Effects of phospholipid composition on MinD-membrane interactions in vitro and in vivo.

The peripheral membrane ATPase MinD is a component of the Min system responsible for correct placement of the division site in Escherichia coli cells. By rapidly migrating from one cell pole to the other, MinD helps to block unwanted septation events at the poles. MinD is an amphitropic protein that is localized to the membrane in its ATP-bound form. A C-terminal domain essential for membrane localization is predicted to be an amphipathic alpha-helix with hydrophobic residues interacting with lipid acyl chains and cationic residues on the opposite face of the helix interacting with the head groups of anionic phospholipids (Szeto, T. H., Rowland, S. L., Rothfield, L. I., and King, G. F. (2002) Proc. Natl. Acad. Sci. U. S. A. 99, 15693-15698). To investigate whether E. coli MinD displays a preference for anionic phospholipids, we first examined the localization dynamics of a green fluorescent protein-tagged derivative of MinD expressed in a mutant of E. coli that lacks phosphatidylethanolamine. In these cells, which contain only anionic phospholipids (phosphatidylglycerol and cardiolipin), green fluorescent protein-MinD assembled into dynamic focal clusters instead of the broad zones typical of cells with normal phospholipid content. In experiments with liposomes composed of only zwitterionic, only anionic, or a mixture of anionic and zwitterionic phospholipids, purified MinD bound to these liposomes in the presence of ATP with positive cooperativity with respect to the protein concentration and exhibited Hill coefficients of about 2. Oligomerization of MinD on the liposome surface also was detected by fluorescence resonance energy transfer between MinD molecules labeled with different fluorescent probes. The affinity of MinD-ATP for anionic liposomes as well as liposomes composed of both anionic and zwitterionic phospholipids increased 9- and 2-fold, respectively, relative to zwitterionic liposomes. The degree of acyl chain unsaturation contributed positively to binding strength. These results suggest that MinD has a preference for anionic phospholipids and that MinD oscillation behavior, and therefore cell division site selection, may be regulated by membrane phospholipid composition.

MinD protein, along with MinC and MinE, is required for selection of the correct placement of the division site in bacterial cells (1). During vegetative growth, cell division in rodshaped bacteria occurs at the cell center. The earliest event in this process is the polymerization of the tubulin-like protein FtsZ at mid-cell into an annular structure called the Z-ring (2)(3)(4). In the absence of the min system, Z-rings form at midcell as well as cell poles, resulting in the production of minicells (5). The ATP-bound form of MinD, an amphitropic peripheral membrane protein, is localized to the membrane. The binding of MinE to MinD induces hydrolysis of ATP and the release of MinD into the cytoplasm (6). The ATP binding cycle induced by MinE results in the rapid movement of MinD from one cell pole to the opposite cell pole, forming alternating broad polar zones (7). MinC is a specific inhibitor of Z-ring formation (8,9). Because MinC binds to MinD, the movement of MinC from pole to pole with relatively long polar dwell times and a short transit time blocks the formation of polar Z-rings but not medial rings (10,11). Therefore, the ATPase activity of MinD is presumed to provide the driving force for the pole-to-pole oscillation of the MinC division inhibitor.
The mechanism of MinD binding to the membrane recently has been elucidated (12,13). The C-terminal region of MinD contains a highly conserved motif that is essential for membrane localization. This motif is unstructured in crystals of MinD (14,15). On the other hand, the motif is predicted to be an amphipathic ␣-helix with one side of the helix containing mainly hydrophobic amino acids and the other side containing mainly positively charged amino acids (12). According to the model (12), ATP binding to MinD induces a conformational change in the protein that results in a release of the C-terminal motif and exposure of its hydrophobic residues followed by binding to the phospholipid bilayer and helix formation. Such amphipathic helices usually align parallel to the membrane surface so that the hydrophobic residues interact directly with lipid acyl chains, whereas the cationic residues on the opposite face of the helix interact with the head groups of anionic phospholipids (for review, see Ref. 16). In the present study we use both in vivo and in vitro systems to examine directly whether the Escherichia coli MinD protein has a preference for anionic over zwitterionic phospholipids.

EXPERIMENTAL PROCEDURES
Strains and Growth Conditions-To construct a phosphatidylethanolamine (PE) 1 -deficient strain expressing a green fluorescent protein (GFP) derivative of MinD, plasmid pWM1255 (17) was introduced into strain AD90 (pss93::kan R recA) carrying plasmid pDD72 (pssA ϩ cam R ), which has a thermosensitive origin of replication derived from plasmid pSC101 (18). Plasmid pWM1255 (amp R ColE1 origin) synthesizes MinE and a translational fusion of GFP to the N terminus of MinD, both from the isopropyl-1-thio-␤-D-galactopyranoside-inducible Ptrc90 promoter (19). Strain AD90/pDD72/pWM1255 (PE-containing) exhibits wild-type phospholipid composition. Curing of plasmid pDD72 as described previously (18) resulted in strain AD90/pWM1255, which lacks PE (contains only phosphatidylglycerol (PG) and cardiolipin (CL)) but maintains the same membrane protein-to-phospholipid ratio as wild-type cells made up of solely PG and CL (18). Strains were grown on LB agar or liquid LB medium supplemented with 50 mM MgCl 2 and 50 g/ml ampicillin.
To construct a MinD-overproducing strain, the E. coli minD gene was cloned as an EcoRI-HindIII fragment into plasmid pET28aϩ (Novagen), fusing minD to an N-terminal hexahistidine tag and placing the His-minD construct under the control of the T7 promoter. This plasmid, pET28a-MinD (kan R ), then was introduced into strain BL21(DE3) (20) to make strain WM1682. For His-MinD purification, WM1682 cells were grown overnight in LB medium with kanamycin (50 g/ml) at 37°C. The culture was diluted 100-fold in the same medium, and growth was continued to an A 600 of 0.6. To induce overproduction of His-MinD, isopropyl-1-thio-␤-D-galactopyranoside was added to 1 mM, and cells were grown for an additional 2-3 h. The culture was collected by centrifugation and frozen at Ϫ80°C.
Purification of MinD-Cells from 1 liter of culture of WM1682 were suspended in 25 ml of lysis buffer containing 50 mM NaH 2 PO 4 , 300 mM NaCl, 10 mM imidazole, pH 8.0, and 25 l of protease inhibitor mixture (Calbiochem, set I). Cells were broken using a French press, and cell debris was removed by centrifugation at 10,500 ϫ g av for 10 min (4°C). The supernatant was centrifuged further at 120,000 ϫ g av for 90 min (4°C). Nickel-nitrilotriacetic acid Superflow slurry (Qiagen) was added to the supernatant (3 ml of slurry for 12 ml of supernatant) and mixed by shaking in a tube at 4°C for 1 h. The slurry was poured into a column and washed twice with 8 ml of wash buffer containing 50 mM NaH 2 PO 4 , 300 mM NaCl, and 20 mM imidazole (pH 8.0). His-MinD was eluted with three aliquots of elution buffer (2 ml) containing 50 mM NaH 2 PO 4 , 300 mM NaCl, and 250 mM imidazole (pH 8.0). Fractions (2 ml) were collected into tubes containing 2 ml of 20% glycerol. Protein concentration was determined by the BCA method, and the fractions were analyzed by SDS-PAGE. Fractions containing His-MinD were desalted on Sephadex G-25 columns equilibrated with 50 mM HEPES, pH 7.2, 150 mM KCl, 0.1 mM EDTA, and 10% glycerol (storage buffer) and were frozen in 0.2-ml aliquots at Ϫ80°C.
For some experiments, the His tag was removed from the protein by treatment with biotinylated thrombin (pET28aϩ encodes a thrombin cleavage site) using the thrombin kit (Novagen) according to the kit instructions. Small scale optimization was done with serial dilutions of thrombin and monitoring of cleavage by using SDS-PAGE. After the cleavage reaction, biotinylated thrombin was removed with streptavidin-agarose as recommended in the kit instructions.
Monitoring of GFP-MinD Movement in Cells-To observe MinD movement, cells were grown to exponential phase at 30°C. Leaky expression from the Ptrc90 promoter of pWM1255 (without the addition of isopropyl-1-thio-␤-D-galactopyranoside) resulted in levels of GFP-MinD high enough to be detected by the camera system but low enough to have minimal impact on cell physiology. For microscopic time-lapse studies, cells were immobilized as described by Corbin et al. (17). Briefly, 5 l of cell culture was mixed with an equal volume of molten growth medium containing 2% low melt agarose, was dropped onto a microscope slide, and was covered quickly with a cover glass. Fluorescence images were observed with a ϫ100 oil immersion objective on an Olympus BX60 microscope fitted with a GFP filter cube, and the images were captured with a light-sensitive Photometrics CoolSnap FX cooled charge-coupled device camera driven by QED image capturing software. Time-lapse images were taken every 10 s, each with 2-4-s exposures using 2 ϫ 2 binning to improve sensitivity, and the images were saved as PICT files. Movies were made from these files with iMOVIE 2.2 software.
Preparation of Liposomes-Soybean phosphatidylcholine (PC), CL from heart, synthetic dioleoyl-PG, E. coli PE, and E. coli polar phospholipids were purchased from Avanti Polar Lipids (Alabaster, AL). E. coli PG and CL were purified from E. coli polar phospholipids (21). Liposomes were prepared by water bath sonication for 1-2 h (6-12 times for 10 min each) at 0°C in a buffer containing 25 mM Tris-HCl, pH 7.5, and 50 mM KCl and were stored at Ϫ80°C.
MinD Binding to Liposomes-The sedimentation assay (6) with mi-nor modifications was used to study MinD binding to liposomes. MinD preparations first were centrifuged at 263,000 ϫ g av for 10 min to remove potential MinD aggregates. MinD (0.2-20 M) was mixed with 1 mM ATP or ADP and liposomes (160 g/ml) in 100 l of buffer containing 25 mM Tris-HCl, pH 7.5, 50 mM KCl, and 5 mM MgCl 2 and was incubated for 10 min at 30°C. After incubation, samples were centrifuged at 263,000 ϫ g av at 30°C for 10 min. After centrifugation, the pellets were resuspended in 40 l of sample buffer. Aliquots of 20 l were subjected to 12% SDS-PAGE and stained with Bio-Safe Coomassie Blue (Bio-Rad). The bands were quantified using FluorS-Max Multi-Imager with Quantity One software (Bio-Rad). The experimental data were compiled in KaleidaGraph 3.01. Lines on the graphs (Figs. 4 and 5) are curves calculated by the Hill equation that best fit the data points. Hill coefficients (n H ) and association constants (K a ) were calculated from the graph in Fig. 5.
Labeling of MinD with Fluorescent Probes-MinD (0.8 mg/ml) in 100 l of the storage buffer was incubated at 22°C in the dark with a 20-fold molar excess of 2-(4Ј-maleimidylanilino)naphthalene-6-sulfonic acid (MIANS; Molecular Probes, M-8) or lucifer yellow iodoacetamide (LY; Molecular Probes, L-1338) for 2 h. The conjugate was separated on Sephadex G-25 columns equilibrated with the same buffer. The level of modification, estimated from protein concentration and fluorescence of standard dye solutions, was close to 1:1 for both dyes. Steady-state fluorescence resonance energy transfer (FRET) experiments were conducted at 22°C on a PTI QuantaMaster spectrofluorometer (Photon Technology International). Samples contained 1 M total MinD protein in the same buffer that was used for the liposome sedimentation assay. Additional components were added as indicated. In a single scan, a sample was excited at 322 nm (slit width 4 nm), and emitted light was scanned from 350 to 600 nm (slit width 4 nm, scan rate 100 nm/min).

GFP-MinD in PE-lacking Cells-
We demonstrated previously that E. coli cells lacking PE were filamentous because of inhibition of cell division. The lack of PE did not prevent localization of FtsZ, FtsA, and ZipA in these cells, but the proteins often formed aberrant spiral structures (22). We first examined the dynamic localization of GFP-MinD in the pss93 mutant (AD90/pWM1255) containing the covering pDD72 plasmid, which maintains wild-type phospholipid composition and cell morphology. In these cells, GFP-MinD localized into a typical horseshoe structure on the membrane at the poles of the cells and oscillated from pole to pole in large zones with cycle times similar to those reported previously (data not shown) (7). In contrast, GFP-MinD movement in cells lacking PE was markedly different as shown by a typical time-lapse series of GFP-MinD movements in a filamentous PE-lacking cell ( Fig. 1). Although the change in phospholipid composition still allowed movement of GFP-MinD in the PE-lacking cells, fluorescence localized as a "zigzag" pattern of compact spots along the filamentous cells. No clear back-and-forth movement was observed. As can be seen by following individual spots in Fig. 1, the lifetime of most spots was ϳ30 -50 s, about a factor of 2 longer than the shift-dwell-shift observed for GFP-MinD zones in PE-containing cells (7) (data not shown). Because of the apparent random localization of the spots, it was difficult to establish their lineage during the time course. However, the simultaneous appearance of a new spot and disappearance of an old spot nearby in most cases suggested that material from the old spot was migrating to the new spot. For example, fluorescence in spots B and C, visible in the first panel, appears to have been transferred to spots E and D, respectively. If so, this would represent a distance of ϳ0.5-1 m, considerably less than the migration distances observed in filamentous cells with normal phospholipid composition (7). Interestingly, the spots appeared to move not only along the long axis of the cell but also in the perpendicular direction (for example, see M-Q).
This disturbed pattern of movement suggested that the high level of PG and CL and/or the lack of PE in the membrane results in the alteration of the GFP-MinD assembly-disassembly process on the membrane surface. This effect might be a consequence of the change in the affinity of MinD to the membrane composed of only anionic phospholipids. To test this proposal, we studied the interaction of MinD with liposomes of different phospholipid compositions in vitro.
Interaction of Purified MinD with Liposomes-It was demonstrated that MinD with a His tag at the N terminus is functionally active in E. coli (23). Therefore, we constructed and purified a His-tagged MinD protein for in vitro experiments (see "Experimental Procedures"). To study binding of MinD to liposomes, we used the method developed by Hu et al. (6) in which the protein is incubated in the presence of ATP and liposomes followed by high speed centrifugation (see "Experimental Procedures"). Fig. 2, A and B, represents an experiment in which MinD was bound to liposomes of different phospholipid compositions. As seen in Fig. 2, A and B, MinD at 6 M can bind to PC liposomes in an ATP-dependent manner. Enrichment of PC liposomes with an anionic phospholipid (CL from heart) strongly enhanced both specific (ATP-dependent) and nonspecific (ADP-dependent or without any nucleotide) binding of MinD. At the same time, enrichment of PC liposomes with PE did not change the level of MinD bound to liposomes. Binding to liposomes made from E. coli total phospholipid containing about 20% of anionic phospholipid was also higher than binding to PC or PC-PE liposomes. All these data together indicate that MinD has a higher affinity to liposomes that are enriched with anionic phospholipids.
To investigate whether MinD binding to CL from heart is ATP-dependent, we decreased the concentration of MinD in the incubation mixture to 2.5 M and determined the dependence of MinD binding on levels of CL by using PC-CL liposomes of varying CL concentrations (Fig. 3, A and B). The increase in the content of CL in PC liposomes enhanced ATP-dependent specific binding. However, nonspecific binding without nucleotides also was elevated in liposomes with higher CL concentrations.
To exclude the possibility that the higher affinity to anionic phospholipids is a property of the His-tagged protein but not the native MinD, the tag was removed by thrombin digestion. MinD without the His tag had the same high affinity to PC-CL liposomes as His-tagged MinD (data not shown). All experiments described below were carried out with His-tagged MinD, and in all in vitro experiments described herein, His-MinD is referred as MinD.
To study in more detail whether the affinity of MinD to the membrane depends on its phospholipid composition, we investigated the affinity of MinD to binding to liposomes of different phospholipid compositions as a function of MinD concentration. For each type of liposome, we measured the dependence of MinD binding on MinD concentration in the presence of ATP or ADP. Fig. 4 represents an example of such a titration experiment for PC liposomes enriched with E. coli PG. The amounts of MinD bound to liposomes are expressed in arbitrary units (A.U.). Every experimental point was obtained by quantification of the corresponding band after SDS-PAGE (see "Experimental Procedures"). Protein determination showed that in all experiments the amount of MinD bound to liposomes was no more than 15% of the total amount of MinD present (data not shown). Lines on the graphs represent the best fit to the experimental points as calculated by the Hill equation. The calculated association constant (K a ) and Hill coefficient (n H ) for the ATP form of MinD were 4.5 M and 2, respectively. the one presented in Fig. 4. The values of the association constants and Hill coefficients are summarized in Table I. As shown in Fig. 5 and Table I, the affinity of MinD-ATP to liposomes is twice as high for the liposomes containing a fraction of anionic phospholipids and 9-fold higher for the entirely anionic liposomes than for the pure PC liposomes. The higher affinity of MinD to liposomes composed of PC and CL from heart compared with PC and CL from E. coli might be due to the difference in fatty acid composition of these two species of CL. CL from bovine heart is highly unsaturated and contains four polyunsaturated acyl chains, whereas acyl chains of CL from E. coli are both saturated and monounsaturated (24). This finding is consistent with results (25) showing that the affinity of DnaA, another protein with amphipathic ␣-helices, to phospholipids depends on both the negative charge of head groups and the unsaturation of acyl chains.
As shown in Figs. 4 and 5 and Table I, the binding of MinD to liposomes in the presence of ATP displays a considerable positive cooperativity with respect to the protein concentration. The Hill coefficient of ϳ2 is characteristic of MinD-ATP binding to all types of liposomes used in the experiments. This result suggests that MinD-ATP oligomerizes at least to the level of dimers on the liposome surface. In contrast, binding of MinD-ADP to liposomes ( Fig. 4 and data not shown) was characterized by a Hill coefficient close to 1, which may reflect nonspecific binding without oligomerization on the surface of liposomes. In panel A, samples were processed as in Fig. 2A. In panel B, samples were quantified as in Fig. 2B. A.U., arbitrary units. Samples were processed and analyzed as described for Fig. 4. Each point on the graph represents the amount of MinD bound to liposomes in the presence of ATP as a percentage of theoretical maximal binding calculated by using the Hill equation from graphs similar to the one presented in Fig. 4.

Dimerization of MinD
Probed by FRET-One very useful method for the analysis of protein-protein interactions is fluorescence resonance energy transfer. The specific utility of FRET rests in its ability to provide direct information on molecular proximity. In general, FRET measurements are made by observing the transfer of excited state energy from one chromophore (the donor) to another (the acceptor). In the most favorable instances, FRET can be used to determine accurately molecular distances over a range of 20 -90 Å (26 -28). To address the question of MinD dimerization, we chose MIANS and LY as a donor-acceptor pair because of the very good overlap of their emission and absorption spectra (322/417 and 426/531 nm excitation/emission maxima, respectively). MinD from E. coli contains two cysteines, Cys-52 and Cys-119; the former appears to be accessible for derivatization by these thiol-reactive dyes. Fig. 6 shows that the fluorescent spectrum of the equimolar mixture of MinD fluorescent derivatives in solution (thin continuous line) is identical to the sum of their individual spectra (wide gray line), indicative of the absence of any FRET between them. In contrast, incubation of this mixture with PC liposomes containing 20 mol % CL (heart) caused a decrease of the donor fluorescence and a proportional increase of the fluorescence from the acceptor, which is typical for FRET (Fig. 6, bold continuous line). Moreover, the addition of a 5-fold excess of unlabeled MinD had no further effect on the shape of this spectrum (data not shown).

DISCUSSION
In this study, we investigated the role of anionic phospholipids in the interaction of MinD with the membrane. Our data demonstrate a preferential interaction of MinD with anionic phospholipids accompanied by oligomerization of MinD-ATP on the liposome surface. The observed cooperativity of MinD binding to liposomes, thus indicating oligomerization, is supported by our FRET measurements. The absence of FRET in the mixture of MinD carrying different fluorescent labels (Fig. 6) demonstrates that the protein does not form dimers in solution at least at 1 M concentration. Because only 10 -15% of the total amount of MinD was bound to liposomes (see above), the extent of FRET observed in the presence of anionic liposomes is significant. The appearance of FRET could be simply the result of an increase of the local concentration of the protein on the membrane surface and the consequent increased collision probability. However, when an excess of unlabeled MinD was added to mixtures containing labeled MinD, FRET was unchanged. This suggests that MinD formed stable dimers or oligomers on the liposomes. Our data are consistent with previous results demonstrating self-enhanced binding of MinD to liposomes (23), formation of MinD filament bundles in solution (29), and cryoelectron microscopy images of the MinD tubes (6).
We have found that both the negative charge of the phospho-lipids and the degree of acyl chain unsaturation contribute to binding affinity. These findings are consistent with the model of Szeto et al. (12), in which MinD interacts with the phospholipid membrane by partitioning its C-terminal amphipathic ␣-helix into the membrane bilayer. The helix orientation is likely to be such that hydrophobic residues interact directly with lipid acyl chains, whereas the cationic residues on the opposite face of the helix interact with the head groups of anionic phospholipids. An explanation for the higher affinity of MinD to CL from heart might be that membranes with higher unsaturated acyl chains of phospholipids are packed more loosely and the distance between head groups is greater. This factor, in combination with the higher degree of disorder in unsaturated phospholipid bilayers, might create more favorable conditions for insertion of a short amphipathic ␣-helix oriented parallel to the membrane surface (16,30). Orientation of amphipathic ␣-helices in the membrane can be predicted by a hydrophobic moment plot analysis (31,32). We calculated the sequence hydrophobic moment according to Eisenberg et al. (31) for the putative C-terminal ␣-helix of E. coli MinD based on the putative membrane-targeting sequence of this motif (12,13). The calculated mean of the sequence hydrophobic moment per residue is about 0.6, and the hydrophobicity per residue is about 0.07. On a conventional plot of hydrophobic moment versus hydrophobicity (32), the values place this putative C-terminal ␣-helix of E. coli MinD on the border between a surface-active and a globular peptide. The helix also falls within the area delineating parallel orientation to the plane of the membrane but also is very near the border delineating an oblique orientation. This is consistent with the localization of the C-terminal ␣-helix in the plane parallel to the membrane surface. As such, the hydrophobic residues in-  teract directly with lipid acyl chains, whereas the cationic residues on the opposite face of the helix interact with the head groups of anionic phospholipids. However, a possibility of some oblique orientation is not excluded. This may explain the significant input of hydrophobic interactions into the binding of MinD to liposomes (Ref. 13 and this study), i.e. the significant affinity of MinD for PC liposomes lacking anionic phospholipids.
MinD appears to be associated with a growing set of proteins that are classified as amphitropic and share the same property of reversible binding to membrane lipids. This process regulates their function, and the binding affinity is subject to regulation. Regulatory switches that control membrane affinity include modification of the protein itself, for example, by ligand binding, and modulation of membrane lipid composition. These proteins use an amphipathic ␣-helix to sense the membrane lipid composition. Increases in the proportion of a particular phospholipid in the membrane can trigger protein-membrane binding through the generation of an electrostatic pull to the membrane surface, where the protein searches for intercalation sites (for review, see Ref. 16).
If the membrane binding affinity of MinD depends on lipid composition, changing the protein retention time on the membrane should modulate both spatial and temporal oscillatory characteristics of the protein. The observed aberrant behavior of GFP-MinD in cells lacking PE is consistent with such changes. It has been proposed that the MinD protein forms a lattice on the membrane surface (1). Alternatively, MinD may assemble into two-stranded filaments of limited size, which are dispersed over one-half of the cell membrane in wild-type E. coli (29). We have shown that in PE-lacking cells, in which negative charge density on the membrane surface is very high, MinD polymers are not dispersed over the membrane surface but rather organized in compact clusters. Why might MinD associate in compact areas instead of forming the broad zones observed in PE-containing cells? One possibility is that the increased concentration of anionic phospholipids increases the cooperativity of MinD assembly at the membrane, causing assembly to occur at a higher local MinD concentration and therefore encompassing a smaller area. The time-lapse data suggest that upon disassembly, MinD often reassembles at locations closer than those in normal cells.
The predicted number of MinD monomers per cell with average dimensions of 2 ϫ 0.6 m is 2000 -3000 (33,34). Therefore, the total concentration of MinD in the cell is estimated to be ϳ1-2 M. This is within the range of association constants of MinD binding in our in vitro experiments, suggesting that they are physiologically relevant in terms of MinD concentrations. Therefore, it is reasonable to propose that membrane phospholipid heterogeneity might regulate MinD polymerization events. The mathematical modeling of MinD oscillations in E. coli cells shows that reliable patterning is possible without the need for any prelocalized determinants, but such modeling does not preclude the existence of prelocalized determinants and only demonstrates that they are not necessary for the oscillations. Non-homogeneous distribution of proteins and lipids over the membrane surface resulting in domains enriched in phospholipids has been confirmed experimentally using different fluorescent probes (35)(36)(37)(38)(39) and by analysis of phospholipid composition (24). A higher concentration of CL is found at the cell poles, and CL domains (24, 39) may be important for self-assembly of MinD. The pattern of localization of GFP-MinD in compact zones in PE-lacking cells is consistent with localization of CL domains in this mutant (39). As it was shown recently for CTP:phosphocholine cytidylyltransferase, acidic amino acids of an amphipathic ␣-helix also dictate enzyme selectivity for anionic phospholipids (40). These phospholipids create a low pH environment at the membrane surface that favors protonation of acidic amino acids and consequently decreased repulsion of the helix. CL domains in the E. coli membrane are the best candidates for such a role because they are a natural sink for protons in the membrane (41) and thus could facilitate the membrane insertion of a cluster of glutamic acid residues at one end of the MinD C-terminal motif (12,42). Alternatively, the formation of MinD polymers on the membrane surface may induce the formation of anionic phospholipid microdomain(s). The mutual effect of MinD and anionic phospholipids on MinD polymerization and phospholipid domain formation in the E. coli membrane is under investigation.