Using a bacterial two-hybrid system and a combination of in vivo and in vitro assays that take advantage of the green fluorescent reporter protein (GFP), we have investigated the localization and the protein–protein interaction of several key components of the cytokinetic machinery of cyanobacteria (i.e. the progenitor of chloroplast). We demonstrate that (i) the ftsZ and zipN genes are essential for the viability of the model cyanobacterium Synechocystis sp. PCC 6803, whereas the minCDE cluster is dispensable for cell growth; (ii) the GTP-binding domain of FtsZ is crucial to FtsZ assembly into the septal ring at mid-cell; (iii) the Z-ring of deeply constricted daughter cells is oriented perpendicularly to the mother Z-ring, showing that Synechocystis divides in alternating perpendicular planes; (iv) the MinCDE system affects the morphology of the cell, as well as the position and the shape of FtsZ structures; and (v) MinD is targeted to cell membranes in a process involving its C-terminal amphipathic helix, but not its ATP-binding region. Finally, we have also characterized a novel Z-interacting protein, ZipN, the N-terminal DnaJ domain of which is critical to the decoration of the Z-ring, and we report that this process is independent of MinCDE.
In the past decade, bacteria have no longer been regarded as being too simple organisms to require a spatial regulation and a cytoskeleton (Young, 2003). Bacterial cells, such as those of Escherichia coli, normally divide by binary fission, which occurs at mid-cell after the DNA has been duplicated and segregated into daughter nucleoïds, yielding two equal-sized daughter cells. Cell division is initiated through the polymerization of the FtsZ protein that is homologous, in both sequence and three-dimensional structure, to the cytoskeletal protein tubulin of the eukaryotes (Lowe and Amos, 1998) and occurs in archaeal cells as well (Wang and Lutkenhaus, 1996). In the rod-shaped bacterium E. coli, the GTPase FtsZ protein assembles into a Z-ring that encircles the dividing cell at the division plane (Bi and Lutkenhaus, 1991). The site for FtsZ ring formation is restricted to mid-cell by the MinCDE system (Margolin, 2001; Rothfield et al., 2001; Lutkenhaus, 2002; Errington et al., 2003). Abrogation of the MinCDE system causes a large fraction of cells to divide near the poles instead of the middle, creating one viable bacterium with a double genome content and one minicell that is ‘devoid’ of DNA. The MinCD complex is a binary inhibitor of cell division that oscillates from one end of the cell to the other, thereby blocking the formation of the Z-ring at the cell poles. MinCD oscillation is dependent on MinE that form an off-centre oscillating ring preventing the action of MinCD at mid-cell, thereby allowing proper localization of the Z-ring (Hale et al., 2001). Formation of the Z-ring initiates at a single nucleation site and expands bidirectionally around the circumference of the cell. Then, the Z-ring directs division by recruiting other proteins that participate in septum formation, a process involving co-ordinated inward growth of the cytoplasmic membrane, the rigid murein (peptidoglycan) layer and the outer membrane of the cell envelope.
However, the E. coli model for division site placement in rod-shaped bacteria is not appropriate for spherical cells, such as those of a variety of cyanobacteria that do not have a defined middle, but instead possess an infinite number of potential division planes at the point of greater cell diameter. Furthermore, it is interesting to investigate cell division in cyanobacteria because they are regarded as the ancestor of the chloroplast (Gray, 1993). Consistently, chloroplast division was found to require several prokaryotically derived proteins, including FtsZ1, FtsZ2, MinD and MinE, as well as ARTEMIS and ARC6, which are unique to plants and cyanobacteria (Osteryoung and McAndrew, 2001; Fulgosi et al., 2002; Lutkenhaus, 2002; Vitha et al., 2003). Little is known concerning cell division in cyanobacteria, even though mutants impaired in cytokinesis were isolated 30 years ago (reviewed by Koksharova and Wolk, 2002). In the filamentous strain Anabaena PCC7120, the ftsZ gene was found to be expressed only in vegetative cells, not in heterocysts (i.e. the non-dividing differentiated cells devoted to nitrogen fixation) (Kuhn et al., 2000). In the rod-shaped unicellular strain Synechococcus PCC7942, overexpression of ftsZ halted division and caused the cells to become filamentous (Mori and Johnson, 2001), as found in E. coli. Recently, a novel gene involved in cytokinesis has been identified (it has been termed ftn2) and found to be dispensable to the survival of Synechococcus PCC7942 (the ftn2 null mutant is filamentous), whereas it is critical to the viability of Anabaena PCC7120 (the heteroploïd, i.e. unsegregated mutant, harbours enlarged cells) (Koksharova and Wolk, 2002). The scarcity of molecular data concerning the cytokinesis of cyanobacteria results from the fact that cytological techniques (e.g. fluorescence microscopy) have not yet been used in these organisms. Therefore, we have used the green fluorescent reporter protein (GFP) for both in vivo and in vitro analysis of the localization and the protein–protein interaction of several key components of the cyanobacterial cytokinetic machinery (FtsZ, MinCDE, Ftn2). As the host, we have used the spherical-celled unicellular strain Synechocystis sp. PCC6803 (hereafter referred to as Synechocystis), which possesses a small fully sequenced genome (Kaneko et al., 1996) (http://www.kazusa.or.jp/cyano/cyano.html) that can easily be manipulated with replicating plasmids (Mazouni et al., 1998; Poncelet et al., 1998). We demonstrate here that FtsZ forms a septal ring, the position and shape of which is controlled by the MinCDE system. The Z-ring interacts with the Ftn2 protein (hereafter referred to as ZipN for Z-ring interacting protein N), a process that depends on the conserved central region of FtsZ and on the DnaJ domain of ZipN.
The ftsZ gene is essential for the viability of Synechocystis, whereas the minCDE cluster is dispensable for cell growth
The genome of Synechocystis (Kaneko et al., 1996; Cyanobase: http://www.kazusa.or.jp/cyano/cyano.html), which occurs at about 10 copies per cell (Labarre et al., 1989), encodes several proteins showing significant homology to the E. coli cell division proteins FtsZ (Sll1633 in cyanobase, 40.1% amino acid identity with the E. coli orthologue), MinC (Sll0288, 19.8% amino acid identity), MinD (Sll0289, 42.9% amino acid identity) and MinE (Ssl0546, 30.6% amino acid identity). These genes were independently cloned in an E. coli plasmid, along with their flanking regions (Table 1) to serve as platforms for homologous recombinations mediating targeted gene replacement in Synechocystis (Labarre et al., 1989). Then, the studied coding sequences were replaced (or interrupted) by the kanamycin resistance marker gene (Kmr), yielding the plasmids listed in Table 1. Kmr clones originating from the transformation with the plasmid pΔZ::Kmr (ftsZ interruption) invariably died upon restreaking on Km plates, showing that ftsZ is crucial to cell survival in cyanobacteria. In contrast, the Kmr transformants obtained with any of the five ‘min’ deletion plasmids (pΔCDE::Kmr, pΔDE::Kmr, pΔC::Kmr, pΔD::Kmr and pΔE::Kmr) remained viable, even when restreaked in the presence of Km, which favoured the amplification of the mutant (e.g. Kmr) copies of the chromosome. Polymerase chain reaction (PCR) analyses of these transformants (data not presented) demonstrated that the Kmr cassette had been inserted properly in their genome, thus replacing the corresponding min gene (or cluster) in every chromosome copy. This showed that the min system is dispensable for cell survival, as occurs in E. coli (Margolin, 2001; Lutkenhaus, 2002; Errington et al., 2003).
Table 1. . Characteristics of the plasmids used in the present study.
Purpose and description
The following abbreviations were used for plasmid designation. Following the usual p for plasmid, the second letter (written in upper case) refers to the host cell: C (E. coli ) or S (Synechocystis). GFP stands for the open reading frame encoding the green fluorescent protein. Δ refers to a deletion of part of the studied genes, such as the N-terminus (ΔNTD) or the C-terminus (ΔCTD) of the protein coding sequence.
Gene inactivation in Synechocystis
AT overhang cloning vector (Apr)
Source of the Kmr marker gene
pGEMT with the 1293 bp ftsZ ORF (Sll1633, see Cyanobase)
pGEMT with the Kmr marker from pUC4K inserted at the codon 107 (replaced by a unique EcoRV restriction site)
pGEMT with the 2628 bp region encompassing the 1972 bp minCDE cluster (sll0288, sll0289 and ssl0546, see Cyanobase) flanked by ≈ 0.3 kb neighbouring sequences to serve as platform for homologous recombination mediating targeted gene replacement in Synechocystis
pminCDE with the Kmr marker from pUC4K inserted in place of the sequence from the codon 6 of minC to the stop codon of minE
pminCDE with the Kmr marker from pUC4K inserted in place of the sequence from minD codon 31, to minE stop codon
pminCDE with the Kmr marker from pUC4K inserted in place of the minC sequence from codon 72 to codon 174
pminCDE with the Kmr marker from pUC4K inserted in place of the minD sequence from codon 31 to codon 133
pminCDE with the Kmr marker from pUC4K inserted in place of the full ORF (294 bp) of minE
pGEMT harbouring the 2145 bp ORF of zipN (Sll0169, see Cyanobase)
pzipN with the Kmr marker from pUC4K inserted downstream of the codon 100 of zipN
Analysis of GFP-tagged proteins in vitro
Source of the EGFP (enhanced green fluorescent protein)
Apr KmrE. coli plasmid with a strong promoter
pCR2.1 plasmid (purchased from Invitrogen) harbouring the Synechocystis fedI gene (ssl0020) that possesses a ribosome binding site highly effective in both Synechocystis and E. coli
To be published elsewhere
pCfedI with the fed ORF replaced by the SmaI restriction site for cloning the studied ORF to be overexpressed in E. coli
pC with the 6xHis-EGFP gene fusion in front of the SspI site for in frame cloning of the studied ORF downstream of the tag
pC with the EGFP-6xHis gene fusion behind the SspI site for in frame cloning of the studied ORF upstream of the tag
pC with the Synechocystis ftsZ ORF (sll1633, see Cyanobase)
pC with the Synechocystis ftsZ ORF lacking the 62 N-terminus codons
pC with the Synechocystis ftsZ ORF lacking the 60 C-terminus codons
pCGFPHis with the Synechocystis ftsZ ORF upstream of the EGFP-6xHis tag
pCGFPHis with the Synechocystis ftsZ ORF (lacking its 62 N-terminus codons) upstream of the EGFP-6xHis tag
pCGFPHis with the Synechocystis ftsZ ORF (lacking its 60 C-terminus codons) upstream of the EGFP-6xHis tag
pCGFPHis with the Synechocystis ftsZ ORF (lacking its 62 N-terminus and 60 C-terminus codons), upstream of the EGFP
pCZGGTG to AAAAGFP
Same as pCZGFP, except that the GGTG amino acids (at position 162–165) of FtsZ have been changed to AAAA
pC with the Synechocystis minC ORF (sll0288, see Cyanobase)
pC with the Synechocystis minD ORF (sll0289, see Cyanobase)
pC with the Synechocystis minD ORF (sll0289, see Cyanobase) tagged with 6xHis-EGFP at its N-terminus
Same as pCGFPminD, except that the 20 C-terminus codons of minD have been deleted
pCGFP-minDVVT to AAI
Same as pCGFPminD, except that the VVT amino acids (at position 6, 7 and 8) of MinD have been changed to AAI
pC with the Synechocystis zipN ORF (sll0169, see Cyanobase) tagged with 6xHis-EGPF at its N-terminus
Same as pCGFPZipN, except that the 59 N-terminus codons (from position 4–63) of ZipN has been deleted
Analysis of GFP-tagged proteins in Synechocystis
Kmr Smr Spr plasmid that replicates in Synechocystis where it expresses any gene cloned downstream of its E. coli tac promoter
Kms derivative of pSBTac without the 0.95 kb PvuII–NruI segment carrying the 5′ half of the Kmr gene. All tested genes were cloned, as blunt end DNA segment, downstream of its E. coli tac promoter (i.e. at the unique HpaI site)
pS expressing the EGFP-6xHistag protein
pSGFP with the Synechocystis ftsZ ORF upstream of the EGFP-6xHis coding sequence
pSGFP with the Synechocystis ftsZ ORF (lacking its 62 N-terminus codons) upstream of the EGFP-6xHis tag
pSGFP with the Synechocystis ftsZ ORF (lacking its 60 C-terminus codons) upstream of the EGFP-6xHis tag
pSZGGTG to AAAA-GFP
Same as pSZGFP, except that the GGTG amino acids (at position 162–165) of FtsZ have been changed to AAAA
pSBTac with the Synechocystis minD ORF downstream of the EGFP-6xHis tag
Same as pSGFPminD, except that the 20 C-terminus codons of minD have been deleted
pSGFP-minDVVT to AAI
Same as pSGFPminD, except that the VVT amino acids (at position 6, 7, 8) of MinD have been changed to AAI
pSGFP with the Synechocystis zipN ORF downstream of the EGFP-6xHis tag
Same as pSGFP-ZipN, except that the 59 N-terminus codons (from position 4–63) of ZipN gene have been deleted
E. coli two-hybrid analysis of FtsZ–ZipN interaction
E. coli plasmid encoding the C-terminal T18 fragment (aa 225–339) of the Bordetella pertussis adenylate cyclase (CyaA) in frame with a multiple cloning site
Same as pKT25-zipN, except that the 59 N-terminus codons (from position 4–63) of zipN gene have been deleted
In vivo decoration of the septal Z-ring depends on the GTPase domain, but not on the N- and C-termini of the GFP-tagged FtsZ protein
To investigate the subcellular localization of FtsZ, we have fused its coding sequence (either wild-type or mutant) to the one encoding the GFP and cloned the resulting translational fusions (Table 1) in a Synechocystis expression vector that replicates at 10 copies per cell (i.e. at one copy per chromosome copy; Marraccini et al., 1993). The corresponding pS plasmids (S for Synechocystis) were independently introduced in Synechocystis, yielding merodiploid strains harbouring the ftsZ gene at the normal chromosomal locus and the ftsZ–GFP chimera (or GFP alone) in the pS plasmid. These cells retained normal growth and morphology (Fig. 1A–D). GFP fluorescence was either scattered in the cytoplasm of the cells producing GFP alone (Fig. 1D′) or accumulated at the constriction site of FtsZ–GFP reporter cells (Fig. 1A′) in a structure that appeared to be truly annular (Fig. 1B′). The same Z-ring structure was also observed in cells expressing GFP-tagged FtsZ proteins lacking about 60 amino acids at either or both the N- and C-termini, in agreement with the poor sequence conservation of these FtsZ parts. Indeed, the 60 N-terminal amino acids were observed in no other FtsZ orthologues (even of cyanobacterial source), whereas the 60 C-terminal amino acids were conserved in only a few other cyanobacterial FtsZ proteins (e.g. FtsZ from Anabaena PCC7120 or Thermosynechococcus elongatus BP-1, see CyanoBase: http://www.kazusa.or.jp/cyano/cyano.html). In contrast, no Z-ring structure was observed (data not presented) with the GFP-tagged variant of FtsZ harbouring the GGTG to AAAA mutation in the well-conserved GTPase pocket (at position 162–165 in the Synechocystis FtsZ amino acid sequence). The Z-ring of deeply constricted daughter cells was found to be oriented perpendicularly to the bright and tiny mother Z-ring (Fig. 1C′). This finding indicates that Synechocystis divides in alternating perpendicular planes, much like other naturally occurring spherical bacteria (Tzagoloff and Novick, 1977; Westling-Haggstrom et al., 1977). Cloverleaf-like dividing cells with perpendicular septation sites are not normally observed because the first round of septation is usually completed before the initiation of the second one. However, such aberrant structures can be generated after the depletion of a cell division protein (e.g. FtsQ, to be published elsewhere) or a salt challenge of the osmotic stress-sensitive glucosylglycerol-phosphate synthase mutant of Synechocystis (Ferjani et al., 2003).
The in vitro polymerization of FtsZ depends on its GTP-binding domain, but not on its N- and C-termini
To investigate the polymerization of FtsZ in vitro, we have cloned the ftsZ and ftsZ–GFP genes in a multicopy E. coli vector for mass production of the corresponding reporter proteins, yielding the pC (E. coli) series of plasmids (Table 1). After protein purification, FtsZ polymerization was performed (Experimental procedures) basically as described with the E. coli protein (Hale et al., 2000), except that FtsZ polymers were visualized by phase-contrast microscopy and fluorescence microscopy, instead of electron microscopy. The cyanobacterial FtsZ protein self-assembled into higher order polymers, which were fluorescent when produced with GFP-tagged FtsZ molecules (Fig. 2A and A′). FtsZ polymerization was prevented by the absence of GTP in the buffer or the GGTG to AAAA mutation in the FtsZ GTPase pocket (Fig. 2B and B′). In contrast, FtsZ mutant proteins lacking about 60 amino acids at either or both the N- and the C-terminus (Table 1) retained the ability to self-assemble (Fig. 2A and A′ ), in agreement with the fact that these FtsZ parts have been poorly conserved, unlike the GTPase domain (see above). Collectively, the data obtained in vitro confirmed the results of the in vivo analysis of FtsZ polymerization (see Fig. 1).
MinCDE influence cell morphology, as well as the subcellular position and the shape of FtsZ structures
The mutants ΔminCDE::Kmr, ΔminDE::Kmr, ΔminC::Kmr and ΔminD::Kmr all displayed similar aberrant morphologies (Fig. 3) such as spiralled cells, heart-shaped cells (probably resulting from a strongly asymmetric invagination) and minicells. These data showed that MinC and MinD play an important role in cytokinesis, as occurs in E. coli. In contrast, cell shape was poorly affected upon deletion of the minE gene alone (slight asymmetric placement of the septum see Fig. 3E), which generated rare minicells (twofold smaller than wild-type cells; compare Fig. 3F and G). The mild phenotype of the minE null mutant of Synechocystis is remarkable in that inactivation of minE is lethal in E. coli (de Boer et al., 1989). All Synechocystis min-less minicells remained capable of division (Fig. 3F; data not shown), indicating that they possessed at least one copy of their chromosome. This finding is at variance with that found in E. coli where minCDE deletion generates DNA-less minicells.
To investigate whether the abnormal morphology of the min-less mutant correlated with aberrant FtsZ structures, we introduced the pSZGFP reporter plasmid in the ΔminCDE::Kmr cells (Table 1). As expected, the FtsZ–GFP protein did not modified the phenotype of the ΔminCDE::Kmr mutant that retained spiralled cells (Fig. 4). These cells displayed spiral-like FtsZ structures, indicating that the abnormal constrictions were caused directly by cell wall invagination along the aberrant FtsZ structures. Collectively, these results demonstrated that the cyanobacterial Min system controls cell morphology, probably through the control of the structure and localization of FtsZ polymers, as occurs in the spherical rodA– mutant of E. coli (Corbin et al., 2002).
GFP–MinD is biologically active and localizes to cell membranes
To investigate the subcellular localization of the Min proteins, their coding sequence was translationally fused to that of GFP. minC–GFP, GFP–minC and minE–GFP fusions could not be propagated in Synechocystis, probably because they interfere with cell division (consequently, the corresponding plasmids are not described in Table 1). In contrast, GFP–minD did not interfere with the cytokinetic machinery of Synechocystis as wild-type cells harbouring the corresponding pSGFP-minD plasmid displayed normal morphology (data not shown). Interestingly, ΔminDE::Kmr cells carrying pSGFP-minD no longer display spiralled cells (Fig. 5) but retained minicells such as ΔminE::Kmr cells (Fig. 3). This finding showed that GFP–MinD suppress the formation of the spiralled cells resulting from the lack of MinD, but not the minicelling phenotype promoted by the absence of MinE. As a control, we have verified that GFP alone could not prevent the appearance of spiralled cells in the ΔminDE::Kmr mutant (Fig. 5). Then, the subcellular location of the biologically active GFP–MinD protein was investigated through fluorescence microscopy. In both the wild-type (data not shown) and the ΔminDE::Kmr mutant strains (Fig. 5), GFP–MinD was found to be accumulated at the cell membranes.
The C-terminal domain of MinD is critical to MinD attachment to cell membranes
In E. coli, the C-terminal amphipathic helix of MinD is required for MinD binding to cell membranes (Szeto et al., 2002; Hu and Lutkenhaus, 2003). As this domain is well conserved in the cyanobacterial MinD protein (sequence alignment not shown), we decided to investigate its function. Therefore, we have deleted the 20 C-terminal amino acids of the Synechocystis MinD protein (Table 1). The corresponding GFP–MinDΔCTD mutant protein was unable to compensate for the absence of the native MinD. First, a large proportion of the ΔminDE::Kmr cells harbouring the pSGFP-MinDΔCTD plasmid retained the spiralled shape (Fig. 5C) characteristic of the absence of MinD (Fig. 3B), whereas the control cells (i.e. ΔminDE::Kmr cells carry-ing pSGFP-MinD) displayed the normal morphology (Fig. 5A). Secondly, GFP–MinDΔCTD was unable to accumulate at the cell membrane, unlike GFP–MinD (compare Fig. 5A′ and C′).
In E. coli, the biological activity of MinD is known to depend upon its ATPase activity (Hayashi et al., 2001). To investigate whether the same is true in the Synechocystis MinD protein, we have mutagenized the VVT amino acids (VVT to AAI, at position 6–8, Table 1) of its well-conserved ATP-binding region. Although the corresponding GFP–MinDVVT → AAI protein remained targeted to the cell membranes (compare Fig. 5A′ and B′), it could not compensate for the lack of the native MinD (ΔminDE::Kmr cells harbouring pSGFP-MinDVVT → AAI retained the spiralled shape, Fig. 5B). Collectively, these results demonstrated that both C- and N-terminal domains of MinD are crucial to MinD activity, while the CTD domain, but not the NTD, is required for MinD attachment to cell membranes.
ZipN is crucial to cell viability
Synechocystis harbours a copy of the ftn2 gene that is unique to cyanobacteria and plants where it operates in cytokinesis (Koksharova and Wolk, 2002; Vitha et al., 2003). Ftn2 is dispensable to the growth of the unicellular rod-shaped cyanobacterium Synechoccocus PCC7942 (the ftn2 null mutant was filamentous), whereas it is critical to the viability of the filamentous cyanobacterium Anabaena PCC7120 (the ftn2 heteroploid strain displayed enlarged cells; Koksharova and Wolk, 2002). To analyse the Synechocystis ftn2 gene (sll0169 in cyanobase), hereafter designated zipN (see below), we have replaced its coding sequence by the Kmr marker, yielding the ΔzipN::Kmr cassette (Table 1) that was introduced in Synechocystis. The Kmr transformant clones invariably possessed the Kmr marker properly inserted in their genomes (in place of the zipN gene), but only in half of the chromosome copies (about 10 per cell in Synechocystis; Labarre et al., 1989), as the other copies remained wild type. This finding showed that zipN is essential to cell viability in Synechocystis, as occurs in Anabaena PCC7120 (Koksharova and Wolk, 2002). The ΔzipN::Kmr/zipN+ heteroploid mutant of Synechocystis displayed spiralled cells and minicells (Fig. 6), like the minCDE and minDE null mutants (Fig. 3). However, the aberrant morphology of the ΔzipN::Kmr/zipN+ strain was transient (it disappeared after several rounds of subcultivation), probably because it accumulated suppressor mutations.
GFP–ZipN localizes to the septum, probably through binding on to FtsZ polymers, in a MinCDE-independent way
The pSGFP-zipN plasmid (Table 1) was constructed to investigate the cellular distribution of ZipN. As expected, we found that GFP–ZipN does not interfere with the cytokinetic machinery as wild-type cells harbouring this plasmid retained normal morphology (data not presented). Then, we tested whether GFP–ZipN could compensate for the lack of the native ZipN. Therefore, we introduced the zipN inactivation cassette (ΔzipN::Kmr) in cells carrying either the pSGFP-zipN vector or the pSGFP plasmid (negative control). Cells harbouring pSGFP-zipN could achieve complete chromosome segregation in possessing only mutant copies of the chromosome (i.e. ΔzipN::Kmr), and nevertheless displayed a normal morphology (Fig. 6). In contrast, cells carrying the negative control plasmid pSGFP remained heteroploid (ΔzipN::Kmr/zipN+) and retained the aberrant phenotype characteristic of ZipN depletion (data not shown). Together, these data demonstrate that GFP–ZipN is biologically active. The fact that it was accumulated at the cell septum (Fig. 6) suggested that ZipN might physically interact with the FtsZ ring.
We have also investigated the influence of the Min system on the subcellular localization of ZipN. Therefore, the pSGFP-zipN reporter plasmid was introduced into the ΔminCDE::Kmr mutant, and the resulting cells were analysed by fluorescence microscopy. The results in Fig. 7 show that ZipN networks closely resemble FtsZ structures (Fig. 4), suggesting that ZipN decoration of Z polymers occurs in the absence of the MinCDE proteins.
Collectively, these results demonstrated that ZipN accumulates at the septum of the cell, probably through binding onto the FtsZ ring, in a process that does not require the Min system.
The N-terminal DnaJ domain of ZipN is critical to the septal localization of ZipN
The DnaJ domain located at the extreme N-terminus of the cyanobacterial ZipN protein was analysed through deletion of its codons 4–63 from the GFP-tagged version of this protein (Table 1). The corresponding GFP–ZipNΔNTD mutant protein was found to be unable to compensate for the absence of ZipN, unlike GFP–ZipN. First, ΔzipN::Kmr cells carrying the pSGFP-ZipNΔNTD plasmid retained wild-type (i.e. zipN+) chromosome copies, and displayed the aberrant morphologies characteristic of the decreased level of ZipN (Fig. 6). Secondly, the GFP–ZipNΔNTD mutant protein could not localize at the cell septum (it remained entirely cytoplasmic, Fig. 6). Collectively, these results demonstrated that the DnaJ domain of ZipN is crucial to the biological function of ZipN, including its accumulation at the cell septum.
The DnaJ domain of ZipN is crucial to the binding of ZipN on to FtsZ polymers
Taking advantage of the ability of the scaffold protein FtsZ to self-assemble in vitro (Fig. 1), we have developed a simple assay to study the binding of GFP-tagged proteins on to FtsZ polymers. This test is based on the fact that if, and when, a GFP fusion protein is binding on to preassembled FtsZ polymers, the decorated Z structures become fluorescent. GFP-tagged ZipN proteins were incubated with preformed FtsZ polymers (Experimental procedures), and the mixture was examined by microscopy (Fig. 8). The FtsZ filaments appeared to be fluorescent when GFP–ZipN (Fig. 8A and A′), but not GFP–ZipNΔNTD (Fig. 8B and B′), was used in the test. These findings, together with those of the in vivo analysis of GFP–ZipN (Fig. 6), showed that the N-terminal DnaJ domain of ZipN is required for ZipN binding on to FtsZ network.
Two-hybrid analysis of the FstZ–ZipN interaction: further evidence of the essential role of the DnaJ domain of ZipN
ZipN–FtsZ interaction was investigated with the bacterial two-hybrid system based on the interaction-mediated reconstitution of the functional adenylate cyclase from Bordetella pertussis, in the E. coli strain DHP1 that is deficient in the endogenous cAMP-producing enzyme (Karimova et al., 1998). The Synechocystis ftsZ and zipN protein coding sequences were translationally fused to the adenylate cyclase domains T18 and T25 (Experimental procedures), yielding the plasmids pUT18-ftsZ and pKT25-zipN respectively. Double transformant clones of E. coli DHP1 propagating both the pUT18-ftsZ and the pKT25-zipN plasmids turned dark blue in colour on Xgal–IPTG indicator plates and exhibited a high level of β-galactosidase activity (Table 2). In contrast, every type of negative control clone remained white and displayed only background levels of β-gal activity. Finally, the N-terminal domain of ZipN was found to be required for the successful ZipN–FtsZ interaction, in agreement with the above-mentioned data showing that the NTD domain of ZipN is involved in the ZipN decoration of FtsZ polymers (Fig. 8).
Table 2. . Utilization of a bacterial two-hybrid system to validate the interaction between FtsZ and ZipN.
Coding sequence cloned in pUT18
Coding sequence cloned in pKT25
Colony colour of the double transformants on Xgal plates
β-Galactosidase activity (nmol min−1 mg−1)
The ZipN–FtsZ interaction was confirmed with the bacterial two-hybrid system that is based on the interaction-mediated reconstitution of an adenylate cyclase (CyaA) activity in the enzyme-deficient E. coli strain DHM1 (Karimova et al., 1998). The sequences encoding the tested proteins were cloned (see Experimental procedures) in the pUT18 or pKT25 plasmids encoding the C-terminal T18 fragment or the N-terminal T25 fragment of CyaA respectively. The leucine zipper domain (zip domain) of the yeast GCN4 protein was used as a positive control for protein–protein interaction (Karimova et al., 1998). Reconstitution of the CyaA activity in E. coli DHM1 cells co-transformed with pKT25- and pUT18-derived plasmids was ascertained by the appearance of both the blue colour and the β-galactosidase activity of the cells (Experimental procedures).
4213 ± 385
75 ± 4
94 ± 6
99 ± 6
1377 ± 82
47 ± 1
For the first time, several key components of the cyanobacterial cell division machinery have been analysed through mutagenesis and fusion to the GFP, as well as with a bacterial two-hybrid system (Karimova et al., 1998). Because cytokinesis is poorly known in non-cylindrical cells, we have chosen the spherical host Synechocystis sp. PCC 6803 (Synechocystis). Synechocystis is also attractive because it is well suited to gene manipula-tion as its small genome (3.57 Mb) has been fully sequenced (Kaneko et al., 1996) (see Cyanobase at http://www.kazusa.or.jp/cyano/cyano.html), and it can easily be manipulated with replicating plasmids (Mazouni et al., 1998; Poncelet et al., 1998).
The FtsZ protein (Sll1633 in Cyanobase), an ancestral form of the eukaryotic cytoskeletal protein tubulin, was found to be essential for the viability of Synechocystis where it assembles in a septal ring-like structure at mid-cell (Fig. 1), as occurs in E. coli (Margolin, 2001; Rothfield et al., 2001; Lutkenhaus, 2002; Errington et al., 2003). In Synechocystis, the Z-ring of deeply constricted daughter cells appeared to be oriented perpendicularly to the bright and tiny mother Z-ring (Fig. 1), showing that this organism divides in alternating perpendicular planes like other spherical bacteria (Tzagoloff and Novick, 1977; Westling-Haggstrom et al., 1977). This finding is at strong variance with what occurs in E. coli, which divides within one and the same plane (Young, 2003). With a view to unravelling the molecular basis of this different behaviour in future, we have initiated the analysis of the cyanobacterial MinCDE system. In the cylindrical bacterium E. coli, the Min proteins oscillate from one cell pole to another, thereby preventing placement of the Z-ring at all sites except mid-cell. However, this working model for division site placement cannot be entirely valid in spherical cells, such as those of Synechocystis, that do not have a defined middle but instead possess an infinite number of potential division planes at the point of greatest diameter. The MinCDE proteins of Synechocystis were found to be dispensable for cell growth, although they influence both cell morphology (Fig. 3) and the shape and position of FtsZ structures (Fig. 4). MinC (Sll0288) and MinD (Sll0289) are equally important to the control of both the size and the shape of the cells, whereas MinE (Ssl0546) mainly affects the size of the cell (Fig. 3). This is at strong variance with what occurs in E. coli, where inactivation of minE alone is lethal (de Boer et al., 1989). Focusing our attention on MinD, because it is the best conserved and most widespread protein of the MinCDE trio occurring in all domains of life (eubacteria, archaea and eukaryotes), we found that MinD is accumulated at the cell membranes of Synechocystis in a process that requires the C-terminal amphipathic helix of MinD but not its ATP-binding region (Fig. 5).
We have also analysed the Ftn2 protein (Sll0169) that is unique to the cytokinetic apparatus of cyanobacteria (Koksharova and Wolk, 2002) and plastid (Vitha et al., 2003). Ftn2 was previously found to be essential to the viability of the filamentous (e.g. multicellular) cyanobacterium Anabaena PCC7120, but not to that of the unicellular cyanobacterium Synechococcus PCC7942 (Koksharova and Wolk, 2002). In reporting that Ftn2 is crucial in Synechocystis (unicellular, like Synechococcus), we are ruling out the possibility that the critical versus dispensable role of Ftn2 in cyanobacteria is related to cell status (filamentous versus unicellular). Rather, these findings suggest that Ftn2 might be dispensable only in a cylindrical host (such as Synechococcus), and essential only in spherical organisms (such as Anabaena and Synechocystis). Using a combination of in vivo and in vitro assays, we show in this study that Ftn2 localizes to the septum of Synechocystis (Fig. 6) in a MinCDE-independent process (Fig. 7) through physical interactions with FtsZ (Fig. 8 and Table 2). Consequently, we propose to rename Ftn2 as ZipN for Z-ring-interacting protein N. Finally, the N-terminal DnaJ domain of ZipN was found to be involved in the septal localization of ZipN in Synechocystis and in the physical interaction of this protein with FtsZ.
The present data demonstrate that the naturally occurring cyanobacterium Synechocystis is an appealing alternative to the rodA– mutant of E. coli to investigate the cytokinesis of spherical-celled organisms. In addition, our work will certainly stimulate the analysis of cell division in many other cyanobacteria displaying a variety of cell morphology (Rippka et al., 1979), as well as in plants that share several cytokinetic proteins in common with cyanobacteria (FtsZ, MinC, MinE, ARTEMIS and Arc6, the orthologue of ZipN; Osteryoung and McAndrew, 2001; Fulgosi et al., 2002; Koksharova and Wolk, 2002; Lutkenhaus, 2002; Vitha et al., 2003), some of which are unique to plants and cyanobacteria (ARTEMIS and Arc6/ZipN). Because Arc6 is localized to a septal ring (Vitha et al., 2003), which might turn out to be the Z-ring, it will be especially interesting to test whether Arc6 can physically interact with FtsZ, like its ZipN orthologue (our results, see Figs 6, 7 and 8 and Table 2).
Bacterial strains, growth, plasmids and gene transfer procedures
Synechocystis PCC6803 (Synechocystis) was grown at 30°C in BG11 medium enriched with Na2CO3 (3.78 mM final concentration), under continuous white light of standard fluence (2500 lux, i.e. 31.25 µE m−2 s−1). E. coli strains HB101, XL1Blue MRF′ and CM404, which were grown on LB at 30°C, were used for gene manipulation (HB101 or XL1Blue MRF′) or for conjugative transfer (CM404) to Synechocystis (Mermet-Bouvier et al., 1993) of the plasmids (Table 1) derived from the constitutive expression vector pSB2T (Marraccini et al., 1993). This vector replicates at about 10 copies per cell (i.e. one copy per chromosome) in Synechocystis (Mermet-Bouvier et al., 1993). Occasionally, plasmids were introduced in Synechocystis through electroporation (Mermet-Bouvier et al., 1993) instead of conjugation. The final concentration of the selective antibiotics (see Table 1) was as follows: ampicillin (Ap) 100 µg ml−1, kanamycin (Km) 50 µg ml−1, nalidixic acid 20 µg ml−1 and spectinomycin (Sp) 100 µg ml−1 for E. coli; Km 50–300 µg ml−1 and Sp 10 µg ml−1 for Synechocystis.
Gene cloning and manipulation
All Synechocystis genes presently studied (Table 1) were amplified by PCR from chromosomal DNA of the wild-type strain as the templates, using appropriate specific primers, and cloned as blunt end fragments (unless stated otherwise, see Two-hybrid assay) in which every nucleotide substitution used for site creation was eliminated upon cleavage. Site-directed mutagenesis and deletions were made with the Quick ChangeTM mutagenesis kit (Stratagene). The sequence of every construction was verified with the Big Dye kit (ABI Perkin-Elmer).
The bacterial two-hybrid assay
This system is based on the interaction-mediated reconstitution of an adenylate cyclase activity in the enzyme-deficient E. coli strain DHM1 (Karimova et al., 1998). The Synechocystis ftsZ coding sequence, lacking its stop codon, was cloned as a PstI–BamHI restriction fragment in the PstI–BamHI site of the pUT18 plasmid harbouring the C-terminal T18 fragment (amino acids 225–339) of the Bordetella pertussis adenylate cyclase. This yielded the in frame ftsZ-T18 translational fusion gene. Similarly, a PstI–BamHI restriction fragment carrying the Synechocystis zipN coding sequence without its methionine initiator codon was inserted into the PstI–BamHI site of the pKT25 plasmid encoding the N-terminal T25 fragment (amino acids 1–224) of the B. pertussis adenylate cyclase. This yielded the in frame T25-zipN translational fusion gene. The nucleotide sequence of every fusion genes was verified. Then, the E. coli DHM1 strain was doubly transformed with pKT25- and pUT18-derived plasmids. Reconstitution of the catalytic domains of B. pertussis adenylate cyclase through fused portions was ascertained by two methods. First, we verified that bacterial colonies turned blue in 2 days at 30°C on indicator plates, i.e. Luria–Bertani medium supplemented with 40 µg ml−1 Xgal (Eurobio), 0.5 mM IPTG (Invitrogen), ampicillin, kanamycin and nalidixic acid. Otherwise (absence of interaction or negative control) white colonies appeared. Secondly, the β-galactosidase activity of the bacterial lysate was measured after overnight culture of the cells at 30°C. β-Galactosidase activities are the mean value of three measurements performed on the cellular extracts of two brother E. coli clones. 1 β-gal unit = 1 nmol of ONPG min−1 mg−1 protein.
Each XL1Blue E. coli strain harbouring a recombinant plasmid was spread on a set of 20 LB plates containing both Ap and Km (100 µg ml−1 each) and grown for 2 days at 37°C. Cells were harvested, washed and resuspended in 10 ml of purification buffer (PB), 50 mM Tris, pH 7.5, 150 mM NaCl. Then, cells were rapidly frozen in an Eaton press chamber (Eaton, 1962), cooled in a dry-ice ethanol bath and disrupted (250 MPa). The cell extract was centrifuged at 14 000 g for 20 min at 4°C, and the cleared supernatant (about 7 ml) was collected.
The native cyanobacterial FtsZ protein was purified from the engineered E. coli strain with DEAE sephacel beads (Pharmacia) through a two-round batch procedure as follows. The cleared cell extract was gently mixed with 300 µl of beads for 30 min at 4°C before centrifugation (14 000 g, 5 min) and five washes with 20 ml of PB. FtsZ molecules were eluted with 1 ml of 1 M KCl, diluted with PB to a final concentration of 100 mM KCl and incubated for 60 min at 4°C with 150 µl of beads. After five washes with 20 ml of PB, the purified FtsZ protein was eluted with 200 µl of PB containing KCl at 1 M.
All cyanobacterial cytokinetic proteins that were fused to a 6xHis-GFP tag were purified with Ni-NTA agarose beads (Qiagen) according to the above-described procedure, except that the proteins were eluted with PB containing 1 M imidazole.
Protein concentration was determined with the Bio-Rad protein assay reagent using bovine serum albumin as standard. In every case, the purity of the studied proteins was > 90%, as judged by SDS-PAGE.
In vitro analyses of FstZ polymerization and FtsZ interaction with GFP-tagged Zip proteins
Polymerization reactions were performed with the wild-type or mutant versions of the FtsZ protein (at 5 µM) with or without the GFP-6xHis tag, as described previously (Yu and Margolin, 1997). The mixture was incubated at 30°C for 20 min in the polymerization buffer: 50 mM Tris, pH 7.5, 150 mM NaCl, 10 mM MgCl2 and 1 mM GTP.
Interactions between preassembled FtsZ polymers with either wild-type or mutated MinD or ZipN proteins tagged with GFP (Table 1) were done basically as described previously (Hale et al., 2000), with the following modifications. FtsZ (5 µM) was incubated in polymerization buffer for 20 min at 30°C before the addition of either GFP–MinD or GFP–ZipN (0.5 µM final concentration). The mixtures were incubated for 10 min and then added directly to a glass slide, covered with a coverglass. The polymers were visualized by phase-contrast (FtsZ or FtsZ–GFP) or fluorescence microscopy (FtsZ–GFP or native FtsZ polymers decorated by either GFP–MinD or GFP–ZipN).
Light and fluorescence microscopy
Images of wet mounts of cells, grown on solid medium, were captured with a Leica DMRXA microscope equipped with a Ropper Scientific Micromax cooled CCD camera and metamorph software (Universal Imaging).
K.M. and F.D. are recipients of thesis fellowships from the Ministry for Education of Algeria and from the CEA Saclay France respectively.
We thank G. Karimova and D. Ladant (Institut Pasteur) and L. Selig (Hybrigenics) for sharing the bacterial two-hybrid system with us, and A. Sentenac, M. Werner and A. W. Rutherford for support of our work.