Super-resolution microscopy reveals cell wall dynamics and peptidoglycan architecture in ovococcal bacteria


  • Richard Wheeler,

    1. Krebs Institute, University of Sheffield, Firth Court, Western Bank, Sheffield S10 2TN, UK.
    2. Department of Molecular Biology and Biotechnology, University of Sheffield, Firth Court, Western Bank, Sheffield S10 2TN, UK.
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Stéphane Mesnage,

    1. Krebs Institute, University of Sheffield, Firth Court, Western Bank, Sheffield S10 2TN, UK.
    2. Department of Molecular Biology and Biotechnology, University of Sheffield, Firth Court, Western Bank, Sheffield S10 2TN, UK.
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Ivo G. Boneca,

    1. Institut Pasteur, Group Biology and Genetics of the Bacterial Cell Wall, Paris F-75015, France.
    2. INSERM, Groupe Avenir, Paris F-75015, France.
    Search for more papers by this author
  • Jamie K. Hobbs,

    1. Krebs Institute, University of Sheffield, Firth Court, Western Bank, Sheffield S10 2TN, UK.
    2. Department of Physics and Astronomy, University of Sheffield, The Hicks Building, Hounsfield Road, Sheffield S3 7RH, UK.
    Search for more papers by this author
  • Simon J. Foster

    Corresponding author
    1. Krebs Institute, University of Sheffield, Firth Court, Western Bank, Sheffield S10 2TN, UK.
    2. Department of Molecular Biology and Biotechnology, University of Sheffield, Firth Court, Western Bank, Sheffield S10 2TN, UK.
      E-mail; Tel. (+44) 114 222 4411; Fax (+44) 114 272 8697.
    Search for more papers by this author

E-mail; Tel. (+44) 114 222 4411; Fax (+44) 114 272 8697.


Cell morphology and viability in Eubacteria is dictated by the architecture of peptidoglycan, the major and essential structural component of the cell wall. Although the biochemical composition of peptidoglycan is well understood, how the peptidoglycan architecture can accommodate the dynamics of growth and division while maintaining cell shape remains largely unknown. Here, we elucidate the peptidoglycan architecture and dynamics of bacteria with ovoid cell shape (ovococci), which includes a number of important pathogens, by combining biochemical analyses with atomic force and super-resolution microscopies. Atomic force microscopy analysis showed preferential orientation of the peptidoglycan network parallel to the short axis of the cell, with distinct architectural features associated with septal and peripheral wall synthesis. Super-resolution three-dimensional structured illumination fluorescence microscopy was applied for the first time in bacteria to unravel the dynamics of peptidoglycan assembly in ovococci. The ovococci have a unique peptidoglycan architecture and growth mode not observed in other model organisms.


The major component of the bacterial cell wall, peptidoglycan, is essential for shape determination and cellular integrity (Weidel and Pelzer, 1964). Peptidoglycan is unique to the bacterial kingdom and its biosynthesis is the site of action of some of the most clinically important antibiotics such as the beta-lactams (e.g. penicillin) and glycopeptides (e.g. vancomycin) (Park and Strominger, 1957; Courvalin, 2006). Peptidoglycan is a complex polymer comprising glycan chains of repeating disaccharide residues, cross-linked via peptide side chains (Vollmer et al., 2008). Despite a wealth of knowledge concerning the components required for peptidoglycan synthesis and remodelling, the architecture of the polymer and its dynamics during cell growth and division has remained largely elusive (Vollmer and Seligman, 2010).

Mutually exclusive peptidoglycan architectural models have been proposed: the layered (classical) model, where strands run parallel to the cell membrane surface (Verwer et al., 1978; Koch, 1998) and more recently the scaffold model, where strands are orientated perpendicular to the membrane (Dmitriev et al., 2003). However, several criticisms of the scaffold model have been raised (Vollmer and Höltje, 2004), in particular the maximum glycan chain length that can be accommodated within the cell wall. A further model of peptidoglycan architecture has been proposed for Bacillus subtilis (Hayhurst et al., 2008). In this bacterium, a recent study revealed the presence of extremely long glycan strands of over 500 disaccharides. Atomic force microscopy (AFM) of purified peptidoglycan sacculi suggested that glycan strands were organized as cables running around the cell cylinder (Hayhurst et al., 2008). In S. aureus, AFM analysis has demonstrated a complex, dynamic peptidoglycan architecture of rings of material maturing, via apparent hydrolysis, into a punctate ‘knobbly’ pattern (Turner et al., 2010). Whether any of these models and observed architecture are applicable to other bacteria is unknown. Furthermore, the relationship between the dynamics of peptidoglycan synthesis and the resulting cell wall architecture has been largely ignored (Vollmer and Seligman, 2010). The dynamics of wall assembly is determined by the availability of specific peptidoglycan synthesis machineries. While a single machinery is associated with an approximate spherical morphology (S. aureus), other cell morphologies (such as rod and ovoid) rely on the activity of two distinct peptidoglycan machineries dedicated either to cell elongation or septation (Pérez-Núñez et al., 2011). The fine co-ordination of these machineries underpins the final morphology assumed by the cells. For example, Lactococcus lactis can transition from ovoid to rod-shaped when the cell wall septation machinery is suppressed (Pérez-Núñez et al., 2011).

In this work, we revealed the relationship between cell wall architecture and the dynamics of peptidoglycan synthesis for a group of organisms with ovoid cell shape (‘ovococci’) (Zapun et al., 2008) which includes a number of important pathogens. We combined biochemical analyses with super-resolution microscopy to dissect the cell wall peptidoglycan architecture and dynamics of three species of ovococcus: Streptococcus pneumoniae, Enterococcus faecalis and L. lactis. Chain length analysis revealed the existence of surprisingly long glycan strands more similar to B. subtilis than S. aureus. Extensive AFM analysis showed a preferential orientation of the peptidoglycan network parallel to the short axis of the cells. Finally, we applied for the first time in bacteria super-resolution structured illumination fluorescence microscopy to unravel the dynamics of peptidoglycan assembly in ovococci. Based on our results, we propose that ovococci have a unique peptidoglycan architecture not observed previously in other model organisms.


Length distribution of glycan strands in ovococci

Peptidoglycan was labelled with N-acetyl[14C]glucosamine, purified and digested with the S. aureus Atl amidase (see Experimental procedures). The absence of detectable amounts of remaining cross-linked material was confirmed by muropeptide analysis and digestion with mutanolysin (Fig. S1). Previous studies of glycan chain length have been conducted following separation from peptide stems by cation-exchange chromatography (MonoS) at pH 2.0 (Harz et al., 1990; Boneca et al., 2000; Hayhurst et al., 2008). However, S. pneumoniae in particular has a substantial level (40–80%) of deacetylated GlcNAc residues (Ohno et al., 1982; Vollmer and Tomasz, 2000). Glucosamine residues could bind to the MonoS column under the conditions used here, raising the possibility that a fraction of the glycan chains would be retained on the column and lost to further analysis. We therefore omitted the MonoS step and analysed glycan strand length directly by gel filtration (Fig. 1 and Table S1A). For comparative purposes, we also performed these experiments including a cation-exchange step (Fig. S2C and D and Table S1B). In all cases distribution of glycan strands was comparable, regardless of the method used. In the case of S. pneumoniae approximately 60% of material was lost on inclusion of the MonoS step.

Figure 1.

Glycan strand length distribution in Gram-positive bacteria. Radiolabelled peptidoglycan sacculi were digested with recombinant amidase. Glycan strand distribution was determined by gel filtration HPLC using two columns with distinct separation ranges to resolve long and short glycan strands. Left, TSKSW2000 column with optimal theoretical resolution of dextran standards between 1 and 30 kDa; right, TSKSW4000 with optimal theoretical resolution of dextran standards between 4 and 500 kDa. A. Samples containing 10 000–20 000 cpm were eluted at a flow rate of 1 ml min−1; fractions were collected every 15 s and analysed by scintillation counting. Muramidase hydrolysis of radiolabelled material confirmed that it was glycan (digest of B. subtilis glycan strands is shown). B. To take into account the fact that longer strands incorporate more radioactive N-acetylglucosamine, the radioactivity counts (cpm) were divided by the corresponding theoretical molecular weight (MW). Glycan chain abundance was plotted as a percentage relative to maximal (cpm/MW) ratio.

The total glycan strand population of the ovococci could not be resolved by rp-HPLC (data not shown) suggesting that the average chain length was relatively long. Glycan chain length distribution in ovoccoci was therefore investigated by size-exclusion HPLC, including two model organisms (B. subtilis and S. aureus) reported to have very long and short glycan strands respectively (Boneca et al., 2000; Hayhurst et al., 2008). E. faecalis, L. lactis and S. pneumoniae had similar glycan strand distributions when analysed with a TSKSW2000 column (Fig. 1). The majority of the radioactive material corresponded to glycan strands with an apparent molecular weight (MW) over 25 kDa (50 disaccharides), close to the theoretical resolution limit of the TSKSW2000 column (see Experimental procedures for the rationale behind the analysis). Since long strands contain more N-acetyl[14C]glucosamine residues than short strands, we determined the relative chain length distribution by plotting the ratio between radioactivity and apparent MW (Fig. S2 and Table S1). Between 44% and 57% of the ovococci glycan strand material consisted of glycan strands with an apparent MW of greater than 25 kDa (50 disaccharides). The proportion of long to short glycan strands in ovococci was virtually identical to that of B. subtilis (53% of glycan strands > 50 disaccharides) and significantly larger than that of S. aureus (14% of glycan strands > 50 disaccharides) (Table S1A). We then used a second gel filtration column with optimal resolution of dextrans with masses between 25 and 500 kDa (TSKSW4000; Fig. 1B). A more accurate analysis of high MW material purified from ovococci indicated that approximately 8–14% of the glycan strands had an apparent size > 100 disaccharides. Again, these values were similar to B. subtilis (14% of glycan strands > 100 disaccharides) and much higher than S. aureus (no glycan strands > 100 disaccharides) (Table S1A). The finding that the ovococci have very long glycan strands has important implications for peptidoglycan architecture, as it precludes the possibility of a purely scaffold architecture in ovococci (Dmitriev et al., 2003).

We attempted to observe the glycan chain distribution in situ by using a differential fluorescent lectin binding assay (Fig. S3) (Hayhurst et al., 2008) on whole cells devoid of accessory wall polymers (see Experimental procedures). Cells were labelled using fluorophores conjugated to Wheat Germ Agglutinin (WGA), which binds GlcNAc residues throughout the peptidoglycan, and Griffonia simplicifolia lectin II (GSII), which binds only to non-reducing GlcNAc termini. A strong fluorescent GSII signal will therefore only be obtained in regions of predominantly short glycan chains. Ovococcal cell walls labelled uniformly with fluorescent WGA (Fig. S3A–C). By contrast, no signal (E. faecalis and L. lactis, Fig. S3A and B) or very weak labelling (S. pneumoniae, Fig. S3C) was detected using fluorescent GSII. This labelling pattern contrasted with S. aureus, where uniform strong labelling was detected with both WGA and GSII (Fig. S3D). This suggested that in ovococci, short glycan chains were not present at a specific subcellular localization, but rather distributed throughout the whole cell wall, as seen previously for B. subtilis (Hayhurst et al., 2008).

Characterization of ovococcal cell walls by AFM

To characterize whole sacculi by AFM, cells were gently broken to release the cytoplasmic contents. Non-covalently bound components were then extracted by boiling cells in SDS and covalently bound proteins were removed with protease treatment. Sacculi were dried onto a mica sheet and imaged in ‘tapping mode’ under ambient conditions. Flattened sacculi retained the overall morphological characteristics of the cell (Fig. 2A). Cell morphology and cell dimensions appeared most similar between S. pneumoniae and L. lactis, which were more elongated at the peripheral wall than E. faecalis (Fig. 2B). Thickness measurements of dried sacculi (Fig. 2C) revealed that the average single leaf thickness was similar between E. faecalis JH2-2 (14.2 ± 1.5 nm, n = 22) and L. lactis MG1363 (13.5 ± 1.5 nm, n = 37). L. lactis VES5751, a wall polysaccharide mutant derivative of MG1363 (Chapot-Chartier et al., 2010), was not significantly thinner than the parental strain (12.6 ± 1.2 nm, n = 19), suggesting a minor contribution to average wall thickness by the polysaccharide layer. The average leaf thickness of S. pneumoniae R6 was surprisingly thin at just 7.1 ± 0.7 nm (n = 22). Whole sacculi appeared extremely smooth suggesting either the absence of nanoscale architecture, or that the peptidoglycan architecture is obscured by covalently bound wall polymers (Fig. 2B and D). In all cases annular features associated with growth and division, namely septa and equatorial rings, were observed (see below).

Figure 2.

Architecture of ovococcal cell walls. AFM height (H) and phase (P) images of cell walls extracted from exponential phase cultures. A. Field showing broken and unbroken S. pneumoniae R6 sacculi. B. Representative sacculi architecture of (left to right) S. pneumoniae R6, L. lactis MG1363 and VES5751 [a wall polysaccharide mutant derivative of MG1363 (Chapot-Chartier et al., 2010)], and E. faecalis JH2-2 showing annular features associated with growth and division (septum and equatorial rings). C. Thickness measurement of cell walls (blue bars) or purified peptidoglycan sacculi after removal of wall polymers (red bars). Error bars show standard deviation from the mean. D. S. pneumoniae cell wall surface with both equatorial rings and septum; inset was imaged at high resolution (right panel) revealing a smooth architecture. Scales: (A) H 50 nm, P 17°; (C) H 25 nm, P 33° (R6), H 50 nm, P 17° (MG1363), H 40 nm, P 40° (VES5751), H 85 nm, P 23° (JH2-2); (D) H 35 nm, P 35°, inset H 9 nm, P 27°.

Analysis of peptidoglycan architecture by AFM

Since peptidoglycan is the major structural component of the cell wall, sacculi were further purified to remove all accessory wall polymers covalently anchored to the C6 group of MurNAc, including techoic acids. This was achieved by weak acid treatment using hydrofluoric acid (HF), a compound used to hydrolyse the phosphodiester linkage while leaving the peptidoglycan structure unaffected (Atrih et al., 1999). Treatment of sacculi with HF resulted in a 28–46% decrease in cell wall thickness (Fig. 2C), which was found to be 9.4 ± 1.0 nm for E. faecalis JH2-2 (n = 41), 9.5 ± 1.5 nm for L. lactis MG1363 (n = 75), 6.8 ± 2 nm for L. lactis VES5751 (n = 18) and just 4.3 ± 0.8 nm for S. pneumoniae R6 (n = 42).

Distinct annular features associated with growth and division (septa and equatorial rings) were more clearly observed upon HF treatment (Fig. 3A and B). In S. pneumoniae a thick rib of peptidoglycan parallel to the short axis of the mid-cell was attributed to the septal cross-wall on the sacculus interior which runs around the upper and lower leaflets as a single continuous ring (Fig. 3B). Although such bands were observed in E. faecalis sacculi (Fig. 3B), folds across the mid-cell suggested that characteristic flattening of the sacculus was altered to accommodate a more deeply penetrating septal annulus (Fig. 2B). Sacculi of L. lactis and E. faecalis consistently contained a single growth annulus at the mid-cell flanked by equatorial rings, while S. pneumoniae sacculi carried up to three growth annuli in parallel (Fig. 3B). Equatorial rings were less distinct than features associated with nascent synthesis, appearing as a thickened ring of peptidoglycan, the nanoscale architecture of which was contiguous with the sacculus wall (Fig. S4).

Figure 3.

Architecture of ovococcal peptidoglycan sacculi. AFM height (H) and phase (P) images of purified peptidoglycan sacculi obtained after removal of secondary polymers by HF treatment. A. Field showing broken and unbroken S. pneumoniae R6 sacculi. B. Surface architecture of sacculi from (left to right) S. pneumoniae R6, L. lactis MG1363 and VES5751 [a wall polysaccharide mutant derivative of MG1363 (Chapot-Chartier et al., 2010)], and E. faecalis JH2-2 showing annular features (septum and equatorial rings). C. Architecture of the sacculus interior in S. pneumoniae R6. Left hand images show the edge of a broken sacculus with exposed peripheral wall displaying a circumferential banding pattern running across the short axis of the cell. Right hand images show a broken sacculus fully exposing both smooth, centripetal polar and banded peripheral architectures. D. Left, architecture of an incomplete septum revealing a centripetal pattern. Right, architecture of a septal annulus approaching complete constriction; the three-dimensional representation is shown below. All images are from S. pneumoniae R6. E. Interior peripheral wall architecture of (left to right): L. lactis MG1363, VES5751 and E. faecalis JH2-2. A less pronounced architecture is present in L. lactis and E. faecalis. Scales: (A) H 40 nm, P 13°; (B) H 20 nm, P 6°(R6), H 50 nm, P 11°(MG1363), H 50 nm, P 26°(VES5751), H 40 nm, P 23°(JH2-2); (C) H 15 nm, P 16° (left image), H 10 nm, P 21° (right image); (D) H 8 nm, P 9° (left image), H 17 nm, P 18° (right image); (E) H 30 nm, P 19° (MG1363), H 22 nm, P 10° (VES5751), H 14 nm, P 11° (JH2-2).

Even after HF treatment, the sacculus exterior of all three ovococci was smooth (Figs 3B and S4A–D). This may be due to the action of hydrolases altering peptidoglycan architecture as has been proposed for B. subtilis (Smith et al., 2000; Hayhurst et al., 2008). To investigate the interior sacculus architecture, cells were broken by French press or mechanical sheering. In all ovococci, broken sacculi characteristically fractured parallel to the short axis of the cell, usually adjacent to the septal annulus (Fig. S5A). Intact septa in various stages of completion fragmented as discrete structures in E. faecalis (Fig. S5B), but were rarely observed with S. pneumoniae or L. lactis (Fig. S5C). When the interior architecture of S. pneumoniae was exposed, bands of peptidoglycan with average width 18 ± 4 nm (n = 31) were observed parallel to the short axis of the peripheral wall (Figs 3C and S6A–D). These bands were much narrower in diameter than peptidoglycan cables imaged in B. subtilis and showed no evidence of twisting associated with a cabled wall architecture (Hayhurst et al., 2008). The interior wall of cell poles appeared smooth with a subtle centripetal architecture (Fig. S6A and D) suggesting that in S. pneumoniae the architecture of peptidoglycan synthesized by septation is distinct from that made at the peripheral wall (Fig. 3D). The distinctive peripheral wall banding pattern was occasionally observed on the interior wall of E. faecalis but not in L. lactis samples (Fig. 3E). Although no clear periodicity was observed, the nanoscale architecture suggested that the peptidoglycan network was preferentially orientated in the direction of the short axis (Fig. 3C–E).

Super-resolution structured illumination microscopy of vancomycin labelled ovococci

The cell size of ovococci (∼ 1 µm) is close to the diffraction limit of optical resolution (0.2 µm) and therefore represents a major obstacle to study the dynamics of growth and division. Here we used super-resolution fluorescence microscopy to gain insight into the spatial and temporal dynamics of growth and division in ovococci. Fixed cells, labelled with fluorescent vancomycin, which binds d-ala-D-ala residues in peptidoglycan, were imaged by OMX three-dimensional structured illumination microscopy (3D-SIM; Fig. 4) (Gustafsson, 2000; Posch et al., 2010). Although the gross characteristics of the cell cycle were initially visualized by conventional deconvolution microscopy, OMX imaging revealed novel, subtle differences between growth and division dynamics of the ovococci (Fig. 4A–C). Super-resolution allowed accurate measurement of subcellular features with a resolution of approximately 50 nm in the x–y plane and 125 nm in the z plane. Four distinct dimensions [cell length (L), maximal cell width (W), cross-wall diameter (C) and septal aperture diameter (A); Fig. 4D] were measured in individual cells to follow cell elongation, septum progression and cell constriction.

Figure 4.

Fluorescence microscopy of vancomycin-labelled ovococci. A–C. (A) S. pneumoniae R6, (B) L. lactis MG1363 and (C) E. faecalis JH2-2. Images show a comparison between standard deconvolution microscopy (DeltaVision, DV) and structured illumination (OMX) microscopy of the same sample. D. Schematic of the measurements taken to analyse cell cycle in ovococci; A, septal aperture; C, maximal cross-wall diameter; L, total cell length; W, maximal cell width.

Overall shape and cell dimensions appeared most similar between S. pneumoniae and L. lactis which were more elongated than E. faecalis (Table 1). As expected for cells which divide in a plane perpendicular to the long axis and parallel to the short axis, significant variation was observed in cell length, whereas maximum cell width showed little variation throughout the cell cycle (Table 1).

Table 1.  Average cell length and width of ovococci.
 Length (L)Maximal width (W)
  1. Average cell dimensions and standard deviations are in nm (n = 50).

S. pneumoniae1870 ± 520820 ± 70
E. faecalis1370 ± 320860 ± 40
L. lactis1920 ± 360890 ± 30

Temporal and spatial regulation of the cell cycle in ovococci

We analysed the temporal regulation of successive rounds of division in non-limiting growth conditions, where growth rates were comparable between ovococci (between 48 and 56 min per generation). A complete division began with the appearance of vancomycin labelling at the mid-cell and ended when the diameter of the septal aperture was zero. We recorded the appearance of new labelling foci in presumptive daughter cells during progression of septation, which was followed by plotting septal aperture against cross-wall diameter (Fig. 5). S. pneumoniae displayed overlapping rounds of growth and division, as strong vancomycin labelling was observed in presumptive daughter cells before closure of the septum (Fig. 5A). By contrast, E. faecalis and L. lactis displayed discrete rounds of growth and division. In E. faecalis, vancomycin labelling was detected in presumptive daughter cells only when the septum was closed (diameter of aperture equal to zero, see red circles in Fig. 5B). In L. lactis, vancomycin labelling was never observed in presumptive daughter cells (resulting in the absence of red circles in Fig. 5C), indicating that division must be complete before a new round starts.

Figure 5.

Peptidoglycan synthesis during growth and division of the ovococci. The timing of successive rounds of cell division was followed in (A) S. pneumoniae R6, (B) E. faecalis JH2-2 and (C) L. lactis MG1363 by plotting the diameter of the septal aperture (as an indicator of septum progression; see Fig. 4) against the diameter of the septal cross-wall (as an indicator of cell separation; see Fig. 4). All strains were grown in non-limiting conditions to ensure equivalent doubling times. (•), cells showing a single ring of vancomycin labelling at the septum; (inline image), cells with additional rings of labelling at equators of presumptive daughter cells. S. pneumoniae shows overlapping rounds of synthesis, whereas E. faecalis and L. lactis show discrete rounds of division. Contrary to L. lactis, where complete separation occurs before the next round of division begins, E. faecalis initiates division while separation is continuing (i.e. cross-wall diameter decreases).

The synchronization of septum synthesis and septum hydrolysis varies between the ovococci. In L. lactis, and more prominently in S. pneumoniae, labelled peptidoglycan mainly occurred as thin rings (Movie S1A and B), indicating that in these ovococci septum progression is concomitant with hydrolysis for daughter cell separation. Consistent with AFM observations, broad septal discs were frequently visualized by fluorescence microscopy in E. faecalis samples (Movie S1C) suggesting that septum formation and hydrolysis are not tightly synchronized processes in this organism. In L. lactis (Movie S1B) vancomycin labelling occurs as a thin ring at mid-cell, but also relatively uniformly over the area of the cell wall associated with peripheral wall synthesis. This may suggest longevity of the maintenance of D-ala-D-ala residues which are able to bind vancomycin within the peripheral wall.

We next investigated the dynamics of cell division for each of the ovococci using the ratio between cross-wall diameter and maximal cell width (C/W), plotted against cell length, as an indicator of septum constriction. We defined three developmental phases corresponding to elongation (C/W > 0.95), transition to constriction (0.7 < C/W < 0.95) and constriction (C/W < 0.7).

Lactococcus lactis was the only organism for which we observed a strict elongation phase which accounted for an increase in cell length of approximately 600 nm. This phase was followed by a transition to cell constriction (0.7 < C/W < 0.95) during which the cell length continued to increase by up to 700 nm (Fig. 6A). By contrast, a short cell elongation phase, corresponding to an increase in cell length of approximately 300 nm, was detected in S. pneumoniae before transition to septation (Fig. 6B). Instead, cell constriction occurred simultaneously with elongation at the equatorial rings of the forming daughter cells. The existence of overlapping rounds of cell division and a greater variation in cell width (Table 1) gave rise to a more heterogeneous distribution of cell dimensions. In addition, a proportion of the cells (∼ 6%) did not undergo septation but continued to elongate at multiple vancomycin labelling foci. Similar to S. pneumoniae, a relatively short elongation phase was detected prior to cell constriction in E. faecalis (Fig. 6C). Elongation accounted for an approximate length increase of 300 nm. Septum constriction and cell elongation progressed simultaneously throughout the cell cycle.

Figure 6.

Dynamics of cell growth and division in ovococci. A–C. Cell cycle progression was followed in (A) L. lactis MG1363, (B) S. pneumoniae R6 and (C) E. faecalis JH2-2 by plotting total cell length against the width ratio (cross-wall diameter over maximal cell width, C/W; see Fig. 4) as an indicator of septum constriction. Increasing length with little or no change in C/W ratio indicates an elongation phase. Increasing length with decreasing C/W ratio indicates septum constriction during cell growth. The dotted lines indicate threshold values of C/W, defining the developmental stages described in (D). D. Model of cell growth and division in different ovococci. Cells were grouped in three developmental phases: elongation (C/W > 0.95), transition to constriction (0.7 < C/W < 0.95) and constriction (C/W < 0.7). For each phase, an interpretive model is provided; (inline image), the peripheral peptidoglycan biosynthesis machinery; (•), the septal peptidoglycan biosynthesis machinery. Red rings represent peripheral elongation and black rings represent septal synthesis. Annular features of the peptidoglycan in dark blue include the equatorial rings identified by AFM and the cross-wall. Representative vancomycin-labelled cells accompany each diagram. The percentage of cells in each developmental stage (n = 50) is shown for each ovococcus.

The overall distribution of the number of cells in each phase was similar between the ovococci (Fig. 6). Most of the cell cycle was devoted to elongating (46–54% of cells). The number of cells undergoing transition to constriction was slightly lower in E. faecalis (22%) as compared to L. lactis and S. pneumoniae (30% and 32% respectively), suggesting that in this organism septation occurs more rapidly.


Computational simulations based on the chain length reported for E. coli and S. aureus proposed a scaffold model in these organisms (Dmitriev et al., 2004; 2005). This scaffold organization remains controversial especially in E. coli where most of the glycan chains are composed of 20 to 25 disaccharides (Vollmer and Höltje, 2004), which is inconsistent with the proposed architecture. Whether the scaffold model can be applied to other bacteria is unknown. To address this issue, we investigated the chain length of the glycan strands, a major constraint on peptidoglycan architecture, in three ovococci. Several approaches have been previously described to determine glycan chain length (Ward, 1973; Schindler et al., 1976; Glauner, 1988; Boneca et al., 2000). Unfortunately, each of these methods introduces bias (Vollmer, 2008) or resolves a limited range of strand sizes. For example, rp-HPLC resolved short glycan strands in S. aureus and E. coli, but it was inadequate in the case of B. subtilis. More recently, a size-exclusion chromatography strategy was developed to reveal the glycan chain distribution in B. subtilis (Hayhurst et al., 2008). We applied a similar strategy to re-evaluate the entire complexity of chain sizes in Gram-positive bacteria, using two size-exclusion columns with distinct resolution ranges. We used dextran molecules as standards to emulate the glycan backbone, assuming that their low level of branching (5%, one to two residues) has a marginal impact on their retention time (Fig. S2A and B). Using S. aureus, we found that strand size peaked between 5 and 10 disaccharides (Fig. 1B). This is in agreement with previous rp-HPLC results, which reported a predominant length between 3 and 10 disaccharides (Boneca et al., 2000). All ovococci contained much longer glycan strands than those found in S. aureus. In fact, the glycan length distribution in ovococci was virtually identical to that of B. subtilis, which has markedly long strands (Hayhurst et al., 2008).

Although we did not observe extremely long glycan strands as previously observed in B. subtilis (attributable to the radiolabelling method in rich medium, favouring autolysin activities), a high proportion of long glycan strands (Table S1) seems to be a phenomenon unique to Gram-positives (Vollmer and Seligman, 2010). Consistent with the presence of long glycan strands, labelling of ovococcus cell walls with the fluorescent GSII lectin gave no significant fluorescence signal for L. lactis and E. faecalis and a very weak signal for S. pneumoniae at the limit of detection. The ability of our method to separate long glycan strand material (> 25 disaccharides) that could not be resolved by rp-HPLC allows the glycan strand distribution to be determined. Therefore, the use of size-exclusion chromatography has direct implications for the modelling of peptidoglycan architecture. Considering a length of 1 nm per disaccharide (Vollmer and Höltje, 2004) and a maximum cell envelope thickness of 50 nm (Matias and Beveridge, 2005; Zuber et al., 2006), the significant proportion of chains of over 100 disaccharides in ovococci (between 8% and 14%) is clearly not in favour of a scaffold architecture in these bacteria. In S. aureus, we found that 3–14% of glycan chains contained 50 or more disaccharides, depending on the column used. This represents a significant proportion of strands that were not taken into account when developing the scaffold model proposed for S. aureus (Dmitriev et al., 2004).

The relationship between biochemistry and peptidoglycan architecture is unknown. Extensive AFM analysis of the peptidoglycan architecture of three model ovococci (S. pneumoniae, L. lactis and E. faecalis) suggested that glycan strands are preferentially oriented circumferentially, parallel to the short axis of the cell. This orientation was present at the macro-architectural level as annular features associated with growth and division, and at the nanoscale level. This was particularly evident at the interior peripheral wall of S. pneumoniae which possessed parallel bands of material, the average width suggesting that they were composed of several intertwined glycan strands. Such bands were rarely seen in E. faecalis and were not present in L. lactis, although preferentially circumferential orientation of the smooth nanoscale architecture was observed. The pronounced nascent banding architecture in S. pneumoniae could result from the abundance of sites of peptidoglycan synthesis due to overlapping rounds of division. A lack of cabled peptidoglycan architecture is in agreement with recent work on L. lactis using purified sacculi (Andre et al., 2010). Since peptidoglycan purification by HF extraction alters neither the biochemistry of peptidoglycan nor its architecture (Ries et al., 1997; Hayhurst et al., 2008; Turner et al., 2010), it may be that nascent peptidoglycan architecture is lost via rapid remodelling due to autolytic activity. In support of this hypothesis, the architectural features were always observed in close proximity to annular growth features (Fig. 2) where glycan chain length is expected to be maximal. Previous AFM studies on living L. lactis identified topographic ridges on the outer cell surface (Andre et al., 2010) that are not present on pure peptidoglycan. Thus, other wall components such as teichoic acids, polysaccharides and/or proteins exposed at the surface of L. lactis contribute to the overall cell wall architectural features. However, unless purified material is used, such features cannot be unequivocally assigned to peptidoglycan, the structural integrity determinant of the cell wall.

Our AFM analysis showed that the cell wall architecture in ovococci is distinct from that observed in other Gram-positive species. The rod-shaped organism B. subtilis has a cabled architecture along the cylinder and apparent spiral septa (Hayhurst et al., 2008). In contrast, the round S. aureus has a complex, dynamic architecture of rings of nascent peptidoglycan which mature during growth into a knobbly architecture (Turner et al., 2010). Our analysis of ovococci also revealed architectural variation between species. To explain the differences in cell wall ultrastructure, we followed the dynamics of the ovococcal cell cycle using fluorescent vancomycin labelling in conjunction with 3D-SIM (OMX) super-resolution microscopy. To our knowledge, this is the first reported use of 3D-SIM to study bacterial specimens with a size close to the resolution limit of conventional deconvolution microscopy. We propose a model describing the cell cycle for each of the ovococci, taking into account AFM and fluorescence microscopy data (Fig. 6D). These models account for the activity of the peptidoglycan synthesis machineries involved in elongation (in red) and septation (in black). They highlight major differences in the temporal and spatial regulation of the cell cycle in the ovococci: (i) the initiation of subsequent rounds of division; (ii) the co-ordination of septal constriction with cell separation; and (iii) the timing of peripheral vs. septal peptidoglycan synthesis.

In all the ovococci included in this study, cell separation primarily relies on glucosaminidase activity (Buist et al., 1995; De Las Rivas et al., 2002; Mesnage et al., 2008). A unique feature of E. faecalis was the formation of a deeply penetrating cross-wall prior to cell separation. Such a cross-wall was never seen in the other ovococci, as a result of a tight association between septum constriction and hydrolysis. Our data thus indicate that synchronization of glucosaminidase enzymes with peptidoglycan synthesis is regulated differently amongst ovococci. The exact mechanism responsible for the co-ordination of septum synthesis with hydrolysis awaits further analysis.

A hallmark of the ovococci, identified by AFM, is the presence of equatorial ring features in presumptive daughter cells (Higgins and Shockman, 1970) (Fig. 3B) that coincide with the initiation site of the next round of peptidoglycan synthesis (Fig. 5). No mechanism for targeting the division machinery to the nascent septum has been identified in the ovococci. One possibility is that epigenetic information encoded by the local peptidoglycan architecture at the equatorial rings aids recruitment of the peptidoglycan synthesis machinery. The use of such intergenerational peptidoglycan architectural features to convey information to guide the spatial control of nascent synthesis has recently been suggested for S. aureus (Turner et al., 2010). In the ovococci, following recruitment of the peripheral wall peptidoglycan synthesis machinery at the equatorial rings, a single annulus of peptidoglycan synthesis is detected from which both elongation and septation occur. Elongation and septation are therefore intimately linked processes in ovococci. A very recent study on S. pneumoniae also provides support for the two-state elongation-septation model (Land and Winkler, 2011). Our results are in agreement with recent studies in L. lactis which revealed that in filamentous cells, peripheral wall synthesis was colocalized with FtsK, one of the first proteins to localize at the septal Z-ring (Pérez-Núñez et al., 2011). Interestingly, it has been shown in Caulobacter crescentus that independent of Mre-mediated elongation, FtsZ-dependent elongation also occurs prior to constriction (Aaron et al., 2007). Colocalization of all the high MW penicillin binding proteins, which are responsible for peptidoglycan biosynthesis, with FtsZ has been reported in S. pneumoniae (Zapun et al., 2008) suggesting that a similar mechanism to C. crescentus may exist in the ovococci for the co-ordination of elongation and constriction during the cell cycle.

Together our data provide new insights regarding the peptidoglycan architecture of ovococci, which is distinct from architectures observed in other Gram-positive bacteria. Thus, apparent diversity of peptidoglycan architecture provides solutions to common engineering problems encountered by bacteria of different morphologies and growth/division strategies.

Experimental procedures

Strains and growth conditions

Bacteria strains used in this study were B. subtilis 168, S. aureus COL, S. pneumoniae R6, L. lactis MG1363, L. lactis VES5751 and E. faecalis JH2-2. All bacteria were routinely grown in rich medium: Todd-Hewitt supplemented with 0.5% w/v yeast extract for S. pneumoniae; M17 supplemented with 0.5% w/v glucose for L. lactis; Tryptone-soya broth E. faecalis; Luria–Bertani (LB) for B. subtilis and S. aureus. For cell wall labelling, all strains were grown in LB broth, except S. pneumoniae which was grown in Todd-Hewitt supplemented with yeast extract as above.

Cell wall labelling and purification

For labelling with N-acetyl[14C]glucosamine, cells were grown from overnight cultures until exponential phase (OD600 = 0.3). Cultures were diluted in 50 ml of pre-warmed LB containing 0.185 MBq N-acetyl[14C]glucosamine (1.67 TBq mmol−1; Hartmann Analytic) and 500 ml of non-radioactive medium to give a starting OD600 = 0.04. After three generations (OD600 = 0.3–0.35), the cells were harvested by centrifugation (5 500 g for 5 min at room temperature), and non-covalently bound cell wall components were removed in 4% w/v boiling SDS for 30 min. After repeated washing steps in water and treatment with 2 mg ml−1 pronase to remove covalently bound proteins, secondary wall polymers were extracted by incubation in 48% v/v HF for 48 h at 4°C. Pure peptidoglycan sacculi were extensively washed in water and stored at −20°C. Using this method, 5–20 mg of radiolabelled peptidoglycan was obtained for each species, with approximately 10% of N-acetyl[14C]glucosamine incorporation.

Purification of glycan strands

Radiolabelled peptidoglycan sacculi were digested by the amidase domain of the bifunctional S. aureus autolysin Atl as previously described (Hayhurst et al., 2008). To determine optimal digestion conditions, pilot experiments were carried out with approximately 10 000 cpm of radiolabelled peptidoglycan and a range of enzyme concentrations for 3 h at 37°C in 10 mM Tris-HCl (pH 7.0) containing 1 mM CaCl2. Solubilization levels were determined by comparing total counts in the reaction mixture against those in the soluble fraction obtained after centrifugation (10 min at 20 000 g). These conditions were scaled up to digest sufficient material for glycan chain length analyses. Typically 1 mg of peptidoglycan was digested overnight with an Atl concentration fivefold greater than that required to solubilize over 90% of the peptidoglycan. The sample was then boiled (3 min) to inactivate the enzyme, insoluble material removed by centrifugation (14 000 g for 8 min at 4°C), and the supernatant collected for further analysis. Counts in the supernatant and pellet fractions using N-acetyl[14C]glucosamine radiolabelled material were used to check hydrolysis efficiency (97.5 ± 3.9% hydrolysis, n = 10). For cation exchange separation of glycan strands, the pH of solubilized peptidoglycan fragments was adjusted to 2 and the sample loaded onto a MonoS column (Amersham Biosciences) equilibrated with 100 mM sodium phosphate buffer (pH 2.0). Glycan strands were recovered in the flow-through fraction, and bound material containing peptide stems was eluted with the buffer containing 1 M NaCl.

HPLC separation of glycan strands

Size-exclusion chromatography of glycan strands was performed as described (Hayhurst et al., 2008). The radiolabelled glycan strands were freeze-dried and resuspended in 100 mM phosphate buffer (pH 6.0). Approximately 10 000–20 000 cpm of each glycan strand fraction (corresponding to 50–100 µg peptidoglycan) was injected in a volume of 200 µl onto TSKSW2000 (7.5 × 600 mm) and TSKSW4000 (7.5 × 300 mm) size-exclusion HPLC columns (Tosoh) pre-equilibrated in 100 mM phosphate buffer (pH 6.0). Elution was carried out at a flow rate of 1 ml min−1. The fractions were collected every 15 s (250 µl) and counted after mixing in Ultima GoldTM scintillation cocktail (Perkin Elmer) to detect radiolabelled glycan strands. The gel filtration columns were calibrated using dextran standards (analytical standards grade for GPC, Sigma-Aldrich: 1000 Da, #31416; 1500 Da, #31394; 5000 Da, #00269; 10 000 Da, #D9260; 12 000 Da, #00270; 25 000 Da, #00271; 50 000 Da, #00891; 80 000 Da, #00892; 150 000 Da, #00893). Typically 200 µl of dextran (10 mg ml−1) was injected in phosphate buffer at a flow rate of 1 ml min−1. Fractions were collected every 15 s (250 µl) and neutral sugars were detected using a colorimetric phenol-sulphuric acid assay (Kennedy and Pagliuca, 1994).

For rp-HPLC analyses (Fig. S1), glycan strands were reduced and analysed as previously described (Boneca et al., 2000).

Analysis of glycan strand distribution

To avoid bias due to the exponential distribution of glycan strands as they are eluted from the size-exclusion column (Fig. S2A), we re-plotted the radioactivity elution profile on a linear axis (Fig. S2B). The percentage of radioactivity above and below the threshold value of 25 kDa was determined by measuring the area under the curve.

Radioactivity counts do not reflect the abundance of a particular strand size (for example, one glycan strand of 100 disaccharides gives the same signal as 20 strands of five disaccharides). We therefore divided the radioactivity counts (CPM) by the corresponding theoretical MW deduced from the calibration curves to reveal the abundance of strand lengths. The distribution of glycan chain lengths was plotted as a percentage relative to the maximum CPM/MW ratio observed in each organism (Fig. S2C). To reveal the complete distribution of chain sizes in Gram-positive bacteria, we used two size-exclusion columns with distinct resolutions ranges: TSKSW2000 (Fig. S2, left panels) and TSKSW4000 (Fig. S2, right panels).

Purification of sacculi for AFM

Bacterial cultures were grown to exponential phase (OD600∼ 0.5) and peptidoglycan purified as described previously (Turner et al., 2010). Briefly, cells were broken either by French press at a pressure of approximately 1000 psi, or by mechanical shearing using a FastPrep. After breakage, extraction by boiling in SDS (4% w/v) and pronase (2 mg ml−1) treatment, removal of accessory polymers was achieved by incubation in 48% v/v HF at 4°C. Purified sacculi were washed at least six times in LC-MS CHROMASOLV® grade water (Fluka) at room temperature and stored at −20°C.

AFM of ovococcus sacculi

Samples were adjusted to an appropriate working concentration for imaging of well-dispersed sacculi. Aggregates were dispersed by gentle sonication. The sacculi suspension was dried onto a clean mica sheet under nitrogen gas. Sacculi were imaged using a Dimension or Multimode AFM with an Extended Nanoscope IIIa controller (Veeco Instruments). Imaging was carried out in tapping mode using silicon tips (Olympus) under ambient conditions. Image processing and measurements were performed as described previously using Gwyddeon v2.19 software (Turner et al., 2010). Over 600 height and phase images were obtained for S. pneumoniae R6, over 100 images for E. faecalis JH2-2 and for L. lactis MG1363 and VES5751, in excess of 300 and 120 images respectively.

Labelling of peptidoglycan with fluorescent lectins

Cells were grown from overnight cultures for approximately five to six generations until exponential phase (final OD600∼ 0.3) and immediately added to excess boiling 5% w/v SDS. Cells were then incubated in 5% w/v SDS for 30 min at 50°C, harvested and washed with water by centrifugation and incubated overnight in 48% v/v HF at 4°C. Extracted cell walls were then washed thoroughly with water and resuspended to 250 µl volume with 1.0 mM CaCl2 and either 100 µg ml−1 WGA AlexaFluor 488 conjugate (Invitrogen) or 100 µg ml−1 GSII AlexaFluor 594 conjugate (Invitrogen) for 30 min. Labelled pellets were washed by centrifugation to remove excess lectin.

Labelling of peptidoglycan with fluorescent vancomycin

Cells were grown from overnight cultures to exponential phase (final OD600 ∼ 0.3). A 1 ml aliquot of culture was removed and incubated for 5 min with 1.0 µg ml−1 of a 1:1 mix of vancomycin (Sigma) : vancomycin-BODIPY (Invitrogen) in water. Aliquots were chilled rapidly to 4°C and washed twice by centrifugation with ice-cold distilled water. Cells were then fixed in PBS with 6% w/v paraformaldehyde and 0.005% v/v glutaraldehyde for 30 min at room temperature and washed at least twice in distilled water by centrifugation.

Fluorescence microscopy

Cells were mounted onto poly-l-lysine coated slides and imaged using a DeltaVision deconvolution microscope (Applied precision). Images were deconvolved using SoftWoRx suite v.3.5.1. Three-dimensional structured illumination microscopy (3D-SIM) was carried out at the University of Dundee Light Microscope Facility using an OMX v.3 system (Applied Precision). Images were reconstructed and aligned using SoftWoRx suite v.4.5.0 (unreleased development version). For both DeltaVision and OMX images, contrast optimization, 3D reconstruction and cell measurements were performed using ImageJ v.1.43u.


This work was funded by the Biotechnology and Biological Sciences Research Council (BBSRC). Work in the Boneca laboratory is supported by a European Research Council starting grant (n°202283 PGNfromSHAPEtoVIR). The light microscopy facility was supported by the European Union (Grant N°LSHM-CT-2006–019064), a University of Sheffield Capital Infrastructure Grant and by a Wellcome Trust Grant. We thank an anonymous reviewer for insightful comments on the manuscript. We are grateful to Marcus Posch for help and valuable advice with OMX and to the Scottish Universities Life Sciences Alliance (SULSA) and the Light Microscope Facility, University of Dundee for access to the 3D-SIM microscope. S. Mesnage is supported by a Marie Curie Intra-European Fellowship (Grant N°251336) and INSERM. We are indebted to Saulius Kulakauskas for providing strains and Bob Turner for critical reading of the manuscript.