Bacterial Ion Channels

EMMA L. R. COMPTON AND JOSEPH A. MINDELL*

[SECTION EDITOR: PETER MALONEY]

Posted 01 June, 2010

Membrane Transport Biophysics Unit, Porter Neuroscience Research Center, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD 20892

*Corresponding author. Mailing address: 35 Convent Drive, Bldg. 35, MSC 3701, Bethesda, MD 20892. Phone: (301) 402–3473, Fax: (301) 480–1693, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

For many years bacteria and other prokaryotes were thought to have extremely “tight” plasma membranes, with only minimal permeabilities to most ions, to facilitate the large voltage and pH differences needed for oxidative respiration. The idea that a bacterium might have an ion channel, a high-turnover passageway for gradient-dissipating salts, was a heretical thought. Bacterial ion channels were known, but only in special cases, such as outer membrane porins in Escherichia coli and bacterial toxins that form pores in their target (bacterial or mammalian) membranes. Early hints of more pervasive bacterial channel formation came from the work of Kung and colleagues, who developed methods for patch clamping bacteria and discovered several classes of channels involved in osmoregulation. Ultimately, the exhaustive coverage provided by a decade of bacterial genome sequencing has revealed that ion channels are actually widespread in bacteria, with homologs of a broad range of mammalian channel proteins coded throughout the bacterial and archaeal kingdoms. In this latter case, even though many of these proteins have been reported and they have generated a frenzy of activity among structural biologists, extremely little is known about their contributions to bacterial physiology.

Ion channels are subtly but critically distinct from the transporters usually thought of in bacterial membranes. Both groups of proteins catalyze the transmembrane movement of hydrophilic (and thus membrane-impermeant) molecules and ions; however, channels form a continuous aqueous pathway across the membrane, whereas transporters alternately expose substrate-binding sites to one side or the other. Substrates move through channels at rates limited by diffusion, often as high as 10⁷/s. Movement though transporters, however, is limited by the need for some kind of conformational change to occur, allowing rates of usually only 10¹ to 10³/s. Ion channels in bacterial plasma membranes must therefore be highly regulated—a single channel is sufficient to rapidly deplete the gradients needed for respiration (see “Toxic channels [colicin],” below). When open, channels are purely dissipative—they can only allow ions to move down their electrochemical gradients. Perhaps because having channels is “risky” to a bacterium's integrity they are comparatively uncommon. Nevertheless, various channels do play important roles; here, we will discuss four groups of bacterial channels: porins, mechanosensitive channels, channel-forming toxins, and bacterial homologs of mammalian channels.

PORINS

The outer membrane (OM) of gram-negative bacteria forms an impermeable protective barrier between the cell and its environment; this barrier is especially important for enteric bacteria because the gut is an inhospitable environment, containing digestive enzymes and bile salts. The OM also forms an obstruction to antibiotics and host defense factors (86). However, while clearly beneficial to cells, the OM also blocks access of essential nutrients; to survive, the cell needs to provide a mechanism for nutrients to penetrate the OM. Porin channels provide this access by forming large, nonspecific aqueous pores in the OM that allow ions and vital nutrients to cross it and enter the periplasm. Porins’ importance is underlined by their presence as major proteins in the OM and by their occurrence in all gram-negative bacteria. Extensive reviews of the porin literature are available (11, 30, 52, 59, 84, 86, 97, 109).

The first indication of the presence of pores in the OM was the finding that the outer membrane could behave as a “molecular sieve,” filtering solutes largely on the basis of size. Thus, a wide range of radiolabeled molecules, including both natural (like ions, di- and trisaccharides) and nonnative (like polyethylene glycols) solutes, with molecular masses ≤600 Da can diffuse through the OM of Salmonella enterica serovar Typhimurium, while larger solutes cannot (28, 80). Whole OM, but not its nonprotein constituents (purified phospholipid or lipopolysaccharides), reconstituted into membrane vesicles retains similar permeability profiles, indicating that a protein is necessary for this process (80). Three proteins responsible for these behaviors were identified in serovar Typhimurium by following this permeability activity in liposome-reconstituted OM fractions (81). The physiological importance of these proteins was demonstrated by the use of porin-deficient mutants that showed a variety of phenotypes reflecting their impaired outer membrane permeability, including decreased abilities to utilize metabolites (67) and reduced uptake of 5′-AMP (9). The term “porin” was proposed to describe the nonspecific diffusion channels (pores) formed by these proteins in the outer membrane (81).

Solutes pass through porin channels by passive diffusion and are excluded primarily by their size; they do not accumulate in the periplasmic space because active transport processes rapidly pump them into the cytoplasm (86). In contrast to these “nonselective holes,” another group of “specific” porins selectively pass particular larger solutes (e.g., LamB, which is specific for maltose and maltodextrins) although they remain broadly selective for smaller solutes (52). Both porin groups are structurally similar, sharing the same β-barrel fold. Here we will only discuss those proteins for which the term was initially intended: the nonspecific porins. These so-called classical porins have been studied thoroughly; they compose a highly homologous family of E. coli proteins known as OmpC, OmpF, and PhoE.

Although proteins conferring solute permeability on the outer membrane might be expected to express constitutively, porin gene expression is regulated by several mechanisms (reviewed in reference 97). OmpC and OmpF in E. coli are expressed under normal culture conditions. Although OmpF expression is under tighter control, the two are regulated in a roughly reciprocal fashion; OmpC production is favored and OmpF repressed by anaerobic conditions as well as high osmotic strength and high temperature (those found in the gut). OmpF production, however, is favored and OmpC repressed under the opposite circumstances: low osmotic strength, low temperature, and aerobic conditions (those found perhaps in open water). The regulation of expression by osmolarity is controlled by the EnvZ-OmpR two-component regulatory system. The mRNA interfering complementary RNA, micF, is involved in the response to temperature and oxidative stress and may also act as a “fine tuner” in the control of OmpF by osmolarity (40, 97). OmpF expression has also been shown to be repressed by the presence of environmental chemicals, such as salicylate, weak acids, and antibiotics via the XylS-AraC family of regulatory proteins and again by micF. RpoS has also been shown to play a role in OmpF regulation, with a decrease in synthesis of OmpF upon entry to the stationary phase of growth. The complex regulation of OmpC and, in particular, OmpF makes sense when the pore size of the two proteins is considered—OmpC has a slightly narrower pore than OmpF (1.1 nm and 1.2 nm, respectively) (85). In the gut, where nutrients are plentiful, but so are toxins, a large pore is unnecessary and a potential weakness; E. coli mutants with mutations in the ompF gene are easily selected for by use of β-lactam antibiotics (42). There are many examples where loss of a porin increases antiobiotic resistance (87). In open water, however, there are fewer toxins and the larger pore of OmpF is probably beneficial in gathering the scarcer nutrients. PhoE is expressed under low-phosphate conditions because it is a member of the phosphate regulon, a large group of genes (potentially up to 400) controlled by the PhoR/PhoB two-component regulatory system. The phosphate regulon controls the phosphate homeostasis of the cell, sensing environmental levels of inorganic phosphate, and has recently been shown to be involved in bacterial virulence (63). Thus, by the careful control at multiple levels of porin expression and function, the cells attempt a perfect balance of nutrient gathering with defense against the outside world.

Structure/Mechanism

A plethora of techniques have been used to assess the permeability properties of porins, including studies of cell growth, measurements of radioactive solute flux, assays of swelling in liposomes (11, 87), and examinations of electrophysiological properties (30). While confirming that these proteins are largely nonselective, these experiments have revealed subtleties in porin permeability properties: OmpC and OmpF prefer cations, while PhoE prefers anions (12, 13, 62, 85); OmpF allows slightly larger solutes through than OmpC (87). In addition to these details regarding permeability, electrophysiological approaches have also led to other insights regarding porin function.

Initially porins were thought to be merely ubiquitously open “holes” through the membrane without the ability to close (13, 62). However, other studies (using gentler purification procedures) showed that porins actually open and close in a complex fashion and can be inhibited by certain compounds. In patch-clamp experiments they display a basal open state with fast extra channel openings and two types of closings that are slower (30): one lasting a few milliseconds and a longer “inactivated” state that is possibly irreversible (15, 32). The entry into the closed states is faster at higher voltages and the threshold voltage can be modulated by pH (30). The closures are also cooperative, both within a trimer (see below) and within a population; in planar lipid bilayers, channels are often seen to close in threes, possibly representing successive closures of the monomers of a trimer (33), but in patch-clamp recordings any number of closures occur concurrently (15, 32). Although the subject of much discussion and some contention in the literature (31, 87), this voltage-dependent gating has not yet been shown to have any physiological relevance. It has been suggested that it may serve as protection against the channels misincorporating into the cytoplasmic membrane (87), but no data have yet been provided to prove this. Periplasmic membrane-derived oligosaccharides (MDOs) and polyamines, such as cadaverine, stabilize the closed state (29). Secretion of cadaverine in response to low pH causes a decrease in the number of open porins, thus decreasing the permeability of the OM (102). This form of closure, by molecules that naturally occur in bacteria, seems much more likely to have relevance in vivo.

Structurally, porins are very unusual membrane proteins (Fig. 1a). This was immediately apparent when their amino acid sequences were determined (48, 78, 91). Porins do not contain stretches of hydrophobic residues long enough to form the characteristic transmembrane α-helices expected of membrane proteins (at one time, it was thought that only α-helices could cross the membrane [52]). However, circular dichroism and infrared spectroscopy experiments revealed that porins instead contain β-sheets (101). The determination of a porin structure (from Rhodobacter capsulatus) using X-ray diffraction (124, 125) solved the mystery: each monomer consists of an antiparallel β-sheet wrapping upon itself to form a “β-barrel.” Thus, the membrane-spanning regions are β-strands, not α-helices. Through each β-barrel monomer a central hole forms the conducting aqueous pore. The X-ray structures of OmpF and PhoE from E. coli confirm that they too form trimeric β-barrels (26) (Fig. 1a). This β-barrel fold is not confined to porins; it is now known to be a characteristic of outer membrane proteins in general. The number of β-strands varies for other outer membrane proteins, but is 16 for porins. These strands are joined by short turns on the periplasmic face of the membrane and longer loops on the cell surface. The loops have high sequence variability, as they rapidly mutate in response to interactions with antibodies, phages, and colicins (87), and are usually mobile (108). One loop, L3, is essential to porin function. It folds back inside the barrel constricting its waist to form an “eyelet.” The charges in this loop and on the barrel wall combined with steric effects determine size and valence preferences of the individual porins and help them exclude nonpolar molecules. The porins are timeric complexes; the interfaces between porin monomers composing a trimer are stabilized by close packing of hydrophobic residues along the walls of the monomers and loop L2, which bends over the adjacent monomer aiding the stability of the complex (25).

Fig. 1Structures of a porin and mechanosensitive channels. Views along (left) and perpendicular (right) to the plane of the lipid bilayer are shown, colored by subunit. (a) OmpF porin (PDB ID code 2zfg) (128) is shown. Loop L2 (red) stabilizes the trimer and loop L3 (orange) constricts the pore and contributes to determining ion selectivity. (b) MscL (PDB ID code 2oar) (20) is shown in the closed state. Residues that occlude the permeation pathway are shown as blue and red space-filling representations. (c) MscS (PDB ID code 2vv5) (123) shown in the open state. Residues lining the narrow part of the transmembrane pore are shown as red and blue space-filling representations.

Overall, porins are not the simple, ubiquitously open pores they once seemed but are channels capable of exhibiting complex behavior. Their expression and function is modulated at multiple levels to allow rapid, transient responses as well as slower more stable effects. They are an example of the elegance of the systems bacteria use to balance gathering of nutrients with survival in inhospitable and unpredictable environments.

MECHANOSENSITIVE CHANNELS

The conventional view of ion channels dictates that they open and close in response to the composition of the intra- and/or extracellular milieu, be it binding ligands or sensing changes in membrane potential, all the while ignorant of the sea of lipid surrounding them. Upon hypoosmotic shock (when transferred from a medium of high osmolarity to one of low osmolarity) cells rapidly swell because of an influx of water; without adaptive measures they would burst. Mechanosensitive (MS) channels act as “emergency release valves,” allowing solutes to rapidly exit the cytoplasm and to dissipate the large osmotic disparity between the internal and external environments. MS channels are remarkable in that they do this by responding to forces exerted by the membrane itself. Reviews of MS channels can be found in references 24, 41, 69, 70, and 90.

In notably the first demonstration of bacterial ion channel activity in vivo, ion channel currents were elicited from E. coli giant spheroplasts by application of slight hydrostatic pressure upon the membrane, by sucking or blowing through the pipette during electrophysiological recordings (72). These large currents represent channels opening in response to the pressure applied to the membrane. So far, these currents have been attributed to four proteins in E. coli, grouped by their single-channel conductance into three families: MscL, large (~3 nS) (14); MscS/MscK, small/kalium, i.e., potassium (~1 nS) (64, 74, 116); and MscM, mini (~0.15 nS) (14). Although all respond to similar stimuli, the proteins themselves are turning out to be very different entities.

Indirect evidence for the involvement of MS channels in E. coli's survival of hypoosmotic shock was provided when it was shown that Gadolinium, a known MS channel blocker (129) also blocks the release of solutes from the cytoplasm during hypoosmotic stress [16]). Definitive evidence came when double-knockout E. coli cells, lacking both MscL (gene mscl) and MscS (YggB), show a loss of viability upon hypoosmotic shock. However, single mutants with MscL, MscS, or MscK knocked out individually, double mutants with MscL and MscK or MscS and MscK knocked out, or the wild-type (WT) cells were unaffected (64). Hence, MscL and MscS together are essential for the protection of the cell against hypoosmotic shock but MscK and, by inference, MscM are not, at least under these experimental conditions.

MS channels reside in the plasma membrane (17). When cells are transferred into a solution of lower osmolarity, water rushes into the cell down its osmotic gradient. The subsequent rapid increase in the turgor pressure of the cell causes the membrane to stretch; this in turn causes lateral pressure changes within the membrane that activate MS channels (41). This must occur on a very rapid timescale to save the cells from lysis, too fast for de novo gene expression to accomplish. Instead, a reserve of quiescent channels is needed. When E. coli are grown in high-osmolarity media (i.e., when the likelihood of hypoosmotic shock increases) they enhance their level of MS channel expression, an effective preparation in case an osmotic challenge occurs (115). An increase in expression is also seen upon entry to the stationary growth phase. The expression of MscL and MscS is regulated by stress-sigma factor (RpoS); rpoS knockout cells show a greater sensitivity to hypoosmotic shock in stationary phase than WT cells (115). It has also been suggested that MscS may play a role in bacterial cell division because mutations of various residues doubled the number of cells (after isopropyl-β-D-thiogalactopyranoside induction) compared with WT growth (121). Hence, MS channels may play a role not only in the cells’ response to extreme conditions, but also in ensuring the structural integrity of the cells during “normal” growth.

MS channels open in response to changes in the tension in the bilayer rather than sensing an osmotic gradient. This was demonstrated by using molecules, such as amphipaths (molecules both hydrophilic and hydrophobic in nature) and lysophosphatidtylcholine, which can insert into one or other leaflet of the bilayer causing the whole bilayer to curve and altering the membrane tension in the absence of an osmotic gradient. The resulting changes in membrane tension indeed cause MS channels to open, proving that mechanical forces in the lipid bilayer itself are responsible for MS channel gating (71, 95). Computational studies have sought to explain how these changes in the bilayer forces are translated into protein conformational changes (reviewed in reference 24). They suggest that, at least in silico, it may be a combination of the weak, transient Van der Waals forces between the protein and lipid combined with hydrogen bonding between the protein and lipid head groups, but as yet there is no clear, definitive answer.

Subtle differences in the activation and function of MS channels have been discovered. The conductances of the channels correlate with the pressure needed for them to open, MscM opening at the lowest pressures and MscL at the highest (14). MscS and MscK were initially thought to be one protein; however, two gene loci, YggB and KefA, have been identified as responsible for MscS activity. MscK activity can be distinguished from MscS because it does not inactivate upon continuous suction as MscS does (64). MscK, uniquely among MS channels, shows a sensitivity to the extracellular environment, requiring high external concentrations of K⁺, NH₄⁺, Rb⁺, or Cs⁺ to open (65). MscM and MscS/MscK have slow gating kinetics, remaining open for long periods of time, whereas MscL has much faster kinetics, opening only for brief bursts. All the channels close as soon as the pressure is removed (14, 117). MscS is also sensitive to voltage. Voltage and pressure together increase the open probability of the channel; as the voltage increases, it opens at lower pressures. The converse is also true, at higher pressures less voltage is needed to open more channels. At high voltages MscS actually closes (72). Thus, the bacterium can carefully tune its response to be appropriate to the degree of osmotic shock it experiences, so as not to unnecessarily waste precious solutes.

Structure/Mechanism

Although in electrophysiological studies MS channels appeared to be similar to each other, as soon as their genes were cloned, it became apparent that they were, in fact, very different molecules. There is no sequence homology between MscL and MscS/MscK. MscL forms a 136-amino-acid protein (116); MscS and MscK form 286- and 1,120-amino-acid proteins, respectively (64). While MscK is much larger than MscS, it contains an MscS-like sequence. So far, MscM has not been isolated. A major breakthrough in understanding these differences came with the publication of the X-ray diffraction structures of an MscL homolog (20) and MscS (8). It was perhaps no surprise that, while these proteins have clearly different structures, they share some generalities. Both are predominantly α-helical homomultimers (MscL, a homopentamer; MscS, a homoheptamer) with transmembrane α-helices surrounding a central pore and a C-terminal cytoplasmic domain. Both proteins have constrictions inside the pore that are thought to act as channel gates. The cytoplasmic domains may act as size exclusion filters, stopping certain essential solutes from leaving the cytoplasm when the channel is open (24). The MscL C-terminal cytoplasmic domain shows some pH sensitivity and has effects on channel gating (58). The N-terminal region of MscL also resides in the cytoplasm and may act as a second gate for the channel (24). The N terminus of MscS is located on the periplasmic side of the membrane suggesting that the extra bulk of MscK will reside in the periplasm, perhaps even interacting with the outer membrane (74).

The first MS channel structure to be published was that of a MscL homolog (37% sequence identity with E. coli MscL) from Mycobacterium tuberculosis in a closed conformation (20) (Fig. 1b). A few years later the structure of E. coli MscS was published, causing some controversy (8). It was unclear as to the conductance state of the protein—was it open, closed, or somewhere in between? Some of the confusion has recently cleared up with the publication of MscS in the open state (123) (Fig. 1c), and combined electron paramagnetic resonance spectroscopy (EPR) and molecular dynamics studies have proposed structures for both the open and closed states (120, 121). It appears that the original structure was somewhere in between, perhaps representing an inactivated or desensitized state. The structure of MscL in the open state has been proposed from EPR and site-directed spin-labeling studies (27, 94). Molecular dynamics simulations using several different techniques (24) agree with the experimental data. The details of how the two channels open and close are somewhat different, but the end result is essentially the same: the α-helices lining the pore move so that it opens like the iris of a camera lens, becoming a large water-filled channel through the membrane.

TOXIN CHANNELS (COLICINS)

In addition to proteins that form pores in cis in bacteria (i.e., in the cell that synthesized them), some bacteria produce toxic proteins that form pores in trans, attacking and killing other organisms by virtue of their pore formation. Although such toxins can target both eukaryotes (e.g., anthrax toxin, diphtheria toxin) and prokaryotes (colicin) we will focus here on those bacterial toxins that kill other bacteria, specifically the class of proteins called colicins. Colicins are protein toxins produced by some strains of E. coli that target other strains of the same species. These proteins are thought to confer a selective advantage on the strains that produce them by killing related strains that might compete for nutrients. The producing strain is always immune to its own colicin, assuring its own survival (19).

First described in 1925, colicins were shown to be proteins in 1946 (19). In a seminal work, Jacob et al (50) demonstrated that a single colicin molecule (of colicin E1) is sufficient to kill an individual E. coli bacterium. The term “colicin” is broad, describing a range of toxins that act by various mechanisms; a subset, including colicins A, K, E1, Ia, and Ib, kills by forming ion channels in the inner membranes of their targets (other colicins kill by degrading DNA or inhibiting protein synthesis).

Hints as to colicin E1's mechanism of action initially came from Levinthal and Levinthal's observation that the toxin is ineffective on E. coli grown in a strict anaerobic environment (66). Follow-up of this observation revealed that the toxin affects only energy-requiring transport mechanisms, uptake by facilitated diffusion is preserved (38), and the metabolic changes in colicin E1-treated cells are consistent with specific inhibition of oxidative phosphorylation (39). A specific increase in membrane permeability was suggested by observations that treatment with colicin K rapidly compromises membrane potential and causes depletion of intracellular K⁺ (126). All of these results were elegantly tied together by the observation that colicin K spontaneously forms ion-conducting channels in planar lipid bilayers (104); a single such channel has a high enough conductance to rapidly deplete the ion gradients in an E. coli cell and short circuit its respiratory chain.

All colicins must be delivered to their sites of action to be effective: they bind to a series of receptor proteins on the outer membrane, are then translocated across this membrane and the periplasmic space, and finally exert their toxic activity on the inner membrane or in the cytoplasm. Colicin proteins have three domains corresponding to these three functions: receptor binding, translocation, and effector (channel-forming) domains. Since the receptor-binding and translocation processes are not directly related to ion channel function and have recently been reviewed (19), we will focus here on the salient features of colicin ion channels.

Structure/Mechanism

Like many channel-forming bacterial toxins, the colicin channels live a dual life: they are secreted and delivered to their targets as soluble proteins, but they must transform into membrane proteins to effect their toxicity. Several structures of soluble forms of colicin channel-forming domains have been solved, all displaying similar architectures (37, 93, 127). Each has a highly hydrophobic α-helical hairpin (long enough to span a lipid bilayer membrane) buried in a sandwich of amphipathic α-helices.

Biophysical studies suggest that, upon exposure to low pH, these proteins partially unfold to a so-called molten-globule state that largely preserves secondary structure but disrupts the tertiary architecture of the protein. First intimations regarding the nature of the unfolding step came from biochemical experiments. Upon exposure to low pH, the channel-forming domain of colicin E1 shows (i) increased protease sensitivity and detergent partitioning and (ii) increased solvent exposure of a fluorescently labeled cysteine (75). However, such pH changes result in limited changes in hydrodynamic radius (i.e., the effective size of the protein) (75). Together these results suggested a limited unfolding of the protein in these conditions. These inferences were refined by further work using circular dichroism to demonstrate that the low-pH forms of colicin channels largely retain their secondary structure while losing the primary features of tertiary structure (119). Similarly, Fourier transform infrared spectroscopy indicated that colicin E1 retains its α-helical secondary structure when reconstituted into a lipid membrane (99).

The unfolding process exposes the buried hydrophobic hairpin in the colicin channel-forming domains, facilitating its insertion into the lipid membrane. The channel-forming peptide of colicin E1 binds tightly to liposomes with K_d < 10⁻⁷ M. Before the channel opens, most of its protein mass remains on the side of the bilayer from which it inserted (the cis side), where it is sensitive to proteases in the cis solution. Mutagenesis studies revealed significantly compromised toxin activity when charged residues were introduced into the hydrophobic helical hairpin in colicin E1, suggesting that insertion of this hairpin into the lipid membrane is essential for toxicity (113). Elegant electrophysiological experiments using streptavidin bound to biotinylated cysteines at positions in the hairpin revealed that the hairpin initially binds to the membrane in a fashion that keeps its tip exposed to the cis compartment, followed by its full insertion across the bilayer and subsequent channel opening (57). However, the three-dimensional structure of the membrane-inserted colicin channel remains unclear because no X-ray structure has been solved for this form of the protein.

Although the insertion of the hydrophobic hairpin plays an important role in associating channel-forming colicins with the membrane, the process of opening a colicin channel involves dramatic but reversible rearrangements of the protein. Strong voltage dependence of channel opening suggests the movement of multiple charges on the protein at least partway across the lipid bilayer. Opening colicin E1 channels with voltage steps in planar lipid bilayers or in proteoliposomes protects the protein from proteolysis by added proteases (111), suggesting that the open channel has more protein buried in the lipid bilayer than the closed form. Mutagenesis experiments focused attention on a histidine in a loop connecting two α-helices (His440) in colicin E1, revealing that this residue carries charge across the membrane as the channel opens and closes, implying that its movement is completely across the bilayer span (1). Direct evidence of the large-scale conformational change underlying colicin channel gating came from experiments using the biotin/streptavidin system described above. A series of experiments revealed that biotin attached at various points in the colicin Ia protein loses its accessibility to cis-streptavidin but becomes accessible to trans-streptavidin upon voltage-driven channel opening, revealing that substantial portions of the protein, including partial and whole α-helices, translocate fully across the lipid bilayer to open the channel (98, 112). Indeed, heterologous peptides (51), small folded peptides (56), or even functional proteins (110) inserted into the translocated region of colicin Ia can be carried along, even when they contain multiple charged residues, suggesting that the translocation pathway may be at least partly hydrophilic in character.

BACTERIAL HOMOLOGS OF EUKARYOTIC CHANNELS

The proteins described above represent the most thoroughly studied groups of bacterial ion channels and, until fairly recently, represented the only known prokaryotic proteins in this class. However, the explosion of DNA-sequencing data resulting from the various genome projects has revealed that these proteins only scratch the surface; indeed, a plethora of previously unexpected ion channels have been identified in prokaryotes by virtue of their sequence homology with previously characterized eukaryotic channel families. Because these proteins are generally easier to express in heterologous systems and are often more stable than their eukaryotic counterparts, bacterial channels have proven extremely useful for structural studies, but their physiological functions in bacterial cells have been more evasive.

Potassium Channels

Potassium channels were among the first ion channels identified in animals, and were a major focus of Hodgkin and Huxley's original analysis of neuronal action potentials. These channels are extremely selective for K⁺ over Na⁺ and are regulated by a wide range of stimuli, from G-proteins to intracellular Ca²⁺, to the transmembrane voltage. Initial cloning revealed a characteristic architecture for these proteins (92), and subsequent analysis revealed a “signature” amino acid sequence that can help identify K⁺ channels based on sequence alone (45). Such information proved useful in identifying a K⁺ channel homolog in E. coli (Kch) (76) and another in Streptomyces lividans (KcsA) that are amenable to heterologous expression and purification (105). Subsequent genome-sequencing projects have revealed that nearly all prokaryotes carry genes coding for putative K⁺ proteins. Similarly to its mammalian counterparts, KcsA forms homotetrameric, highly K⁺-selective ion channels (43, 44, 46). KcsA established a pattern, providing the first X-ray structure of a potassium channel (34); analyses of KcsA structures have yielded profound insights into the mechanisms of ion conduction and selectivity (55, 79, 130, 131, 132, 133, 134) (Fig. 2a) in these exquisite proteins (as well as the 2003 Nobel Prize for Rod Mackinnon). Further work from the Mackinnon laboratory has revealed the structures of Ca²⁺-activated (53) (Fig. 2b) and voltage-activated (54) bacterial K⁺ channels.

Fig. 2Tetrameric cation channels. (a) KcsA (PDB ID code 1k4c) (133). (b) MethK (PDB ID code 1lnq) (53). (c) NaK (PDB ID code 3e86) (7). (d) MlotK1 (PDB ID code 3beh) (23). All structures are shown parallel to the plane of the membrane, colored by subunit. In panels a, c, and d, individual ions are shown in the channel pores.

Despite the tremendous insights gained into K⁺ channel biophysics from work on bacterial homologs, relatively little is known about the roles of these proteins in bacterial biology. E. coli growing in exponential phase maintains a strong K⁺ gradient across its plasma membrane with intracellular [K⁺] ~ 100 to 200 mM even when the extracellular [K⁺] is low (5 mM) (106) with several (nonchannel) transport systems contributing to K⁺ homeostasis in E. coli (reviewed in reference 60) including a P-type ATPase (Kdp), a low-affinity, high-rate transporter (Trk), and a third system (Kup). No phenotype has yet been found for knockouts of the E. coli K⁺ channel, Kch, although gain-of-function mutations have been used to suggest that it forms active, but closed, channels in native E. coli (61); thus, the physiological function of these proteins remains unknown. Recent experiments in Helicobacter pylori (which lacks Kdp, Trk, and Kup) suggest that the K⁺ channel in this bacterium is an uptake system essential for survival in the acidic environment of the stomach (114).

Chloride Channels (and Transporters)

Although Cl⁻ channels in eukaryotes lack the storied history of their positively charged counterions, they play crucial roles in a range of physiological processes. The success of bacterial genome projects in revealing K⁺ channels spurred a similar effort to find prokaryotic Cl⁻ channels, with similar successes. In particular, members of the CLC family of Cl⁻ channels are ubiquitous in prokaryotic genomes, with many organisms even harboring two CLC family members (68). A family member from E. coli, ClC-ec1, is easily expressed in functional form (68) and its structure has been solved by electron microscopy and X-ray crystallography (35, 77) (Fig. 3a). Only after ClC-ec1's structure had been solved was it found not to be a channel at all, but a Cl⁻/H⁺ antiporter, using the energy stored in a transmembrane gradient of either of these ions to pump the other uphill, against its gradient (2, 4). This critical insight in a bacterial system led to the reevaluation of several mammalian CLC family members, which also turned out to be Cl⁻/H⁺ antiporters (96, 103) (although several eukaryotic CLCs are certainly ion channels, not antiporters). In addition to its utility for structural analysis, ClC-ec1 has proven to be an outstanding system for mechanistic study of CLC-mediated antiport (3, 5, 6, 10, 73, 82, 83, 122). In contrast to the K⁺ channels, however, at least one biological function of a bacterial CLC transporter is known. Knocking out ClC-ec1 in E. coli impairs the bacterium's ability to withstand an extreme acid challenge (pH 2.5) (49). The CLC knockout impaired resistance mediated by two well-characterized acid resistance mechanisms and was proposed to operate by dissipating the transmembrane voltage generated by these systems (49). Even though the known bacterial forms are transporters, we include the CLCs in this section because it is likely that some bacterial forms are indeed Cl⁻ channels. A conserved residue on the transport pathway has been noted to be a valine in all of the known channel CLCs but a glutamate in all the known antiporter CLCs (5); since several predicted bacterial CLCs contain a valine at this position, they are likely to operate as CLC channels.

Fig. 3A ClC transporter and a Cys-loop receptor. (a) ClC-ec1 (PDB ID code 1ots) (36). (b) pGLIC (PDB ID code 2VL0) (47). Structures are viewed parallel to the plane of the membrane, colored by subunit. Bound Cl⁻ ions are shown in red in ClC-ec1 and possible “gate” residues are shown in blue and red for pGLIC.

Other Channels

A number of other bacterial homologs of mammalian channels have been reported. Similar to the bacterial K⁺ channels, these proteins have proven remarkably useful for structure determination and mechanistic analysis; also, similar to the bacterial K⁺ channels, the physiological functions of these proteins remain unproven. These proteins include cation channels (NachBac [100], NaK [107] [Fig. 2c]), cyclic-nucleotide-gated channels (MLotiK1) (22, 88, 89) (Fig. 2d), and glutamate receptors (GluR0) (21). Recently, prokaryotic homologs of the cys-loop receptor family (which includes the acetylcholine receptor, γ-aminobutyric acid receptor, and glycine receptor in mammals) were reported (118). A recent X-ray structure of a homolog (from Erwinia chrysanthemi) reveals a pentameric protein with strong structural similarity to the mammalian cys-loop receptors (47) (Fig. 3b). Expression and functional characterization of one such cys-loop receptor, from Gloeobacter violaceus, revealed it to be a cation-conducting channel gated (opened) by protons (18). Given that Gloeobacter is a photosynthetic organism that uses its plasma membrane for photosynthesis (it has no thylakoids), this channel was suggested to function in the organism's adaptation to different proton concentrations (18). Analysis of the gene neighborhoods surrounding the bacterial cys-loop receptors revealed, in some cases, a nearby periplasmic binding protein gene, an observation interpreted to suggest that these proteins are involved in chemotaxis (118).

CONCLUSIONS

In summary, even though in most cases ion channels represent risky propositions for bacteria, because their high-throughput rates can rapidly deplete a cell's vital ionic gradients, these proteins have still been put to important uses. Porins permeabilize the outer membrane to small solutes, an essential role that poses no threat to the integrity of the plasma membrane. MS channels reside in the plasma membrane, but their opening is tightly regulated so that they provide an “escape valve” only during life-threatening osmotic shocks. Colicins reveal the dangers of channel formation in the plasma membrane, since they kill their targets with exactly that approach. For the many other channels found in bacteria, we have yet to learn about their modes of regulation and even about their basic physiological functions. It is clear, however, that we must revise our original premise: ion channels are not only present in bacteria but abundant. The wide range of channel forms parallels those found in eukaryotes and suggests that the coming years will reveal plentiful new roles for these proteins in bacterial physiology.