Introduction

In the age of antibiotic resistance, multi-resistant bacteria pose a serious threat to global health. This calls for novel strategies to fight the infections¹. Contemporary means to treat bacterial infections rely on antibiotics. One of the most common mechanisms of action of antibiotics is inhibition of the peptidoglycan (PG) synthesis, a major macromolecular structure in the bacterial cell wall^2,3. Alarmingly, the Gram-positive bacterium Staphylococcus aureus (S. aureus) is notorious in developing resistance towards β-lactam based antibiotics e.g., penicillin and their derivatives, which target penicillin binding proteins (PBPs) that are vital for PG synthesis⁴. These methicillin-resistant strains of S. aureus (MRSA) can cause life-threatening infections which are very difficult to eradicate⁵. Moreover, although still rarely occurring, outbreaks of infections caused by vancomycin intermediate-level and fully resistant S. aureus (VISA/VRSA) are lurking on the horizon⁶. Development of alternative means to treat multi-resistant bacterial infections is therefore needed.

Cell division, cell shape determination, PG remodelling and recycling is administrated and executed by peptidoglycan hydrolases (PGHs), enzymes produced to function as (auto)lysins in the regulation of the cell wall during growth and division^7,8. On the other hand, PGHs may also function as defensive weapons against other bacterial species as in the case of lysostaphin, an exolysin secreted by S. simulans biovar staphylolyticus, which displays bacteriocidic action against competing S. aureus⁹. The potential chemotherapeutic role of PGHs is based on the targeted destruction of the protecting PG by hydrolysis, which leads to growing turgor pressure and thus lysis of the bacterial cells^7,9. Given that PGHs are promising bacteriocins as well as druggable targets for the treatment of multidrug-resistant S. aureus infections, profound knowledge of their structure, function as well as substrate specificity is instrumental to harness their full potential as a new breed of antibiotics¹⁰.

S. aureus and other Gram-positive bacteria are protected by a thick PG layer that is composed of repeating β-1,4 linked N-acetylmuramic acid (MurNAc) and N-acetylglucosamine (GlcNAc) disaccharide units, forming the conserved glycan backbone of murein^7,11 (Fig. 1A-B). Each MurNAc carboxyl group is linked to a stem peptide (L-Ala-D-iso-Gln-L-Lys-D-Ala-D-Ala) and two stem peptides are connected via a cross-bridge structure, the exact composition and length of which depends on the bacterial species in question. In S. aureus the cross-bridge is characteristically composed of five glycine residues⁷. Cross-bridging provides an integral structural support for the entire PG wall and allows PG to reach a thickness of over 40 layers (20–80 nm) in Gram-positive bacterial species^7,12.

Figure 1.

While cell wall composition and biosynthesis of PG are relatively well understood, the perception of PG maintenance and hydrolysis is vague¹⁰. One prominent PGH family are the lysostaphin-like endopeptidases, which specifically target the cross-bridge peptide bonds of S. aureus PG^13,14. Here we compared LytM and lysostaphin (LSS), which both belong to the zinc-dependent M23 family of metalloendopeptidases and are designated as glycyl–glycine hydrolases. They have been shown to cleave the peptide bonds between glycine residues in the pentaglycine cross-bridge of S. aureus PG (Fig. 1B)^9,15–20.

S. simulans LSS, the founding member of the M23 endopeptidase family, is an exolysin possessing bacteriolytic properties towards staphylococci with a pentaglycine cross-bridge structure (e.g. S. aureus, S. carnosus, S. cohnii) (Schindler & Schuhardt, 1964; Schleifer & Fischer, 1982). LytM is a S. aureus autolysin with a catalytic M23 domain structurally very similar to that of LSS (Fig. S1). The common fold consists of a characteristic narrow groove formed by a β sheet and four surrounding loops. At one end of the groove resides the catalytic site in which a zinc cation is coordinated by two conserved histidines and an aspartate. The zinc cation, which polarises the peptide bond, and a nucleophilic water molecule activated by two other conserved histidines act in concert to hydrolyse the substrate glycyl-glycine bond²².

Despite tens of years of effort in studying substrate specificities of LSS-like M23 endopeptidase family members, the exact molecular targets of these enzymes have remained enigmatic, contradictory, and yet imprecise. Interpretation of results is further complicated by diverse conventions used for the numbering of cross-bridge residues. Indeed, the exact cross-bridge bond that LSS targets remains controversial; Browder et al. (1965) reported LSS cleavage between glycines 4 and 5 while Sloan et al. (1977) observed cleavage of all glycyl-glycine bonds with several substrates of different sizes^23,24. Recently, using mass spectrometry (MS), Schneewind et al. (1995) observed LSS-mediated cleavage between glycines 3 and 4²⁵. This observation was supported by a recent NMR spectroscopic study highlighting that hydrolysis by LSS produced fragments in which either two or three glycines are interlinked to the lysine sidechain²⁶. On the other hand, Xu et al. (1997) reported cleavage at several sites with digested muropeptides: between glycines 1 and 2, 2 and 3 as well as 3 and 4²⁷. Additional studies carried out with FRET peptides reported cleavage between glycines 2 and 3 (∼60 %) and glycines 3 and 4 (∼40 %)²⁰.

Much less is understood about the functional role of S. aureus autolysin LytM. This is partially due to lack of profound knowledge of its substrate specificity. Indeed, based on its high sequence homology with LSS catalytic domain as well as its association with S. aureus cell wall maintenance, LytM has implicitly been envisaged to target pGly cross-bridges in the S. aureus cell wall. LytM has been shown to hydrolyse pGly into di- and triglycine, and tetraglycine into diglycine²⁸. Understanding the substrate specificity is of utmost importance for deciphering the functional role of endogenous LytM in cell division, PG remodelling, antimicrobial resistance, and PG synthesis/hydrolysis in general.

Thus far studies of PG hydrolase specificity have typically utilised end-point kinetics together with LC/MS analysis of digested PG fragments or turbidity reduction assays to determine the site(s) and efficiency of hydrolysis^29,30. However, these approaches have several pitfalls e.g., they do not provide information on reaction kinetics or atomic resolution of reaction products, which severely limits determination of substrate specificity. In seeking to understand the functional role of these enzymes we have deciphered substrate specificity of LSS and LytM catalytic domains by taking a systematic approach by monitoring hydrolysis of various synthetised PG fragments as well as muropeptides from purified sacculus extracted from S. aureus cells using solution-state NMR spectroscopy and utilising turbidity reduction assay on living S. aureus cells. NMR enables following the hydrolysis reaction in real time as well as identification of hydrolysis products.

Here we show that the substrate specificities as well as catalytic efficiencies of M23 family PGHs targeting S. aureus PG are different from what has been earlier anticipated. Moreover, our results reveal that quite unexpectedly LytM substrate specificity extends beyond glycyl-glycine endopeptidase activity, which calls for revision of their classification.

Results

Substrate specificities of Lysostaphin M23 enzyme family

Due to inconsistent nomenclature used in the literature to annotate S. aureus PG cross-bridge structure and hence substrates used in the enzymatic assays and specificity studies, we conducted a systematic bottom-up approach to delineate the significance of various PG structural elements for the specific activities of LSS and LytM. We prepared a panoply of synthetic peptides faithfully replicating the chemical structure of the recognised S. aureus PG fragments (Fig. 2J, Table S1). The common denominator - the scaffold structure - in the PG fragments used in this study, is the pGly cross-bridge as it is uniquely present in S. aureus PG (Fig. 1B).

Figure 2.

Figure 2 highlights the strategy employed to study substrate specificity and kinetics of LSS and LytM using two different approaches; monitoring hydrolysis of synthetic peptides mimicking PG fragments and muropeptides extracted from S. aureus sacculus using solution-state NMR spectroscopy, and monitoring lysis of bacterial cell wall using turbidity reduction assay. In the turbidity reduction assay lytic efficiencies of externally administered LSS and LytM against S. aureus USA300 (MRSA) strain were compared (Fig. 2D). These data show that OD₆₀₀ of late stationary cells reduced from 100 % to 25 % in 16 hours for LytM and LSS.

Given that pGly has long been recognised as the common, non-redundant and yet physiological substrate unit for M23 family endopeptidases, we wanted to accurately define substrate specificities, catalytic efficiencies, and the sites of cutting for the catalytic domains of LSS and LytM. To measure the rate of pGly hydrolysis in vitro, we added the corresponding enzymes and monitored decaying substrate concentration with respect to reaction time using quantitative ₁H NMR spectroscopy (Fig. 2A). Results indicate that LSS hydrolyses pGly 15-fold faster than LytM. In vitro hydrolysis rates of pGly (Fig. 2B) and reduction in S. aureus USA300 cell turbidities (Fig. 2D) display consistent differences in catalytic efficiencies of LytM and LSS. However, we see that overall, the lytic efficiency of LSS and LytM against USA300 cells is higher than the rate of pGly hydrolysis in vitro. This suggests that S. aureus cell wall as a macroscopic substrate or cellular milieu differs significantly from the conditions used for the kinetic assay in vitro i.e., pGly as a substrate represents a poor model for describing substrate specificity and functional differences of M23 family endopeptidases. As the next step, we therefore determined the scissile bonds in pGly as well as extended substrate specificity studies of these enzymes beyond the pGly cross-bridge.

LSS and LytM preferentially hydrolyse the Gly₂-Gly₃ bond in pentaglycine

We recently showed using a two-dimensional ¹³C-HMBC NMR experiment that LytU hydrolyses pGly into di- and triglycine¹⁸. Whether this was the result of hydrolysis of the bond between Gly₂ and Gly₃, and/or between Gly₃ and Gly₄ in pGly remained undefined because the products are the same in both reactions. Selective ¹⁵N,¹³C-labelling (Fig. 2E) breaks the isotopic symmetry of pGly without introducing per se any non-physiological tags to the substrate and enables to define the preferred cleavage sites for LSS and LytM in pGly using ¹H-¹³C NMR spectroscopy. As can be appreciated in Figure 2G,H, isotopic labelling of Gly₂ at Cα and CO carbons in pGly allows unambiguous determination of cleavage site, because the characteristic chemical shift of a C-terminal ¹³CO resonance (179.4 ppm) is markedly different from a non-terminal ¹³CO chemical shift (173.8 ppm) at physiological pH. Representative two-dimensional ¹Hα-¹³CO correlation maps, collected with the glycine-optimised 2D HA(CA)CO NMR experiment (Fig. 2F) from the selectively Gly₂ ¹³C-labeled pGly are highlighted in (Fig. 2H). These data clearly show that LytM and LSS are highly specific for the bond between Gly₂and Gly₃ (> 85-94 %, Fig. 2H). However, residual cleavage activity towards the bond between Gly₃ and Gly₄ is also evident (< 15 %).

The substrate specificity of LytM and LSS is determined by the D-Ala-Gly cross-link between adjacent PG monomers

In the next phase, we sought to understand the role of plausible auxiliary contacts arising from stem peptides flanking the pGly cross-bridge for the substrate specificity of LSS and LytM. To this end, several larger peptides with PG specific extensions around the pGly scaffold were utilised (Fig. 2J). To alleviate complications arising from multiple sites of hydrolysis in pGly, we decided to monitor the rate of substrate hydrolysis rather than the rate of product formation, and to use a similar substrate concentration for all the PG fragments (∼0.4 mM) (Fig. S2). These data using substrates 1-7 are shown in Figure 2K. Normalisation of the absolute rates with respect to the common scaffold structure, pGly, should reveal on the substrate level the preference of LSS and LytM towards different PG fragments (Fig. 2L).

Clearly, the overall rate of hydrolysis for LytM and LSS increases with peptides that mimic cross-linked PG fragments (Fig. 2K). Indeed, cross-linking L-Lys-D-Ala dipeptide N-terminally to pGly in 2 increased the rate of hydrolysis with respect to 1 by a factor of 12 and 7.3 for LSS and LytM, respectively (Fig. 2L). Further elongation towards the N-terminus (ADiQKDA-GGGGG, 3) had only incremental contribution to the rate of overall substrate hydrolysis, 16.8 and 3.3 for LSS and LytM, respectively.

Next, we inspected the influence of a stem peptide linked to the C-terminus of pGly on the substrate hydrolysis rate. In PG fragment 4 the stem peptide is linked to the pGly via an isopeptide bond between lysine ε amino group and the C-terminus of pGly. The PG fragment 5 is similar but corresponds to the peptide moiety in a PG monomer (or pentaglycyl-Lipid II) (Fig. 2J). For LSS, the rate of hydrolysis increased by 5- and 4.5-fold for substrates 4 and 5, respectively. This increase in overall rate of substrate hydrolysis with respect to pGly is roughly two times smaller than for the linear PG fragments 2 and 3 that contain a D-Ala-Gly cross-link. For LytM, the outcome was in stark contrast to the results observed with LSS. Indeed, LytM activity towards PG fragments 4 and 5 dropped to 69 and 49 %, respectively, of that found for 1.

Next, we studied the significance of capping the pGly moiety from both its termini. PG fragments 6 and 7 both contain an N-terminal D-Ala-Gly cross-linkage in their structure but differ in 6 having a C-terminal tetra- and 7 a pentapeptide stem i.e., the latter has a D-Ala-D-Ala moiety in its structure. For LSS, 10-fold rate enhancements with respect to 1 for substrate hydrolysis were observed, comparable to enhancements with substrates 2 and 3, indicating that the moiety C-terminal to pGly does not contribute to LSS specificity. Hence, based on these results, it can be deduced that LSS strongly favors substrates that contain a D-Ala-Gly cross-link i.e., 2, 3, 6 and 7. In the case of LytM, rate enhancements of PG fragments 6 and 7 with respect to 1 were 5.5- and 3-fold, respectively, meaning that the branched stem peptide linked to the C-terminal end of pGly has a slight negative effect on LytM specificity. These data indicate that similarly to LSS, LytM is more specific towards substrates 2, 3, 6 and 7 i.e., PG fragments that contain a D-Ala-Gly cross-link.

In all, overall rates of substrate hydrolysis using comparative real time kinetics convincingly indicate that LytM as well as LSS prefer mature PG fragments as substrates i.e., PG units that contain a D-Ala-Gly cross-link between two stem peptides. In contrast to the mere end-point kinetics carried out in the past for LytM and LSS, our present results clearly demonstrate the preference of these enzymes for cross-linked PG fragments beyond redundant glycyl-glycine endopeptidase activity. However, the substrate specific real time kinetics assay does not tell anything about the site(s) of hydrolysis in the substrates. To understand substrate specificity and the physiological role of these PGHs, we inspected the outcome of these reaction assays in more detail.

LSS and LytM display different substrate specificities than earlier anticipated

NMR spectroscopy allows identification of atom connectivities within the substrate and products, enabling determination of site(s) of hydrolysis at atomic resolution. Figure 3A displays the LSS-catalysed hydrolysis of substrate 2, the simplest of our PG fragments that contains the cross-link between D-Ala and Gly₁ of pGly synthetised by the transpeptidase. Expansion of the region corresponding to the ¹Hα resonance of D-Ala in the ¹H spectrum of 2 shows that upon hydrolysis of 2, two products are formed as manifested themselves by increasing concentrations of KDAG₁ and KDAG₁G₂ (Fig. S3 shows the identification of the product peaks based on ¹³C-HMBC spectra). Owing to their different chemical structures, the products display separate peaks, resolved at 800 MHz ¹H field, which allows determination of individual reaction rates. Given that the ¹H method we used is quantitative the preferred site of hydrolysis in the PG cross-bridge can be identified based on product concentrations (Fig. 3C-E). These data show that in 2 LSS hydrolyses amide bonds between Gly₁-Gly₂ as well as Gly₂-Gly₃ with a clear preference for the first one.

Figure 3.

By coupling real time kinetics data with identification of reaction products, we determined the cutting sites and the reaction rates of the corresponding products for a total of seven PG fragments for LSS and the combined results of their analyses are shown in Figure 3E. The difference in panels C and E is that for the latter we considered only reactions which dominate for the particular PG fragment whereas in the heatmap representation the relative product concentrations at the end of the hydrolysis reaction are shown. These results are consistent with analyses of substrate hydrolysis rates in that they show that LSS clearly favors PG fragments with a D-Ala-Gly cross-link. Pentaglycine or other PG monomers in which Gly₁ is not cross-linked to the stem peptide of the adjacent PG fragment are hydrolysed at slower rate and the cutting site is shifted by one residue downstream compared to the cross-linked ones. In addition, the bond specificity of the hydrolysis reaction increases with D-Ala-Gly cross-linked PG fragments. In this case, LSS favors cutting the amide bond between Gly₁ and Gly₂ and the hydrolysis rate of this bond is several times higher than that of the Gly₂-Gly₃ bond, or those of Gly₂-Gly₃ and Gly₃-Gly₄ bonds in PG monomers devoid of D-Ala-Gly cross-link.

As we reckoned that D-Ala-Gly cross-link is critical for efficient hydrolysis of LSS substrates, we wanted to study how the length of the glycine containing cross-bridge influences substrate hydrolysis (Figs. S4, S5). Substrate 12, containing four consecutive glycines attached to Lys-D-Ala was hydrolysed between Gly₁ and Gly₂, as well as Gly₂ and Gly₃, exactly like substrate 2. However, the hydrolysis rate of the Gly₂-Gly₃ bond was 3-fold faster than in the case of 2. This is probably due to the absence of tetraglycine as a secondary substrate in the reaction i.e., the reaction products resulting from hydrolysis of 12 do not significantly compete with the primary substrate for binding to the LSS catalytic center. Substrate 13 (KDAGGG) was hydrolysed very slowly and 14 (KDAGG) not at all.

We carried out a similar analysis for LytM using the same set of PG fragments (Fig. 4B). The most striking observation is clearly visible in Figure 3B showing hydrolysis of 2 by LytM, which results in formation of products KDA and KDAG₁. Indeed, given that LytM is a well-established glycyl-glycine endopeptidase^17,22,28, we were taken by surprise to observe that LytM, in addition to Gly₁-Gly₂ bond hydrolysis, is also able to cleave the D-Ala-Gly₁ amide bond in the S. aureus PG cross-bridge whenever the cross-linked D-Ala-Gly structure is available.

Figure 4.

We verified that the observed D-alanyl-glycine bond hydrolysis, which has not been reported before this study, is stereospecific and linked to LytM activity. We performed an identical kinetic assay using KLAGGGGG as a negative control. It confirmed that the enzyme is indeed D-enantiomer specific as no cleavage between alanine and glycine was observed (Fig. S6). Further, using a protocol identical to that of LytM we engineered, expressed, and purified the well-established inactive H291A mutant of LytM and tested it with substrate 2. As expected, LytM H291A showed no activity. The results confirm that in addition to being a glycyl-glycine endopeptidase as reported earlier^17,22,28, LytM belongs to a small family of endopeptidases which cleave the peptide bond between D-Ala and Gly.

In substrates 2 and 6 LytM cleaves between D-Ala-Gly and Gly₁-Gly₂ with approximately equal rates. Using substrate 2 we did not observe a significant concentration-dependent change in the relative reaction rates of D-Ala-Gly and Gly₁-Gly₂ hydrolysis in the [S]/[E] range tested, from 725:1 to 1:2, when [LytM] = 50 μM. For the N-terminally longer substrates 3 and 7, the balance is shifted in favor of hydrolysis of Gly₁-Gly₂. Regarding substrates lacking the D-Ala-Gly cross-link, LytM seems to disfavor the stem peptide in the C-terminus of pGly and the overall rate of hydrolysis with respect to pGly is diminished more drastically in comparison to LSS.

Analogously to LSS, we compared the rate of hydrolysis of shorter PG fragments with LytM to map the critical length of the glycine-bridge in the substrate needed for the hydrolysis (Figs. S4, S5). Compared to 2, the rate of hydrolysis did not change drastically when shortening the glycine chain from five glycines to three. However, LytM strongly favored D-Ala-Gly cleavage in 13 (KDAGGG) whereas it preferred Gly₁-Gly₂ cleavage in 12 (KDAGGGG). Unlike LSS, LytM can still hydrolyse 14 (KDAGG). Only the D-alanyl-glycine bond is cleaved, and the reaction is five to ten times slower than hydrolysis of 2, 12 and 13. Owing to its D-Ala-Gly cleavage efficiency, we tested whether LytM could also hydrolyse KDAAGGGGG and KDAAA type substrates. However, the former was hydrolysed like pGly that is, no D-Ala-L-Ala cleavage occurred, but KDAAGG and KDAAGGG products were observed instead (not shown). Enterococci cell wall PG mimicking peptide KDAAA was not cleaved at all. This data further confirms that LytM is not only a glycyl-glycine endopeptidase but also acts as a D-alanyl-glycine hydrolase.

Hydrolyses of PG fragments mirror scissions in muropeptides extracted from S. aureus USA300 sacculus

To confirm the scissile bond specificity determined for LSS and LytM using synthetic PG fragment mimicking peptides, we extracted and purified muropeptides from S. aureus USA300 sacculus using the established protocol³¹. After administering muropeptide samples with LSS and LytM, we observed hydrolysis of the same amide bonds as with corresponding synthetic PG fragments (Fig. 2J). LSS hydrolyses the peptide bond between Gly₁ and Gly₂, recognised as appearance of peak of the C-terminal Gly₁ ¹Hα–¹³C’ resonance correlated with the corresponding ¹Hα–¹³C’ resonance of D-Ala (Fig. 3H). Identical correlations are observed for LytM, indicating that it also hydrolyses the amide bond between Gly₁ and Gly₂ (Fig. 3H). However, additional resonances stemming from the C-terminal D-Ala that is correlated with the neighboring Lys are also observed. This indicates that LytM is cutting both glycyl-glycine and D-alanyl-glycine bonds also in muropeptides extracted from S. aureus sacculus.

Differences in susceptibilities of S. aureus mutants towards Lysostaphin and LytM can be explained by their substrate specificities

Composition of the cell wall PG of Staphylococci, including e.g., S. aureus, S. simulans, S. epidermidis, can be modulated by intracellular enzymes. For instance, gene products of fem (factor essential to methicillin resistance) family members, i.e., femX, femA and femB catalyse nonribosomal insertion of mono-(Gly₅), di-(Gly₄-Gly₃) and di-(Gly₂-Gly₁) glycines into the glycine cross-bridge in S. aureus PG, respectively³². Reduced susceptibility of S. aureus βfemB and βfemAB mutant strains towards LSS has been observed earlier^30,33. We pondered whether the length of the cross-bridge of S. aureus could influence the catalytic efficiency or target bond specificity of LSS and LytM. To this end, we studied hydrolysis of PG fragments 8 and 9 mimicking tri- and monoGly cross-bridge composition of PG in S. aureus ΔfemB and βfemAB mutants, respectively (Fig. 4). Interestingly, LSS hydrolyses 8 with an increased rate as compared to 7. Furthermore, specificity increases i.e., cutting of glycyl-glycine bond, corresponding to the site “2” in 7 and site “6” in 8, is not observed. Quite surprisingly, we also observed that LSS is able to hydrolyse substrate 9, devoid of glycyl-glycine bond. Instead, LSS hydrolyses the Lys Nζ-monoGly C isopeptide bond. The rate of hydrolysis is, however, drastically lower (1.3%) than that of the amide bond between glycines in 7. This result is excellent agreement with earlier studies on S. aureus ΔfemB and βfemAB mutants and muropeptides extracted from these strains. Indeed, lytic efficiency of LSS against ΔfemB mutant is not drastically reduced while the minimum inhibitory concentration (MIC) is three orders of magnitude higher for the βfemAB mutant³². A dramatic reduction of LSS lytic performance against βfemAB mutants was likewise observed in a turbidity reduction assay³⁰. However, residual lytic activity of LSS can be explained by its ability to hydrolyse the Lys Nζ-monoGly isopeptide bond existing in the βfemAB mutant.

Similar results were obtained with LytM except for the scissile bond specificity. Even if the number of glycines in the cross-bridge is reduced from five to three, LytM is still able to hydrolyse both D-Ala-Gly and glycyl-glycine bonds (sites “1” or “5”) (Fig. 4). This is in accordance with results on shortened linear peptides (Fig. S5) as well as with previous turbidity reduction assay data on ΔfemB mutant³⁰. However, if stem peptides are crosslinked with only a single glycine in the cross-bridge (9), LytM displays diminished activity by 8-fold in comparison to 7. Yet somewhat surprisingly, considering the cleavage of D-Ala-Gly bond in substrate 14 (KDAGG, Fig. S5), LytM hydrolysed the Lys Nζ -monoGly C isopeptide bond. This again provides rationale for non-negligible lytic activity observed for LytM on βfemAB mutants although devoid of glycyl-glycine peptide bonds in their cross-bridge³⁰.

Next we tested the activity of LSS and LytM on PG fragments originating from so-called lysostaphin immunity factor (Lif/epr)-containing strains of S. aureus, that is the significance of serine substitutions in the PG cross-bridge on substrate hydrolysis^34–36. Of the two PG fragments, 10 and 11, having Gly₂ or Gly₃ replaced by a serine, only 11 was hydrolysed by LSS. LSS cleaved the bond between Gly₁ and Gly₂ in 11 although with a rate 30-fold slower than that for substrate 2. The results with LSS correlate well with earlier observations that show only fractional lytic activity of LSS towards Staphylococcal strains with serine in the cross-bridge³⁰. LytM was able to hydrolyse both PG fragments. 11 was cleaved similarly to the PG fragment 2, whereas LytM specifically cleaved the D-Ala-Gly amide bond in 10 as it contains serine in the second position. The rates of substrate hydrolysis were not drastically lower i.e., by a factor of 4.3 (10) and 1.5 (11) in comparison to 2. These data provide rationale to the observed differences in lytic efficiencies of LSS and LytM on S. aureus cells having serine substitutions in the cross-bridge³⁰. Hence, given that LytM exhibits significant catalytic activity towards D-Ala-Gly cleavage, it is less susceptible to serine substitutions in the cell wall PG. LytU hydrolysed only the amide bond between Gly₃-Gly₄ in 10 and no activity was detected with 11 as serine substitution for Gly₃ results in non-scissile bonds Gly₂-Ser₃ and Ser₃-Gly₄ in 11. The effect of serine in a fragment containing an isopeptide bond was not tested in this study.

Structural level perspective on deviating substrate specificities for LSS and LytM

We combined the detailed substrate-level information in terms of real time reaction kinetics and scissile bond specificities together with the existing structural models available for LSS and LytM to glean structural-level understanding of enzyme specificities. Based on these data and using the nomenclature formulated by Schechter and Berger³⁷, we delineated substrate specificities for LSS and LytM (Fig. 5). It is clear that LSS is a glycyl-glycine endopeptidase as P1 and P1’ positions are invariably occupied by Gly residues. However, the rate of hydrolysis increases when D-Ala occupies the P2 position, that is when LSS recognises a D-Ala-Gly cross-link in the cell wall. LytM is flexible regarding the P1 site, it can be accommodated either by D-Ala or Gly, whereas the P1’ position is invariably occupied by Gly.

Figure 5.

Why does LytM hydrolyse a D-Ala-Gly bond but LSS does not? To address this intriguing question, we utilised the existing structure of LytM in complex with the transition state analog tetraglycine-phosphinate ²² and used molecular modelling approach to dock PG fragment 2 into the active sites of LSS and LytM. As has been proposed earlier, the Zn²⁺ at the active site polarises the scissile bond by coordinating the carbonyl oxygen of a residue in the P1 position. For LytM this is either D-Ala or Gly, and for LSS it is Gly (Fig. 5). D-Ala in the P1/S1 position in the LSS active site results in a steric clash with loop 1, most notably its residues L272-I274, thus preventing hydrolysis of a D-alanyl-glycine peptide bond. In LytM the corresponding loop is shorter allowing accommodation of D-Ala in the P1/S1 position and hence cleavage of D-alanyl-glycine cross-bridge or alternatively, if the substrate has a Gly in the P1 position, the peptide bond between Gly₁ and Gly₂.

The same loop 1 gates the LSS active site and establishes the structural basis for Lif function. Serine in the P2’ position severely hinders accommodation of the substrate to the LSS active site, which results in a drastic drop in the rate of hydrolysis (Fig. 5). In LytM the three residues shorter loop renders the active site more voluminous, which permits the short polar sidechain of Ser to fit in to the P2’ or P3’ positions. Consequently, substrates 10 (KDAGSGGG) and 11 (KDAGGSGG) can still be hydrolysed with relatively high efficiency (ca. 50%) in comparison to PG fragment 2 which contains a pGly bridge. Interestingly, as the P1’ position only allows Gly, 10 cannot be hydrolysed between Gly₁ and Ser₂. However, LytM is able to hydrolyse the D-alanyl-glycine bond due to the positioning of the Ser into the P2’ position. LSS cannot hydrolyse 10 as it would require the disallowed P1’ position to accommodate Ser or alternatively P1 to be occupied by D-Ala, both of which are prevented by the extended loop 1 structure in LSS.

Interestingly, when the D-Ala-Gly cross-link is missing e.g., in pGly or PG fragments 4 and 5, the latter mimicking a non-cross-linked PG monomer, the cleavage site is shifted within the glycine cross-bridge that is, cutting occurs between Gly₂ and Gly₃ since in minimum a substrate with four residues occupying P2, P1, P1’ and P2’ sites is required for both LSS and LytM. Owing to this limitation in substrate size and distinct difference in the substrate specificity of LSS and LytM, we observe that while both enzymes can hydrolyse tetraglycine, only LytM can cleave 14 (KDAGG) since it accepts D-Ala in the P1 position. We observed significant reduction in rate of hydrolysis for both LSS and LytM when the substrate is too short i.e., it cannot fill the P3’/S3’ position. This is probably due to the stabilising hydrogen bonds that are formed between the longer glycine chain and a key asparagine residue in loop 4, Asn372 and Asn303 for LSS and LytM, respectively (Fig. 5B, F).

Discussion

To gain novel insight into lysostaphin enzyme family specificities, we employed NMR spectroscopy to determine the substrate specificities using real-time kinetics at atomic resolution on substrates mimicking PG fragments. In addition, we verified the sites of hydrolysis on muropeptides purified from S. aureus USA300 sacculus and confirmed lytic efficiencies of LSS and LytM with turbidity reduction assays on bacterial S. aureus USA300 cells.

Previous research on substrate specificity is largely based on testing several bacterial strains with varying PG composition^17,30 either with lysis assays or purified PG. Thus, determination of the site of hydrolysis often relies on applying abductive reasoning to either include or exclude different PG structures according to the assay results. Because this type of experimental set up often does not reveal the direct cleavage site, nor if there are several cleavage sites it should be supplemented with methods allowing direct observation – such as NMR. Yet, the end-point kinetics type of enzymatic assays e.g., administering chromatographically purified muropeptides with a lytic catalyst and quenching the reaction after overnight incubation, followed by mass spectrometric analysis of reaction products, do not necessarily reflect substrate specificity accurately because they only detect reaction products at the end and do not consider plausible secondary reactions.

Our method is superior to the previous approaches because real-time kinetics data combined with a precise determination of the bond cut-off point reveal the true substrate specificity of LSS. Our data also explain the inconsistencies between results of previous studies, that is, our extensive set of PG fragments allowed a more comprehensive interpretation than previous studies with fewer substrates^{15,20,23–27,30,32,38,39}. LSS recognises the D-Ala-Gly cross-link. In such a substrate, D-Ala occupies the P2 position, which increases the rate of hydrolysis by 10-fold in comparison to substrates which position glycine into P2 (Fig. 5). As mature PG in S. aureus cell wall is highly (D-Ala-Gly) cross-linked, our results are in excellent agreement with the observed efficiency of LSS towards S. aureus cells in numerous studies in the literature ^30,40. We also showed that LSS can hydrolyse cell wall in the S. aureus ΔfemB mutants³³, having three glycines in the cross-bridge, as they contain cross-linked D-Ala-Gly which occupy the P2-P1 positions and two glycines accommodated in the P1’-P2’ positions (Fig. 5A). However, our data also demonstrate that LSS is capable of efficiently hydrolysing glycyl-glycine bonds in PG fragments with different levels of cross-linking, including also non-cross-linked PG monomers. In such a case, Gly₁ and Gly₂ house the P2 and P1 positions and the scissile bond is shifted from the preferable Gly₁-Gly₂ bond one residue forward to the Gly₂-Gly₃ amide bond. This confirms the findings made by Maya-Martinez and colleagues with non-cross-linked PG fragments i.e., LSS leaves two or three glycines connected to Lys sidechain²⁶.

Hence, our results provide rationale to the previous observations which show discrepancy regarding the LSS site of hydrolysis. To summarise, LSS prefers cutting between Gly₁ and Gly₂, whenever Gly₁ is cross-linked to D-Ala of neighboring stem, whereas it hydrolyses the amide bond between Gly₂ and Gly₃ in non-cross-linked (devoid of D-Ala-Gly bond) PG fragments.

LytM was originally categorised as a glycyl–glycine endopeptidase based on lytic experiments performed using purified cell walls, which showed that it is active against S. aureus and S. carnosus but not against Micrococcus luteus¹⁷. M. luteus PG has the following structure: Ala-(D-γGlu-Gly)-Lys[D-Ala-Lys-D-Glu(-Gly)-Ala-D-Ala-Lys-D-γGlu(-Gly)-Ala]-D-Ala and thus contains neither Gly-Gly nor D-Ala-Gly bonds¹⁴. As was suggested by Ramadurai et al. (1999)¹⁷ M. luteus lacked the bond necessary for LytM cleavage, and here we have identified that bond to be D-Ala-Gly in addition to the known specificity towards glycyl–glycine bonds.

The only other recognised D-Ala–Gly hydrolase in S. aureus is LytN. Contrary to LytM, it contains a catalytic CHAP domain, which cleaves both D-Ala-Gly and MurNAc-L-Ala bonds, in other words is a D-alanyl-glycine endopeptidase as well as an N-acetylmuramoyl-L-alanine amidase domain²⁹. D-Ala-Gly cleavage activity has also been found from a CHAP domain of a S. aureus Phage ϕ11 murein hydrolase. In this hydrolase amidase activity is contained in another domain of the modular protein ³⁹. Recently also Enterococcus faecalis EnpA, which belongs to the M23 metalloendopeptidase family, has been shown to cleave D-Ala-Ala bond in E. faecalis. D-Ala-Gly cleaving activity was also reported, although without quantitative information regarding catalytic efficiency. Given that the P1 site in EnpA can accommodate D-Ala, it exhibits substrate specificity similar to that of LytM, which allows occupation of the P1 position by D-Ala (and Gly). In this way, LytM and EnpA deviate from LSS which requires invariably glycine in the P1 position. On the other hand, similar to LSS, LytM allows only glycine at the P1’ site, whereas EnpA is more promiscuous and can accommodate Gly, L-Ala and L-Ser in this position^30,38.

The pivotal findings in this paper are the discovery of the D-Ala-Gly hydrolysis activity of LytM, previously designated solely as a glycyl-glycine endopeptidase. Yet, we show for the first time that the substrate specificity of LSS is defined by the D-Ala-Gly cross-link, which increases catalytic efficiency 10-fold with respect to pentaglycine.

Methods and materials

Reagents

Peptides were synthetised by CASLO as Trifluoroacetic acid (TFA) salts. Escherichia coli BL21 (DE3)pLysS cells were obtained from Novagen (Novagen). Staphylococcus aureus USA300 and Newman D2C RN4220 cells were purchased from American Type Culture Collection (ATCC). The S. aureus Newman conditional overexpressing LytU RH7781 mutant was previously described¹⁸.

Protein expression and purification

The catalytic domain of LSS (LSS_cat residues 248-384) was produced as a His-tagged GB1-fusion protein. LSS_catGB1 was cloned in pET15b vector (Novagen) and heat-shock transformed into E. coli BL21(DE3) pLysS (Novagen). The cells were grown at 37 °C with 250 rpm orbital shaking in Luria-Bertani broth supplemented with 100 μg/ml ampicillin until the OD₆₀₀ reached 0.55-0.6. Then the temperature was lowered to 20 °C and protein expression was induced by adding 0.4 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). The overexpression of the proteins occurred at 20 °C for 20 hours. The cells were harvested by centrifugation at 6000 g for 20 min at 20 °C followed by cell resuspension in 1x PBS buffer and stored in -80 °C until purification. Complete lysis of the cells was achieved using Emulsiflex-C3 homogeniser (Avestin) and lysates were cleared by centrifugation at 35,000g for 30 min at 4 °C. The 6x-His-tagged proteins were captured using Ni-NTA Superflow (QIAGEN) and the His-tag was cleaved at 4 °C for 16-18 hours by human thrombin (Biopharm laboratories). HiLoad 26/60 Superdex 75 pg column (GE Healthcare) was utilised in the size exclusion chromatography of the desired fragments in 20 mM Sodium-phosphate buffer pH 6.5, 50 mM NaCl using an ÄKTA pure system (GE Healthcare).

LytM catalytic subunits (LytM_cat residues 185-316) were cloned into pGEX-2T vector and heat-shock transformed into E. coli BL21(DE3)pLysS (Novagen). The GST fusion LytM_cat protein was produced the same way as LSS_catGB1. The GST fusion proteins were purified with Protino Glutathione Agarose 4B (Macherey-Nagel) and GST was cleaved by human thrombin (Biopharm laboratories) at 4 °C for 16 h. The protein was further purified by size exclusion chromatography (SEC) with a HiLoad 26/60 Superdex 75 pg column (GE Healthcare) in SEC buffer (20 mM Sodium-phosphate buffer pH 6.5, 50 mM NaCl) using an ÄKTA pure chromatography system (GE Healthcare). Eluted fractions were concentrated using Amicon^® Ultra to 1 mM final concentration.

The inactive mutant of LytMcat (LytMcat_H291A) was generated using QuikChange II Site-Directed Mutagenesis kit (Agilent Technologies) and the mutation was verified by sequencing. The recombinant proteins were produced and purified same as the wildtype.

The ¹⁵N-labelled LytM_cat, and LytM_cat_H291A were expressed in E. coli BL21(DE3)pLysS cells in standard M9 minimal medium using 1 g/l ¹⁵N NH₄Cl (Cambridge Isotope Laboratories) as a sole nitrogen source. The proteins were purified in 50 mM NaH₂PO₄, pH 6.5, 50 mM NaCl using the same protocol as described above for the unlabelled proteins.

Muropeptide extraction

Muropeptides were extracted as before described with modifications ³¹. 2 mL of overnight bacteria culture were centrifuged at 10,000 rpm in table microcentrifuge, supernatant was discarded, and bacteria were resuspended in 4 mL of 100 mM Tris HCl pH 6.8, 0.25% Sodium dodecyl sulfate (SDS) to reach OD₆₀₀ equal to 10. After boiling SDS was removed with extensive washes. Cells were then solubilised with 1 mL of dH₂O and incubated for 30 min at room temperature in sonifier waterbath. 500 μL of 15 μg/mL DNase in 0.1 M Tris HCl pH 6.8 solution was added and sample was incubated 1 hour at 37 °C, 150 rpm. 500 μL of 4 mg/mL Pronase (Sigma) were added and sample was incubated overnight at 37 °C, 150 rpm. After enzymes inactivation, the pellet was resuspended with 500 μL of 1 M HCl solution and incubated for 4 hours at 37 °C, 150 rpm to release the teichoic acid. The sample was then washed with dH₂O until pH was between 5 and 6. Finally, the pellet was resuspended with 12.5 mM sodium dihydrogen-phosphate, pH 5.5 or 20 mM sodium phosphate buffer pH 6.5 and adjusted to OD₅₇₈ equal to 3; Mutanolysin (Sigma) solution 5000 U/mL was added and incubated for 16 hours at 37 °C, 150 rpm. After mutanolysin inactivation, sample was centrifuged at 10,000 rpm in table microcentrifuge for 10 min, pellet was discarded and muropeptides were in the supernatant. Muropeptides were dried using Savant SC110A SpeedVac (Thermo Fisher Scientific), resuspended with D₂O and pH was adjusted to 7.5 using deuterated sodium hydroxide. NMR sample was prepared by diluting the muropeptides stock solution in 50 mM dTRIS 7.5, and 0.1 mM DSS as reference compound.

Sample preparation for NMR kinetics

The synthetic peptides were purchased from CASLO. The peptides were resuspended in D₂O at the concentration of 20-40 mM, except for Gly₂ position ¹³C-labeled pGly which was 6.5 mM. For all peptides the pH was adjusted to 7.5 using deuterated sodium hydroxide. Peptides were diluted to a desired concentration in presence of 50 mM deuterated TRIS pH 7.5. For each sample, prepared to 3 mm OD round bottom NMR tube, 0.1 mM Sodium trimethylsilylpropanesulfonate (DSS, Chenomx Internal standard, 5 mM 99,9 % D, lot PS20190624) was added as a reference compound. Reactions were initiated by adding the enzyme to a final concentration of 2 μM or 50 μM.

NMR-based kinetics and resonance assignment of synthetic PG fragments and muropeptides from S. aureus sacculus

All NMR experiments were carried out at 25 °C, and at the field strength of 800 MHz of ¹H frequency on a Bruker Avance III HD NMR spectrometer, equipped with cryogenically cooled ¹H, ¹³C, ¹⁵N triple-resonance TCI probehead. NMR data collection for kinetics measurements employed a standard ¹H pulse program (zgpr) having a selective radiofrequency field for residual HDO signal presaturation during the recycle delay. To ensure quantitative detection of substrate and product concentrations, a 20 second long recycle delay was used between the transients used for the signal averaging. 0.1 mM DSS was used as a reference compound both for the peak integration, chemical shift referencing as well as lineshape optimisation. For each time point, experiment was accumulated with 24 transients, yielding an experimental time of 8 minutes per time point.

First, a reference ¹H spectrum was acquired. The sample was then removed from the magnet, and the enzyme was added. The time of enzyme addition was recorded as t=0. After the enzyme addition the sample was placed back into the magnet, and the shim was manually readjusted before the start of acquisition of consecutive ¹H spectra to follow the hydrolysis. This preparatory work resulted in a delay of about 5-10 minutes between enzyme addition (t=0) and the end of the acquisition of the first spectrum. This delay was accounted for in the data analysis for each experiment.

For the resonance assignment of NMR ¹Hα, ¹³Cα and ¹³CO chemical shifts in selectively ¹³C-labeled Gly₂ position in pentaglycine (1), a glycine Hα-detection optimised HCACO -type NMR experiment was devised (Fig. S7). The spectrum was collected as a 2D H(CA)CO ¹H-¹³C correlation experiment to establish connectivities between ¹Hα and ¹³C’ resonances in Gly₂. The experiment was measured using 2 transients per FID and the overall experimental time was 200 seconds per time point.

Assignment of chemical shift resonances in different synthetic PG substrates and products as well as muropeptides from S. aureus USA300 sacculus was based on the measurement of ¹H as well as 2D ¹H-¹³C HSQC, ¹H-¹³C HMBC, and ¹H-¹³C’ selective HMBC spectra. Typically, ¹H-¹³C HMBC experiments were measured overnight and at the end of the reaction to warrant the highest sensitivity for product chemical shift assignment.

Data analysis of NMR kinetics

To obtain substrate and product concentrations at each time point, NMR resonances were integrated together with reference compound using Topspin 3.6.5 software package (Bruker). Rates of reactions were calculated using linear regression of the first 40-60 min of the reaction. Goodness of fitting was evaluated by using the R² value. All the data fitting had R² >= 0.9. Rates of the reaction versus substrate concentrations were plotted and Michaelis-Menten fitting was performed using GraphPad Prism software version 9.5.1 (GraphPad Software Inc.).

Staphylococcus aureus cell growth conditions

Staphylococcus aureus USA 300 cells were grown overnight in Tryptic Soy Broth (TSB) and Newman RN4220 and RH7781 were grown in Lysogeny Broth (LB) at 37 °C, 200 rpm by inoculating one single colony in 3 mL of medium. Bacteria were then diluted 1:100 in prewarmed TSB and grown until desired OD₆₀₀.

Turbidity reduction assay

Turbidity assay was performed as previously described with some modifications ¹⁸. S. aureus USA 300 bacterial cells were incubated at 37 °C, 200 rpm until OD₆₀₀ between 6 to 8 corresponding to late stationary phase. Bacteria were then washed twice using 20 mM Tris HCl pH 7.5, 50 mM NaCl, and resuspended at OD₆₀₀ equal to 5. Bacteria were plated in round-bottom 96-well plate (Thermo Fisher Scientific), and the reaction was started by adding the enzymes at final concentration of 50 μg/mL. The assay was carried out in final volume of 100 μL. Bacteria without any enzyme was used as control sample. Reduction of turbidly of bacteria suspension was followed using BioTek Epoch2 Microplate Spectrophotometer (Agilent Technologies), at 25 °C for 16 hours with continuous shaking at 500 cpm. Data were expressed as normalised reduction of the turbidity over the time. Each reaction was carried out in quadruplicate.

Docking

Substrate 2 for docking was built with Maestro molecular modeling software (Schrödinger, 2021) available in Schrödinger (LLC, New York, NY) and Ligprep ligand preparation tool was used to refine the structure with force field OPLS4. The crystal structures of LytM (PDB code: 4ZYB, (23)) and lysostaphin (PDB code: 4QPB, ³⁹) were retrieved from the Protein Data Bank. The protein structures were first prepared with the Protein preparation wizard available in the Schrödinger suite. Protein preparation wizard was used to add missing hydrogen atoms, delete water molecules, and assign correct bond orders with force field OPLS4. Glide ^42–44 was used for docking. The receptor grid was generated with Glide and the Standard Precision (SP) Peptide mode was used for docking. It was noticed that the scoring function favors the strong interaction between the carboxylic acid of the substrate C-terminus and zinc ion, and thus the C-terminus was capped with an N-methylacetamide (NMA) residue. The serine containing peptides were modeled in the binding site by mutating the appropriate residues with Pymol ⁴⁵.

Acknowledgements

We thank Dr. Maarit Hellman for expert technical assistance. This work was supported by the grants from the Academy of Finland and Jane and Aatos Erkko foundation. Part of the figures were made using BioRender.com.

Conflict of Interests

The authors declare that they have no conflict of interest.