0.09	0.984 ± 0.005	5.37 ± 0.06	0.704 ± 0.005
NCM3306/pTR14b	AqpZ	7.08 ± 0.06	29.19 ± 0.36	7.76 ± 0.04	35.08 ± 0.31
NCM3306/pTR15b	MVCCPGCC-AqpZ	7.07 ± 0.06	29.38 ± 0.51	7.95 ± 0.03	29.38 ± 0.51

^aRate constants (k_w) for the optical density change following osmotic upshocks (0.5 mol/kg) imposed on bacteria at 5°C were determined by stopped-flow light-scattering spectroscopy as described in Materials and Methods.

^bExpression of aqpZ was induced with 1 mM arabinose for the last 2 h of growth.

The abilities of AqpZ and its FlAsH-tagged derivative to complement the aqpZ deficiency during expression from plasmid pBAD24 were compared. As before, k_w did not vary widely for bacteria from exponential-phase cultures (Table 4). In contrast, plasmid-based expression of either AqpZ or MVCCPGCC-AqpZ clearly complemented the AqpZ deficiency in cells from stationary-phase cultures (Table 4 and Fig. 7). This suggests that the particularly large effects of AqpZ on the osmotic stress response in stationary-phase bacteria result from factors other than stationary-phase (RpoS-dependent) induction of aqpZ expression. Comparable responses were observed for bacteria expressing AqpZ without arabinose induction and for bacteria expressing MVCCPGCC-AqpZ after induction with 1 mM arabinose. Insertion of the oligonucleotide encoding the N-terminal FlAsH tag may have impaired expression of the tagged variant (as seen for ProP and LacY) and/or the FlAsH tag may have impaired channel function.

FIG. 7.

Analysis of AqpZ function by stopped-flow spectroscopy. E. coli strains NCM3306/pBAD24 (AqpZ⁻), NCM3306/pTR14 (AqpZ⁺), and NCM3306/pTR15 (MVCCPGCC-AqpZ) were grown in LB medium to an OD₆₀₀ of 6.5 (stationary phase) and then upshocked with 0.5 M NaCl (A) or 0.5 M proline (B) in PBS as described in Materials and Methods. Strain NCM3306/pTR15 was induced with 1 mM arabinose. The time course of the change in absorbance after osmotic upshock was measured at 600 nm. The curves represent averages of five replicates, and data from 0.1 to 5 s were fitted to a single exponential equation to obtain k_w (see Materials and Methods and Table 4). a.u., absorbance units.

DISCUSSION

We have examined the localization of the proteins ProP, LacY, ProW, MscS, MscL, and AqpZ in E. coli by using reagent FlAsH-EDT₂ to specifically label an introduced hexapeptide tag (51, 52; this study). This study addressed a number of technical issues important for lipid and protein localization studies and provided evidence that ProP and MscS are unusual in showing CL-dependent localization at the cell poles.

Care is required to maximize the probability that protein localization determined with an introduced tag reflects localization of the native protein expressed at levels in the physiological range. We employed FlAsH labeling in this study because we anticipated that the small “CCPGCC” tag would perturb protein structure, function, and localization less than an added domain comprised of an entire fluorescent protein, like GFP. Our CCPGCC-tagged proteins were functional (Tables 2 to 4; Fig. 4 to 7). As observed in previous work employing GFP, we found that LacY was concentrated at the cell poles in a large proportion of bacteria, whereas MscL was not (Fig. 1 and 2; see Table S1 in the supplemental material). This lends credence to both visualization techniques, but it is still possible that the introduced tag and FlAsH labeling influenced the protein localization observed during this study.

Artifacts due to anomalous protein expression levels are a particular concern for protein localization studies. Overexpressed proteins can accumulate at the cell poles as inclusion bodies (28). The appearance of the regions of membrane protein fluorescence concentrated at the cell poles that we report here is clearly different from the appearance of inclusion bodies that can be visualized by phase-contrast microscopy. However, it is difficult to rule out artifacts due to the polar accumulation of protein aggregates visible by fluorescence microscopy but not by phase-contrast microscopy. Conversely, an apparently uniform distribution of a tagged protein on the cell surface could mask more specific localization of a native protein that is normally present at lower levels.

Our studies included two approaches to deal with this problem. Since anti-ProP antibodies are available, we compared the localization of tagged ProP when it was expressed from the native proP promoters in the chromosome to its localization when it was expressed at lower, comparable, or higher levels from the arabinose-inducible P_BAD promoter of a pBAD24-derived plasmid (Fig. 2A and B; see Table S1 in the supplemental material). (Note that the highest levels of ProP expression employed during this study were only approximately 2-fold higher than the levels of the native protein present after osmotic induction.) Polar concentration of ProP was seen at comparable frequencies, and the frequencies were similarly correlated with bacterial CL content at each expression level. We concluded that neither concentration of ProP at the cell poles nor its correlation with cellular CL content is an artifact of the ProP expression level.

Antibodies were not available for the other proteins included in our study. We therefore determined the conditions under which each tagged protein, expressed from a pBAD24-derived plasmid, restored function to a physiological level. We then determined protein localization at the minimum level attainable with plasmid pBAD24 (i.e., without arabinose induction) (Fig. 1 and 2) and, if it was higher, the expression level required to restore physiological function (see Table S1 in the supplemental material). For each protein we found the same frequency of polar localization and the same correlation (or lack of correlation) with CL levels at various protein expression levels. However, we cannot rule out the possibility that the native expression levels of these proteins are outside the range employed in our study.

Taking the approach described above, we observed polar localization of FlAsH-labeled ProW, MscL, and AqpZ at low frequencies (less than 20%) in cls or cls⁺ bacteria cultivated at a variety of osmolalities (Fig. 2; see Table S1 in the supplemental material). Cellular processes that are not protein specific may cause these proteins to be concentrated at the poles in a small proportion of cells. In contrast, LacY, ProP, and MscS were concentrated more frequently at the cell poles, and for ProP and MscS this behavior correlated with cellular CL content (Fig. 2 and 3; see Table S1).

At least two questions arise from observations of lipid and protein localization in bacterial cells. How are proteins targeted to particular subcellular locations, and what is the functional significance of lipid and protein localization? The mechanisms targeting proteins to cell poles remain unclear, but our data suggest a role for CL in the polar localization of ProP and MscS (51, 52) (Fig. 1 to 3; see Table S1 in the supplemental material). We previously reported that concentration of ProP at the cell poles is CL dependent and suggested that CL controls the polar localization of ProP (51, 52). Other workers proposed that, when it reaches a critical mole fraction, CL may concentrate spontaneously at the highly curved cell poles (22, 41). Here we compared the proportions of cells with CL, ProP, or MscS concentrated at the poles of bacteria with various CL contents. Concentration of CL at the poles was detected 24-fold more frequently in cls⁺ bacteria than in cls bacteria, but the frequencies did not vary significantly as the CL content was further modulated by varying the growth osmolality (Fig. 2G and H). We do not know whether the frequency of polar localization would be modulated in the critical but experimentally inaccessible range of CL contents between that of cls bacteria cultivated at high osmolality (0.8 mol%) and that of cls⁺ bacteria cultivated at low osmolality (4 mol%).

In contrast to the results for CL, the frequencies with which ProP and MscS were concentrated at the poles varied with both bacterial genotype (cls⁺ or cls) and growth osmolality (Fig. 2A to D). Concentration of ProP at the poles was detected 16-fold more frequently in cls⁺ bacteria cultivated at high osmolality than in cls bacteria cultivated at low osmolality, whereas the corresponding ratio for MscS was only 2-fold. The different correlations of CL localization, ProP localization, and MscS localization with CL content may indicate that they do not reflect direct interaction of ProP or MscS with CL or even the same underlying phenomenon. Other changes are known to accompany the cls mutation (e.g., changes in PG content) and growth in media with variable osmolality (e.g., widespread induction or repression of gene expression [70]). ProP and/or MscS may interact with unknown proteins targeted to the cell poles in a CL-dependent manner rather than with CL itself. The molecular basis for colocalization of membrane proteins with CL at the cell poles will be understood better when the structures of the polar membrane domains are better characterized.

The functional importance of protein localization is not always evident, and phospholipid-dependent protein function is not always simply correlated with protein-lipid colocalization. Polar localization of some proteins has been correlated with functional changes. For example, the localization of ProP in the CL-rich environment at the cell poles tunes its sensitivity to osmotic pressure (51), the chemoreceptor arrays located at cell poles function in chemotactic signal integration (23, 61), and concentration at cell poles increased when the sensor kinases DcuS and CitA were provided with their effectors (fumarate and citrate, respectively) (55). The osmolality at which ProP attains one-half its maximal activity correlates with the anionic phospholipid content (CL plus PG) in both cells and proteoliposomes, although ProP function is more sensitive to CL than to PG (51). However, the proportion of PG increased with growth medium osmolality in cls bacteria, and ProP function was influenced by the varying PG content even though ProP was not concentrated at the poles of these cells. Thus, ProP function depends on its phospholipid environment and not on its polar localization per se.

This study provides the first evidence that MscS concentrates at the poles of E. coli cells in a CL-dependent manner. Such localization would be expected to influence channel function because the pressure sensitivity of MscS is known to depend on membrane curvature in vitro (18). Aggregation of MscS at the cell poles could influence both its channel function and the impact of its structural transitions on membrane strain (20, 40).

Future work will determine whether other proteins included in this study function in a lipid-dependent manner, but this study shows that their function is not simply correlated with their subcellular localization. For example CL, PG, and phosphatidylserine can bind to a specific site on MscL (47, 48), and CL, PG, and phosphatidic acid influenced the rate and extent of calcein flux through E. coli MscL in vitro (46). The effects of phospholipid headgroup composition on purified and reconstituted MscL were also examined by patch clamping in vitro (39). The data suggested that the lipid headgroup composition altered MscL activity by changing the biophysical properties of the membrane rather than by interacting specifically with MscL. Thus, further investigation is required to define the impact of lipids on MscL function in vivo, but MscL does not appear to colocalize with CL at the cell poles. Stopped-flow spectroscopy suggests that aqpZ E. coli cells from stationary-phase cultures shrink more slowly in response to osmotic upshifts than bacteria from log-phase cultures and that the impact of AqpZ on this response is much greater in bacteria from stationary-phase cultures than in bacteria from log-phase cultures, even when aqpZ is expressed from an RpoS-independent promoter (Table 4). Elevated CL levels are characteristic of E. coli cells from stationary-phase cultures (52), so AqpZ function may be CL dependent, even though AqpZ does not colocalize with CL at the cell poles. Overall, lipid-dependent function and lipid-dependent subcellular localization are not always linked.

This study revealed that CL-dependent localization at cell poles is not a general characteristic of transporters, channels, or osmoregulatory proteins in E. coli. Polar localization can be frequent and CL independent (as observed for LacY), frequent and CL dependent (as observed for ProP and MscS), or infrequent (as observed for AqpZ, ProW, and MscL).