Module 5.4.5

Osmotic Stress

KARLHEINZ ALTENDORF,¹ IAN R. BOOTH,² JAY GRALLA,³ JÖRG-CHRISTIAN GREIE,¹ ADAM Z. ROSENTHAL,³ AND JANET M. WOOD^4*

[SECTION EDITOR: JOHN FOSTER]

Posted November 19, 2009

Universität Osnabrück, Fachbereich Biologie/Chemie, Abteilung Mikrobiologie, Barbarastrasse 11, D-49069 Osnabrück, Germany¹; School of Medical Sciences, University of Aberdeen, Institute of Medical Sciences, Foresterhill, Aberdeen, AB25 2ZD, United Kingdom²; Department of Chemistry and Biochemistry and the Molecular Biology Institute, University of California, Los Angeles, PO Box 951569, Los Angeles, CA 90095³; and Department of Molecular and Cellular Biology, University of Guelph, Guelph, ON, Canada N1G 2W1⁴

*Corresponding author. Mailing address: Department of Molecular and Cellular Biology, University of Guelph, Guelph, ON Canada N1G 2W1. Phone: (519) 822–9922, ext. 53866, Fax: (519) 837–1802, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it .

OSMOTIC STRESS TOLERANCE: ESCHERICHIA COLI AND SALMONELLA IN THEIR NATURAL ENVIRONMENTS

E. coli lives as a commensal organism within the intestinal tracts of mammals. However, some E. coli, Shigellae and Salmonellae are intestinal or extraintestinal pathogens (EcoSal chapter 6.4.5). These organisms experience diverse and changing environments within human or animal hosts, outside those hosts on plants, in soil, or in water and during processing and storage of contaminated feed, food, or water (references 1, 161, 203, 274 and EcoSal chapter 8.2). Their survival and growth may be impaired by osmotic stress (fluctuations and extremes of osmotic pressure) and by desiccation or variations in salinity, temperature, or pH. Evidence reviewed below indicates that even small fluctuations in osmotic pressure affect cellular functions and elicit regulatory responses. These responses may result in osmotic stress tolerance and influence the incidence and severity of infections caused by E. coli and Salmonella. For example, osmoregulatory mechanisms may permit uropathogenic E. coli to grow rapidly and extensively in mammalian urinary tracts despite the widely fluctuating osmotic pressure of urine (61, 73, 78, 161) (see “Perspectives,” below).

Studies of E. coli and Salmonella serve as paradigms and provide experimental tools for the study of cellular osmoregulation (314). As a result, the osmotic stress responses of these bacteria have been reviewed extensively (16, 28, 32, 36, 40, 46, 58, 69, 71, 99, 107, 117, 146, 179, 186, 233, 241, 242, 284, 285, 312) and were covered by previous print editions (70, 136, 149). Table 1 provides a glossary of the terms used to describe bacterial osmotic stress tolerance and Table 2 lists quantities relevant to osmotic calculations. This chapter refers primarily to laboratory strains of E. coli K-12 and Salmonella LT-2 because most studies of osmotic stress tolerance have been based on those organisms. It emphasizes new information, in particular, structure-function analyses of osmoregulatory proteins conducted since the last print edition was published.

WHY DO OSMOTIC PRESSURE VARIATIONS CONSTITUTE “STRESS?”

Osmotic pressure changes and extremes are expected to alter bacteria in ways that hinge on the compositions of the bacterial cell and its environment, the different permeabilities of the cell wall and cytoplasmic membrane, and the mechanical properties of the cell wall. We assume that cells respond to osmotic stress because it dramatically affects their physical and chemical nature, yet the effects of osmotic stress on bacteria have been hard to delineate and this section of the osmotic stress module, for the most part, is predictive. Our knowledge of this topic is now improving as biophysical and cell biological tools are applied to bacteria (see Perspectives).

The osmotic pressure of an aqueous solution is proportional to the water activity (a_w, unitless) which is 1.0 for pure water and falls toward zero as solutes are added. The osmotic pressure increases as water activity decreases, the osmotic pressure (Π, atm) being defined by:

Π = −(RT/V_w) ln a_w   (1)
where R is the gas constant (0.082054 liter atm mol⁻¹ K⁻¹), T is the temperature (degrees Kelvin) and V_w is the partial molar volume of water (0.01801 liter mol⁻¹). The osmolality or osmotic pressure at a particular temperature is defined as:

Osmolality = Π/RT   (2)
The units of osmolality are osmolal, or osmoles per kilogram of solvent, in this case, water. The osmolarity, a measure of solute concentration, is the number of osmoles per liter of solution (osmolar, or osmoles per liter of solution). The osmolarity of a solution can be calculated as:

Osmolarity = Σ_i φ_i n_i C_i   (3)
where φ is the osmotic coefficient (discussed below), n is the number of particles (e.g., ions) into which a molecule dissociates, C is the molar concentration of the solute, and the index i represents the identity of a particular solute. The osmolarity can also be calculated from the osmolality as follows:

Osmolarity = osmolality (ρ_sol − c_a)   (4)
where ρ_sol is the density of the solution in grams per milliliter and c_a is the total mass concentration of solute(s) in grams per milliliter.

Osmotic pressure and osmolality can be measured but not calculated. Sweeney and Beuchat reviewed available techniques and endorsed vapor pressure osmometry as the method of choice for osmotic pressure/osmolality determination (292). Osmolarity can be calculated only if the following information is available: the mass or molar concentration of each solution constituent, the degree to which each constituent dissociates and the osmotic coefficient (φ) for each constituent (Equation 3) or the solution osmolality, density, and total mass concentration (Equation 4). Since this information is often not available, osmotic effects are usually reported in terms of measured osmolalities. Sometimes osmolarities are calculated by using Equation 3 and assuming that all osmotic coefficients (φ) are 1 (that is, the solution is “ideal”). This assumption is seldom justified. For example, it yields osmolarities of 0.5 and 1.0 OsM for 0.25 M and 0.50 M NaCl solutions, whereas the osmolalities of those solutions have been measured as 0.47 and 0.93 mol/kg, respectively, at 20°C (1). It yields osmolarities of 0.5 and 1.0 OsM for 0.5 M and 1.0 M sucrose solutions that have measured osmolalities of 0.59 and 1.45 mol/kg, respectively, at 20°C (177). The term osmolality is used below because bacteria respond to the osmolality of their environment.

E. coli is reported to grow in media with osmolalities that range from 0.015 to 3.0 mol/kg, with significant variation in osmotolerance among E. coli isolates (42, 162, 241). For example, the osmolalities of MOPS medium and LB, both standard laboratory culture media for E. coli and Salmonella, are 0.2 mol/kg and 0.4 mol/kg, respectively, at 25°C (314). The osmolality of MOPS medium, which lacks osmoprotectants, was adjusted with NaCl to optimize the growth of E. coli B (214). The NaCl content (and hence the osmolality) of LB varies, but LB always contains osmoprotectants provided by yeast extract and tryptone. The cited osmolality applies to the LB formulation of Miller (202) which contains 10 g of NaCl per liter.

Structure and Permeability of the Cytoplasmic Membrane

Phospholipid membranes are highly permeable to water but intrinsically impermeable to inorganic ions and polar organic solutes like amino acids, sugars, and nucleotides. Cellular dehydration occurs within less than a second when solutes are added to the external medium, increasing its osmotic pressure (7). Conversely, water flows into cells when the extracellular osmotic pressure decreases. The specific impacts of osmotic shifts on E. coli and Salmonella cells are determined by the properties of their walls and cytoplasmic membranes, as follows.

Liposomes prepared with a polar lipid extract from E. coli are freely permeable to water and other uncharged polar molecules with molecular weights below 250 g/mol {e.g., ethanol, glycerol, poly(ethyleneglycol) with a degree of polymerization below 5 [238]}. In addition the aquaporin (AqpZ), which mediates water flux, is present in the cytoplasmic membrane of E. coli and the (aqua)glyceroporin (GlpF), which mediates water, glycerol, urea, and glycine fluxes, is present in the cytoplasmic membranes of E. coli and Salmonella (47, 283). AqpZ and GlpF, composed of bundled transmembrane α-helices, are members of the Major Intrinsic Protein (MIP) family. AqpZ was shown to enhance the osmotic water permeability of E. coli in experiments based on cryoelectron microscopy (84) and light-scattering spectroscopy (184), but no growth phenotype has been associated with AqpZ deficiency (48, 184, 277; T. Romantsov, A. R. Battle, J. M. Hendel, B. Martinae, and J. M. Wood, submitted for publication). In principle, the levels of AqpZ and GlpF could be adjusted to modulate bacterial membrane permeabilities and, hence, the rates of osmotically induced water or glycerol fluxes. Indeed, stationary phase sigma factor RpoS (σ³⁸) mediates growth-phase-dependent aqpZ transcription. Light-scattering experiments suggest that AqpZ has little impact on the water permeability of cells from exponential-phase cultures, whereas the passive water permeability of the E. coli membrane decreases while the rate of AqpZ-mediated water flux increases as E. coli enters stationary phase (48, 184, 277; Romantsov et al., submitted). GlpF is encoded by the glpFKX operon (GlpK is glycerol kinase and GlpX is fructose-1,6-bisphosphatase II) and regulation of glpFKX transcription by the catabolite repressor protein (CRP) and GlpR reflects the use of glycerol as a carbon and energy source (297, 309). Neither channel has yet been implicated in an osmotic stress response, perhaps because phospholipid bilayers are intrinsically permeable to water and glycerol or because the roles of AqpZ and GlpF in the kinetics of osmoregulatory responses have not been investigated.

The mole fractions of the major phospholipids in the cytoplasmic and outer membranes of E. coli are usually cited as 0.75 for phosphatidylethanolamine (PE), 0.20 for phosphatidylglycerol (PG), and 0.05 for cardiolipin (CL, also known as diphosphatidylglycerol) (66). The relative compositions of the membrane leaflets are not known. CL and PG are anionic, whereas PE is zwitterionic, and each CL molecule contains two phosphates, whereas each PE or PG contains only one phosphate. The proportions of these lipids vary with growth medium osmolality and growth phase. The proportion of CL increases approximately twofold at the expense of PE, and PG remains constant, as E. coli is cultured to exponential phase in minimal media of increasing osmolality (298). The CL synthase encoded by the osmotically inducible cls gene is responsible for most CL synthesis. In cls⁻ bacteria, a trace of CL remains, the PG:PE ratio is elevated and that ratio increases further as the bacteria are cultivated to exponential phase in minimal media of increasing osmolality (66, 250). Thus, PG synthesis is also osmotically induced and/or PE synthesis is osmotically repressed, and the cls defect does not significantly alter the proportion of phosphate in anionic phospholipid head groups.

The proportion of anionic lipid (PG plus CL) decreases 2-fold on entry to stationary phase, but the proportion of phosphate in anionic lipid head groups decreases only 1.5-fold because CL rises as PG falls (250). The cyclopropane fatty acid (CFA) synthase encoded by cfa adds methyl groups from S-adenosylmethionine across double bonds of unsaturated fatty acids in existing phospholipid molecules of E. coli (113). Transcription of gene cfa depends on RpoS (σ³⁸) and is osmotically inducible (113). Bacterial CFA content increases as cultures enter stationary phase (308). Neither the impact of osmotic stress on CFA content nor the impacts of changing head group and fatty acid composition on intrinsic or channel-mediated membrane permeability have been reported.

Phospholipid composition is important to osmoadaptation because it affects passive membrane permeability (e.g., for water). In addition, phospholipid composition determines the membrane's lateral pressure profile, and some osmoregulatory proteins may sense effects of osmotic shifts on the lateral pressure profile (232, 312). A planar phospholipid bilayer represents a compromise between the tendency of each monolayer to be curved (intrinsic curvature, arising if phospholipid molecules are conical rather than cylindrical in shape) and the tendency for both monolayers to be flat so that all acyl chains are sequestered from water. The frustration of intrinsic curvature within planar phospholipid bilayers creates a lateral pressure that varies across the membrane and can affect membrane protein structure (51, 185). The lateral pressure profile is influenced by the relative cross-sectional areas of the phospholipid headgroup and acyl chain regions as well as the solution environment of the phospholipid headgroups (312). Among the predominant lipids in E. coli and Salmonella membranes, PG is cylindrical and PE is conical because the ethanolamine headgroup is compact. CL may behave like a cylindrical or a conical lipid depending on its acyl chain composition and the divalent cation content of its environment (83). Osmotically induced changes in phospholipid headgroup and fatty acid composition, outlined above, may affect osmosensing by altering the lateral pressure profile of the cytoplasmic membrane and its sensitivity to the periplasmic and cytoplasmic environments (232, 312).

Structure and Permeability of the Cell Wall

In E. coli and Salmonella the murein layer is a thin mesh with holes approximately 2 to 5 nm in diameter, expected to allow passage of globular proteins with molecular masses in the range of 25 to 55 kDa (88). The phospholipid-lipopolysaccharide bilayer of the outer membrane has important barrier properties, but outer membrane porins such as OmpF and OmpC provide transmembrane aqueous channels approximately 1 nm in diameter (216). As β-barrel proteins, OmpF and OmpC are structurally distinct from AqpZ and GlpF. These porins have a slight preference for cations over anions, and OmpF allows the permeation of slightly larger solutes than OmpC (216). Thus, the outer membrane and murein layer of E. coli and Salmonella are much more permeable than the cytoplasmic membrane to the inorganic ions (e.g., Na⁺ and Cl⁻) and small organic solutes (e.g., mono- or disaccharides) that typically impose osmotic stress in vivo and in vitro. If such small solutes are added to the external medium they are expected to passively equilibrate across the periplasmic membrane so that the resulting osmotic stress will dehydrate the cytoplasm but not the periplasm.

In E. coli K-12 the proportions of OmpF and OmpC present in the outer membrane are regulated at the transcriptional level by two-component system EnvZ/OmpR and at the translational level by small, antisense RNA micF, each a paradigm for its mode of regulation (115, 216, 273). OmpF and OmpC are controlled by extracellular pH, periplasmic membrane-derived oligosaccharides, and polyamines (in particular, polyamines delivered to the periplasm from the cytoplasm) (85). Porin regulation has been shown to confer bile salt resistance and possibly acid tolerance, but not osmotic stress tolerance (216).

Turgor Pressure, Membrane Strain, and Cell Volume

Turgor pressure (ΔP) is the hydrostatic pressure that balances the osmotic pressure difference between the cell interior and exterior:

ΔP = Π_i – Π_o   (5)
where Π_i and Π_o are the cytoplasmic and extracellular osmotic pressures, respectively. Turgor pressure arises as the cell wall limits osmotically induced water influx, so it is usually assumed that turgor presses the cytoplasmic membrane against the murein sacculus. However, this view has been questioned (see further discussion below).

Physicists define “strain” as a change in size or shape of an object in response to an applied stress; for example, an increase in cell wall area (strain) in response to increasing turgor pressure (stress). In quantitative terms:

Stress = (elastic modulus) (strain)   (6)
The elastic modulus captures the quantitative relationship between strain and stress, reflecting physical properties of the object to which stress is applied (312). The impacts of osmotic pressure changes on E. coli and Salmonella reflect the differing permeabilities and stress-strain relations of the murein layer (which is elastic but rigid) and the cytoplasmic membrane (which is inelastic but deformable or fluid) (71, 269, 312). The stress of changing turgor pressure strains the sacculus, causing it to shrink or stretch. Thus, small changes in osmotic pressure may alter cell volume without substantially altering turgor pressure. The relative surface areas (or surface area ranges) of the murein and cytoplasmic membrane are not known. The cytoplasmic membrane is expected to have a high elastic modulus so that its area will change little in response to the stress of turgor pressure. However, the membrane may become more or less wrinkled as water leaves or enters the cytoplasm. Large osmotic upshifts eliminate turgor pressure and cause plasmolysis, or shrinkage of the membrane-bounded cytoplasm away from the fully retracted murein sacculus (155, 269). Plasmolysis clearly redistributes cell volume by diluting the periplasm and concentrating the cytoplasm. Effects of osmotic downshifts on cell structure may also be complex. Gradual dilution of the external medium causes liposomes to become strained, gradually leaking solutes while retaining their integrity (101, 119, 310, 311). Large, abrupt osmotic downshifts cause cells to lyse if the elastic limits of the membrane and the sacculus are exceeded (173).

Osmoregulatory mechanisms (discussed below) cause Π_i to rise and fall in parallel with Π_o. E. coli is often reported to maintain a turgor pressure of a few atmospheres (210). (A turgor pressure of 1 atm would correspond to a cytoplasmic osmolality of 0.24 mol/kg at an extracellular osmolality of 0.20 mol/kg.) According to the surface stress theory, bacteria maintain turgor pressure to ensure extension of the murein sacculus during growth (154). Experiments substantiate a role for turgor pressure in plant and fungal cell growth (122). This relationship remains controversial for E. coli and Salmonella because these bacteria are too small for direct measurement of turgor pressure (made by impaling cells with pressure probes) and indirect, physiologically relevant turgor pressure measurements are difficult (312). Osmoregulatory mechanisms may not be designed to “fix” turgor at an optimal set point since turgor pressure changes are not essential for activation of some osmoregulatory systems (see below) and the turgor pressure of E. coli decreased from approximately 3.1 atm to less than 0.5 atm as the growth medium osmolality increased from 0.03 to 0.5 mol/kg (54). In addition, Meury (198) showed that E. coli cells continue to grow, even though they do not divide, after a substantial osmotic upshift. Thus, cell division is more sensitive to osmolality than cell expansion is.

The composition of the periplasm is expected to reflect the composition of the bacterial environment because the outer membrane and murein layer are porous, at least for low-molecular-weight solutes. However periplasmic composition can also be modulated. Anionic β-D-glucans (also denoted membrane-derived oligosaccharides [MDOs]) accumulate in the periplasm when E. coli is cultivated in very-low-osmolality media (less than 0.05 mol/kg) (149). These glucans are glucose oligomers substituted with phospho-sn-1-glycerol, phosphoethanolamine, and O-succinyl ester residues. MDO accumulation can raise the osmotic pressure of the periplasm by causing counterion accumulation (the Donnan effect, discussed previously (see “Osmoregulation of the periplasm” in reference 70). This would be expected to favor water influx from any neighboring compartment of lower osmolality (in this case, the external medium).

Some data suggest that, at least for bacteria cultivated at high osmolality, periplasmic and cytoplasmic osmolality are equal (281). This implies that turgor pressure is exerted on the covalently linked outer membrane-murein complex, that the cytoplasmic membrane defines the interface between the cytoplasm and periplasm without being part of the stress-bearing cell wall (54), and that changes in relative osmolality of the cytoplasm and periplasm due to osmoregulatory processes will adjust their relative volumes while also altering turgor pressure. Cytoplasmic and periplasmic volumes are difficult to measure and are usually reported relative to total cell mass or protein, not per cell (312). Cells vary in size and shape as a function of culture growth phase and medium osmolality (e.g., reference 250), and both absolute and relative periplasmic and cytoplasmic volumes will vary with cell size and shape. Thus, uncertainties remain and further studies of the relationships among osmoregulatory processes and cell size, shape, compartmentation, and turgor pressure are warranted.

Physical Chemistry of the Cytoplasm

Osmotic pressure changes profoundly affect cellular physical chemistry because osmotically induced water fluxes and ensuing cellular responses alter cytoplasmic composition. Indeed, chemists wishing to elucidate complex interactions among biological solutes and water have chosen experimental systems based on cellular osmoregulation (34, 58, 243, 294). The cytoplasm and periplasm are gels that comprise water, inorganic ions, uncharged and charged organic solutes with broad molecular size distributions, and extended macromolecular complexes (e.g., ribosomes, the nucleoid, the murein). The physical and chemical properties of these gels, conferred by all of their solutes, profoundly influence the folding, interactions, and functions of each constituent biopolymer (protein or nucleic acid). Key properties are conferred by solute “collectives.” Biopolymers exclude one another from the space they occupy (a phenomenon also known as macromolecular crowding), all solutes contribute to the osmotic pressure, and all ions contribute to the ionic strength.

The biopolymer volume fraction (Φ, the fraction of solution volume from which biopolymers exclude one another, a measure of macromolecular crowding) can be calculated as:

Φ = Σ c_i v_i   (7)
where c_i is the mass concentration of the ith biopolymer (grams of biopolymer per milliliter of solution), v_i is its partial specific volume (milliliters of biopolymer per gram of biopolymer), and Φ is summed over all biopolymers. Most proteins have a partial specific volume of approximately 0.7 ml/g (50). The biopolymer volume fraction of the nonnucleoid cytoplasm of E. coli has been estimated as 0.3 to 0.4. The corresponding degree of crowding could be simulated by a globular protein with a molecular mass of approximately 75 kDa at a concentration of 0.34 g/ml (321, 324). Macromolecular crowding slows biopolymer diffusion, favors biopolymer folding and association, and can mitigate effects of otherwise unfavorable conditions on biopolymer folding, interactions, and functions (209). Thus, osmotic stress is likely to affect intracellular processes by altering crowding.

Changes in osmotic pressure and ionic strength also influence cellular processes. Biopolymer structures, interactions, and functions may be osmolality sensitive if solutes are sterically excluded from biopolymer-associated water, creating osmotic stress on a molecular scale. Processes known to be modulated by osmotic stress in this way include enzyme-catalyzed reactions, channel-mediated transmembrane solute fluxes, and DNA-protein interactions (240). The ionic strength (Ι) is a measure of the total ion content of a solution:

Ι = ½ Σ (c_i z_i²)   (8)
where c_i is the concentration of the ith species, z_i is its net charge, and the sum extends over all ions present. Electrostatic interactions weaken as ionic strength increases. Such effects may be particularly important for interactions of proteins with nucleic acids and the anionic membrane surfaces of E. coli and Salmonella.

In addition to these collective effects, each solute has specific effects on biopolymer properties. Some effects of ions on biopolymers follow the empirical Hofmeister series, an ion ranking based on diverse physical processes (65). Physiologically relevant ions such as Na⁺, K⁺, Cl⁻, phosphate, and glutamate fall near the middle of the Hofmeister series. This means that they are neither extremely kosmotropic (decreasing protein solubility while favoring protein stability and aggregation) nor extremely chaotropic (increasing protein solubility while favoring protein dissociation and denaturation). However, as shown below, differences among these ions can profoundly affect physiological processes.

The impacts of nonionic and zwitterionic organic solutes on biopolymer folding, structures, and interactions are also under intense investigation (34, 243, 294). They include “compatible solutes” which are osmolytes that accumulate to high levels, without impairing cellular functions, when cells are cultivated in high osmotic pressure media (see “Accumulation of compatible solutes,” below). For example E. coli and Salmonella use trehalose, proline, glycine betaine, ectoine, and other compounds as compatible solutes (Fig. 1). Thermodynamic analyses and calculations show that effects of osmolytes on protein folding arise from the interdependent interactions of water and osmolytes with the peptide backbone and the amino acid side chains (15, 123, 257, 282). The physiological relevance of these phenomena was demonstrated by using spectrofluorimetry to probe the impacts of proline, glycine betaine, trehalose, and urea on the folding of a heterologous model protein in the E. coli cytoplasm (134, 135).

Fig. 1Solutes involved in osmoregulation by Escherichia coli and Salmonella. Osmotic upshifts stimulate K⁺ uptake, putrescine efflux, and the accumulation of glutamate as K⁺ counterion. Glutathione, MOPS (3-[N-morpholino]propanesulfonate), and other anions can also accumulate in cells under osmotic stress. Later, trehalose is synthesized and potassium glutamate levels fall. Uptake of available osmoprotectants attenuates those responses. Osmoprotectant choline is taken up and oxidized to glycine betaine. Osmoprotectants glycine betaine, proline, proline betaine, taurine, pipecolate, ectoine, and carnitine are transported into the cytoplasm where they accumulate unmodified as “compatible solutes.” The mechanisms responsible for solute accumulation and release are listed in Table 3 and illustrated in Fig. 2.

Fig. 2Osmoregulatory systems of Escherichia coli and Salmonella (see also Table 3 and the text). Aquaporin AqpZ mediates transmembrane water flux. K⁺ transporter TrkAE(G/H) mediates K⁺ accumulation in response to high osmotic pressure (TrkE is also known as SapD); K⁺ transporter KdpFABC contributes to a lesser extent. KdpDE is a two-component regulatory system that controls kdpFABC transcription in response to K⁺ supply and osmotic stress. Suppression of glutamate catabolism leads to its accumulation as K⁺ counterion. At high osmotic pressure trehalase (TreA) hydrolyzes extracellular trehalose. OtsA and OtsB mediate trehalose synthesis. Expression of treB (encoding trehalose-specific Enzyme IIBC^Tre) and treC (encoding trehalose-6-phosphate phosphatase) is repressed. Transporters ProP, ProU, BetT, and BetU mediate organic osmolyte accumulation at high osmotic pressure. Their properties are summarized in Table 4. Enzymes BetA and BetB mediate glycine betaine synthesis from choline. Mechanosensitive channels MscL and MscS mediate solute efflux in response to decreasing osmotic pressure. Factors regulating transcription of the genes encoding these systems are listed in Table 3.

Adjustment of the physicochemical properties of the cytoplasm in response to osmotic stress may equal or exceed the adjustment of turgor pressure in physiological significance. Indeed, osmosensors are of interest, in part, because they may detect osmotically induced alterations to the physical chemistry of the cytoplasm. Such behavior would fundamentally differentiate osmosensors from the solute-specific chemosensors that are ubiquitous in biology.

OSMOTIC ADAPTATION

Osmotic shifts cause water to flow across bacterial cytoplasmic membranes within milliseconds. Bacteria respond within minutes by actively adjusting the distributions of selected osmoregulatory solutes across the cytoplasmic membrane (312). Solute accumulation in response to increasing or high external osmotic pressure forestalls or reverses cytoplasmic dehydration, whereas solute release in response to decreasing external osmotic pressure prevents cytoplasmic dilution and cell lysis. E. coli and Salmonella use K⁺, organic anions like glutamate, and organic osmolytes with no net charge as osmoregulatory solutes (312) (Fig. 1). Potassium glutamate accumulates rapidly in response to osmotic upshifts and it induces osmoregulatory responses. However, organic osmolyte accumulation is the preferred mechanism for the maintenance of cellular hydration, presumably because the osmolytes (also designated “compatible solutes”) elicit more complete cytoplasmic rehydration while favoring native macromolecular structures and interactions (241, 242). Accumulation of compatible solutes attenuates potassium glutamate accumulation and is associated with the release of potassium glutamate (90). Consequently, effects of potassium glutamate accumulation on macromolecular structures and interactions are avoided or relieved. Thus, osmoregulatory responses counter imposed water fluxes and mitigate their consequences.

Arrays of osmoregulatory systems mediate solute accumulation and release by E. coli and Salmonella (Fig. 2 and Table 3). Current studies focus on regulation of their activities and cellular levels. Future work will continue to elucidate how these mechanisms, which seem redundant, are integrated over time and space to deliver effective osmoadaptation for bacteria experiencing diverse and changing natural environments. This review integrates data obtained with a range of strains (and mutants), different media, in vitro versus in vivo systems, and different time intervals. Each of these parameters can influence the experimental outcome and cause an apparent contradiction with earlier data. Not all of these instances can be documented or discussed. We encourage the reader to treat this module as an introduction and to use the original literature as the basis for the development of their further understanding.

RESPONSES TO INCREASING OSMOTIC PRESSURE

Bacteria respond to increasing osmotic pressure by rapidly activating mechanisms that are already present and more slowly implementing those that require osmotically induced gene expression (312). The detailed, temporal sequence of the osmoregulatory responses is not fully understood. Some are too rapid for kinetic analysis with existing techniques (e.g., reference 299), and the relationship between the magnitude of an applied osmotic stress and the selection and ordering of responses has not been studied systematically. We do know that substantial osmotic upshocks (1.2 OsM) halt macromolecule synthesis and inhibit protein functions while selectively activating osmoregulatory transporters (200, 201). If organic osmoprotectants are not available, potassium glutamate accumulates to be replaced later by trehalose. Alternatively, potassium glutamate accumulation is attenuated as available osmoprotectants are transported into the cell where they accumulate unchanged (e.g., proline, proline betaine, and ectoine) or are converted to compatible solutes (e.g. choline, which is converted to glycine betaine in E. coli). Activation of enzymes and transporters is integrated with regulation of gene expression to bring about these responses. Consequently, adaptation to high osmolality must be seen as a temporal sequence in which the constitution of the cytoplasm (and potentially periplasm) changes continuously. Osmoregulatory responses occur in that context and multiple cellular properties, summarized above, may signal a need for adaptation (312).

Global Transcriptional Responses to Osmotic Upshift

Upon osmotic upshock, the majority of transcription in E. coli is rapidly altered (63, 111, 143, 308). When cells are growing exponentially, the production of stable RNA species (rRNA and tRNA) dominates, because 60 to 90% of the RNA produced is devoted to forming the translation apparatus (reviewed in reference 224). When an osmotic shock halts growth temporarily ribosomal RNA transcription is rapidly repressed (111). This forms an essential part of the adaptive response by conserving resources (109). Ribosomal RNA synthesis only resumes after protective responses permit cell growth to resume (111).

Microarray studies suggest that the adaptive response involves the upregulation of 175 to 400 transcripts (63, 308). Most of these are products of the alternative σ³⁸ (RpoS) form of RNA polymerase, which is rare prior to osmotic shock, and transcripts are produced in a temporal sequence of events (17) (see “Accumulation of compatible solutes: osmotic induction of transcription,” below). The array of genes induced varies with the experimental conditions, and not all relevant conditions have been explored. For example, only 60% of the genes upregulated by equivalent, acute salt and sucrose stress are the same (bacteria cultivated in minimal medium adjusted to 2.7 mo/kg) (270). In addition, long-term growth in salt yields a slightly different pattern from osmotic upshift (116). Further work will show which of the identified genes encode proteins that provide osmoprotection. However, it is clear that osmotic stress results in an impressive global reprogramming of transcription in which transcription mediated by σ³⁸ RNA polymerase complements, and, in some cases, replaces, stable RNA production by the housekeeping σ⁷⁰ RNA polymerase.

Increases in the level of the σ³⁸ protein account for much of the upregulation. These increases are achieved primarily by a combination of increases in the rate of rpoS gene translation and in σ³⁸ protein stability (for review see reference 127). The activation pattern is not fixed and can depend on the nature of the growth medium (143) because genes subject to σ³⁸ transcription do not need to be activated either at the same time or to any fixed extent (17, 63, 308). The reprogramming occurs in the context of changing concentrations of potassium glutamate that accumulates rapidly upon osmotic shock (90). Proteins that import or produce other solutes to replace potassium glutamate may be synthesized at low osmolality and available before an osmotic upshock (e.g., ProP) and/or among the products of the transcriptional reprogramming (e.g., ProP, ProU, OtsAB). The mechanisms governing potassium glutamate accumulation and replacement are discussed below.

Overall, the global reprogramming of transcription can be considered to occur in four phases (Fig. 3). In the first phase, ribosomal component synthesis is inhibited, and then in a second phase σ⁷⁰ RNA polymerase is selectively activated at genes such as those for transporters ProU and ProP that scavenge for osmoprotectants (see below). During this second phase σ³⁸ is produced, which, in a third phase, directs transcription of a large number of genes to allow osmotic adaptation and potentiate survival (126). Finally, many of these changes are reversed or attenuated as potassium glutamate is replaced with compatible solutes and the cell approaches full adaptation.

Fig. 3Temporal response of transcription to osmotic upshift.

Both activation of RpoS-dependent gene transcription and repression of ribosomal RNA synthesis are triggered by increasing potassium glutamate levels via mechanisms that are not completely understood. The consequences of potassium glutamate accumulation include both a direct influence on RNA polymerase activity (see below) and a derepression of individual genes regulated by proteins such as H-NS and CRP (107, 316). Differential reliance on DNA supercoiling might also contribute to changing patterns of expression (63, 129, 164). DNA supercoiling increases upon osmotic shock, and its artificial suppression leads to a reduction in the transcription of osmotically responsive genes (63, 108, 129, 215). Last, many genes can be transcribed by both the σ³⁸ and σ⁷⁰ forms of RNA polymerase. In those cases transcript levels would correspond to the net effect of the new intracellular environment on transcription mediated by both forms of the polymerase (128, 254). RNA array analyses, often used to classify osmotically induced genes, focus on large changes and may not detect changes effected at the translational level or changes in the transcription start site that would characterize genes whose transcription largely shifts from σ⁷⁰ to σ³⁸.

The primary transcription signaling mechanism is thought to involve responses to elevated levels of potassium glutamate. The addition of potassium glutamate alone to purified transcription systems can mimic a substantial fraction of the observed differential transcription (111, 129, 133, 164, 169). Models of activation relying on increases in ionic strength or turgor pressure have also been suggested. However, no regulatory proteins that bind potassium glutamate have been identified. Recent evidence suggests that it is not the ionic strength effects of potassium glutamate that are important, but rather that the glutamate anion has the capability of remodeling and activating RNA polymerases poised at some osmotic promoters by acting as a Hofmeister salt (110).

OsmY is a periplasmic protein of unknown function, and osmY transcription is osmoregulated. In vivo, the target of activation at the osmY promoter is a poised, inactive RNA polymerase as shown by chromatin immunoprecipitation studies (253). In vitro, the polymerase contains σ³⁸ and is inactive because of tight DNA wrapping. Potassium glutamate activation is accompanied by the unwrapping of the osmY promoter DNA, releasing RNA polymerase for transcription (169). The RNA polymerase has also been shown to be poised and inactive at the otsB promoter in vitro in the absence of potassium glutamate. Activation at both promoters in vitro depends on the addition of the glutamate anion and can be achieved by other anions according to their ability to act as Hofmeister salts (110). Glutamate-dependent release of RNA polymerase for transcription is accompanied by conformational changes involving the unique C-terminal tail region of the σ³⁸ subunit.

For all genes that have been tested, σ³⁸-mediated transcription is stimulated by the levels of glutamate that exist shortly after shock, although not all are stimulated equally well and it is probable that diverse mechanisms are involved (89, 107, 169, 170, 252). By contrast, the effects of glutamate on σ⁷⁰ transcription range from severe inhibition to stimulation (111, 151, 169, 170). The most important of these effects is the inhibition of ribosomal transcription, which occurs because σ⁷⁰ RNA polymerase engages ribosomal promoters poorly in the presence of an elevated concentration of potassium glutamate (111). Thus, differential transcription is largely achieved by osmotic shock triggering the accumulation of σ³⁸ protein and potassium glutamate, which act in concert to produce global changes.

Transcription patterns may change again after the cell completes this temporal response to osmotic upshift and adapts to the changed environment. This has not been studied systematically, but effects of continuous thermal and osmotic stress on transcription were recently compared (116). Most studies observe the cells as they transition to stationary phase in which different mechanisms repress ribosomal transcription and activate σ³⁸ transcription (109, 128, 224). It is known that cells tend to reverse the transcription program as they resume growth (17, 111, 143). This is consistent with the observation that the uncharged solutes that replace potassium glutamate do not inhibit transcription by either form of RNA polymerase (169). Thus, the direct effects of changing potassium glutamate levels, coupled with the effects of changing levels of gene-specific regulators and DNA supercoiling, can account substantially for the changing global patterns of transcription following osmotic upshock.

Accumulation of Potassium and Organic Anions

Control of K⁺ Influx.

E. coli and Salmonella accumulate K⁺ within minutes after an osmotic upshock (100). K⁺ constitutes the main monovalent cation in the bacterial cell. However, not all of the K⁺ is osmotically active since it also serves other functions, for example, as a constituent of ribosomes. In addition, the total charge of fixed or nondiffusible macromolecular anions (primarily nucleic acids [55, 194]) cannot be balanced by the total charge of macromolecular cations. Electroneutrality requires that a fraction of the small cations balance the excess negative charge of the macromolecular anions. Cations balancing the negative charge of macromolecules are often referred to as bound, but this balancing represents ionic interaction rather than site-specific binding, and such ions can be readily replaced during osmotic transitions. Thus, ions associated with macromolecules have low osmotic activity (8). Only the fraction of the total ion concentration that is balanced by other small ions makes a large contribution to osmotic pressure.

The strong dependence of intracellular K⁺ on medium osmolality in the absence of osmoprotectants readily suggests that K⁺ is a major contributor to cytoplasmic osmolality (55, 56, 100). Whereas the osmotic dependence of cytoplasmic K⁺ is a direct result of the control of K⁺ uptake and efflux via membrane-embedded transporters, the signals triggering the activation or synthesis of the transport complexes are still unclear (see below). There are three saturable K⁺ uptake systems in both E. coli and Salmonella, as well as in many other bacteria, namely Trk, Kdp, and Kup (see also reference 272). Although any one of these suffices to support growth in media with moderate K⁺ concentrations, K⁺ uptake via all three is increased upon osmotic upshock, thereby demonstrating that these K⁺-scavenging systems fulfill a dual function in essential K⁺ supply and osmoprotection.

The predominant K⁺ uptake system is the Trk, which is expressed constitutively and has a modest affinity for K⁺ (K_m = 1 mM). Trk is sufficient for K⁺ uptake and allows cells to thrive in media with K⁺ concentrations above 200 μM. The system is composed of a membrane-integral K⁺/H⁺ symporter module, which in E. coli K-12 strains is present as two separate homologous polypeptides, TrkG and TrkH, comprising 41% identical amino acid residues (267). Many E. coli strains and other bacteria contain just one copy of this Trk module (equivalent to TrkH). Based on sequence comparison, the Trk belongs to a superfamily of K⁺ transporters present in plants, fungi, bacteria, and archaea, that is suggested to have evolved from an ancestral potassium channel by gene duplication and fusion events (93, 94). The system also contains two peripheral, cytoplasmic proteins, TrkA and TrkE, which have NAD⁺- and ATP-binding properties, respectively (121, 285), although the major driving force for transport is the proton motive force. Osmotic upshock only stimulates influx while efflux remains constant. K⁺ uptake increases immediately (within 5 s) due to the high maximum transport rate of Trk (10⁴ to 10⁵ K⁺ complex^-1 s^-1) (267, 285) and persists for about a minute (247). Later, influx falls to again approach the efflux rate as the cells attain a new steady state that is characteristic of the higher osmolality (200).

The other K⁺ uptake system is the Kdp system (reviewed in references 40 and 112). Kdp has a 500-fold higher K⁺ affinity (K_m = 2 μM) than Trk, and although its maximal transport rate is significantly lower than that of Trk, it enables the cells to grow in media with K⁺ concentrations of approximately 50 nM. The Kdp system consists of four subunits that build the functional KdpFABC complex. Synthesis of this high-affinity K⁺ transport complex is significantly induced (approximately 1,000-fold) if the external K⁺ concentration drops below 200 μM, so the primary role of this transporter is the scavenging of K⁺ under K⁺-limiting growth conditions. In addition, in the case of an osmotic upshock imposed with NaCl, the synthesis of the KdpFABC complex is induced to a lower extent (see below). The driving force for K⁺ uptake, even against a large opposing gradient, is ATP hydrolysis. Based on the ATP-hydrolyzing subunit KdpB, the KdpFABC complex can be grouped in the family of P-type ATPases, which undergo a conserved reaction cycle and can be inhibited by ortho-vanadate. However, the KdpFABC complex represents a unique member of this family, since the sites of K⁺ translocation and ATP hydrolysis are spatially separated on two different subunits, KdpB and KdpA, respectively. In this context, KdpA has sequence similarities to potassium channels (92). The third subunit of the complex, KdpC, is thought to be involved in cooperative nucleotide binding together with KdpB and, thus, acts as a catalytic chaperone (4, 112). The KdpF subunit comprises just one transmembrane span and stabilizes the KdpFABC complex assembly.

Kup is a minor K⁺ uptake system, which is so overshadowed by the Trk and Kdp systems that it can only be studied in kdp and trk deletion strains. The system consists of just one polypeptide. The V_max for net K⁺ uptake via Kup is relatively low, and its K_m is only slightly lower than that of Trk (296). Kup operates as a secondary porter using only Δμ̃_H+. Although it is not involved in the osmoregulatory response to elevated NaCl concentrations, it serves as the dominant K⁺ uptake system upon sugar-induced hyperosmotic stress at low pH values and nonlimiting K⁺ concentrations, i.e., conditions, under which Trk activity is strongly decreased and the Kdp system is not induced (296).

A relatively mild osmotic upshock suffices to stimulate K⁺ influx. The rate of influx is constant for upshock of 0.2 osM or greater (201, 246). The mechanisms by which the Trk and Kup systems sense and respond to osmotic upshocks have not yet been identified. However, the net K⁺ uptake by all these systems is electrogenic. Electroneutrality during rapid K⁺ uptake in case of osmotic upshock is maintained by the extrusion of protons (90, 194). The resulting alkalinization of the cytoplasm is reversed within approximately 10 min as organic anions accumulate (see below). Alkalinization of the cytoplasm after osmotic upshock must require K⁺ uptake since the cytoplasmic pH dropped when an upshock was imposed on cells in K⁺-free medium (194).

Control of kdpFABC expression.

Synthesis of the KdpFABC complex upon K⁺ limitation and, to a much lower extent, hyperosmotic stress is governed by a two-component sensor kinase/response regulator system. This mechanism is important for osmoregulation only in media of low K⁺ concentration, since kdpFABC is not expressed in the presence of moderate or high concentrations of K⁺. Expression of kdpFABC depends on the products of the kdpDE operon, which is downstream of and overlaps with the kdpFABC operon (231, 306).

The two regulatory proteins, KdpD and KdpE, are members of the so-called two-component family of sensor kinase, response regulator proteins (306). The sensor kinase KdpD is a homodimeric integral membrane protein of 99 kDa (256, 322). It can be functionally divided into an input domain (residues 1 to 660) and a transmitter domain (residues 661 to 894). The input domain comprises an unusually large cytoplasmic N-terminal domain (residues 1 to 395), four narrow-spaced transmembrane helices followed by an arginine-rich region and an additional 140 residues of the cytoplasmic C-terminal part. The C-terminal cytoplasmic transmitter region consists of a catalytic domain carrying all features characteristic of bacterial histidine kinases, including a conserved histidine residue (His 673), a dimerization domain, and several conserved motifs involved in the binding and hydrolysis of ATP (223). The input and transmitter domains are linked by a putative coiled-coil region. The 395-amino-acid-long cytoplasmic N-terminal domain is the largest among the N-terminal domains of bacterial sensor kinases. It contains a slightly modified Walker A and B motif (144) that was shown to bind 8-azido-ATP (124) and a Usp domain of the ATP-binding type (271).

Upon stimulus perception, KdpD autophosphorylates at conserved residue His673 in one subunit of the C-terminal catalytic domain, and then one monomer translocates that γ-phosphate of ATP to the corresponding histidine on the other subunit (303). Subsequently, the phosphoryl group is transferred in a kinase reaction to the cytoplasmic response regulator KdpE. KdpE~P dimerizes and serves as a positive transcriptional regulator for the genes of the kdpFABC operon (287). At normal growth conditions, the phospho-KdpE-specific phosphatase activity of KdpD prevails over its kinase activity (41) and, as a consequence, KdpE~P is dephosphorylated and dissociates from the promoter binding region, thereby terminating the expression of the kdpFABC operon.

The discovery that kdpFABC is an osmotically upregulated operon under the control of KdpD/KdpE led to the hypothesis that KdpD senses changes in turgor or a related parameter (165). As an integral membrane protein, KdpD could be affected by a change in the physicochemical state of the membrane (for example, strain or shape) related to turgor. At first glance, the induction of kdpFABC expression under potassium-limiting conditions fit well into this picture, since the level of intracellular potassium was expected to play an important role in the maintenance of turgor. However, it has so far not been demonstrated that turgor level changes under potassium-limiting conditions, since turgor can, at best, be estimated roughly, but not measured directly in gram-negative bacteria. The idea that KdpD senses changes in turgor was challenged by the fact that osmotic upshifts imposed with nonionic solutes like sugars did not elicit the same kdpFABC expression level as salts like NaCl at iso-osmolal concentrations (14). Therefore, it was suggested that KdpD somehow senses the cell's need for potassium to maintain turgor. Sugiura et al. (286) suggested mechanistically different sensing mechanisms for potassium-limiting conditions and osmotic stress. However, Zimmann et al. (323) showed that the extension of the fourth transmembrane helix encompassing the arginine cluster is involved in sensing both stimuli, which may not be separable. Furthermore, Heermann et al. (124) proposed an additional, more complex regulatory network for kdpFABC expression. Malli and Epstein (183) showed that the cytoplasmic potassium concentration is not the stimulus for KdpD. However, based on deletion mutant analysis of kdpD, Rothenbücher et al. (256) proposed recently that intracellular potassium is sensed by the C-terminal domain of KdpD. KdpD is expected to sense one or more intracellular parameters since kdpFABC expression does not correlate with external potassium concentration (98, 183) and a cytoplasmic KdpD derivative lacking all four transmembrane helices can activate some kdpFABC expression in response to K⁺ limitation (125). ATP could be such a candidate, since ATP binding to the N-terminal domain regulates the phosphatase activity of KdpD (145). Furthermore, a significant increase in the ATP level occurs in osmotically stressed cells (219), but other solutes involved in the osmotic stress response of E. coli (e.g., glutamate or compatible solutes) might also play a role in signal transduction elicited by potassium-limiting conditions (312; reviewed in reference 69).

Recently, real-time RT-PCR analyses clearly showed that expression of kdpFABC rises quickly and stays high as long as potassium-limiting conditions prevail (120). In contrast, kdpFABC is only transiently expressed under osmotic stress elicited by NaCl or sucrose, rising much more slowly and to a much lower level. This observation suggests that changes in turgor play a much smaller role than K⁺ limitation in stimulating KdpD. However, one has to keep in mind that the rates of nucleic acid and protein synthesis are known to be much lower immediately after osmotic stress is imposed (111, 143, 165). Since the synthesis of new mRNA is inhibited, it is probably not appropriate to normalize the levels of kdp mRNA by comparison with mRNA of a housekeeping gene. Therefore, the issue of onset and extent of kdpFABC expression under K⁺ limitation compared with that under osmotic stress can only be settled satisfactorily when appropriate controls for the determination of mRNA levels can be established. Similar problems arise when lacZ fusions are used to determine levels of gene expression under these totally different conditions.

To test more directly whether changes in turgor could be the stimulus for KdpD, Hamann et al. (120) compared changes in cytoplasmic volume under K⁺-limiting conditions with those resulting from salt and sugar stress. Surprisingly, a potassium downshift had little effect on the cytoplasmic volume of the cell, whereas salt or sugar stress elicited the expected transient decrease in cytoplasmic volume, indicative of a change in turgor. Furthermore, the internal K⁺ and ATP levels hardly changed under potassium downshift. Since K⁺ downshift doesn't decrease cytoplasmic volume, a reduction in turgor is probably not the stimulus for KdpD. The low level of kdpFABC expression under salt stress imposed by sodium chloride might be a secondary effect due to the inhibitory effect of sodium ions on the potassium uptake systems (W. Epstein, personal communication), thereby mimicking K⁺ limitation. Since the method of measuring the cytoplasmic volume is limited in sensitivity, it is still conceivable that small changes in turgor, for which no appropriate assay is so far available, might be sensed by KdpD.

How else could the cell's need for potassium be sensed? One possibility would be the interaction of KdpD with another protein (accessory sensor). The interaction of histidine kinases with transporters has already been shown for DcuS (sensor for fumarate) and DcuB (fumarate/succinate antiporter) (152). Furthermore, Steyn et al. (280) showed that, in Mycobacterium tuberculosis, the N-terminal domain of KdpD interacts with the lipoproteins LprF and LprJ. Although no homologous proteins have been found in E. coli, it is still conceivable that they have eluded identification due to low homology. In vitro transphosphorylation experiments between non-cognate sensor kinase/response regulator pairs in E. coli revealed no cross-talk partner for the C-terminal domain of KdpD (317).

So far, the major difficulty in establishing the stimulus for KdpD is that the proposed models are either inherently difficult to test, are not very specific, or have not been described in sufficient detail to allow for experimental tests.

Glutamate and other anions.

The most abundant osmotically regulated anion in enteric bacteria is glutamate (195, 293) (Fig. 1). Glutamate accumulates as the transient alkalinization of the cytoplasm resulting from K⁺ uptake dissipates (90, 293). When an increase in osmolality occurs in defined minimal medium, glutamate is synthesized in the cell. Glutamate accumulation starts at approximately 1 min after osmotic stress and is reduced by 90% when K⁺ uptake is blocked (194). In enteric bacteria, two enzymes mediate de novo synthesis of glutamate: glutamate dehydrogenase and glutamate synthase (chapter 3.6). Mutational loss of either enzyme does not alter the kinetics of glutamate accumulation following osmotic stress (38, 194). Since the flow of nitrogen through glutamate during growth is more rapid than is the accumulation of glutamate after osmotic stress, the accumulation of glutamate during osmotic stress could result from inhibition of its utilization for biosynthesis rather than from stimulation of its synthesis.

Mutants lacking glutamate synthase require high concentrations of NH₄⁺ for growth because glutamate dehydrogenase has a relatively high K_m for NH₄⁺ (chapter 3.6). When such mutants are grown in medium containing 1 mM NH₄⁺, they cannot generate high-glutamate pools and, therefore, are inhibited by high-medium osmolality (72). It was suggested that high-glutamate pools might have an additional function, besides serving as an osmotic solute, because this growth inhibition at high osmolality was not reversed by glycine betaine. The internal pH rises but then falls upon osmotic stress, whether glutamate is made or glutamate accumulation is blocked. Thus, some other acid must accumulate to account for the restoration of internal pH. In this context it is interesting to note that both γ-aminobutyrate and glutamate accumulate in osmotically stressed cells and their accumulation depends on external pH (218).

3-[N-morpholino]propanesulfonate (MOPS) (pictured in Fig. 1), a buffer commonly used in growth media, accumulates in E. coli at high osmolality (59) but does not act as an osmoprotectant (181). The levels of two other anions, glutathione (Fig. 1) and γ-glutamyl-glutamine, increase substantially upon a shift to high osmolality (193, 212). The growth of mutants unable to synthesize glutathione was restricted to minimal media with osmolarities below 1.5 OsM, but addition of glutathione allowed them to grow at osmolarities as high as 2 OsM (193). Neither the reason for the osmotic impairment of the glutathione-deficient mutant nor the role, if any, of γ-glutamyl-glutamine in osmotic adaptation is known. Upshock produces a significant increase in the concentration of ATP (120, 219) and of other phosphorylated metabolites, including nucleotides and two glycolytic intermediates, dihydroxyacetone-phosphate and 1,3-bisphosphoglycerate (D. McLaggan and W. Epstein, unpublished data).

Putrescine.

Putrescine is the major organic divalent cation in many bacteria (see Fig. 1). It is not expected to contribute much to cytoplasmic osmolality because most of it is probably bound to nucleic acids (53). Putrescine pools are approximately 50 mM in medium of low osmolality and fall to approximately 5 mM upon a shift to high osmolality as a result of efflux by a specific export system (53, 212). Pools of spermidine, the other divalent organic cation, are approximately 10 mM and are indifferent to medium osmolality. Like the accumulation of glutamate, the fall in putrescine requires K⁺ uptake. It has been suggested that putrescine bound to nucleic acids is displaced by K⁺ when K⁺ is present at high concentrations. Together with an increase in macromolecular crowding, this is proposed to maintain the kinetics and thermodynamics of protein-nucleic acid interactions in the range required for in vivo function (53, 54, 212). This view has been challenged by Schiller et al. (263) who demonstrated that glutamate synthesis together with putrescine efflux could not fully account for potassium uptake in response to hyperosmotic stress. However, the quantity of K⁺ required to replace putrescine as a countercation for fixed anions exceeds the quantity of K⁺ that accumulates with increasing medium osmolality as K⁺ glutamate (53, 194). Putrescine synthesis increases when most cell K⁺ is replaced with Na⁺, but putrescine does not accumulate because the excess is excreted (258). More fundamental work on polyamine transport is required, including identification of putrescine transporters involved in growth or osmoregulation.

Accumulation of Compatible Solutes

Potassium glutamate accumulation serves as a first response to cellular dehydration, but it does not restore bacterial growth to the rate observed in the absence of osmotic stress (57). Indeed, protein production is progressively inhibited as the concentrations of these solutes in the cytoplasm increase to very high levels. In fact, E. coli cells cultivated without organic osmoprotectants respond to an osmotic upshock by accumulating potassium glutamate, then replacing it with trehalose (α-D-glucopyranosyl-(1→1) α-D-glucopyranoside; see Fig. 1) over a period of 30 to 120 min (90). Mechanisms coordinating the use of trehalose as a carbon and energy source with its use as an osmolyte are reviewed elsewhere (70) (chapter 3.4.1). In brief, osmoregulatory trehalose synthesis from UDP-glucose and glucose 6-phosphate requires trehalose-6-phosphate synthase (OtsA) and trehalose-6-phosphate phosphatase (OtsB). It occurs slowly, at least in part because osmotic induction of rpoS must precede σ³⁸-mediated otsAB expression. Trehalose can also be used as a carbon and energy source by cells growing at low or high osmolality. At low osmolality, trehalose is taken up and phosphorylated by the phosphoenolpyruvate-dependent sugar phosphotransferase system, specifically trehalose-specific Enzyme IIBC^Tre (TreB) with Enzyme IIA^Glc as phosphate donor. It is then cleaved by trehalose-6-phosphate phosphatase (TreC) for further metabolism. At high osmolality a periplasmic trehalase (TreA/OsmA) is induced and treBC is repressed. Trehalose synthesized under osmotic stress enters the periplasm by an unknown mechanism to be hydrolyzed by TreA, then metabolized via Enzyme II^Glc. TreA thereby prevents the loss of a valuable nutrient.

E. coli and Salmonella can also respond to osmotic upshifts by accumulating organic zwitterions that are fully compatible with cell functions (often called osmolytes or compatible solutes) (Fig. 1). Increasing or high osmolality stimulates the accumulation of these compounds by modulating the activities of existing osmoregulatory proteins and by inducing gene expression (Fig. 2 and Table 3). Four transporters mediate osmoprotectant uptake: BetT, BetU, ProP, and ProU (Table 4). Neither BetT nor BetU is present in Salmonella and BetU is present only in some E. coli strains (see “E. coli and Salmonella in their natural environments,” below). Most osmoprotectants accumulate in the cytoplasm as compatible solutes, but some are converted to compatible solutes. The oxidation of osmoprotectant choline to compatible solute glycine betaine in E. coli was reviewed in the second (print) edition (70). In brief, choline is taken up primarily via transporter BetT. In the cytoplasm it is oxidized to glycine betaine by choline dehydrogenase (BetA, with oxygen as electron acceptor) and betaine aldehyde dehydrogenase (BetA and BetB, the latter with NAD⁺ as electron acceptor). Divergent, σ⁷⁰-mediated transcription of betT and the betIBA operon responds to the supplies of choline (via repressor BetI) and oxygen (via global regulator ArcA) as well as to osmotic pressure (166, 248). Thus, choline oxidation to glycine betaine contributes to both energy metabolism and osmotolerance under aerobic conditions.

Osmosensory transporters.

Osmotic upshifts stimulate osmoprotectant uptake by E. coli and Salmonella as they inhibit respiration and other transport (79, 132). Experiments and structural predictions demonstrate that, despite their functional similarity, osmoprotectant transporters do not constitute a distinct structural class. They represent diverse sequence families and use diverse energy sources (Table 4). ProP, a broad specificity H⁺-osmoprotectant symporter, is functionally related to LacY, a well characterized H⁺-lactose symporter (79). Experimental tests conducted to date support a homology model obtained by fitting the ProP sequence to the crystal structures of GlpT and LacY (81, 178, 315) (Fig. 5A). BetT (choline-specific) and BetU (betaine-specific) have 35% sequence identity with Na⁺-betaine symporter BetP of Corynebacterium glutamicum (BetPCg). The crystal structure of BetPCg is remarkably similar to the structures of Na⁺ symporters from diverse sequence families (245). BetT comprises 12 transmembrane segments with cytoplasmic N and C termini (295) as predicted for BetPCg. ProU comprises periplasmic binding protein ProX, integral membrane protein ProW, and ATP-binding cassette protein ProV. Crystal structures for proline betaine- and glycine betaine-liganded ProX (261) are serving as prototypes for crystallographic analyses of related proteins from other species. They reveal that cation-pi interactions are a major determinant of high-affinity osmolyte binding. Although proline and ectoine are low-affinity substrates for ProU in vivo, ProX does not bind proline or ectoine in vitro (18, 118, 139). This suggests that the ProU system can mediate ProX-independent proline and ectoine uptake.

Fig. 4Osmoregulation of transporter activity. (A) LacY activity decreases as ProP activity increases in membrane vesicles under osmotic stress. Membrane vesicles containing both LacY and ProP were prepared, then proline (circles) and lactose (triangles) uptake activities were measured in K phosphate supplemented with sucrose to adjust the osmolality (and hence, the water activity). (B) Impacts of ΔΨ on the osmotic activation of ProP in proteoliposomes. The impact of osmolality on ProP-His6 activity in proteoliposomes was determined with ΔpH fixed at 1 unit (−59 mV) as ΔΨ varied from 0 to −137 mV. The initial rate of proline uptake via ProP-His6 is plotted versus osmolality, with ΔΨ at −137 mV (filled circles), −120 mV (squares), −100 mV (triangles), and 0 (open circles). The initial rate of proline uptake at high osmolality increases with ΔΨ because the proton motive force powers H⁺-proline symport. In addition, the initial rate of proline uptake at low osmolality decreases as the magnitude of ΔΨ increases, showing that ΔΨ is required for the osmoregulation of ProP activity.

Source: These figures were reproduced with permission from Culham et al. (2008), Biochemistry47:8176–8185, ©2008, American Chemical Society.

Fig. 5Structure-function analysis of osmosensory transporter ProP. ProP is a member of the Major Facilitator Superfamily (77) that catalyzes osmolality-dependent H⁺-osmoprotectant symport (181). (A) ProP structure. ProP is a 500-residue protein with 12 transmembrane α-helices, a cytoplasmic N terminus, and an extended, cytoplasmic C-terminal domain (315). The illustrations show a homology model for residues 4 to 236 and 246 to 452 of ProP (left, Protein Data Bank [PDB] ID code 1Y8S [315]) and an NMR structure of the homodimeric, antiparallel α-helical coiled-coil formed by a peptide replica of ProP residues 468 to 497 (right, PDB ID code 1R48 [131, 325]). The structures of the intervening peptides (residues 237 to 245 and 453 to 467) are unknown. Residues in ribbon format are yellow (nonpolar), green (polar), red (acidic), or blue (basic). Primes denote residues in the second copy of the homodimeric coiled-coil. The homology model (left) is based on the crystal structure of anion antiporter GlpT (PDB ID code 1PW4). ProP and other symporters are believed to function via an alternating access mechanism in which substrate binding sites alternately face the periplasm and cytoplasm (114, 171). Similarly to the GlpT structure, the ProP model would represent the “inward facing” conformation of the transporter. The black oval encircles the cluster of intramembrane polar residues within the N-terminal helix bundle that may be involved in H⁺-osmoprotectant symport. No polar residues are found within the intramembrane portion of the C-terminal helix bundle in this model. Single-cysteine variants facilitate structure-function analysis of ProP (76, 81, 130, 178). A bifunctional alkylating reagent covalently cross-linked dimers of ProP variants with cysteine only at position 419, 420, 422, 439, or 480 in vivo. Variants with cysteine at other positions, including 133, 241, or 473, were not cross-linked. This and other evidence indicates that antiparallel coiled-coils link ProP monomers in vivo (76, 130, 178). Arg 488 (left, space-filling representation with CPK color), embedded in the hydrophobic coiled-coil interface, stabilizes the antiparallel structure by extending to the hydrophilic surface to interact with acidic residues of the opposite strand (80, 299, 325). (B) ProP function. The homology model (315) is used to illustrate the locations of functionally important residues in ProP. Alkylating agents have been used to probe the solvent exposure of cysteines in single-cysteine ProP variants (81, 178, 315). Like ProP activity, the reactivities of residues C59 (periplasmic loop 1) and C415-C418 (periplasmic loop 6) increased with osmolality. The reactivities of other periplasmic residues did not. Amino acid replacements Y44M, S62C, and Y423C render ProP activity osmolality-insensitive (81, 178, 315). Replacement Y44M also rendered alkylation of C59 or C415 osmolality-insensitive, as did dissipation of the proton motive force (81). According to the structural model, Y44 caps the cluster of polar residues within the N-terminal bundle of ProP that may be involved in H⁺-osmoprotectant symport (ball-and-stick structures encircled by black oval).

Osmotic stress activates each of the osmoprotectant transporters, but ProP is likely to serve as the first responder because it is ubiquitous, expressed by cells cultivated at low osmotic pressure and broad in substrate specificity (Table 3). BetT is present only in E. coli and BetU is present in only some E. coli isolates. BetT and BetU are narrow in substrate specificity and BetT is present only when choline is supplied under aerobic conditions. ProU is ubiquitous and broad in substrate specificity, but it is present only when cells are cultivated at high osmolality. ProP senses and responds to osmolality changes in a relatively narrow range around the optimal osmolality for bacterial growth (e.g., a range of approximately 0.2 mol/kg, after growth in MOPS medium for which the growth rate is optimal at 0.2 mol/kg) (75). Thus, cells may rely on ProP to quickly address limited osmolality fluctuations due to metabolic activity and environmental change; other systems may be invoked to address more extreme conditions or circumstances in which ATP is available, but Δμ̃_H+ and Δμ̃_Na+ are limited.

Three osmosensing transporters are being investigated intensively to elucidate the mechanisms of osmosensing and the osmoregulation of transporter activity: ProP of E. coli, BetP of C. glutamicum (159) (BetPCg) and OpuA of Lactococcus lactis (233) (OpuALl). BetPCg is related to BetT and BetU (as noted above), whereas OpuALl is related to ProU. Subunit OpuAALl corresponds with ProU subunit ProV, whereas subunit OpuABCLl corresponds with ProU subunits ProW and ProX. OpuAALl is 50% identical in sequence to E. coli ProV, whereas the relevant regions of OpuABCLl are 47% identical to E. coli ProW and 21% identical to E. coli ProX. BetPCg and OpuALl originate from gram-positive organisms that differ from E. coli and Salmonella in both cell wall structure and phospholipid composition. However, their properties are outlined below because the corresponding systems from E. coli and Salmonella have not been examined intensively.

Increasing osmotic pressure alters many cellular properties (discussed above). In principle, an osmosensory transporter could respond to osmotic pressure directly or it could respond to a secondary effect of osmotic stress on cell structure or composition (312). Secondary effects could include changes to turgor pressure, membrane properties (including the lateral pressure profile, macroscopic strain, thickness, headgroup charge density, hydrogen bonding and hydration, and lateral or intermonolayer distribution of phospholipid species) and cytoplasmic properties (including composition, ionic strength, and macromolecular crowding) (312).

The rate of osmoprotectant uptake via ProP increases with medium osmolality in intact E. coli, cytoplasmic membrane vesicles (bacterial membrane ghosts), and proteoliposomes reconstituted with the purified protein (206, 239) (Fig. 4). BetPCg and OpuALl are osmotically activated in E. coli cells and in proteoliposomes (159), and the osmotic activation of ProPCg in C. glutamicum has also been characterized (158). ProP and BetPCg activate within seconds after an osmotic upshift, although activation can be delayed if respiration is also inhibited (103, 299). Diverse, membrane-impermeant solutes activate all three systems to the same extent when they raise the osmotic pressure of the external medium to the same level (233). Thus, all respond to changes in extracellular osmotic pressure, not to a particular extracellular solute. Each can act, without other proteins, to detect an increase in osmotic pressure and respond by transporting osmoprotectants. Turgor pressure is not essential for osmosensing or the osmoregulation of these transporters (159, 233).

Proteoliposomes have been used to further explore the requirements for osmosensing. ProP and BetPCg reconstitute with a predominantly “outside out” orientation and a functionally “outside out” orientation can be imposed on OpuALl by supplying ATP in the proteoliposome lumen; so, in each case, the proteoliposome lumen corresponds to the cytoplasm of an intact cell. Each transporter can be activated by concentrating luminal electrolytes without altering membrane strain or turgor pressure (159, 233). Each fails to activate when low-molecular-weight organic compounds are concentrated in the proteoliposome lumen, attaining an osmolality that would activate transport if the same, membrane-impermeant compounds were provided externally and electrolytes were concentrated luminally (159, 233). Further proteoliposome studies suggest that BetPCg and OpuALl respond to osmotically induced changes in luminal K⁺ concentration (259, 262, 265) and ionic strength (26, 182), respectively. In these cases the semipermeable membrane would transduce the osmotic signal, and the resulting cytoplasmic property would act on the transporter's cytoplasmic surface to influence its function.

K⁺, Na⁺, Li⁺, and Cs⁺ chlorides are equally effective activators of ProP in proteoliposomes (75). Although a K⁺ requirement for ProP activation was observed in cells (157, 181), increasing the osmolality activated ProP in K⁺-free membrane vesicles. Furthermore, the activities of ProP and LacY, the latter not an osmoregulatory protein, were reduced by similar fractions at each osmolality when Na⁺ replaced K⁺ (79). Although low-molecular-weight organic compounds did not activate ProP, this transporter was activated when poly(ethyleneglycol)s were concentrated in the proteoliposome lumen (75). Thus, ProP activation does not require concentration of a specific electrolyte (e.g., K⁺) in the proteoliposome lumen, and ProP activity may depend on ProP hydration and/or macromolecular crowding (75, 313). ProP activity is osmolality insensitive when the magnitude of the membrane potential (ΔΨ) falls to approximately 100 mV, and it is suppressed at low osmolality only when the magnitude of ΔΨ exceeds approximately 120 mV (79). Thus, ΔΨ contributes to the driving force for osmoprotectant transport via ProP (Δμ̃_H+) and works with the osmolality to regulate ProP activity (Fig. 4).

If the properties of OpuALl represent those of ProU and the properties of BetPCg represent those of BetT and BetU, the data summarized above suggest that the osmosensory transporters of E. coli and Salmonella respond to different, osmotically induced changes in cytoplasmic composition (i.e., K⁺ concentration [BetT, BetU], ionic strength [ProU], macromolecular crowding and/or competition for water of hydration [ProP]) (313). These hypotheses should be tested via direct comparison of ProP, BetT, BetU, and ProU.

There is intense interest in the structural features that lead ProP, BetP, and OpuA to activate as other transporters become inactive when osmolality increases (132, 313). Sequence comparisons identified the extended, C-terminal domains of these proteins as putative osmosensors (27, 77, 228). The cytoplasmic C-terminal domain of E. coli ProP (residues 438 to 500) is approximately 55 residues longer than those of the closest E. coli paralogues that are not osmosensors (ShiA and KgtP) (178, 233, 315). This C-terminal extension terminates in a 30-residue coiled-coil domain of known structure (80, 325) (Fig. 5A). Homodimeric, antiparallel coiled-coils form between peptide replicas of the ProP C terminus (131, 325) and link intact ProP molecules in vivo (80, 130, 178, 299). However, some ProP orthologues lack the coiled-coil sequence (298) and at least one of these orthologues is an osmosensory transporter (C. glutamicum ProP [229, 298]). Disruption of the coiled-coil raises the osmolality at which ProP from E. coli or Agrobacterium tumefaciens becomes active but does not render them insensitive to osmolality- (298, 299). Recent evidence suggests that the coiled-coil contributes to colocalization of ProP with cardiolipin at E. coli cell poles (250) and modulates the osmolality range over which ProP can become active (251, 298) (see “Cell structure and osmotic adaptation,” below). Thus, the C-terminal domain of ProP functions in osmotic adaptation but is not an osmosensor, per se.

Where is the osmosensor in ProP? Amino acid substitutions that render ProP partly or fully osmolality insensitive do not define a localized osmosensing “site” (81, 178, 315). Recently, Culham et al. showed that increasing osmolality enhanced the N-ethylmaleimide (NEM) reactivity of cysteine residues that replaced certain periplasmic residues in ProP (81) (Fig. 5B). The NEM reactivities of these Cys became less osmolality dependent in the absence of Δμ̃_H+ and in ProP variants with osmolality-insensitive activity. NEM permeates the cell wall and cytoplasmic membrane of E. coli and reacts with cysteines only if they are accessible and hydrated (147). Thus, these experiments provide the first evidence for a conformational change associated with osmosensing by ProP. Increasing osmolality could influence ProP activity by altering the hydration of the transporter's membrane-integral domain and/or by shifting the equilibrium between ProP conformers with cytoplasm- and periplasm-facing substrate binding sites. Structural data show that, despite their membrane integration, MFS members like LacY and ProP are flexible and highly hydrated with more extensive cytoplasmic than periplasmic surface exposure (2, 3, 178). These proteins are likely to share an alternating access mechanism in which two helix bundles rock around an axis in the membrane plane, opening the substrate binding site sequentially to the periplasm and cytoplasm (114). In LacY, lactose-coupled H⁺ transport likely involves water molecules associated with key acidic and basic residues (3). Water serves as a cofactor for some H⁺ transport reactions as illustrated by bacteriorhodopsin, an osmotic pressure-sensitive, light-driven H⁺ pump (52,168). Most proteins are inhibited as they dehydrate with increasing osmotic pressure. In contrast, dehydration may eliminate competing reactions to create tightly coupled H⁺ transport via ProP.

The C-terminal domains of BetPCg and OpuALl are believed to serve as a K⁺ and an ionic strength sensor, respectively. Residues 541 to 595 constitute the C-terminal domain of BetPCg (265) and Glu572 is critical for its K⁺-dependent osmotic activation (264). Residues 555 to 572 are predicted to form an α-helix and have been shown to interact with the N-terminal domain and Loop 8 (221). However, the C-terminal sequence of BetPCg is not similar to the longer C termini of BetT and BetU, and Glu572 is not conserved. In addition, C-terminal deletions have different effects on the osmoregulation of BetPCg and BetT (228, 295). In the case of OpuALl, ionic strength sensing is ascribed to dual cystathionine-β-synthase (CBS) domains located near the C termini of the cytoplasmic ATP-binding OpuAALl subunits (26). Both ionic strength and an anionic C-terminal tail are proposed to modulate electrostatic interactions of the CBS domains with the membrane surface (26). ATP-binding subunit ProV of the E. coli ProU system also has dual C-terminal CBS domains but lacks the anionic extension present on OpuAALl.

Osmotic induction of transcription: proP and proU.

Transcription of genes encoding osmoprotectant transporters is upregulated following osmotic upshifts (17, 63, 308). ProP is present in bacteria cultivated at low osmolality, and osmotic induction increases proP transcription approximately 5-fold. In contrast, proU is expressed only when bacteria are cultivated in high-osmolality media and proU transcription is subject to more than 100-fold induction. Transcription of proP and the proU operon apparently reaches a maximum after a few minutes (17), which might be thought of as a second phase of regulation following the rapid inhibition of ribosomal RNA transcription, described above (111) (Fig. 3). This second phase produces transporters, increasing the cell's capacity to accumulate osmoprotectants available in the external environment. Expression of proP and proU relies on the same transcription machinery used before osmotic upshift (196, 197) and delays associated with the production of new transcription factors are thus avoided. The third wave of expression largely relies on the production of σ³⁸ (RpoS) (308), which accumulates roughly coordinately with proU and proP transcription (17) (Fig. 3). RpoS-dependent genes encode a wide variety of proteins. As discussed above, all of these effects are reversed as the cell adapts to its new environment. Many studies have compared transcription in cells fully adapted to growth in media of different osmolalities. Such results will not reveal changes that occur transiently during osmotic adaptation.

The proU operon can be transcribed from ^multiple promoters, but the most downstream of these is selectively responsive to osmotic upshift (196). In vivo footprinting shows that RNA polymerases are poised to transcribe proU but are largely held silent prior to osmotic shock (213). This situation is formally similar to the poised RNA polymerase that exists at the osmY promoter with two differences: osmY is occupied by σ³⁸ RNA polymerase and has opened the DNA (169, 253), whereas proU is occupied by σ⁷⁰ RNA polymerase and the DNA remains closed (39, 142, 213). In the proU case, as for osmY (and for proP, below), no macromolecular activators of transcription have been detected.

The RNA polymerase poised at proU is held in check by the apparently simultaneous binding of protein H-NS. H-NS binds cooperatively to regions both upstream and downstream of the transcription start site (39, 142, 213). This is believed to form a repression loop that keeps proU transcription inactive (39, 213). Upon potassium glutamate accumulation, H-NS dissociates, allowing transcription to occur (39, 213). However an hns (osmZ) defect alters the level of proU expression without affecting osmotic control (129). The reported effects of changes in DNA topology on proU transcription (129) may be explained by the influence of supercoiling on DNA loop formation (39, 213). The increase in DNA supercoiling that occurs upon osmotic upshift (129) could also play a role.

Transcription of proP can occur from two promoters, and transcription from the upstream promoter is selectively stimulated upon osmotic upshift (197). Still, no gene-specific macromolecular activators have been detected, although the genome-wide activator Fis can activate in vitro (300). Transcription relies on the σ⁷⁰ form of RNA polymerase. In this case the RNA polymerase is not poised but rather needs to be recruited to the promoter. Recruitment prior to osmotic upshift is prevented by the binding of protein Crp to a region required for promoter recognition (316). Upon potassium glutamate accumulation Crp is released, freeing the promoter for RNA polymerase recognition and transcription (167). The use of Crp would suggest some involvement of external glucose in regulation, but this has not been investigated.

Expression of proP may also be controlled at the translational level. Mutations at proQ, which is not linked to proP, diminish ProP activity approximately 5-fold without altering its dependence on osmolality (163, 208, 278). Early studies identified no effect of proQ lesions on proP expression (163, 208), but recently proQ lesions were found to decrease ProP protein levels (M. N. Smith and J. M. Wood, unpublished data). Amino acid sequence analyses predict that the ProQ protein comprises a FinO-like N-terminal domain linked to an Hfq-like C-terminal domain by an unstructured linker (275, 276; Smith and Wood, unpublished data). ProQ and its putative domains were purified so the sequence-based predictions could be tested. As expected, ProQ is a soluble protein, the linker is more susceptible to tryptic proteolysis than the N- or C-terminal domain, and the N- and C-terminal domains are composed predominantly of α-helices and β-sheets, respectively (275, 276). FinO and Hfq are both RNA-binding translational regulators; FinO is a predominantly α-helical protein implicated in F-pilus biogenesis (13) and Hfq is a predominantly β-sheet protein implicated in diverse processes including regulation of RpoS (301). ProQ was also among ribosome-associated E. coli proteins found during a proteomic screen (140). These observations have led to the hypothesis that ProQ may be an RNA-binding translational regulator.

Overall, the transcriptional induction of osmosensory transporters ProU and ProP appears to occur when DNA-binding proteins are released as potassium glutamate accumulates in response to osmotic stress. Induction can be quite rapid because potassium glutamate accumulates rapidly and the response relies on common, pre-existing transcription factors. The role of the ProQ protein remains to be delineated.

RESPONSES TO OSMOTIC DOWNSHOCKS

Bacteria that have previously adjusted to high osmolality through the accumulation of salts (predominantly K⁺ glutamate) or compatible solutes (trehalose, proline, betaine, ectoine, hydroxyectoine, and peptides) may encounter severe turgor stress as they pass into environments of lower osmolality (24, 210, 266). The turgor stress arises from the inrush of water associated with increased transmembrane osmotic gradient that is simply a consequence of the dilution of the surrounding medium. Pressure can rise by up to 10 atmospheres in milliseconds (22). The principal mechanism of relieving this stress is the activation of mechanosensitive (MS) channels (24, 173).

In vitro studies have shown that MS channels sense the state of the lipids and, in particular, deformation of the membrane that can be imposed either via a sudden increase in transmembrane pressure or the asymmetric insertion of small amphipaths into the two leaflets of the lipid bilayer (187, 188). Amphipaths are small organic molecules that dissolve in the lipid bilayer but have a slow rate of passage from one leaflet to the other, such that they exist at different concentrations in the two leaflets of the bilayer. This asymmetry modifies the bilayer tension leading to channel activation. These two different events cause increased tension in the lipid bilayer (288, 291), and MS channels undergo a conformational change that transiently creates protein-lined pores in the membrane. The pores have large diameters (~8 to 35 Å, MscK being the smallest and MscL the largest) (67, 288) and, in broad terms, are nondiscriminating in the solutes that they pass other than on the basis of solute size (67).

In E. coli and Salmonella three different classes of channel activity are defined by their sensitivity to tension, their conductance, and their structure. The channel classes are called MscM (mini), MscS (small), and MscL (large), where the names in brackets refer to the magnitude of their conductance. This definition is based on the distinguishable electrical characteristics of the channels assayed by patch-clamp electrophysiology and does not disclose the genetic diversity of the structural genes (173). To date, MscM has proved to be the most elusive channel to characterize. Its conductance is quite small and it is not consistently present in membrane patches, which is the “gold standard” for defining MS channels. No gene has been shown to encode the structural protein; consequently, little is known. In contrast, MscL is known to be the product of a single gene (mscL, 74.06 min on the E. coli K-12 genetic map) (290). The structure of the closed form of this protein from M. tuberculosis has been deduced by X-ray crystallography at 3.5 Å (62, 279) and the channel has been subjected to very detailed genetic and biophysical analysis that has resulted in a good working model for the transition of the channel from the closed to the open state (11, 19, 20, 22, 29, 30, 31, 137, 138, 176, 189, 190, 191, 192, 225, 226, 227, 289). The MscS activity is now known to comprise at least two channels: MscS (yggB; 66.11 min) and MscK (kefA; 10.47 min) (173). However, there are two further homologues for both the kefA (ybiO, 18.16 min; yjeP, 94 min) and yggB (ybdG, 13 min; ynaI, 30 min) genes. The precise functions of these homologues are at present unknown; however, several of them are known to be expressed in a regulated fashion and at least one, YbdG, has physiological activity (U. Schumann, M. D. Edwards, W. Barlett, S. Miller, and I. R. Booth, unpublished data).

The two most important channels for survival of osmotic downshock by E. coli are MscL and MscS (173). However, minor roles can be detected for MscK and YbdG when MscL and MscS have been deleted. This assessment is based on the ability to create single, double, and multiple mutants of E. coli K-12 and assay their ability to survive hypoosmotic shock. Thus, deletion of any one of the channels does not affect the ability to survive a large hypoosmotic shock challenge. In contrast, double MscS, MscL deletions exhibit a 70 to 95% loss in colony-forming ability after hypoosmotic shock. Additional mutations in MscK and YbdG decrease survival to <1% and change the threshold osmolality step at which loss of viability is observed (Schumann et al., unpublished). Therefore, these two channels play ancillary roles in protection, whereas survival is mainly predicated on the activity of MscL and MscS. This observation correlates with the relatively greater abundance of MscL and MscS in cells, compared with MscK and YbdG, but also with the ease with which the former, but not the latter, are gated (opened under stress) in membrane patches. MscK activity is frequently observed in patches derived from E. coli strains lacking MscS (173). In contrast, YbdG activity can be observed in cells by use of physiological assays, but has not been consistently detected by electrophysiology (Schumann et al., unpublished).

Mechanosensitive channel proteins occur at relatively low abundance in the cell membrane. Calculations based on the surface area of a cell, the surface area of a patch, and the number of channels observed to open simultaneously under pressure suggest that as few as 5 MscL channels (25 protein subunits) and ~30 to 40 MscS channels are present per cell. This is supported by the observation that very low expression of the channels from plasmids (that observed in the absence of an inducer) is sufficient to protect from downshock (96, 173, 204, 205). Moreover, cells are tolerant of the expression of mutant forms of the channels that gate at low pressure when the expression is either from the chromosomal locus or from the uninduced state of the plasmid (44). In contrast, induced expression of the same mutant is often lethal (222). These observations have two significant aspects. First, the conductance of these channels is so high that relatively few channels can cause quantitative release of the solute pools of the cells in a few seconds (279). Second, the actual phenotype seen with cloned channels is the product of the expression level and the activity of the specific channel. This can be misleading—a channel that has suffered a significant loss of activity can still protect cells against hypoosmotic shock when expressed at a sufficiently high level. The assay is principally qualitative, although, with care, quantitative analysis can be performed (35).

Each channel exhibits a unique threshold for activation by tension, with MscM opening at the lowest tensions, followed “sequentially” by MscK, MscS, and MscL (174). Demonstrations of this sequential gating in vivo are limited. However, Batiza and colleagues (22) reported that a MscL L19C mutant channel exhibits a significantly higher probability of gating in response to small downshock steps in the absence of MscS than in its presence. The implication is that the buildup of membrane tension is normally offset by MscS (and MscK) activity. In vivo function of the channels remains one of the most complex areas of this field, since the turgor pressure across the membrane is reported to be 4 atm (210), whereas MscS and MscL gate in vitro at membrane tensions that are predicted to occur in cells with turgor pressures of approximately 0.1 atm (288, 291). A dynamic interplay occurs between the resistance offered by the cell wall (peptidoglycan and lipopolysaccharide) and the outwardly directed turgor so that the net tension on the bilayer is less than that needed to activate the channels. This may explain the high frequency of “osmotic sensitivity” phenotypes among mutations that affect the structural integrity of the cell wall.

MscL

The MscL crystal structure is a homopentamer of 136 amino acid subunits (62, 279) (Fig. 6). A minor difficulty in understanding all of the data for MscL is the fact that the crystal structure was obtained with a M. tuberculosis homologue that does not gate effectively in E. coli membranes (211), but most of the genetic and biochemical analysis has been performed with the E. coli protein. A deep understanding of the basic mechanism of the channel has been achieved despite such problems. The channel has the largest conductance of all characterized MS channels (~3 nS) and opens in response to tension close to that at which the membrane ruptures (291). A large pore size (34 to 46 Å) for the open MscL channel has been estimated from the conductance and measurements of permeability to oligomeric molecules. The open channel was found to pass 1,1′-bis-(3-(1′-methyl-(4,4′-bipyridinium)-1-yl)-propyl)-4,4′-bipyridinium ions that have a size of ~30 Å, but conductance was impaired by poly(l-lysine) molecules >37 Å. Thus, by a variety of measurements an open channel of ~40-Å diameter is considered likely (67, 291). The MscL channel has been proposed to allow the passage of small proteins, but this remains controversial (25). Despite this, biotechnological applications for delivery of small peptides and organic molecules have been developed (104, 105).

Fig. 6Mechanosensitive channels MscL and MscS. (A) Crystal structures of the closed (nonconducting) forms of MscL (a) and MscS (b) from Mycobacterium tuberculosis (62, 279) and Escherichia coli (21, 279). Color coding: (a) MscL: amino-terminal helix (red), TM1 (blue), TM2 (yellow), carboxy-terminal helix (green). (b) MscS: TM1-TM2 (yellow), pore-lining helix TM3a (blue), tensor helix TM3b (yellow), and carboxy-terminal domain (green). Although the color coding implies functional similarities between the domains sharing the same color in the two proteins, this cannot be used too literally. Thus, although TM2 of MscL and TM1-TM2 of MscS are the major lipid facing domains, mutations that affect the response to changes in lateral tension are found in both TM1 and TM2 of MscL. (B) Model for the structural transition for MscS. The model is based on the crystal structures of MscS in the nonconducting (a) (21, 279) and the conducting state (307) (b). The color coding of the helices is as in panel A. A cartoon representation of the movements required to attain the open, conducting state (c) shows that the TM3a helices rotate clockwise (as viewed from the periplasm) and move out from the axis of the pore as these helices straighten and separate.

Each subunit has two transmembrane (TM) helices and also contributes an amino-terminal segment and a carboxy-terminal helix to the pentamer (62). The TM1 helices cross the membrane at an angle of ~30° to the membrane normal and converge close to the cytoplasmic face of the membrane, forming an inverted tepee, meeting at Val23 in E. coli (Fig. 7). The five Val23 residues form the principal component of a hydrophobic seal that keeps the structure closed. Formation of the large ~30-Å pore in the open state requires the separation of the TM1 helices (see below). The TM2 helices primarily interact with lipid and with TM1, thus acting as the frame for the pore-forming helices (62). Considerable genetic analysis has elucidated both the importance of the seal and the significance of residues that pack the helices together around the seal region. In a key development, Blount and colleagues demonstrated that hydrophilic substitutions at, or close to, the hydrophobic seal modify the gating behavior of the channel such that the open probability at low tension is considerably increased (222, 318, 319). Such mutations confer on the channel the ability to inhibit growth when the channel protein is expressed and assembled into the membrane. The most severe of these mutations affected the residues that constitute the seal or modify the packing of the subunits around the seal (Val23, Gly22, Gly26, and Gly30). Introduction of smaller (Gly, Ala) or hydrophilic (Asp, Asn, Arg, Ser) residues at these positions generated channels with extremely low energy barriers for the closed to open structural transition. Many other changes to the amino acid sequence, some quite subtle, significantly modified gating. The N15D mutation is particularly interesting. Expressed from the chromosome, MscL-N15D was identified as a suppresser of K⁺ transport defects (44), but induced expression of this protein from a plasmid was profoundly growth inhibitory (222). This reinforces the observation that expression level interacts with activity to produce the observed phenotype.

Fig. 7Positions of critical residues in MscL (a, b) and MscS (c, d). Two views are shown for each of the channels: the side view and the view from the periplasm down the axis of the channel. Some of the residues that are critical for function are highlighted. (a, b) MscL Val21 (equivalent of Val23 in MscLEc [MscL from E. coli]), is blue; the positions equivalent to MscLEc residues Gly24, Gly26, and Gly30 are red. Residues that, when mutated to Asn or to Trp, cause loss of channel function, which is believed to be the result of altered interaction with lipid headgroups (64, 320), are green (TM1) or yellow (TM2). (c) A side view of MscSEc is shown in the closed state in which the seal residues (L105 and L109) are blue. The critical Gly residues (101, 104, and 108), which are central to the closed-open transition, are orange, and Ala98 and Ala106 are green. The Ala102 residue that is critical for the stable open state of the channel is shown in vermillion. Note that in this representation TM1-TM2 of one subunit has been removed to allow sight of the interactions between Gly and Ala in the TM3a helices. The residues forming the carboxy-terminal domain have been removed for clarity. (d) MscSEc viewed from the periplasm. The missing TM1-TM2 segment has been restored and is pink to enable the viewer to see that TM3b of one subunit (pink) interacts with TM1–2 (silver) of the next subunit (clockwise). The other five subunits are shown in dark gray. Note that in MscL the amino terminus of the pore-lining TM1 helix is cytoplasmic, whereas in MscS the amino terminus of the pore-lining TM3a helix is periplasmic.

The transition from the closed to the open state has been characterized by the introduction of single Cys residues and the application of electron paramagnetic resonance (EPR) spectroscopy using site-specific nitroxide spin labels (225, 226, 227). These studies allowed the analysis of the relative roles of hydrophobic mismatch and membrane tension in gating. In response to reconstitution of MscL into lipid bilayers constituted from short-chain lipids (C-10, dicapryl-sn-3-phosphorylcholine) the structure of the protein was perturbed, but the channel did not show a significant propensity to gate spontaneously (227). Changing the lipid modified the gating pressure consistent with conformational modification as the complex changed shape to maintain interactions with the shorter lipids, but large-scale conformational changes were not observed. In contrast, insertion of lysophosphatidyl choline (LPC), which inserts into the outer exposed leaflet of a liposome bilayer and is only slowly translocated to the inner leaflet, provoked large-scale changes in protein organization (225). The TM1 helices had separated and the spin probes attached to them increased in mobility, consistent with fewer spatial constraints. Moreover, spin labels attached to Cys residues close to the hydrophobic seal exhibited increased exposure to an aqueous probe (Ni ethylenediamine-N,N′-diacetate) consistent with separation of the helices at this point. The closed to open transition is thought to involve the rotation and tilting of TM1 helices such that they lie at a very acute angle relative to the membrane normal. The final effect of the tilting motion is to break the seal created by the five Val23 residues and thereby create a large-diameter pore.

Critical elements of the interaction with the lipids are also being identified (64, 234, 235, 236, 320). The rim of the channel protein, where it contacts lipid headgroups, is enriched in hydrophobic amino acids, lacks residues that exhibit strong hydrogen-bonding potential through their side chains, and has few of the Trp and Tyr residues that are frequently found at these positions in other membrane proteins. Trp and Tyr residues are often considered to anchor proteins to the lipid head groups to prevent excessive conformational change (150, 304). Mutations in either TM1 or TM2 that introduced either Trp or Asn strengthened the interaction with lipid headgroups and increased the pressure that is needed to open the channel (64, 320). The correlation is that strong interactions with the lipid headgroups would prevent movements of the helices that are required during the closed to open transition.

The principal movements of the TM1 helices to create the open pore are well established. Other regions of the protein play critical roles in either stability or gating but their precise roles are poorly understood. In particular, the roles of the N- and C-terminal α-helices are not well understood (279). It is possible that the C-terminal helical bundle forms a mesh through which solutes exiting through the channel must pass to leave the cell (11). However, it is not clear how this fits with the measurements of pore diameter using large hydrophilic solutes. More recently, a proposed role for the N-terminal helices emerged on revision and extension of the MscL structure (279). With the use of data refinement tools the α-helices have been located in the M. tuberculosis crystal structure and are proposed to lie in the plane of the membrane and to make contact with the second TM1 helix (viewed from above) (279). It may be that these helices act as a restraint on the structural transition that occurs during opening to ensure that the channel can revert to the closed state without unfolding (see below for MscS).

MscS

MscS is a much more complex protein than MscL (21, 204) (Fig. 6). The E. coli protein has been crystallized in two states: native, proposed to reflect the closed or nonconducting state (21), and a mutant A106V, proposed to trap an open state (307). In both crystal structures the protein is a homoheptamer of 286 amino acid subunits. In contrast to MscL, MscS has three TM segments: TM1 and TM2 form the tension sensor, whereas TM3 is in two parts, TM3a and TM3b. TM3a is an extremely hydrophobic helix and lines the pore, while TM3b is amphipathic and lies against the bilayer surface. TM3b probably stabilizes the open state and guides a smooth open to closed transition (307). In the native state the seven TM3 helices are packed very closely in a predominantly symmetrical array, such that there is always a visible “pore” of approximately 6 Ådiameter (305). This apparent pore led to the original proposal that this was the “open” state (21). Subsequent molecular dynamics and biophysical analysis reversed this view and led to the proposal that this “pore” was nonconducting because of a “vapor lock”—a situation that arises in very narrow hydrophobic pores when the surface cannot be wetted and, therefore, no water column is filling the pore (12). Consequently, ion conduction is prevented. The narrowest point of the pore, considered to be the seal, is provided by two rings of residues, Leu105 and Leu109, which lie close to the cytoplasmic end of the channel (21) (Fig. 7). Substitution of these residues with either a small amino acid (Ala, Gly) or a hydrophilic amino acid (Ser) creates channels that gate at low pressure (204). The transition to the open state, therefore, must involve helix separation to create a pore with an ~14-Å diameter (based on conductance measurements) (288).

The carboxy-terminal domains of MscS and MscL differ in complexity. Whereas MscL has a “simple” bundle of helices, MscS has a complex multidomain fold that creates a large cytoplasmic vestibule perforated by seven lateral portals (21). These portals are critical since all of the solutes that leave the cell during hypoosmotic shock must pass through the portal to gain access to the pore. The portals may impose kinetic limitations on ion movements through the channel. Deletions that remove the extreme carboxy-terminal residues and the base of the vestibule increase the conductivity of the channel (95). The cytoplasmic domain is required for maximum stability of the MscS channel and may also allow a moderate degree of ion and solute specificity (95, 268). MscS from E. coli exhibits a slight preference for anions, but MscK is not ion selective (175), whereas a homologue from Methanococcus jannaschii is slightly cation selective (153). The MscS-A106V crystal structure has an essentially unchanged structure for the cytoplasmic domain, which contradicts earlier evidence for substantial movements during gating of the channel (307).

A critical feature of the TM3a packing in the pore is the conserved pattern of Ala and Gly residues that facilitates rotational movements of the helices during gating by creating complementary surfaces on adjacent helices (21, 96). Adjacent TM3a helices are as closely packed as is feasible while retaining symmetry, and the Leu105 and Leu109 side chains extend into the lumen of the pore. Replacement of the Ala with Gly residues decreases the tension required to gate the channel, and conversely, introduction of Ala residues in place of Gly increases the pressure for gating (174, 204). The crystal structures of MscS-A106V and the native protein differ in two remarkable ways (307). In the former, the TM3a helices have rotated and separated. Arising from these changes, the only protein-protein contacts between adjacent TM3a helices are at the lower (A110-L115 interaction) and upper extremities. The diameter of the channel pore is now ~14 Å and this is sufficiently large to account for the known conductance of the channel (1 to 1.25 nS). This large diameter is achieved by TM3a helix separation coupled with rotation that moves the Leu105 and Leu109 side chains away from the center of the pore.

Comparison of the two MscS structures also reveals three other fundamental aspects of the gating mechanism (307). First, TM1–2 moves as a rigid body that has rotated so that it is now at an acute angle to the membrane plane. The effect of this rotation is to change the conformation of the linker between TM2 and TM3a, which provokes the initial rotation of TM3a to start the gating movement (307). Second, the structure of TM3b shows no significant change, in contrast to recent hypotheses based on genetic analysis (5, 10). The tension associated with the rotation of TM3a is absorbed by a change in the bond angle at the “bend” between TM3a and TM3b. The rigidity of the TM3b structure prevents the further movement or separation of the TM3a helices, thus stabilizing the open state. Finally, the new structure informs our understanding of channel inactivation. Unlike MscL, MscS displays at least three states: closed, open, and inactivated. Inactivation occurs when MscS channels are opened under submaximal tension in patch-clamp analysis (5, 6, 95). The new structure suggests that the path from open to closed includes intermediates that can misfold, yielding inactivated channels. This is undoubtedly a complex phenomenon that is still poorly understood.

As with MscL much remains to be determined about the mechanism of tension sensing and the interactions with lipids. The crystal structure of MscS-A106V, taken to represent the closed channel, suggests that TM1 and TM2 interact with lipids since they occupy a peripheral position relative to TM3a. Recent studies suggest parallels with MscL (217, 320). MscS channels became harder to gate when hydrophobic residues close to the region that would interact with the phospholipid headgroups were substituted with Asn (217). These data are consistent with transmission of tension through TM1-TM2 to the TM3a pore-lining helices that then rotate and straighten to generate the open pore. Consistent with this, mutations that make MscS easier to gate have been found to map to the TM2-TM3 linker peptide (204). Moreover, the earliest mutations in MscS, V40K, and V40D, which caused the channel to gate at low tension (220), were also substitutions of hydrophilic residues into the region just below the zone where Asn insertion was found to be an effective blocking mutation. Val40 is ideally placed, when mutated, to stabilize an intermediate structure that is poised toward the open state as a consequence of hydrogen bonding to lipid headgroups. Thus, the precise location within TM1-TM2 of residues that can interact with phospholipid headgroups is a major determinant of tension sensing by fixing the conformation so that the channel is either poised to gate or prevented from gating.

In conclusion, we now have considerable insight into the mechanisms of gating of MS channels, in particular, for MscS and MscL, and how they help bacteria to survive hypoosmotic shock. These channels may also play a role in normal vegetative growth, and their activities may change as lipid composition is modified. These are the areas that will, in the future, provide much needed advances in understanding of the physiology of osmoregulating cells.

PERSPECTIVES

Much has been learned about the impact of osmotic stress on E. coli and Salmonella, and about their responses to osmotic stress. However, many important questions remain to be answered. What are the immediate and long-term impacts of small and large osmotic shifts on cell chemistry and physics, on the transcriptome and the proteome? Why do some physiological processes slow or cease, while others accelerate, in response to osmotic shifts? What molecular mechanisms enable proteins to sense osmotic pressure, linking osmosensing to the osmotic activation of osmosensing transporters and mechanosensitive channels? How do osmoregulatory responses evolve over time after an osmotic shift and how does that program depend on the magnitude of the shift. How are responses to osmotic stress related to variations in other physicochemical and metabolic parameters, including temperature, pressure, pH, and aeration? What is the physiological rationale for osmolyte selection? What roles do osmoregulatory mechanisms play in wild-type E. coli and Salmonella, within and during transmission between their mammalian hosts.

Cell Structure and Osmotic Adaptation

Until recently, technical limitations prevented researchers from determining how osmotic stress changes cell structure and organization. Biophysical and imaging techniques are now beginning to reveal interdependent changes to bacterial cell structure and chemistry in response to both osmotic shifts and long-term exposure to media of varying osmolality (156, 250, 302). This progress occurs despite technical challenges associated with the small size of E. coli and Salmonella cells. The transcriptomes and proteomes of these bacteria are altered dramatically by osmotic shifts and by long-term cultivation in media with varying osmolalities (63, 116). Many genes and proteins of unknown function have been associated with the osmotic stress response in this way. However, osmotic stress affects multiple cell structures, processes, and properties, including the growth rate. Thus, it is not clear which of these genes and proteins mitigate the effects of osmotic stress on bacteria and which have other functions. For example, two-component regulatory system EnvZ/OmpR is triggered by osmolality changes and other stimuli to influence transcription of many genes, including ompF and ompC (237). Glycine betaine must cross the outer membrane of E. coli via OmpF and OmpC since ompF and ompC defect slowed glycine betaine uptake when it was supplied at a very low concentration (0.7 μM) (102). However, there is no evidence that osmoregulation of porin gene expression influences the ability of E. coli to use glycine betaine as an osmoprotectant.

In contrast, osmoregulation of membrane phospholipid composition clearly influences the osmotic stress response of E. coli (250, 251, 298). As noted above, the mole fraction of cardiolipin (CL) increases at the expense of phosphatidylethanolamine (PE), while the mole fraction of phosphatidylglycerol (PG) remains constant, as E. coli is cultivated in media of increasing osmolality (298). In addition, the PG content of CL-deficient (cls) bacteria rises in response to osmotic stress (250). Transporters ProP and LacY are concentrated with CL (and not PG) near the cell poles and septa, yet the polar concentration of ProP, but not LacY, is CL-dependent (250, 251). The osmolality required to activate ProP is proportional to the CL content of wild-type bacteria, the PG content of CL-deficient bacteria, and the anionic lipid content of cells and proteoliposomes. CL is effective at a lower concentration in cells than in proteoliposomes, and at a much lower concentration than PG in either system. Thus, in wild-type bacteria, osmotic induction of CL synthesis and concentration of ProP with CL at the cell poles adjust the osmotic activation threshold of ProP to match ambient conditions. In some ProP orthologues, this “tuning” also involves the C-terminal coiled-coil (251, 298).

Osmosensory and Osmoregulatory Mechanisms

Osmoregulatory processes accelerate in response to osmotic upshifts as other cellular processes are attenuated. For example, the transcription of ribosomal RNA required for cell growth is initially inhibited while transcription of genes encoding osmoprotectant transporters (proP and proU) is activated^. The osmotic activation of osmosensory transporter ProP occurs in stark contrast to the osmotic inhibition of its paralogue, LacY (Fig. 4, top). Mechanistic insights can be gained by comparing analogous processes, like solute-H⁺ symport via ProP and LacY, that respond in opposite ways to the same osmotic stimulus (79), or the remarkably different responses to potassium glutamate accumulation by RNA polymerases that are bound to different sigma factors. Most studies of osmoregulation have used bacteria supported by aerobic respiration, which in turn generates the transmembrane electrochemical potential gradients and ATP required to power osmoregulatory solute uptake. Yet osmotic upshifts inhibit respiration, and this inhibition is exacerbated by K⁺ deficiency (79, 199). It would be helpful to know how osmotic upshifts inhibit respiration and how K⁺ protects the respiratory system from osmotic inhibition.

Most studies discussed above delineate events that occur within minutes of an osmotic upshift or downshift. Through the application of biochemical tools, sensor kinase KdpD, osmosensing transporter ProP, and mechanosensitive channels MscL and MscS from E. coli have become paradigms for the study of osmosensing (146, 313, 314), and orthologues of osmosensing transporters BetT/BetU (C. glutamicum BetP) and ProU (L. lactis OpuA) have attained similar status (159, 233, 313). The next phase of this work, already well underway in the case of MscL and MscS, will include detailed structure-function analysis and the application of biophysical tools to delineate events that occur on shorter time scales. Its aim is to correlate physical and chemical changes to the periplasm, cytoplasmic membrane, and cytoplasm with changes in conformation and function for each protein. Precise delineation of the cellular properties sensed and controlled by each protein remains an important goal of this work.

Osmoregulatory Programs and Solute Selection

Studies outlined above indicate that K glutamate accumulation is central to osmoregulation and, in particular, to transcriptional reprogramming in response to osmotic stress. It is tempting to conclude that K glutamate accumulation must precede other osmoregulatory responses and that K glutamate is an osmoregulatory “second messenger” (37). Yet, bacteria prefer to accumulate compatible solutes like glycine betaine rather than K glutamate, compatible solutes and glutamate are expected to affect protein structures and interactions differently, and compatible solutes are more effective than K glutamate as accelerators of bacterial growth (57). The uptake of osmoprotectants like glycine betaine depends on transporter ProP, which is present in bacteria cultivated at low osmolality, without K glutamate accumulation, and active in the absence of K⁺ glutamate (79). In fact, K glutamate accumulation in response to a large osmotic upshift was attenuated (although not eliminated) when bacteria were provided with ProP substrate proline or glycine betaine (90). Thus, alternative osmoregulatory programs, contingent on the osmoprotectant supply, are available to E. coli and Salmonella. However, the mechanisms balancing K glutamate and compatible solute accumulation, in particular, in response to osmotic upshifts of various magnitudes and durations, are not fully understood.

Diverse organic compounds serve as osmoprotectants and/or compatible solutes for Bacteria and Archaea (106, 230, 249, 260). E. coli and Salmonella must select these compounds on the basis of their availability (see below). However, the osmolytes have varying abilities to stabilize cytoplasmic macromolecules, in particular, proteins (33). Some can attenuate, others exacerbate, the effects of membrane-permeant solute urea, high or low temperature, and oxidative stress on bacteria. For example, trehalose can confer osmotic stress tolerance, thermotolerance, and oxidative stress tolerance (23, 49, 126). Further, bacterial pH homeostasis and salinity tolerance mechanisms must be intertwined. Thus, it is possible that bacteria select osmolytes to address multiple properties of their environment and, hence, their cytoplasm and periplasm.

E. coli and Salmonella in Their Natural Environments

Genomic sequencing is revealing extensive genetic diversity among naturally occurring E. coli, Shigella, and Salmonella isolates (chapter 6.3). Most osmotic stress tolerance mechanisms discussed above are ubiquitous in E. coli (Table 3) so they are believed to be encoded by a core E. coli genome. Locus betTIBA is present in E. coli but not Salmonella. Genes encoding isolate-specific mechanisms also exist and appear to have arisen via lateral gene transfer as exemplified by trkG (which resides in the cryptic rac prophage) and betU (which is flanked by insertion sequences) (180). TrkG is present in only two-thirds of E. coli isolates, but this does not impair osmoregulatory K⁺ uptake, because TrkG is redundant in function to TrkH (78). The identification and properties of BetU are discussed further below.

It seems obvious that osmoregulatory mechanisms must contribute to the survival and growth of E. coli and Salmonella within, and during, transitions between their mammalian hosts. Particular attention has been paid to osmoregulation by uropathogenic E. coli, the agent of most ascending urinary tract infections (141), because fluctuating osmolality distinguishes the urinary tract from other mammalian tissues. The osmolality of human urine normally varies in the range 0.5 to 0.8 mol/kg, but it can vary from 0.04 to 1.4 mol/kg (160, 255). Glycine betaine is used as an osmoprotectant by kidney medullary cells (43) and, hence, is consistently available to bacteria growing in mammalian urinary tracts (172). The availability of other osmoprotectants depends on diet. Glycine betaine and proline betaine were identified as urinary factors that permit E. coli to grow in high-osmolality media (60), leading to the identification of osmoprotectant transporters as potential antibiotic targets and analysis of their substrate specificities (61).

Mixed inorganic salts are major contributors to the osmolality of human urine, but urea (approximately 0.5 M) is the largest individual contributor. Urea is membrane permeant for E. coli and does not activate osmoregulatory responses (78). Thus, high-osmolality human urine (1 mol/kg) imposes moderate urea stress and moderate osmotic stress (equivalent toapproximately 0.5 mol/kg) on E. coli. Examining 301 E. coli isolates from blood, urine, or stool and 12 representative enteric strains, Kunin et al. found no correlation between the maximal salinity for growth in the absence of osmoprotectants and site of origin (162). However, poor ability to use glycine betaine as an osmoprotectant was correlated with poor growth in hypertonic urine. The sensitivity of E. coli to toxic proline analogues increases with osmotic stress, but decreased under anaerobic conditions (86, 244). This is an important observation since urine and the renal medulla are low in oxygen.

Analysis of the osmotic stress response was extended to uropathogenic E. coli via characterization of representative pyelonephritis isolates HU734 and CFT073 (73, 74, 78). HU734 was used to establish a murine model for an ascending urinary tract infection. CFT073 is both more virulent and more similar to other urinary tract infection isolates, and its genomic sequence is available. HU734 encodes a defective RpoS variant, abrogating biosynthetic trehalose accumulation via system OtsAB (78, 82) and undoubtedly affecting the expression of other osmoregulatory loci. This may explain why E. coli K-12 and isolate CFT073 grow at higher osmolalities in minimal medium devoid of osmoprotectants than HU734.

Surprisingly, deletion of genes proP and proU from HU734 compromised growth in high-osmolality human urine without eliminating glycine betaine uptake activity, whereas deletion of those loci from an rpoS-deficient CFT073 derivative eliminated glycine betaine uptake activity but did not retard growth in high-osmolality urine (73, 78). BetU, a new betaine-specific osmolyte transporter, is responsible for the residual glycine betaine uptake activity in HU734 and absent from CFT073 (180). Locus betU is present in one-third of E. coli isolates, but its distribution provides no evidence of selection during the evolution of virulence (180). Unknown osmoprotectants may be taken up via unknown transporters to enhance the growth of CFT073 in high-osmolality urine.

Despite the impact of betaine supply and osmoregulatory defects on bacterial growth in urine, rpoS, proP, and/or proU defects did not significantly influence the number of bacteria recovered from the bladders or kidneys of mice inoculated with E. coli HU734, CFT073, or their derivatives (73, 78). However, these experiments entail transurethral inoculation of mice with a dense bacterial suspension (approximately 10⁹ bacteria/ml), bypassing any need for the bacteria to proliferate during ascent and colonization of the bladder and kidneys. Characterization of the remaining osmoregulatory mechanisms in uropathogenic E. coli will further elucidate the role of osmotic adaptation in urinary tract infection. It will also pave the way toward understanding their roles in survival, transmission, growth, and pathogenesis of commensal E. coli and strains with other pathotypes. Food scientists frequently examine the impact of osmotic pressure on the survival of food- and water-borne pathogens, but the contributions of known osmoregulatory mechanisms to the survival of E. coli and Salmonella in their natural environments are not yet known.

ACKNOWLEDGMENTS

We thank Doreen E Culham, Charles Deutch, Tatyana Romantsov, Michelle Smith, John Foster, and an anonymous reviewer for helpful comments on this chapter.