A key event in bacterial cytokinesis is the formation of the Z ring, which serves as a mechanical scaffold that recruits other cytokinetic proteins to establish functional divisomes. This scaffolding function of Z rings is essential throughout cytokinesis, but the underlying molecular interactions are poorly understood. Here we report that a widely conserved FtsZ binding protein, ZapA, has cytological, biochemical and biophysical properties that argue for the importance of cross-linking interactions between FtsZ polymers in the coherence of Z rings. Escherichia coli zapA null mutant cells have Z rings that are structurally looser and many helical precursors of Z rings fail to coalesce into coherent rings. Biophysical behaviour of FtsZ in the presence of ZapA reveals that ZapA not only bundles, but also cross-links FtsZ polymers, which makes it the first cross-linking protein of the bacterial cytoskeleton. Cross-linking in vitro occurs at the stoichiometry of FtsZ–ZapA interaction at the Z rings in vivo, where nearly all intracellular ZapA is dynamically associated. ZapA also stabilizes longitudinal bonds between FtsZ monomers since it promotes the polymerization of FtsZ mutants with lesions at the polymerization interface and since it reverses the inhibitory effects of SulA, a known antagonist of FtsZ longitudinal interactions.
Bacterial cell division occurs through the coordinated assembly of the cytokinetic apparatus, which ultimately constricts to produce two daughter cells (Dajkovic and Lutkenhaus, 2006). The first and key event in the formation of this division apparatus, or divisome, is the attachment of FtsZ polymers to the cell membrane and their concentration in the middle of the cell to form a dynamic structure termed the Z ring. Because FtsZ does not directly bind phospholipids, FtsZ polymers interact with other proteins to efficiently form coherent Z rings. In Escherichia coli, membrane associated proteins FtsA and ZipA are required for this process (Pichoff and Lutkenhaus, 2002).
Once formed as a mechanically coherent structure, the Z ring serves as a scaffold where the remaining cell division proteins are recruited to establish a functional divisome and to effect cytokinesis (Goehring and Beckwith, 2005; Dajkovic and Lutkenhaus, 2006). The scaffolding function of FtsZ polymers is essential for the persistence of the divisome as a coherent structure throughout the cell division process (Addinall et al., 1997). The Z ring is highly dynamic and the subunits of FtsZ cycle rapidly between the Z ring and the surrounding cytoplasmic environment (Stricker et al., 2002; Anderson et al., 2004). These dynamics do not seem to be essential for cell division (Mukherjee et al., 2001; Dajkovic and Lutkenhaus, 2006), but they do raise the question of how the Z ring can maintain mechanical coherence over the entire cell division cycle while at the same time being unstable.
One proposal concerning the nature of the mechanical coherence of Z rings holds that FtsZ polymers in the Z ring must maintain a critical number of connections between each other to effect the scaffolding function (Dajkovic and Lutkenhaus, 2006; Dajkovic et al., 2008a,b). At least some of these connections are mediated by direct interactions between FtsZ molecules themselves (Osawa et al., 2008; Dajkovic et al., 2008a). FtsZ molecules can interact with each other through two sets of bonds with different energies. GTP-dependent longitudinal bonds lead to the formation of linear polymers of FtsZ and have high energies on a per subunit basis (Lan et al., 2008; Dajkovic et al., 2008a). Lateral bonds are weaker on a per subunit basis but are important because they allow the formation of bundles of polymers and polymer networks that lead to the gelation of FtsZ into mechanically stiff structures (Dajkovic et al., 2008a; Lan et al., 2008).
Lateral interactions between FtsZ polymers are important for the scaffolding function of FtsZ (Dajkovic et al., 2008a; Lan et al., 2008; Monahan et al., 2009), but the extent of their physiological significance is controversial (Lutkenhaus, 2008; Monahan et al., 2009). Electron cryotomographic reconstructions of cytokinetic rings in bacteria (Li et al., 2007) show a collection of disordered polymers at the division site without a substantial presence of bundling (shown diagrammatically in Fig. 8C). It is possible that lateral interactions are easily disrupted by sample preparation for cryoelectron microscopy. Another possibility is that lateral interactions between FtsZ polymers play a significant but ancillary role in the coherence of Z rings and that molecular interactions of a different nature hold Z rings together.
What molecular interactions between FtsZ polymers are responsible for the coherence of Z rings is a major question in bacterial cell biology. In addition to direct interactions between FtsZ molecules, the mechanical coherence of Z rings in vivo appears to be aided by interactions of FtsZ polymers with other molecular species such as other proteins and lipids (Geissler and Margolin, 2005; Goehring et al., 2007b; Dajkovic et al., 2008a; Osawa et al., 2008).
One widely conserved FtsZ binding protein is ZapA, first identified in Bacillus subtilis as a positive regulator of FtsZ assembly. There is no discernible phenotype of inactivation of zapA in B. subtilis, except when cell division is compromised with additional mutations or a decrease in the level of FtsZ (Gueiros-Filho and Losick, 2002). In E. coli absence of ZapA is associated with defects in cell division (Mohammadi et al., 2009; Galli and Gerdes, 2010). ZapA proteins from E. coli and Pseudomonas aeuroginosa exist in a dimer–tetramer equilibrium in vitro, but it is unclear which form is physiologically relevant (Low et al., 2004; Small et al., 2007). ZapA proteins from all organisms studied thus far have been shown to promote the bundling of FtsZ polymers (Gueiros-Filho and Losick, 2002; Low et al., 2004; Small et al., 2007). While other activities have not yet been established for ZapA, Mohammadi et al. (2009) suggested that ZapA might have cross-linking activity. Here we use a combination of cytological, biophysical and biochemical methods to study the function of E. coli ZapA and the organization of the Z ring.
ZapA is required for viability of a subset of exponentially growing cells
To study the function of ZapA in E. coli, we investigated the phenotype of the zapA deletion strain PS620 (W3110 zapA::CAT). We find that the zapA deletion strain has statistically significant cytokinesis defects that can be complemented with H6–ZapA provided from a plasmid (see Fig. S1 in the Supporting information).
In the process of investigating the zapA::CAT phenotype we noticed that the phenotype of PS620 cells was much less pronounced in stationary phase cultures (Fig. 1A and B). The mean cell length of stationary phase PS620 was still somewhat greater than in the parent W3110 cells, as was the width of the cell length distribution (Fig. 1C), but the effects were much smaller than in the exponential phase. Furthermore, filaments were absent from stationary phase cultures. This might indicate that the filaments observed in the exponential phase eventually divided to produce cells of normal size. This would suggest that ZapA may cause a delay in cell division. Alternatively, filaments could disappear from the exponentially growing population by dying.
To distinguish between these two possibilities we microscopically observed the division of lineages of single PS620 cells using a microcolony assay. Lineages of individual cells can be followed up to eight generations using this assay, as described previously (Stewart et al., 2005). We followed the division of 30 individual cells in this way.
Cells that were of normal size at the beginning of the experiment gave rise to cells of normal size as well as to filaments (Movie S4). Some of these filaments divided normal sized cells off the ends for several generations (see also Movie S1). These normal sized cells in turn divided again to give cells of normal size as well as filaments. Most filaments stopped dividing and increasing in size and many eventually lost phase contrast, indicating cell death in the course of the experiment (see also Movie S2). Other cells that started as filaments simply did not grow at all and eventually lost phase contrast and died (Movie S3). When the moment of the loss of contrast was caught on film it appeared that the cells lysed at the pole and the cytoplasmic contents leaked out in a wave (see the right-hand pole of the cell on the left in Movie S2).
To confirm that the cells of the PS620 population were dying at a greater rate than the cells of the parent strain W3110, we stained exponentially growing cells of both strains with propidium iodide and analysed them using flow cytometry. There was an approximate 10-fold increase in the percentage of dead cells in populations of exponentially growing PS620 cells compared with the wild-type parent strain W3110 (Fig. 1D). Microscopic analysis of the PS620 cells stained with propidium iodide showed that 90% of the stained cells were filaments (examples in Fig. 1E and F), consistent with the increased mortality rates observed by flow cytometry.
These results indicate that ZapA is required for the viability of a subset of exponentially growing cells in which the failure of cytokinesis is fatal.
Synergy between ZapA and early cell division proteins FtsZ, FtsA and ZipA
The assembly of the bacterial cytokinetic apparatus is known to occur in two temporally distinct stages (Aarsman et al., 2005). In the first stage, FtsZ forms the Z ring with the aid of FtsA and ZipA. In the second stage, the remaining cell division proteins are recruited to the Z ring to form a functional cytokinetic apparatus capable of catalysing septation. ZapA is recruited to Z rings in the first stage (Goehring et al., 2005; Galli and Gerdes, 2010). The deletion of zapA in E. coli causes a cytokinesis defect (Mohammadi et al., 2009; Galli and Gerdes, 2010), but it is not clear at which stage ZapA functions.
To address this question, we tested whether ZapA acts synergistically with early- or late-stage cell division proteins. We introduced the zapA::CAT allele from PS620 into strains carrying temperature sensitive alleles of the early cell division genes ftsZ, ftsA, zipA and one late cell division gene, ftsI. In multiple experiments the zapA deletion enhanced the temperature sensitivity of strains carrying ts mutations in all early cell division proteins FtsZ, FtsA and ZipA (representative experiment is shown in Fig. 2). In contrast, it had no discernible effect on the temperature sensitivity of FtsI, a late-stage protein (Fig. 2). This suggests that ZapA might function in the first stage of the assembly of the cytokinetic apparatus.
ZapA promotes the coherence of Z rings
To investigate the effects of ZapA on the organization of FtsZ in the cell, we used a previously described construct (Dajkovic et al., 2008b) to expresses FtsZ–GFP from a chromosomal locus under the control of a modified lac promoter in the wild-type and zapA::CAT strains. Expressing FtsZ–GFP under these conditions has no discernible effect on cell division (Dajkovic et al., 2008b).
Cells of the zapA deletion strain AND831 (zapA::CAT att::plac–ftsZ–gfp) contained significantly fewer coherent Z rings compared with the wild-type strain AND101 (W3110 att::plac–ftsZ–gfp) (Fig. 3A, B and C). Twenty-eight per cent of discrete localizations of FtsZ–GFP in AND831 were in helical formations as opposed to coherent rings, compared with less than 3% for the AND101 strain (Fig. 3D). These quantifications are congruent with a recent report on the localization of FtsZ–GFP in the zapA deletion strain (Galli and Gerdes, 2010). These helices were dynamic and moved along the long axis of the cell sometimes becoming transiently concentrated in ring-like structures before reverting again to helices (see Movie S5). Interestingly, the longer cells in the AND831 (zapA::CAT att::plac–ftsZ–gfp) population often contained coherent rings at one potential division site (PDS) and helices of FtsZ around another PDS (Fig. 3A and B). This suggests that the failure of ring formation is not a global physiological defect, but that it is localized to individual PDSs. This is further reinforced with results from timelapse imaging of PS620 cells that show that many filaments are able to divide (see Supporting information and Movies S1–S4).
We noticed that many apparently coherent Z rings in AND831 were appreciably wider compared with the rings of the wild-type strain AND101 (Fig. 3A, B and C). We hypothesized that even the Z rings that appear coherent by conventional fluorescence microscopy have a looser structure in the absence of ZapA. Even though it is not possible to see the fine structure of the Z ring using conventional fluorescence microscopy, the relative widths of Z rings in different strain backgrounds can be directly compared by measuring and comparing their apparent widths. The apparent widths of the seemingly coherent Z rings in AND831 strain were significantly (t-test, P < 0.01) larger than in the wild-type AND101 strain (Fig. 3E). This indicates that even the apparently coherent Z rings are structured differently in the absence of ZapA since they occupy more space than the rings of the wild-type strain.
We wondered whether the FtsZ–GFP localized in helices instead of rings in the approximately 3% of wild-type cells represented bona fide precursors of coherent Z rings. To investigate this, we followed FtsZ–GFP in newborn wild-type cells. As previously described (Sun and Margolin, 1998; Ben-Yehuda and Losick, 2002; Peters et al., 2007), incipient Z rings in newborn cells form after the constricting ring from the mother cell sends out helical formations that coalesce into a Z ring structure. We found that some incipient Z rings did not persist as coherent structures. In a subset of the population of newly born cells, the incipient Z rings of one of the daughter cells reverted to the helical precursors (Fig. 3F) for a period of time before coalescing back into coherent Z rings. In some cells this fluctuation was repeated one or two times. Even in these cells, the cytokinesis process was invariably completed successfully as evidenced by the eventual formation of the septum and cell separation.
These results suggest that the incipient Z rings are meta-stable structures that can fluctuate between coherent rings and helices. Coherent rings and helices therefore represent two minima of the energy landscape of FtsZ in the cell that are separated by a small energy barrier (Fig. 8B). Since the absence of ZapA makes the helical formations more prevalent, we conclude that ZapA holds FtsZ polymers together and thereby makes the organization of FtsZ into coherent Z rings more favourable.
ZapA cross-links FtsZ polymers
To gain insight into the nature of interactions by which ZapA holds FtsZ structures together, we studied the effects of ZapA on the mechanical properties of FtsZ gels. We used quantitative rheometry to measure the elastic modulus of FtsZ structures in the presence of H6–ZapA in vitro. The magnitude of the elastic modulus measures the ability of structures to rebound after deformations and it characterizes their mechanical stiffness. Similar to our previous study (Dajkovic et al., 2008a), we chose to examine the mechanical properties of FtsZ at concentrations where FtsZ forms a gel (Esue et al., 2005; Dajkovic et al., 2008a). The sol state is physiologically less relevant for the study of the scaffolding function of FtsZ because FtsZ is incapable of functioning as a mechanical scaffold in this state and is consequently unable to support cytokinesis (Dajkovic et al., 2008a).
FtsZ was polymerized with GTP in the presence of varying concentrations of H6–ZapA. ZapA had no effect on the elastic modulus of FtsZ gels at ZapA/FtsZ molar ratios lower than 0.5. In the presence of increasing concentrations of H6–ZapA, for molar ratios of ZapA/FtsZ of up to 4:1, the steady state elastic modulus of FtsZ–ZapA structures increased significantly (Fig. 4A and B). At the highest concentration of H6–ZapA, the stiffness of FtsZ–ZapA structures was approximately 30 greater than in the absence of ZapA at the same concentration of FtsZ. This effect is specific to ZapA since random proteins like MalE or BSA have no effect on the elasticity of FtsZ networks even at elevated concentrations (Dajkovic et al., 2008a). In the presence of GDP, there was no appreciable elasticity at any concentration of ZapA.
Theoretical considerations from polymer physics suggest that the stimulation of bundling alone should have only slight effects on the elasticity of polymer networks (Ferry, 1980; Tseng et al., 2001). Indeed, the bundling of FtsZ observed in the presence of Ca2+ leads only to a twofold increase in the stiffness of FtsZ structures at the same concentration of FtsZ (Fig. 4F). To further investigate whether bundling of FtsZ polymers could alone explain the large increase in the mechanical stiffness of FtsZ gels, we took advantage of an FtsZ mutant, FtsZ320, which lacks the disordered C-terminal peptide dispensable for polymerization. This truncation mutant hydrolyses GTP like the wild-type protein (Wang et al., 1997), but it preferentially polymerizes into bundles (Fig. 4C, D and E) as evidenced by electron microscopy (EM). To complement the EM studies we confirmed that FtsZ320 also forms bundles in solution (Supporting information and Fig. S2). The elastic modulus of FtsZ320 was not significantly different from the elastic modulus of the wild-type FtsZ protein at the same concentration (Fig. 4F). Therefore, neither an exogenous factor like Ca2+, which is known to promote bundling, nor an endogenous change in the FtsZ protein that likewise greatly increases bundling, can significantly enhance the stiffness of FtsZ structures (Fig. 4F). Taken together, these results suggest that bundling alone is not sufficient to explain the increase in the stiffness of FtsZ structures in the presence of ZapA.
Mohammadi et al. (2009) reported that the effect of ZapA on FtsZ in sedimentation assays is affected by pH and Mg2+ concentration in the buffer. In contrast, the stiffening observed in the rheology experiments in the present study also occurred when the experiments were done in a buffer at pH 7.5 and 5 mM Mg (Fig. S6), suggesting that the observed effect is robust. Nevertheless, the stiffening was slightly smaller at pH 7.5 and 5 mM Mg. The differences observed by Mohammadi et al. (2009) are perhaps due to the limitations of the sedimentation method used in their study, as discussed for the case of the study of Low et al. (2004) in the Supporting information (see Supplemental Discussion and also Fig. S3).
Cross-linking interactions between polymer species, in contrast to bundling, are known to significantly increase the stiffness of gels. Polymer species in a network with cross-linking interactions are not simply entangled, but also stick to each other. Therefore, cross-linking interactions are different from entanglement in the sense that they are not just steric but contain an enthalpic component. A functional distinction between cross-linking and bundling interactions of polymers is clearly made in the literature on the eukaryotic cytoskeleton (Matsudaira, 1991; 1994; Wachsstock et al., 1993; Xu et al., 2000; Tseng et al., 2004). However, there seems to be some confusion in the literature on the bacterial cytoskeleton and the terms cross-linking and bundling are frequently used interchangeably. For a diagrammatic representation of the difference between cross-linking and bundling interactions between polymer species see Fig. 8A. No FtsZ cross-linking proteins have been described thus far, but in the case of actin the presence of cross-linking factors such as α-actinin and filamin greatly increases the stiffness of actin gels and renders them more solid-like (Wachsstock et al., 1993; Tseng et al., 2004).
Cross-linking interactions cannot be detected by EM, but their effects can be studied rheologically (Wachsstock et al., 1993; Tseng et al., 2004). Because of the increased binding between polymer species in a cross-linked gel, the mobility of polymer species is decreased. Indeed, for a given concentration of polymer species of a certain length their mobility in a gel depends largely on their propensity to interact through cross-linking interactions (Ferry, 1980).
To study the mobility of FtsZ polymer species we subjected FtsZ gels formed in the presence of ZapA to oscillatory shear deformations of increasing frequencies. Decreased mobility of polymer species correlates with a decrease in the dependence of the elastic modulus on the frequency of deformation. Indeed, with increasing concentrations of ZapA, the elastic modulus of FtsZ gels depended less on the frequency of deformation (Fig. 4G). To further quantify the inhibition of movement of FtsZ polymers in the structures formed in the presence of ZapA, we fit the frequency-dependent profiles of the elastic modulus (Fig. 4G) to power laws of the type G′(ω) ∼ ωa, where the exponent a describes the degree of frequency dependence of the elastic modulus (Tseng and Wirtz, 2001). The exponent a decreased for increasing H6–ZapA concentrations (Fig. 4G, inset), which indicates that the movement of polymer species in structures formed with increasing concentrations of ZapA was correspondingly inhibited. This trend in the frequency dependence of the elastic modulus was also evident at pH 7.5 and 5 mM Mg2+. This suggests that FtsZ polymer species in the presence of ZapA are indeed less mobile.
Since cross-linking interactions also increase the solid-like properties of gels (Wachsstock et al., 1993; Tseng et al., 2004), we measured the phase angle of FtsZ structures in the presence of ZapA as a complementary approach. The phase angle quantifies the solid-like versus liquid-like nature of viscoelastic materials, i.e. it is an indication of how much elastic behaviour is present. On the other hand, it does not quantify the stiffness of a material.
To illustrate, a completely solid material like steel has a phase angle of 0° whereas a complete liquid like water has a phase angle of 90°. Most viscoelastic materials fall somewhere between these two values, such as actin gels cross-linked by α-actinin that have phase angles of approximately 13° (Wachsstock et al., 1993). ZapA lead to a small but significant (t-test, P < 0.00001) decrease in the phase angle of FtsZ structures compared with the same structures without ZapA (Fig. 4H). This is consistent with ZapA conferring a more solid-like character to FtsZ gels due to cross-linking interactions.
Taken together, these results suggest that ZapA mediates cross-linking interactions between FtsZ polymer species.
ZapA promotes the polymerization of FtsZ by stabilizing longitudinal bonds between FtsZ molecules
The in vitro data presented thus far suggest that ZapA has cross-linking activity in addition to its known ability to bundle FtsZ polymer species. It has been hypothesized that ZapA may have an additional activity, that it may stabilize the longitudinal contacts between subunits in a polymer by acting as a molecular splint (Gueiros-Filho and Losick, 2002; Low et al., 2004).
To test this hypothesis, we investigated whether ZapA could promote the polymerization of otherwise polymerization deficient mutants of FtsZ. We examined FtsZ1(A70T) and FtsZ2(D212G) mutants (Mukherjee et al., 2001). Both mutants are compromised for assembly due to changes in the amino acids at the polymerization interface (Fig. 5A). They are unable to assemble under standard polymerization conditions, but do assemble in the presence of DEAE dextran, a molecule that promotes the polymerization of FtsZ. This suggests that they retain the ability to polymerize but that the molecular lesions in these mutant proteins raise the energy barrier for assembly, most likely at the level of nucleation. The argument for this notion is that FtsZ2(D212G) co-assembles with the wild-type protein if the latter is present above the critical concentration for polymerization (Mukherjee et al., 2001).
To test whether ZapA could lower this energy barrier and promote the nucleation of FtsZ1(A70T) and FtsZ2(D212G), we first examined the polymerization of these mutants using the standard sedimentation assay (Mukherjee and Lutkenhaus, 1998a). FtsZ1(A70T) and FtsZ2(D212G) did not pellet on their own in the presence of GTP, but ZapA promoted the pelleting of both mutant proteins (data not shown). However, in these assays, ZapA promoted the pelleting of FtsZ1(A70T) and FtsZ2(D212G) both in the presence of GDP and in the presence of GTP. This confirms that ZapA has the capacity to interact with both GDP-bound and GTP-bound forms of FtsZ1(A70T) and FtsZ2(D212G) to form large sedimentable complexes as is the case with the wild-type FtsZ (see Supporting information and Fig. S4).
No polymers were found among the products of the GDP reactions examined by EM. Only large aggregates were present for both FtsZ1(A70T) and FtsZ2(D212G) (Fig. 5B and E). Interestingly, in the presence of GTP, linear polymers were abundant on the EM grid (Fig. 5C and F). The polymers appeared to be mostly single filaments of FtsZ since the distribution of the widths of polymer species was bimodal and narrowly clustered around 4 and 8 nm, with 70% of polymers in both cases falling in the 4 nm bin (Fig. 5D).
To test whether the observed stabilization of longitudinal bonds between monomers of these mutant proteins affected the mechanical properties of FtsZ structures, we measured the elasticity of FtsZ1(A70T) and FtsZ2(D212G) polymer networks with and without ZapA. In the presence of GTP, but with no ZapA present, neither protein showed appreciable elasticity (Fig. 5G). However, in the presence of ZapA, the elasticity of networks composed of FtsZ1(A70T) and FtsZ2(D212G) reached values measured for wild-type FtsZ without any accessory proteins (Fig. 5G). These results suggest that ZapA stabilizes the longitudinal bonds between FtsZ subunits and that this activity results in the formation of elastic structures for proteins otherwise unable to assemble into filaments.
As a complementary approach, we showed that ZapA reverses the inhibitory effects of SulA, a factor known to destabilize the longitudinal bonds between FtsZ molecules (see Supporting information and Fig. S5).
Nearly all intracellular ZapA is dynamically associated with the cytokinetic ring
We next investigated whether ZapA is localized at the division site throughout the cell cycle. To that end, we examined exponentially growing MG1655 cells expressing ZapA–GFP. The localization of ZapA–GFP was very similar to the localization of FtsZ–GFP. Cells at different stages of the cell cycle, indicated by their differing lengths, had medial localizations of ZapA–GFP (Fig. 6A). Even in cells with very deep constrictions ZapA–GFP was localized to the cytokinetic ring (Fig. 6A). We conclude that ZapA is associated with the cytokinetic ring throughout the cell cycle.
To determine the proportion of intracellular ZapA that is localized at the Z ring, we analysed the fluorescence intensity scans along the long axes of MG1655 cells expressing ZapA–GFP. In > 85% of the cases, the fluorescence outside the Z ring was not above the fluorescence background of cells that did not express ZapA–GFP. A typical scan is shown in Fig. 6B. In the remaining cases, less than 10% of fluorescence was found outside the ring. A recent publication reported that the intracellular concentration of ZapA in E. coli is equal to the intracellular concentration of FtsZ (Mohammadi et al., 2009). Given that 25–30% of intracellular FtsZ is located in the Z ring (Anderson et al., 2004), our results indicate that there are on average 3–4 molecules of ZapA per molecule of FtsZ in the Z ring. This is consistent with the highest stoichiometries used in the rheology experiments where the evidence for ZapA cross-linking of FtsZ polymers is the strongest.
To determine whether the localization of ZapA to the cytokinetic ring was dynamic, we selectively bleached parts of rings in MG1655 cells expressing ZapA–GFP and measured the kinetics of the recovery of fluorescence. Figure 6C shows a typical recovery curve. Fluorescence recovered rapidly with the t1/2 of 11 ± 4 s (n = 23), indicating that the association of ZapA with the cytokinetic ring is dynamic.
Suppression of the cytokinesis defect of zapA deletion
To find potential overlaps between the function of ZapA and the functions of other cell division proteins, we explored whether the cell division defect caused by the absence of ZapA could be suppressed by other cell division proteins.
First, we tested whether the increase in the intracellular concentrations of FtsZ, FtsA and FtsQ could suppress the zapA deletion. Overexpression of this combination of proteins has been previously shown to suppress a number of cell division defects (Geissler et al., 2003; Geissler and Margolin, 2005; Reddy, 2007). We introduced the plasmid pBS58 that carries the ftsQ, ftsA and ftsZ genes on the pGB2 backbone. It raises the level of FtsZ in the cell about fivefold that is sufficient to suppress inhibition of cell division by SulA or MinCD (Bi and Lutkenhaus, 1990). This plasmid was not able to suppress the cytokinesis defect of PS620 (Fig. 7A and B). Furthermore, this plasmid did not bring the mean cell length nor the cell length distribution of PS620 to the level of the wild-type strain (Fig. 7C).
Next, we tested ftsA*, a hypermorphic allele of ftsA. This allele bypasses the requirement for ZipA in cell division (Geissler et al., 2003) and has also been shown to suppress a number of cell division defects (Geissler and Margolin, 2005; Goehring et al., 2007a). We transduced the zapA::CAT allele into the strain W3110*, which carries ftsA* in place of the wild-type ftsA. The resulting strain AND1124 (ftsA*leu::Tn10 zapA::CAT) had a significantly (t-test, P < 0.0001) lower mean cell length than the parent PS620 strain, though there were still some filaments present in exponentially growing populations (Fig. 7D, E and F). Likewise, the range of cell widths was significantly smaller in AND1124 than in the PS620 strain (Fig. 7F). The percentage of dead cells in AND1124 was also significantly reduced with respect to PS620, and was near the wild-type level (Fig. 7G). One possible explanation is that the ftsA* allele suppresses both the cell division defect and the mortality caused by the absence of ZapA. This suggests that FtsA* protein may share a molecular activity with ZapA.
In this study, we examined the function of ZapA by studying the consequences of the deletion of the zapA gene from E. coli and by examining the effects of the ZapA protein on the mechanical and biochemical properties of FtsZ structures in vitro. We found that the deletion of zapA results in a significant cytokinesis defect and a 10-fold increase in cell mortality. We observed that ZapA has two additional activities, besides the known bundling activity. It cross-links FtsZ polymers and stabilizes longitudinal bonds between FtsZ monomers. The loss of these activities in the zapA mutant reduces the efficiency of the formation of coherent Z rings at the site of septation and their maintenance as mechanically coherent structures.
Building coherent Z rings with ZapA
A previous study reported that the inactivation of zapA in B. subtilis causes no cytokinesis defects unless other cell division genes are inactivated or the level of FtsZ in the cell is decreased. The level of ZapA in B. subtilis is 5% of the level of FtsZ (Gueiros-Filho and Losick, 2002). In contrast, the intracellular level of ZapA in E. coli is 20-fold greater and is equal to the level of FtsZ (Mohammadi et al., 2009). This suggests that ZapA may be more important for cytokinesis in this organism.
In accordance with this, the inactivation of zapA in E. coli causes cytokinesis defects that result in the filamentation of many E. coli cells (Mohammadi et al., 2009; Galli and Gerdes, 2010). These cytokinesis defects are due to the inability of FtsZ to coalesce and concentrate into functional Z rings, being instead mislocalized in helices. Fluorescence microscopy revealed that about one third of all discrete localizations of FtsZ in the zapA cells is found in helices instead of coherent Z rings. Even the Z rings that appeared coherent have a seemingly looser structure since they are wider than the rings of the wild-type strain. Some of these rings may in fact be helices of FtsZ that cannot be resolved by fluorescence microscopy.
Our observations of FtsZ–GFP in the newborn wild-type cells revealed that some incipient Z rings fluctuate between an organized ring and a helix before finally stabilizing into rings. Other laboratories have observed the same phenomenon (H. Erickson, pers. comm.). This natural fluctuation of FtsZ even in the presence of ZapA suggests that the energy minima represented by helices and coherent rings are close and may not be separated by a large barrier (Fig. 8B). So even though FtsZ helices have the ability to spontaneously assemble into rings in vitro (Osawa et al., 2008), their efficient concentration into coherent Z rings in vivo is aided by other cell division proteins, such as ZapA. ZapA appears to hold FtsZ polymers together in a mechanically coherent structure thereby favouring the natural tendency of some FtsZ helices to organize into rings (Fig. 8B). How might ZapA do that?
Cross-linking interactions and the mechanical coherence of the Z ring
The nature of the molecular interactions between FtsZ polymers that maintain the mechanical coherence of Z rings is somewhat enigmatic. Several studies suggested that bundling of FtsZ polymers is required for the stability of Z rings (Lu et al., 2001; Dajkovic et al., 2008a; Monahan et al., 2009).
There are at least two lines of evidence that point to the fact that bundling of FtsZ polymers may play a significant but ancillary role in maintaining the organization of the FtsZ scaffold. The first is empirical. Cryoelectron tomography reconstructions of cytokinetic rings have failed to reveal extensive presence of bundles of FtsZ polymers. What they show, instead, is a compilation of somewhat disordered polymers at the division site whose long axes are on average oriented perpendicular to the long axis of the cell (Li et al., 2007) (shown diagrammatically in Fig. 8C).
The second line of evidence consists of geometric considerations from the known interactions of FtsZ at the division site. FtsZ is known to interact with a number of proteins to form functional divisomes (Goehring and Beckwith, 2005; Dajkovic and Lutkenhaus, 2006), and these interactions may not be possible or may have insufficient avidity if all FtsZ polymers were arranged in tightly ordered bundles. This notion anticipates that some essential cytokinetic proteins may be involved in antagonizing the inherent bundling activity of FtsZ.
Indeed, the inherent tendency of FtsZ to bundle seems greater than previously appreciated. This inherent bundling of FtsZ appears to be regulated by the disordered C-terminal domain of the molecule where FtsA, ZipA and MinC bind (Wang et al., 1997; Din et al., 1998; Ma and Margolin, 1999; Mosyak et al., 2000; Shen and Lutkenhaus, 2009). MinC and ZipA are known to affect the bundling of FtsZ (Hale et al., 2000; Dajkovic et al., 2008a). When this tail is deleted, as in the case of the FtsZ320 mutant, GTP binding and subsequent polymerization favour formation of bundles. This suggests that the actual magnitude of the energy of the lateral bonds between FtsZ molecules may be obscured in in vitro studies by the presence of this tail.
The tendency of FtsZ polymers to bundle is dramatically enhanced in crowded environments, such as the physical environment of the cytoplasm (Gonzalez et al., 2003; Popp et al., 2009). While bundling might help provide the force for constriction (Lan et al., 2009) and help hold Z rings together (Dajkovic et al., 2008a; Monahan et al., 2009), an excess of bundling could interfere with other essential interactions of FtsZ. In support of this, bundled FtsZ is known to be insensitive to two known regulators, MinC (Dajkovic et al., 2008a) and SulA. Furthermore, the expression of FtsZ320 in the cell is much more toxic than the expression of the same level of wild-type FtsZ (Wang et al., 1997). Taken together, these considerations suggest that while bundling may have a role in the coherence of Z rings, excessive bundling of FtsZ may have to be held in check by other factors to assure the functionality of Z rings.
How might then FtsZ polymers be organized in the Z ring? The interaction of FtsZ polymers with the membrane may play a role. Local changes in membrane curvature produced by the binding of FtsZ polymers may provide a part of the driving force for the concentration of FtsZ polymers into rings (Shlomovitz and Gov, 2009). Nevertheless, this is not sufficient, since membrane targeted FtsZ only inefficiently organizes into rings in vivo (Osawa et al., 2008). Such inefficiencies can lead to death in an appreciable fraction of the cell population, as is the case with the deletion of zapA (see Supporting information, Fig. 1 and Movies S1–S4).
Cross-linking interactions can confer mechanical coherence on a polymeric structure even in the absence of lateral ordering of polymers (Ferry, 1980; Wachsstock et al., 1993; Tseng et al., 2004). Cross-links bridge polymers and interconnect them (Fig. 8A), thus giving rigidity to a structure without imposing a liquid crystalline-like order that characterizes bundles of polymers. The advantage of this type of molecular interconnection is that the polymers involved have greater freedom to interact with other relevant factors without geometric constraints.
Cross-linking interactions between polymers are known to play an important role in the actin cytoskeleton, and a number of actin cross-linking proteins are known (Pollard and Cooper, 1986; Matsudaira, 1991; 1994; Wachsstock et al., 1993; Tseng et al., 2004). As is the case with actin cross-linking proteins (Wachsstock et al., 1993; Tseng et al., 2004), ZapA greatly increases the stiffness of FtsZ polymer structures and it bestows upon them a more solid-like character. Bundling mediated by ZapA is not sufficient to account for these effects since other known bundling factors, such as Ca2+, as well as a mutant of FtsZ with a greatly increased propensity to bundle (FtsZ320), do not produce the same effects. ZapA represents, to our knowledge, the first cross-linking protein of the bacterial cytoskeleton. Interestingly, FtsA* suppresses the cytokinesis defect caused by the absence of ZapA. It is tempting to speculate that FtsA* has cross-linking activity that allows it to suppress some cell division defects.
Further evidence that the cross-linking activity of ZapA, and not its bundling activity, is involved in the helix to ring transition and the mechanical coherence of Z rings comes from a recent study of a FtsZ mutant from B. subtilis. This mutant, FtsZ(Ts1), is unable to coalesce into rings at the non-permissive temperature but stays organized as a helix (Monahan et al., 2009). Overexpression of ZapA was shown to suppress this cytokinesis defect. Because the only known activity of ZapA at that point was its bundling activity, this finding was interpreted to suggest the importance of lateral interactions for the helix to ring transition. Nevertheless, there are several inconsistencies with this interpretation. First, the overexpression of another known FtsZ bundling protein from B. subtilis, SepF, was not able to suppress this defect. In addition, the deletion of the minC gene that codes for a known debundling factor (Dajkovic et al., 2008a), also failed to compensate for the cytokinesis defect of the FtsZ(Ts1) mutant. We suggest that the salient activity of ZapA for the stabilization of Z ring structures in this mutant is cross-linking, not bundling, and that this is the reason why the overexpression of ZapA suppressed the cell division defect, but the deletion of minC and the overexpression of SepF had no effect.
A recent study has suggested that another cell division protein, ZapB, may mediate cross-linking interactions between FtsZ polymer by binding to ZapA (Galli and Gerdes, 2010). Our results suggest that ZapA itself has cross-linking activity so it remains to be determined how much ZapB enhances this effect.
Stoichiometry and valence of FtsZ–ZapA interaction
In addition to the novel cross-linking activity, we found that ZapA promotes the nucleation of FtsZ polymerization by stabilizing longitudinal bonds between FtsZ molecules. How might ZapA have such functional plasticity?
The contours of the answer come from the finding that that ZapA mediates the formation of large complexes with monomeric FtsZ (Supporting information and Figs S4 and S5). This argues for the existence of multiple binding sites for ZapA on the FtsZ molecule. ZapA exists in the dimer–tetramer equilibrium in vitro (Low et al., 2004; Small et al., 2007). If only one binding site for ZapA existed on the FtsZ molecule, only four monomers of FtsZ could complex with tetramers of ZapA. This complex could not be detected by sedimentation assays. In contrast, two binding sites on the FtsZ molecule permit the formation of large interconnected networks between monomers of FtsZ and dimers or tetramers of ZapA that could easily be sedimented. EM studies corroborate the existence of such networks whose nature does not appear polymeric, but rather rope-like. In addition, SulA and MinC, two molecules with disparate binding surfaces on FtsZ, inhibit the formation of these large complexes, further arguing for the presence of multiple binding sites of ZapA on FtsZ. Such versatility of binding is not without precedent in the eukaryotic cytoskeleton (Egelman, 2004)
This finding is consistent with observations for the B. subtilis ZapA protein, which also forms large complexes with monomeric FtsZ in the presence of GDP (Gueiros-Filho and Losick, 2002). This raises the possibility that the polyvalence of interaction between ZapA and FtsZ may be an evolutionarily conserved character. The polyvalence of ZapA–FtsZ interaction is also consistent with the observed cross-linking of FtsZ polymers. In cross-linking interactions, ZapA could interact with one FtsZ molecule at one of the binding sites and with another FtsZ molecule in a different polymer on another binding site and thereby not impose a bundled arrangement of FtsZ structures.
The dynamics of ZapA–GFP associated with the cytokinetic ring are very similar to the dynamics of FtsZ–GFP reported by other studies. We measured the half-life of ZapA to be approximately 11 s while the half-life of FtsZ molecules in the ring has been reported to be approximately 30 s (Stricker et al., 2002) and approximately 8 s (Chen and Erickson, 2005). These very similar half-lives suggest that the dynamics of ZapA in the Z ring may be driven through its interaction with FtsZ.
The stoichiometry of the ZapA–FtsZ interaction in vitro that produced the stiffest and most solid-like structures of FtsZ by cross-linking FtsZ polymers is nearly identical to the stoichiometry of FtsZ and ZapA estimated for the cytokinetic ring. This further strengthens the notion that ZapA is involved in cross-linking interactions in vivo as suggested by a previous study (Mohammadi et al., 2009) and that these cross-linking interactions play a role in the mechanical coherence of Z rings.
Construction of the zapA::CAT strain
To interrupt the zapA gene, we started by amplifying the chloramphenicol acetyl transferase (CAT) gene by polymerase chain reaction (PCR) from the pACYC184 plasmid using the oligos 5′BamHI–Cm (5′actaggatcc-cagtagctgaacagg) and 3′BssSI–Cm (5′actacacgag-aggcgtagcaccagg). This gave a 1049 bp DNA fragment containing the CAT gene flanked by the BssSI and BamHI restriction sites. Throughout, the (–) in the oligo sequence indicates the beginning of the homology with the relevant gene. The sequences before (–) are primer extensions included to engineer the restriction sites.
Two DNA fragments with homology to the chromosomal region on each side of the zapA (ygfE) gene were also amplified by PCR. The 438 bp DNA fragment that carries the 5′ of zapA and the 3′ of the upstream gene (ygfB) was amplified using the oligos 5′EcoRI–ZapAChr (5′gggtgaattc-ggtggcagagtgcattttac) and 3′BamHI–ZapAChr (5′acgtggatcc-tggatatcgacgggttgtgc). The 309 bp DNA fragment that carries the 3′ end of zapA was amplified using the oligos 5′BamHI–ZapAChr (5′gcaaggatcc-gcaaagactcgtgactacgc) and 3′HindIII–ZapAChr (5′aaggaagctt-tgtcgtcgcagttttaaggc).
pSEB240 was constructed by cloning the 438 bp fragment containing the 5′ end of zapA into pHGB2 using the EcoRI and BamHI restriction sites. The plasmid pHGB2 is a derivative of the pGB2 plasmid (SpcR), which is temperature sensitive for replication. The 309 bp fragment containing the 3′ end of zapA was then cloned into pSEB240 using the BamHI and HindIII sites to give pSEB241. Finally, the 1049 bp fragment containing the CAT gene was cloned on pSEB241 between the two regions of homology to the zapA gene using the BamHI and BssSI sites to give pSEB242. This plasmid contained the CAT gene inserted between the 5′ and 3′ ends of zapA.
To integrate the pSEB24 plasmid on the chromosome at the zapA locus of the wild-type JS219 strain, JS219 cells containing pSEB242 were grown in Luria–Bertani (LB) containing chloramphenicol (20 µg ml−1) for 3 h before being diluted and plated at 30°C and 42°C on chloramphenicol-containing plates. The integration frequency was approximately 2.6 × 10−5 as judged by the ratio of the number of colonies obtained at 30°C and 42°C.
In order to resolve the plasmid out of the chromosome and retain the CAT gene in zapA, a mix of 10 chloramphenicol resistant colonies from the 42°C plates was used to inoculate a liquid culture that was grown at 30°C with no selection. When this culture reached the stationary phase (OD540 about 2.5), it was diluted 100 times into fresh LB and grown at 42°C. This cycle of culture at 30°C and 42°C in fresh LB with no selection was repeated and the cells were then diluted and plated on chloramphenicol at 37°C. The obtained colonies were then tested for chloramphenicol resistance and spectinomycin sensitivity (100% of the 50 colonies tested).
To verify that the plasmid was resolved out of the chromosome and that the CAT gene remained in zapA, ten colonies were selected and the insertion of the CAT cassette in the zapA gene was verified by PCR using the following oligos: 5′Xba–ZapA (5′cgtctaga-ggcatgtctgcacaacccgtc) and 3′HindIII–ZapA (5′tgaagctt-gttactctaccacagtaaaccg). The genomic DNA of wild-type JS219 cells was used as control. The genomic DNA extracted from all 10 chloramphenicol resistant colonies tested gave the expected 1.3 kb fragment. The genomic DNA extracted from the two control JS219 colonies gave the expected 0.38 kb fragment. One of the verified clones was renamed PS576 and kept in the strain collection. PS620 and all other strains carrying the zapA::CAT allele were constructed by P1 transduction.
The strains and plasmids used in this study are detailed in Table S1.
For studies of the coherence of Z rings, we utilized the strains AND101 and AND831, which have a copy of FtsZ–GFP fusion on the chromosome expressed from a modified lac promoter (Dajkovic et al., 2008b). We induced the expression of FtsZ–GFP with 40 µM IPTG in exponentially growing liquid cultures. After 2 h, the Z rings could be clearly seen by fluorescence. Under these conditions, the FtsZ–GFP represents 15–25% of the total cellular level of FtsZ (Dajkovic et al., 2008b) and does not detectably affect division. ZapA–GFP was expressed from the pZapA–GFP plasmid (Kitagawa et al., 2005) by induction with 5 µM IPTG for 1–2 h.
The cells were placed in a microculture on top of a 2 mm film of 7% low melting point agarose spread on a microscope slide. This was covered with a coverslip and observed microscopically. Experiments with cell division of lineages of single cells were done as described previously (Stewart et al., 2005).
The quantitative fluorescence microscopy and FRAP experiments were done on a Nikon Eclipse TE2000 Inverted Microscope equipped with Piezo Flexure Objective Scanner and 100 × CFI Plan Apo VC objective with a numerical aperture of 1.49. The sample was illuminated by a 491 nm laser and bleached with laser pulses of full laser intensity for 10 ms. Image acquisition was done with exposure times of 100 ms on an EMCCD camera. This combination of laser illumination, electron multiplied CCD camera and high numerical aperture lens gives single molecule sensitivity.
To compare the Z ring widths of the wild-type and the zapA::CAT strains, we took advantage of the fact that microscopic objects are blurred beyond their edges by the point spread function of the microscope. It is therefore possible for the purposes of comparison to measure and compare the apparent sizes of the objects. Smaller objects will have a smaller apparent size than bigger objects, even though their actual size may be fall below the diffraction limit of optical microscopy (Leake et al., 2006; Lenn et al., 2008).
Wild-type FtsZ protein as well as the mutants FtsZ1(A70T) and FtsZ2(D212G) were purified as described previously (Mukherjee and Lutkenhaus, 1998b; Dajkovic et al., 2008a). The cloning and purification of H6–ZapA has been described in a previous paper (Dajkovic et al., 2008a). Briefly, we overexpressed the H6–ZapA protein in the E. coli strain W3110 strain by inducing an exponentially growing culture (OD540 = 0.4) supplemented with 100 µg ml−1 of ampicillin by addition of 1 mM IPTG. After 3 h of growth at 37°C, the cells were pelleted and washed with 10 mM HEPES (pH 7.2) and frozen at −80°C. For purification, the cells were thawed and resuspended in lysis buffer (25 mM Tris pH 7.8, 100 mM NaCl, 10 mM imidazole) and passed three times through a French press. Cell debris was removed by centrifugation at 12 000 r.p.m. for 15 min. The lysate was then loaded on Ni-NTA resin (QIAGEN) pre-equilibrated with lysis buffer. The column was washed with five column volumes of high salt wash buffer (25 mM Tris pH 7.8, 500 mM NaCl, 50 mM imidazole) followed by five column volumes of low salt wash buffer (25 mM Tris pH 7.8, 300 mM NaCl, 30 mM imidazole). The bound protein was eluted with elution buffer (25 mM Tris pH 7.8, 300 mM NaCl, 300 mM imidazole). The peak fractions were dialysed against the storage buffer (20 mM HEPES pH 7.2, 0.1 mM EDTA and 100 mM KCl), concentrated and frozen at −80°C until use. We were able to obtain milligram quantities of protein from 4 l of culture. The purification of MalE–SulA and MalE–MinC was done as previously described (Dajkovic et al., 2008a,b).
For quantitative rheometry experiments, we used a strain-controlled rheometer (ARES-100; Rheometrics, Piscattaway, NJ). The rheometer is equipped with a lower plate and an upper cone (Xu et al., 2000; Tseng and Wirtz, 2001, Tseng et al., 2002). The polymerization reactions are placed between the plate and the cone. The plate is connected to a computer-controlled motor that applies either steady or oscillatory deformations of defined frequencies and amplitudes. The forces transmitted through the network as a result of the applied deformation are measured by the torque-transducer connected to the cone. The degree to which the mechanical energy from the applied deformation is maintained in the network, and not dissipated, gives the value of the reported elastic modulus (in units of force per area: dyn cm−2). In other words, the elastic modulus measures the tendency of materials to rebound to their initial configuration after imposed deformations. A purely liquid material (like water) dissipates all the energy from the imposed deformation. A purely elastic material (like rubber) maintains the energy from the imposed deformation and rebounds to its initial configuration (Heidemann and Wirtz, 2004; Janmey et al., 2007).
The gelation reactions for polymer networks were conducted at FtsZ concentrations of 25 µM. This is a concentration where FtsZ forms a gel and the elastic modulus is appreciable and reproducibly measurable. The reactions were initiated by adding buffered GTP (10 mM final concentration) and the mixtures were placed on the plate of the rheometer with a micropipette whose tip was cut off to minimize shear during transfer of the mixture to the plate. Dead time was approximately 30 s. Note that the dead time is longer than the kinetics of FtsZ polymerization, which reaches a steady state in less than 10 s (Chen and Erickson, 2005). The cone was brought in contact with the mixture and the elastic modulus was measured as a function of time. The measurements of gelation kinetics were done at low amplitudes of deformations (1%), which do not perturb the structure of the polymer networks, and at low frequencies (1 rad s−1), where the elastic modulus is independent of frequency. To examine the effects of ZapA, H6–ZapA at the indicated concentrations was added to reactions containing FtsZ. The reactions were polymerized by addition of GTP and loaded on the rheometer. All reactions were done in the standard buffer used for FtsZ polymerization, the Pol buffer (50 mM MES pH 6.5, 50 mM KCl, 10 mM MgCl2) (Mukherjee and Lutkenhaus, 1998b; Mukherjee et al., 1998; Dajkovic et al., 2008a,b). All rheology experiments were repeated at least three times to assure reproducibility and assess variability between experiments. The reactions were examined by EM to verify that the rheological behaviour is due to ordered structures and not due to non-specific aggregates. We observed no differences in the final rheological properties of FtsZ/ZapA networks when ZapA was added after the polymerization of FtsZ.
To study the dynamic properties of FtsZ polymer networks in the presence of ZapA, oscillatory deformations of small amplitude (1%) and of frequencies between 0.01 and 100 rad s−1 were applied to measure the elastic modulus.
To examine the resilience polymer networks, oscillatory deformations of fixed frequencies (1 rad s−1) and strain amplitudes between 0.1% and 500% were applied.
EM and quantification of polymer width
FtsZ polymers were visualized by negative staining on carbon-coated copper grids (300 mesh). Ten microlitres of a polymerization reaction was carefully transferred from the tube to the grid by a micropipette whose tip was cut off to minimize shear during transfer. The mixture was immediately blotted away and 10 µl of 1% uranyl acetate was dropped on the grid and immediately blotted away. The grids were air-dried and imaged. The images were captured on film and the negatives were drum-scanned (Custom Color, Kansas City, MO) at 2500 pixels per inch. For quantification of thickness of polymer species, the digitized images were analysed using ImageJ software [National Institutes of Health (NIH)]. A minimum of 200 measurements were done for each reported value.
The authors would like to thank the Nikon Imaging Center at the Curie Institute in Paris for the use of their microscopes and Marina Elez for a critical reading of the manuscript. This work was supported by NIH CA143868 to Denis Writz, by NIH GM075305 to Denis Wirtz and Sean X. Sun and by NIH GM29764 to Joe Lutkenhaus. Alex Dajkovic benefits from an EMBO Long Term Fellowship and would especially like to express gratitude for the support of Jan Tiplick and the EMBO Long Term Fellowship Committee.