Turnover chemistry and structural characterization of the Cj0843c lytic transglycosylase of Campylobacter jejuni

Abstract

The soluble lytic transglycosylase Cj0843c from Campylobacter jejuni breaks down cell-wall peptidoglycan (PG). Its non-hydrolytic activity sustains cell-wall remodeling and repair. We report herein our structure-function studies probing the substrate preferences and recognition by this enzyme. Our studies show that Cj0843c exhibits both exolytic and endolytic activities and forms the N-acetyl-1,6-anhydromuramyl (anhMurNAc) peptidoglycan termini, the typical transformation catalyzed by lytic transglycosylase. Cj0843c shows a trend toward a preference of substrates with anhMurNAc ends and those with peptide stems. Mutagenesis revealed that the catalytic E390 is critical for activity. In addition, mutagenesis showed that R388 and K505, located in the positively charged pocket near E390, also serve important roles. Mutation of R326, on the opposite side of this positively charged pocket, enhanced activity. Our data point to different roles for positively charged residues in this pocket for productive binding of the predominantly negatively charged PG. We also show by X-ray crystallography and by MD simulations that the active site of Cj0843c is still capable of binding GlcNAc containing di- and trisaccharides without MurNAc moieties, without peptide stems, and without the anhMurNAc ends.

Graphical Abstract

Introduction

Lytic transglycosylases (LTs) are key bacterial periplasmic enzymes responsible for the degradation of cell-wall peptidoglycan (PG) strands ^{1, 2}. This activity can be either endolytic or exolytic, resulting in cleavage in either the middle or at the ends of PG strands, respectively ^{3, 4}. Bacteria need such lytic enzymes for the purpose of remodeling the cell wall (for assembly of large protein machines such as the flagellum and secretion systems), cell division, repair of a damaged cell wall, etc. LTs can be either membrane-bound or soluble. Some soluble LTs are doughnut-shaped with the active-site groove facing toward the central opening of the doughnut ^4–7. The active sites of LTs harbor a catalytic glutamate or aspartate residue ¹. The catalytic cycle of this class of enzymes has been investigated at the quantum mechanical/molecular mechanical (QM/MM) level ⁸. The process entails protonation of the glycosidic linkage by the aforementioned acidic residue and the formation of an oxocarbenium-ion transition state, which entraps the C-6 hydroxyl of the muramic acid to give rise to the N-acetyl-1,6-anhydromuramyl (AnhMurNAc) product.

The present study focuses on the soluble LT Cj0843c of Campylobacter jejuni, a major foodborne pathogen. We recently determined its crystal structure in the absence and presence of the inhibitor bulgecin A ⁷. In the same study, we also probed PG binding in the active site of Cj0843c using molecular-dynamics (MD) simulations. These simulations pointed to a possible role for a positively charged pocket, pocket 2, adjacent to the active-site groove. This pocket was found to form salt-bridge interactions with the carboxyl groups of the peptide stem of the terminal disaccharide unit both when the PG is bound in the substrate-binding position (occupying saccharide subsites +2, +1, −1, −2, −3, etc.) and the product-binding position (occupying subsites −1, −2, −3, etc.). In addition, the MD-simulation insights showed that pocket 2 could home in a PG strand traveling through the doughnut opening by attracting the terminal peptide stem, thereby aligning the terminal disaccharide with respect to the catalytic E390 to favor the exolytic activity ⁷. We herein tested these MD-derived hypotheses using PG from different sources, including synthetic PG fragments, to probe for exolytic- vs. endolytic-activity preference and the importance of peptide-stem attached to the PG, in particular as a part of the terminal anhMurNAc moiety. We also carried out site-directed mutagenesis experiments to probe the importance of pocket 2 amino acids on their postulated roles in peptide-moiety recognition. Finally, we explored the binding of substrate mimetics to the Cj0843c active site by X-ray crystallography.

Materials and Methods

Reactions of Cj0843c with sacculus and synthetic peptidoglycans

Cj0843c, expressed and purified as described previously ⁷, was incubated with Pseudomonal sacculus ³ in PBS for 22 h at 37 °C. The sacculus of P. aeruginosa is a good model substrate for Cj0843c as it shares many similarities with the C. jejuni sacculus yet does not have MurNAc O6-acetylation that C. jejuni has at ~12% of its MurNAc moieties; this acetylation makes PG resistant to LT action at that MurNAc position ^{9, 10}. The reaction was stopped by heating at 100 °C for 5 min. The resulting suspension was centrifuged at 13,000 g for 10 min, and the supernatant was analyzed by Ultra performance liquid chromatography mass spectrometry (UPLC/MS). The reactions of Cj0843c with six synthetic peptidoglycans were done at room temperature for 2 h in PBS. The reactions were stopped by the addition of 0.2% trifluoroacetic acid, and the resultant solutions were analyzed by UPLC/MS. The UPLC/MS condition and synthetic methodologies of six peptidoglycans were performed per previously reported methods ^{3, 4, 11–14}. Structure elucidation of muropeptide products from reaction of MltA with the sacculus was reported previously in detail ¹⁵ and the product profile of MltA was used as a standard for that of Cj0843c. The retention times, and high-resolution accurate masses (HRMS) of each product was used to assign the structure in this study.

Products from the reactions of Cj0843c with synthetic peptidoglycans were characterized by comparison of the retention times of LC and HRMS to those of authentic synthetic standards, when available. When the authentic synthetic standard was not available, we used HRMS, retention time and LC/MS/MS to deduce the structure. A similar strategy was used previously ¹¹ to analyze the reaction profile of lytic transglycosylase MltB from E. coli and RlpA from P. aeruginosa.

Differential scanning fluorimetry (DSF)

For the crystallography and DSF thermal shift assays, the Cj0843c protein was expressed and purified as above, except that Co-NTA beads (Thermo Fisher) were used instead of Ni-NTA beads for the initial His-tag purification step. The DSF ligand binding experiments were carried out similarly as described for bulgecin A binding ⁷. In the 30 μl reaction volume, the Cj0843c protein concentration was 10 μM in 100 mM HEPES pH 8.0, 200 mM Ammonium Acetate, and 10X SYPRO Orange dye. N,N′-diacetylchitobiose, and N,N′,N″-triacetylchitotriose (purchased from Sigma-Aldrich) concentrations were varied from 1–9 mM. Experiments were done in duplicate, and the fluorescence signal was read out on a CFX96 Touch ThermoCycler (Bio-rad).

Data collection and crystallographic refinement

Crystals of Cj0843c were grown using the sitting drop method. The protein was concentrated to 10 mg/ml in 10 mM HEPES pH 8.0 and 200 mM ammonium acetate; 1 μl of protein was mixed with 1 μl of the reservoir containing 100 sodium citrate buffer with pH ranging from 5.1–6.0, and 25–33% PEG 600. Crystals grew over several days and were used for soaking the disaccharide N,N′-diacetylchitobiose (5 mM for 60 minutes), and trisaccharide N,N′,N″-triacetylchitotriose (10 mM for 105 minutes) in mother liquor prior to freezing the crystals in liquid nitrogen for data collection. Data were collected at the NSLS beamline FMX and processed using XDS ¹⁶ (see Table 1 for data collection statistics). The structures were solved via molecular replacement with Phaser ¹⁷ using the apo Cj0843c coordinates as the search model (PDB ID 6CF9 ⁷). The disaccharide and trisaccharide Cj0843c complexes were refined to 2.58 and 2.31 Å resolution, respectively. The former complex contains Cj0843c residues 18–533, and the latter includes residues 17–533. The structure was refined using Refmac ¹⁸; the model building was done using COOT ¹⁹. After initial refinement, active site density for the saccharides was located in two separate active regions in both structures. Refinement parameters for the ligands were generated using AceDRG ²⁰, and the ligands were included in subsequent refinement steps. An additional acetate (binding near R392) and citrate molecule (distant from the active site) were observed in the density maps and included in the refinement. Molecular figures were generated using Pymol (www.pymol.org). Coordinates and structure factors for the di- and trisaccharide Cj0843c complexes have been deposited with the Protein Data Bank with PDB ID codes 7LAQ and 7LAM, respectively.

Table 1.

X-ray diffraction data collection and crystallographic refinement statistics for the Cj0843c complexes with the tri- and disaccharides.

	Cj0843c trisaccharide-soaked complex structure	Cj0843c disaccharide-soaked complex structure
Wavelength (Å)	0.97935	0.97935
Resolution range (Å)	29.71–2.31 (2.37–2.31)	29.50–2.58 (2.65–2.58)
Space group	I23	I23
Unit cell (Å, °)	178.2 178.2 178.2 90 90 90	177.0 177.0 177.0 90 90 90
Total reflections	551,805	403,267
Unique reflections	41,293 (2,871)	28,913 (2,029)
Multiplicity	13.4 (11,5)	13.9 (11.7)
Completeness (%)	99.5 (94.3)	99.5 (94.6)
Mean I/sigma (I)	13.0 (0.9)	13.0 (2.5)
CC1/2	0.998 (0.593)	0.997 (0.782)
R-merge (%)	14.9 (234.6)	17.0 (101.8)
Resolution refinement (Å)	29.71–2.31 (2.39 – 2.31)	29.50 – 2.58 (2.68 – 2.58)
Reflections used in refinement	41,183 (4,023)	28,908 (2,782)
Reflections used for R-free	2,039 (204)	1,574 (166)
R-work	0.187 (0.310)	0.182 (0.240)
R-free	0.223 (0.366)	0.226 (0.321)
Number of non-hydrogen atoms	4,617	4,544
Macromolecules	4,286	4,276
Ligand	103	75
Solvent	228	193
Protein residues	517	516
RMS (bonds, Å)	0.008	0.008
RMS (angles, °)	1.51	1.52
Ramachandran favored (%)	97.7	97.3
Ramachandran allowed (%)	2.1	2.3
Ramachandran outliers (%)	0.2	0.4
Average B-factor (Å2)	53.8	45.1
Macromolecules (Å2)	54.0	45.3
Ligands (Å2)	60.4	50.4
Solvent (Å2)	48.6	38.2

Statistics in parentheses are for the highest-resolution shell.

MD simulations

The coordinates of the di- and trisaccharide complexes of Cj0843c were used as starting models for MD simulation. The crystallographically determined water molecules were kept in this starting model. Hydrogen atoms were added using the Schrodinger Maestro protein preparation step with the catalytic E390 set as protonated/neutral, per previous studies ^{7, 8, 21}, and separate simulations with this residue set as deprotonated/charged. The protein was put in a box of explicit SPC water molecules with a 10 Å buffer in each direction. NaCl ions were added to neutralize the overall charge of the protein, and additional NaCl ions were added to set the ionic strength to 0.15 M NaCl. The model was minimized and subsequently relaxed using default Desmond protocols prior to the 10 ns simulation step (at 310 K). The Desmond software as part of the Schrodinger 2018–4 version was used, and the default OPLS3e force field ²² was selected.

Mutagenesis

Mutagenesis to generate the E390Q, E390A, R326A, R388A, and K505A mutants of Cj0843c was done using the Q5 mutagenesis kit (New England Biolabs) with the wild-type (wt) cj0843c-bearing pET28b expression plasmids ⁷. The primers used for site-directed amino acid substitutions are described in Table S1. The desired mutations in the Cj0843c containing pET28b expression plasmids were validated using DNA sequencing of the entire coding sequence. Purification of wt Cj0843c and its variant proteins using His-tag purification and size-exclusion chromatography were carried out as previously described ⁷. Activity assays were carried out using the lysozyme EnzChek assay kit (ThermoFisher) at the 0.23 – 15 μM protein concentration range at pH 7.5. The fluorescence signal was measured per assay kit instructions using an iD3 Molecular Devices plate reader. Experiments were carried out in duplicate.

Results and discussion

The catalytic activity of Cj0843c toward several different PG substrates was probed to gain insights into its substrate specificity (Figures 1, 2, S1–S9, and Table S2). Cj0843c has been considered a lytic transglycosylase based on the structural similarity to Slt70 of E. coli and to Slt of P. aeruginosa. Its cell-wall degrading activity had been demonstrated only by turbidometry ⁷, but its chemistry was not demonstrated with either peptidoglycan or synthetic samples.

Figure 1.

Lytic transglycosylase activity of Cj0843c. (A) LC/MS traces of the reaction of Cj0843c with pseudomonal sacculus and of (B) MltA with pseudomonal sacculus. (C) The chemical structures of the ten most abundant reaction products. For TriTetraA2, the moiety in the box with dotted line of TetraTriA2 structure is replaced by the give structure.

Figure 2.

The lytic transglycosylase activity of Cj0843c with synthetic peptidoglycans. LC/MS traces of synthetic peptidoglycans (VIIIp, IVp, IVp*a, VIII, IV, and IVa) before (B, D, F, I, K, and M), after (C, E, G, J, L, and N) addition of the enzyme, and of the synthetic standards of products (H and O). Reactions of substrates with peptide stems (VIIIp, IVp, and IVp*a) are shown in left panels (B-H). Reactions of substrates without peptide stems (VIII, IV, and IVa) are shown in the right panels (I-O). VIIIp, IVp, and their products contain pentapeptide stem(s) while IVp*a, and its reaction product IIp*a contain tetrapeptide stem(s). Substrates without peptide stem(s) and their reaction products are given in parenthesis in panel A. p: pentapeptide, l-Ala-γ-d-Glu-m-Dap-d-Ala-d-Ala. p*:tetrapeptide, l-Ala-γ-d-Glu-m-Dap-d-Ala.

An initial attempt at using the C. jejuni sacculus for documentation of turnover products of Cj0843c was unsuccessful in that the products were being isolated in minute quantities, not amenable to definitive characterization. A complication for the C. jejuni sacculus reaction is that ~12% of the 6-hydroxyl group in muramic acid in C. jejuni are acetylated ¹⁰. 6-O-Acetylated muramic acid cannot form 1,6-anhydromuramic acid, the product of LT reaction. Similar findings have been reported in the characterization of the reactions of LtgA, and LtgD from Neisseria gonorrhoeae, whose sacculus is also extensively acetylated. The formation of LT products was significantly decreased with sacculus from the wild-type strain containing O-acetylation compared to that from strains lacking O-acetylation ²³. This O-acetylation at a particular site in the PG is thought to site-specifically control the LT activity ²⁴. Due to this complication, a sacculus preparation from P. aeruginosa was used as previously reported ³. Cj0843c turned over the pseudomonal sacculus well, which allowed for characterization of the products containing anhMurNAc with non-crosslinked and crosslinked peptide stems (Figures 1A, 1C, S1, and Table S2). TetraA1 and Tetra2A2 were two abundant products from the exolytic reaction. The product profile of MltA of E. coli is shown in Figure 1B as a control of an LT with an exolytic activity. This enzyme exhibits high specific activity with both non-crosslinked and crosslinked peptide-containing muropeptide, and its product whose structure elucidation was given in detail previously ¹⁵. Note that MltA and Cj0843c are not related in terms of sequence similarity, nor are they in the same LT family classification.

We further characterized the enzyme using six synthetic peptidoglycan compounds (VIIIp, IVp, IVp*a, VIII, IV, and IVa; Roman numerals specify the number of sugars, “p” for pentapeptide stem, “p*” for tetrapeptide and “a” for anhMurNAc). Six peptidoglycans were specifically designed and synthesized ^{3, 4, 11–14} to explore which structural moiety in PG plays an important role for substrate recognition by Cj0843c. The length of sugar (4 vs. 8), the presence/absence of peptide stem, and the nature of the terminal sugar (MurNAc vs. anhMurNAc) were attributes that were studied. Overall, Cj0843c accepts all octa- and tetra-saccharides as substrate, and the presence of peptide stems is not critical for the reaction (Figures 2, and S2–S9). Reaction of octasaccharide VIIIp would produce three different pairs of products (IIpa + VIp, IVpa + IVp and VIpa + IIp) by reactions at its three glycosidic bonds (colored in red, blue, and green in Figure 2A). The reaction of Cj0843c with VIIIp produced all six possible products (Figures 2B, 2C, S2, and S3), while IIpa is the abundant product. Intermediary products like IVp*a and IVp (products of endolytic reaction) were further converted to IIpa and IIp, in separate experiments (Figures 2D–2G, S4–S5). Turnover of the substrate containing anhMurNAc (IVp*a) appeared faster than the ones with MurNAc (IVp). Note that the IVp*a and IVp have different length peptide stems (tetrapeptide and pentapeptide stems, respectively), which would not appreciably influence the outcome ⁷. The enzyme also turned over the substrate without peptide stem(s) (VIII, IV, and IVa) shown in Figures 2I - 2N and S6–S9. Although Cj0843c was found to have some in vitro endolytic activity with synthetic PG substrates, it is not known whether this endolytic activity also occurs physiologically with heavily crosslinked PG. Perhaps under certain physiological conditions, such as under the challenge of β-lactam antibiotics, the resulting decreased PG cross-linking results in longer non-crosslinked PG strands that might allow endolytic cleavage by Cj0843c (in addition to exolytic cleavage).

Structural insights into PG recognition by LTs can often be had by using GlcNAc polymers of various lengths mimicking the PG saccharide backbone ^{25, 26}. Therefore, we first tested whether disaccharide N,N′-diacetylchitobiose and trisaccharide N,N′,N″-triacetylchitotriose could bind to Cj0843c by measuring protein thermal shifts upon binding. The ligands are di- and tri-GlcNAc polymers, respectively. These compounds are mimetics of the PG in that the backbone of repeating N-acetylated GlcNAc is present. The limitation is that the alternating MurNAc is approximated by a unit of GlcNAc. That is to say that the lactyl moiety at the 3-position is absent, to which the stem peptide would be appended. Differential-scanning fluorimetry (DSF) measurements showed that the melting temperate (T_m) of Cj0843c increased from 52.0 ± 0.28 °C in the absence of ligand, to 52.8 ± 0.28, 53.6 ± 0.28, 55.6 ± 0.28 °C in the presence of 1, 3, and 9 mM trisaccharide, respectively (Figure 3). For the disaccharide at these concentrations, the T_m values were 52.2 ± 0.0, 52.6 ± 0.0, and 53.2 ± 0.28°C, respectively (Figure 3). The trisaccharide thus stabilizes Cj0843c to a larger extent compared to the disaccharide. For comparison, bulgecin A was shown to induce a larger T_m increase of 9 °C upon binding Cj0843c ⁷.

Figure 3.

Differential scanning fluorimetry thermal shift assay of disaccharide and trisaccharide binding to Cj0843c. Experiments were done in duplicate, and a representative fluorescence derivative curve is shown for each saccharide concentration.

The crystal structures of the disaccharide and trisaccharide complexed with Cj0843c were determined to 2.58 and 2.31 Å resolution, respectively (Table 1). The crystal structure for the trisaccharide in complex with Cj0843c reveals two bound trisaccharides in two separate locations in the extended active-site region: trisaccharide 1 binds near F412, and trisaccharide 2 binds near R392 (Figure 4; Figure S10 for stereo). The same two binding sites are also observed for the disaccharide when complexed to Cj0843c (Figures 5 and S10). The tri- and disaccharide-binding modes have almost identical interactions in the Cj0843c active site (Figure S11A). The disaccharide positions correspond to the trisaccharides’ two GlcNAc residues closest to the catalytic E390 in both binding sites.

Figure 4.

Trisaccharides bound to the Cj0843c. (A) Cj0843c with two trisaccharide molecules bound in the active site. The two trisaccharides (labeled “1” and “2”) are shown in stick representation with carbon atoms in cyan. The N-terminal NU-domain (teal), U-domain (light blue), L-domain (pale red), and catalytic C-domain (grey) are shown. The catalytic E390 is depicted in spheres with carbon atoms in purple. (B) Unbiased omit |Fo|-|Fc| electron density showing two trisaccharide molecules bound. Electron density was calculated after removal of the trisaccharides from the model followed by 10 cycles of Refmac refinement. Density is contoured at the 3 σ level; the trisaccharide molecules (labeled “1” and “2” in cyan) are depicted in ball-and-stick model with carbon atoms in cyan. The saccharide subsite numbers are shown in bold. (C) interactions of trisaccharide 1. The carbon atom numbering 1–6 for GlcNAc in subsite −2 is shown. Hydrogen bonds are shown as dashed lines, and saccharide subsites are labeled in bold. Water molecules are depicted as red spheres. (D) interactions of trisaccharide 2. The carbon atom numbering 1–6 for GlcNAc in subsite +1 are indicated.

Figure 5.

Disaccharide electron density in the Cj0843c active site. An unbiased omit |Fo|-|Fc| electron density is shown, indicating two bound disaccharide molecules. Electron density was calculated after removal of the trisaccharides from the model, followed by 10 cycles of Refmac refinement. Density is contoured at the 3 σ level. The disaccharides (labeled ‘1’ and ‘2’ in green) are colored with green carbon atoms, and the saccharide subsites are labeled in bold.

The two separate trisaccharides bound in the active site together mimic Cj0843c binding a longer PG strand before cleavage. Previous MD simulation of PG fragments in the active site of Cj0843c provided insights into the PG binding sites ⁷. Superimposition of one of the MD snapshots with the trisaccharide-bound Cj0843c structure indicates that the trisaccharide 1 GlcNAc moieties are bound in the −4, −3, and −2 subsites, respectively; trisaccharide 2 GlcNAc moieties are situated in +1, +2, and +3 subsites, respectively (Figure 6). The space between the two trisaccharides, subsite −1, would be filled by a MurNAc moiety in the native PG strand. Instead, this space is occupied by five water molecules in the trisaccharide Cj0843c complex (Figures 4C and 4D).

Figure 6.

Superimposition of Cj0843c bound with two trisaccharide molecules and with the computational model of a PG fragment bound to the enzyme. The model of the Cj0843c with PG fragment was obtained from a snapshot of an MD simulation of Cj0843c with a decasaccharide PG substrate in the active site ⁷. The trisaccharides molecules are shown in ball-and-stick with carbon atoms in cyan and labeled as in Figures 4C and 4D. The PG strand is shown in stick model and contains alternating GlcNAc (‘G’) and MurNAc (‘M’) moiety with carbon atoms colored in dark blue; the terminal anhMurNAc is labeled ‘AnhM’. The peptide stems are colored with yellow carbon atoms. The saccharide subsites −6 → +3 are labeled below the saccharides in bold. Superimposition of MD simulation snapshot coordinates and trisaccharide Cj0843c complex involved catalytic-domain residues spanning 300–530. The r.m.s.d. was 1.0 Å for the 231 Cα atoms.

The trisaccharides make multiple interactions in the active site (Figures 4C and 4D). The N-acetyl moiety of GlcNAc in the −2 subsite interacts with the backbone nitrogen and oxygen atoms of M410 and Y463, respectively (Figure 4C). This GlcNAc also makes a water-mediated interaction via its O3 hydroxyl; this hydroxyl also hydrogen bonds to the ring oxygen of the adjacent GlcNAc in the −3 subsite. The GlcNAc in the −3 subsite hydrogen bonds to the backbone nitrogens of G468 and F469 via its N-acetyl oxygen atom. This GlcNAc makes two additional water-mediated interactions. The GlcNAc in the −4 subsite of this trisaccharide does not make any direct protein interactions. The GlcNAc moieties in the −3 and −2 subsites occupy the active-site groove forming hydrophobic interactions via both faces of the saccharides interacting with F412 and M410 on one side and G465 and G466 on the other side of the active-site groove.

Trisaccharide 2 also makes hydrogen bonds, but, unlike trisaccharide 1, only one of the saccharide hydrophobic faces makes van der Waals interactions in the active site (Figures 4D and S10). GlcNAc in the +1 subsite makes four hydrogen bonds: with E390 (via O4), Q408 (via O3), Q389 (via O6), and the backbone oxygen of E390 (via the N-acetyl nitrogen; Figures 4D and S10). This GlcNAc makes several additional water-mediated interactions. The GlcNAc in the +2 subsite hydrogen bonds with K505 via the N-acetyl oxygen and makes a water-mediated interaction (via O6). The GlcNAc moiety in the +3 subsite makes only hydrophobic interactions with the non-polar face of R392. Also, the non-polar face of the N-acetyl moiety of this GlcNAc makes van der Waals interactions with Y351 (Figures 4D and S10).

The observation that the GlcNAc moieties closest to E390 make the most interactions is also evident in the refined B-factors (Figure S11B). Trisaccharide 1 in the −2 subsite has an average B factor of 42.2 Ų, which increases to 61.4 Ų for GlcNAc in the −4 subsite being the farthest from E390. Trisaccharide 2 GlcNAc in the +1 subsite has an average B-factor of 41.9 Ų, which increases to 75.4 Ų for GlcNAc in the +3 subsite being the farthest from E390 in this trisaccharide (Figure S11B). This analysis indicates that the saccharide moieties closest to E390 are the most ordered, with adjacent saccharide moieties being progressively less ordered.

The binding of the trisaccharides and disaccharides caused some conformational changes in the active site. Trisaccharide 1 binding causes the active site to clamp down on this trisaccharide, particularly near the N-acetyl moiety of GlcNAc in the −2 subsite. The distance between the interacting backbone oxygen of Y463 and backbone nitrogen of M410 shortens from 8.3–8.4 Å in the apo Cj0843c structures ⁷ to 7.6 Å in the di- and trisaccharide bound structures. This movement is similar in the bulgecin A Cj0843c complex, where this distance is 7.5 Å; a similar movement is also observed in P. aeruginosa Slt binding PG fragments ⁴. F412 swings over and forms a hydrophobic edge-on interaction with the hydrophobic faces of the GlcNAc moieties in the −3 and −2 subsites (Figure 4C); this was also observed with bulgecin A binding (Figure S11C). An additional conformational change in the active site is observed only in the trisaccharide complex. To accommodate trisaccharide 2 GlcNAc in the +3 subsite, residues Y351, R355, and F359 shift away to provide space for this GlcNAc moiety (Figure S11A). The saccharide binding and concomitant conformational changes increase the T_m of Cj0843c (Figure 3).

To complement the static crystallographic snapshots, we carried out MD simulation of the di- and trisaccharide Cj0843c complexes to probe the dynamic nature of the interactions. We carried out the simulations in the presence of explicit water molecules for 10 ns at 310 K (Figure S12). Two simulation runs were done for each saccharide. In one run, the catalytic E390 residue is protonated/neutral as previously stipulated for the corresponding catalytic Glu in the substrate-bound state in the homologous Slt70 ²¹ and MltE ⁸ as well as MD simulations with Cj0843c ⁷ (Figures S12A–S12D). A second run was carried out with a deprotonated (i.e., charged) E390 (Figures S12E–S12H) since it is not known if this glutamate is also protonated without PG substrate in the active site which would position a MurNAc in the −1 subsite near E390. The GlcNAc trisaccharide/disaccharide bound Cj0843c structures likely more closely represent the ground-state of this enzyme. The simulations at 310 K showed that trisaccharide 1 GlcNAc in the −2 subsite maintains its key hydrogen bonds with M410 and Y463, and GlcNAc in the −3 subsite maintains its hydrogen bond with F469 (Figure S12). GlcNAc in subsite +1 of trisaccharide 2 maintains relative stable hydrogen bonds with the backbone oxygen of E390 and the Q408 side chain. The hydrogen bond with Q389 by this GlcNAc is also present during a significant portion of the simulation in both sets of runs. However, the crystallographically observed (2.7 Å) hydrogen bond of the O4 hydroxyl with the E390 side chain (Figure 4) is only present in the MD runs with the deprotonated E390 (Figures S12E–S12H). These MD results suggests that the E390 is not protonated in the crystal thereby being a better hydrogen-bond acceptor for the O4 hydroxyl of the GlcNAc in the +1 subsite. The E390 might thus also be deprotonated in the ground-state of Cj0843c; perhaps the catalytic glutamate becomes protonated, needed for catalysis⁸, upon productive PG substrate binding such that the concomitant change in dielectric constant in the active site affects the pKa of E390.

Comparison with previously determined PG complexes of homologous LTs reveals that trisaccharide 2 GlcNAc in the +1 subsite is in the same position bound as the GlcNAc segment of a 1,6-anhydromuropeptide when bound to Slt70 (PDB ID 1QTE; ²⁷). A similar interaction was observed of muropeptide (GlcNAc-anhMurNAc-pentapeptide) bound to P. aeruginosa Slt (PDB ID 6FBT; ⁴) and in the chitotetraose/GlcNAc complex of N. meningitidis LtgA (PDB ID 5O2N; ²⁵). A structure-based sequence alignment, including the three above-mentioned related LTs shows the sequence conservation of a number of residues interacting with the trisaccharides (Figure S13). Regarding trisaccharide 1, its sugar in the −2 subsite superimposed well onto the GlcNAc moiety of bulgecin A when bound to Cj0843c (Figure S11C). The N-acetyl moiety of this GlcNAc in trisaccharide 1 and bulgecin A are in the same position; the ring of the attached GlcNAc in bulgecin A is, however, slightly shifted (Figure S11C). These sugar positions of trisaccharide 1 are in similar positions as the corresponding moieties in the PG-bound structure of P. aeruginosa Slt (PDB ID 6FCU; ⁴; Figure S14A) and chitotetraose bound structure of N. meningitidis LtgA (PDB ID 5O2N; ²⁵; Figure S14B). These latter superimpositions also show similarities for trisaccharide 2 subsite binding, although the PG strand bound to P. aeruginosa Slt bends in a different direction (Figure S14A). Cj0843c was found to have endolytic activity with substrates VIIIp/VIII (Figure 2), which would likely require the PG in the +2, +3, and +4 subsites to bend to avoid steric clashes with R355 and F359 (Figures 4D and S11A); such PG bending was observed in the PG-bound structure of P. aeruginosa Slt (PDB ID 6FCU; ⁴; Figure S14A). Alternatively, R355 and F359 from the L-domain could shift even further than was needed to accommodate the trisaccharide (Figure S11A).

The observation that Cj0843c can accept PG substrates with or without anhMurNAc end, and with or without peptide stems (Figures 1 and 2) agrees with the crystallographic results with the di- and trisaccharides of the present work (Figures 4 and 5). Our structural results show that the +1, +2, and even +3 subsites of the Cj0843c active site can accommodate GlcNAc moieties. GlcNAc binding in the +1 and +2 positions would mimic the last two saccharide moieties of PG substrate without anhMurNAc, except that GlcNAc in the +2 position is a MurNAc, a GlcNAc analog. We anticipate a conformational change to accommodate the MurNAc lactate group at the GlcNAc O3 position (Figure 4).

Longer synthetic substrates (VIIIp/VIII) are better substrates for Cj0843c compared to shorter ones (IVp/IV; Figure 2). The shorter ones would occupy only the active-site subsites −2 −1 +1 +2, which would leave the −3 subsite empty; this latter subsite can provide stabilizing saccharide interactions, as observed in the di- and trisaccharide Cj0843c complexes (Figures 4, 5, S10, and S11A), and could perhaps aid in productive PG binding for longer PG substrates. Compared with VIII, the VIIIp substrate is converted more readily to the major product IIpa (Figure 2). This indicates that PG with peptide moieties serve as better substrates for Cj0843c. This is in agreement with previous Cj0843c MD simulation results, which suggested a role for PG peptide stems in enzymatic activity ⁷. The PG peptide stem carboxyl moieties of anhMurNAc in the MD simulations interacted with positively charged pocket 2 at various steps of Cj0843c PG recognition. This pocket 2 not only had a role in anchoring of the PG strand for the exolytic activity but also maintains interactions with PG strands before and after cleavage to potentially prevent the PG strand from leaving the active site after cleavage. Overall, this could potentially promote processivity by having the (now shorter) PG strand slide further in the active site groove for the next exolytic reaction.

Furthermore, the presence of an anhMurNAc end would appear to speed up the reaction for the tetrasaccharide PG substrates (IVp*a/IVa vs. IVp/IV; Figure 2). This finding can be explained by the crystallographic observation that trisaccharides can bind to two separate regions of the active site, including to subsites −2, −3, and −4 (Figures 4 and 5). Thus, it is conceivable that the IVp and IV could occupy these same subsites and have the 4^th (MurNac) saccharide occupy the adjacent −1 subsite near E390. This binding mode would be unproductive and would thus compete with the productive binding mode occupying the −2 −1 +1 +2 subsites (thus occupying the trisaccharide 2 subsites +1 +2). In the unproductive binding mode, the terminal MurNAc of IVp/IV situated in the −1 subsite is anticipated to make favorable active-site interactions via its O6 hydroxyl (with the catalytic Glu) as observed in the octasaccharide P. aeruginosa Slt complex ⁴ and Cj0843c PG MD simulations ⁷. This interaction would not be possible for a PG substrate with an anhMurNAc end (i.e., IVpa/IVa) since the O6 atom is no longer a free hydroxyl. Substrates IVpa/IVa could thus favor binding to the −2 −1 +1 +2 subsites over the (non-productive) −4 −3 −2 −1 subsites, thereby possibly explaining their faster reaction compared to that of IVp/IV.

The data in Figures 1 and 2 are suggestive that the peptide stems on PG substrate have an enhancing effect on catalytic activity. Previously, MD simulations showed that the carboxyl moieties of the peptide stem of the anhMurNAc end of the PG could readily interact with the positively charged pocket 2 containing 8 Arg/Lys residues, situated near the active site (Figures 7 and S15). We performed site-directed mutagenesis to probe the importance of three of these Arg/Lys residues on catalytic activity.

Figure 7.

Cj0843c interactions with PG. (A) MD snapshot of Cj0843c in complex with PG decasaccharide PG fragment bound in the active site (from our previous study ⁷). Domain coloring is as in Figure 4A, and PG moiety coloring is as in Figure 6. The peptide moieties of the 5^th and terminal disaccharide moiety attached to AnhMurNAc are labeled Ala51, d-Glu52, DAP53, and Ala54. The catalytic E390 is labeled. The pocket 2 Arg/Lys residues are colored with carbon atoms in green. Key hydrogen bonds between the PG and Cj0843c are indicated by black dashed lines. (B) zoomed-in view of A focusing on regions harboring the PG bound in the −1→ +2 saccharide subsites (labeled in bold). This panel depicts the interactions of the terminal tetrapeptide stem (Ala51, d-Glu52, m-Dap53, and Ala54) attached to the GlcNac-AnhMurNAc at the +1 and +2 subsites, respectively. Key stipulated salt-bridge interactions of the peptide stem carboxyl moieties with R392 and K505 are shown as dashed lines. (C) Superimposition of Slt70 bound to GlcNAc-anhMurNAc- l-Ala-γ-d-Glu ²⁷ onto the Cj0843c MD snapshot. Slt70 protein carbon atoms are colored light orange. The saccharide carbon atoms of the peptidoglycan fragment to Slt70 are colored orange; the attached peptide moieties’ carbon atoms are colored gray. The Cj0843c and bound PG are colored as in Figure 7B. The (backbone) carboxyl of the peptide stem of d-Glu52 (Cj0843c) and d-Glu2 (Slt70) are labeled 1 with colors black and dark grey, respectively. The amide bond of the AnhMurNac (in +2) to which stem peptide d-Ala51 is attached is labeled 2. The following residues were superimpositioned: Cj0843c 381–409 and 497–508 onto Slt70 469–497 and 582–593, respectively (PDB ID 1QTE), yielding an r.m.s.d. of 0.64 Å for 58 Cα atoms.

Residues R388 was selected since it is adjacent to the catalytic E390 and forms guanidinium-group-stacking interactions with R392 and together form a localized strongly positive patch (Figures 4D, 7, and the stereo image in Figure S15). Such Arg-stacking interactions were previously observed in other proteins ²⁸. Among the nine homologous LT sequences, R388 is conserved (exception is a Lys in Nitratiruptor LT, but there too the functional conservation is present); R392 is present in five of the nine LT sequences with one other sequence having a Lys at this position (Figure S13). In previous MD simulations of Cj0843c with PG (Figures 7A–7B and S15) as well as the structure of Slt70 in complex with a peptidoglycan fragment, the arginine in Slt70 equivalent to R388 of Cj0843c was found to interact with the main-chain carboxyl of d-Glu in the tripeptide segment of GlcNAc-anhMurNAc-l-Ala-γ-d-Glu-m-Dap (Figures 7C and S15C; PDB ID 1QTE; ²⁷). These observations suggest a role for this arginine; a similar interaction was also seen in the muropeptide (GlcNAc-anhMurNAc-l-Ala-γ-d-Glu-m-Dap-d-Ala-d-Ala) of P. aeruginosa Slt complex (PDB ID 6FBT; ⁴). This arginine actually has shifted position in Slt70, and its guanidinium group is now structurally more equivalent to R392 of Cj0843c. On the same face of the active site and near R388/R392 is the conserved K505 (present in five out of the nine LTs, including N. meningitidis LtgA ²⁵; Figure S13), which was also targeted for mutagenesis. We selected the third residue on the other side of the pocket 2 active-site region, R326. This residue is part of a 3-residue cluster of positively charged amino acids on the L-domain (K323, R326, and R355); since R326 is in the middle of this cluster, it was chosen as a representative of this cluster. R326 is not conserved, yet six out of nine homologous LTs have a positively charged Lys residue at this position; furthermore, both P. aeruginosa Slt and N. meningitidis LtgA structures have different arginines (R459 and R408, respectively; Figure S13) yet all their side-chain guanidium moieties occupy the same space in their respective active sites. This ‘conservation’ of the guanidium group placement is facilitated by shifts in these structures’ L-domain helices that harbor these arginines; in addition, Cj0843c also contains fewer helices in the L-domain ^{4, 25}(Figure S13). Finally, we also targeted the catalytic E390 for mutagenesis.

The catalytic activity of wild-type Cj0843c was measured by fluorescence using the EnzChek assay. The mutations targeting the catalytic glutamic acid E390A and E390Q mutations resulted in a drastic loss of activity, as expected (Figures 8 and S16).

Figure 8.

Activity assays of the wild-type Cj0843c and active-site variants. Bar graph showing the 60 min time-point activity comparisons of the wild-type Cj0843c and variants, all at 1.88 μM protein concentration. Data show the mean ± SEM of two independent experiments. *P<0.05, **P<0.01, and ***P<0.001 (Student’s t-test).

The variants R388A and K505A showed a decrease in activity, whereas R326A unexpectedly showed enhancement of activity, compared with wild-type Cj0843c (Figures 8 and S16). The larger decrease in activity observed for R388A, compared with K505A, is likely due to this R388/R392 region being key for anchoring of the peptidoglycan peptide-stem carboxylate (Figures 7B–7C, S15B–S15C) such that the saccharides in the +2 and +1 subsites are stabilized and aligned for efficient scission at the seat of reaction (−1/+1 position), initiated by E390. K505A also has a negative effect on activity, suggesting an important role. K505 is in ~6 Å of R392, and this lysine could interact with the PG carboxyl(s) or anhMurNAc (such as in Figures 7B–7C, S15B–S15C).

The increased activity for the R326A mutant was unexpected. A possible explanation is that this residue is on the other side of the positively charged pocket 2 region of the active site (Figures 7A and S15) of which the consequences will be discussed below. Previous MD simulations showed that some of the carboxyl groups of the terminal peptide stem of a PG decasaccharide in the active site could form salt-bridge interactions with R326 (albeit not as frequently as with R388, R392, and K505)⁷. However, the positive electrostatic surface potential in pocket 2 is most prominent near residues R388, R392, and K505 (Figure S17A). On this site, key carboxyl groups of the terminal peptide stem can form salt-bridge interactions (Figures 7B–7C, S15B–S15C). These observations, combined with the strong conservation of these positively charged residues, suggest roles for R388/R392/K505 in productive binding of PG strands. A calculated electrostatic surface potential map of a modeled R326A Cj0843c variant shows that the region around residue 326 is changed from slightly positive to more negatively charged since the R326A change has left the interacting D327 electrostatically unpaired and more solvent-exposed (Figure S17B). Although speculative, it is conceivable that the R326A change not only prevents the peptide-stem carboxyl moieties from interacting with residue 326, but the concomitant change in electrostatic potential near 326 might even electrostatically steer the peptide stem away from 326. Although R326 seems to have a negative effect on overall lytic activity, it could play a positive role in facilitating the binding of larger, crosslinked PG substrates. Such substrates are known to be recognized by Cj0843c as we observed the resulting Tetra2A2, TetraTriA2, and Tetra3A3 products (Figure 1).

In conclusion, our studies show that Cj0843c is indeed a lytic transglycosylase that exhibits both endolytic and exolytic activities, in vitro. The enzyme prefers PG substrates with anhMurNAc ends and with peptide stems. Regarding the peptide-containing PG, we identified active-site residues that might be important for forming salt-bridge interactions with PG carboxyl groups for substrate recognition, as assessed by mutagenesis studies. Mutation of a more distant positively charged residue on the other side of the active site, R326, caused an increase in lytic activity. We speculate that this residue could be more important for interactions with the polymeric PG substrate. We also show by crystallography and by MD simulations that the active site of Cj0843c is capable of binding GlcNAc containing disaccharide and trisaccharide without MurNAc moieties, without peptide stems, and without the anhMurNAc ends.