By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
The recent development of specific probes for lipid molecules has led to the discovery of lipid domains in bacterial membranes, that is, of membrane areas differing in lipid composition. A view of the membrane as a patchwork is replacing the assumption of lipid homogeneity inherent in the fluid mosaic model of Singer and Nicolson (Science 1972, 175: 720–731). If thus membranes have complex lipid structure, questions arise about how it is generated and maintained, and what its function might be. How do lipid domains relate to the functionally distinct regions in bacterial cells as they are identified by protein localization techniques? This review assesses the current knowledge on the existence of cardiolipin (CL) and phosphatidylethanolamine (PE) domains in bacterial cell membranes and on the specific cellular localization of certain membrane proteins, which include phospholipid synthases, and discusses possible mechanisms, both chemical and physiological, for the formation of the lipid domains. We propose that bacterial membranes contain a mosaic of microdomains of CL and PE, which are to a significant extent self-assembled according to their respective intrinsic chemical characteristics. We extend the discussion to the possible relevance of the domains to specific cellular processes, including cell division and sporulation.
Until fairly recently, the lipids in the membranes of bacterial cells were assumed to be homogeneously distributed, and the fluidity of biological membranes had become generally accepted, according to the fluid mosaic model by Singer and Nicolson (1972). However, contradicting the assumed homogeneity, it is apparent that cell membranes must be laterally polarized to produce specific environments for certain membrane proteins, in particular the polar chemoreceptor proteins and host actin-polymerizing proteins, and proteins involved in cell division at the midcell and at asymmetrically positioned septa (for a review, see Shapiro et al., 2002). Furthermore, results indicating lateral heterogeneity of lipid molecules or lipid domains in the membranes have emerged from a variety of studies both in eukaryotic and in prokaryotic cells (for reviews, see Vereb et al., 2003; Dowhan et al., 2004). In addition to the studies employing biophysical techniques, microscopic visualization of membrane lipids in bacterial cells has reinforced the view that bacterial membranes do possess structural heterogeneity: uneven distribution of fluorescent lipophilic dyes and selective staining of septal regions has been observed in mycobacteria (Christensen et al., 1999) and the distribution of fluorescence in Escherichia coli cells stained with a lipophilic dye is distinctly uneven (Fishov and Woldringh, 1999). Finally, unequivocal visualization of cardiolipin (CL) domains in E. coli and Bacillus subtilis cells has been accomplished by means of a CL-specific fluorescent dye (Mileykovskaya and Dowhan, 2000; Kawai et al., 2004). Thus, studies with dyes have confirmed the existence of heterogeneity of phospholipids in bacterial membranes and have prompted further work on the phospholipid domains and on the relevance of this heterogeneity to physiological function.
Laterally heterogeneous distribution of phospholipids in bacterial membranes
Cardiolipin-rich domains were visualized with the CL-specific fluorescent dye 10-N-nonyl acridine orange (NAO) in the septal and on the polar membrane regions of E. coli cells by Mileykovskaya and Dowhan (2000). The same group proposed a model for the mechanism of CL-specific staining in which the nonyl group of NAO inserts between the phosphate groups at the hydrophobic surface generated by the two outer acyl chains of CL (Mileykovskaya et al., 2001). The dye forms an array of parallel skewed stacks on the surface of the hexagonalarray that comprises all (four per molecule) acyl chains of CL (Fig. 1). Septal and polar localization of the fluorescent domains was observed in B. subtilis cells during exponential growth, but not in cells carrying a clsA null mutation blocking CL synthase and lacking measurable levels of CL (Kawai et al., 2004). In sporulating cells, fluorescent domains were clearly observed in the polar septal and engulfment membranes and subsequently in forespore membranes at different stages during the course of sporulation. Interestingly, spore membranes have a quite high CL content (Kawai et al., 2006), although its localization in the membranes is not yet known. The fluorescence images of NAO in B. subtilis cells seem to be clearer than those obtained with E. coli, probably due to the simpler envelope structure of the former. The preferential localization of CL at the poles of E. coli and B. subtilis cells is consistent with its enrichment in minicells, which are formed by aberrant cell division close to the pole (Koppelman et al., 2001 and our unpublished results).
These findings encouraged us to examine the localization of another major phospholipid, phosphatidylethanolamine (PE) with the cyclic peptide probe Ro09-0198 (Ro), which binds specifically to PE (Emoto and Umeda, 2001). Treatment with biotinylated Ro followed by detection with tetramethylrhodamine-conjugated streptavidin revealed that PE is localized in the septal membranes of exponential growth-phase cells of B. subtilis and in the membranes of the polar septal and the engulfment membranes and forespore membranes at various stages in sporulating cells (Nishibori et al., 2005). As mutant cells lacking PE were not stained, one can be confident that the fluorescence reflects the localization of PE-rich domains in the septal membranes. A typical example of PE localization in B. subtilis cells detected by fluoresceine-labelled Ro is shown in Fig. 2A. It turns out that the application of fluoresceine-labelled Ro produces clearer images of the localization of PE than biotinylated Ro-streptavidin conjugated with tetramethylrhodamine (Nishibori et al., 2005). Figure 2B shows a series of thin sections along the z-axis. Note the band of intense fluorescence at the septal region of every thin section, indicating that the intense fluorescence band is not an artefact from piling up of weak fluorescence images but is actually present in the septal membranes.
In E. coli cells, the fluorescence signal of Ro-bound PE is distributed uniformly over the cell surface, suggesting that PE is uniformly distributed over the whole cell membrane (Nishibori et al., 2005). A uniform signal was also observed in many other Gram-negative bacteria, e.g. certain strains of Salmonella typhimurium, Pseudomonas putida, Azotobacter vinelandii and Proteus vulgaris. In many Bacillus species, including Bacillus polymixa, Bacillus amyloliquefaciens and Brevibacillus brevis, Ro-bound PE showed septal PE localization, as in B. subtilis. The different distribution of the signal between the Gram-negative and Gram-positive bacteria might imply that PE plays different physiological roles in the two bacterial types.
The Ro probe binds at the cleavage furrow of dividing Chinese hamster ovary (CHO) cells and it has been suggested that PE, which usually resides in the inner leaflet of the plasma membrane, is exposed on the outer leaflet of the membrane of the cleavage furrow at the final stage of cytokinesis (Emoto and Umeda, 2001). The outer leaflet of the plasma membrane of yeasts has been probed for PE using similar techniques (Iwamoto et al., 2004). The PE signals are located at the bud neck of late mitotic stage, large-budded Saccharomyces cerevisiae cells. In the fission yeast Schizosaccharomyces pombe, PE is located at the division plane of late mitotic cells and at one or both poles of mononucleated cells, suggesting that PE exposure at the region of cell division is a common feature. In addition, the use of a fluorescent probe, filipin, indicated sterol localization to the site of cell division in the fission yeast (Wachtler et al., 2003).
Septal localization of phospholipid synthases and generation of lipid domains
The septal localization of both PE- and CL-rich domains directed our interest to the subcellular localization of the enzymes involved in PE and CL synthesis. The committed step in PE synthesis in B. subtilis is catalysed by phosphatidylserine synthase (PssA). Its reaction product, phosphatidylserine, is then converted to PE. Phosphatidylglycerophosphate synthase (PgsA) catalyses the committed step for the synthesis of phosphatidylglycerol (PG), which is then used by CL synthase (ClsA) to produce CL. All GFP fusions to these enzymes were septally localized, even when expression levels (and hence fluorescence intensity) were low (Nishibori et al., 2005). Thus, these enzymes probably concentrate at the septum under natural conditions.
Attempts to localize other B. subtilis enzymes involved in lipid synthesis (Table 1) have yielded interesting results (Nishibori et al., 2005). GFP fusions to several phospholipid synthases were localized to the septum in a thick, bright fluorescence band. These synthases include CdsA, which produces CDP-diacylglycerol, Psd, which converts phosphatidylserine into PE, MprF, which transfers lysine to PG to produce lysyl-PG, and UgtP, which is responsible for glucolipid synthesis. Their distribution thus differed from the uniform distribution of the membrane proteins AtpC, a subunit of ATP synthase, and SecY, the uniform membrane distribution of GFP fusions of which has been shown by H. Takamatsu and T. Kobayashi (pers. comm.) and has also been confirmed in our laboratory. It also differed from the uniform cytoplasmic localization of GpsA, which catalyses the production of glycerol 3-phosphate. It has recently been suggested that a complex of lipid synthesis factors, including the acylcarrier protein, YbgC (which exhibits thioesterase activity on acyl-CoA derivatives), PssA and sn-glycerol 3-phosphate acyltransferase (PlsB), resides in the E. coli inner membrane (Gully and Bouveret, 2006). It seems likely that the septally localized lipid synthases are also integrated in such a lipid synthesis complex for co-ordinated lipid metabolism.
Table 1. Cellular localization of the product of the genes involved in lipid synthesis in B. subtilis and E. coli.
Localization in B. subtilis cells
Genes and their functions and cellular localization of the products are from Nishibori et al. (2005). An asterisk indicates that the gene is absent in B. subtilis and double asterisk indicates absence in E. coli. The triple asterisk indicates that the septal localization appears as a two-dot structure that is thickest near the edge of the septal face. The question mark indicates that the localization may not be confined to the septal membranes (unpublished results). Note that no specific localization of the enzymes in E. coli cells has been shown.
Although polar and septal localization of CL has been demonstrated in E. coli, efforts to localize CL synthase have been less successful. GFP-CL synthase chimeras constructed so far were not functional and their fluorescence was observed in the cytoplasm. No specific localization of PgsA, the enzyme responsible for the preceding reaction in E. coli cells, has been found either; the fluorescent GFP-PgsA chimera was observed as dots distributed around the periphery of the cells (our unpublished data).
How can phospholipid synthases be targeted to the septal membranes? FtsZ-depletion experiments indicate that the phospholipid synthase localization depends on FtsZ and exclude the possibility that they become localized before FtsZ ring assembly (Nishibori et al., 2005). Thus, localization probably follows or is concurrent with the assembly of cell division proteins in order to synthesize phospholipid membranes in concert with the synthesis of peptidoglycan at the leading edge of the invaginating envelope. Two-hybrid analyses of B. subtilis lipid synthases with cell division proteins and envelope proteins are in progress. Preliminary results suggest possible interactions of the lipid synthases with some cell division proteins and envelope proteins (our unpublished results in collaboration with the group of Dr H. Yoshikawa, Tokyo University of Agriculture). The lipid synthases presumably have a specific region(s) responsible for the septal localization (or interaction with certain cell division proteins), leading one to wonder about the consequences of their inactivation. Furthermore, as the septal localization of the phospholipid synthases in B. subtilis cells implies that most phospholipids are synthesized mainly at the septal membranes, there must be a mechanism that prevents PE and CL diffusion into the lateral membranes. This mechanism need not to be a physical barrier. It could be that the diffusion of lipid molecules is essentially free but that some lipids associate preferentially with much larger structures that themselves are localized while other lipids go elsewhere (V. Norris, pers. comm.).
It should be noted that the GFP-labelled version of UgtP, responsible for glucolipid synthesis, appears as a ‘two dot’ structure that is thickest near the edge of the septal face. It seems likely that it actually forms a ring structure, like FtsZ, and thus differs from the phospholipid synthases, which form a band on the septal membranes. UgtP is probably not an integral membrane protein, as it does not have a membrane-spanning region (Nishibori et al., 2005), in contrast to phospholipid synthases, which have several membrane-spanning regions. This property of UgtP might have some relation to the difference in its localization pattern on the septal membranes.
The chemical basis for the generation of lipid domains
How do the lipid molecules form domains in membranes that are fluid? Both lipid–lipid interactions and lipid–membrane protein interactions are suspected to induce the formation of microdomains, which comprise at most some tens of molecules of a specific lipid (for a review, see Edidin, 1997). The following properties of CL and PE might account for the formation of labile microdomains through lipid–lipid interactions.
The polar head of the PE molecule has both a cationic amine residue and an anionic phosphate residue. Each amine and unesterified phosphate oxygen can participate in two short distance intermolecular hydrogen bonds. The ethanolamine groups thus form a linkage between phosphorous groups of adjacent PE molecules producing a very compact, rigid head-group network at the bilayer surface (Elder et al., 1977; Hauser et al., 1981; Boggs, 1987), giving PE substantially higher Tm values than phosphatidylcholine, which has an identical acyl chain structure (reviewed in London and Brown, 2000). This compact head-group network of PE might well suffice to explain PE microdomain formation. Interactions with membrane proteins (both transmembrane and peripheral membrane proteins) might then modify the lipid organization and further stabilize the head-group network of PE, as suggested by Edidin (1997).
Cardiolipin, which has a double-glycerophospholipid structure connected with a glycerol residue, has a rather unique head group, with a tightly locked, surprisingly small configuration and only one negative charge (Haines and Dencher, 2002; this result contradicts the depiction found in textbooks). The two phosphates in the head group trap a proton and are locked in a bicyclic array held together by the hydroxyl residue of the glycerol, which connects the two halves of the CL molecule. This small polar head group makes for a tighter packing of hydrophobic acyl chains between CL molecules by van der Waals force interactions than is found in lipids with larger polar head groups. It has also been suggested (Haines and Dencher, 2002) that interaction between adjacent head groups creates a compact array of CL molecules (Fig. 1), which becomes manifest in the presence of associated NAO arrays (Mileykovskaya et al., 2001). Head groups of PG, the third major phospholipid in E. coli, interact by an extensive network of hydrogen bonds, ionic bonds and co-ordination bonds between glycerol hydroxyls and the unesterified phosphate oxygen both in the anhydrous crystal and in the hydrated gel state (Boggs, 1987; Pascher et al., 1987). In fact, PG is perhaps segregated into distinct domains, which are different in their composition and proteo-lipid interaction, according to studies using pyren-labelled phospholipids in both B. subtilis and E. coli membranes (Vanounou et al., 2003). This extensive and tight network of PG molecules might cause exclusion of CL molecules to produce patches of CL in the membranes, as the small and tightly locked head group of CL cannot interact with PG. Certain membrane proteins with an affinity to CL might also have a role in stabilizing the patches of CL. The head-group array formed within patches of CL might allow these membrane proteins to bind specifically to the patches. The observed CL-rich domains of NAO fluorescence in the polar and septal regions should be rich in such microdomains stabilized by membrane proteins.
A growing number of proteins have been shown to localize to the septal and polar membranes in bacterial cells (Lybarger and Maddock, 2001; Lai et al., 2004). It is possible therefore that some of them, having an affinity for the head-group array of CL, might help stabilize patches of the lipid. E. coli protein MurG, a peripheral membrane N-acetylglucosamine transferase involved in murein synthesis, interacts preferentially with CL, and its overexpression results in formation of vesicles enriched with CL at the poles of cells (van den Brink-van der Laan et al., 2003). MurG is thus a candidate CL microdomain-stabilizing protein. Another one is monoglucosyl diacylglycerol synthase, a glucosyltransferase from Acholeplasma laidlawii with an affinity for acidic phospholipids that preferentially localizes to the cell poles when it is expressed in E. coli cells (Wikström et al., 2004).
In summary, the observations and considerations above give us a new view of bacterial membrane surfaces. They consist of a mosaic of small patches or microdomains of phospholipids with a specific polar head group, the maintenance of which might be supported by certain membrane proteins. The conditions required for their formation are probably similar to those for the formation in eukaryotic cell membranes of sphingolipid rafts that are promoted by both hydrogen bonding between –OH and –NH in the polar head groups (Boggs, 1987) and the van der Waals forces of long, saturated acyl chains (Edidin, 1997; London and Brown, 2000; Kobayashi et al., 2001).
Physiological roles of lipid domains
The propensity of CL and PE to form non-bilayers in cell division and sporulation
Phosphatidylethanolamine and CL domains in B. subtilis cells tend to occur in the same regions (Nishibori et al., 2005). What, then, might be the roles of the CL and PE domains in the septal membranes and in the membranes of sporulating cells? It appears that the propensity to form non-bilayer structures conferred by small polar head groups (Dowhan, 1997) is essential here. At the initial stage of cell division, the small radius of curvature of the division site, on the leading edge, requires a lipid with a small head group in the concave region of the outer monolayer (Fig. 3). However, as invagination proceeds to decrease the diameter of the ring, the constraints become dominated by the convexity of the monolayer. However, the packing constraints (and, hence, the nature of the lipids) in the inner monolayer of the bilayer membranes are totally opposite (Norris et al., 2002). Fusion and fission of bilayer membranes might require a lipid to take on a non-bilayer structure (Cullis and de Kruijff, 1979). If this is correct, then cells are faced with the problem of ensuring a supply of appropriate lipids at the division site. It seems likely that the septal localization of phospholipid synthases can meet this need by serving lipids of appropriate shape at proper times and sites during the cell division process; indeed, this might, in fact, be the major reason for the septal localization of the majority of the phospholipid synthases.
In sporulating cells, PE and CL domains are also found in the polar septa and on the engulfment and forespore membranes. During the sporulation process, membranes undergo dynamic transformations, including the formation of the asymmetric septal membrane, engulfment and finally fusion (Fig. 4). Development of these sporulation-specific membranes probably requires that the lipids involved be able to form non-bilayers as described above. The significance in sporulation of CL has been demonstrated by examining mutants lacking CL, which show retarded emergence of polar septal and engulfment membranes and a reduced frequency of heat-resistant spores (Kawai et al., 2004). The process of activation of sigma E and sigma F during the sporulation phase involves the asymmetric septal membranes (Feucht et al., 1996; Schujman et al., 1998), which should be rich in CL, and the activation of the former has been shown to require de novo fatty acid synthesis (Schujman et al., 1998). Thus, some of the factors responsible for the activation, which are on the septal membranes, might require CL to work properly (this requirement is not necessarily the propensity to form non-bilayer structures).
The significance of PE, however, has not been clarified in B. subtilis cells as PE-defective mutant cells show no obvious change in phenotype (Matsumoto et al., 1998). Whatever the role of PE might be, it appears that CL or other lipids with similar properties can replace it.
The acidic nature of CL and recruitment of membrane proteins
The negative charge of CL recruits peripheral membrane proteins to the membranes. This has been amply illustrated by the examples of DnaA, FtsY, GlpD, PssA and SecA (Dowhan, 1997; Matsumoto, 2001; Walz et al., 2002). Basic residues of these proteins are assumed to interact with the negative charge of the acidic phospholipid. The MinD regulator of cell division site selection is also included in this category of proteins (Mileykovskaya et al., 2003; Dowhan et al., 2004) and will be discussed briefly in the next section.
Many proteins have been localized at the polar membranes (Lybarger and Maddock, 2001; Lai et al., 2004), and some of them might also be recruited by the negative charge of CL. It has been suggested that in the invasive bacterium Shigella, where IcsA localized at the pole nucleates host actin filaments to form a comet-like tail that propels the bacterium forward to penetrate a neighbouring cell, certain specific receptors exist on the membrane at the old pole, which induce the nascent IcsA polypeptide to be secreted there (Brandon et al., 2003; for a review see Pugsley and Buddelmeijer, 2004). Such receptor proteins and other proteins involved in similar polar targeting might be recruited to the poles by the negative charge of CL.
Possible relevance to regulation of cell division machinery
The septal localization of domains enriched in particular phospholipids suggests their relevance to the cell division process. Besides meeting membrane curvature requirements and forming non-bilayer structures during the process (as discussed above), they might be involved in the functional regulation of the cell division machinery. The cell division process begins with the formation of a ring composed of polymerized FtsZ protein (Weiss, 2004). The position of the ring formation is confined to the midcell by a combination of nucleoid occlusion and Min systems, which inhibit the Z-ring assembly in the vicinity of the nucleoids and of the poles respectively.
The MinC division inhibitor is recruited to the cytoplasmic membrane by the ATP-bound form of the MinD protein. In E. coli cells, the MinE topological specificity factor drives MinD to oscillate rapidly from pole to pole along helical arrays by stimulating the ATPase activity of MinD. This oscillation maximizes the time-averaged concentration of MinC at the poles and minimizes it at the midcell. MinD interacts preferentially with acidic phospholipids (Mileykovskaya et al., 2003). The acidic phospholipid content of the membrane affects the affinity of MinD for the membrane and modulates its dynamic behaviour. In a pssA-null mutant that completely lacks zwitterionic PE and contains only acidic phospholipids, MinD does not oscillate from pole to pole but forms compact spots which randomly migrate around the thus partially division-defective filamentous cell (Mileykovskaya et al., 2003). This leads to the conclusion that polar localization of CL-rich domains in the wild-type cell plays an important role in MinD oscillation by promoting assembly of MinD oligomers at the poles (Dowhan et al., 2004; Mileykovskaya and Dowhan, 2005), although it does not simply define the MinD distribution in the absence of MinE, as MinD is distributed almost evenly around the cell periphery in a minE mutant. We note that the interaction of MinD with the membrane depends on its conserved C-terminal membrane-targeting sequence (MTS). This sequence forms an amphipathic α-helix with a preference for acidic phospholipids because of basic residues on the hydrophilic side (Szeto et al., 2003).
The FtsA protein, which functions to tether FtsZ polymers to the membrane, has the conserved amphipathic MTS in its C-terminus. This MTS also has basic residues on the hydrophilic side of the helix and is functionally interchangeable with that of MinD (Pichoff and Lutkenhaus, 2005). Although the question has not yet been examined, it can be reasonably expected that FtsA also preferentially interacts with acidic phospholipids. This might have significance for the function of FtsA, considering the localization of the CL-rich domains in the septal region.
In B. subtilis, which does not have MinE, the topological specificity factor of the Min system is the coiled-coil protein DivIVA. It is targeted to the cell poles by an as yet unknown mechanism, independent of FtsZ and other division proteins (Harry and Lewis, 2003) and recruits MinD, which then recruits MinC, to the poles. These proteins persist there without oscillation and block polar division. Surprisingly, DivIVA is targeted to the cell poles in E. coli and even in S. pombe (Edwards et al., 2000). It has been proposed that this protein might be attracted by a physical property of the poles and that the targeting signal for division sites is conserved across eukaryotes and prokaryotes. It is intriguing that the nascent division sites and poles of these organisms are enriched in CL and PE (B. subtilis), CL (E. coli), and PE and sterols (S. pombe; cf. Wachtler et al., 2003; Iwamoto et al., 2004), all conferring a non-bilayer-forming propensity on the membrane.
The standard concept of the ‘fluid’ and ‘mosaic’ architecture of membranes (Singer and Nicolson, 1972) that has been depicted in the textbooks of biochemistry assumes that lipid molecules are homogeneously distributed in the membranes and that integral membrane proteins resemble icebergs floating unencumbered in a two-dimensional lipid sea. However, the evidence that lipid molecules in eukaryotic membranes are segregated into regions (microdomains) of specific lipid molecules or specific composition is accumulating (see the latest review, Engelman, 2005). The view of prokaryotic membranes is also changing to one where a mosaic of small patches or microdomains of phospholipids with a specific polar head group, CL and PE, is to a significant extent self-assembled according to their respective intrinsic chemical characteristics, and possibly maintained with the support of certain membrane proteins. Thus, contradicting the previously assumed homogeneous distribution, bacterial membranes are patchy with regions of specific lipid molecules and one must admit that, for lipid membranes, the fluid is a mosaic.
We thank Professor Emeritus Isao Shibuya and Professor Yoshito Sadaie for discussion and encouragement. We also thank Tony Pugsley for a critical reading of this manuscript. Thanks are also due to Yoshinori Hara, Tomohiro Hayakawa, Toshihide Kobayashi, Eugenia Mileykovskaya, Vic Norris, Akinori Ohta, Satoshi Shuto, Hiromu Takamatsu, Masato Umeda, Akihiro Yoshida and Hirofumi Yoshikawa for their helpful discussions, suggestions and encouragement. Due to the concise nature of this Microreview, we were unable to cite all relevant studies regarding this area of research.