Structure-Based Modification of a Clostridium difficile-Targeting Endolysin Affects Activity and Host Range

Abstract

Endolysin CD27L causes cell lysis of the pathogen Clostridium difficile, a major cause of nosocomial infection. We report a structural and functional analysis of the catalytic activity of CD27L against C. difficile and other bacterial strains. We show that truncation of the endolysin to the N-terminal domain, CD27L_1–179, gave an increased lytic activity against cells of C. difficile, while the C-terminal region, CD27L_180–270, failed to produce lysis. CD27L_1–179 also has increased activity against other bacterial species that are targeted by the full-length protein and in addition was able to lyse some CD27L-insensitive strains. However, CD27L_1–179 retained a measure of specificity, failing to lyse a wide range of bacteria. The use of green fluorescent protein (GFP)-labeled proteins demonstrated that both CD27L and CD27L_1–179 bound to C. difficile cell walls. The crystal structure of CD27L_1–179 confirms that the enzyme is a zinc-dependent N-acetylmuramoyl-l-alanine amidase. A structure-based sequence analysis allowed us to identify four catalytic residues, a proton relay cascade, and a substrate binding pocket. A BLAST search shows that the closest-related amidases almost exclusively target Clostridia. This implied that the catalytic domain alone contained features that target a specific bacterial species. To test this hypothesis, we modified Leu 98 to a Trp residue which is found in an endolysin from a bacteriophage of Listeria monocytogenes (PlyPSA). This mutation in CD27L resulted in an increased activity against selected serotypes of L. monocytogenes, demonstrating the potential to tune the species specificity of the catalytic domain of an endolysin.

INTRODUCTION

Clostridium difficile is the causative agent of C. difficile infection (CDI), a commonly hospital-acquired infection which is gaining in notoriety and severity (24, 30). Although currently treatable by antibiotic therapy, the mosaic nature of the C. difficile genome highlights its strong ability to acquire antibiotic resistances (38). The organism produces spores which are long lived and resistant to some disinfectants, contributing to its persistence in the environment. Further, C. difficile is present in animals and has been detected in meat (33). Thus, novel approaches are required not only for therapy in the patient but also for elimination and detection of organisms in hospital or community environments and within the food chain.

Bacteriophage endolysins represent potential antimicrobials via their ability to lyse Gram-positive cell walls when applied externally (11, 28). They have shown promising potential in therapy, disinfection, and detection (15, 34). Endolysins exhibit species-specific targeting of the enzyme to its substrate (24); this targeting is mediated via specific cell wall binding domains which recognize certain cell wall components and are thought to be the key to the restricted host range of the enzymes. Thus, endolysins encompass not only a mechanism for killing the cells but also the basis for a specific detection system. In some cases, it has been demonstrated that these cell wall binding moieties reside in the C-terminal part of the endolysin (19, 20, 25, 30). The N-terminal part of the endolysin commonly houses the catalytic domain, which attacks one or more of the various bonds within the peptidoglycan layer, for example, N-acetylmuramoyl-l-alanine amidase, N-acetyl-β-d-glucosaminidase, N-acetyl-β-d-muramidase, or endopeptidases (21).

We have recently isolated the endolysin of ΦCD27, a bacteriophage infecting Clostridium difficile, and demonstrated that it can cause cell lysis when applied externally (26). CD27L is a 270-amino-acid protein and shows homology to N-acetylmuramoyl-l-alanine amidases in the N-terminal part of the protein (residues 2 to 143), while the C-terminal portion (residues 180 to 270) has no identifiable domains. CD27L was not active against a range of environmental and commensal bacteria, an important factor in the treatment of CDI, as the condition is closely related to the disturbance of the native gut microbiota during antibiotic treatment. However, although CD27L is active under in vitro conditions, a higher rate of activity would be required to successfully lower C. difficile numbers in a growing population, especially in the challenging environment of the gastrointestinal tract. Further optimization of these enzymes could help in the treatment of acute CDI. The production of truncated peptidoglycan hydrolases has been demonstrated to produce active lysins (10). However, removal of the putative cell wall binding region raises questions about the maintenance of specificity.

Although the specificities of endolysins are believed to reside in the C-terminal domain, the catalytic N-terminal domain does contain some specificity based on its enzymatic substrate. The enzyme also has to extract its substrate from a complex peptidoglycan layer. The specificity of a truncated peptidoglycan hydrolase is therefore rather complex. In this paper, we explored the structural and functional characteristics of truncated CD27L_1–179, and we show that the substrate specificity of the catalytic portion of the endolysin alone can be used to target a subpopulation of bacteria.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

C. difficile strains were obtained from culture collections (NCTC, Central Public Health Laboratory, London, United Kingdom, strains NCTC 11204, 11205, 11206, 11207, 11208, 11209, 11223, 12726, 12727, 12728, 12731, 11382, 13287, 13307, and 13366; DSMZ, Braunschweig, Germany, strains DSMZ 12056 and 12057) or were kindly donated by Jonathan Brazier, Anaerobe Reference Unit, University Hospital of Wales, Cardiff, United Kingdom (strains R23 521, R23 524, R23 613, R23 614, R23 621, R23 635, R23 639, R23 642, R23 720, R23 727, R23 732, R23 737, G83/03). Strains were grown anaerobically at 37°C in brain heart infusion (BHI) broth with complements as described previously (26). Escherichia coli strains were grown in L broth shaken at 37°C. Commensal and clostridial strains were obtained from the in-house IFR culture collection, NCIMB (Aberdeen), or DSMZ and were grown as recommended by DSMZ or in BHI broth with complements. Strains used in addition to those mentioned previously (26) were Bacillus amyloliquefaciens 0880, Bacillus cereus NCIMB 11796, Bacillus subtilis ATCC 6633, Clostridium acetobutylicum BL75141, Clostridium bifermentans NCTC 13019, Clostridium sordellii NCTC 13356, Clostridium sporogenes ATCC 17886, Clostridium tyrobutyricum NCIMB 9582, Leuconostoc mesenteroides subsp. mesenteroides ATCC 8293, Listeria ivanovii NCTC 11007, and Listeria monocytogenes NCTC 5214 (serovar 4a), NCTC 5348 (serovar 1/2c), SLCC 2540 (serovar 3b), SLCC 2479 (serovar 3c), SLCC 2374 (serovar 4a), SLCC 2376 (serovar 4c), SLCC 2377 (serovar 4d), SLCC 2378 (serovar 4e), SLCC 2482 (serovar 7), and F6861 (serovar 4b).

Truncation and site-directed mutation of CD27L.

The N-terminal portion (coding for CD27L_1–179) and the C-terminal portion (coding for CD27L_180–270) of CD27L were produced by PCR from plasmid CD27L-pET15b (26) using GoTaq polymerase (Promega). CD27L_1–179 was amplified using primer T7P (5′-TAA TAC GAC TCA CTA TAG GG) from the pET15b vector (Novagen) and a primer to create a stop codon after Asn 179, CD27L EAD (5′-TTA ACT CCC TCC TAA TTT ATA TT [bolding represents altered nucleotides throughout unless otherwise noted]), giving a 698-bp product. CD27L_180–270 was amplified to start from Asn 180, using a primer to create an NdeI site encompassing an initiation Met immediately upstream of Asn 180, CD27L CBD (5′-CAT ATG AAT GAG GGA GTT AAA CAG ATG) and T7T (GCT AGT TAT TGC TCA GCG G) from the pET15b vector, giving 370 bp. Both products were subcloned into pCR2.1 by using the TA Cloning kit (Invitrogen) and FastLink DNA ligase (Epicentre). The constructs were transformed into E. coli TOP10 cells (Invitrogen) for sequence and orientation confirmation. Products were excised with NdeI and XhoI and subcloned into pET15b.

The L98WCD27L and L98WCD27L_1–179 mutants were produced by splice overlap PCR from CD27L-pET15b and CD27L_1–179-pET15b, respectively, using Phusion polymerase (Finnzymes). To create the L98WCD27L mutant, two PCR products were produced using primers T7P2 (5′-TGA GCG GAT AAC AAT TCC C) and CD27L_L98WR (5′-ATA ATA CCA GAC TTC TGA ACC TTT ACC) to give 434 bp and primers T7T and CD27L_L98WF (5′-GAA GTC TGG TAT TAT AGT AAT AAA GGC T) to give 616 bp from CD27L-pET15b and 378 bp from CD27L_1–179-pET15b. Products were spliced and amplified with T7P2 and T7T to give 1,022 bp and 784 bp, respectively. Products were restricted with NdeI and XhoI and subcloned into pET15b. All pET15b constructs were transformed into E. coli TOP10 cells for sequence confirmation and then into E. coli BL21(DE3) cells (Invitrogen) for protein expression, as described previously (26).

Production of GFP-labeled endolysins.

Green fluorescent protein (GFP)-labeled endolysins were produced as translational fusions. The sequence for a red-shifted variant GFP mutant 3 (7) was subcloned with CD27L and CD27L_1–179 in pET15b to produce His-tagged N-terminally labeled proteins. To facilitate subcloning, an internal NdeI site in gfp3 was first removed by splice overlap PCR to alter T231 to C231. For this, gfp3 was amplified by PCR from pSB2030 (32) in two parts using primers GFP_NDE (5′-GGA ATA ACA TAT GAG TAA AGG CGA AG) to create an NdeI site around the start Met codon and GFPSPLICEGTG (5′-TTT CAT GTG ATC TGG GTA TCT CGC) to produce a 247-bp product and primers GFPSPLICECAC (5′-CCA GAT CAC ATG AAA CAG CAT GAC) and GFP_TAC (5′-GTA TTT GTA TAG TTC ATC CAT GGC) to change the TAA stop site to TAC, giving a 495-bp product. PCR was performed with Phusion polymerase; products were purified and spliced together by amplification with primers GFP_NDE and NGFP_LINK, which overlaps the end of the coding sequence and adds a linker to code for seven Gly and Ser residues to allow efficient protein folding, followed by sites for NdeI and EcoRI to facilitate future subcloning (5′-GGA TGA ACT ATA CAA ATA CGG TAG TGG ATC AGG TAG TGG ACA TAT GAA TTC T [linker given in bold]) to give the expected 760-bp product (gfp-linker). The PCR product was given 3′ A overhangs using GoTaq polymerase and subcloned into pCR2.1 for sequence confirmation. The modified gfp-linker was restricted from this construct using NdeI and subcloned into vector pET15b and constructs CD27L-pET15b and CD27L_1–179-pET15b, all of which had been restricted with NdeI and dephosphorylated with Antarctic phosphatase (New England BioLabs).

Protein expression and analysis.

His-tagged endolysins CD27L, CD27L_1–179, and CD27L_180–270, His-tagged GFP-endolysin fusions, and the negative His-tagged GFP-linker control expressed from pET15b were expressed from IPTG (isopropyl-β-d-thiogalactopyranoside)-induced E. coli BL21(DE3), partially purified using Qiagen Fast-Start Ni-nitrilotriacetic acid (NTA) columns, analyzed by SDS-PAGE and Western hybridization, and assayed for lytic activity as described previously (26). Lysis assays were commonly performed with equimolar concentrations of approximately 1 μM (10 μg Ni-NTA-purified full-length endolysin per 300 μl assay and 6.8 μg truncated endolysin to take account of the difference in protein molecular weight). Species specificity was confirmed with turbidity reduction assays on a variety of species and C. difficile strains obtained from culture collections as listed previously (26). E. coli K-12 cells were treated with choloroform to remove the outer membrane by the method of Nakimbugwe et al. (27), and their sensitivity was tested in lysis assays with 30 μg crude protein extracts prepared in 20 mM sodium phosphate buffer, pH 6 (26), using chicken egg white lysozyme (Sigma) as a positive control. Lysis assays were performed in duplicate in 300-μl volumes. Lytic activity was measured as the percentage drop in optical density (OD)/minute during linear lysis over a 10.5-min window, corrected for any loss in OD in buffer controls. A comparison of the effects of CD27L and CD27L_1–179 on C. difficile viability was conducted as described previously (26) under anaerobic conditions using 10 μM Ni-NTA-purified protein in elution buffer in a volume of 300 μl with cell enumeration on BHI plates.

Preparation and structural determination of the catalytic domain of CD27L.

The CD27L_1–179-expressed protein was affinity purified using Ni-NTA beads as described above. The His tag was cleaved using 100 U thrombin, and the cleaved protein was purified by running the sample again over an Ni-NTA column. The eluted protein was concentrated to 1 mg/ml and further purified by size exclusion chromatography using a Sephadex 200 10/30 column (GE Healthcare). During the size exclusion step, the running buffer used was 0.1 M HEPES, pH 7.4, and 100 mM NaCl. The protein eluted under the peak was concentrated to 7 mg/ml and crystallized in a hanging drop configuration. Crystals were obtained with a mother liquor of 0.1 M Tris (pH 8.2), 0.2 M lithium sulfate, and 20% polyethylene glycol (PEG) 4000.

A single crystal was soaked in a buffer containing 5% glycerol, 25% PEG 4000, 0.2 M lithium sulfate, and 0.1 M Tris (pH 8.2) and then flash frozen to 100 K. The crystal diffracted X rays to a resolution of 2.0 Å at the EMBL Hamburg X12 beamline (Table 1). The X-ray data were collected with an Mar225 charge-coupled device (CCD) camera and processed using Denzo and Scalepack (29). The structure was solved by molecular replacement with MolRep (39) using the catalytic domain of the PlyPSA endolysin (PDB code 1XOV) as a search model. Four molecules were identified, and the structure was rebuilt by Buccaneer (8) and further refined with Refmac5 (41) to an R factor of 18.5% (R_free, 24.5%). The structure was inspected with Coot (12), and the final structure was validated with MolProbity (5). All structural figures were produced with PyMOL (http://www.pymol.org/).

Table 1.

Crystallographic statistics

Parametera	Value(s)b
Data collection statistics
Space group	P1
Unit cell dimensions:
a, b, c (Å)	36.5, 69.7, 78.8
α, β, γ (degrees)	101.7, 102.0, 105.2
Resolution range (Å)	20–1.87 (1.90–1.87)
No. of unique reflections	50,589 (3,987)
Completeness (%)	88.3 (69.8)
Avg intensity [I/σ(I)]	8.9 (2.3)
Redundancy	1.8 (1.5)
Rmerge	13.7 (31.0)
Refinement statistics
Resolution range (Å)	20–2.0 (2.05–2.00)
No. of unique reflections	41,298 (2,875)
Completeness (%)	91.2 (85.6)
R/Rfree (%)	18.5/24.5 (17.8/28.9)
No. of protein residues	716
No. of sulfate ions	14
No. of zinc ions	4
No. of solvent molecules	1,468
MolProbity Ramachandran plot (%)
Favored area	98.3
Outliers	0

^aRedundancy is the average number of equivalent reflections measured for each unique reflection. R_merge = ΣΣ|I_i − <I>|/ΣΣ(I), where I_i is the intensity for the jth measurement of a reflection with indices hkl and <I> is the weighted mean of the reflection intensity. R_factor = Σ F_o(hkl) − F_c(hkl)/Σ F_o(hkl), where F_o and F_c are the observed and calculated structure factors, respectively. R_free is the crystallographic R factor calculated with 5.0% of the data that were excluded from the structure refinement.

^bNumbers in parentheses are for the outermost resolution shell.

Analysis of cell wall binding.

Binding of GFP-labeled endolysin to cells of C. difficile was assessed using the method of Loessner et al. (24) using Ni-NTA-purified protein at a final concentration of 7.5 μM and incubating for 20 min at 37°C. After being washed, cells were viewed by fluorescence microscopy using an Olympus BX60 microscope with ProgRes Capture Pro 2.1 software (Jenoptik, Germany) and an NB filter cube (U-MNB, exciter filter BP470-490, barrier filter BA515) with the ×100 magnification oil immersion lens.

Peptidoglycan was extracted from C. difficile cells grown to an optical density at 600 nm (OD₆₀₀) of 0.5 by the method of Atrih et al. (2) with an added step after boiling in which the cells were disrupted with 0.1-mm acid-washed glass beads (Sigma) in 50 mM Tris-HCl (pH 7) by using a FastPrep FP120 cell disrupter (Savant) with four 30-s bursts (speed 10), with incubation on ice for 5 to 10 min between bursts. Peptidoglycan from Bacillus subtilis and Staphylococcus aureus was generously provided by Emma Hayhurst and Simon Foster (University of Sheffield).

Atomic coordinates and structure factors have been deposited in the Protein Data Bank (PDB) (http://www.pdb.org) under the accession code 3QAY.

RESULTS

Lytic activity of CD27L domains.

The truncated endolysin N-terminal domain CD27L_1–179 was able to lyse all 32 C. difficile strains tested, and in all cases CD27L_1–179 caused more rapid lysis than the full-length CD27L enzyme when the proteins were added to fresh or frozen cells (Fig. 1). The C-terminal half, CD27L_180–270, failed to cause specific lysis (data not shown). As with CD27L, there was a short lag before the onset of linear lysis from CD27L_1–179 (Fig. 1).

Fig. 1.

Effect of truncation on lytic activity and specificity. (a) N- and C-terminal truncations of endolysin CD27L were produced as His-tagged proteins. CD27L_1–179 gave more rapid lysis of C. difficile than the full-length CD27L (b), but both proteins failed to lyse C. tyrobutyricum (c). Lysis assays comprised fresh cells incubated with 10 μg Ni-NTA-purified protein or buffer. Results are the means ± standard deviations (SD) from duplicate assays. ▴, CD27L; •, CD27L_1–179; □, buffer control; N-AAA, N-acetylmuramoyl-l-alanine amidase.

The effect of the full-length and truncated endolysins on cell viability was tested at a 10 μM concentration. In assays containing ca. 5 × 10⁴ CFU, 2 h of incubation with CD27L reduced surviving cell numbers to ca. 1 × 10² and 6 × 10¹ CFU in each of the duplicate assays, almost a 3-log reduction in cell numbers. As noted previously, in assays containing ca. 5 × 10⁵ CFU, CD27L reduced cell numbers by 2 log. In the same assays, truncated endolysin CD27L_1–179 was much more effective, with no live cells being recovered. At higher cell densities of ca. 1 × 10⁸ or 1 × 10⁷ CFU, CD27L gave reductions of ca. 0.5 log or ca. 1 log, respectively, while CD27L_1–179 gave a reduction of more than 4 log from 1 × 10⁸ CFU and an almost 6-log drop from 1 × 10⁷ CFU to ca. 2 × 10¹ CFU; those colonies that survived the endolysin treatment showed a slow growth phenotype, requiring 2 days of incubation, suggesting cell damage.

Binding of GFP-labeled endolysin.

GFP-labeled CD27L and CD27L_1–179 produced as translational fusions both showed clear, strong binding to cells of C. difficile 11204 (Fig. 2a and b). No labeling was seen with the GFP-linker protein produced from pET15b, although the cells were visible due to native autofluorescence. Binding was also demonstrated using lower concentrations of labeled proteins (0.2 μM, 2 μM) and shorter incubation times (5, 10, 15 min at 37°C), but labeling increased with both protein concentration and time.

Fig. 2.

Binding and lytic activity of GFP translational fusions with C. difficile. Binding assays contained 7.5 μM Ni-NTA-purified GFP-CD27L (a) or GFP-CD27L_1–179 (b); lysis assays (c) contained 10 μg protein incubated with fresh cells. Values are the means ± SD from duplicate assays. ▴, GFP-CD27L; ○, GFP-CD27L_1–179; □, GFP-linker; ×, buffer control.

Lysis of cells was visible in samples containing GFP-CD27L and GFP-CD27L_1–179, and the activities of these protein samples were confirmed by turbidity reduction assays (Fig. 2c). The lytic activity was considerably lower than that seen with the unlabeled endolysins. It was notable that GFP-CD27L_1–179 did not give consistently faster lysis than GFP-CD27L, indicating that size and structure are important in the increased lytic activity of the truncation. No lytic activity was observed from the GFP-linker alone, although this sample and the buffer control did exhibit the gradual autolysis that is commonly seen upon extended incubation of C. difficile in these oxygenated conditions.

Specificity of the truncated endolysin.

CD27L_1–179 retained the same specificity as that of the full-length endolysin when tested against a range of other clostridial species (Fig. 1c, Table 2). Two species, C. bifermentans and C. sordellii, were identified which were sensitive to lysis by both CD27L and CD27L_1–179, and both are closely related to C. difficile. These species lysed rapidly in buffer after a short lag, but the onset of lysis was faster with CD27L, and the length of the lag was reduced further with CD27L_1–179. Despite the specificity of lysis observed against Clostridiales, CD27L and CD27L_1–179 were surprisingly able to lyse a number of less closely related species (Table 2). B. amyloliquefaciens, B. cereus, and B. subtilis were sensitive to lysis with CD27L, and in all cases CD27L_1–179 was more effective. This difference was particularly evident with B. subtilis: with this species, lysis of fresh cells with CD27L commonly occurred only after a lengthy lag of ca. 35 min, while lysis from truncated CD27L_1–179 began immediately. Lysis of Listeria ivanovii NCTC 11007 with CD27L was only slightly faster than autolysis in buffer but was more noticeable with CD27L_1–179. In addition, the truncation also caused a slight increase in the host range: CD27L was previously shown to be ineffective against both Listeria innocua NCTC 11288 and L. monocytogenes NCTC 5214 (serovar 4a) (26), but turbidity reduction assays showed that CD27L_1–179 was able to cause lysis of both strains, albeit more slowly than that with C. difficile.

Table 2.

Lytic activity of the full-length and truncated endolysins CD27L and CD27L_1–179 on Gram-positive species

Bacterium	Peptidoglycan typea	Lysin activityb
		CD27L	CD27L1–179
Bacillus amyloliquefaciens	A1γ	+	++
Bacillus cereus	A1γ	+	++
Bacillus subtilis	A1γ	++	+++
Clostridium bifermentans	A1γ	++	+++
Clostridium sordellii	A1γ	++	+++
Clostridium acetobutylicum	A1γ	−	−
Clostridium cellobioparum		−	−
Clostridium coccoides		−	−
Clostridium innocuum	A1γ	−	−
Clostridium perfringens	A1γ	−	−
Clostridium ramosum	A1γ	−	−
Clostridium sporogenes	A1γ	−	−
Clostridium tyrobutyricum	A1γ	−	−
Lactobacillus plantarum	A1γ	−	−
Listeria innocua	A1γ	−	+
Listeria ivanovii	A1γ	+	++
Listeria monocytogenes serovar 4a	A1γ	−	++
Micrococcus luteus	A2	−	−
Enterococcus faecalis	A3α	−	−
Leuconostoc mesenteroides	A3α	−	−
Staphylococcus aureus	A3α	−	−
Bifidobacterium longum	A3β	−	−
Anaerococcus hydrogenalis	A4α	−	−
Enterococcus hirae	A4α	−	−
Lactobacillus casei	A4α	−	−
Lactobacillus johnsonii	A4α	−	−
Lactococcus lactis	A4α	−	−
Pediococcus pentosaceus	A4α	−	−
Bifidobacterium bifidum	A4β	−	−
Eubacterium barkeri	B	−	−

^aFrom references 13, 14, 37, and 40.

^b−, no lysis; +, limited lysis; ++, medium lysis; +++, rapid lysis.

All of these sensitive species have been characterized as having peptidoglycan type A1γ (37), and no species with a range of other peptidoglycan types showed any drop in optical density upon incubation with either lysin. Peptidoglycan type A1γ is also found in Gram-negative cells (36), and although untreated E. coli cells were insensitive to both lysins and lysozyme, when E. coli cells that had been treated with chloroform to remove the outer membrane were tested with crude protein extracts of CD27L and CD27L_1–179, rapid lysis was observed, while extracts of an empty vector control had no effect. The percent drop in OD₆₀₀/minute from the CD27L_1–179 extract on E. coli (3.06 ± 0.17) was similar to that on fresh cells of C. difficile (3.00 ± 0.02), although the lag period commonly seen with C. difficile cells was absent. The full-length CD27L extract was not only less effective on both species but also considerably slower on C. difficile (0.31 ± 0) than on treated E. coli (2.50 ± 0.08). Similarly, the action of 10 U lysozyme was also slower in C. difficile (0.58 ± 0.01) than on treated E. coli (2.69 ± 0.05). Nevertheless, like some Clostridiales, Lactobacillus plantarum, which also has peptidoglycan type A1γ, was resistant to lysis by both the full-length and the truncated lysins.

In addition to lysis of fresh cells, turbidity reduction assays were also performed on peptidoglycan extracts. Peptidoglycan from C. difficile degraded very rapidly in buffer, and a difference between CD27L or CD27L_1–179 and buffer controls was not discernible. Peptidoglycan from S. aureus showed no reduction in optical density upon incubation with CD27L or CD27L_1–179, but peptidoglycan from B. subtilis showed a drop in turbidity upon incubation with both CD27L (percent drop in OD₆₀₀/minute of 1.21 ± 0.04) and, more rapidly, CD27L_1–179 (2.95 ± 0.12).

The structure of the catalytic domain of the endolysin from CD27L.

The crystal structure of the catalytic domain of endolysin CD27L was determined to a resolution of 2.0 Å. The structure comprises residues 1 to 179 and one residue from the N-terminal His tag, a metal ion, and three or four sulfate ions per monomer. There are four protein molecules in the asymmetric unit, and the electron densities of all four monomers are well defined. The average root mean square deviation (RMSD) for the main chain Cα atoms is 0.1 Ų for all four monomers, and the structure is described using monomer A.

The catalytic domain of CD27L consists of an α/β fold of globular shape, with a core of six beta strands surrounded by five alpha helices (Fig. 3). The active site of CD27L is situated around a metal ion close to the surface and is partly exposed to the solvent. Since the X-ray diffraction data were collected at an energy (9,700 eV) close to the K edge of zinc (9,670 eV), it was possible to confirm the presence of zinc from a phased anomalous difference map. The fluorescence spectrum also shows an excitation peak at the zinc edge. The zinc ion is coordinated by two histidines (His 9 and His 84) and a glutamic acid (Glu 26). There is a sulfate ion that coordinates to the zinc ion, and there are two water molecules in the vicinity of the zinc ion. One water molecule coordinates to the metal, and the other is situated at about 2.8 Å. All four monomers contain a metal-bound sulfate, but in monomer C, the sulfate is oriented such that the water molecule that is more distant from the zinc ion is displaced. There are 15 sulfate ions in total, which is probably the result of the presence of lithium sulfate in the crystallization solution. The other sulfates bind on the surface of the protein in different clefts without a consistent secondary binding site between the four monomers.

Fig. 3.

Structural overview of the catalytic domain of CD27L, the position of the active site, and the residues conserved in the substrate binding site. The catalytic domain of CD27L is shown in stereo view. The backbone of the catalytic domain of CD27L is colored yellow. The active site is shown with the zinc ion as a gray sphere, and the protein residues (H9, E26, and H84) that coordinate the zinc ion are shown as sticks. The residues E96, R122, and E144 that are part of the conserved hydrogen bonding network are shown as sticks, as well as the residues L130 and Y131, which are conserved within all Clostridia-targeting amidases. The position of the L98-W98 mutant is shown, with the side chain of tryptophan represented as sticks superimposed, as observed in the PlyPSA structure.

Structural relationship with other amidases.

The backbone structure of CD27L_1–179 is similar to the PlyPSA amidase of the endolysin against L. monocytogenes (Fig. 4a). A DALI analysis searching for structural homologues (18a) shows that the two structures can be superimposed with an RMSD of 1.3 Ų for 171 residues (20) and a Z score of 24.2. This places the structure clearly within the family of N-acetylmuramoyl-l-alanine amidases. The sequence identity between PlyPSA and CD27L_1–179 is 30% for the entire catalytic domain, and the topologies are identical except for the C-terminal helix. CD27L_1–179 has one extended and curved alpha helix at the C terminus, whereas PlyPSA has two helices with a small loop inserted in the middle.

Fig. 4.

Structural comparison and surface representation of the active site. (a) The ribbon diagrams of the catalytic domains of CD27L (green) and PlyPSA (yellow) are overlaid to display the similarities in secondary structures. The loop extensions that contribute to an increase in substrate binding area are indicated. (b) Molecular surface representation of CD27L_1–179, in the same orientation as that shown in panel a, showing the electrostatic potential displayed on the surface-accessible area, as calculated with APBS (3). (c) Molecular surface representation of PlyPSA in the same orientation as that of CD27L_1–179.

The DALI analysis further revealed that the CD27L_1–179 structure is homologous to several other members of the amidase-3 family. Other significant structural homologues are the autolysin amidase CwlV (PDB code 1JWQ), with a Z score of 21.3 and an RMSD of 2.2 Ų, a putative amidase (PDB code 3CZX) with a Z score of 19.0 and an RMSD of 2.1 Ų, and an amidase from Bartonella henselae strain Houston-1 (PDB code 3NE8) with a Z score of 18.5 and an RMSD of 2.2 Ų. All these structural homologues have a sequence identity below 20% for the entire amidase domain.

The substrate binding area of CD27L_1–179 is larger than that in PlyPSA due to loop extensions between secondary structure elements (Fig. 4b and c). There is a loop extension of six residues in CD27L_1–179 just after His 9, which coordinates to the zinc ion. In the same area, there is a loop extension between residues 54 and 61 that stacks behind the loop, spanning residues 11 to 16. Another loop extension occurs on the other side of the active site, covering residues 88 to 92. A surface representation showing the electrostatic potential around the active site reveals that the active site itself is negatively charged, whereas the rim of the active site is positively charged. Compared to the electrostatic isosurface of PlyPSA, the rim is more positively charged. The accumulated positive charge will have a repulsive effect on the peptidoglycan layer, which may be needed to get access to the buried substrate of CD27L.

Structure based sequence analysis of related phage endolysins with amidase activity that target Clostridia.

A BLAST search (1) for the sequence of the catalytic domain of CD27L against the nonredundant database revealed a cluster of sequences of high homology that are all representing either endolysins of phages that target C. difficile or cell wall hydrolases and putative phage endolysins from bacterial genome sequences of principally C. difficile, Clostridium botulinum, and C. sporogenes, potentially representing either prophage genes or autolysins. We took all the amidase sequences that had a BLAST score that was higher than the score for the PlyPSA endolysin, an amidase that targets L. monocytogenes and whose structure is known (20). From a total of 40 sequences, 38 were annotated as endolysins targeting Clostridia. One sequence was attributed to an amidase from the Gram-negative Leptotrichia goodfellowii F0264. The other sequence was annotated as a hypothetical protein from the bacterium Anaerostipes caccae of the Clostridium coccoides group. After curation of the sequence cluster to remove duplicates, the sequences of 14 unique Clostridia-targeting amidases were aligned and analyzed for sequence conservation (Fig. 5).

Fig. 5.

Sequence of the catalytic domain of CD27L aligned with Clostridia-targeting endolysins identified by BLAST analysis. Alignment was produced with ESPRIPT (17), and the secondary structure of CD27L_1–179 is shown, with arrows for beta strands and ribbons for alpha helices. Conserved residues are white on a red background, and residues that are filled by amino acids with similar properties throughout are colored red.

The conserved residues contributing to the catalytic site are the three amino acids that coordinate to the zinc ion (His 9, Glu 26, and His 84) as well as Glu 144. The latter residue is found in most zinc-dependent peptidases and is thought to serve as a proton acceptor during the nucleophilic attack of a water molecule on the carbonyl group of the substrate (20).

A number of residues that are conserved among the endolysins targeting Clostridia are found in the area that has been associated with the substrate binding site (Fig. 4). There are two adjacent residues (Leu 130 and Tyr 131) at about 10 Å from the catalytic zinc ion that are conserved among the endolysins that target Clostridia. The residue pointing directly at the active site is Leu 130, and there are a number of water molecules that fill the cavity between the leucine side chain and the zinc ion. This leucine is also found in the CwlV endolysin, but in the PlyPSA endolysin, it is replaced by a phenylalanine.

Three residues that cluster together on the other side of the active site are Asn 86, Glu 96, and Arg 122. The three residues form a hydrogen bonding network. The Asn 86 residue forms a hydrogen bond through atom OD1 with the OE1 atom of catalytic residue Glu 144. The same Asn 86 forms a hydrogen bond through atom ND2 with the OE2 atom of Glu 96. This residue in turn also makes a hydrogen bond through atom OE1 with the NE atom of residue Arg 122. These three residues are conserved not only among the Clostridia-targeting endolysins but also in PlyPSA and CwlV. The connection between the proton donor Glu 144 and this cluster indicates that it may play a role in proton relay (Fig. 3).

Between these conserved clusters of residues lies one residue (Leu 98 in CD27L) that differs in sequence but is in general of hydrophobic nature. Among the Clostridia-targeting endolysins, this position is occupied by a leucine, valine, isoleucine, phenylalanine, tyrosine, and in one case a glutamic acid. In the PlyPSA structure, the position is occupied by a tryptophan that is kept in place by a hydrogen bond to the conserved Glu 96 residue (20). In the CwlV autolysin, this position is occupied by a tyrosine, and it again makes a hydrogen bond to Glu 96. The residue Glu 96 is part of the conserved hydrogen bonding network that is connected to the catalytic residue Glu 144. Therefore, it seems that the tryptophan and the tyrosine are affecting the hydrogen bond network, and they are kept in a specific orientation to interact with the substrate.

The residue 98 lies directly above the active site, pointing to the outside solvent, and is likely to interact with the substrate from the bacterial cell wall (Fig. 3). The residue L98 in the CD27L endolysin does not affect the conserved hydrogen bond network, and it lacks the bulk of an aromatic ring. The L98W mutant could affect both the proton relay and the substrate binding. We decided to target this residue to see if we could broaden the substrate specificity of the CD27L endolysin so that it would be optimized for digestion of Listeria cell walls.

Modification of the proposed substrate binding site.

Leucine 98 was mutated into tryptophan in both CD27L and CD27L_1–179 to produce the L98WCD27L and L98WCD27L_1–179 mutants, respectively. The lytic activity of these mutants was compared to that of the unmodified full-length and truncated endolysins using partially purified protein extracted contemporaneously. Both mutants retained activity against C. difficile 11204 (Fig. 6). The biological activity of the L98WCD27L mutant was similar to CD27L, while the L98WCD27L_1–179 mutant was almost as effective as CD27L_1–179. This pattern was also seen with the other sensitive species, B. amyloliquefaciens, B. cereus, B. subtilis, C. bifermentans, and C. sordellii. With B. subtilis, the usual lag period seen with CD27L was increased to ca. 50 min with the L98WCD27L mutant, while no lag was present with the L98WCD27L_1–179 mutant.

Fig. 6.

Lysis of Gram-positive species (a, b) and L. monocytogenes serovars (c, d) by unmodified and L98W mutant versions of CD27L (a, c) and CD27L_1–179 (b, d). Turbidity reduction assays contained 10 μg Ni-NTA-purified full-length endolysins or 6.8 μg truncated endolysins. Figures represent percentage drops in OD₆₀₀ per minute, measured over 10.5 min of linear lysis, and are the means ± SD from duplicate assays.

A cumulative rise in lysis efficiency is observed when truncation is coupled with the L98W mutation in L. monocytogenes NCTC 5348 (serovar 1/2c), SLCC 2540 (serovar 3b), and SLCC 2479 (serovar 3c) (Fig. 6). These strains show an increase in lysis efficiency for truncated proteins, and the mutation has an additive effect both in the full-length and the truncated endolysins. In L. ivanovii, there is a mixed response. The truncated wild type and L98W mutant do lyse these bacteria, but the full-length mutant was barely active.

L. monocytogenes 5214 (serovar 4a), SLCC 2374 (serovar 4a), F6861 (serovar 4b), SLCC 2376 (serovar 4c), SLCC 2377 (serovar 4d), SLCC 2378 (serovar 4e), and SLCC 2482 (serovar 7) and L. innocua were not specifically lysed by the full CD27L endolysin or the full-length L98WCD27L mutant, although limited autolysis was observed in these samples and in buffer controls. All were, however, sensitive to the two truncated versions, to different extents depending on the strain. The L98W mutation did not give a large increase in activity in the truncated version in these strains but was often slightly more effective than CD27L_1–179. Serovar 4b was the most resistant, with assays taking over 100 min of incubation before a difference in optical density was seen compared to that of buffer controls.

DISCUSSION

We have shown that the catalytic domain of CD27L is an N-acetylmuramoyl-l-alanine amidase and that the catalytic domain alone causes faster lysis of C. difficile than the full-length endolysin but does not lyse the majority of other Clostridiales tested, suggesting that a measure of substrate specificity remains in the N-terminal domain. CD27L and CD27L_1–179 were also able to cause lysis in several other species, all of which have been described as peptidoglycan type A1γ (37), but other species of the same peptidoglycan type were resistant to lysis. There are several variations in amino acid composition within the type A1γ Clostridiales; however, lysin-sensitive strains C. bifermentans and C. sordellii have the same composition as insensitive C. acetobutylicum, Clostridium ramosum, C. sporogenes, and C. tyrobutyricum (37). The difference in sensitivities may be due to other mediators of specificity; for example, wall sugar patterns have been shown to differ between species (9), and the presence of particular structures may be a requirement for susceptibility. Alternatively, the lack of lysis may be due to a problem in accessing the peptidoglycan from the outside of the cell due to the presence of other cell wall structures. In all cases, the truncated CD27L_1–179 produced faster lysis, but in the majority of cases, removal of the C-terminal part of the protein, classically assumed to represent the cell wall binding domain and the mediator of specificity, did not greatly widen the host range. The exceptions were L. innocua and L. monocytogenes, and the production of mutants demonstrated that subtle changes in species specificity could be produced by modifying part of the enzymatic domain that was separate from the catalytic site. Becker et al. (4) found that chimeric fusions of a streptococcal endopeptidase domain with staphylococcal SH3b domains not only produced increased activity against staphylococci but also maintained significant streptolytic activity, arguing either for a broader host recognition of the SH3b domain or for some native ability of the catalytic domain to recognize its natural host. Truncation of PlyGBS, which acts on group B streptococci, also produced a small increase in its activity spectra to include B. cereus (6), while truncation of endolysin P16, which acted only on B. cereus, extended its host range to include B. subtilis (25). All of these studies argue for further mediators of substrate specificity existing in the catalytic domain, as has previously been suggested (31).

The structure of the N-terminal domain CD27L_1–179 reveals a cascade of conserved residues surrounding the active site that are aligned along the substrate channel that lies on the surface of the amidase. The precise extent of the substrate binding pocket is not clear, because there is no structure of a complex to date between an amidase and its complete peptidoglycan substrate. It is assumed that the peptidoglycan is likely to stretch along the surface next to the active site (18). A BLAST search against the sequence of the catalytic domain produced a cluster of endolysins that all target Clostridia and have a high sequence similarity. Based on a structural-based sequence alignment, several conserved features were determined. The substrate binding area can be divided into three compartments. The most conserved compartment consists of a hydrogen bonding network between residues Asn 86, Glu 96, and Arg 122. This hydrogen bonding network is connected to the active site residue Glu 144, which has been designated a proton donor in the zinc peptidase reaction mechanism. Since Glu 144 is probably buried by the substrate and therefore not directly solvent accessible, the extended hydrogen bonding network probably acts as a proton relay cascade to channel protons from the protein surface to the active site. Since this compartment is conserved in many amidases, it is not expected to contribute much to substrate specificity.

On the other side of the active site, there are two residues that are conserved among the Clostridia-targeting amidases, namely, Leu 130 and Tyr 131. They are 10 Å away from the active site, and they might either bind the substrate directly or interact with other cell wall components that are closely associated with the substrate. This region is relatively similar in the structure of PlyPSA, where the leucine is conserved and the tyrosine is replaced by an arginine that points away from the active site. The side chain of the arginine from PlyPSA coincides in the CD27L_1–179 structure with Lys 65. The substrate binding compartment that lies in the middle and is closest to the active site is not entirely conserved among Clostridia-targeting amidases. However, when comparing the structures of CD27L_1–179 and PlyPSA in this region, we observed the most effective differences in side-chain compositions and orientations, which can be addressed by a point mutation. The position of residue Leu 98 in CD27L is occupied by Trp 91 in PlyPSA. This side chain points toward the active site, and there is a difference of 30 Ų in accessible surface area between the two structures at this position. The Leu 98Trp mutation did result in a higher lysis activity for Listeria species, which is the native target of PlyPSA.

As a result of these observations, we would like to propose that the substrate specificity of the catalytic domain of the CD27L endolysin for certain bacteria is directed not only toward the catalytic substrate alone but also toward the composition of the surrounding cell wall. In order to cleave the substrate, the enzyme needs to penetrate the peptidoglycan layer. In some cell walls, the substrate is situated at the surface and relatively easy to access. In other cell walls, it might be buried, and the enzyme has to be able to dig in and expose the peptidoglycan substrate.

We observed that the absence of the C-terminal domain increased lytic activity and that we could steer activity with the catalytic domain alone. A similar pattern was observed for other endolysins where removal of C-terminal cell wall binding domains gave more effective lysins (6, 16, 22, 23, 25). It seems that the cell wall binding domain helps mainly the endolysin to localize to the cell wall. Once the endolysin is attached to the cell wall, the cell wall binding domain hinders the full activity of the catalytic domain. The efficiency of the endolysin depends on the binding kinetics of the cell wall binding domain, and if the binding to the cell wall is strong, then the enzyme can digest only the peptidoglycan in the immediate surroundings. There are, however, a number of endolysins that show the opposite effect, namely, a decrease in activity when the cell wall binding domain is removed (19, 24, 31, 35, 36). We propose that in this case, the cell wall binding domain is needed to help the catalytic domain access the substrate, and the two domains act together to initiate cleavage. This could be the case, for example, for choline binding lysins, where it was proposed that the substrate of the cell wall binding domain is tethered to the substrate of the catalytic domain (18). In short, we think that some endolysins share a substrate or attached substrates between the cell wall binding domain and catalytic site. In this case, the cell wall binding domain enhances the catalytic activity. Other endolysins, like CD27L, do not need a cell wall binding domain to assist in accessing the peptidoglycan substrate. In this case, the absence of a cell wall binding domain increases the effectivity of the catalytic domain. One important point to remember is that the endolysins have evolved to act from the inside of the cell to allow the release of free bacteriophages, and a cell wall binding domain could have more importance within this context than when acting externally. In addition to being designed to approach the peptidoglycan from the inner membrane, another purpose of the cell wall binding domain may be not to affect activity but to tether the lytic domain to the cell wall and prevent large-scale lysis of surrounding cells, which the emerging bacteriophages may seek to infect (24). In this case, the incorporation of such binding domains into endolysins aimed for therapeutic use may be inappropriate, and the use of truncated lysins would be more beneficial. Further examination of the relationships between the cell wall and the catalytic domain and the contribution of the C-terminal domain to activity and host range may facilitate further engineering to improve activity against C. difficile in a clinical environment while maintaining specificity.