A bacterial cell protects itself from rupture by the internal osmotic pressure (turgor) by surrounding its cytoplasmic membrane with a single, giant molecule called the peptidoglycan (PG, or murein) sacculus (Weidel and Pelzer, 1964). This remarkable structure, made of glycan chains cross-linked by short peptides, forms a net-like monolayer in Gram-negative bacteria and a thicker, multilayered ‘wall’ in Gram-positive species. PG is strong enough to withstand the turgor and to maintain cell shape but, unlike the more rigid wall of a plant cell, it is elastic and can significantly expand or relax depending on the osmotic conditions. While the sacculus is essential to protect the cell's integrity, the presence of this stretched exoskeleton made of covalently linked subunits poses a challenge for the cell when it comes to growth and cell division. Not only must the cell enlarge the surface area of its stress-bearing sacculus by a safe mechanism to avoid defects that could lead to lysis, it also must execute a spatial and temporal control on sacculus growth to ensure that the daughter cells inherit the species-specific cell shape and size. Despite recent progress in PG research the molecular details of sacculus growth have remained largely unknown (Typas et al., 2012).
Over the past decades different models of sacculus architecture and growth have been proposed and discussed with respect to the phenotype of mutants lacking PG enzymes or cell morphogenesis proteins (Vollmer and Seligman, 2010; Typas et al., 2012). All PG growth models assume not only synthetic but also hydrolytic activities. This is because common sense tells us that merely synthesizing new PG and attaching it to the sacculus could just thicken the layer without increasing its surface area. Only if hydrolases open the PG mesh can the new material be inserted into the layer for surface growth (Fig. 1A). Very surprisingly, this logical assumption has never been proven by experimentation until recently. Presumably, this is due to the large number of hydrolases with redundant functions present in many species (Vollmer et al., 2008). For example, until recently Escherichia coli has been known to have 13 periplasmic hydrolases, including four N-acetylmuramyl-L-alanine amidases, seven lytic transglycosylases and three endopeptidases, and no single mutant nor any multiple mutant lacking up to seven hydrolases showed a growth defect – apart from a cell chaining phenotype of multiple amidase mutants indicating a role of these hydrolases in cleaving the septum PG for cell separation (Heidrich et al., 2001). Interestingly, the group of Manjula Reddy (Hyderabad, India) has now published in this issue of Molecular Microbiology that three novel, redundant PG hydrolases, the DD-endopeptidases Spr, YdhO and YebA, are collectively required for cell growth and PG incorporation in E. coli (Fig. 1A) (Singh et al., 2012). Moreover, it was found recently that in Bacillus subtilis, a species with as many as 35 known or hypothetical PG hydrolases, the disruption of two DL-endopeptidase genes, cwlO and lytE, is lethal (Bisicchia et al., 2007) and that the catalytic activity of at least one of the corresponding PG hydrolases is required to support cell growth (Hashimoto et al., 2012). These important discoveries establish, to my knowledge, for the first time that PG hydrolases are indeed essential for bacterial growth.
Peptidoglycan endopeptidases hydrolyse amide bonds within the cross-linked peptides that connect two or more glycan chains resulting in larger holes in the PG (Fig. 1B) (Vollmer et al., 2008). They are capable of degrading insoluble, high-molecular-weight PG sacculi releasing soluble glycan chains that carry un-cross-linked peptides. The PG peptide cross-links contain D- and L-amino acids and, in Gram-negative bacteria, meso-diaminopimelic acid (meso-A2pm) rather than the usual L-amino acids found in proteins. Hence, PG endopeptidases are sometimes and, in fact, more correctly called ‘amidases’, although the terms DD-endopeptidase, DL-endopeptidase and LD-endopeptidase denoting the cleavage site within a cross-link are widely used in the literature (Fig. 1B). PG endopeptidases are members of different protein families, including class C penicillin-binding proteins (PBPs, for example: E. coli PBP4 and PBP7) (Sauvage et al., 2008), the LAS metallopeptidases (E. coli MepA) (Marcyjaniak et al., 2004), MEROPS family M23/LytM metallopeptidases (lysostaphin; E. coli NlpD and the newly identified YebA; Helicobacter pylori Csd1–3) (Bochtler et al., 2004) and NlpC/P60 peptidases which belong to the cysteine, histidine-dependent amidohydrolase/peptidase (CHAP) superfamily (the new E. coli enzymes Spr and YdhO; B. subtilis LytE, LytF and CwlO; the type VI secretion system effectors Tse1, Tae2, Tae3 and Tae4) (Bateman and Rawlings, 2003; Rigden et al., 2003).
Like PG hydrolases in general, endopeptidases have multiple and often poorly characterized functions that differ among bacterial species. Several interesting recent findings highlight their importance for various aspects of bacterial lifestyle, for example in bacterial warfare: Pseudomonas aeruginosa and many other Gram-negative bacteria deliver NlpC/P60-type endopeptidases (called Tse and Tae) via type VI secretion systems into the periplasm of adjacent Gram-negative rivals causing their lysis, unless the target cell has specific immunity proteins to inactivate the incoming hydrolases (Russell et al., 2011). These secreted endopeptidases belong to four subfamilies, characterized by the founding members Tse1, Tae2, Tae3 and Tae4. They have different cleavage sites in the peptide cross-link, albeit always next to a meso-A2pm residue, and are either DL-endopeptidases (Tse1, Tae4) or DD-endopeptidases (Tae2, Tae3) (Russell et al., 2012). Consistent with its aggressive function, the Tse1 DL-endopeptidase has a compact structure with a wide open active site, unlike ‘housekeeping’ enzymes of the same family (like E. coli Spr) with apparently more tightly regulated activity (Chou et al., 2012). The recently published Bdellovibrio bacteriovorus PBP4 homologues Bd0816 and Bd3459 are further examples of endopeptidases that are used to attack other bacteria; in this case the endopeptidases promote the rounding of prey cells (Lerner et al., 2012). Many bacterial PG hydrolases including endopeptidases are not secreted to kill rivals but function inside their own cells. In H. pylori several M23-type DD-endopeptidases (named Csd1–3) have an essential role together with the DL-carboxypeptidase Csd4 and other proteins of unknown function in generating the cell's helical shape, which is required for efficient colonization of the mucus layer of the stomach (Sycuro et al., 2010; 2012). Although the precise mechanism is not known, it is possible that Csd1–3 are spatially well regulated to relax cross-links in the stretched PG sacculus enabling to maintain helical curvature and twist of the cell during growth.
A H. pylori Δcsd1–3 triple endopeptidase mutant is unable to produce a helical cell shape but is still able to grow and divide (Sycuro et al., 2010). Not so E. coli cells depleted of spr, ydhO and yebA which, interestingly, cannot elongate nor divide but swell to become oblong spherical shaped before they suddenly burst (Fig. 1A) (Singh et al., 2012). The single mutants or the ydhO spr double mutant are viable, and the yebA spr double mutant grows only in minimal media. The simultaneous deletion of other known or hypothetical endopeptidases did not exacerbate the phenotypes of the spr single or ydhO spr double mutant, indicating that spr, ydhO and yebA alone form a set of redundant genes providing an essential function for growth of E. coli. As would be expected from sequence similarity, the purified proteins showed DD-endopeptidase activity against cross-linked dimeric and trimeric muropeptides, which are soluble fragments obtained by muramidase digestion of PG. YebA and YdhO also hydrolyse DD-cross-links in intact PG sacculi, albeit with apparently weaker activity as compared to soluble muropeptides. Spr was not active against sacculi and also exhibits a weak LD-carboxypeptidase activity against muropeptides. Two further observations highlight the importance of the novel endopeptidases for PG growth in E. coli. First, it is the endopeptidase activity that is required for growth, because the catalytically inactive versions fail to support growth just like the gene deletions. This observation implies that cleavage of cross-links in the PG is essential for growth. Second, radioactive labelling experiments showed that depletion of spr, yebA and ydhO prevents further attachment or incorporation of new PG into the sacculus. This is different to previous in vitro experiments showing that the E. coli PG synthases PBP1A and PBP2 can attach newly synthesized PG to isolated sacculi by forming new peptide cross-links in the absence of a hydrolase (Born et al., 2006; Banzhaf et al., 2012). Interestingly, in the cell the attachment of new glycan chains (or patches of them) to the sacculus, which is know to occur by formation of peptide cross-links (de Jonge et al., 1989), appears to depend on the hydrolysis of other, presumably old cross-links by Spr, YebA and/or YdhO. Therefore, the identification of the three collectively essential DD-endopeptidases is a big step towards understanding the mechanism of PG growth in E. coli and likely other Gram-negative species. Moreover, the recent identification of essential B. subtilis DL-endopeptidases (CwlO and LytE) suggests that PG hydrolases could also be required for growth of a much thicker PG in Gram-positives (Bisicchia et al., 2007; Hashimoto et al., 2012). Also in this case the catalytically inactive versions of CwlO and LytE failed to support cell wall elongation indicating that the peptide cleavage activity of these hydrolases is required for growth (Hashimoto et al., 2012). It remains to be investigated whether or not the cwlO lytE depleted cells fail to incorporate new PG as do the spr yebA ydhO depleted E. coli cells.
How are these endopeptidases regulated? Are they positioned at designated sacculus growth sites and then activated? Are the activities of these and other hydrolases co-ordinated with those of PG synthesis enzymes, as has been proposed in the three-for-one growth mechanism by Höltje (Höltje, 1998)? Answering these questions will likely provide important insights into the molecular mechanisms of PG growth and its regulation, which occurs at multiple levels (Typas et al., 2012). In the case of B. subtilis, the expression of cwlO and lytE is controlled by the essential YycFG two-component system indicating that cell wall hydrolysis might be regulated by a yet unknown signal sensed by YycFG. The signal for YycFG could be, as has been suggested, the PG precursor lipid II or the process of its incorporation into PG (Bisicchia et al., 2007). LytE requires the actin homologue MreBH for proper localization and functioning in the cell wall, consistent with the role of both MreBH and LytE in cell elongation (Carballido-Lopez et al., 2006). It is not yet known how the new E. coli endopeptidases required for cell growth are positioned and regulated, and how their depletion results in a halt in PG incorporation. Unlike in cell elongation, in cell division the inhibition of PG hydrolysis does not prevent further PG synthesis, allowing the cell to form a complete PG septum between the daughter cells which remain connected due to the lack of septum PG cleavage activity (Fig. 1A) (Heidrich et al., 2001). Recent work from T. Bernhardt's laboratory provided insights into the activation of PG hydrolases active during cell division. The cell division proteins FtsN and FtsEX recruit and/or activate EnvC and NlpD, respectively, both of which are catalytically inactive versions of LytM endopeptidases that activate three septum-splitting amidases depending on ongoing septum PG synthesis (Uehara et al., 2010). EnvC activates AmiA and AmiB, and NlpD activates AmiC. Remarkably, the ATPase function of FtsE, which constitutes an ABC transporter with FtsX, is required for EnvC-mediated activation of AmiA and AmiB, indicating that cleavage events in the PG are coupled to the cytoplasmic hydrolysis of ATP (Yang et al., 2011). A similar mechanism exists in the Gram-positive Streptococcus pneumoniae where FtsEX activates the predicted PG hydrolase PcsB (Sham et al., 2011). The allosteric activation of E. coli AmiB requires the removal of an α-helix to open the active site cleft for substrate binding, a mechanism that appears to be conserved between septum PG cleaving amidases (Yang et al., 2012). Perhaps also the new hydrolases essential for cell elongation, Spr, YdhO and YebA, require similar activation by (an)other protein(s) explaining their apparently low in vitro activity against PG sacculi (Singh et al., 2012). This possibility is supported by the solution structure determined by nuclear magnetic resonance (NMR) spectroscopy of Spr showing the catalytic Cys68 residue is partly covered by loops surrounding the active site (Aramini et al., 2008), in contrast to the wide open, unregulated active site of the homologous Tse1 which functions to lyse target cells (Chou et al., 2012). Perhaps Spr can be activated by allosteric displacement of loops surrounding its active site, similarly to the way in which AmiB is activated. In future work it will be important to dissect how the endopeptidases and other PG hydrolases are regulated and how they contribute to PG growth in E. coli and other bacteria. Potentially, some of the essential PG hydrolases represent new targets for antimicrobial drugs.