The N-Terminal Amphipathic Helix of the Topological Specificity Factor MinE Is Associated with Shaping Membrane Curvature

Pole-to-pole oscillations of the Min proteins in Escherichia coli are required for the proper placement of the division septum. Direct interaction of MinE with the cell membrane is critical for the dynamic behavior of the Min system. In vitro, this MinE-membrane interaction led to membrane deformation; however, the underlying mechanism remained unclear. Here we report that MinE-induced membrane deformation involves the formation of an amphipathic helix of MinE2–9, which, together with the adjacent basic residues, function as membrane anchors. Biochemical evidence suggested that the membrane association induces formation of the helix, with the helical face, consisting of A2, L3, and F6, inserted into the membrane. Insertion of this helix into the cell membrane can influence local membrane curvature and lead to drastic changes in membrane topology. Accordingly, MinE showed characteristic features of protein-induced membrane tubulation and lipid clustering in in vitro reconstituted systems. In conclusion, MinE shares common protein signatures with a group of membrane trafficking proteins in eukaryotic cells. These MinE signatures appear to affect membrane curvature.


Introduction
Targeting of proteins to specific destinations at the appropriate time is crucial for cell function. This process often involves specific protein motifs, and requires the intricate regulation and coordination of different cellular components. Protein targeting is involved in prokaryotic cell division, during which a series of proteins are assembled in a hierarchical order to form a division septum at the correct mid-cell position. An essential component of the division apparatus is the tubulin homolog FtsZ; this is precisely located at the midpoint of the cell, where it forms a ring-like structure underneath the membrane and recruits other division proteins (reviewed in [1]). In Escherichia coli (Ec), the position of the FtsZ ring is regulated by the Min system [2], which is composed of three proteins, MinC, MinD, and MinE; these cooperate to form a dynamic oscillator that guides the placement of the FtsZ assembly. MinC is a negative regulator of the FtsZ ring [3,4], and MinD associates with the cell membrane and undergoes a pole-to-pole oscillatory localization cycle in the presence of MinE and ATP [5,6]. The Min system is a simple but dynamic and functional unit that has received attention from researchers involved in a variety of scientific disciplines [7,8,9]. However, the underlying mechanisms responsible for the membrane-association properties of the Min system require further investigation.
Correct functioning of the Min system involves the formation of membrane-associated polymeric structures of MinD [10,11,12]. MinD accumulates in the membrane at a polar zone at one end of the cell. It associates with the cell membrane as a MinD-ATP complex through its C-terminal amino acids, which fold into an amphipathic helix [13,14]. Upon membrane association, MinD polymerizes into a tightly coiled helix extending from the originating pole almost to the midpoint of the cell [15]. MinE forms a ring-like structure at the mid-cell and stimulates MinD's ATPase activity. This drives its release from the membrane and causes retraction of the leading edge of the MinD polar zone back towards the pole [12]. Recently, we demonstrated that MinE is capable of associating with the cell membrane through its Nterminal domain [16]. A mutant MinE containing residue substitutions at positions R10, K11, and K12 was deficient in membrane binding and unable to support normal MinD/E localization and oscillation cycles; however, MinE's ability to stimulate MinD ATPase activity was unaffected. This suggests that direct MinE interaction with membranes is critical for the functioning of the Min system, and that stimulation of the MinD ATPase activity alone is not sufficient. Interestingly, under a transmission electron microscope, purified MinE caused phospholipid vesicles reconstituted from E. coli lipids to deform into tubules that were surrounded with a discrete coat. These data indicate that MinE can induce membrane deformation, change membrane topology, and provide a physical force. This force may act with ATP hydrolysis in MinD to remove MinD molecules from membranes during the disassembly stage of the oscillation cycle [16].
Examples of protein-induced membrane deformation in prokaryotes are limited. MinD is known to form arrays of helical filaments surrounding membrane tubules [10], but the function of this phenomenon is not fully understood. It was proposed that the dynamics of the FtsZ ring generate a force that constricts the membrane at the division site [17]. In vitro evidence also suggests that the constriction force of the FtsZ ring is caused by filament bending. The intrinsic curvature of FtsZ protofilaments is known to generate bulges and convex depressions in membranes and to deform liposomes following fusion with the amphipathic helix of MinD [18]. The bacterial dynamin-like protein (BDLP) of Nostoc punctiforme showed helical self-assembly and tubulation of a lipid bilayer in vitro, which may represent a transitional stage of BDLPmediated membrane fission and fusion [19,20]. MinD and BDLP share common features of self-assembly on the membrane and nucleotide-mediated conformational changes; however, BDLP is anchored to the membrane by a hydrophobic paddle, while MinD is attached by an amphipathic helix.
In this work, we have identified an additional functional motif of MinE that is associated with MinE-induced membrane deformation. We have provided direct evidence that the extreme Nterminus of MinE from E. coli folds into an amphipathic a-helix when associated with a membrane. This property differed from MinE from Neisseria gonorrhoeae (Ng), which showed a stable Nterminal helix in solution [21]. Meanwhile, we have further monitored MinE-induced membrane deformation using in vitro systems of synthetic giant liposomes and supported lipid bilayers (SLBs) via time-lapse fluorescence microscopy. This MinE-induced membrane deformation required both the earlier identified charged residues R10, K11, and K12 [16] and the amphipathic motif identified in this report. Disturbing the amphipathicity in this region not only led to failure to deform the membrane in vitro, but also caused alterations in protein stability, which may serve as a control mechanism for the regulation of the cellular concentration of MinE. In summary, this study of MinE illustrates the universal mechanisms involved in the targeting of peripheral membrane proteins that are capable of causing membrane deformation; such mechanisms have prokaryotic and eukaryotic origins.

MinE 2-9 inserts into the membranes as an amphipathic helix
To investigate whether other mechanisms besides the electrostatic interaction are involved in mediating the MinE-induced membrane deformation, we analyzed the MinE protein sequence using helical wheel projection programs. We found that residues 2-9 were capable of forming an amphipathic helix of 1-2 helical turns ( Figure 1a). Residues A2, L3, L4, F6, F7, and L8 formed a large non-polar, hydrophobic face, and residues D5 and S9 were located on a hydrophilic surface. The extreme N-terminus of MinE from 11 other bacterial species showed propensities to form amphipathic helices, and had 4-6 residues located on a hydrophobic surface ( Figure S1). The high conservation of amphipathic helix formation was suggestive of its importance, and led us to hypothesize that this amphipathic helix, along with the basic residues R10, K11, and K12 [16], served as a membrane anchor that sustains the peripheral association of MinE.
To explore this hypothesis, we took advantage of the characteristic spectral shift of tryptophan fluorescence emission that occurs as a function of solvent polarity and serves as a measure of peptide-membrane interactions [22]. A single tryptophan substitution was introduced in MinE 1-31 during peptide synthesis to replace residues A2, L3, L4, F6, F7, or L8. A tryptophan residue added to the C-terminus of MinE 1-31 served as a control.  Figure S2). This was higher than the peptides bearing L4W (4.7861.57 nm), F7W (4.0060.88 nm), L8W (4.4460.20 nm), and W32 (3.5661.36 nm) substitutions. Interestingly, peptide MinE 1-12 with the F6W substitution showed a mild blue shift (260.77 nm), indicating an indispensible role for residues 13-31 in stabilizing the peptide-membrane interaction. Taken together, the results suggest that the helical face of MinE, consisting of A2, L3, and F6, forms a hydrophobic surface that is oriented to interact with the hydrophobic regions of the phospholipid bilayer.
A helical conformation of MinE 2-9 is induced upon association with the membrane To further investigate the helix forming ability of MinE and its association with the membrane, we measured the far-UV circular dichroism (CD) spectra of MinE 1-12 and MinE 1-31 in the presence or absence of liposomes (PE:PG:CL = 36:14:50 mol%; Figure 1bd). Interestingly, MinE 1-12 and MinE 1-31 in buffer may have adopted a polyproline II (P II )-like conformation, as suggested by strong negative values near 200 nm and elevated readings at 220 nm in the spectra (Figure 1b). The P II conformation is a lefthanded threefold helix of nominally unordered peptides in their charged forms. By the addition of 50% trifluoroethanol (TFE), which is known to stabilize the helical structures of proteins and peptides, spectra of both MinE 1-12 and MinE 1-31 showed characteristic features of a high helical content, i.e. the troughs around 208 and 222 nm (Figure 1b). MinE 1-12 showed typical features of high helical contents when 100 mM liposomes were added in the reaction (Figure 1c). This further expanded a previous theory that a nascent helix of MinE  in solution [23] may be stabilized by interacting with the cell membrane. We also detected significant changes in the CD spectrum of MinE 1-31 with liposomes ( Figure 1d), but the overall secondary structure was more complicated. Part of the reason may be because of aggregation of the peptide when associated with the liposomes [16], as indicated by reduction of the signal. In summary, our results suggest that the extreme N-terminal region of MinE has a strong propensity to fold into a helix during membrane association.

Molecular dynamics simulation of interactions between MinE 2-12 and membranes
In addition, we used the molecular dynamics simulation to model how MinE 2-12 was positioned in the membrane (Figure 2, S3). We studied MinE 2-12 because the first methionine residue of MinE was cleaved off in E. coli [16]. The starting model of MinE 2-12 was constructed based on the NMR structure of NgMinE 2-12 , in which residues 2-8 showed an a-helical conformation and the rest of residues are in a loop region [21]. The procedure of adding a virtual membrane of 30 Å thickness generated a model of the peptide sitting at the interface region of the membrane. Information from the tryptophan blue shift assays allowed us to manually adjust the orientation of the MinE 2-12 molecule so that the side chains of A2, L3, and F6 were positioned in the membrane in the initial model. The side chains of D5, S9, and R10 were also positioned in the membrane through this operation ( Figure S3a: starting model). This peptide-membrane complex was then simulated using an implicit solvent model, as suggested for studying the peptide-membrane association [24,25].
The conformation trajectory of a 10 ns simulation ( Figure 2, S3) suggested that the major conformational changes occurred in the loop region, where the side chains of R10 and K11 were repositioned out and in the membrane, respectively (Figure 2a, S3a). The side chains of residues 2-8 showed constant locomotion because of their interactions with the membrane environment, but their relative orientations to the membrane were unchanged. The charge coming from the side chain of D5 was neutralized by the formation of a salt bridge with the N-terminal amino group of A2. The conformation trajectories also suggested that the interface localization of MinE 2-12 was maintained by hydrophobic interactions between side chains of A2, L3, and F6 and the membrane (Figure 2, S3a). The benzyl group of F6 appeared to insert deeper into the phospholipid bilayer. The presence of side chains of D5, S9, and K11 in the membrane may be explained by polar interactions with the head groups of the bilayer. This simulation provided a specific view of the folding and positioning of MinE 2-12 when associated with the membrane. It should be noted that the simulation process did not account for the bending flexibility of the membrane; in reality, insertion of such a helix into a membrane is likely to induce bending [26].

MinE induced liposome deformation in real-time
MinE was found to induce liposome deformation in association with direct MinE-membrane interactions [16]. To better characterize this deformation process, and establish the correlation between insertion of an amphipathic helix and membrane deformation, we set up an imaging system to simultaneously   (Figure 3a, S4a). In each time-lapse sequence acquisition, we imaged an isolated liposome for a short period of time before addition of the protein.
In reactions with wild-type MinE, the periphery of the spherical liposomes gradually became coated with green fluorescence (Figure 3a; yellow arrow). Moments later, the liposomes either suddenly burst or gradually deformed into membrane tubules from a confined area (Figure 3a; white arrow). Membrane tubules emanating from a liposome were also observed with an electron microscope (Figure 3b,c). Fluorescent MinE was colocalized with the membrane tubules ( Figure 3a; green arrow), indicating that tubule formation is associated with MinE. We also examined reactions that were incubated for 10 min prior to mounting on clean glass slides. In these experiments, adsorption of lipids to the glass surface simplified the imaging process. MinE colocalized with various parts of the lipid tubules and liposomes (Figure 3d; white arrow), or concentrated at the tips of the tubules (Figure 3d; green arrow). This is consistent with electron microscope observations, which showed electron dense caps on buds sprouting from liposomes (Figure 3e; arrows). Membrane deformation and tubulation have been associated with some membrane trafficking proteins in eukaryotes (Table S1); here, we demonstrate that a prokaryotic protein has the same activity.

MinE 1-31 is fully capable of inducing membrane tubule formation
Interestingly, we found that N-terminal MinE 1-31 was able to induce membrane deformation of the giant liposomes in our experimental setup (Figure 3f, S4b). The initiating points of the deformation process were significantly different for the wild-type protein and MinE 1-31 . The full-length protein induced liposome deformation at a focal point ( Figure 3a; white arrow); MinE 1-31 initiated tubule formation around the entire periphery of the liposome ( Figure 3f). These data suggest that the C-terminal domain of MinE is required for localizing the deformation activity to a specific area of the membrane environment. This may involve dimerization [27] or a higher-order pattern of organization of the C-terminal domain [28].
The control experiment showed that no membrane deformation occurred with the addition of

MinE-induced deformation of the supported lipid bilayers
We further examined MinE-induced membrane deformation using supported lipid bilayers (SLBs) prepared with E. coli polar lipids (PE:PG:CL = 65:25:10 mol%; Figure 4). The fluidity of the bilayer was demonstrated to show its functionality under our experimental conditions ( Figure S6). Before addition of the protein we identified an area on the labeled SLBs that showed even  unchanged, but the intensity of others increased as they developed into tubules (Figure 4a) or spread laterally to form larger fluorescent patches (Figure 4b). These patches may have originated as membrane tubules laid down on the mica surface during image acquisition, and subsequent enlargement might have been due to the diffusion of phospholipids from accumulation sites. Membrane tubules induced by MinE were coiled and bent ( Figure  S7a), this differed from the smooth contour of those caused by the external force of buffer purposely blown over the SLBs ( Figure  S7b). The images of fluorescently labeled MinE colocalized with membrane tubules indicated that tubule formation was associated with MinE ( Figure S7a).
Replacing wild-type MinE with MinE 1-31 in the SLB experiments resulted in the formation of fluorescent foci, but no obvious membrane tubules were seen (Figure 4c, S7c,d). Atto488-labeled anti-MinE antibody was used to identify MinE 1-31 on the fluorescent patches. MinE 1-31 was found at the vicinity of the lipid patches, but was not completely superimposed on them ( Figure S7c; cyan arrow). We also identified arcs ( Figure S7c; yellow arrow) and enclosed rings ( Figure S7c; white arrows) of MinE 1-31 surrounding larger lipid patches. These data suggest that the association of MinE 1-31 with membranes resulted in the local accumulation of surrounding phospholipids. The number of phospholipids between the accumulation points significantly decreased and contributed to the reduction in background fluorescence (Figure 4b,c). The differences between MinE 1-31 induced membrane deformation of giant vesicles and SLBs may reside in the continuity of the lipid supplies. Lipids were continuously drawn into the growing tubules in the giant vesicles until transformation was complete. The initiation points for tubule formation on SLBs were scattered and lipids were drawn independently into separate foci. This resulted in a shortage of lipids, which was not able to support tubule growth. These data indicate that MinE is able to cause membrane deformation and induce tubule formation in a flat membrane, which further confirms our observation using the giant liposome system.

Importance of MinE 1-12 in membrane association and protein stability
We constructed a mutant MinE protein by substituting F6 with aspartic acid to weaken the amphipathicity of MinE 2-9 . In the sedimentation assays, the purified mutant protein MinE F6D only retained 45% of the ability to co-sediment with liposomes (PE:PG:CL = 36:14:50 mol%; Figure 5a,b), indicating the importance of this residue in supporting the protein-membrane interaction. The remaining hydrophobic residues, A2 and L3, and the charged residues R10, K11, and K12 may have sustained part of the interaction. In addition, the large hydrophobic face might have allowed the mutant helix to rotate and associate with the membrane. Time-lapse fluorescence microscopy was used to examine liposome deformation induced by the mutant MinE proteins C1 (R10G/K11E/K12E) and MinE F6D , which were defective in membrane association. We acquired images for a minimum of 20 min for each experiment and observed five liposomes . 15 mm in diameter for each mutant protein. All five liposomes studied for wild-type MinE showed complete (4/5) or partial (1/5) deformation ( Figure S4a); the partially deformed liposomes were likely to progress to full deformation. There was no liposome deformation with the C1 and MinE F6D mutant proteins ( Figure S8a,b). Interestingly, although MinE F6D retained approximately half of the membrane binding activity in the sedimentation assay (composition of liposomes PE:PG:CL = 36:14:50 mol%), it failed to bind and deform liposomes (PE:PG:CL = 65:25:10 mol%) under the fluorescence microscope ( Figure S8b,c). We conclude that the C1 and MinE F6D mutant proteins are defective in both membrane-association and liposome deformation.
The pSOT169 (P lac -yfp::minD minE A2E/L3S/F6D ::cfp) construct was generated to further investigate the physiological relevance of the extreme N-terminal helix. The triple mutant was created because the single substitution mutant F6D still retained approximately half of its membrane association ability, even though it failed to deform liposomes. This resulted in no significant changes in MinDE localization when the mutant MinE F6D was expressed in cells. The defect detected in the sedimentation assay may be overcome by the complexity of the cellular environment, including MinD's recruitment of MinE to the membrane location and enrichment of cardiolipin at the division site. When the triple mutant MinDE A2E/L3S/F6D expression was induced in a Dmin strain YLS1, MinD was delocalized from the polar zone into a peripheral pattern and MinE A2E/L3S/F6D was dispersed or accumulated as punctuates in the cells (Figure 5c). Western blot analysis detected a low abundance of the MinE A2E/L3S/F6D -CFP fusion protein in cells, indicating that the mutant protein was unstable. This instability was more severe than that of the C1 mutant, which was stable when fused to CFP, but unstable when expressed alone [16]. Although the results did not allow us to draw an apparent link with cellular localization, they suggest that proper folding of MinE 2-12 and membrane association may serve as a control mechanism for the regulation of the cellular concentration of MinE, which is critical for sustaining the oscillation cycles of the Min proteins [29].

Discussion
Amphipathic helices are widely found in proteins participating in membrane-associated biological activities, such as vesicle trafficking, viral fusion, and toxin-induced membrane lysis. The amphipathic nature of the helix serves as a membrane-anchoring motif that locates near the interface region of the cell membrane, often leading to modification of the protein function and the membrane properties.
A generalized mechanism for peripheral membrane association has been proposed [30]. Primary adsorption of a protein onto a membrane is facilitated by non-specific charge interaction and diffusion. This is subsequently stabilized through membrane penetration by protein motifs and binding to specific lipids. Here, we identified the necessary elements in MinE that fulfill this paradigm. In addition to the charged residues R10, K11, and K12 characterized in our previous study [16], MinE 2-9 has the tendency to fold into an amphipathic helix upon association with a membrane, as determined by the circular dichroism measurements. Tryptophan blue shift assays suggested that the helical face of residues A2, L3, and F6 are positioned in the membrane. The molecular dynamics simulation provided information on the peptide-membrane interaction, which showed specific conformations when it encountered the cell membrane. Meanwhile, deeper insertion of the side chain of F6 may act as a structural landmark to effectively create membrane defects or to target membranes with positive curvature.
To support the importance of the MinE 2-9 helix for proper function of the Min system, the mutant protein MinE F6D was engineered to weaken the amphipathicity, which significantly reduced the ability of MinE to associate with membranes in vitro. A triple mutant MinE A2E/L3S/F6D affected the protein stability in vivo. An unbalanced ratio of MinD to MinE resulted in mislocalization of the proteins [29]. Interestingly, an earlier study showed that an N-terminally truncated MinE (MinE 6-88 ) retained its ability to suppress division inhibition by MinCD, but still resulted in a minicelling phenotype [31]. This indicates that the extreme Nterminus of MinE is important for the function of the Min system, but does not affect the interaction of MinE with MinD. Therefore, the membrane anchoring mechanism of MinE, including the charge interaction, the formation of an amphipathic helix of MinE 2-9 , and the preference for cardiolipin [16], is independent of the mechanism that regulates the interaction of MinE with MinD.
A recent solved NMR structure of the full-length MinE from N. gonorrhoeae showed that the N-terminal helix of residues 228 is exposed and connected by an extended loop region to the integral part of the MinE dimer (PDB code: 2KXO) [21]. This structure suggested that the N-terminal amphipathic helix is highly flexible for interactions with other binding partners. Interestingly, the hydrophobic face of the helix, which may be involved in the membrane interaction, was oriented away from the protein surface, suggesting a rotation of the helix is necessary for association with a membrane. Previously, structure determinations of the extreme N-termini of the MinE proteins from E. coli and Helicobacter pylori were inconclusive [23,28], which was possibly due to the nature of the MinE proteins from different bacterial species. Our current study demonstrated that the helical conformation of EcMinE 2-9 was stabilized by the presence of the membrane. Based on the structure information of NgMinE, we modeled the structure of EcMinE for a suggestive view of the N-terminal domain when it forms (Figure 6a2c). In this model, most residues (A2, D5, F6, S9) on the membrane interacting face of the N-terminal helix of EcMinE is exposed on the protein surface and appears accessible for membrane interactions. Therefore, the control mechanism for EcMinE interaction with a membrane may rely on the induced folding property and an interaction between the N-terminal and C-terminal domains to sequester the membrane interaction [16]. Moreover, the similarity of the side chain orientations of residues A2, L3, D5, F6, and S9 in both the molecular dynamics simulation model and the predicted model based on NgMinE, indicated the reliability of the approaches. It will be interesting to see whether targeting the N-terminal domain of MinE to the membrane may trigger conformational changes that expose the MinD interacting sites located on the b-face of the MinE dimer.
The in vitro membrane deformation activity of MinE reported in this study, including budding, tubulation, and lipid clustering, is similar to that of several proteins involved in membrane trafficking in eukaryotic cells. Membrane trafficking is a process that allows membranes from different sources to exchange their lipids, proteins, and interior contents. Examples include dynamin, Bar domain proteins (amphiphysin, epsin, endophilin, and nexin), ENTH domain proteins (epsin, Ent3, Ent5), Arf, SarI, Septin, and C2 domain proteins (such as synaptotagmin) (Table S1). Common features shared by these membrane-associating proteins include (1) an amphipathic helix or simply a hydrophobic surface that can insert shallowly into a membrane bilayer, (2) a patch of charged residues that support electrostatic interaction with the membrane, and (3) the unique folding of specific protein domains or a curved shape maintained through self-association to sculpt the cellular membranes [32][33] (Table S1). Interestingly, although some of these proteins possess nucleotide triphosphatase activity, there is no evidence, thus far, to demonstrate coupling of nucleotide hydrolysis with induction of membrane tubule formation.
In this study, we present evidence that MinE, the topological specificity determinant of the E. coli's divisome, has membrane deformation activity in vitro and possesses signature motifs relating to membrane deformation. By analogy to other membrane curvature sensing and induction mechanisms, we propose a model to explain the involvement of an amphipathic helix in the MinE protein-membrane interaction and MinE-induced membrane deformation. The insertion of the amphipathic helix of MinE into membranes may lead to a local change in curvature that acts as an initiation point for membrane deformation (Figure 6d). This local change in curvature may be propagated through the accumulation of high concentrations of MinE by a process that may or may not involve self-association of MinE, and results in drastic membrane deformation. The enrichment of cardiolipin at the division site of an E. coli cell [34,35] and MinE's higher affinity to anionic phospholipids [16] may in turn contribute to formation of a MinE ring at the midcell, which arrests growth of the MinD polar zone [12]. Both MinE's ability to stimulate MinD ATPase activity and to deform the membrane may contribute to removal of MinD from the membrane location.
In contrast to the list of eukaryotic proteins that possess membrane deformation properties, to our knowledge, MinE, MinD, and BDLP are the only documented prokaryotic proteins that have in vitro membrane tubulation activities [10,16,20]. These findings indicate that protein-mediated membrane remodeling may occur in bacteria. Although the purpose of such an activity in prokaryotes is not yet fully understood, the activity may contribute to membrane recycling and restructuring during cell growth and development. In plant and animal cells, evidence suggests that membrane trafficking can act as a developmental control during cleavage furrow formation and abscission of daughter cells [36]. Membrane trafficking may involve delivery and sorting of cargo, and deposition of membranes that are linked to the dynamics of the cytoskeleton. The only known equivalent membrane trafficking systems in prokaryotes are an ESCRT-like machinery in wallless Crenarchaea that has been correlated to membrane abscission during cell division [37,38], and a simple endocytotic system in Gemmata obscuriglobus [39]. Further investigations are required to determine whether the protein-induced membrane deformation contributes to effective removal of incorrectly placed septal machinery, and serves as a developmental control in bacteria.

Tryptophan blue shift assay
The tryptophan blue shift assay was conducted by incubating 6 mM MinE 1-31 carrying a tryptophan residue at various positions, and 10 mM liposomes (with a diameter of 400 nm) in buffer A (20 mM Tris-Cl, pH 7.5; 200 mM sucrose), at room temperature for 10 min. Control reactions were incubated without liposomes. The mixtures were excited with 280 nm UV light and scanned for Step 1: MinE can directly target existing defects on membranes through its N-terminal amphipathic helix (residues 2-9) and the adjacent basic residues (R10, K11, K12). Alternatively, MinE may directly target to a membrane and cause a membrane defect to occur. Both membrane defects and high concentrations of anionic phospholipids will stabilize the initial protein membrane interaction. Step 2: A ''wedging effect'' on the membranes occurs when an amphipathic helix intercalates into the membranes.
Step 3: Localized high density wedges due to self-association of MinE or accumulation of large numbers of wedges can lead to the propagation of membrane defects and more drastic changes in local curvature. This process will overcome an energy barrier and destabilize the membranes, leading to tubule formation. doi:10.1371/journal.pone.0021425.g006 fluorescence emission at 300-400 nm on a Fluorolog-3 spectrofluorometer (HORIBA Scientific, Inc.). The statistical value of the blue shift at the maximal emission wavelength was averaged from at least three independent experiments; three continuous scans were repeated in each experiment. We found that the fluorescence intensity gradually decayed during continuous scans of the same sample, thus we did not use fluorescence intensity as an indication of oligomerization.

Circular dichroism (CD)
The MinE 1-31 and MinE 1-12 peptides were dissolved in 20 mM Tris-Cl, pH 7.5 and purified by passing through a 0.22 mm filter and diluted to suitable concentrations before the experiments. Sucrose generates a strong spectral signal at 190 nm; therefore, buffer A was not suitable for this experiment. CD spectra of the peptides were measured in the far UV range (190-250 nm) on a JASCO J-715 spectrometer (JASCO, Japan). The bandwidth and the step resolution were set to 2 nm and 0.2 nm respectively. A quartz cuvette was cleaned by soaking in potassium dichromate and rinsed before use. The optical path of the cuvette was 0.1 cm. For each sample, three scans were performed to obtain an averaged spectrum; this was subtracted from the spectrum of the buffer to provide a baseline correction. When appropriate, 100 mM liposomes were supplied in the reaction.

Molecular dynamics simulation
The molecular dynamics simulation study was performed using the Discovery Studio 2.5 (Accelrys Inc., San Diego, CA, USA). The starting model of MinE 2-12 was constructed by replacing the amino-acid side chains of NgMinE 2-12 (PBD code: 2KXO) with the side chains at EcMinE 2-12 . The CHARMm Polar H force field was applied to the molecule for subsequent simulation. Prior to simulation, a virtual membrane of 30 Å -thickness was added to the molecule using an Implicit Solvent Model GBSW (Generalized Born with a simple SWitching). This step created a model of the EcMinE 2-12 helix sitting on the interface region of the membrane. The helical face containing A2, L3, and F6 was then manually rotated to face down the membrane, based on the knowledge learned from our experimental data. The resulting model was simulated using the ''Standard Dynamics Cascade'' protocol consisted of steps of two rounds of minimization, heating, equilibration, and production for 10 ns. During simulation, the backbone carbon atoms of MinE 2-9 were constrained by a harmonic force with a constant of 1 kcal mole 21 Å 22 . The heating temperature was set between 50 to 300uK. All other settings for the simulation were the same as those for adding a membrane.

Strains and Plasmids
Strains and procedures for overproduction and purification of MinE, preparation of giant liposomes, and electron microscopy were as previously described [16]. pSOT164 [P T7 ::minE F6D -his] was generated for protein overproduction by introducing a point mutation by a long-range PCR reaction with pSOT13 [16] as the template DNA. pSOT169 [P lac ::yfp-minD minE A2E/L3S/F6D -cfp] was constructed by subcloning minE F6D from pSOT164 into pYLS68 [P lac ::yfp-minD minE-cfp] [12] followed by a long-range PCR reaction to introduce additional point mutations.

Fluorescence microscopy
For observing giant liposome deformation, glass slides and cover slips were cleaned by sonication in ddH 2 O, ethanol, acetone, 1 M KOH, and ddH 2 O sequentially for 30 min each, before being soaked in methanol and dried before use. An o-ring was placed on each clean slide to create a sample-holding chamber, and 100 ml 20 mM Tris-Cl, pH 7.5 were added followed by 100 ml of liposome suspension in buffer A. An upright Olympus BX61 microscope equipped with Chroma ET-mCherry and ET-GFP filter sets, a Hamamatsu Orca-AG Cool Charge-Coupled Digital camera, and Volocity (Improvision, PerkinElmer) was used for image acquisition and analysis. A water-immersion objective (Olympus LUMPlanF1 60X/0.9W) was attached to the microscope to view an isolated liposome for a few minutes prior to protein addition. Liposomes were labeled by the addition of 0.2 mol% Texas Red-DHPE (Texas Red 1,2-dihexadecanoyl-snglycero-3-phosphoethanolamine, Invitrogen), and purified MinE or MinE F6D was labeled using Alexa FluorH 488 reactive dye with a tetrafluorophenyl (TFP) ester moiety (Invitrogen). The degree of labeling was estimated according to the manufacturer's instructions; every protein molecule was averaged to carry 0.2-0.7 fluorescent dye molecules (mol dye/mol MinE). Liposomes formed in buffer A were diluted two fold in 20 mM Tris-Cl, pH 7.5 in the observation chamber. Protein was added to a final concentration of 6 mM from the side of the chamber. The protein was allowed to diffuse toward the targeted liposome. Simultaneous imaging using both the Gfp and mCherry channels was conducted for at least 20 min or until the liposomes burst. Time zero was defined as the first frame acquired after the addition of the protein. The acquisition interval was fixed as 6 s, unless otherwise specified. The acquired image sequences were processed in Volocity, Matlab, and/or Photoshop for figure presentations. Additional reactions were incubated for 10 min before spotting on the glass slides for single time point observations.

Preparation of supported lipid bilayers
A total of 0.5 mg/ml E. coli polar lipids (Avanti) mixed with 0.4 mol% Texas Red-DHPE in chloroform in a small glass vial were dried under nitrogen and then in a vacuum for an additional 1-2 h. The dried lipid layers were rehydrated in 1 ml buffer A and kept in the dark with intermittent gentle shaking for an hour at room temperature. The liposome suspension was subjected to five to eight freeze-thaw cycles of 1 min in liquid nitrogen and 5 min in water at room temperature. The freezing step fragmented the bilayers, thus enhancing the reformation of unilamellar vesicles in the thawing step. The vesicle suspension was passed through an extruder with filters of pore size 400 and 100 nm, in sequential steps of 21 passages each, to generate small unilamellar vesicles (SUVs) of uniform size. The SUV suspension was diluted 10 fold with buffer A and applied to a chamber with a freshly cleaved piece of mica mounted on a glass slide. After incubation at 37uC for 30 min, 20 mM CaCl 2 and 100 mM NaCl were added to the chamber to facilitate vesicle fusion and adsorption onto the mica. After incubation at 37uC for an additional 30-60 min, the suspension was carefully drawn out followed by four gentle washes with buffer A. The supported lipid bilayer (SLB) on the mica was immersed in 200 ml buffer A. All studied areas of SLBs were observed for one minute before addition of purified MinE (final concentration, 24 mM) or synthesized MinE 1-31 peptide (24.5 mM) using the microscopy system described previously. The image sequences were acquired continuously at 3-s intervals without stopping when the proteins were applied. The acquired image sequences were processed as previously described. Fluorescence intensity maps were generated in Matlab; 16-bit images (grey scale range, 0-65535) were analyzed and the color bar was normalized using the lowest and highest intensity values in each image sequence.

Immunofluorescence detection of MinE 1-31 on the deformed bilayer
We identified the presence of MinE 1-31 on the lipid clusters by hybridization. We purified crude anti-MinE antisera by adsorbing anti-MinE antibodies onto purified MinE proteins immobilized on a PVDF membrane by SDS-PAGE separation and western blotting. The adsorbed antibodies were stripped off the membrane in 1 ml pre-chilled 0.2 M glycine, pH 2.5 with gentle shaking for 45 s. The purified antibody solution was immediately neutralized with 1 ml 1 M Tris, pH 9.0 and concentrated to 1.2 mg/ml. This was then conjugated with Atto488 following the manufacturer's instructions [Lightning-Link TM Atto488 Conjugation Kit; Innova Biosciences]. The Atto488-conjugated antibody was exchanged into buffer A before use.
For probing MinE 1-31 on the SLB, Texas Red-DHPE in the SLB recipe was reduced to 0.04 mol%. We followed the previously described protocol to induce membrane deformation, then slowly withdrew all the solution from the chamber to remove unbound proteins, and immediately applied 200 ml fresh buffer A carefully back into the chamber. A control experiment was performed in parallel with buffer in place of the MinE solution. In both the test and control samples, 10 ml Atto488 conjugated anti-MinE antibodies were added into the chamber. The chambered slides were placed in a moisture box and incubated at 4uC overnight with gentle shaking. Prior to image acquisition, the bilayer was washed by slowly withdrawing 150 ml solution and replacing with the same volume of fresh buffer A. This step was repeated five times to remove excess antibody. Samples were then ready for image acquisition.