Bacterial division requires the co-ordination of membrane invagination, driven by the constriction of the FtsZ-ring, and concomitant cell wall synthesis, performed by the high-molecular-weight penicillin-binding proteins (HMW PBPs). Using immunofluorescence techniques, we show in Streptococcus pneumoniae that this co-ordination requires PBP3, a d,d-carboxypeptidase that degrades the substrate of the HMW PBPs. In a mutant deprived of PBP3, the apparent rings of HMW PBPs and that of FtsZ are no longer co-localized. In wild-type cells, PBP3 is absent at the future division site and present over the rest of the cell surface, implying that the localization of the HMW PBPs at mid-cell depends on the availability of their substrate. FtsW, a putative translocase of the substrate of the PBPs, forms an apparent ring that is co-localized with the septal HMW PBPs throughout the cell cycle of wild-type cells. In particular, the constriction of the FtsW-ring occurs after that of the FtsZ-ring, with the same delay as the constriction of the septal PBP-rings. However, in the absence of PBP3, FtsW remains co-localized with FtsZ in contrast to the HMW PBPs. Our work reveals an unexpected complexity in the relationships between the division proteins. The consequences of the absence of PBP3 indicate that the peptidoglycan composition is central to the co-ordination of the division process.
Bacterial division is a complex and poorly understood process, which encompasses the segregation of chromosomes, the membrane invagination and final fusion event and the concomitant synthesis of the septal cell wall. Genetic and localization experiments conducted mainly in Escherichia coli and Bacillus subtilis have shown that a conserved set of proteins is recruited to the site of division in the following conditional order: FtsZ, FtsA, FtsK, FtsQ, FtsL/FtsB, FtsW and FtsI (Buddelmeijer and Beckwith, 2002; Errington et al., 2003). These essential proteins are termed Fts, as thermosensitive mutants of E. coli in the corresponding genes do not divide and adopt a filamentous phenotype at restrictive temperature. FtsZ polymerizes at mid-cell on the inner side of the cytoplasmic membrane as a ring, the constriction of which is thought to drive membrane invagination (Addinall and Holland, 2002). FtsA binds to both FtsZ and the membrane and possibly mediates the interaction between them (Buddelmeijer and Beckwith, 2002; Errington et al., 2003). FtsK is a membrane protein with a cytoplasmic domain that is involved in chromosome segregation (Buddelmeijer and Beckwith, 2002; Errington et al., 2003) and possibly in membrane fusion (Sharp and Pogliano, 1999). FtsQ, FtsL and FtsB are bitopic membrane proteins of unknown function (Buddelmeijer and Beckwith, 2002; Errington et al., 2003). FtsW has 10 transmembrane segments (Gérard et al., 2002) and has been proposed to translocate the precursor of the peptidoglycan to the extracellular side of the membrane where it is incorporated in the cell wall by several penicillin-binding proteins (PBPs), including FtsI (Matsuhashi et al., 1990).
Although the Fts proteins are recruited in a conditional succession, several lines of evidence indicate that a complex comprising all the division proteins does not exist throughout the septation process. Indeed, immunolocalization of FtsZ and PBP2x (FtsI) during the cell cycle of Streptococcus pneumoniae has shown that constriction of the FtsZ-ring is well under way while PBP2x remains at the periphery of the septum (Morlot et al., 2003). Similarly, in B. subtilis, the localization of PBP2B was found to span the whole cell diameter, whereas clear signs of membrane constriction were observed (Daniel et al., 2000). These findings raise the question of how the processes of membrane invagination and septal cell wall formation are co-ordinated.
d,d-Carboxypeptidases are not essential and have not hitherto been considered as major participants of the division mechanism, although a mutant of S. pneumoniae (dacA) deficient in its single d,d-carboxypeptidase PBP3 exhibits defects in division with multiple septa initiated at aberrant locations (Schuster et al., 1990). Similarly, the dacA mutant of E. coli deficient in PBP5 has an abnormal morphology (Nelson and Young, 2000; 2001). We demonstrate here that the d,d-carboxypeptidase PBP3 is central to the organization of the division process in S. pneumoniae.
d,d-Carboxypeptidases are low-molecular-weight PBPs. In contrast to the high-molecular-weight (HMW) PBPs that synthesize the cell wall peptidoglycan, the d,d-carboxypeptidases are thought to regulate its degree of cross-linking. The peptidoglycan consists of disaccharide chains that are cross-linked by peptide bridges. These bridges are formed by a transpeptidation reaction between a primary amine from an acceptor stem-peptide attached to one glycan chain and a donor stem-pentapeptide attached to an adjacent chain, with the departure of the last residue of the pentapeptide. Trimming of the last residue of the pentapeptides by d,d-carboxypeptidases reduces the availability of donors for the transpeptidation reaction and thus limits the reticulation of the peptidoglycan.
To investigate the role of PBP3 in the growth and division of S. pneumoniae, the localization of FtsZ, FtsW and the HMW PBPs was determined in a wild-type and a dacA mutant strain, using various combinations of double immunolabelling. In the PBP3-deficient cells that are not engaged in division, we show that the rings formed by FtsZ and FtsW are spatially dissociated from the rings formed by the HMW PBPs, including PBP2x (FtsI).
The localization of PBP3 in a wild-type strain of S. pneumoniae suggests that it exerts its role in division not through direct protein–protein interaction, but probably through its enzymatic activity, which modifies the composition of the peptidoglycan.
Localization of the HMW PBPs in a PBP3-deficient strain
In a previous work, we have studied the localization of all the HMW PBPs and FtsZ during the cell cycle of a wild-type strain of S. pneumoniae. The reported phenotype of the dacA mutant, which shows severe disorders in septa formation and a thickened cell wall (Schuster et al., 1990), prompted us to study the localization of the HMW PBPs and FtsZ in this strain. The dacA mutant expresses a PBP3 variant without its 51 C-terminal residues, which include the membrane anchor. The truncated PBP3 of this mutant is therefore secreted in the medium. No PBP3 can be detected in whole-cell preparations (Schuster et al., 1990), and we could not detect any by immunofluorescence (data not shown).
Immunofluorescence pictures were obtained with rabbit antisera against the PBPs together with a mouse antiserum against FtsZ. Data were more difficult to obtain with the dacA mutant than with the wild-type strain, particularly regarding staining of the PBPs. The thick cell wall of this mutant may impair fixation or staining, or the localization of the PBPs may be more diffuse. Thus, < 60% of the mutant cells displayed defined staining patterns for the different PBPs. Nearly all cells with a clear localization of the PBPs had a single nucleoid. Among those, about 30% displayed the typical localization found with the wild-type strain, whereas 70% exhibited unexpected fluorescence patterns.
The patterns observed with the dacA mutant stained for DNA, FtsZ and PBP2x are shown in Fig. 1. These patterns are representative of the immunolabelling observed with the other four HMW PBPs (PBP2b, PBP1a, PBP2a, PBP1b; data not shown). Some cells displayed perfect superposition of FtsZ and PBP2x (Fig. 1A) as observed with the wild-type strain (Morlot et al., 2003). In other cells, FtsZ and PBP2x appeared as superimposed circles or ellipses (Fig. 1B), which have not been observed with wild-type pneumococci. Whereas wild-type cells are football-shaped and always lie along their long axis, Gram staining (not shown) and electron micrographs (Schuster et al., 1990) have shown that dacA mutant cells are more spherical. With this rounded morphology, cells can lie on the slide randomly, including in orientations that allow a frontal view of the septal rings of FtsZ and PBPs. Unfortunately, this opportunity to view the septum frontally concerns only isolated cells with a single nucleoid, as they become elongated as division proceeds.
Most interestingly, some mutant cells displayed a circle of FtsZ crossed by a band of PBP2x and vice versa (Fig. 1C and D). Other cells showed crossed bands of FtsZ and PBP2x (Fig. 1E). These patterns, also observed with the other HMW PBPs (not shown), indicate that their annular localization does not depend on the FtsZ-ring in the absence of PBP3.
We could not identify unexpected localization for stages of the cell cycle in which division is under way. Therefore, it was impossible to determine whether cells that do not display initial co-localizations of FtsZ and the PBPs (Fig. 1C–E) are able to divide.
Localization of FtsW in wild-type and dacA mutant strains
We have shown previously that septal PBPs (PBP2x and PBP1a) are not co-localized with FtsZ throughout the division process. Instead, constriction of the FtsZ-ring precedes that of PBP2x and PBP1a by about 5 min, when cells are growing with a generation time of 35 min.
Several lines of evidence point to a close functional relationship between the class B PBPs, which include PBP2x and the family of proteins that includes FtsW. First, in E. coli and B. subtilis, the localization of the septal class B PBPs depends on FtsW (Mercer and Weiss, 2002). Secondly, E. coli cells deficient in PBP2 or RodA (its cognate FtsW-like protein) have the same phenotype (Asoh et al., 1983). Thirdly, in each complete bacterial genome, there are an equivalent number of genes encoding class B PBPs and FtsW-like proteins, with their respective genes often grouped in operons. These considerations prompted us to investigate whether the localizations of FtsW and the septal PBPs are identical.
Mouse and rabbit antisera were raised against FtsW and FtsZ, respectively, to allow double immunolabelling. Figure 2 shows the localization of FtsW and FtsZ during the cell cycle of a wild-type strain. As every cell displayed one of the represented patterns (Fig. 2), and assuming that all cells are undergoing growth and division, the relative number of cells displaying each pattern is directly proportional to the duration of the corresponding stage of the cell cycle (Morlot et al., 2003) (Table 1). A delay of about 5 min is observed between the onset of FtsZ-ring constriction and the apparent diminution of the distribution of FtsW (Fig. 2C), with cells growing with a generation time of 35 min. Near the end of the division process, FtsZ is already relocalized to the equator of the daughter cells, whereas FtsW remains at the septum (Fig. 2D). FtsW even appears to persist at the poles (Fig. 2A–C). Thus, the localization of FtsW during the cell cycle appears to be identical to that of PBP2x and PBP1a, with the exception of the persistence of a polar localization in some cells. This was confirmed by the double immunolabelling of FtsW and PBP2x, which shows the co-localization of the two proteins at the septum in every cell (Fig. 3) and some additional polar distribution of FtsW (Fig. 3A, D and E).
Table 1. . Relative amount (%) of the localization patterns displayed by the different combinations of Fts and PBPs in the wild-type strain and duration of the corresponding stages of the cell cycle [in brackets (min)].
a. 2A−2F, 3A−3E and 6A−6D refer to the fluorescence patterns shown in Figs 2, 3 and 6 respectively.
2A + 2F
41 ± 3 (15)
15 ± 2 (5)
14 ± 2 (5)
14 ± 2 (5)
15 ± 2 (5)
3A + 3E
43 ± 3 (15)
15 ± 2 (5)
27 ± 3 (10)
15 ± 2 (5)
6A + 6D
56 ± 3 (15)
29 ± 3 (10)
14 ± 2 (5)
In the dacA mutant, the annular distribution of FtsW could be observed in some cells on account of their spherical shape. In contrast to the PBPs, FtsW always appears to be co-localized with FtsZ (Fig. 4), except for the delay of the constriction of the annular distribution. Thus, the deficiency in PBP3 does not result in a spatial dissociation of the FtsW- and FtsZ-rings. Consistently, in the absence of PBP3, the annular localizations of PBP2x and FtsW are uncoupled in cells that are not undergoing division. Indeed, in the dacA mutant, the double immunolabelling of PBP2x and FtsW (Fig. 5) shows fluorescent patterns similar to those observed with the labelling of the PBPs and FtsZ (Fig. 1). Note that the size of the observed rings, ellipses or bands varies, which reflects the irregular size and morphology of the dacA mutant cells (Schuster et al., 1990).
Localization of the d,d-carboxypeptidase PBP3
Considering the dramatic consequences of the absence of PBP3, we determined its localization in wild-type cells to understand its function better. In isolated cells or diplococci with a single central nucleoid and an equatorially localized FtsZ-ring (Fig. 6A and D), PBP3 labelling was found to be evenly distributed on both hemispheres, but absent at the equator, the site of future division. Thus, the localization of PBP3 in non-dividing cells appears to be the reverse of that of the HMW PBPs (Morlot et al., 2003).
In the next staining pattern (Fig. 6B), PBP3 immunofluorescence appears to be lighter on the old hemispheres of the dividing bacteria, whereas it is brighter in a central zone. The greater labelling of PBP3 between the parting old hemispheres is even more visible in cells with a constricted FtsZ-ring, which appears as a dot between the nucleoids (Fig. 6C).
As for the labelling of the HMW PBPs and FtsW, the relative number of the various observed patterns could be translated into the duration of the corresponding stages of the cell cycles. Values are given in Table 1.
The d,d-carboxypeptidase PBP3 is a major determinant of the division site
The localization of the various HMW PBPs, FtsZ and FtsW during the cell cycle is represented in Fig. 7, according to the observations presented in this and a previous study (Morlot et al., 2003). It is assumed that the new peptidoglycan is the product of the observed PBPs.
Our main observation is the spatial dissociation of the rings formed by FtsZ and FtsW from those formed by the HMW PBPs, in a strain deficient in PBP3 (Figs 1 and 5). Thus, the d,d-carboxypeptidase PBP3 plays a major role in organizing the growth and division processes. This finding adds to the growing evidence that the d,d-carboxypeptidases are key elements in the bacterial morphogenetic apparatus (Young, 2003).
HMW PBPs catalyse the synthesis of the peptidoglycan
Stem-peptides attached to the disaccharide units of the glycan chain are cross-linked by a transpeptidation reaction. In S. pneumoniae, the major donor stem-peptide is the l-Ala-d-iGln-l-Lys-d-Ala-d-Ala pentapeptide, which loses its terminal d-Ala residue while forming a new peptide bond with the amine of the lysine of an acceptor stem-peptide (Severin and Tomasz, 2000). The presence of PBP3 on the surface of both hemispheres of wild-type cells (Fig. 6A and D) ensures that no donor stem-peptides are available for transpeptidation in these normally inert regions of the cell envelope. Inversely, the absence of PBP3 at the equator (Fig. 6A and D) ensures that it is the only place where donor pentapeptides are available for transpeptidation. The localization of the HMW PBPs would thus be mediated by the availability of substrate for their transpeptidase activity. Indeed, the non-overlapping distribution of PBP3 and the HMW PBPs rules out a direct interaction between these proteins. Also, the dacA mutant in E. coli can be complemented by another low-molecular-weight PBP with an active site modified to mimic that of the product of dacA (Ghosh and Young, 2003). This result demonstrates that the enzymatic activity encoded by dacA, rather than the protein supporting this activity, is crucial for its cellular function (Nelson et al., 2002; Ghosh and Young, 2003).
The mechanism of exclusion of PBP3 from the equator remains to be investigated. PBP3 is anchored to the cell surface by a C-terminal amphipathic helix that lies on the plasma membrane. It is therefore possible that the local lipid composition of the membrane influences the localization of PBP3. Cardiolipin, for example, has been shown to be concentrated at the site of division in E. coli (Mileykovskaya et al., 1998). Another intriguing possibility is that FtsK plays a role in the exclusion of PBP3 at mid-cell. Two different point mutations in the N-terminal, membrane-embedded portion of FtsK cause a late block of the division of E. coli, which is suppressed by the deletion of dacA (Begg et al., 1995; Draper et al., 1998). It is conceivable that, in these ftsK mutants, the dacA gene product is not excluded from the division site and that degradation of the pentapeptide at mid-cell prevents the completion of division.
Interestingly, once growth and division are under way, PBP3 appears to be more concentrated in the central zone, the region of young cell wall (Fig. 6B and C). This young peptidoglycan presumably presents a greater concentration of pentapeptides to be trimmed, in order to mature into inert hemispheres. Then, the greater concentration of PBP3 in the young region of the cell surface may simply result from the affinity for its substrate.
HMW PBPs can form autonomous ring-like structures
A second unexpected conclusion stems from the non-co-localization of the rings of HMW PBPs and of Fts proteins in the dacA mutant. The HMW PBPs can exhibit a ring-like distribution independently from the FtsZ-ring scaffold or other conspicuous annular structures of the cell wall. Two possibilities can account for the formation of these rings of HMW PBPs. They may have an intrinsic propensity to assemble as rings surrounding the plasma membrane, following a nucleation event that can occur anywhere on the cell surface in the absence of PBP3, or only at the equator in its presence. Alternatively, rings of HMW PBPs may assemble on the FtsZ-ring scaffold, but later drift around the cell in the absence of PBP3 or be retained at the equator in its presence.
Three considerations suggest that, in the absence of PBP3, the rings of HMW PBPs can move around the cell. First, in the dacA mutant, the great thickness of the cell wall and the aborted septa may result from the activity of the PBP rings as they skim the cell surface. Secondly, some cells show multiple incomplete septa, whereas we observe single rings of PBP2x and PBP1a, indicating that the rings can move. Thirdly, the mild effect of the dacA mutation on the growth rate, and the great proportion of cells displaying non-co-localized rings of FtsZ and HMW PBPs (70%), suggest that the latter are viable and may divide if the various rings can drift and eventually adopt a relative orientation that is compatible with the division process (Fig. 7B).
How is the annular distribution of the PBPs maintained in the absence of known supporting structure? The simplest explanation is that membrane-bound PBPs polymerize around the cell. Dividing the perimeter of the cell by the size of the PBPs shows that about 500 molecules are required to surround the cell once. Small numbers of molecules per cell have been reported for the PBPs from E. coli and Staphylococcus aureus (Dougherty et al., 1996; Pucci and Dougherty, 2002). These numbers of molecules fall short of allowing a contiguous polymer of PBP around the cell. Moreover, no tendency to aggregate has been noticed for the various recombinant extracytoplasmic domains of class B PBPs that have been purified. Also, in the various crystal structures of PBP2x that were obtained, the crystallographic contacts differ, arguing against a favoured mechanism of autoassociation. The alternative explanation to the autoassociation of the PBPs is their assembly on a scaffold formed by other unidentified proteins or along a special feature of the peptidoglycan that is not visible to our methods of investigation. For example, the cell wall outgrowth that marks the equator of wild-type cells may persist in some form in the dacA mutant strain, although it is not visible on the available electron micrographs (Schuster et al., 1990).
FtsW does not form a stable complex with either FtsZ or the septal PBPs
The genetic and functional relationship between FtsW-like proteins and class B HMW PBPs led to the long-held, but still unproven, hypothesis that FtsW-like proteins are flippases, which transfer the peptidoglycan precursor from the cytoplasm to the periplasm for use by the PBPs (Matsuhashi et al., 1990). It was therefore of interest to check whether FtsW was associated with the septal PBPs.
Indeed, in wild-type cells, FtsW is co-localized with PBP2x, showing the same delay of constriction after that of the FtsZ-ring. This finding is consistent with a functional partnership between FtsW and septal PBPs. This observation also indicates that FtsW does not belong to a complex comprising FtsZ throughout the division process, although FtsW depends on FtsZ proteins for its septal localization in E. coli and B. subtilis (Buddelmeijer and Beckwith, 2002; Errington et al., 2003).
Against our expectations, in PBP3-deficient cells that are not engaged in division, FtsW was found to co-localize with FtsZ, but not always with PBP2x. Thus, FtsW does not form an obligatory complex with septal PBPs. However, in dacA mutant cells that are undergoing division, constriction of the FtsW-ring is indistinguishable from that of PBP2x. That is, constriction of the FtsW- and PBP2x-rings occurs later than constriction of the FtsZ-ring. Therefore, we must conclude that, unlike the HMW PBPs, the localization at mid-cell of FtsW is independent of PBP3, but that constriction of the FtsW-ring is somehow linked to the synthesis of the septal cell wall by the HMW PBPs (Fig. 7B).
The present work has revealed a hitherto hidden complexity in the relationships between the division proteins. Although numerous studies in E. coli and B. subtilis have determined the localization of the different division proteins in various genetic backgrounds, few works have investigated two or more proteins simultaneously. We think that such multiprotein investigations, carried out in different types of bacteria, will provide a more detailed picture of the fundamental processes of bacterial growth and division.
Pneumococcal strains and cultures
Streptococcus pneumoniae R6 is a non-encapsulated derivative of the Rockefeller University strain R36A. The PBP3 inactivated strain (Schuster et al., 1990) is an insertion duplication mutant generously provided by Professor Regine Hakenbeck (University of Kaiserslautern, Germany). Strains were grown anaerobically to an optical density of 0.4 at 600 nm at 37°C in glucose-buffered broth (Diagnostics Pasteur), before preparation for immunofluorescence microscopy.
Plasmids, proteins, antibodies and immunofluorescence microscopy
The sequence encoding a truncated form of PBP3 (Gly-15 to Lys-394) was amplified by polymerase chain reaction (PCR) using the primers 5′-CGGGATCCGAGAATCTTTATTTTCAG GGCGGGGGTGTTTCTACTGCTG-3′ and 5′-GGCTCGAGT CATTTTTCAATTTTCTTG-3′ and inserted as a BamHI–XhoI fragment into the plasmid pGEX-4T1 (Pharmacia) to generate pGEX-sPBP3*.
PBP3 was expressed in E. coli MC1061 as a fusion protein with the glutathione-S-transferase (GST) and a TEV protease cleavage site. Expression induced with of 0.5 mM IPTG was overnight at 16°C. After sonication in 50 mM Tris-HCl, pH 8.0, 200 mM NaCl, 1 mM EDTA, the lysate was loaded onto a 5 ml glutathione-Sepharose 4B affinity column (Pharmacia) equilibrated with the same buffer. PBP3 was eluted with 10 mM glutathione and dialysed against 50 mM Tris-HCl, pH 8.0, 50 mM NaCl, 1 mM EDTA. Cleavage of the GST was performed overnight at 37°C with 10 U of TEV protease (Invitrogen) mg−1 protein. After retention of the GST on a glutathione-Sepharose resin, PBP3 was further submitted to ion exchange chromatography on a ResourceQ resin (Pharmacia) equilibrated in the same buffer and eluted with a NaCl gradient. Homogeneity was finally achieved by gel filtration on a Superdex 200 HR (Pharmacia) equilibrated with the dialysis buffer.
We are very grateful to Professor R. Hakenbeck (University of Kaiserslautern) for the generous gift of the dacA mutant strain. We thank Dr I. Attree-Delic for providing access to fluorescence microscope. P. Meresse (Hybrisère) is acknowledged for the production of mouse anti-antisera. We thank A. M. Di Guilmi, E. Pagliero and L. Chesnel for providing purified high-molecular-weight PBPs. C.M. was supported by a CFR fellowship, and M.N.-S. by a postdoctoral fellowship, both from CEA.