Dynamic Interaction of the Escherichia coli Cell Division ZipA and FtsZ Proteins Evidenced in Nanodiscs

Background: ZipA provides membrane tethering to septation FtsZ protein.

Results: ZipA in nanodiscs moderately binds FtsZ oligomers and polymers equally. FtsZ-binding sequence peptides inhibit binding. The transmembrane ZipA segment has no role in ZipA·FtsZ complex formation.

Conclusion: Tethering of FtsZ to the membrane through ZipA shows plasticity.

Significance: Acellular system partly reproduces assembly of cell division components.

Abstract

The full-length ZipA protein from Escherichia coli, one of the essential components of the division proto-ring that provides membrane tethering to the septation FtsZ protein, has been incorporated in single copy into nanodiscs formed by a membrane scaffold protein encircling an E. coli phospholipid mixture. This is an acellular system that reproduces the assembly of part of the cell division components. ZipA contained in nanodiscs (Nd-ZipA) retains the ability to interact with FtsZ oligomers and with FtsZ polymers. Interactions with FtsZ occur at similar strengths as those involved in the binding of the soluble form of ZipA, lacking the transmembrane region, suggesting that the transmembrane region of ZipA has little influence on the formation of the ZipA·FtsZ complex. Peptides containing partial sequences of the C terminus of FtsZ compete with FtsZ polymers for binding to Nd-ZipA. The affinity of Nd-ZipA for the FtsZ polymer formed with GTP or GMPCPP (a slowly hydrolyzable analog of GTP) is moderate (micromolar range) and of similar magnitude as for FtsZ-GDP oligomers. Polymerization does not stabilize the binding of FtsZ to ZipA. This supports the role of ZipA as a passive anchoring device for the proto-ring with little implication, if any, in the regulation of its assembly. Furthermore, it indicates that the tethering of FtsZ to the membrane shows sufficient plasticity to allow for its release from noncentral regions of the cytoplasmic membrane and its subsequent relocation to midcell when demanded by the assembly of a division ring.

Introduction

Bacterial division involves the assembly of a macromolecular complex, the divisome, formed by several proteins, 10 of them essential (1). The nature of a large part of the divisome components is known, as it is their assembly pathway and some of the biochemical activities and protein-protein interactions involved (1, 2). The elements of the divisome follow an assembly pathway in which both sequential and concerted stages are involved. Initially, three proteins, FtsZ, FtsA, and ZipA, assemble together, forming a proto-ring into which the other components are added (1). The Z-ring is highly dynamic, and its position is regulated by two negative control systems that inhibit ring formation at wrong places (3). FtsZ is among the most phylogenetically conserved cell division proteins and is present in most bacteria. It is a 40-kDa soluble GTPase homolog of eukaryotic tubulin, whose GTP-linked assembly-disassembly cycle is thought to be important for Z-ring formation (Refs. 4 and 5 and references therein). ZipA is a 36.4-kDa protein that contains an amino-terminal helix that is integrated into the membrane and connected to a cytoplasmic FtsZ-interacting domain via a flexible linker (6). For the attachment to the membrane, FtsA and ZipA are interchangeable to some extent, but no localization of FtsZ at the membrane occurs in the simultaneous absence of both (7).

As most of the available knowledge has been derived from the behavior of mutants, in which one or more components of the divisome have been genetically modified or impaired, it becomes essential to corroborate the commonly accepted hypothesis using a bottom-up synthetic approach. One major advantage of in vitro reconstitution experiments in biomimetic membranes is that the biochemical parameters can be controlled precisely, providing the opportunity to investigate complex membrane-associated reactions under defined experimental conditions (8, 9). Along these lines, the septum placement mechanism, mediated by the MinCDE proteins, has been partly reconstructed on a surface (10) and an artificially membrane-attached FtsZ has been shown to mimic rings assimilated to the Z-ring in artificially flattened and elongated liposomes (11), whereas the interacting FtsZ-FtsA components of the proto-ring have been introduced onto giant unilamellar vesicles obtained from natural membrane, and their interaction has been studied upon the addition of GTP (12).

Our knowledge on the association of the proto-ring proteins FtsZ and ZipA is up to now mainly based on genetic studies (2), with only a limited number of structural and biophysical studies on ZipA-FtsZ associations being available (13–16). Previous results to measure the binding of a variant of ZipA, lacking the transmembrane region (sZipA) to FtsZ-GDP (16), indicated that independently of the oligomer size (up to hexamers), the binding affinity was moderate, and only one sZipA was bound to any of the different FtsZ species present in the solution.

We describe here a synthetic system based on the integration of ZipA, one component of the division proto-ring, in nanodiscs. These structures are formed by a membrane scaffold protein encircling a phospholipid bilayer (17). When a target membrane protein (e.g. ZipA) is included in the phospholipid mixture, each nanodisc self-assembles and incorporates the target molecule, preserving its structural and functional properties. Even if they provide a topologically restricted environment, nanodiscs remain soluble, allowing the use of a broad arsenal of biochemical and biophysical tools to quantitatively characterize the embedded protein and its interactions.

Incorporation of a single copy of full-length ZipA in nanodiscs has allowed us to measure the binding of Nd-ZipA⁴ to different forms of FtsZ assembled in the presence of GDP, GTP, or GTP analogs and compare them with the results obtained for the binding of a soluble variant of ZipA, lacking the transmembrane region, to FtsZ. In addition, we have measured the competition of two peptides derived from the C terminus of FtsZ in the binding of Nd-ZipA to FtsZ polymers.

EXPERIMENTAL PROCEDURES

See the supplemental Methods for a detailed description of the production of the proteins used in this study (membrane scaffold protein, FtsZ, and ZipA), the incorporation of ZipA into nanodiscs, and the synthesis of inhibitor peptides. Assays (sedimentation velocity and equilibrium, fluorescence correlation spectroscopy, and dynamic light scattering) are also described in the supplemental Methods.

Model 1: Binding of Isodesmically Oligomerizing FtsZ-GDP to ZipA Incorporated in Nanodiscs

It is known that FtsZ (denoted here by Z) undergoes isodesmic self-association in the presence of GDP (18, 19): Z_i−1 + Z ↔ Z_i, with equilibrium association constant K_ZZ for all i. Let any i-mer of Z bind to a molecule of ZipA embedded in a nanodisc (denoted here by A) with equal association constant: Z_i + A ↔ AZ_i, with equilibrium association constant K_AZ for all i.

It follows that

for all i and

for all i, where c_Z and c_A denote the concentrations of monomeric Z and A, respectively.

The conservation of mass is expressed as

Given the values of c_A,tot, c_Z,tot, K_ZZ, and K_AZ, Equations 3 and 4 can be solved numerically to obtain the values of c_A and c_Z. Given these values, Equations 1 and 2 can be used to calculate the values of all c_Zi and c_AZi. Then the average number of molecules of FtsZ bound per molecule of NdZipA can be calculated according to

A MATLAB model script was written for fitting this model to the sedimentation equilibrium data (supplemental Equation S1). Using the best-fit parameter values, it is possible to calculate the fractional distribution of bound Nd-ZipA to all the species containing FtsZ with significant abundance (as it is plotted in Fig. 5C).

FIGURE 5.

Fluorescence correlation spectroscopy analysis of Nd-ZipA binding to FtsZ polymers in the presence of GMPCPP. A, normalized FCS autocorrelation profiles of Nd*-ZipA (45 nm tracer, 1 μm total Nd-ZipA) in the absence (empty circles) and in the presence (solid circles) of FtsZ (50 μm). Lines correspond to the best fit of Equations S2 and S3 to the data as described under supplemental Methods. The same results were obtained in the absence of the unlabeled Nd-ZipA. B, fraction of Nd-ZipA bound to FtsZ polymers as a function of FtsZ concentration measured by FCS. The line represents the best fit of Equations 8 and 9 with the best-fit parameter of log K_AZ = 6.3. Error bars indicate S.D. C, molar fractions of bound Nd-ZipA as a function of the concentration of the average FtsZ oligomer for GDP (solid line), GTP (dashed line), and GMPCPP (dotted line) at 1 μm Nd-ZipA, assuming 2, 100, and 140 FtsZ molecules per FtsZ-GDP, -GTP-, or -GMPCPP oligomer, respectively. The distributions were calculated using the best-fit parameters values given in the legend for Fig. 3B (GDP) and for panel B of the current figure (GMPCPP and GTP). To facilitate the comparative analysis, the fraction of bound Nd*-ZipA to FtsZ-GDP represents the sum of all the species of Nd-ZipA containing FtsZ with significant abundance.

Model 2: Binding of FtsZ Polymers to Nd-ZipA

To determine the association constant for the interaction of Nd-ZipA with FtsZ polymers, the FCS binding data (see Fig. 5B) were fitted with the following simple mechanistic model. It was assumed that for a given total FtsZ concentration, FtsZ may exist as an equilibrium mixture of monomer and a single species of oligomeric fibrils with stoichiometry n_olig. The equilibrium weight fraction of FtsZ existing as oligomers as a function of [FtsZ]_tot has been previously determined by hydrodynamic and scattering measurements (16, 20). The molar concentration of oligomeric fibril is equal to

Assuming that a single fibril may bind to Nd-ZipA with affinity constant K_AZ, it follows that the fraction of Nd-ZipA bound to FtsZ will be

In the limit that all of the FtsZ exists as the oligomer, Equation 7 reduces to

where

It must be noted that Equations 7–9 are valid only when a single fibril binds only a single Nd-ZipA. In principle, a fibril can bind more than one Nd-ZipA, and in fact, we have found that it does so at high total Nd-ZipA concentrations (see below). However, we have demonstrated that under the conditions of our FCS experiments (i.e. low total Nd-ZipA concentration), a maximum of one Nd-ZipA is bound per oligomeric fibril (see below). It follows that Equation 9 may be utilized to model the FCS data and obtain a valid equilibrium association constant.

RESULTS

Incorporation of a Single Copy of ZipA Protein into Phospholipid Bilayer Nanodiscs

The complete ZipA protein, which contains a single transmembrane helix, was incorporated into 10-nm nanodiscs (21) formed by an Escherichia coli polar lipid extract wrapped by a dimer of the membrane scaffold protein variant (MSP1D1). This environment reproduces the lipid composition found in the E. coli inner membrane. Upon detergent removal, the resulting mixture was fractionated by gel filtration to separate the ZipA-containing nanodiscs from empty nanodiscs and free MSP1D1 protein (Fig. 1A). Electrophoretic analysis showed that the main peak, eluting at around 11 ml, contained the fractions with ZipA and MSP1D1, whereas the following two peaks corresponded to empty nanodiscs and MSP1D1 (Fig. 1B). To obtain nanodiscs containing one single ZipA molecule (Nd-ZipA), the fractions containing a ZipA:MSP1D1 ratio close to 1:2 according to densitometry were pooled and fractionated again in a second gel filtration step. The purity of the Nd-ZipA complexes was checked using native gel electrophoresis (Fig. 1C). A single band was found with a migration rate slower than empty nanodiscs or MSP1D1. The samples containing ZipA not incorporated into nanodiscs could not enter into this gel as they were presumably formed by large aggregates.

FIGURE 1.

Analysis of purified nanodiscs containing ZipA (Nd-ZipA). A, gel filtration analysis of the sample obtained after removal of detergent from the initial mixture of self-assembled nanodiscs. OD, optical density; mAU, milliabsorbance units. B, SDS-PAGE analysis of fractions obtained from A corresponding to elution volume 9.5–14.5 ml. C, colorless native PAGE analysis of the final preparation containing purified Nd-ZipA (0.9 μg). For comparison, the detergent-soluble ZipA (14 μg), the empty nanodiscs (15 μg), and MSP1D1 (15 μg) were loaded in the same gel. Conditions are described under supplemental Methods.

Analytical ultracentrifugation was used to verify that the final purified Nd-ZipA fraction contained one ZipA molecule per nanodisc (Fig. 2). Sedimentation velocity analysis of empty nanodiscs (Nd) and Nd-ZipA showed that both sedimented as single species (>95% of loading concentration) with standard s-values of 4.3 and 4.9 S, respectively (Fig. 2A). Sedimentation equilibrium was done in parallel to determine the size of the Nd samples and the stoichiometry of the Nd-ZipA complexes (Fig. 2B). Single species with buoyant molecular weight values of 25,200 ± 2000, for Nd, or 34,100 ± 3000, for Nd-ZipA, fit well with the experimental gradients obtained. The difference in buoyant mass between the two species corresponded to the theoretical value expected for the addition of one ZipA monomer (10,200), confirming that the Nd-ZipA complex contains one molecule of ZipA incorporated into each nanodisc. Variation in the protein concentration within the 1–15 μm range had no significant impact on the sedimentation velocity and equilibrium behavior of the Nd and Nd-ZipA samples.

FIGURE 2.

Biophysical characterization of Nd-ZipA. A, sedimentation velocity analysis of empty nanodiscs (empty circles) and Nd-ZipA (solid circles). B, sedimentation equilibrium absorbance profiles of 6.4 μm empty nanodiscs (empty circles) and 8.7 μm Nd-ZipA (solid circles) obtained at 8000 rpm and 20 °C. Lines indicate best fit of a single species model, yielding buoyant molecular weights for empty and ZipA-nanodiscs of 25,200 ± 2000 and 34,100 ± 3000, respectively. Conditions are described under supplemental Methods. OD, optical density. C, normalized fluorescence autocorrelation profiles for Nd* (empty circles) and Nd*-ZipA (solid circles). The concentration of Nd* and Nd*-ZipA was 45 nm. Lines correspond to the best fit of supplemental Equation S2 to the data with the best-fit parameters given in Table 1.

The homogeneity of the nanodiscs (Nd or Nd-ZipA) was also confirmed by using fluorescence correlation spectroscopy to measure the values of diffusion coefficient of two preparations, labeled empty nanodiscs (Nd*) and nanodiscs containing ZipA in which a trace amount of fluorescent lipid was incorporated (Nd*-ZipA) (Fig. 2C and Table 1). The translational diffusion coefficients did not change at nanodisc concentrations in the 45 nm to 1 μm range. Values for the apparent translational diffusion coefficients for unlabeled Nd and Nd-ZipA obtained from dynamic light scattering measurements were also compatible with homogeneous species (Table 1).

TABLE 1.

Molecular properties of ZipA incorporated into nanodiscs

DLS, dynamic light scattering.

	s20,wa	bMwb	Stokes diameter		Diffusion coefficient
			Gel filtrationc	DLS	FCSd,e	DLSe
			nm		cm2/s
Nd-ZipA	4.9 ± 0.1 S	34,100 ± 3000	13.4 ± 0.3	13.0	3.2 ± 0.4 × 10−7	3.5 ± 0.2 × 10−7
Nd	4.3 ± 0.1 S	25,200 ± 2000	10.3 ± 0.4	10.0	4.5 ± 0.2 × 10−7	4.0 ± 0.5 × 10−7

^a Sedimentation coefficient in standard conditions (water, 20 °C and infinite dilution) obtained from the experiment in Fig. 2A.

^b Average buoyant molecular weight obtained from the experiment in Fig. 2B.

^c The data represent mean ± S.D. for at least four experiments.

^d Calculated from FCS autocorrelation curves displayed on Fig. 2C.

^e Values represent mean ± S.D. for at least three independent experiments.

The hydrodynamic diameter of an empty nanodisc calculated from the translational diffusion coefficient value measured, using the Stokes-Einstein equation, was ∼10 nm, which is the same value obtained by gel filtration chromatography (Table 1) and in excellent agreement with the diameter expected for MSP1D1, the specific MSP variant used. A hydrodynamic diameter of ∼13 nm was obtained for Nd-ZipA, also in good agreement with the value obtained from gel filtration. This increment in hydrodynamic apparent size found for Nd-ZipA may be due to the large unstructured domain in ZipA (6), which may protrude far from the disc surface, increasing the hydrodynamic volume of the structure.

Interaction of Nd-ZipA with FtsZ in the Presence of GDP

The ZipA protein attached to nanodiscs was found to bind with moderate affinity to FtsZ-GDP when measured by sedimentation equilibrium. Under the conditions used in our assays and in the presence of GDP, the FtsZ protein monomers have been shown to self-associate in a noncooperative fashion forming short oligomers, with hexamers being the largest class that can be detectable (16). As expected, Nd*-ZipA (15 μm) sedimented at equilibrium with a buoyant molecular weight of 36,000 ± 1000 (Fig. 3A), similar to Nd-ZipA (Fig. 2B). Upon the addition of FtsZ (30 μm) in the presence of GDP, the buoyant molecular weight of Nd*-ZipA was found to increase to 55,000 ± 2000 (Fig. 3A). The increase in mass is due to the mass contributed by FtsZ bound as Nd*-ZipA is the only component that can be measured in the visible region of optical detection system of the ultracentrifuge (FtsZ being invisible). Control experiments confirmed that FtsZ-GDP did not bind to Nd* without ZipA.

FIGURE 3.

Interaction of Nd-ZipA with FtsZ-GDP. A, sedimentation equilibrium gradients of Nd*-ZipA (10 μm) in the absence (empty circles) or presence (solid circles) of FtsZ (50 μm), obtained at 8000 rpm and 20 °C. Solid lines indicate the best-fit gradients of a single species at equilibrium as described under supplemental Methods. OD, optical density. B, number of FtsZ molecules bound per Nd-ZipA (〈n〉) as a function of FtsZ concentration calculated using supplemental Equation S1 from the sedimentation equilibrium data. The line represents the best fit of Equation 5 with the following best-fit parameters: log K_ZZ = 5.2 and log K_AZ = 6.1. Error bars indicate S.D. C, fractional distribution of total Nd-ZipA (10 μm) as various species plotted as a function of total concentration of FtsZ, calculated using best-fit parameter values given in B. Species are labeled with the number of FtsZ molecules per Nd*-ZipA.

Upon the addition of increasing concentrations of FtsZ, the binding of FtsZ-GDP oligomers to Nd*-ZipA resulted in complexes of higher buoyant mass. The number of FtsZ molecules found in each complex can be derived from supplemental Equation S1. In this manner, the dependence of Nd-ZipA·FtsZ complex formation on the concentration of FtsZ (Fig. 3B) was measured and found to follow a similar association model (model 1) as the one describing the binding of FtsZ-GDP to sZipA (16), a soluble ZipA variant lacking the transmembrane region.

In this treatment, it is assumed that a single Nd-ZipA molecule can only bind to one FtsZ oligomer. We have then assumed, as a first approximation, that the Nd-ZipA molecule can bind any oligomer with equal affinity on a per mole of oligomer basis, as expressed in Equations 1 and 2. This does not have to be the case, but it turns out that this simplifying assumption accounts quantitatively for the data.

The best-fit values of logK_ZZ (5.2 ± 0.4) and logK_AZ (6.1 ± 0.5) measuring respectively the affinity between two FtsZ monomers and between Nd-ZipA and FtsZ are similar to the values published for the binding of the FtsZ and sZipA molecules in solution (logK_ZZ = 4.8, logK_AZ = 5.7 (16)). These results suggest that the transmembrane region is not a critical factor involved in the formation of the ZipA·FtsZ complex. Moreover, the values of the equilibrium constants derived from this analysis, in which a mixed association mode is assumed, indicate that in the presence of GDP, the mixtures of Nd-ZipA and FtsZ are distributed in populations in which higher order oligomers of FtsZ (detectable up to hexamer) bind to each Nd-ZipA as the concentration of FtsZ increases (Fig. 3C).

Interaction of ZipA Inserted in Nanodiscs with FtsZ Polymers in the Presence of GTP or GMPCPP

In the presence of GTP or the slowly hydrolyzable analog GMPCPP, which both promote FtsZ polymerization, heterocomplexes between ZipA incorporated in nanodiscs and FtsZ polymers were formed. These findings have been verified by sedimentation velocity and fluorescence correlation spectroscopy assays.

Although no binding of fluorescent nanodiscs without ZipA to FtsZ polymers has been detected (supplemental Fig. S1), the sedimentation velocity profile of a mixture of Nd*-ZipA (5 μm) and FtsZ (50 μm) in the presence of GMPCPP yielded two separate peaks in the sedimentation coefficient distributions (Fig. 4). The slower sedimenting peak, with an s-value of 4.3 S, corresponds to free Nd*-ZipA species, in good agreement with the s-value measured for Nd-ZipA (without fluorescent lipid tracer, see above). The faster peak corresponds to the complex formed by Nd*-ZipA and FtsZ, with an s-value of 22 S, significantly larger than the s-value measured for FtsZ-GMPCPP polymers alone (19 S (20)). The data (Fig. 4) indicate that approximately one-half of the total amount of Nd-ZipA (5 μm) is binding to the oligomeric fibril (50 μm FtsZ ∼0.5 μm oligomer), indicating an average number of approximately five Nd-ZipA bound per oligomer.

FIGURE 4.

Sedimentation velocity analysis of Nd-ZipA binding to FtsZ polymers in the presence of GMPCPP. Sedimentation coefficient c(s) distributions of Nd*-ZipA (5 μm) in the absence (dashed line) and in the presence (solid line) of FtsZ (50 μm) obtained from absorbance data. The dotted line represents the c(s) distribution of FtsZ-GMPCPP alone from interference data in the same experiment. OD, optical density.

These results suggest that FtsZ polymers bind to the ZipA nanodisc with moderate affinity in the micromolar range. The size of the complex, as reflected in the s-value, was independent of the FtsZ concentration (5–50 μm), excluding higher order forms of the Nd*-ZipA·FtsZ complex.

Similarly, the binding of Nd*-ZipA to FtsZ polymers formed in the presence of GTP and a GTP-regenerating system was also measured by sedimentation velocity (supplemental Fig. S2A). As the FtsZ polymers formed in the presence of GTP (12–13 S) are significantly smaller than those formed in GMPCPP (19 S (20)), their association to the ZipA nanodiscs resulted in complexes with s-values (16 S) smaller than those obtained in GMPCPP (22 S). As in the case of GMPCPP, the size of the complexes formed in GTP did not change upon increasing the concentration of FtsZ.

The association of FtsZ polymers to Nd*-ZipA in the presence of GMPCPP was also evidenced by the lower diffusion coefficient of the labeled nanodisc in complex with the polymer relative to the values obtained for free Nd*-ZipA, as measured by fluorescence correlation spectroscopy (Fig. 5A). The diffusion coefficient of empty nanodiscs did not change upon the addition of FtsZ, thus ruling out interaction of FtsZ with the nanodiscs in the absence of ZipA. Fluorescence cross-correlation spectroscopy confirmed heterocomplex formation (supplemental Methods and supplemental Fig. S2B). Similar results were obtained for FtsZ polymers formed in the presence of GTP and a GTP enzymatic regenerating system (supplemental Fig. S2, C and D).

The equilibrium constant for binding of a single Nd-ZipA to an oligomeric fibril was determined as follows. A fixed dilute concentration (45 nm) of fluorescently labeled Nd-ZipA (in the absence or in the presence of unlabeled Nd-ZipA up to 1 μm) was titrated with varying concentrations of FtsZ (GMPCPP or GTP forms). As described in the supplemental material, the resulting autocorrelation functions were analyzed using an empirical model according to which the fluorescent tracer exists only in one of two states, either free or bound to the oligomer. The results were independent of the total Nd-ZipA concentration used. No assumption is made regarding the number of Nd-ZipA bound to a fibril. The fraction of bound Nd-ZipA is plotted as a function of FtsZ-GMPCPP in Fig. 5B, and the comparable results for FtsZ-GTP are plotted in supplemental Fig. S2. It may be seen that at the highest concentration of FtsZ (50 μm) and Nd-ZipA (1 μm) examined, only half of the ZipA (0.5 μm) is bound to the oligomeric fibril (∼0.5 μm in oligomer concentration units), and therefore the average number of Nd-ZipA bound per oligomer is only about 1, validating the assumptions underlying Equations 7–9. The best fit of Equations 8 and 9 to the data rendered a value of the affinity constant for Nd-ZipA binding to FtsZ polymers of log K_AZ = 6.3 ± 0.3.

The fractional binding of Nd-ZipA to oligomers of both FtsZ-GDP and FtsZ-GTP forms as a function of oligomer concentration (Fig. 5C) shows that although the two types of oligomers are of very different average size (∼2 versus ∼100), the binding affinity per mole of oligomer is almost identical. A similar result was obtained for the binding of Nd-ZipA to FtsZ-GMPCPP polymers (Fig. 5C).

Inhibition of the Binding of Nd-ZipA to FtsZ Polymers by C-terminal FtsZ Peptides

A co-crystal containing a 17-amino acid peptide, corresponding to FtsZ C-terminal residues 367–383 (FtsZ(367–383)) and a C-terminal domain of ZipA (ZipA(185–328)), has been used to determine that the interaction between FtsZ and ZipA is established through those regions (13). Fluorescence polarization competition assays using two different peptides (CTZ-WT and CTZ-MUT, as described under supplemental Methods) designed to mimic a segment of the C-terminal region of FtsZ demonstrated that they could bind with high affinity to the FtsZ-binding domain of ZipA (15). The binding affinity of CTZ-MUT to the C-terminal region of ZipA is an order of magnitude higher than the binding of CTZ-WT (15). We have used fluorescence correlation spectroscopy competition assays to measure the ability of these peptides to inhibit the binding of Nd-ZipA to FtsZ polymers. A 10-fold molar excess of either of the two peptides, relative to FtsZ, was added to mixtures of Nd*-ZipA bound to FtsZ polymers. The autocorrelation curves obtained yield a faster diffusion rate compatible with a complete dissociation of Nd*-ZipA from FtsZ-GMPCPP polymers (Fig. 6A). Sedimentation velocity analysis (Fig. 6B) shows that, upon the addition of the CTZ-MUT peptide, the Nd*-ZipA present in the mixture with FtsZ-GMPCPP sediments with a similar s-value as the free Nd*-ZipA and, as expected, no traces of the complex with FtsZ polymers were observed. At lower peptide concentration (equivalent to a 1:1 peptide:FtsZ molar ratio), the CTZ-MUT peptide still caused a full dissociation. However, only a 60% decrease in the fraction of bound Nd-ZipA was found upon the addition of CTZ-WT. These observations are in accordance with the different affinities of both peptides (15). As expected, the peptide-mediated dissociation of the Nd*-ZipA·FtsZ polymer complex was similarly observed when the FtsZ polymer was formed in the presence of GTP and a GTP-regenerating system.

FIGURE 6.

Inhibition of Nd-ZipA binding to FtsZ polymers by C-terminal FtsZ peptides. A, normalized autocorrelation functions of Nd*-ZipA in the absence (solid circles) and in the presence (solid triangles) of 25 μm FtsZ. The FCS profiles of the Nd*-ZipA:FtsZ mixtures upon the addition of 25 μm CTZ-MUT (empty triangles) and CTZ-WT (empty circles) are shown. Lines correspond to the fit of Equations S2 and S3 to the data as described under supplemental Methods. FtsZ polymerization was triggered upon the addition of 1 mm GMPCPP. B, sedimentation coefficient c(s) distributions of mixtures of Nd*-ZipA (3 μm) and FtsZ-GMPCPP (12.5 μm) in the presence (solid line) of CTZ-MUT peptide (12.5 μm). The theoretical profile corresponding to Nd*-ZipA fully bound to FtsZ-GMPCPP is also shown (dashed line).

DISCUSSION

Using the nanodisc technology (22), we have investigated the interaction between the membrane-bound ZipA and the soluble FtsZ, two proteins that assemble in the E. coli proto-ring. In addition to their capacity to incorporate membrane proteins into a lipid environment while maintaining them in a soluble form, nanodiscs can be tagged with fluorescently labeled lipid molecules, which allowed us to use the full-length ZipA protein with no modification. Nanodiscs also allow formulating assays to measure the inhibition of protein-protein interactions in solution, which is useful to screen for inhibitors of divisome assembly and develop new antimicrobials.

Single copies of the full-length ZipA were easily incorporated into nanodiscs produced by mixtures of an E. coli polar lipid extract and the membrane scaffold protein MSPD1. Using these nanodisc preparations (Nd-ZipA), we have first measured their interaction with different forms of the FtsZ protein, either FtsZ oligomers formed in the presence of GDP or FtsZ polymers formed in the presence of GTP or GMPCPP. The affinity of Nd-ZipA for FtsZ-GDP oligomers is of the same magnitude as the affinity to FtsZ polymers (logK_AZ of the order of 6). These values are of a similar magnitude as those found for the interaction of the soluble form of ZipA (lacking the 1–25 transmembrane region) with each of the two FtsZ forms.

From these results, we can conclude that the transmembrane region of the ZipA protein plays little, if any, role in the interaction with FtsZ and lends support to the idea that it simply provides an anchoring point to the membrane where the E. coli proto-ring can assemble. Although this conclusion could be expected, our results now clearly exclude any undesired effects on the folding and function of the remainder fragment resulting from the truncation of the anchoring segment of the membrane-bound ZipA. Moreover, it is hard to know whether detaching the protein from its native lipid-bound environment will result in a protein having the same properties, in particular those pertaining to binding. Our results, by using a full form of the protein, therefore provide further support to the conclusions previously advanced when a truncated form of ZipA lacking the transmembrane anchoring domain was used to measure its binding to FtsZ (16).

We have confirmed a model in which the affinity of binding of Nd-ZipA to FtsZ is insensitive to the state of nucleotide binding or oligomerization of the FtsZ. This result is certainly remarkable, taking into account the possible presence of allosteric effects, which surely exist because they lead to very large differences between the tendency of FtsZ-GDP and FtsZ-GTP to self-assemble (4, 5).

Our results provide some clues on mechanisms to remodel the architecture of the division ring during constriction. During the E. coli division cycle, ZipA is distributed along the full inner surface of the cytoplasmic membrane and localizes in part at the septal position at the time of proto-ring assembly. In contrast to FtsA, the large unstructured cytoplasmic domain of ZipA provides a flexible anchoring structure. As only a 30% of the FtsZ protein seems to localize at the division ring, it has been assumed that a large amount of FtsZ should be dispersed in the cytoplasm. As measured in nanodiscs containing ZipA, the affinity of FtsZ-GDP oligomers for ZipA is of a similar magnitude as the affinity of larger FtsZ-GTP polymers. This indicates that the attachment of FtsZ to ZipA is likely to play a minor or even negligible role in FtsZ polymerization and its assembly into the proto-ring. However, as the septum selection proteins MinC and SlmA act by preventing polymerization outside midcell, FtsZ, given its relatively weak association to ZipA, can easily migrate from other regions of the cytoplasmic membrane to occupy a central position when required at division. The risk of abortive Z-ring collapse could nevertheless be maintained at lower levels if FtsZ polymers could establish interactions with more than one ZipA molecule or to another tethering element, such as FtsA. Recent evidence suggests that part of the ZipA protein may be inserted into the membrane as dimers (23), which could provide some stability to the flexible attachment of FtsZ if temporal interactions were alternatively established. The final result would be a structurally stable proto-ring that would simultaneously be sufficiently dynamic to rearrange adopting a variable diameter. This would allow constriction once the elements of the divisome are assembled. This plasticity of the proto-ring connection to the cytoplasmic membrane is reflected, for example, in the inability of already formed proto-rings to remain assembled when other downstream elements in the assembly pathway, such as FtsN, are removed (24).

In summary, we have exploited new reconstitution technologies to test the properties of the elements of the bacterial division machinery and their interactions under well defined lipid environments to gain insights into their precise functions. In addition, these nanodisc systems have been used to generate a fluorescence-based assay to screen for inhibitors of ZipA·FtsZ complex formation, potentially applicable to interactions between other divisome components.