The different shapes of cocci

Authors


  • Editor: Jacques Coyette

Correspondence: Thierry Vernet, Laboratoire d'Ingénierie des Macromolécules, Institut de Biologie Structurale, (CEA/CNRS/UJF UMR5075), 41 rue Jules Horowitz, 38027 Grenoble, France. Tel.: +33 0 4 38 78 96 81; fax: +33 0 4 38 78 54 94; e-mail: thierry.vernet@ibs.fr

Abstract

The shape of bacteria is determined by their cell wall and can be very diverse. Even among genera with the suffix ‘cocci’, which are the focus of this review, different shapes exist. While staphylococci or Neisseria cells, for example, are truly round-shaped, streptococci, lactococci or enterococci have an ovoid shape. Interestingly, there seems to be a correlation between the shape of an organism and its set of penicillin-binding proteins – the enzymes that assemble the peptidoglycan, the main constituent of the cell wall. While only one peptidoglycan biosynthesis machinery seems to exist in staphylococci, two of these machineries are proposed to function in ovoid-shaped bacteria, reinforcing the intrinsic differences regarding the morphogenesis of different classes of cocci. The present review aims to integrate older ultra-structural data with recent localization studies, in order to clarify the relation between the mechanisms of cell wall synthesis and the determination of cell shape in various cocci.

Introduction

Most studies of cell division and morphogenesis have been centered on the two rod-shaped laboratory workhorses: Escherichia coli and Bacillus subtilis, due mainly to the wide array of genetic tools available to probe the life and death of these organisms (Errington et al., 2003; Goehring & Beckwith, 2005). Caulobacter crescentus, a bacterium that undergoes a developmental cycle, has also emerged as a powerful model system to investigate morphogenesis (England & Gober, 2001; Briegel et al., 2006). However, much can be learned from comparative studies of morphologically diverse bacteria.

Historically, determining the morphology of bacterial cells has been an important phylogenetic tool. Yet, in the current era of molecular phylogenetics, comparative analysis of small subunit RNA sequences indicates that bacteria with different morphologies exist within single branches of phylogenetic trees and species with coccus morphology are present in clusters with a predominantly rod morphology (Siefert & Fox, 1998). Interestingly, once a particular lineage exhibits coccus morphology, clusters that result from that lineage become homogeneous for coccus morphology. This indicates that coccal morphology, which appears to have evolved multiple times during bacterial history, is evolutionarily irreversible (Siefert & Fox, 1998). The stability of coccal shape in evolution can be due to the inherent difficulty to regain genes for rod morphology, and/or to the absence of selective pressure for rod shape. Genetically, it is also easy to convert a rod into a coccus, for example by loss (rodA, pbpA) or overexpression (bolA) of a gene (Aldea et al., 1988; Murray et al., 1997; Henriques et al., 1998), but there is no report of genetic alterations that convert coccal cells into stable rod-shaped cells.

This review will address the question of the origin of cocci, not from an evolutionary point of view, but from a morphogenetic point of view, i.e. what determines the shape of a coccus cell?

Before answering this question, it is useful to distinguish between two classes of cocci: (1) organisms with truly round cells such as pediococci, micrococci, deinococci, staphylococci or Neisseria (except Neisseria elongata), which usually divide in either two or three alternating perpendicular planes during consecutive division cycles, leading to arrangements in tetrads or in three-dimensional cuboidal packets of eight cells, respectively (Fig. 1a and b); (2) organisms whose cells are elongated ellipsoids, such as enterococci, streptococci or lactococci (and some other genera such as Lactovum, Leuconostoc, Weissella and Oenococcus, Melissococcus, Vagococcus), which divide in successive parallel planes, perpendicular to their long axis (Fig. 1c). These can be observed as isolated cells, diplococci or small chains, depending on the degree of cell separation.

Figure 1.

 Bacterial division over successive division cycles in (a) two perpendicular planes (e.g. Neisseria or pediococci), (b) three perpendicular planes (e.g. staphylococci) or (c) one parallel plane (e.g. enterococci or streptococci).

To the authors' knowledge, there is no widely accepted term to distinguish ellipsoid bacteria from spherical cocci and therefore the term ovococcus is proposed to designate these bacteria with ovoid shape. Note that this designation is only morphological, as common shape is not strictly correlated with evolutionary relationship. In this review, mainly staphylococci will be considered as a model for spherical cocci and enterococci and streptococci as models for ovococci.

Lack of cylindrical elongation in cocci

Bacterial morphology was traditionally assumed to be determined by the cell wall, which acts as an external scaffolding or exoskeleton. This was based on the facts that isolated cell wall sacculi retained the specific shape of a particular cell (Höltje, 1998) and that most mutations originally described as affecting cell shape were located in genes involved in cell wall synthesis (Spratt, 1975; Tamaki et al., 1980; Honeyman & Stewart, 1989; Henriques et al., 1998). Recently, an internal helical scaffolding or cytoskeleton was identified in rod-shaped bacteria, which is composed of actin homologues – encoded by mreB and mreB-like genes – that have a direct role in cell shape determination. These internal scaffolds, rather than behaving like rigid skeletons, may determine bacterial shape by directing peptidoglycan-synthesizing machineries involved in elongation of the cell, which in turn construct the cell's peptidoglycan exoskeleton (Daniel & Errington, 2003; Dye et al., 2005; Carballido-Lopez, 2006), although this is still debated (Tiyanont et al., 2006) and although the key effectors of cell wall biosynthesis controlled by MreBs have not yet been identified.

A second cytoskeleton element is composed of the tubulin homologue FtsZ, which polymerizes as a ring at the division site and is widely distributed in bacteria (Errington et al., 2003; Goehring & Beckwith, 2005). The recruitment of other division proteins in Escherichia coli and B. subtilis is subordinated to the presence of FtsZ at the division site. FtsZ has therefore been considered to be the prime organizer of the division process, onto which other components are assembled to form the so-called divisome (Buddelmeijer & Beckwith, 2002; Errington et al., 2003). Among the later division proteins that are recruited at the division site is FtsI, one of the penicillin-binding proteins (PBPs). These enzymes are responsible for the assembly of peptidoglycan, the main component of the bacterial cell wall (Sauvage et al., 2008). FtsZ-directed cell wall synthesis at the division site of rod-shaped bacteria results in the formation of the division septum, which, after cell division is completed, is converted into the new pole of each of the two daughter cells.

The presence in rods of the two families of cytoskeleton proteins, MreB and FtsZ, is therefore associated with the two main phases of cell wall growth: the elongation and the division, respectively. MreB proteins are not encoded in most genomes of cocci. The few exceptions include some cyanobacteria and the Chlamydia, which have a spherical shape but have the ability to undergo morphological differentiation during their life cycle (Carballido-Lopez, 2006). Thus, the truism of this section's title, the absence of cylindrical side-wall in cocci, may find its molecular basis in the absence of an actin-like cytoskeleton, which precludes a true elongation. In cocci, FtsZ-dependent cell wall synthesis is therefore predominant and determinant for morphogenesis, where it can account for the synthesis of the entire new hemisphere of each daughter cell.

FtsZ-dependent cell wall synthesis at the division site can be observed in Staphylococcus aureus, by visualization of peptidoglycan synthesis using fluorescent vancomycin, under conditions where it binds only to pentapeptides that are present on the new peptidoglycan or its precursor (Pinho & Errington, 2003). This method has shown that staphylococcal cell wall synthesis occurs mainly, if not exclusively, at the division site. However, if FtsZ is depleted from Staphylococcus aureus cells, the septum no longer forms and cell wall synthesis becomes delocalized over the entire surface of the cell, allowing it to enlarge up to eight times its normal volume, before lysing (Pinho & Errington, 2003).

Ovococci also synthesize the cell wall mostly at the division site, with the new hemispheres of the daughter cells being synthesized between the two parting old hemispheres and, accordingly, labeling of Streptococcus pneumoniae with fluorescent vancomycin shows that peptidoglycan synthesis occurs mostly at mid cell (Daniel & Errington, 2003; Ng et al., 2004), confirming the previous pulse chase experiments that did not directly monitor peptidoglycan incorporation, but rather the incorporation of additional components of the cell wall, such as protein antigens or teichoic acids, which are attached to the peptidoglycan (Cole & Hahn, 1962; Tomasz et al., 1975). FtsZ-depletion studies have not been carried out in ovococci. However, colocalization of all high-molecular-weight (HMW) PBPs with FtsZ at the onset of division is consistent with an FtsZ-triggered cell wall synthesis. Also, zantrins, which are small molecules that perturb FtsZ function, inhibit division in Streptococcus pneumoniae, resulting in cell enlargement (Margalit et al., 2004).

One or two types of division-specific cell wall synthesis in cocci

Previous ultrastructural studies using Staphylococcus aureus, carried out primarily by Giesbrecht et al. (1998), have shown that the synthesis of the septum proceeds by centripetal growth resembling a closing iris, until the inner edge of the growing septum eventually fuses in the center of the cell. After the septum or cross-wall is complete, the two daughter cells are still held together within a single sphere, whose cell wall is a homogeneous-looking structure, with no signs of invagination (Fig. 2a). Only at the end of cell division is the septum split into two surfaces that become the new hemispheres of each of the daughter cells (Giesbrecht et al., 1998). Recently, cryo-electron microscopy of frozen-hydrated thin sections of Staphylococcus aureus revealed a differentiated cell wall at the septum with two zones of high density, which correspond to two adjacent cross-walls, located between two low-density zones, and separated by a third zone of low-density (Matias & Beveridge, 2007). The two low density zones adjacent to the cell membrane appear as an extension of the periplasmic space described recently in a number of gram-positive bacteria, including the cocci Enterococcus gallinarum, Streptococcus gordonii and Staphylococcus aureus (Matias & Beveridge, 2006; Zuber et al., 2006). The low-density region between the two cross-walls could correspond to a highly fragile septal region that would facilitate or result from the action of the autolysins responsible for splitting of the septum (Matias & Beveridge, 2007). After splitting, some remodeling of the peptidoglycan may occur when the flat septum becomes spherical, but this alteration of conformation could result merely from exposure of the new hemisphere to high internal osmotic pressure, immediately after cell separation.

Figure 2.

 Electron micrographs of cocci. (a) Thin section of Staphylococcus aureus spherical cells that have completed septum formation. (b) Longitudinal thin section of dividing ovococcus Streptococcus pneumoniae R6.

When a second and third round of division occur in Staphylococcus aureus, the septum is synthesized in consecutive perpendicular planes, similar to the mode of division of Sarcina or some species of cyanobacteria Synechocystis (Giesbrecht et al., 1998) (Fig. 1b). This was initially inferred from scanning electron microscopy results, which showed regular cuboidal packets of eight Staphylococcus aureus cells most likely resulting from three divisions along orthogonal planes (Tzagoloff & Novick, 1977). However, under the light microscope, Staphylococcus aureus appears as clusters of cells without an obvious geometric arrangement. This is probably due to the activity of lytic enzymes responsible for the splitting of the division septum that seem to cause a postfissional movement of the cells, leading to the formation of irregular clusters (Koyama et al., 1977). Despite the fact that these observations were made three decades ago, the mechanism that determines the precise placement of the septum in alternating perpendicular planes is far from understood.

A detailed description of the growth and division process of ovococci was provided in an impressive series of ultrastructural papers from the 1970s by Gerry Shockman, Michael Higgins and colleagues on the morphogenesis of Enterococcus hirae ATCC 9790, as well as early work on Streptococcus pneumoniae by Alexander Tomasz. These landmark studies, based mostly on electron microscopy of negatively stained thin sections, showed that ovococcal cells are surrounded in their middle by an annular outgrowth of peptidoglycan, often referred to as the equatorial ring (Tomasz et al., 1964; Higgins & Shockman, 1970). The first cell wall-related event in cell division is the appearance of a small ingrowth below the equatorial ring. The equatorial ring is then split, more or less in its middle, and the two resulting rings separate while a new peripheral cell wall appears in between. The small annular ingrowth remains equidistant from the two new parting equatorial rings, and at some point starts to grow centripetally to form a septal disc. The septal disc undergoes gradual and simultaneous closure at its center and splitting at its periphery (Fig. 2b).

The first model for the growth of ovococci envisioned a single site of centripetal septal synthesis, with gradual splitting of the septum feeding the peripheral growth (Higgins & Shockman, 1970). However, careful measurements of the septal and peripheral wall surface area showed that the latter was growing faster than allowed by septal splitting alone. A second model was proposed, which included a second site of peptidoglycan synthesis (Higgins & Shockman, 1976). This peripheral synthesis was supposed to occur diffusely along a gradient, with the greatest activity near the site of septal splitting.

The idea of ovococci having two types of division-specific cell wall synthesis, and spherical cocci only one, was later supported by extensive mutagenesis and drug treatment studies by Satta et al. (Higgins et al., 1974; Gibson et al., 1983; Lleo et al., 1990). They found that temperature-sensitive mutants, which grow longer and fail to septate at a nonpermissive temperature, could be isolated from most ovococcal species investigated. These species underwent the same morphological changes upon treatment with antibiotics supposed to block only septation. In contrast, no longitudinal growth was obtained from truly spherical cocci such as staphylococci or certain Neisseria species, either by mutation or with drugs (Higgins et al., 1974; Gibson et al., 1983; Lleo et al., 1990). In Enterococcus hirae, treatment with penicillin at division-inhibiting concentrations led to the formation of long filaments under some growth conditions of temperature and medium (Fontana et al., 1983). The concept of two competing sites of peptidoglycan synthesis in ovococci was then introduced, based on the observation that suppression of septation apparently led to unchecked longitudinal expansion (Higgins et al., 1974; Gibson et al., 1983; Lleo et al., 1990).

Interestingly, Streptococcus mutans displays an odd behavior regarding the morphology of some strains as its length to width ratio can span values from 1 to 5 due to variations of the ratio of K+/bicarbonate in the growth medium (Tao et al., 1988). As pointed out by the authors, the morphological variations exhibited by some strains of Streptococcus mutans would be consistent with the two-site model of cell wall synthesis, where the relative activity of the peripheral and septal synthesis, and/or the timing of septation, would be influenced by the ratio of some ions (Tao et al., 1993).

Following the terms introduced by Higgins & Shockman (1976), these types of synthesis shall be called as septal for the formation of the cross-wall, perpendicular to the main axis of the cell, and peripheral for the longitudinal component. Spherical cocci have only septal synthesis. Note that both septal and peripheral cell wall synthesis are considered as division specific, and that peripheral synthesis is distinct from elongation of rod-shaped bacteria.

However, besides septal and elongation modes of cell wall synthesis, rods also seem to have an additional mode of lateral cell wall synthesis, less well characterized, which occurs in the periphery of the cell, near mid cell, but is dependent on FtsZ. Evidence comes from the observation of areas of high, FtsZ-dependent, peptidoglycan synthesis at the division site of Escherichia coli, even when septal synthesis is specifically suppressed, for example by inactivation of the cell division-specific transpeptidase (Wientjes & Nanninga, 1989; de Pedro et al., 1997). The existence of a significant time gap between FtsZ ring formation and assembly of the septal peptidoglycan synthesis machinery in Escherichia coli raised the possibility that preseptal peptidoglycan synthesis could occur during that period (Aarsman et al., 2005). Furthermore, PBPs known to participate in the elongation of Escherichia coli and B. subtilis were found at the division site of these rod-shaped bacteria (Den Blaauwen et al., 2003; Scheffers et al., 2004). In Escherichia coli, the preseptal synthesis was found to be ‘penicillin’-insensitive. Indeed, a number of tested β-lactams did not inhibit this FtsZ-dependent synthesis, including drugs specific of class A PBPs or of the elongation-specific PBP2. These observations suggested that a monofunctional glycosyltransferase of Escherichia coli may participate in the preseptal synthesis (Aarsman et al., 2005). Alternatively, not all the HMW PBPs of Escherichia coli may have been inhibited with the tested drugs.

In C. crescentus, this mode of cell wall synthesis at the mid cell is more evident and was shown to be dependent on FtsZ but independent of MreB and to occur before constriction of the cell takes place (Aaron et al., 2007). Thus, it is speculated that the division of rods includes the same two types of synthesis as ovococci: peripheral and septal, both FtsZ dependent. In addition to these division-specific modes of cell wall assembly, rods have MreB, dependent elongation, which is absent from cocci.

Similar sets of PBPs for similar shapes – one or two machineries for peptidoglycan assembly

PBPs are the enzymes responsible for the synthesis of long chains of tandemly repeated disaccharide units that make up the glycan strands of the peptidoglycan, as well as for their cross-linking via peptide bridges. They can be classified as HMW class A PBPs, which include bifunctional proteins that have both glycosyl transferase activity (for the synthesis of the glycan strands) and transpeptidase activity (for the cross-linking of the peptidoglycan); HMW class B PBPs, which include proteins that have a N-terminal domain with unknown function and a C-terminal transpeptidase domain; and low molecular weight (LMW) PBPs, which usually have carboxypeptidase or endopeptidase activity (Goffin & Ghuysen, 1998). β-Lactam antibiotics such as penicillin inhibit the transpeptidase, carboxypeptidase or endopeptidase activities.

The number of PBPs varies considerably among bacteria, with the rod-shaped model organisms Escherichia coli and B. subtilis having 12 and 16 PBPs, respectively, while cocci tend to have a lower number, usually from four to seven PBPs. These numbers were initially determined by incubating cells or membranes with radio-labeled penicillin before gel electrophoresis. This procedure often revealed band patterns that are difficult to correlate with known pbp genes present in the sequenced genomes. Although the presence of additional PBPs in some strains cannot be ruled out, the detection of unexpected bands can most easily be explained by various proteolytic cleavages of known PBPs (Coyette et al., 1980). The fact that many PBPs have redundant functions in an organism has made it difficult to assign a specific function for each PBP.

It has been suggested that PBPs work in multi-enzymatic complexes that would include not only peptidoglycan synthetic enzymes but also peptidoglycan-degrading or lytic enzymes, in order to co-ordinate both processes, preserving the integrity of the cell, and that different machineries would be responsible for each mode of peptidoglycan synthesis (Höltje, 1996, 1998).

Staphylococcus aureus cells have only four PBPs. The two essential staphylococcal PBPs (HMW class B PBP1 and HMW class A PBP2, the only bifunctional PBP in Staphylococcus aureus) localize at the division site (Fig. 3a) (Pinho & Errington, 2005; Pereira et al., 2007). PBP4, a LMW PBP, also seems to localize at that place (P.M. Pereira and M.G. Pinho, unpublished data). The localization of PBP3 (HMW class B) has not yet been defined, but it seems unlikely that it would catalyze de novo cell wall synthesis around the entire periphery of the cell without the assistance of a HMW class A PBP. Furthermore, a PBP3 mutant has no major observable morphological defects (Pinho et al., 2000). The small number of PBPs and their septal localization is in accordance with the fact that Staphylococcus aureus synthesizes the cell wall at the division site using one septal synthetic machinery (Fig. 4a), although it cannot be excluded that another type of synthesis involving the second, nonessential, HMW class B PBP3 may exist. Methicillin-resistant Staphylococcus aureus (MRSA) strains have acquired an additional PBP – PBP2A – which has a transpeptidase domain with a very low affinity for β-lactam antibiotics and, together with the transglycosylase domain of PBP2, catalyzes cell wall synthesis in the presence of a high concentration of antibiotics (Pinho et al., 2001). The localization of this protein as well as its activity in the absence of antibiotics remains unknown.

Figure 3.

 Localization of PBPs in cocci. (a) Septal localization of PBP1 (localized by immunofluorescence) and PBP2 (localized using an N-terminal green fluorescent protein fusion), the two essential PBPs of Staphylococcus aureus. (b) Immunofluorescence microscopy of dividing Streptococcus pneumoniae cells. PBP2b was revealed with mouse antiserum and Cy3-coupled goat anti-mouse antibodies. PBP2x was revealed with rabbit antiserum and Cy2-coupled goat anti-rabbit antibodies. DNA was diamidino-2-phenylindole stained. Images are raw. The scale bar is 1 μm.

Figure 4.

 Proposed models for cell wall assembly in spherical cocci (a) and ovococci (b). The parental cell wall is in black; the new cell wall of peripheral and septal origin is in dark or light gray, respectively. Peripheral and septal peptidoglycan synthesis machineries are in red and green, respectively. Black contours indicate active machineries. The splitting machinery is in cyan, with a jaw to indicate activity. In pathway I (b), both machineries are localized at the leading edge of the septum, but are functioning successively, while in pathway II, the peripheral machinery remains at the periphery of the septal disc and contributes with material to the inner face of the cell wall.

PBPs from other spherical cocci have not been localized, but the fact that Neisseria meningitidis or Neisseria gonorrhoeae have only four described PBPs (Nolan & Hildebrandt, 1979; Barbour, 1981; Stefanova et al., 2004), which include only one representative of HMW class A (PBP1) and one of HMW class B (PBP2) PBPs, may indicate that these species also have only one cell wall synthetic machinery.

As mentioned earlier and in contrast to spherical cocci, ovococci have two types of division-specific cell wall synthesis, which are necessary for their specific ovoid shape. To achieve this ovoid shape, a common set of PBPs may be required (Table 1), which includes three class A (PBP1a, 1b and 2a) and two class B (PBP2b and 2x) PBPs. A LMW PBP (PBP3) has carboxypeptidase activity that trims the free end of peptidoglycan pentapeptides, which then become of limited use for the transpeptidation reaction, as they can only serve as acceptors.

Table 1.   Presence of PBPs in ovococci with fully sequenced genome
 Class A (bifunctional)ClassB (monofunctional)dd-Carboxypeptidase
PBP1aPBP1bPBP2aPBP2bPBP2xPBP5PBP3
  1. Locus names of the various sequencing projects are given as well as alternative names found in the literature or databases. Genomes were searched by nucleotide and protein blast.

Enterococcus faecalisV583EF1148EF1740EF0680EF2857PBPC EF0991PBP4 EF2476dacA EF3129  
Enterococcus faecium DOponA EfaeDRAFT_0179pbpZ EfaeDRAFT_0375pbpF EfaeDRAFT_2538pbpA EfaeDRAFT_1462pbpB EfaeDRAFT_2224pbp5 EfaeDRAFT_0894Contig588  
Lactococcus lactis IL1403ponA LL0543LL0386LL2108LL0339pbpX LL0867 dacA LL2263  
Streptococcus pneumoniae TIGR4ponA SP0369SP2099SP2010penA SP1673pbpX SP0336 SP0872  
Streptococcus pneumoniae R6spr0329spr1909spr1823penA spr1517pbpX spr0304 spr0776  
Streptococcus thermophilus CNRZ1066str0230str1869str0212str0613str1701 dacA1 str0080  
Streptococcus thermophilus LMG18311stu0230stu1869stu0212stu0613stu1701 dacA1 stu0080  
Streptococcus mutans UA159SMU.467SMU.1991SMU.1949SMU.597SMU.455 SMU.253  
Streptococcus agalactiae NEM316gbs0288gbs0155gbs2020gbs0785gbs0277 gbs0143gbs0142 
Streptococcus agalactiae 2603V/RSAG0298SAG0159SAG2066SAG0765pbpX SAG0287 SAG0147SAG0146 
Streptococcus pyogenes SF370Spy1649Spy0097Spy2059 pbpX ftsI Spy1664 dacA Spy0292 Spy1093
Streptococcus pyogenes MGAS315SpyM3_1390SpyM3_0074SpyM3_1758 SpyM3_1401 dacA SpyM3_0214SpyM3_0213SpyM3_1390
Streptococcus pyogenes MGAS10394M6_Spy1401M6_Spy0130M6_Spy1753 M6_Spy1413 M6_Spy0279M6_Spy0278M6_Spy0815
Streptococcus pyogenes MGAS8232SpyM18_1661SpyM18_0098SpyM18_2120 SpyM18_1674 dacA SpyM18_0280SpyM18_0279SpyM18_1051
Streptococcus pyogenes MGAS6180M28_Spy1396M28_Spy0080M28_Spy1739 M28_Spy1408 dacA2 M28_Spy0243dacA1 M28_Spy0242M28_Spy0794

Enterococci possess a supplementary class B PBP (PBP5), with a low affinity for penicillins, which confers an intrinsic resistance to these antibiotics. However, in the absence of penicillins, this PBP can be deleted without notable consequences on the peptidoglycan composition or cell morphology (Sifaoui et al., 2001). A true exception to the uniform number of basic PBPs in ovococci is the absence of PBP2b in Streptococcus pyogenes, which will be discussed below. Others are the presence in Streptococcus pyogenes or Streptococcus agalactiae of one or two additional genes coding for uncharacterized LMW PBPs.

The common set of PBPs present in ovococci may be organized in two cell wall synthetic machineries, which would catalyze the two modes of division-specific cell wall synthesis. Based on data from immunofluorescence microscopy of Streptococcus pneumoniae, an initial model was proposed, in which two cell wall synthesis machineries would have different localizations (Morlot et al., 2003), but later proved to be incorrect, as shown below. New data obtained with better batches of antisera against some PBPs have revealed that all the PBPs have the same localization at mid cell, at the resolution of the immunofluorescence microscopy, throughout the cell cycle (Fig. 3b). Is it possible to reconcile these new data showing that all the PBPs are colocalized, with the existence of two modes of synthesis, implied by the ultrastructural, mutational and drug treatment studies (Higgins & Shockman, 1976; Lleo et al., 1990)? Here, a revised model is proposed that accommodates two machineries of peptidoglycan synthesis with a single localization (Fig. 4b). The machinery responsible for the peripheral component of cell wall synthesis would start to add material at mid cell, on the inner face of the cell wall, with the concomittant splitting of the new wall, resulting in the observed initial phase of longitudinal growth. A second phase would start somewhat later, with the activity of a second machinery, to build the septal cross-wall. Septal assembly of the peptidoglycan should occur faster than septal splitting to produce the observed cross-wall. The two types of synthesis could occur strictly successively, or partially simultaneously during the septal phase. After closure of the septum, splitting goes on to separate the daughter cells. Both machineries could be located at the leading edge of the closing septum, but activated at different points of the division process. Alternatively, the machinery for peripheral synthesis might not follow the leading edge of the septum, but remain at its periphery. Thus, once septum formation starts, the peripheral machinery might continue adding material on the inner face of the cell wall. This model, which is essentially that proposed thirty years ago by Higgins and Shockman, can account for the measured thickening of the cell wall near the splitting site (Higgins & Shockman, 1976; Lleo et al., 1990).

Other proteins involved in cell wall synthesis of cocci

In addition to PBPs, the synthesis of peptidoglycan is likely to require the presence of shape, elongation, division and sporulation (SEDS) proteins (Henriques et al., 1998). SEDS proteins are integral membrane proteins, with 10 membrane-spanning segments (Gerard et al., 2002), which presumably participate in peptidoglycan synthesis, as implied by the frequent organization in operons of the genes for SEDS proteins and class B PBPs, and the identical phenotypes resulting from deletion of class B PBPs or their cognate SEDS protein (de Pedro et al., 2001; Thibessard et al., 2002). Although the biochemical function of these proteins remains unknown, it has been proposed (but never experimentally shown) that SEDS proteins might be the flippases required to translocate the peptidoglycan precursors from the cytoplasm, through the plasma membrane, to the extracellular site (Ehlert & Holtje, 1996; Ghuysen & Goffin, 1999). The phenotypes of FtsW mutants and RodA depletion in rod-shaped bacteria indicate that FtsW is the SEDS protein involved in division, whereas RodA is required for elongation (e.g. Ishino et al., 1989; Boyle et al., 1997; Henriques et al., 1998).

Interestingly, Staphylococcus aureus has two SEDS proteins, despite apparently having only septal cell wall synthesis. The prediction is that FtsW would work with the essential class B PBP1 within the main septal machinery of cell wall synthesis. Correspondingly, staphylococcal RodA would play a minor role, working with the nonessential class B PBP3. Neisseria lack RodA, in accordance with the presence of a single HMW class B PBP.

Ovococci also possess two SEDS proteins. In Streptococcus pneumoniae, FtsW was found to have the same localization at mid cell as the PBPs (Morlot et al., 2004a). The pbp2x gene is downstream of ftsL as is ftsI in most genomes (Massidda et al., 1998), indicating that PBP2x is likely the class B PBP participating in the septal machinery with FtsW (Fig. 5). To further support this identification, PBP2x and PBP2b from Streptococcus pneumoniae show better similarity to PBP3 (FtsI) and PBP2 in Escherichia coli, respectively. PBP2b and RodA would therefore be part of the peripheral machinery (Fig. 5).

Figure 5.

 Established and putative proteins participating in the peripheral and septal peptidoglycan synthesis machineries of ovococci. The identity of the bifunctional class A PBPs participating in each phase remains unknown.

Streptococcus pyogenes, which lacks both PBP2b and RodA, often appears less elongated and more spherical than other ovococci. It is conceivable that it performs only septal synthesis, giving rise to cells lacking the longitudinal component provided by the peripheral synthesis. In support of this theory is the observation that opaque colony variants of Streptococcus pyogenes, which apparently fail to split their cross-wall, form stacks of flattened cells with very little non-septal wall (Swanson & McCarty, 1969). Mutants of Streptococcus thermophilus defective in RodA or PBP2b are also more spherical than the parental strain, with a longitudinal cellular length about half that of the wild type (Stingele & Mollet, 1996; Thibessard et al., 2002). In Streptococcus gordonii, the deletion of PBP2b results in more complex morphological defects with aberrant septa in some cells and an increased susceptibility to lysis (Haenni et al., 2006).

Besides PBPs and SEDS proteins, other proteins may contribute to cell wall construction. Among division proteins, DivIB(FtsQ), FtsL and DivIC(FtsB) are bitopic membrane proteins, with their major domains being extracellular (Fig. 5) (Errington et al., 2003; Goehring & Beckwith, 2005). The gene encoding DivIB is often in an operon with genes involved in the synthesis of peptidoglycan precursors. The gene coding for FtsL is often adjacent to that encoding the septal class B PBP. The three genes are completely absent from wall-less bacteria (Margolin, 2000). These arguments suggest that the unknown functions of DivIB, FtsL and DivIC are related to cell wall formation. FtsL and DivIC are small proteins. Sequence-based predictions indicate that the extracellular domains adopt essentially a coiled-coil conformation. The extracellular region of DivIB appears to consist of two autonomously folding domains (Robson & King, 2006). A complex of the extracellular domains of the three proteins from Streptococcus pneumoniae was reconstituted in vitro, provided that heterodimerization of the extracellular domains of FtsL and DivIC was artificially constrained, presumably to compensate for the deletion of the transmembrane segment (Noirclerc-Savoye et al., 2005). A similar complex is formed in Escherichia coli (Buddelmeijer & Beckwith, 2004). In vivo, the three proteins appeared to colocalize at the division site during septation.

What could be the function of DivIB, FtsL and DivIC? Assuming that they function in cell wall assembly, this role may be restricted to the division, as no paralogues exist for the elongation synthesis of rod-shaped bacteria. Therefore, the function of these three proteins may be found in features that distinguish division- and elongation-specific peptidoglycan synthesis. Two such characteristics come to mind. Firstly, the elongation of bacilli produces a cylinder; the amount of cell wall generated at different stages of the formation of a cylinder is constant. In contrast, the amount of synthesis required during division diminishes progressively until extinction, as the septal hole closes. DivIB, FtsL and DivIC might intervene in the regulation of this diminishing synthesis. A second difference between the septal and the lateral cell wall is that the lateral wall is subjected to osmotic pressure, whereas the septal cross-wall is not. In the absence of osmotic pressure, septal synthesis might require an additional scaffolding function, which could be provided by DivIB, FtsL and DivIC.

MreC and MreD may also be key proteins involved in cell wall synthesis that have not been well studied in cocci yet. MreC, a bitopic membrane protein with a major extracellular domain, and MreD, an integral membrane protein (Fig. 5), are both involved in the elongation phase of rod-shaped bacterial morphogenesis, by interacting with Mbl and MreB in B. subtilis, and MreB in Escherichia coli, which lacks Mbl (reviewed by Stewart, 2005). Mbl is a cytoskeleton actin-like protein from the MreB family. Through the MreB/Mbl connection, MreC and MreD are somehow involved in the control of peptidoglycan synthesis (Leaver & Errington, 2005). As both MreB and Mbl are missing in cocci, it will be of interest to find out whether MreC and MreD have a role in peptidoglycan synthesis by themselves. In ovococci, it is suggested that MreC and MreD might take part in peripheral synthesis (Fig. 5). Deletion of MreD in Streptococcus thermophilus produced smaller cells (Thibessard et al., 2004). Note that MreC and MreD are absent in Streptococcus pyogenes and Streptococcus agalactiae. Interestingly, these two species have additional LMW PBPs. This correlation may be coincidental or may indicate that two different solutions have evolved to regulate some aspects of murein metabolism in ovococci.

How do cell wall synthesis machineries localize? Co-ordinating cell wall synthesis and the cell cycle

To produce cells of proper shape and size, the activity and localization of the cell wall synthesis and degradation machineries must be correctly regulated. Most importantly, septal cell wall formation during division must be co-ordinated with the other processes of cell division, such as membrane invagination, DNA replication or chromosome segregation, to avoid catastrophic breakage of the chromosome by division through the nucleoid, and must occur exactly at the middle of the cell to ensure the generation of two identical daughter cells.

Early investigations of Enterococcus hirae by Higgins's group showed that initiation of division, defined as the duplication of the equatorial ring, occurs at constant cell volume, independently of the growth rate (Gibson et al., 1983). Thus, at a fast growth rate, a second round of division can start, before closure of the former septum. Using inhibitors of DNA synthesis and cell shape analysis, initiation of cell wall formation (peripheral and septal) was shown to be independent of the end of chromosome replication, but closure of the septum was not (Higgins et al., 1974; Gibson et al., 1983; Lleo et al., 1990). How cells sense their volume and signal the start of division is entirely unknown.

However it is known that the initiation of division requires the formation of a properly localized FtsZ ring. Rod-shaped bacteria have two main systems to ensure that the FtsZ ring assembles at mid cell: the nucleoid occlusion effect and the Min system.

The nucleoid occlusion effect, first postulated by Woldringh et al. (1990), relies on the ability of the nucleoid, the bacterial equivalent of the nucleus, to prevent division in its vicinity. Recently Noc, a specific effector of nucleoid occlusion, was identified in B. subtilis (Wu & Errington, 2004), and a functional analogue, SlmA, was identified in Escherichia coli (Bernhardt & de Boer, 2005). The Noc protein is a nonspecific DNA-binding protein that prevents the division machinery from assembling in the vicinity of the nucleoid. The Min system acts to prevent the formation of septa near the cell poles (for a recent review see Lutkenhaus, 2007). MinC and MinD proteins form a complex that acts as an antagonist of the FtsZ ring assembly. In B. subtilis, the activity of MinCD is restricted to the vicinity of the poles by the topological specificity factor DivIVA, which directly recruits the inhibitor to that place. In Escherichia coli, MinCD oscillate between the poles, giving rise to a gradient with the lowest concentration at the middle of the cell, and MinE is the topological marker that restricts the activity of MinCD at the middle of the cell. The net effect of both systems is that polymerization of FtsZ in rod-shaped bacteria occurs exclusively at the only place that results in the division of a cylinder with a constant diameter in two equal daughter cells, i.e., the middle of the cell, and it occurs when this place has already been cleared from the nucleoid.

Min proteins are present in various species of cocci such as N. gonorrhoeae, N. meningitidis, Synechocystis sp. or Deinococcus radiodurans but are missing from other cocci, such as Staphylococcus, Enterococcus or Streptococcus. So how do these cocci find their middle? Contrary to rod-shaped bacteria, spherical cocci have not one, but an infinite number of potential cell division planes that can give rise to two equal daughter cells, as the middle of the cell can be defined by any circumference with maximum diameter. Therefore, although the existence of a protein-based system for the selection of the division site in cocci cannot be ruled out, spherical cells may not need a Min-like mechanism to determine the future division site. The zone of maximum diameter is likely to be the site where a circular polymer lying against the cell membrane, such as the FtsZ ring, would be most stable. However, from the infinite number of division planes that divide a coccus in to two identical cells, only a few will not overlap the segregating chromosomes. Round cocci divide into two (e.g. Neisseria) or three (e.g. Staphylococcus) perpendicular division planes. Therefore, the chromosome also has to segregate along two or three perpendicular axes that are perpendicular to the septal planes. The mechanism that determines the choice of axis for chromosome segregation and plane of division remains totally unknown.

Ovococci always divide in the same plane, perpendicular to the longer axis of the cell. However, they also miss the Min system and nucleoid occlusion does not seem to operate in these bacteria. Indeed, the equator, where most division proteins are found at the start of division in Streptococcus pneumoniae, is located precisely around the nucleoid (Morlot et al., 2003, 2004a, b). In these organisms, equatorial ring duplication ensures that the division site is always marked. This equatorial ring marks the cellular site with the largest circular perimeter, which, as mentioned above, may be the most stable place for the FtsZ ring, therefore possibly circumventing the requirement for a Min-like system for determination of the division site. Interestingly, in Streptococcus pneumoniae, the division proteins are always localized, mostly at mid cell and sometimes at the poles (Morlot et al., 2003, 2004a, b). This is in marked contrast with the behavior of division proteins in rods, as in Escherichia coli and B. subtilis the division site is not marked during the elongation phase but instead requires the complex nucleoid occlusion and Min systems to be defined at the time of division.

It is worth mentioning that although staphylococci and streptococci lack MinCD, they have DivIVA. In B. subtilis, DivIVA appears to be involved not only in the positioning of the division site through a functional interaction with the MinCD system but also in chromosome segregation during sporulation (Marston & Errington, 1999; Wu & Errington, 2003).

In Streptococcus pneumoniae, a DivIVA deletion mutant shows division and morphological defects, with aberrant shapes and incomplete septa, as well as some cells devoid of nucleoid, indicating problems in chromosome segregation (Fadda et al., 2003). A similar phenotype was observed in Enterococcus faecalis following disruption of divIVA (Ramirez-Arcos et al., 2005). In contrast, in Staphylococcus aureus, deletion of DivIVA resulted in no distinctive phenotype (Pinho & Errington, 2004). In Streptococcus pneumoniae, DivIVA is found at the division site and the poles, whereas in Staphylococcus aureus, DivIVA is mainly at the septum (Pinho & Errington, 2004; Fadda et al., 2007). Results from two-hybrid experiments in Escherichia coli indicate that pneumococcal DivIVA interacts with many proteins of the division machinery (Pinho & Errington, 2004; Fadda et al., 2007), but it is still unclear whether the morphological defects resulting from the deletion of DivIVA are due to a direct role of DivIVA in the control of cell wall synthesis, or due to a more general function in cell division.

Another feature that may mark the division site is a local difference in the lipid composition of the cell membrane. In Escherichia coli and B. subtilis, cardiolipin was found to be concentrated at the division site (Mileykovskaya & Dowhan, 2000; Koppelman et al., 2001; Kawai et al., 2004). The same is true of phosphatidylethanolamine and several lipid synthesis enzymes in B. subtilis (Nishibori et al., 2005). Interestingly, the drug cerulenin, which inhibits acyl carrier protein synthases (enzymes that mature proteins involved in fatty acid synthesis), appears to block division and giving rise to elongated cells in Enterococcus hirae (Higgins et al., 1980), indicating that membrane lipid composition may also be important for cell division in this organism.

Proper localization of the FtsZ ring is followed by the recruitment of other division proteins at mid cell, including the division-specific PBPs (Errington et al., 2003; Goehring & Beckwith, 2005). PBPs were initially thought to localize by direct or indirect protein–protein interaction with the FtsZ ring. However, substrate recognition may also operate in recruiting PBPs to the septum. Two lines of evidence point towards the existence of such a mechanism. The first is the role of the dd-carboxypeptidase PBP3 in co-ordinating the division process in Streptococcus pneumoniae. By trimming the pentapeptides, PBP3 degrades the substrate of the transpeptidase activity of other PBPs (Severin et al., 1992; Morlot et al., 2004a, b). Before division starts, PBP3 appears to be distributed over the whole cell surface, except at the equator (Morlot et al., 2004a, b). This distribution of PBP3 would permit the presence of pentapeptides only at the equator, to where other PBPs could be recruited by affinity for their substrate. Interestingly, a Streptococcus pneumoniae mutant defective in PBP3 shows aberrant morphology and defects in division, with multiple aborted septa (Schuster et al., 1990), and in the colocalization of PBPs with the FtsZ ring (Morlot et al., 2004a, b). The reason for the exclusion of PBP3 from the equator is unknown, but the lipid composition at the division site might conceivably influence the localization of PBP3, as this protein is thought to be attached to the membrane by a C-terminal amphipathic helix lying in the outer-leaflet of the lipid bilayer (Morlot et al., 2004a, b).

The second line of evidence came from observations that in Staphylococcus aureus, proper localization of the single HMW class A PBP2 was lost after treatment with a β-lactam, which inactivates the enzyme active site. The d-Ala-d-Ala dipeptide at the free end of the stem peptides of the peptidoglycan or of its precursor lipid II is the donor substrate of the transpeptidation reaction catalyzed by PBPs. Proper localization of PBP2 was also lost when the presence of d-Ala-d-Ala was abolished by addition of d-cycloserine or when the access of PBPs to the d-Ala-d-Ala moieties was blocked by vancomycin (Pinho & Errington, 2005). Substrate recognition may also contribute to the localization of the PBPs in noncocci organisms, as in Escherichia coli treatment with a β-lactam thought to be specific for the septal class B PBP3 abrogates the septal localization of this PBP (Wang et al., 1998).

Note that substrate-mediated localization may partly solve the following problem of how the low-affinity PBPs (PBP2a in methicillin-resistant Staphylococcus aureus or PBP5 in enterococci) can replace other PBPs when these are inhibited by β-lactams but are still present. Inhibited PBPs may not find their proper localization, leaving their place within machineries to the low-affinity PBPs. The low-affinity PBPs, in turn, may recruit the required class A PBPs, such as PBP2 in Staphylococcus aureus (Pinho & Errington, 2005).

An interesting finding from immunofluorescence studies in Streptococcus pneumoniae was the existence of a delay between constriction of the FtsZ ring and constriction of the distribution of the cell wall synthesis machineries (Morlot et al., 2003, 2004a, b). Similarly, FtsZ is relocated to the future division site before PBPs and FtsW. In rods, the FtsZ ring is also assembled at mid cell at the onset of the division process, some time before the recruitment of other components, which can be enrolled as pre-exisiting subcomplexes (e.g. FtsQ/DivB, FtsL, FtsB/DivIC, Den Blaauwen et al., 2008, and references therein). The observation that the septal peptidoglycan synthesis machinery and FtsZ are not strictly colocalized throughout the cell cycle showed that co-ordination of the diverse aspects of division is not achieved through the constitution of a super complex comprising all the division proteins, which would remain together during the whole septation process. Note that the delay between constriction of the FtsZ ring and the septal cell wall assembly machinery may differ in various species. Cryo-electron microscopy in Enterococcus gallinarum has shown that the leading edge of the growing septum is very close to a structure identified as the FtsZ ring, separated only by the plasma membrane. In contrast, in Streptococcus gordonii, the invaginating membrane, presumably following the constriction of the FtsZ ring, is largely decoupled and in advance of the synthesis of the septal wall (Zuber et al., 2006).

Besides FtsZ, FtsA is the only conserved essential division protein to be cytoplasmic, although it is associated with the membrane. The cellular localization of FtsA in rods parallels that of FtsZ, it associates with FtsZ and a proper ratio of FtsA to FtsZ molecules is required for efficient division (Errington et al., 2003; Den Blaauwen et al., 2008). An intriguing effect of an FtsA mutant in Escherichia coli on the reactivity of the division specific PBP3 with penicillin has suggested a role of FtsA in cell wall synthesis during division (Tormo et al., 1986). In Streptococcus pneumoniae, FtsA was shown by immunofluorescence and immuno-gold electron microscopy to adopt a localization similar to that of FtsZ (Lara et al., 2005). Most interestingly, recombinant pneumococcal FtsA was found to polymerize in vitro, unlike FtsA from other microorganisms (Lara et al., 2005). The significance of this finding remains unknown, as is the precise function of FtsA.

Finally, septum closure has to be tightly regulated with chromosome segregation to avoid chromosome breakage. In the absence of a nucleoid occlusion effect in some cocci, it is possible that an orthologue of the division protein FtsK/SpoIIIE has a particularly important role in co-ordinating septum closure and chromosome segregation. This protein has a cytoplasmic domain with DNA translocase activity and is of particular importance during sporulation of B. subtilis, a process that requires packaging of the DNA in a much smaller compartment than during vegetative growth. Some cocci cells are not much larger than a Bacillus spore. Therefore, a DNA pump may be essential in some cocci for successful chromosome segregation before septum completion.

Open questions

The models presented above for peptidoglycan metabolism and its role in cell morphogenesis do not address some important aspects of this process. The first aspect is cell wall thickening of ovococci. The careful measurements of dividing Enterococcus hirae by Higgins & Shockman (1976) showed that both septal and peripheral walls, in addition to increasing their surfaces, are also thickened during the cell cycle. Strictly localized machineries cannot account for such thickening. One proposal is that enzymatic complexes involved in division are leaving the leading edge of the septum progressively as it closes. These machineries could add new layers of material to the inner face of the cell wall while drifting from the septum edge to be relocated at the equator. The drift could be a random diffusion, with the machineries reaching the equator being trapped by the presence of FtsZ. The possibility of additional layering synthesis is attractive, as it would also explain the fact that the cell shape, cell volume and cell surface are constant at various growth rates, but that the amount of synthesized peptidoglycan is not (Edelstein et al., 1980). The thicker cell wall of slow-growing cells would result from the longer time available for the PBPs to relocate at the equator of ovococci.

The second aspect is related to peptidoglycan hydrolases. Although these proteins are essential for proper peptidoglycan metabolism, the precise function of each of the many hydrolases present in different bacteria is still unknown, including the identity of the enzyme(s) responsible for splitting the septum, neither is it known in cocci as to which (if any) hydrolases are part of the multienzymatic complexes responsible for cell wall synthesis. In Staphylococcus aureus, the major autolysin is encoded by the atl gene, which encodes a precursor protein that undergoes cleavage to generate a mature amidase and glucosaminidase (Oshida et al., 1995). These proteins localize at the septal region, which is in agreement with their proposed function in the hydrolysis of peptidoglycan for separation of daughter cells after division (Yamada et al., 1996). However, Staphylococcus aureus genome encodes for 16 other peptidoglycan hydrolases (S. Foster, pers. commun.) and the function of the majority of them remains unclear.

The enzyme that splits the septal cell wall of streptococci is possibly PcsB. Inactivation of the gene encoding PcsB in Streptococcus agalactiae and Streptococcus mutans produces cells that grow in clumps, with unsplit cross-wall joining adjacent cells and other size and shape aberrations (Chia et al., 2001; Reinscheid et al., 2001). In Streptococcus pneumoniae, the pcsB gene is essential, but reduced expression also generates joined cells with abnormal morphologies (Ng et al., 2004). PcsB contains a CHAP domain (for cysteine, histidine-dependent amidohydrolase/peptidase). However, a degradative biochemical activity of PcsB on the peptidoglycan has not been demonstrated to date. Although PcsB orthologues are found in streptococci and Lactococcus lactis, a true homologue of PcsB is absent from enterococci. Instead, enterococci have a protein named SagA [Enterococcus faecium; (Teng et al., 2003)], SagBb (Enterococcus hirae) or SalA (Enterococcus faecalis). The N-terminal parts of these proteins and PcsB are similar. In contrast, the C-terminal parts of SagA/Bb and SalA are not CHAP domains but belong to other distantly related groups of the NlpC/P60 superfamily of enzymes (Anantharaman & Aravind, 2003). The similar chromosomal environment of the genes encoding PcsB, SagA/Bb and SalA suggests that they perform the same function, although some domain shuffling has occurred, possibly in relation to diverse peptidoglycan composition.

Note that in Streptococcus pneumoniae, once the septum is split, cells remain attached by the tip of their poles and form long chains without the action of a peptidoglycan hydrolase termed LytB, which is required for cell separation (De Las Rivas et al., 2002). Another hydrolase termed LytA, which is responsible for the spontaneous lysis of Streptococcus pneumoniae, may play a minor role in cell separation as the deletion of lytA causes the formation of short chains of six to eight cells (Sanchez-Puelles et al., 1986).

Lastly, much has been speculated about the existence of multienzymatic complexes responsible for cell wall synthesis in bacteria. However it is still not known as to which enzymes are part of such complexes, or how they are regulated in time and space. Most studies regarding the composition of such complexes have been performed in either Escherichia coli or B. subtilis, both of which have a large number of PBPs. Owing to the lower number of PBPs, cocci are probably better model organisms for these studies, which are essential for understanding the process of cell wall synthesis. This knowledge is important not only in the context of cell morphogenesis and division but also in the context of antibiotic resistance, as cell wall synthesis is the target of a large number of very effective antimicrobial agents.

The valuable insights into bacterial morphogenesis that have been gained by a limited number of investigations in cocci indicate that pursuing comparative studies of morphologically distinct organisms, including clinically relevant pathogens, is the way ahead to gain a better understanding of how bacteria grow and divide.

Acknowledgements

Work in the Vernet's laboratory is supported by grants from the 6th European Framework Program (COBRA LSHM-CT-2003-503335 and EUR-INTAFAR LSHM-CT-2004-512138). Work in M. Pinho's laboratory is supported by grant POCI/BIA-BCM/56493/2004 from Fundação para a Ciência e Tecnologia. The authors thank Anne Marie Di Guilmi for the thin section of pneumococcus. The authors are grateful to Orietta Massidda for pointing out the difficulties with the immunofluorescence data and to Sergio Filipe and Dirk-Jan Scheffers for critically reading the manuscript.

Ancillary