The Osmotic Activation of Transporter ProP Is Tuned by Both Its C-terminal Coiled-coil and Osmotically Induced Changes in Phospholipid Composition*

Transporter ProP of Escherichia coli (ProPEc) senses extracellular osmolality and mediates osmoprotectant uptake when it is rising or high. A replica of the ProPEc C terminus (Asp468–Arg497) forms an intermolecular α-helical coiled-coil. This structure is implicated in the osmoregulation of intact ProPEc, in vivo. Like that from Corynebacterium glutamicum (ProPCg), the ProP orthologue from Agrobacterium tumefaciens (ProPAt) sensed and responded to extracellular osmolality after expression in E. coli. The osmotic activation profiles of all three orthologues depended on the osmolality of the bacterial growth medium, the osmolality required for activation rising as the growth osmolality approached 0.7 mol/kg. Thus, each could undergo osmotic adaptation. The proportion of cardiolipin in a polar lipid extract from E. coli increased with extracellular osmolality so that the osmolality activating ProPEc was a direct function of membrane cardiolipin content. Group A ProP orthologues (ProPEc, ProPAt) share the C-terminal coiled-coil domain and were activated at low osmolalities. Like variant ProPEc-R488I, in which the C-terminal coiled-coil is disrupted, ProPEc derivatives that lack the coiled-coil and Group B orthologue ProPCg required a higher osmolality to activate. The amplitude of ProPEc activation was reduced 10-fold in its deletion derivatives. The coiled-coil structure is not essential for osmotic activation of ProP per se. However, it tunes Group A orthologues to osmoregulate over a low osmolality range. Coiled-coil lesions may impair both coiled-coil formation and interaction of ProPEc with amplifier protein ProQ. Cardiolipin may contribute to ProP adaptation by altering bulk membrane properties or by acting as a ProP ligand.

Bacteria respond to changes in medium osmolality by modulating cytoplasmic composition (1)(2)(3). Osmoregulatory transporters and biosynthetic enzymes mediate the accumulation of K ϩ , glutamate, and selected organic solutes as extracellular osmolality increases. Mechanosensitive channels release solutes as osmolality decreases. Three osmoprotectant transporters were shown to act as both osmosensors and osmoregulators after purification and reconstitution in proteolipo-somes: H ϩ symporter ProP of Escherichia coli (4), Na ϩ symporter BetP of Corynebacterium glutamicum (5), and ATP-binding cassette transporter OpuA of Lactococcus lactis (6). Each was activated as electrolytes were concentrated in the lumen of proteoliposomes, with or without osmotically induced proteoliposome shrinkage (7). In addition, the osmotic upshift required to activate BetP (5) or OpuA (8) increased with the mole fraction of the anionic phospholipid phosphatidylglycerol (PG) 4 in the proteoliposome membrane. It was thus proposed that osmosensing occurs when osmotically induced changes in cytoplasmic ionic strength or K ϩ concentration alter transporter-lipid interactions (7,9).
Transporter ProP of E. coli (ProPEc) mediates the uptake of zwitterionic osmoprotectants such as proline and glycine betaine (N-trimethyl glycine) when osmolality is rising or consistently high (10). ProPEc is a 500-amino acid integral membrane protein and a member of the major facilitator superfamily (11). Our homology model of ProPEc (12) is based on the crystal structure of 12-transmembrane helix transporter GlpT (13), which shares a common fold with major facilitator superfamily members OxlT and LacY (14). Protein ProQ of E. coli amplifies ProPEc activity by acting post-translationally (15,16). ProQ is a basic, cytoplasmic protein that may act directly or indirectly on ProPEc (17).
The central cytoplasmic loop (C3) and C terminus of ProPEc are longer than those of its paralogues (7,12) (Fig. 1A). The latter terminates in six or seven of the heptad repeats that characterize ␣-helical coiledcoil-forming proteins (11). Studies of synthetic peptides corresponding to residues 456 -500 (18) or 468 -497 (19,20) of ProPEc showed that residues 468 -497 (encompassing four heptads) form an antiparallel, homodimeric ␣-helical coiled-coil of low stability (Fig. 1, B and C). The low stability of this structure was expected, because basic residues are present at core heptad "a" positions His 495 and Arg 488 (11) (Fig. 1C). Unexpectedly, replacement R488I disrupted coiled-coil formation by this peptide replica, providing the first evidence that the orientation of the coiled-coil might be antiparallel (18). The antiparallel orientation was substantiated by the NMR solution structure, which appears to be stabilized by interactions of Arg 488 with Asp 475 and Asp 478 on the opposing monomer strand (Fig. 1B) (20). This antiparallel structure was also detected in intact ProPEc, in vivo, by chemical cross-linking of introduced Cys residues (21). A higher osmolality was required to activate the R488I variant than wild type ProPEc, and the R488I variant was activated only transiently, whereas activation of wild type ProPEc was sustained indefinitely (18). These results suggested that the C-terminal coiled-coil of ProPEc plays a role in its osmotic activation.
Amino acid sequence comparisons reveal two groups of bacterial ProP orthologues. All have longer C termini than ProP paralogues with known functions not related to osmosensing or osmoregulation (e.g. ShiA and KgtP) (7) (Fig. 1C). Group A orthologues (typified by ProPEc) include a C-terminal ␣-helical coiled-coil domain and Group B orthologues (typified by C. glutamicum ProP (ProPCg)) do not. The coiledcoil domain unique to Group A orthologues is not essential for osmosensing, since ProPCg was found to act as an osmosensor and osmoregulator after expression in E. coli (22). The initial aim of this study was to further elucidate the role of the C-terminal coiled-coil in the osmoregulation of ProP activity. Here we show that the coiled-coil structure shared by Group A ProP orthologues tunes the transporter so that it can be activated in media of low osmolality. We have also discovered that the osmolality required to activate ProPEc in vivo is modulated by the osmolality at which E. coli is cultured. This osmotic adaptation, which correlates with changing membrane cardiolipin (CL) content, ensures that ProP is poised to respond to ambient osmotic conditions. Thus, both the coiled-coil domain shared by Group A orthologues and the membrane cardiolipin content are involved in tuning the functional response range of ProP.

EXPERIMENTAL PROCEDURES
Culture Media-E. coli strains were cultivated at 37°C, whereas the Agrobacterium tumefaciens strain was cultivated at 30°C, in LB medium (23) or in NaCl-free MOPS medium, a variant of the MOPS medium described by Neidhardt et al. (24), from which all NaCl was omitted. MOPS medium was supplemented with NH 4 Cl (9.5 mM) as a nitrogen source and glycerol (0.4% v/v) as a carbon source. L-Tryptophan (245 M) and thiamine hydrochloride (1 g/ml) were added to meet auxotrophic requirements, and NaCl or sucrose was added to adjust the osmolality as indicated. Ampicillin (100 g/ml) was included to maintain plasmids, and arabinose was added as specified to adjust ProP expression.
Bacteria, Plasmids, and Molecular Biological Manipulations-Basic molecular biological techniques were as described by Sambrook and Russell (25). Chromosomal DNA was isolated as described by Bayliss et al. (26). The PCR was carried out as described by Brown and Wood (27). Site-directed mutagenesis was performed using the QuikChange Mutagenesis Kit (Stratagene, La Jolla, CA) as described by Culham et al. (28). Oligonucleotides were purchased from Cortec DNA Services (Kingston, Canada). Each recombinant plasmid was recovered from a ligation mixture by transformation of E. coli DH5␣ (29), and the entire sequence of the encoded proP variant was confirmed (GenAlyTiC, Guelph, Canada) before the plasmid was expressed in E. coli WG350.
Isolation of Genes proPCg and proPAt-A BglII site overlapping the proPEc stop codon in pDC79 was introduced by site-directed mutagenesis, the resulting plasmid (pYT6) was cleaved with EcoRI and BglII to excise proPEc, and the vector fragment was purified. The gene encoding ProPCg was PCR-amplified using plasmid pHP5 (22) as template, and the gene encoding the putative Group A ProP orthologue from A. tumefaciens (ProPAt) was PCR-amplified using the linear chromosome of A. tumefaciens C58 (ATCC number 33970) (American Type Culture Collection (Manassas, VA)) as template. During amplification, an EcoRI site was introduced 5Ј to each open reading frame, and a BglII restriction site was introduced overlapping the stop codon. Each PCR product was digested with EcoRI and BglII, purified, mixed with the vector fragment of pYT6, and ligated with T4 DNA ligase to create plasmids pYT12 (encoding ProPCg) and pYT13 (encoding ProPAt).
C-terminal Truncation of ProPEc-The proPEc gene encoded by pYT1 was PCR-amplified so that the introduced EcoRI site 5Ј to the open reading frame was included, and a stop codon, with an overlapping BglII site, was introduced after the codon for Ala 482 or Thr 489 . The resulting PCR products and plasmid pYT6 were cleaved with EcoRI and BglII. The desired DNA fragments were purified, mixed, and ligated, creating plasmids pMD2 (encoding His 6 -ProPEc-⌬11, truncated at Ala 482 ) and pMD3 (encoding His 6 -ProPEc-⌬18, ProPEc truncated at Thr 489 ).
Transport Assays-Bacteria were cultivated, and assays were performed as described by Culham et al. (10) using buffers prepared as described by Racher et al. (33). Osmolalities of culture media and buffers were adjusted with NaCl or sucrose, as specified, and measured with a Wescor vapor pressure osmometer (Wescor, Logan, UT). Initial rates of proline uptake were measured using L-[U-14 C]proline (Amersham Biosciences) at 0.2 mM. Protein concentrations were determined by the bicinchoninic acid assay (34), using the BCA kit from Pierce with bovine serum albumin as a standard. All assays were done in triplicate, and all experiments were performed at least twice. The rates are cited as mean Ϯ S.D. Regression lines were obtained by fitting the data to empirical Equation 1 (10) using nonlinear regression performed by Sigma Plot 8.0, where ⌸ is the osmotic pressure of the transport assay medium, a 0 is the initial rate of proline uptake measured with medium osmolality ⌸/RT, A max is the uptake rate that would be observed at infinite medium osmolality, R is the gas constant, T is the temperature, 1 ⁄ 2 /RT is the medium osmolality yielding half-maximal activity, and B is a constant inversely proportional to the slope of the response curve. This process yielded estimates for parameters A max , ⌸1 ⁄ 2 /RT, and B.
Base pairs 1318 -1500 of proPEc (encoding ProPEc (Glu 440 -Glu 500 )) were PCR amplified using primers that created flanking BamHI and HindIII restriction sites. The primers were designed to facilitate insertion of the resulting oligonucleotide in vector pQE82L (Qiagen Inc., Valencia, CA), fusing the proP-derived open reading frame to the upstream vector sequence encoding the MRGSH 6 tag (32). The amplicon and vector were cleaved, purified, mixed, and ligated, and recombinant plasmids were recovered. A primer-encoded mutation (P496T) was corrected by site-directed mutagenesis, yielding the desired plasmid (pJKK2) (32), which was introduced to E. coli MG1655 (35) to create E. coli WG864.
To produce peptide MRGSH 6 -ProPEc (Glu 440 -Glu 500 ), E. coli WG864 was cultivated in LB medium supplemented with ampicillin (100 g/ml). Isopropyl-␤-D-thiogalacto-pyranoside (final concentra-tion 1 mM) was added at a culture A 600 of 0.5-0.6, and the cells were harvested by centrifugation when the A 600 reached 2. The resulting pellet was washed twice with 0.1 M potassium phosphate, pH 7.4, and resuspended in 5 ml of lysis buffer (50 mM sodium phosphate, 0.3 M NaCl, 5 mM imidazole, 1 mM Na-EDTA, pH 8) per g, wet weight. All subsequent steps were performed at 4°C. The cells were disrupted by two passages through a French pressure cell (AMINCO, Silver Spring, MD) at 1600 bars pressure. The lysate was centrifuged in the Sorvall SS34 rotor at 12,100 ϫ g for 20 min and then in the Beckman Ti45 rotor at 100,000 ϫ g for 2 h. MRGSH 6 -ProPEc (Glu 440 -Glu 500 ) was purified from the resulting supernatant by Ni 2ϩ -nitrilotriacetic acid affinity chromatography (Qiagen Inc., Valencia, CA) according to the manufacturer's instructions for protein purification under nondenaturing conditions. It was further purified by size exclusion chromatography using a Superdex-75 HR 10/30 column and an Amersham Biosciences fast protein liquid chromatography component system. The resulting 8.5-kDa peptide was homogeneous as determined by Tricine SDS-PAGE (36).
Western Immunoblotting Analysis-Whole cell proteins were prepared for Western Immunoblotting, and it was performed as described above (see "Transport Assays") and by Culham et al. (18), using the procedure of Towbin et al. (37) and selective anti-ProP antibodies, prepared as follows. Anti-ProP antibodies were recovered from 4 ml of adsorbed anti-ProP serum (4) by affinity purification as described by Salamitou et al. (38) using ProP-His 6 (0.3 mg) as ligand. Peptide MRG-SHis 6 -ProPEc (Glu 440 -Glu 500 ) (0.6 mg) was bound to Ni 2ϩ -nitrilotriacetic acid affinity resin (0.2 ml; Qiagen Inc.) in lysis buffer containing 10 mM imidazole and no EDTA. The loaded resin was recovered in a Micro Bio-Spin chromatography column (Bio-Rad), establishing a 0.1-ml column bed. The purified anti-ProP (1 ml) was added to the column, mixed with the resin by pipetting, transferred to a 2-ml vial, and incubated at 20°C, shaking, for 60 min. It was transferred back to the chromatography column, and the column flow-through was collected as selective anti-ProP. The recovered antibodies recognized full-length ProPEc but not MRGSH 6 -ProPEc (Glu 440 -Glu 500 ) on a Western blot.
Determination of Phospholipid Head Group Composition-E. coli cells expressing ProPEc (strain WG350 pDC79) were grown as described above (see "Transport Assays") in MOPS medium supplemented with [ 32 P]phosphate at 5 Ci/ml. Polar lipids were extracted with chloroform/methanol, and thin layer chromatography was performed as described by Wikstrom et al. (39). The relative amounts of the lipid species, identified by comparison with standards, were determined with a Bio-Rad Fluor-S MultiImager.

RESULTS
Osmotic Adaptation of ProPEc-We previously reported that the transport assay medium osmolality required to activate ProPEc-His 6 was independent of the osmolality of the medium in which E. coli was grown (NaCl-supplemented MOPS media with osmolalities in the range 0.12-0.32 mol/kg; Fig. 1B of Ref. 10). A more complex picture emerged when the bacteria were grown at higher osmolalities (up to 0.7 mol/kg). For cells grown in the higher osmolality range, the osmolality required to activate ProPEc depended on the osmolalities of both the growth medium and the assay medium (Fig. 2).
The response of ProPEc to assay medium osmolality fits an empirical relationship (see "Experimental Procedures") that supports extraction of parameters quantitatively describing its osmotic activation (10). Measurements of the initial rate of proline uptake (a 0 ) as a function of osmolality (⌸/RT) are used to determine the uptake rate that would be observed at infinite osmolality (A max ), the osmolality yielding half-max- DECEMBER 16, 2005 • VOLUME 280 • NUMBER 50 imal activity (⌸1 ⁄ 2 /RT), and the slope of the activation curve (inversely proportional to parameter B). Such analysis showed that, for bacteria grown in NaCl-supplemented MOPS media, the osmolality required to activate ProPEc and parameter B were direct functions of growth medium osmolality (Fig. 3, circles) whereas the ProPEc activity attained upon full osmotic activation (A max ) was not (Fig. 2).

Tuning and Osmotic Activation of Transporter ProP
To determine whether growth medium salinity or osmolality determined the osmolality at which ProPEc would activate, cells expressing ProPEc were grown in a sucrose-supplemented, high osmolality medium (0.72 mol/kg). A higher osmolality was required to activate ProPEc in cells grown in this sucrose-supplemented medium than in those grown without added osmolyte (Fig. 2, inset). Furthermore, the assay medium osmolalities yielding half maximal ProPEc activity (⌸1 ⁄ 2 / RT) and the slopes of the activation curves (indicated by B) were similar for cells cultivated in NaCl-and sucrose-supplemented media of similar osmolality (compare circles and triangles in Fig. 3). An osmotic adaptation process appeared to modulate the osmosensory range of ProPEc so that its activity would vary over an osmolality range relevant to ambient conditions.
Osmotic Activation and Adaptation of ProP Orthologues-Two groups of ProP orthologues are found in bacteria, all with extended C termini (7) (Fig. 1C). Group A orthologues, typified by ProPEc, include a C-terminal ␣-helical coiled-coil domain, and Group B orthologues, typified by ProPCg, do not (Fig. 1C). Putative Group A orthologue ProPAt and Group B orthologue ProPCg were expressed in E. coli, and their osmotic activation profiles were examined to determine whether the ability to undergo osmotic adaptation was correlated with the structure of the C-terminal domain.
To assess the function and osmotic sensitivity of ProPAt, E. coli cells in which that transporter was expressed from plasmid vector pBAD24 were cultivated, harvested, and resuspended in low osmolality medium (0.14 mol/kg). Proline uptake rates of these bacteria increased substantially from base line levels as the assay medium osmolality approached 0.51 mol/kg (0.2 M NaCl) (data not shown), with a maximum proline uptake rate of 5.6 nmol min Ϫ1 mg Ϫ1 protein. This suggested an osmotic response for ProPAt, although the absolute activity observed was low. Arabinose was used to induce ProPAt expression so that its activity could be more directly compared with that of ProPEc. The proline uptake rate increased with increasing arabinose concentration, as expected, and a rate of 67 nmol min Ϫ1 mg Ϫ1 protein (comparable with the activity of ProPEc without arabinose induction) was achieved at an arabinose concentration of 0.4 mM (data not shown). When the impact of osmolality on ProPAt activity was again determined after such arabinose induction, the resulting activity profile was similar to that of ProPEc (Fig. 4A).
To determine whether ProPAt would undergo osmotic adaptation, bacteria expressing ProPAt were grown at culture osmolalities of 0.14 and 0.62 mol/kg, and proline uptake rates were measured. As for FIGURE 2. The osmolality required to activate ProPEc increases with growth medium osmolality. E. coli strain WG350 containing pDC79 was prepared as described under "Experimental Procedures" in NaCl-free MOPS medium (0.14 mol/kg (white circles)); in the same medium adjusted with NaCl to attain osmolalities of 0.43, 0.52, 0.60, or 0.70 mol/kg (represented by increasingly dark gray circles); or in the same medium adjusted with sucrose to attain an osmolality of 0.72 mol/kg (triangles, inset). The initial rate of proline uptake via ProPEc was measured using assay media adjusted with NaCl to the indicated osmolalities, and lines were created by regression analysis as described under "Experimental Procedures."  (encoding ProPAt; squares), or pYT12 (encoding ProPCg; diamonds) was cultured, and its initial rate of proline uptake was measured using growth and assay media adjusted with NaCl to the indicated osmolalities. Expression of ProPAt was induced by including 0.3 mM arabinose in the medium. Lines were created by regression analysis as described under "Experimental Procedures." A, bacteria were prepared in NaCl-free MOPS medium (0.14 mol/kg). B, bacteria were prepared in MOPS media adjusted with NaCl to attain osmolalities of 0.62 mol/kg (ProPAt) or 0.60 mol/kg (ProPEc and ProPCg).
Peter et al. (22) reported that ProPCg transports proline and ectoine and that it can sense and respond to osmotic changes after expression in E. coli. E. coli WG350 expressing ProPCg was grown at culture osmolalities of 0.14 and 0.60 mol/kg. As for ProPEc and ProPAt, ProPCg activity depended on both the growth and assay medium osmolalities (Fig. 4,  compare A and B). However, the osmolalities required for half-maximal ProPCg activation (0.45 Ϯ 0.02 and 0.56 Ϯ 0.03 mol/kg, respectively) were up to 2-fold higher than those required to activate ProPEc and ProPAt, both of which include the coiled-coil domain. These observations showed that the C-terminal coiled-coil is not essential for the osmotic adaptation of ProP but did not rule out involvement of the extended, C-terminal sequences shared by all ProP orthologues (illustrated by the gray box in Fig. 1A and the sequence alignment in Fig. 1C).
Changes in Lipid Composition May Cause the Osmotic Adaptation of ProP-The osmolalities required to activate osmoregulated proteins BetP of C. glutamicum (5,40) and OpuA of L. lactis (8) increase as the PG content of the membrane increases at the expense of zwitterionic lipid (phosphatidylethanolamine (PE) or phosphatidylcholine (PC)). We hypothesized that the alteration in osmotic activation threshold for the ProP orthologues might result from alterations in the anionic lipid content of the bacterial membrane due to growth in media of varying osmolalities. The phospholipid head group composition of E. coli strain WG350 expressing ProPEc did change as the salinity of its growth medium varied (Fig. 5A, solid symbols). The CL content increased, PE content decreased, and PG content remained unchanged as cells were grown at increasing osmolalities. Similar changes were observed when the bacteria were grown with and without sucrose as osmolyte (Fig. 5B,  open symbols). Thus, the anionic lipid content (CL plus PG) increased as the zwitterionic lipid content decreased, and the osmolalities required for half-maximal activation of ProPEc in cells grown at varying osmolalities (⌸1 ⁄ 2 /RT) correlated directly with their CL content (Fig. 5C).
The C-terminal Coiled-coil Controls the Osmolality Required to Activate ProP-Although all three ProP orthologues examined in this study showed osmotic adaptation (Fig. 3), the absolute ⌸1 ⁄ 2 /RT values for ProPCg (Group B) were consistently higher than those for ProPEc and ProPAt (Group A). Thus, the transporters that share the C-terminal coiled-coil motif (ProPEc and ProPAt) activated fully at lower assay osmolalities (near 0.4 mol/kg) than the one that lacks the coiled-coil (ProPCg, full activation at 0.6 mol/kg). This led us to hypothesize that the osmolality required to activate ProP depends on the C-terminal domain. In other words, a ProP protein with a coiled-coil structure at the C terminus will activate at a lower osmolality than a ProP protein without one.
ProPEc was truncated to remove almost two (His 6 -ProPEc-⌬11) or three (His 6 -ProPEc-⌬18) C-terminal heptads (see "Experimental Procedures"). These truncations would preclude formation of the antiparallel, four-heptad ProPEc coiled-coil (20). Bacteria expressing these deletion proteins were cultivated, harvested, and resuspended in low osmolality medium (0.13 mol/kg) supplemented with arabinose to elevate their expression. The proline uptake rates of these bacteria, measured in media adjusted to the appropriate osmolality with NaCl, increased from base-line levels as the assay medium osmolality approached 0.6 mol/kg (Fig. 6A). Thus, as predicted, these proteins required a higher osmolality than wild type His 6 -ProPEc to activate (⌸1 ⁄ 2 /RT values of ϳ0.45 and 0.18 mol/kg, respectively).
The maximum proline uptake rate attained by the truncated transporters (8 -10 nmol min Ϫ1 mg Ϫ1 protein) was much lower than that of their wild type control (His 6 -ProPEc, 95 nmol min Ϫ1 mg of protein Ϫ1 ). In principle, these low activities could be due to low expression levels. The anti-His 5 antibodies normally used to assess expression levels of ProPEc-His 6 and its derivatives (28) did not react with His 6 -ProPEc or its derivatives on Western blots. Selective anti-ProPEc antibodies that do not recogize epitopes present in a C-terminal ProPEc fragment (Glu 440 -Glu 500 ) were therefore used to determine the expression levels of His 6 -ProPEc-⌬11 and His 6 -ProPEc-⌬18. Western blots revealed that, with the specified arabinose induction, both deletion proteins and ProPEc were expressed to similar levels (Fig. 6B). Thus, in addition to altering the osmolality at which they became active, deletion of the C-terminal sequence that characterizes Group A ProP orthologues dramatically reduced the amplitude of their osmotic activation. It is possible that the deleted C-terminal sequences are also required for interactions between ProPEc and amplifier protein ProQ.

DISCUSSION
ProPEc and ProPCg were previously shown to act as both osmosensors and osmoregulators (4,22). In this study, ProPAt was found to function in the same manner (Fig. 4A). In addition, all three transporters were found to undergo osmotic adaptation (Fig. 3). Namely, the osmolality required to activate each transporter was proportional to the osmolality of the culture medium in which the bacteria were grown. This adaptive phenomenon broadens the osmolality range over which ProP can promote bacterial osmotolerance, ensuring that the transporter is poised to respond to ambient osmolality.
The osmolalities required to activate osmoregulatory transporters BetP of C. glutamicum and OpuA of L. lactis increase with the PG content of the bacterial (40) or proteoliposome (5,8) membranes in which they reside. We therefore speculated that the osmotic adaptation of the ProP orthologues in E. coli may correlate with changes in lipid composition that would occur due to osmoregulation of phospholipid metabolism, since the cells expressing the transporters were grown in media with increasing osmolalities. Although effects of other parameters (e.g. temperature) on bacterial membrane lipid composition were defined prior to this study, the dependence of E. coli lipid composition on growth medium osmolality (and salinity) had received limited attention (41). The phospholipid composition of other bacteria is affected by growth medium salinity, but the physiological consequences of those changes are not known (42)(43)(44). In fact, the CL content of E. coli cells increased significantly, whereas the PE content decreased, as cells were grown in media with osmolalities in the range pertinent to ProPEc adaptation (Fig. 5). The CL content of E. coli is also known to rise during the transition to stationary phase (41).
In E. coli, CL (or diphosphatidylglycerol) is produced when CL synthase catalyzes the condensation of two PG molecules, releasing glycerol. Thus, bacteria unable to synthesize PG also lack CL (and hence most anionic lipid species). CL synthase activity is attributed to the cls gene product, but other enzymes may also contribute, since cls mutants contain residual CL. Transcription of cls begins immediately downstream from a classical 70 promoter (45,46) and is enhanced in stationary phase (47). Tropp predicted that cls expression may also be controlled by stationary phase factor RpoS (48). Our data support that notion, since RpoS mediates transcription of multiple osmoregulatory genes including proP (49). However, cls was not identified as an RpoSregulated gene during multiple screens designed to identify members of the RpoS regulon (50); nor was it identified as responding at the transcriptional level to osmotic stress (51)(52)(53)(54).
We observed a direct relationship between the mole fraction of CL among E. coli polar lipids and the osmolality required to activate ProPEc (Fig. 5B). A causal link between changing cardiolipin content and altered osmoregulation of ProP activity remains to be demonstrated. However, the adaptation of ProP to growth osmolality could result from the impact of increased CL and decreased PE on bulk membrane properties (e.g. increased anionic surface charge or altered potential to form the nonbilayer HII phase (55)). Alternatively, ProP could interact with and respond directly to CL.
CL constitutes a small proportion of polar lipids in E. coli, and the absolute change in CL content in response to increasing osmolality is quite small (increase of 3-4 mol %). Polar lipid extracts from E. coli include phospholipids derived from both the cytoplasmic membrane and the inner leaflet of the outer membrane. The inner leaflet of the outer membrane contains a higher proportion of PE and a lower proportion of PG than does the cytoplasmic membrane (56). Thus, the impact of increasing osmolality on the proportion of CL in the cytoplasmic membrane may exceed that observed for the total polar lipid pool. Furthermore, CL contains two negatively charged phosphate groups versus only one in the other phospholipids. In addition, dye-binding studies suggest that CL is concentrated near the septa and poles of E. coli FIGURE 6. ProPEc derivatives lacking the C-terminal coiled-coil can be osmotically activated. A, E. coli strain WG350 containing pMD2 (encoding His 6 -ProPEc-⌬11; black circles) or pMD3 (encoding His 6 -ProPEc-⌬18; gray circles) was prepared as described under "Experimental Procedures" in NaCl-free MOPS medium (0.13 mol/kg) supplemented with 0.1 mM arabinose. The initial rate of proline uptake was measured using assay media adjusted with NaCl to the indicated osmolalities. B, the expression levels of the truncated ProPEc variants were compared with that of full-length ProPEc by Western blotting as described under "Experimental Procedures." The lanes of the gel were loaded with equal quantities of solubilized whole cell protein. The primary antibody (selective anti-ProPEc) did not recognize epitopes constituted by residues Glu 440 -Glu 500 of ProP (see "Experimental Procedures"). Marker, 45 kDa. cells (57). Thus, the ProPEc environment could be more strongly influenced by changing CL levels if it were similarly localized. For example, the osmolality required for ProPEc activation could be determined in part by the relative affinities of the ProPEc C terminus for itself (homodimeric coiled-coil formation) and a cytoplasmic membrane surface of varying CL content (7). Bacteria with null mutations in cls contain less than 0.1% CL yet show few phenotypes. It is therefore likely that PG can assume many functions of CL (55). Nevertheless, certain enzymes are specifically CL-dependent (58), and CL is a structural component of certain membrane proteins, in some cases creating a deformable "cushion" between subunits (59). For example, CL is required for the formation of respiratory enzyme supercomplexes in the inner mitochondrial membrane of yeast (60,61). An antiparallel coiled-coil structure links the subunits in ProPEc dimers within the cytoplasmic membrane of E. coli, and it is associated with activation of ProPEc at low osmolality (18) (Fig. 6). Since the osmolality required for ProPEc activation rises as the membrane CL content rises (Fig. 5), CL may intercalate between ProPEc monomers, obstructing the conformational changes necessary for transporter activation.
Our work with ProP orthologues revealed another interesting phenomenon. When grown at low osmolality, Group A ProP orthologues (ProPEc and ProPAt (Fig. 3) as well as OusA from Erwinia chrysanthemi (18), all with C-terminal coiled-coil motifs) could be activated at much lower osmolalities than a Group B orthologue (ProPCg, which lacks that motif) (see Figs. 1 and 3). Furthermore, substitution R488I, which disrupted coiled-coil formation by a peptide replica of the ProPEc C terminus, elevated the osmolality required to activate ProPEc (18). ProPEc variants with C-terminal deletions were characterized to further test the correlation between the coiled-coil structure and the osmotic activation threshold. Earlier we reported that removal of 26 C-terminal residues inactivated ProPEc in vivo (4). However, that study was not designed to detect limited residual activation of the transporter at very high assay osmolality. In this study, removal of sufficient sequence to preclude antiparallel coiled-coil formation (11 or 18 C-terminal residues; see Fig.  1) attenuated the maximum activity attained by the transporters and dramatically increased the osmotic activation threshold (Fig. 6). This result corroborated our hypothesis that antiparallel coiled-coil formation by the C-terminal domains of adjacent ProPEc molecules is required for its activation at low osmolality. Although other mechanisms are possible, attenuation of ProPEc activity by these deletions may indicate that the C terminus is required for interaction of ProPEc with amplifier protein ProQ.
In addition to the structure of the C-terminal domain, the membrane lipid compositions of the bacteria that encode ProP orthologues may modulate their osmotic activation. Group B orthologue ProPCg originates in a membrane composed entirely of anionic lipid, whereas the Group A orthologues examined in this study (ProPEc and ProPAt) originate in membranes containing much less anionic lipid, as does the putative Group A orthologue from Pseudomonas putida (TABLE  ONE). Indeed, even higher osmolalities were required to activate ProPCg (and Na ϩ -betaine symporter BetP) upon expression in C. glutamicum than upon expression in E. coli (22,40).
It has been proposed that osmosensing occurs when changes in cytoplasmic ionic strength or K ϩ concentration immediately alter interactions between existing phospholipid head groups and particular osmosensor domains or change the conformation of an osmosensory protein within the membrane (2,7). We propose that, on a longer time scale, extracellular osmolality changes elicit physiologically relevant alterations in phospholipid head group composition. The altered membrane lipid composition may alter the structures of embedded proteins (e.g. changing the conformation of ProP in a manner that elevates the extracellular osmolality required for its activation). We therefore propose that at least some effects of phospholipid composition on ProP activation are relevant to its osmotic adaptation, not to osmosensing.