In vivo organization of the FtsZ-ring by ZapA and ZapB revealed by quantitative super-resolution microscopy


  • Jackson Buss,

    1. Department of Biophysics and Biophysical Chemistry, The Johns Hopkins University School of Medicine, Baltimore, MD, USA
    Search for more papers by this author
  • Carla Coltharp,

    1. Department of Biophysics and Biophysical Chemistry, The Johns Hopkins University School of Medicine, Baltimore, MD, USA
    Search for more papers by this author
  • Tao Huang,

    1. Department of Biophysics and Biophysical Chemistry, The Johns Hopkins University School of Medicine, Baltimore, MD, USA
    Current affiliation:
    1. Department of Biomedical Engineering, The Oregon Health and Science University, Portland, OR, USA
    Search for more papers by this author
  • Chris Pohlmeyer,

    1. Department of Biophysics and Biophysical Chemistry, The Johns Hopkins University School of Medicine, Baltimore, MD, USA
    Current affiliation:
    1. Department of Molecular Biology and Genetics, The Johns Hopkins University School of Medicine, Baltimore, MD, USA
    Search for more papers by this author
  • Shih-Chin Wang,

    1. Department of Biophysics and Biophysical Chemistry, The Johns Hopkins University School of Medicine, Baltimore, MD, USA
    Search for more papers by this author
  • Christine Hatem,

    1. Department of Biophysics and Biophysical Chemistry, The Johns Hopkins University School of Medicine, Baltimore, MD, USA
    Current affiliation:
    1. Department of Biophysics, The Johns Hopkins University, Baltimore, MD, USA
    Search for more papers by this author
  • Jie Xiao

    Corresponding author
    • Department of Biophysics and Biophysical Chemistry, The Johns Hopkins University School of Medicine, Baltimore, MD, USA
    Search for more papers by this author

For correspondence. E-mail; Tel. (+1) 410 614 1760; Fax (+1) 410 502 7221.


In most bacterial cells, cell division is dependent on the polymerization of the FtsZ protein to form a ring-like structure (Z-ring) at the midcell. Despite its essential role, the molecular architecture of the Z-ring remains elusive. In this work we examine the roles of two FtsZ-associated proteins, ZapA and ZapB, in the assembly dynamics and structure of the Z-ring in Escherichia coli cells. In cells deleted of zapA or zapB, we observed abnormal septa and highly dynamic FtsZ structures. While details of these FtsZ structures are difficult to discern under conventional fluorescence microscopy, single-molecule-based super-resolution imaging method Photoactivated Localization Microscopy (PALM) reveals that these FtsZ structures arise from disordered arrangements of FtsZ clusters. Quantitative analysis finds these clusters are larger and comprise more molecules than a single FtsZ protofilament, and likely represent a distinct polymeric species that is inherent to the assembly pathway of the Z-ring. Furthermore, we find these clusters are not due to the loss of ZapB–MatP interaction in ΔzapA and ΔzapB cells. Our results suggest that the main function of ZapA and ZapB in vivo may not be to promote the association of individual protofilaments but to align FtsZ clusters that consist of multiple FtsZ protofilaments.


Cell division in Escherichia coli requires the assembly of the divisome, a macromolecular complex that is composed of more than 10 essential proteins (Harry et al., 2006; Erickson et al., 2010). Central to divisome assembly is the formation of a ring-like structure (Z-ring) at midcell through the polymerization of the prokaryotic tubulin homologue, FtsZ (Lutkenhaus et al., 1980; Bi and Lutkenhaus, 1991; Dai and Lutkenhaus, 1991; Ma et al., 1996; Addinall et al., 1997; Nogales et al., 1998). The Z-ring serves as a scaffold for the recruitment of all other essential division proteins and possibly generates the constrictive force required for septation (Addinall et al., 1996; Buddelmeijer and Beckwith, 2002; Osawa et al., 2009). Given its essential role in the cell division process and its highly conserved nature in prokaryotes (Lutkenhaus, 1993), it is important to understand the molecular architecture and assembly pathway underlying the Z-ring structure.

In vitro studies have shown that FtsZ can self-polymerize into short, single-stranded protofilaments, which are believed to serve as the basic structural unit of the Z-ring in vivo (Bramhill and Thompson, 1994; Mukherjee and Lutkenhaus, 1994; Romberg et al., 2001). Under different conditions these single-stranded protofilaments can further associate into a variety of multi-stranded structures – sheets, rings, bundles, toroids and helices (Erickson et al., 1996b; Popp et al., 2009). In a reconstituted liposome system, membrane-tethered FtsZ were found to form ring-like structures (Osawa et al., 2008); corresponding electron microscopy (EM) images showed that these membrane-tethered FtsZ predominantly assembled into tightly packed, ribbon-like configurations (Milam et al., 2012).

While the in vitro polymerization properties of FtsZ have been extensively characterized (Erickson et al., 1996a; Mukherjee and Lutkenhaus, 1999; Caplan and Erickson, 2003; Oliva et al., 2004; Chen et al., 2005; Esue et al., 2005), attempts to determine the in vivo structure of the Z-ring have been difficult. Recently a series of studies using advanced high-resolution microscopy methods have begun to shed light on the in vivo structure of the Z-ring. In Caulobacter crescentus, electron cryotomography has shown that the Z-ring consists of a few short, single FtsZ protofilaments scattered around the division site (Li et al., 2007). In E. coli, we used a single-molecule-based fluorescence method, Photoactivated Localization Microscopy (PALM) (Betzig et al., 2006), to image the Z-ring. We found that the apparent width of the Z-ring is ∼ 110 nm and in some regions of the Z-ring the molecule density is higher than a single layer of protofilaments (Fu et al., 2010). These results suggested that the Z-ring consists of loosely associated protofilaments randomly overlapping in both the longitudinal and radial directions. A later three-dimensional super-resolution study found that the Z-ring in C. crescentus has a similar width (60–110 nm) and radial thickness (Biteen et al., 2012). Finally, in Bacillus subtilis and Staphylococcus aureus, stimulated emission depletion microscopy (STED) and structured illumination microscopy (SIM) illustrated the highly irregular and discontinuous nature of the Z-ring, whereby regions of high and low densities were interspaced along the midcell plane (Jennings et al., 2011; Strauss et al., 2012).

One observation common to all of these in vivo studies is that FtsZ protofilaments appear to adopt a loose, heterogeneous arrangement within the Z-ring. This observation is in contrast to the largely ordered and aligned arrangements of protofilaments observed in vitro, where protofilaments associate with each other through intrinsic lateral interactions (Erickson et al., 1996a; Milam et al., 2012). Hence, it is likely that factors other than the intrinsic lateral interactions are responsible for organizing the loose, heterogeneous arrangement of FtsZ protofilaments in vivo.

Recently, a group of FtsZ-associated proteins (ZapA, ZapB, ZapC and ZapD) have been identified and found to promote Z-ring assembly in vivo (Gueiros-Filho and Losick, 2002; Small et al., 2007; Ebersbach et al., 2008; Durand-Heredia et al., 2011; 2012; Hale et al., 2011). Deletion of a single zap gene is not lethal, but often leads to abnormal Z-ring morphology visible to conventional fluorescence microscopy (Gueiros-Filho and Losick, 2002; Ebersbach et al., 2008; Dajkovic et al., 2010). Deletion of multiple zap genes results in severe cell division defects, suggesting that these Zap proteins have overlapping, but important functions in cell division (Small et al., 2007; Durand-Heredia et al., 2011; 2012; Hale et al., 2011). To gain a deeper understanding of the in vivo contribution of Zap proteins to Z-ring structure and function, in this study we provide a quantitative characterization of the Z-ring in the absence of ZapA and ZapB.

ZapA is a small cytoplasmic protein (∼ 12 kDa) that forms a dimer, which further associates into a tetramer upon increased concentration (Low et al., 2004; Pacheco-Gómez et al., 2013). ZapA binds FtsZ directly with high affinity (Galli and Gerdes, 2012) and promotes bundling of FtsZ protofilaments in vitro (Gueiros-Filho and Losick, 2002; Small et al., 2007; Mohammadi et al., 2009; Monahan et al., 2009; Dajkovic et al., 2010). ZapB is also a small cytoplasmic protein (∼ 10 kDa) that can self-associate; its α-helical nature enables extensive coiled-coil interactions, resulting in the formation of long ZapB filaments and bundles in vitro (Ebersbach et al., 2008; Galli and Gerdes, 2012). ZapB does not bind FtsZ directly (Galli and Gerdes, 2012) but is associated with the Z-ring through its direct interaction with ZapA (Galli and Gerdes, 2010). Deletion of zapA or zapB results in cells that, while viable, display similar abnormal FtsZ structures, such as arcs, spirals and broad, diffusive bands (Ebersbach et al., 2008; Dajkovic et al., 2010). The similarity in deletion phenotype and ability to associate directly with each other suggests that these two proteins may function as a complex or in the same pathway to promote the assembly of the Z-ring. Due to the limited spatial resolution and qualitative nature of conventional fluorescence microscopy, the structural details of abnormal Z-rings formed in the absence of zapA or zapB are difficult to discern. In this study we investigated cell division defects and characterized the structure and dynamics of the Z-ring in the absence of zapA or zapB using a combination of high-resolution imaging techniques and quantitative analyses.


ΔzapA and ΔzapB cells display abnormal septa

We first characterized the effect of zapA or zapB deletion on cell length and division rates. We grew ΔzapA and ΔzapB cells (Baba et al., 2006) in liquid M9+ media and found that cells doubled with rates similar to the wild-type (wt) parental strain BW25113 (Datsenko and Wanner, 2000) (Supplemental Fig. S1). However, when observed under a microscope, ΔzapA and ΔzapB cells exhibited significant increases in cell length (4.3 ± 1.1 μm, 4.2 ± 1.2 μm and 3.5 ± 0.6 μm for ΔzapA, ΔzapB and wt respectively, Supplemental Fig. S2). These findings are consistent with previous reports (Ebersbach et al., 2008; Dajkovic et al., 2010; Durand-Heredia et al., 2011).

We then used scanning electron microscopy (SEM) to examine septum morphology. We found that ΔzapA and ΔzapB cells often displayed abnormal septa that were oriented non-perpendicular to the cell's long axis, or not placed precisely at the midcell (Fig. 1). In addition, about 20% of ΔzapA (n = 132) or ΔzapB cells (n = 197) contained more than one septum (Fig. 1). In contrast, all constricting wt cells possessed a single furrow, oriented perpendicular to the long axis and aligned at or near midcell. This multi-septa phenotype is qualitatively different from the highly twisted septa caused by FtsZ GTPase variants FtsZ2 and FtsZ26 (Bi and Lutkenhaus, 1992; Addinall et al., 1997), or the mini-cell septa produced by FtsZ overexpression (Ward and Lutkenhaus, 1985). Because it was shown that septal geometry is determined by the geometry of underlying FtsZ structures (Addinall and Lutkenhaus, 1996), the abnormal morphology and positioning of septa observed for ΔzapA and ΔzapB cells suggest that the morphology and positioning of FtsZ structures in the absence of ZapA and ZapB are also abnormal.

Figure 1.

Scanning electron micrographs of dividing E. coli cells. (A) BW25113 cells show single, midcell septa oriented perpendicular to the cells’ long axes. In ΔzapA (B, C) and ΔzapB (D, E, F) cells, slanted (▲), non-midcell (△) and multiple (↑) septa are observed. Cells were imaged on PLL-treated coverslips (A–E) or on 0.2 μm filters in the absence of PLL (F). Scale bars, 500 nm.

FtsZ structures are highly dynamic in ΔzapA or ΔzapB cells

To investigate the effect of ZapA and ZapB on the morphology and positioning of FtsZ structures, we ectopically expressed a partially functional FtsZ–GFP fusion protein (Kitagawa et al., 2005; Fu et al., 2010) in the ΔzapA or ΔzapB strains. We monitored the green fluorescence of FtsZ–GFP during cell division using conventional fluorescence light microscopy. Due to the leaky expression of the T5-lac promoter, FtsZ–GFP fluorescence was clearly visible in uninduced cells and was found to localize to midcell in ∼ 75% of the wt population (Fig. 2Ai). Although more than 60% of the ΔzapA and ΔzapB cells also showed distinct FtsZ midcell localization, a significant fraction (> 35%) of these structures were qualitatively different from the sharp band typically observed in wt cells; instead, these structures appeared to be broad and diffusive, or contained multiple bands (Fig. 2Aii–iii). Similar observations have been reported previously (Ebersbach et al., 2008; Dajkovic et al., 2010).

Figure 2.

Localization and dynamics of FtsZ–GFP in wt, ΔzapA and ΔzapB cells.

A. Snapshots of BW25113 (Ai), ΔzapA (Aii) and ΔzapB (Aiii) cells expressing FtsZ–GFP. Bright-field images and corresponding fluorescence images are displayed side by side.

B. Montages from time-lapse movies of BW25113 (Bi), ΔzapA (Bii) and ΔzapB (Biii) cells expressing FtsZ–GFP during cell division. At each time point, the fluorescence image (green) was overlaid with the corresponding bright-field image (grey) with the time (min) indicated in the bottom corner. Corresponding movies are shown in Supplemental Movies S1, S3 and S4 respectively. In wt cells, FtsZ forms a clear band at midcell early and remains there throughout the division process. In contrast, the FtsZ structures in ΔzapA and ΔzapB appear more dynamic, transitioning back-and-forth between multiple active septa.

C. Distributions of cell division times for BW25113 (red) ΔzapA (blue) or ΔzapB (black) cells determined from time-lapse movies. The average division time of wt cells is significantly different from that of ΔzapA and ΔzapB (ks-test, P < 0.01). Subpopulations of ΔzapA and ΔzapB cells in which FtsZ relocalized to visibly constricted sites in newborn daughter cells (e.g. Bii 145′, 190′ and 245′ top cells; and Biii 135′, 225′ and 230′ middle cells) divided significantly faster than the general populations with division times of 83 ± 11 min (n = 15) and 77 ± 13 min (n = 6) respectively. Mean ± standard deviation (number of division events).

Scale bars, 1 μm.

To monitor the dynamics of FtsZ structures during cell division in the absence of ZapA or ZapB, we observed exponentially growing ΔzapA and ΔzapB cells harbouring the FtsZ–GFP plasmid without induction on a microscope stage for multiple generations. In contrast to the relatively stationary localization of FtsZ–GFP in wt cells (Fig. 2Bi, Supplemental Movies S1 and S2), we observed highly dynamic FtsZ–GFP structures transitioning back and forth between multiple sites in ΔzapA and ΔzapB cells (Fig. 2Bii and 2Biii, Supplemental Movies S3–S5). These dynamic FtsZ–GFP structures often led to visible cell wall indentations, likely corresponding to the multiple septa observed by SEM (Fig. 1). At late stages of the cell cycle these FtsZ–GFP structures appeared to consolidate to a single site, where the cell completed septation (Fig. 2Bii 175′ and 2Biii 50′). Once division completed, FtsZ–GFP quickly localized to the midcell region of both daughter cells. Interestingly, in cases where FtsZ–GFP localized to a site in a newborn daughter cell that already possessed a clear indentation, presumably due to previous occupation by FtsZ, cell division ensued more than twofold faster than those with no visible indentation (Fig. 2Bii–iii and C, Supplemental Fig. S3). Consequently, the distributions of cell generation times for actively dividing ΔzapA and ΔzapB cells are shifted to shorter times relative to the wt population (Fig. 2C). We note that the difference in relative division rates observed by time-lapse microscopy (Fig. 2C) versus liquid media growth (Supplemental Fig. S1) can be explained by the increased death rate caused by zapA (and presumably zapB) deletion (Dajkovic et al., 2010). The death rate contributes to the bulk doubling time measured in liquid media, but not to the division time measured from our time-lapse imaging analysis, which only included cells that divided successfully. Thus, the higher death rate of ΔzapA and ΔzapB cells may be compensated by faster division times to result in bulk doubling times that are similar to wt cells.

Ring-like FtsZ structures in ΔzapA and ΔzapB cells resemble wild-type Z-rings

We showed in Fig. 2A that in the absence of ZapA and ZapB, FtsZ forms broad and diffusive, or extended, multi-band structures in addition to the single-band structure characteristic of the Z-ring. However, details of these FtsZ structures are not discernible under conventional fluorescence light microscopy due to diffraction-limited resolution. We employed the single-molecule-based super-resolution imaging method PALM (Betzig et al., 2006), which enables quantitative structural measurements (Ulbrich and Isacoff, 2007; Annibale et al., 2011; Sengupta et al., 2011; Coltharp et al., 2012; Lando et al., 2012; Renz et al., 2012), to investigate these FtsZ structures. Comparing FtsZ structures formed in a wt background to those assembled in the absence of ZapA or ZapB should provide insight into how ZapA and ZapB help promote Z-ring assembly.

We performed PALM imaging using an FtsZ-mEos2 construct described previously (Fu et al., 2010; Coltharp et al., 2012). We imaged live cells where FtsZ-mEos2 was ectopically expressed at ∼ 30% of the total cellular FtsZ level in BW25113, ΔzapA and ΔzapB (Supplemental Fig. S4). Since only slowly moving molecules could be detected at our imaging frame rate, live-cell PALM images comprise molecules that are either membrane-associated or incorporated into large superstructures. We found that of cells displaying distinct FtsZ structures, 36% of ΔzapA (n = 99) and 35% of ΔzapB (n = 103) cells showed a single, sharp band at midcell, reminiscent of the typical Z-ring observed in wt cells (Fig. 3). These Z-rings contained 69% of the total detected FtsZ-mEos2 molecules present within cells. In contrast, 50% of wt cells (n = 117) displayed ring-like structures at midcell and 76% of the detected cellular FtsZ-mEos2 molecules localized to these structures (Table 1). Despite these differences, we found that the ring-like structures in all three strains had similar widths (∼ 110 nm, P > 0.05) and diameters (∼ 1 μm, P > 0.05), suggesting that the structural organization of the ring-like FtsZ structure formed in the absence of ZapA or ZapB is similar to that in wt cells. We note that with a working resolution of ∼ 50 nm under live-cell imaging conditions (Supplemental Fig. S5), we cannot exclude differences at the molecular level.

Figure 3.

Live-cell PALM imaging of ring-like FtsZ structures. Images of live BW25113 (A–C), ΔzapA (D–F) and ΔzapB (G–I) cells expressing FtsZ-mEos2 from pJB042 are shown in the order of bright-field (i), ensemble green fluorescence image (ii) and PALM image (iii). The dotted yellow line is a general indicator of the cell outline. All cells shown here display a single band at midcell, indicative of a normal Z-ring. Scale bars, 500 nm.

Table 1. Dimensional analysis of ring-like FtsZ structures
 No. of cellsRinga width (nm)Ringb diameter (nm)% Molecules in midcell structure
  1. aFull width half maximum.
  2. bNon-constricting cells (BW25113 n = 37, ΔzapA n = 20, ΔzapB n = 22).
  3. Mean ± standard error.
BW2511358113 ± 3960 ± 1576 ± 2
ΔzapA36117 ± 51000 ± 2069 ± 3
ΔzapB36104 ± 4960 ± 2069 ± 3

Non-ring FtsZ structures in ΔzapA and ΔzapB cells are composed of dispersed clusters

In PALM images of ΔzapA and ΔzapB cells, 65% and 64% of cells showed a variety of non-ring structures respectively (Fig. 4D–O). While these non-ring FtsZ-mEos2 structures appeared to adopt broad, diffusive or extended, multi-band structures in conventional fluorescence images, the corresponding PALM images all revealed a widely dispersed, clustered appearance (Fig. 4D–O). Since only FtsZ-mEos2 molecules that moved slowly, presumably tethered to the membrane or confined in large superstructures, were detectable by live-cell PALM, we suspect that the clusters in the PALM images are composed of membrane-tethered FtsZ molecules and/or polymers.

Figure 4.

Live-cell PALM imaging of non-ring FtsZ structures. Images of live BW25113 (A–C), ΔzapA (D–I) and ΔzapB (J–O) cells expressing FtsZ-mEos2 from pJB042 are displayed in the order of bright-field (i), ensemble green fluorescence image (ii) and PALM image (iii). Although the ring-like band conformations shown in Fig. 3 are the predominant structures formed in wt cells, PALM imaging reveals that some wt cells possess punctate structures, characterized by a single focus (A), two foci (B) or multiple foci reminiscent of a compact helix (C). For ΔzapA and ΔzapB cells, while the ensemble green fluorescence images show broad, diffusive or extended multi-band structures, the corresponding PALM images all resolve into various arrangements of clusters. Scale bars, 500 nm.

In 50% of wt cells we also observed non-ring FtsZ structures that had clustered appearances (Fig. 4A–C). However, these non-ring FtsZ structures appeared qualitatively different from those observed in ΔzapA and ΔzapB cells (Fig. 4D–O); in general, FtsZ-mEos2 clusters in wt cells were either a single focus at midcell, a pair of foci at the periphery of the midcell plane, or multiple foci tightly distributed around midcell, reminiscent of the compact helical structure observed previously (Fu et al., 2010). Furthermore, the corresponding ensemble images of these wt non-ring structures resembled the single-band or peripheral dot structures typically attributed to normal Z-rings (Den Blaauwen et al., 1999; Dajkovic et al., 2010; Erickson et al., 2010), and are in stark contrast to the broad or extended ensemble images observed in the deletion strains. These observations suggest that these FtsZ clusters may be inherent to the assembly of the Z-ring, and that ZapA or ZapB may affect the composition and spatial distribution of these clusters.

To ascertain whether the clustered appearance of FtsZ-mEos2 in ΔzapA and ΔzapB cells was caused by differences in FtsZ expression levels relative to wt cells, we performed quantitative immunoblotting. We found that both the endogenous FtsZ and the FtsZ-mEos2 expression levels were similar in all three strains (Supplemental Figs S4 and S6). We also found that when ZapA or ZapB was expressed in trans in ΔzapA or ΔzapB, wt Z-ring morphology was restored (Supplemental Fig. S7). These observations suggest that FtsZ-mEos2 clusters were not produced as a result of FtsZ or FtsZ-mEos2 overexpression, and were specific to the deletion of zapA and zapB.

Next, we examined whether the observed FtsZ-mEos2 clusters in ΔzapA and ΔzapB cells were an artefact of the self-aggregation of the mEos2 protein (Landgraf et al., 2012; Swulius and Jensen, 2012). We replaced mEos2 with two other photoactivatable fluorescent proteins, Dronpa (Ando et al., 2004) and mEos3 (Zhang et al., 2012), which have been shown to be truly monomeric. We observed similarly dispersed clusters in ΔzapA and ΔzapB cells with both fluorescent proteins (Supplemental Figs S8 and S9). Additionally, we performed super-resolution immunofluorescence imaging on ΔzapA and ΔzapB cells expressing native, untagged FtsZ using stochastic optical reconstruction microscopy (STORM) (Rust et al., 2006). Again we observed dispersed FtsZ clusters similar to what were observed in FtsZ-mEos2 PALM imaging (Supplemental Fig. S10). These observations demonstrate that the FtsZ-mEos2 clusters in ΔzapA and ΔzapB cells are not fluorescent-protein specific but reflect the intrinsic property of native FtsZ in the absence of ZapA or ZapB.

To test whether the FtsZ-mEos2 clusters were dependent on FtsZ depolymerization, we examined the effect of SulA, a negative regulator of FtsZ polymerization that sequesters FtsZ monomers (Dajkovic et al., 2008; Chen et al., 2012). In cells overexpressing SulA, we observed a homogenous, diffusive fluorescence signal for FtsZ-mEos2 by conventional fluorescence and detected very few FtsZ-mEos2 molecules via PALM (Supplemental Fig. S11B). This observation is consistent with the notion that SulA-sequestered FtsZ monomers are largely cytoplasmic, and thus undetectable by live-cell PALM. Finally, when wt cells were treated with cinnamaldehyde, a known inhibitor of Z-ring assembly (Domadia et al., 2007), we found that FtsZ-mEos2 assembled into large, cohesive clusters at the cell poles (Supplemental Fig. S11C), in contrast to the dispersed FtsZ-mEos2 clusters observed in the ΔzapA and ΔzapB strains (Fig. 4D–O). These results suggest that the widely dispersed FtsZ-mEos2 clusters are likely composed of polymerized FtsZ molecules.

FtsZ clusters in ΔzapA and ΔzapB cells are quantitatively different from those in wt cells

We have previously shown that quantitative measurements of dimensions and molecule density of the Z-ring can be obtained from PALM images, as PALM is a single-molecule-based super-resolution method (Fu et al., 2010; Coltharp et al., 2012). To reduce the measurement uncertainty associated with two-dimensional projections of three-dimensional cellular structures in wide-field illumination, we performed total internal reflection (TIR) PALM imaging on the three strains. TIR-PALM confines activation and excitation of FtsZ-mEos2 molecules to a thin layer (∼ 200 nm) at the interface of the coverslip and cell, hence only FtsZ-mEos2 molecules at the bottom membrane of the cell are selectively imaged, avoiding contributions of FtsZ-mEos2 molecules from the cytoplasm and top membrane (Fu et al., 2010).

Using TIR-PALM, we observed similar ring and non-ring structures for FtsZ-mEos2 (Supplemental Fig. S12) as in wide-field PALM (Figs 3 and 4). We then applied the same thresholding algorithm to all three strains to isolate individual clusters (Supplemental Fig. S13A). Since ring structures were also heterogeneous and slightly punctated, we included both ring and non-ring structures in the analysis to avoid bias. For all three strains, we found that 80% of FtsZ-mEos2 localizations were contained within 23–33% of the identified clusters (Supplemental Fig. S13B), suggesting that most detected FtsZ-mEos2 molecules were confined in a few large clusters and were not evenly distributed. We selected these large clusters and determined the following four parameters for each cluster: number of detected FtsZ-mEos2 molecules, area occupied, molecule density (number of molecules per unit area), and displacement from the midcell plane (Fig. 5A).

Figure 5.

Cluster analysis of FtsZ-mEos2 TIR-PALM images.

A. (i–iii) Images of a BW25113 cell expressing FtsZ-mEos2 are shown in the order of bright-field (i), ensemble green fluorescence image (ii) and TIR-PALM image (iii). A threshold was applied to the TIR-PALM image to generate a binary image (iv) where the white region indicates an identified cluster. Using the cluster coordinates we determined the number of molecules localized within each identified region and the overall size of the region. (v) Expanded view of the boxed area in iv illustrates additional cluster measurements: major axis length (a), minor axis length (b), displacement of centroid (c) from midcell plane and orientation of major axis relative to midcell plane (d).

B–I. FtsZ-mEos2 clusters observed in BW25113 (red), ΔzapA (blue) and ΔzapB (black) cells were compared for molecule counts (B), size (C), density (D), location (E), shape (F–G) and orientation (I). Histograms for each measurement were plotted using the same bin size (max/10) and normalized by the total number of counts. These measurements are summarized in Table 2. All the measured properties of clusters in ΔzapA and ΔzapB showed significant differences (P < 1e-5) from wt, with ΔzapB clusters closer to wt values. This observation was true independent of threshold value (Supplemental Table S1).

Scale bars, 500 nm.

Table 2. Summary of PALM cluster analysis
StrainNo. of clustersNo. of moleculesArea (nm2)Density (mol nm−2)Distance from midcell (nm)Major axis length (nm)Minor axis length (nm)Orientation (deg)
  1. Mean ± standard error. Standard errors were determined by bootstrapping with 1000 simulations.
BW25113158154 ± 869 000 ± 24000.0022 ± 0.000181 ± 8508 ± 18206 ± 415 ± 1
ΔzapA24648 ± 328 000 ± 12000.0017 ± 0.0001280 ± 23255 ± 9154 ± 334 ± 2
ΔzapB18877 ± 639 000 ± 16000.0018 ± 0.0001194 ± 18327 ± 12170 ± 327 ± 2

We found that on average FtsZ-mEos2 clusters of ΔzapA and ΔzapB cells contained fewer molecules (< 50% wt), occupied less area (< 60% wt), were less dense (∼ 80% wt), and were displaced threefold farther away from the midcell plane than wt clusters (Fig. 5B–E, Table 2). All differences were statistically significant (P < 1e-5). We also observed small, but significant differences (P < 0.01) in all measurements between ΔzapA and ΔzapB, except for the molecule density measurement (P = 0.06). Interestingly, all measurements of ΔzapB were closer to wt than those of ΔzapA were to wt. None of the four measurements showed a strong correlation with cell length (Supplemental Fig. S14), suggesting that a cell cycle-dependent progression of cluster formation is unlikely.

Next, we analysed the shapes of these clusters by comparing the lengths and orientations of the major and minor axes of individual clusters among the three strains (Fig. 5A and F–I, Table 2, Supplemental Fig. S15). In wt cells, a ring-like FtsZ-mEos2 structure would have a major axis length proportional to the ring diameter, a minor axis length proportional to the width of the Z-ring, and an orientation parallel to the cell's short axis. We found that on average the major axis lengths of ΔzapA and ΔzapB clusters were ∼ 50% the length of wt clusters, whereas the minor axis lengths of ΔzapA and ΔzapB clusters were only reduced by ∼ 20%. Interestingly, the major axis length distributions for all three strains exhibited a twofold larger variation inline image than the minor axis length distribution, suggesting that clusters predominantly differ from each other in length rather than width (Fig. 5F and G). Furthermore, we found the ratio of major:minor axis length was significantly larger than one for all strains (Fig. 5H), indicating that clusters preferentially adopt elongated shapes. Finally, in contrast to wt clusters where the major axis is largely parallel to the short axis of the cell (within ± 15°), that of ΔzapA and ΔzapB clusters were tilted farther away (± 30°) from the short axis of the cell (Fig. 5I).

FtsZ-mEos2 molecules in clusters are stationary

We have shown that FtsZ clusters formed in the absence of ZapA or ZapB, while smaller than those in wt cells, still contain large numbers of FtsZ molecules (∼ 50–70 molecules per cluster). To further investigate the polymerization state of FtsZ within these clusters, we measured and compared the mobility of individual FtsZ-mEos2 molecules in the wt and deletion strains using single molecule tracking. We reasoned that if these FtsZ molecules are in polymerized forms, their mobility would be significantly restricted relative to individual, membrane-tethered FtsZ monomers. We used very low activation power to turn on one FtsZ-mEos2 molecule at a time and tracked its movement every 200 ms with an exposure time of 50 ms. Figure 6A shows two typical tracking trajectories of FtsZ-mEos2 molecules in ΔzapA cells. We then used trajectories that lasted for at least 10 frames (2 s in total) to compute the mean squared displacement (MSD) at different time lags (Fig. 6B, n = 136, 431, and 211 for wt, ΔzapA, ΔzapB respectively). The MSD plots of all three strains can be well described by a simple random Brownian diffusion model on the timescale from 0.2 to 2 s: the MSD scales linearly with the time lags; the slope is proportional to the apparent 2D diffusion coefficient and the y-intercept is determined by the experimental spatial resolution. The observed apparent 2D diffusion coefficients for FtsZ-mEos2 in wt, ΔzapA and ΔzapB cells were 0.0004 ± 0.00003, 0.0005 ± 0.00001 and 0.0006 ± 0.00003 μm2 s−1 respectively. While there is a slight increase in the mobility of FtsZ-mEos2 molecules in the absence of ZapA or ZapB, these diffusion coefficients are orders of magnitude smaller than that expected for a typical freely diffusing inner membrane protein (0.01–0.1 μm2 s−1) (Deich et al., 2004; Kim et al., 2006; Leake et al., 2006; Mullineaux et al., 2006). The slow, essentially immobile FtsZ-mEos2 molecules in the clusters are also consistent with the stationary FtsZ-Dendra2 population reported previously (Niu and Yu, 2008), suggesting that these clusters are likely composed of membrane-tethered FtsZ polymers. Consistent with this notion, the single-step displacement (200 ms) histogram showed primarily a single population that centred around 75 nm (Fig. 6C), suggesting that a significant population of mobile FtsZ-mEos2 does not exist in the absence of ZapA or ZapB.

Figure 6.

Single molecule tracking of FtsZ-mEos2.

A. Representative single-molecule trajectories of consecutive frames for a mobile (Ai) and immobile (Aii) FtsZ-mEos2 molecule in ΔzapA. Fluorescence images were acquired every 200 ms.

B. Mean squared displacements (closed circles) are plotted at different time lags for BW25113 (red), ΔzapA (blue) and ΔzapB (black) cells. Error bars indicate standard error. The data were fit to a linear equation (y = 4Dx + A). The diffusion coefficients for BW25113, ΔzapA and ΔzapB strains determined from the fits are 0.0004 ± 0.00003, 0.0005 ± 0.00001, 0.0006 ± 0.00003 μm2 s−1 respectively.

C. Histograms of single-step (200 ms) displacements of the three strains showing predominate immobile populations centred around 75 nm.

Scale bars, 1 μm.

Abnormal Z-ring localization observed in ΔmatP only under fast growth condition

It was recently reported that MatP, a DNA-binding protein that condenses the Ter macrodomain (MD), directly interacts with ZapB (Mercier et al., 2008; Espeli et al., 2012). Since deletion of ZapA or ZapB results in early segregation of the Ter MD through the loss of the ZapB–MatP interaction (Espeli et al., 2012), it is possible that the dispersed FtsZ clusters we observed in ΔzapA and ΔzapB cells did not result directly from the loss of ZapA or ZapB, but were instead caused by the loss of ZapB–MatP interaction. In this scenario, the loss of ZapB–MatP interaction would lead to an abnormally positioned nucleoid, resulting in aberrant distribution pattern of the nucleoid occlusion protein, SlmA, and consequently mislocalized FtsZ structures (Bernhardt and de Boer, 2005; Tonthat et al., 2011; Cho and Bernhardt, 2013).

To examine this possibility, we first investigated the localization of FtsZ–GFP in a ΔmatP strain. We found that under our slow growth condition (M9+ at RT, T ≈ 160 min), FtsZ–GFP localization in ΔmatP cells was indistinguishable from that in wt cells (Fig. 7A). PALM imaging of FtsZ-mEos2 under the same growth condition confirmed this observation (Supplemental Fig. S16A–D). These results are consistent with previous reports that the deletion of matP does not result in any distinguishable phenotype under slow growth conditions (Mercier et al., 2008) and suggest that the dispersed FtsZ clusters we observed in ΔzapA and ΔzapB cells are not caused by the loss of ZapB–MatP interaction.

Figure 7.

FtsZ and SlmA localization under slow and fast growth.

A and B. Representative images of live ΔmatP cells expressing FtsZ–GFP under slow (A) or fast (B) growth conditions. A bright-field image is displayed atop an ensemble fluorescence image.

C and D. Individual ΔslmA, ΔzapA, ΔzapB and ΔmatP cells expressing mEos2–SlmA in the absence of inducer under the slow growth condition (C) and GFP–SlmA under the fast growth condition (D). Each bright-field image is displayed next to the corresponding ensemble fluorescence image (full dynamic range), which is adjacent to an overlaid, intensity-adjusted image.

Scale bars, 1 μm.

Previous reports have shown that under fast growth conditions, deletion of matP results in severe chromosome segregation and cell division defects (Mercier et al., 2008). Therefore, we next investigated the localization of FtsZ–GFP in ΔmatP, ΔzapA and ΔzapB cells under a fast growth condition (rich defined media at 37°C, T ≈ 45 min). We found that in addition to a slight increase in cell length (Supplemental Fig. S16E), 17% (n = 69) of ΔmatP cells also displayed abnormal FtsZ–GFP localizations (Fig. 7B). In contrast, we did not observe a significant increase in the percentage of cells displaying abnormal FtsZ structures in ΔzapA (44%, n = 52) and ΔzapB (35%, n = 161) cells compared with that under the slow growth condition. We were unable to resolve these abnormal FtsZ structures via PALM imaging due to the maturation defects of mEos2 at 37°C (McKinney et al., 2009).

To further investigate whether the abnormal localizations of FtsZ in ΔmatP cells observed under the fast growth condition were caused by the altered spatial distribution of SlmA, we constructed an mEos2–SlmA fusion protein. When this fusion protein is expressed in a ΔslmA strain, we found that it formed a punctated, nucleoid-dependent localization pattern (Fig. 7C), similar to what was observed previously for a fully functional GFP–SlmA construct (Bernhardt and de Boer, 2005). We then used this construct to determine the SlmA distribution pattern in ΔmatP, ΔzapA and ΔzapB cells under both growth conditions. We found that under our normal slow growth condition, the localization patterns observed for mEos2–SlmA in ΔmatP cells were indistinguishable from ΔslmA (Fig. 7C). In ΔzapA and ΔzapB cells, we found that although there were often multiple fluorescent puncta per cell, the segregated, nucleoid-dependent localization pattern of SlmA was conserved (Fig. 7C). The observation of multiple puncta in ΔzapA and ΔzapB cells is consistent with a cytokinetically defective mutant possessing multiple chromosomes. PALM imaging confirmed that under the slow growth condition, the localization pattern of SlmA was generally similar across the four strains (Supplemental Fig. S17A–D). Lastly, when we imaged FtsZ-mEos2 in two double-deletion strains, ΔslmAΔzapA and ΔslmAΔzapB, under the same slow growth condition, we found that the percentage of cells (> 30%) displaying abnormal FtsZ structures and the appearance of dispersed FtsZ clusters under PALM imaging remained similar to that in the single-deletion strains, ΔzapA and ΔzapB (Supplemental Fig. S18A–C). These observations further suggest that under slow growth, SlmA does not mislocalize and does not contribute significantly to the FtsZ phenotype in ΔzapA and ΔzapB cells.

Under a fast growth condition, however, we found that ∼ 30% of ΔmatP (n = 122) cells displayed a significantly more diffusive SlmA localization pattern (Fig. 7D). In particular, elongated cells of ΔmatP typically displayed a more homogenous, non-segregated SlmA distribution, likely resulting from non-segregated or disorganized nucleoids. This is in contrast to what was observed in elongated ΔzapA and ΔzapB cells grown under the same fast growth condition – SlmA still forms segregated puncta (Fig. 7D). These results suggest that the mislocalization of FtsZ in ΔmatP cells under the fast growth condition is likely due to abnormal distributions of SlmA on incorrectly segregated or organized nucleoids.

ΔzapA or ΔzapB does not affect the arrival time of downstream division proteins

Previous studies have shown that ZapA and ZapB localize to the midcell in an FtsZ-dependent manner early in cell division (Gueiros-Filho and Losick, 2002; Aarsman et al., 2005; Goehring et al., 2005; Ebersbach et al., 2008) and that ZapA directly interacts with a number of other essential division proteins, including FtsA, FtsI and FtsN (Di Lallo et al., 2003; Maggi et al., 2008; Alexeeva et al., 2010). As the assembly of the divisome in E. coli follows a largely linear order, in which the arrival of most division proteins is dependent on that of an earlier one (Errington et al., 2003; Aarsman et al., 2005; Gamba et al., 2009), we investigated whether the arrival times of the essential division proteins are altered in the absence of ZapA or ZapB. Furthermore, since ΔzapA and ΔzapB cells display a more dynamic FtsZ structure and more dispersed FtsZ clusters, we were interested to find out whether the formation of these abnormal structures affected downstream protein recruitment.

To examine the arrival time of division proteins in the absence of ZapA and ZapB, we used a previously developed assay (Aarsman et al., 2005), in which we scored and compared the percentages of cells showing midcell localization of each division protein in a steady-state population undergoing exponential growth. Proteins that arrive to the divisome early will be localized to the midcell in a larger percentage of cells than those that arrive late. We fused a fast-maturing yellow fluorescent protein, Venus (V) (Nagai et al., 2002), to the N- or C- terminus of 10 essential division proteins, to generate FtsZ-V, V-FtsA, V-ZipA, V-FtsK, V-FtsQ, FtsL-V, V-FtsB, FtsW-V, V-FtsI and V-FtsN. We expressed these fusion proteins at extremely low levels from the lac promoter on a miniF plasmid without an inducer. Cells expressing these fusion proteins showed essentially the same division rate and cell length as wt cells under identical growth conditions (Supplemental Table S2). All fusion proteins localized correctly to the midcell in a wt background (Supplemental Fig. S19). In addition, V-FtsA, V-FtsK, V-FtsQ, FtsW-V, V-FtsI and V-FtsN were capable of rescuing the lethal phenotype of the corresponding conditional mutants under non-permissive growth conditions (Supplemental Fig. S20, Supplemental Table S3).

We transformed each of the 10 plasmids carrying fusion proteins into BW25113, ΔzapA and ΔzapB strains and imaged these cells on a microscope stage. We found that in ΔzapA and ΔzapB cells, the localization patterns of several division proteins became difficult to score, likely due to the altered localization pattern of FtsZ in the deletion strains, and extremely low expression levels of these fusion proteins without induction. Therefore, we only scored the midcell localization for FtsZ-V, ZipA-V, V-FtsK and V-FtsI. We used the fraction of cells displaying midcell localization to calculate the arrival time of each division protein relative to the cell cycle of each strain (Supplemental Fig. S21). We found that the localization pattern of the early division protein, ZipA, also formed broad or diffusive bands at the midcell, whereas the late division proteins FtsK and FtsI largely formed single, sharp bands at the midcell. Nevertheless, the arrival time of the four division proteins did not vary significantly between the three strains (P > 0.25). Therefore, we conclude that under our experimental conditions, the arrival times of these essential division proteins do not appear to depend on ZapA or ZapB and that the recruitment of these proteins does not appear to depend on the formation of a coherent Z-ring.


We previously showed that in E. coli the Z-ring is likely composed of a loose bundle of FtsZ protofilaments that randomly overlap with each other in space (Fu et al., 2010). We proposed that lateral interactions between FtsZ protofilaments, mediated either by the intrinsic lateral affinity of protofilaments or by other protein factors that bind FtsZ, are required to organize such a loose bundle. In this work, we investigated the role of ZapA and ZapB in promoting the assembly of the Z-ring.

Deletion of zapA or zapB affects septum morphology and FtsZ dynamics

In cells deleted of zapA or zapB, we observed a minor cell division defect similar to what was reported previously: on average ΔzapA and ΔzapB cells are about 20% longer than wt cells and grow with a similar rate as wt cells in liquid media (Gueiros-Filho and Losick, 2002; Ebersbach et al., 2008). However, when observed under SEM, we found that many ΔzapA or ΔzapB cells showed abnormal septa that were not precisely aligned at the midcell or parallel to the cell's short axis (Fig. 1). Additionally about 20% of cells showed more than one septum around the midcell region. These observations argue that ZapA and ZapB have larger roles than previously thought in cell division, and that they may facilitate the correct positioning of cell division plane, possibly by organizing the structure and localization of the Z-ring (see below).

When investigated using time-lapse fluorescence imaging, we observed dynamic FtsZ structures moving back and forth between multiple sites in ΔzapA and ΔzapB cells, often resulting in visible cell wall indentations at these sites (Fig. 2B). These multiple constriction sites likely correspond to the multi-septa observed in ΔzapA and ΔzapB cells by SEM. At later stages of cell division FtsZ structures coalesced to a single site to achieve complete septation. As a result, the corresponding daughter cells contained visible constriction sites that were previously occupied by FtsZ in their parent cell. Interestingly, in a subpopulation of these daughter cells FtsZ relocalized to the partially constricted sites and completed cell division at a significantly faster rate than the average population (Fig. 2C).

The ability of FtsZ to relocalize to aborted constriction sites in ΔzapA and ΔzapB cells conflicts with a previous study performed with a temperature-sensitive FtsZ GTPase mutant, ftsZ84 (Addinall et al., 1997). Addinall et al. showed that FtsZ84 does not return to aborted division sites that had initiated constriction, and proposed that the localization signal for FtsZ is lost once cell wall constriction begins (Addinall et al., 1997). As recent studies have shown that the localization of FtsZ to midcell is the result of the combined mechanisms of nucleoid occlusion, MinCDE antagonism and membrane properties, a specific localization signal may not be involved (Bernhardt and de Boer, 2005; Mileykovskaya and Dowhan, 2005; Lutkenhaus, 2007). Therefore, we suspect that the conflicting observations may be due to the inefficient polymerization of the FtsZ84 mutant (Bramhill and Thompson, 1994).

The question of why reinitiated constriction sites complete septation faster than the general population is intriguing. One possible explanation is that some division proteins, such as peptidoglycan modification enzymes, are not completely disassembled upon the departure of FtsZ from these sites, hence when FtsZ returns, the reassembly of the divisome and reinitiation of cell wall constriction occurs more rapidly. While many genetic studies have shown that the midcell recruitment of all other division proteins are dependent on the presence of FtsZ (Addinall et al., 1997; Buddelmeijer and Beckwith, 2002; Errington et al., 2003), it has not been systematically shown that all the division proteins leave immediately upon the disassembly of the Z-ring. In fact, a few studies observed that at the end of septation, FtsK and FtsN persist at the midcell even after FtsZ leaves (Wang et al., 2005; Möll et al., 2010). However, it is not known how long these late division proteins persist after the departure of FtsZ. If they persist at aborted septa in ΔzapA and ΔzapB cells until the beginning of the next cell cycle, we would expect their calculated arrival times to be shorter in the deletion strains than in wt cells. In our own study (Supplemental Fig. S21), we did not observe significant changes in the arrival time of the division proteins we investigated (ZipA, FtsK and FtsI), possibly due to our limited temporal resolution.

Enhanced global dynamics of FtsZ in ΔzapA and ΔzapB cells are likely due to weakened structural stability

We observed more dynamic, global movement of FtsZ structures in the absence of ZapA and ZapB than in wt cells (Fig. 2). However, we showed through single molecule tracking experiments that this apparent increase in global movement is not due to the enhanced mobility of individual FtsZ molecules. FtsZ molecules are known to turn over rapidly inside the Z-ring on the timescale of tens of seconds (Stricker et al., 2002), resulting in a constant flux of molecules dissociating from and reassociating into the ring. This turnover process may drive the global movement of FtsZ structures and is consistent with the observed density fluctuations within the Z-ring recently reported for wt B. subtilis cells (Strauss et al., 2012). As FRAP experiments have shown that the turnover rate of FtsZ molecules in Bacillus subtilis is not affected by the absence of ZapA (Anderson et al., 2004), we reason that the global dynamics of FtsZ may only appear to be enhanced in ΔzapA and ΔzapB cells because of the increased ability of FtsZ molecules to reassociate at positions farther away from midcell. This could be caused by the removal of some structural constraints imposed by ZapA and ZapB (see below). Consistent with this suggestion we observed using PALM imaging that FtsZ clusters in ΔzapA and ΔzapB cells were more widely dispersed than in wt cells (Fig. 4, Supplemental Figs S8–S10).

ZapA and ZapB likely function together in organizing Z-ring structure

In vitro biochemical studies have shown that ZapA is capable of promoting FtsZ bundling by itself (Gueiros-Filho and Losick, 2002; Small et al., 2007; Mohammadi et al., 2009; Dajkovic et al., 2010; Galli and Gerdes, 2012). Other studies on ZapB have shown that its ability to promote Z-ring assembly is entirely dependent on ZapA, as it directly binds to ZapA but not FtsZ (Ebersbach et al., 2008; Galli and Gerdes, 2012). Therefore, we expected that in vivo ZapA may be able to function alone in the absence of ZapB and that deletion of zapB would have a milder effect than deletion of zapA. Although statistically different, our results show that deletion of zapB results in a similar adverse phenotype as the deletion of zapA. Both deletion strains displayed increased cell length, abnormal septum morphology, heterogeneous cell cycle time, highly dynamic and abortive Z-ring assembly, and significantly smaller FtsZ clusters compared with wt. These results suggest that ZapA and ZapB's function in promoting Z-ring assembly in vivo may be dependent on each other. Interestingly, although the phenotypes of ΔzapA and ΔzapB were more similar to each other than to wt, we did notice that the effect of ΔzapB on FtsZ clusters was milder than ΔzapA: FtsZ clusters in ΔzapB were bigger, less dispersed and contained more molecules than those in ΔzapA. These observations are consistent with the expectation that ZapA does retain some ability to promote FtsZ assembly in the absence of ZapB, but the presence of ZapB greatly enhances this ability.

Influence of MatP on FtsZ structure

ZapB was initially identified as a plasmid-partitioning factor (Ebersbach et al., 2008) and has recently been implicated in chromosome segregation through its interaction with MatP (Espeli et al., 2012). The loss of ZapB–MatP interaction in ΔzapA and ΔzapB cells could impact the segregation of nucleoid, and hence the localization of the nucleoid occlusion factor, SlmA. We investigated FtsZ and SlmA localization in ΔmatP, ΔzapA and ΔzapB cells. We found that FtsZ and SlmA localizations are only perturbed in ΔmatP cells under a fast growth condition. In contrast, FtsZ and SlmA localization in ΔzapA and ΔzapB cells are similar in both fast and slow growth conditions.

Our results suggest that the abnormal FtsZ localization in ΔmatP cells under the fast growth is likely caused by a nucleoid segregation defect that alters the SlmA distribution, whereas the dispersed FtsZ clusters we observed in ΔzapA and ΔzapB cells are not due to nucleoid segregation defects. These observations are also consistent with previous studies that deletion of ZapB has no effect on nucleoid segregation (Ebersbach et al., 2008) or the mobility of the chromosome outside of the Ter region (Thiel et al., 2012). Note that the loss of ZapB–MatP interaction in ΔzapA or ΔzapB cells is not equivalent to the loss of MatP function in ΔmatP cells. The severely affected FtsZ localization in ΔmatP under fast growth condition but essentially the same FtsZ localization in ΔzapA and ΔzapB under both growth conditions supports a minimal role of ZapB–MatP interaction in organizing the Z-ring, but a significant role of MatP in organizing the nucleoid.

It is known that under fast growth conditions where cell doubling time is close to or shorter than that required for a complete round of chromosome replication (∼ 40 min), bacterial cells usually initiate multiple rounds of chromosome replication before the previous round is finished. As a result, multiple, partially replicated chromosomes present a significant topological challenge for correct and efficient nucleoid segregation. Therefore, it is reasonable to expect that under this condition the coordination between cytokinesis, initiated by the Z-ring assembly at the midcell, and nucleoid segregation, in which MatP has a significant role, is more tightly coupled to ensure proper cell division. Hence, it is not surprising that the loss of MatP function in ΔmatP cells leads to significantly altered FtsZ localization.

FtsZ clusters formed in the absence of ZapA or ZapB contain higher-ordered FtsZ polymers

As described above, we found that FtsZ formed small, widely dispersed clusters around the midcell in the absence of ZapA or ZapB. We showed that these FtsZ clusters are not due to aggregation of the fluorescent proteins, but are specific to the deletion of zapA or zapB and the polymerization of FtsZ.

To gain further molecular insight, we compared these clusters’ composition to that expected for a single FtsZ protofilament, the presumed basic structural unit of the Z-ring (Erickson et al., 1996b; Chen and Erickson, 2005; Li et al., 2007). The average numbers of FtsZ-mEos2 molecules in ΔzapA and ΔzapB clusters were ∼ 50 and ∼ 80, respectively, while that in the wt clusters was ∼ 150 (Table 2). Using individual cellular fluorescence and population-based immunoblotting measurements we showed that on average FtsZ-mEos2 was expressed at ∼ 30% of total cellular FtsZ levels (Supplemental Fig. S4). Therefore, FtsZ clusters in ΔzapA, ΔzapB and wt cells would contain ∼ 160, ∼ 250 and ∼ 500 FtsZ molecules respectively. We note that since not all FtsZ-mEos2 molecules were detected within the limited imaging time, these estimates represent lower bounds. These underestimates are still significantly larger than the estimated ∼ 30 monomers that comprise single FtsZ protofilaments (Romberg et al., 2001; Chen et al., 2005; Huecas et al., 2008). This calculation suggests that FtsZ clusters in all three strains are much larger than a typical single FtsZ protofilament observed in vitro.

Consistent with this suggestion, we found that the measured areas of these clusters were also much larger than that expected for a single protofilament. Given our spatial resolution of imaging (∼ 50 nm, Supplemental Fig. S5), a single FtsZ protofilament based on in vitro EM measurement (120 nm × 5 nm) (Erickson et al., 1996a) would occupy an area of 6500 nm2 (130 nm × 50 nm). In contrast, the average measured area of FtsZ-mEos2 clusters in ΔzapA or ΔzapB cells was ∼ 30 000 nm2, and in wt cells was ∼ 70 000 nm2. These values are more than 5-fold and 10-fold larger than the expected value of a single protofilament respectively. The dimensions of these clusters (∼ 300 nm × 150 nm) were also significantly larger than the respective dimensions of a single protofilament, consistent with the previous observation that these clusters contain more molecules than single protofilaments. Based on the dimensions observed, it is unlikely that FtsZ molecules in these clusters are arranged into a single, long FtsZ filament, as such a long filamentous configuration would give rise to a much larger cluster length (major axis length) and a much smaller cluster width (minor axis length). It is also unlikely that FtsZ protofilaments in these clusters adopt a tightly aligned ribbon configuration, as the cluster width would be significantly smaller than what we observed. Based on these estimations, we suggest that FtsZ clusters formed in wt cells and in the absence of ZapA and ZapB are loosely associated FtsZ polymers that likely consist of multiple FtsZ protofilaments. More work is needed to determine what other factors are required to maintain these loose associations of higher-ordered FtsZ polymers.

ZapA and ZapB mainly function at the level of organizing higher-ordered FtsZ polymers

We observed that the ring-like FtsZ structures formed in the ΔzapA and ΔzapB cells are dimensionally indistinguishable from those observed in wt cells (Fig. 3). This observation is consistent with the fact that a number of other factors, besides ZapA and ZapB, have been shown to promote Z-ring assembly. FtsZ by itself has been shown to form ring-like structures on liposomes in vitro or in a non-native cellular environment (Osawa et al., 2008; Srinivasan et al., 2008), suggesting that intrinsic lateral interactions between protofilaments alone can mediate the formation of large FtsZ assemblies. Additionally, at least two other Zap proteins, ZapC and ZapD, have been identified that function similar to ZapA and ZapB in promoting the association of FtsZ protofilaments in vitro and Z-ring assembly in vivo (Durand-Heredia et al., 2011; 2012; Hale et al., 2011).

Given our observation of higher-ordered FtsZ polymers in ΔzapA and ΔzapB cells, it is plausible that ZapA and ZapB do not function at the level of promoting the association of single protofilaments. However, given the redundant mechanisms described above we cannot formally exclude this possibility. Nevertheless, the most significant defect that we observed on Z-ring assembly in ΔzapA and ΔzapB cells is that FtsZ forms clusters that are smaller, less dense, and more dispersed away from the midcell region. This suggests that although ZapA and ZapB may promote the association of single protofilaments as postulated in previous in vitro studies (Dajkovic et al., 2010), their major contribution in vivo is to facilitate the assembly of these higher-ordered FtsZ polymers into the Z-ring.

To understand how ZapA and ZapB might facilitate the assembly of FtsZ clusters into the Z-ring, we further analysed the shape and orientation of FtsZ clusters. In all strains we found that FtsZ clusters were elongated and generally oriented along the short axis of the cell (Table 2). The length (major axis) of FtsZ clusters in ΔzapA and ΔzapB cells was ∼ 50% shorter than wt clusters, whereas their width (minor axis) was only reduced by ∼ 20%. Considering the preferential alignment of FtsZ clusters, the larger differences in cluster lengths between strains suggests that ZapA and ZapB may have a greater affect on the growth of FtsZ polymers along the short axis of the cell.

Previously, we and other groups have reported that the width of the Z-ring (∼ 110 nm) is largely invariant between different bacterial species and FtsZ expression levels (Fu et al., 2010; Jennings et al., 2011; Biteen et al., 2012). Consistent with these reports, we observed tight, largely symmetric width distributions for clusters in all three strains (Fig. 5G).These observations suggest that the polymerization of FtsZ along the long axis of the cell is tightly regulated by factors other than ZapA and ZapB. Whether this regulation is mediated by negative factors such as SlmA and MinC, which impose physical constraints on FtsZ polymerization, or by positive factors that regulate the lateral associations of FtsZ remains to be seen.

ZapA and ZapB promote Z-ring assembly by aligning FtsZ clusters

Assembly of the Z-ring has been proposed to consist of multiple steps (Margolin, 2005; Adams and Errington, 2009). In the first step, FtsZ monomers polymerize to form single-stranded protofilaments. Next, these single-stranded protofilaments can further grow in length and also associate with others laterally to assemble the Z-ring. In this study, we observed that in the absence of ZapA and ZapB, FtsZ formed small, widely dispersed clusters. These clusters have larger sizes and more molecule counts than what would be expected from a single FtsZ protofilament, suggesting that they are higher-ordered structures composed of multiple FtsZ protofilaments.

Based on these observations, we propose that these clusters, or multi-stranded FtsZ polymers, represent a distinct polymeric species of FtsZ that is inherent to the assembly pathway of the Z-ring (Fig. 8). Furthermore, we propose that the main function of ZapA and ZapB in vivo is to bring these dynamic FtsZ clusters into close proximity at midcell and this coherent alignment of FtsZ clusters enables proper Z-ring assembly (Fig. 8). In the absence of ZapA or ZapB, FtsZ clusters easily break off from each other, leading to dispersed localizations and consequently abnormal septa. This model is consistent with our observation of larger clusters in wt cells, as well as a recent study that used three-dimensional structured illumination to show that the Z-ring in B. subtilis adopts a non-uniform, bead-like configuration (Strauss et al., 2012). It is likely that these beads are analogous to the FtsZ clusters we observed and the in vivo function of ZapA and ZapB is to corral these clusters at midcell. Future characterization of the mechanisms by which these FtsZ clusters are organized will be pivotal to our understanding of Z-ring assembly and function.

Figure 8.

Model of how ZapA and ZapB promote Z-ring assembly. 1. FtsZ monomers associate longitudinally into FtsZ protofilaments. 2. FtsZ protofilaments laterally associate to form FtsZ clusters, or higher-ordered polymers consisting of multiple FtsZ protofilaments. This process may be mediated by the intrinsic properties of FtsZ and/or other protein factors. ZapA and ZapB may participate but are not required for this step. 3. FtsZ clusters are corralled at midcell through the combined function of ZapA and ZapB, resulting in larger continuous structures. 4. In the absence of ZapA and ZapB, FtsZ clusters scatter throughout the midcell region and some unknown mechanism (dashed arrow) is responsible for proper Z-ring assembly.

Experimental procedures

Bacterial strains, growth conditions and materials

Bacterial strains and plasmids are indicated in Supplemental Table S4. Primer sequences are listed in Supplemental Tables S5 and S6. Cells were inoculated into LB media and grown overnight at 37°C, then diluted in M9 minimal media supplemented with 0.4% Glucose, MEM Vitamins and MEM Amino Acids (M9+), and grown at room temperature (RT) for at least 20 h prior to fixation or imaging. When appropriate, 150 μg ml−1 chloramphenicol, 50 μg ml−1 kanamycin or 50 μg ml−1 carbenicillin was added. For all PALM studies, FtsZ-mEos2 expression was induced with 20 μM IPTG for 2 h followed by a washing step and an additional 2–3 h outgrowth at RT. For fast growth conditions, we applied EZ Rich Defined Media (Teknova) supplemented with 0.4% Glucose and incubated at 37°C. All restriction enzymes were from New England Biolabs. PfuUltra II polymerase (Agilent) was used for PCR amplification.

Plasmid construction

The N-terminal His-tag and corresponding linker of FtsZ-mEos2 in pJB004 (Fu et al., 2010) was removed by inverse PCR to generate pJB042. The forward primer #1 contained a SpeI site followed by the 5′ end of ftsZ. The reverse primer #2 was designed complementary to the sequence immediately upstream of the His-tag start codon and also included a novel SpeI site. After PCR amplification, the reaction mixture was treated with DpnI, ligated and transformed into DH5α (Invitrogen), generating pCH027. FtsZ-mEos2 was amplified from pET28-FtsZ-mEos2 with primers 5 and 7, restricted with SpeI and NotI, and ligated into a similarly digested pCH027, generating pJB042. FtsZ-mEos2 from pJB042 contains the GSAGSAAGSGEF linker. Basal expression of FtsZ-mEos2 from pJB042 is significantly reduced relative to pJB004. This phenomenon is likely a result of the incorporation of the novel SpeI site in the region separating the FtsZ-mEos2 start codon from the ribosome binding site.

pJB108 (PT5-lac::FtsZ-Dronpa) was constructed by amplifying the dronpa sequence from pDG1-S1 (Amalgaam) using primers #3 and #4. The fragment was then digested with NotI and ligated into a similarly digested pCA24N-FtsZ, resulting in pJB072 (PT5-lac::6xHis-FtsZ-Dronpa). The ftsZ-dronpa sequence was then amplified with primers #5 and #6, restricted with SpeI and SalI, and inserted into a similarly digested pJB042. The amino acid sequence linking FtsZ to Dronpa is GLCGR.

pJB106 (PT5-lac::FtsZ-mEos3) was constructed by amplifying ftsZ-mEos2 from pJB042 using primers 5 and 7. The sequence was then restricted with SpeI and HindIII, and the larger fragment, corresponding to FtsZ and an N-terminal portion of mEos2, was purified. mEos3 was amplified from pTriEX-HM-mEos3 (a gift from J. Yu) using primers #8 and #9, and digested with HindIII and SalI. The 400 bp fragment was combined with the ftsZ-mEos2 fragment and inserted into a pJB042 digested with SpeI and SalI. The linker sequence is GSAGSAAGSGEF.

pJB044 was constructed by amplifying mEos2 from pJB042 using primers #10 and #7 and inserted into the backbone of pJB042, similarly digested with SpeI and NotI.

pJB095 (PBAD::sulA) was constructed in multiple steps. First, the λPR promoter from pDR175 was replaced with PBAD from pBAD TOPO® (Invitrogen) using a cut-and-paste scheme involving primers #11 and #12, and restriction with EcoRI and NheI, generating pTH323. The sulA gene was amplified from purified E. coli K12 genomic DNA using primers #13 and #14 and joined via NheI and SalI sites. pJB095 was transformed into BW27783, a BW25113 derivative that homogenously expresses from PBAD by decoupling the arabinose transporters, AraF and AraE, from induction (Khlebnikov et al., 2001).

pJB056 (PBAD:zapA) was constructed by amplification of zapA from pJW2878 (Kitagawa et al., 2005) using primers #15 and #16, restriction with NheI and SalI, and ligation with a similarly digested pTH323. pJB065 (PBAD:zapB) was constructed in the same fashion, using primers #17 and #18 to amplify zapB from pJW3899 (Kitagawa et al., 2005).

pJB139 (PT5-lac::mEos2–SlmA) was constructed by amplifying mEos2 and slmA from pJB042 and genomic K12 DNA, using primer pairs #10–19 and #20–21 respectively. The fragments were digested with SpeI and NotI, gel-purified, and ligated with a similarly digested pJB042. The construct was confirmed by sequencing. The linker sequence is ASAGSAAGSSGR.

pJB144 (PT5-lac::GFP–SlmA) was constructed by amplifying GFP from JW0093 (ASKA +) and using primer pairs #22–23. The fragment was digested with SpeI and NheI, gel-purified, and ligated with a similarly digested pJB139. The construct was confirmed by sequencing.

Deletion strain construction

The double-deletion strains, ΔslmAΔzapA and ΔslmAΔzapB, were constructed from slmA::kan using lambda red technology (Datsenko and Wanner, 2000). First, the kanamycin resistance of JW5641 was removed, resulting in slmA::frt. Next, the kanamycin cassette of pKD4 was amplified with the primer sequences complimentary to the termini of zapA or zapB (Baba et al., 2006). The slmA::frt strain, containing pKD46, was then transformed with the appropriate gel-purified fragments. Sequencing and sensitivity to carbenicillin confirmed insertion and pKD46 removal respectively.

Quantitative immunoblotting

FtsZ purification and Western analysis was performed as described previously (Fu et al., 2010) with the following exceptions. M9+ cultures of BW25113, ΔzapA and ΔzapB cells were grown at room temperature to mid-log phase (0D600 ≈ 0.4), counted and then applied to SDS-PAGE. After semi-dry transfer and blocking, primary incubation was performed for 1 h at RT with an affinity-purified anti-FtsZ rabbit antibody (a gift from H. Erickson) diluted 1:3000 in TBS containing 0.05% Tween-20 and 1% BSA. Blots were washed three times with TBS + Tween for 10 min each. Secondary incubation was with Goat anti-rabbit HRP (Bio-Rad) diluted 1:45 000 in TBS + Tween-20. After washing, blots were developed using Immun-Star™ WesternC™ (Bio-Rad) and imaged on a Typhoon Scanner (GE LifeSciences). Band intensities were determined using ImageJ.

Scanning electron microscopy

BW25113, ΔzapA and ΔzapB cells were grown up to mid-log phase in M9+ at RT, concentrated by centrifugation, washed with M9+ and then either applied directly to poly-l-lysine (PLL)-treated 18 mm coverslips (0.01% PLL for 10 min, 5 min dH2O wash) or passed through 13 mm 0.2 μm Polypropylene Membrane Filters (Sterlitech). Coverslips and filters were then submerged in fixative [2% Formaldehyde, 2% Glutaraldehyde, 0.08 M Sorenson's Phosphate Buffer (SPB), 3 mM MgCl2, pH = 7.2] for at least 1 h at RT and then washed with PBS. Samples were then post-fixed in 0.8% potassium ferrocyanide reduced 1% OsO4 in 0.0 8 M SPB and 3 mM MgCl2 for 1 h on ice in the dark. Following a rinse in 0.08 M Maleate Buffer (3 × 5 min), samples were placed in 2% Uranyl Acetate in Maleate for 1 h at RT in the dark. En-bloc staining was followed by dehydration through a graded series of ethanol to 100% ethanol, then passed through a 1:1 solution of 100% ethanol and hexamethyldisiloxane (HMDS), and finally pure HMDS. Coverslips were then placed in a desiccator overnight to dry. The next day, coverslips were attached to aluminium stubs via carbon sticky tabs (Ted Pella), and coated with 20 nm AuPd via a Denton Vacuum Desk II sputter coater. Stubs were viewed and digital images captured on a Leo 1540 FESEM operating at 1 kV.

Fluorescence microscopy

All fluorescence imaging was performed as described previously (Fu et al., 2010; Buss et al., 2013) with the following exceptions. The green and red fluorescence of mEos2 were excited with a solid-state 488 nm and 561 nm laser (Sapphire™, Coherent, Santa Clara, CA) respectively. Green fluorescence images were taken prior to PALM imaging to provide a comparable ensemble fluorescence image and to calibrate for the expression level of individual cells. These images were captured using the same acquisition settings and power density (∼ 500 W cm−2) as before. Live-cell PALM imaging was achieved over the course of 30 s using a frame rate of 100 s−1, a 561 nm excitation of ∼ 1 kW cm−2 and continuous 405 nm activation at ∼ 5 W cm−2. A resolution of ∼ 50 nm (Supplemental Fig. S5) was achieved. Fiducial markers were not used for live-cell PALM imaging, as sample drift during the 30 s of acquisition was undetectable. Construction and analysis of PALM images was performed as described previously (Fu et al., 2010). Briefly, an average of the maximum pixel intensities for a running widow of 150 frames was used to threshold the corresponding images. Pixels above threshold were fitted to a two-dimensional Gaussian function and spots with localization precision better than 50 nm were plotted. Images are rendered in pseudo-colour (‘Red Hot’) via ImageJ. Where indicated, TIR-PALM was performed by translating the beam away from the central axis. Pixel size in PALM images is 15 nm.

Time-lapse imaging of FtsZ–GFP from JW0093 (ASKA +) (Kitagawa et al., 2005) was performed on live cells grown at RT on a 3% agarose gel pad made with M9+ lacking MEM vitamins. Imaging was performed with an integration time of 50 ms. To reduce photobleaching effects, 488 nm excitation power was decreased to ∼ 10 W cm−2. Frames were acquired every 5 min. Data were acquired on multiple days for each sample. Division times were determined empirically by visual inspection of cell morphology and FtsZ–GFP signal. Only actively dividing cells were assayed, as division times were determined by the amount of time separating two constriction events (ex. Fig. 2Biii 50′ and 130′).

Single molecule tracking (SMT) was performed on BW25113, ΔzapA and ΔzapB cells expressing FtsZ-mEos2 from pJB042 in the absence of induction. All SMT imaging employed a 50 ms integration time and ∼ 350 W cm−2 561 nm excitation power density. A frame rate of 5 s−1 was achieved through the insertion of 150 ms dark intervals via an acousto-optic modulator. The applied 405 nm activation power was ∼ 50 mW cm−2. Imaging of individual regions never exceeded 5 min. Data were analysed via the ImageJ plugin Octane (Niu and Yu, 2008). All traces were inspected by eye and any trace suspected of containing multiple overlapping emitters was discarded.

Snap-shots of BW25113, ΔzapA and ΔzapB cells expressing the various Venus fusions were acquired on an inverted microscope (IX-81, Olympus) equipped with a 100× oil-immersion objective lens (Olympus) and captured with a cooled EMCCD camera (Andor Ixon DU888). Excitation was provided by the 514 nm line of an Innova Ion I-308 laser (Coherent) at a power density of ∼ 1 kW cm−2. The fraction of population, F(x), displaying midcell fluorescence was determined by eye and used to calculate the arrival time (tarr) relative to the cell cycle of each strain, using:

display math

Three independent experiments were performed for each fusion in each strain.

Cluster determination and analysis

Analysis of TIR-PALM images was performed using custom MATLAB® software. Spot identification and plotting was carried out as described above. For all cells, a single empirically determined threshold (0.00045) was applied as a percentage of the total integrated signal of the Gaussian-plotted PALM image (Supplemental Fig. S13). Groups of more than five adjacent pixels above the threshold were identified as clusters. The coordinates of the cluster were used to identify the localizations within those regions. Number of molecules per cluster was calculated from the number of localizations using the conversion factor determined in Supplemental Fig. S22. The regionprops function of MATLAB® was used to identify area, centroid position, major axis length, minor axis length and orientation of each cluster (Fig. 5A). The cellular pole positions were determined by eye and used to calculate cell length, cellular axes and midcell plane. Cluster orientations are reported as absolute values relative to the midcell plane, where 0° indicates a parallel alignment. The significant deviation away from 0° observed for wt structures is likely due to the error associated with cell pole selection as well as the irregularity in cluster shape. Midcell position, cellular orientation and cluster centroid position were used to calculate the displacement of clusters away from the midcell plane.


We thank M. Delannoy (Johns Hopkins School of Medicine Microscope Facility) for technical assistance, and E. Goley, P. Levin and H. Erickson for critical reading of the manuscript. We also thank NBRP-E.coli at NIG, the E. coli Genetic Stock Center at Yale University, J. Lutkenhaus, J. Keasling, D. Weiss, J. Beckwith and P. de Boer for strains, L. Looger and J. Yu for mEos2 and mEos3 respectively. This study was supported by grants from the National Institute of Health 1R01GM086447-01A2 and NSF EAGER MCB1019000.