Characterization of Modular Bacteriophage Endolysins from Myoviridae Phages OBP, 201ϕ2-1 and PVP-SE1

Peptidoglycan lytic enzymes (endolysins) induce bacterial host cell lysis in the late phase of the lytic bacteriophage replication cycle. Endolysins OBPgp279 (from Pseudomonas fluorescens phage OBP), PVP-SE1gp146 (Salmonella enterica serovar Enteritidis phage PVP-SE1) and 201ϕ2-1gp229 (Pseudomonas chlororaphis phage 201ϕ2-1) all possess a modular structure with an N-terminal cell wall binding domain and a C-terminal catalytic domain, a unique property for endolysins with a Gram-negative background. All three modular endolysins showed strong muralytic activity on the peptidoglycan of a broad range of Gram-negative bacteria, partly due to the presence of the cell wall binding domain. In the case of PVP-SE1gp146, this domain shows a binding affinity for Salmonella peptidoglycan that falls within the range of typical cell adhesion molecules (Kaff = 1.26×106 M−1). Remarkably, PVP-SE1gp146 turns out to be thermoresistant up to temperatures of 90°C, making it a potential candidate as antibacterial component in hurdle technology for food preservation. OBPgp279, on the other hand, is suggested to intrinsically destabilize the outer membrane of Pseudomonas species, thereby gaining access to their peptidoglycan and exerts an antibacterial activity of 1 logarithmic unit reduction. Addition of 0.5 mM EDTA significantly increases the antibacterial activity of the three modular endolysins up to 2–3 logarithmic units reduction. This research work offers perspectives towards elucidation of the structural differences explaining the unique biochemical and antibacterial properties of OBPgp279, PVP-SE1gp146 and 201ϕ2-1gp229. Furthermore, these endolysins extensively enlarge the pool of potential antibacterial compounds used against multi-drug resistant Gram-negative bacterial infections.

The use of externally applied endolysins from phages infecting Gram-positive bacteria as antibacterial compounds has proven successful in various applications [4]. However, the use of endolysins from Gram-negative origin against Gram-negative pathogens is an unexploited research focus as these bacteria possess an outer membrane (OM) leaflet which prevents efficient peptidoglycan lysis [5].
In this context, Pseudomonas aeruginosa phage endolysins KZ144 (from giant phage QKZ) and EL188 (giant phage EL) were recently shown to have lytic activity on the PG of P. aeruginosa PAO1 [6]. KZ144 and EL188 possess a modular structure composed of an N-terminal cell wall binding domain and a Cterminal catalytic domain which is rare amongst endolysins from Gram-negative origin [6,7]. The presence of this cell wall binding domain is linked with the high muralytic activity of KZ144 and EL188 on OM permeabilized P. aeruginosa [6].
In this paper we expand the pool of modular endolysins from Gram-negative infecting Myoviridae phages with three new candidates (OBPgp279, 201Q2-1gp229 and PVP-SE1gp146) and extensively describe and compare them at the biochemical and antibacterial level. These data help to elucidate their fundamental properties and emphasize their potential as antibacterial agents towards treatment of Gram-negative infections.

Materials and Methods
Bacterial strains P. aeruginosa PAO1 from two different sources was used in this study: a PAO1 strain donated by prof. V. Krylov (Institute of Genetics and Selection of Industrial Microorganisms in Moscow, Russia) for the biochemical characterization and peptidoglycan binding assay, and a PAO1 strain from Pirnay [8] for the in vitro antibacterial assay. Both strains only differ in pili formation, a difference which has no further impact on the obtained results. For this reason, we consistently named the strain PAO1 throughout the manuscript. The clinical P. aeruginosa Br667 strain was a burn wound isolate from the Military Queen Astrid Hospital (Brussels, Belgium) that is multidrug-resistant against 10 out of 11 tested antibiotics due to reduced OM permeability [8]. Escherichia coli XL1-Blue and E. coli BL21(DE3)pLysS are commercially available expression strains (Agilent Technologies, Santa Clara, CA, USA). Salmonella enterica serovar Typhimurium LT2 (further abbreviated as S. Typhimurium LT2), Staphylococcus aureus subsp. aureus Rosenbach ATCC 6538, Micrococcus lysodeikticus ATCC 4698, Bacillus subtilis PSB3 and Lactococcus lactis subsp. lactis were kindly provided by the Centre of Food and Microbial Technology (University of Leuven, Belgium). All these strains were grown at 37uC in Lysogeny Broth (LB).

Isolation and cloning of endolysin open reading frames and derivatives
Open reading frame (ORF) 279 from phage OBP, ORF 229 from phage 201Q2-1, ORF 146 of PVP-SE1 and the sequences encoding the individual catalytic domains of each endolysin were all amplified (Pfu polymerase, ThermoFischer Scientific, Waltham, MA, USA, Table S1) using genomic DNA template isolated from the respective phages. All obtained PCR products were cloned into the pEXP5CT-TOPOH expression vector according to the manufacturer's manual (Life Technologies, Carlsbad, CA, USA) and sequence-verified (Big Dye polymerase, ABI3130 Sequencher, Life Technologies). The pEXP5CT-TOPOH vector provides a Cterminal His 6 -tag for efficient Ni 2+ -NTA purification.
To obtain fusion proteins of the peptidoglycan binding domains (PBD) of endolysins OBPgp279, PVP-SE1gp146 and 201Q2-1gp229 to the Enhanced Green Fluorescent Protein (EGFP), the corresponding sequences encoding the different PG binding moieties of the endolysins were amplified (Pfu polymerase, Table  S1) and spliced to the EGFP encoding sequence (Takara, Otsu, Japan) using a BamHI restriction site. The fusion fragments were cloned into the pEXP5CT-TOPOH expression vector for protein expression and purification.

Large scale expression and purification of endolysins and their derivatives
All proteins were expressed for 4 hours at 37uC or for 18 hours at 16uC in E. coli BL21(DE3)pLysS cells (Life Technologies) upon induction of mid-exponential growing cells (optical density at 600 nm wavelength (OD 600 nm ) of 0.6) using 1 mM of isopropyl-b-D-thiogalactopyranoside (IPTG). After expression, the bacterial cell pellet was resuspended in a lysis buffer (0.5 M NaCl, 20 mM of NaH 2 PO 4 /NaOH, pH 7.4), and disrupted with a combination of freeze-thawing (280uC/room temperature) and sonication (10 cycles of 30 s pulse and 30 s rest, Vibra-Cell TM Sonics and Materials, Newtown, CT, USA). For efficient lysis of expression strains producing the EGFP fusion proteins, 10 mg Hen Egg white Lysozyme (HEWL, Across Organics, Belgium) was added to the lysis buffer. Recombinant endolysin was purified on Ni 2+ -NTA resin stacked HisTrap TM HP 1 ml columns (GE Healthcare, Waukesha, WI, USA) using protein-dependent imidazole concentrations in the washing buffer (0.5 M NaCl, 20 mM of NaH 2 PO 4 / NaOH, 60-70 mM imidazole, pH 7.4) ( Table S1). Purity of the recombinant proteins was estimated to be more than 90% by SDS-PAGE. Protein concentrations were determined spectrophotometrically after dialyzing against a phosphate buffered saline (PBS) pH 7.4 buffer using Slide-A-LyzerH MINI dialysis units (ThermoFischer Scientific, Table S1).

Peptidoglycan binding capacity assay
The binding capacity of the PBDs in OBPgp279, PVP-SE1gp146 and 201Q2-1gp229 was experimentally verified using a fluorescence assay modified for Gram-negative cell walls [6]. For this assay, P. aeruginosa PAO1 or S. Typhimurium LT2 cells with permeabilized OM are used, according to [9]. Summarized, midexponentially growing P. aeruginosa PAO1 or S. Typhimurium LT2 cells (OD 600 nm = 0.6) were spun down and subsequently incubated for 45 min in chloroform-saturated 0.05 M Tris/HCl (pH 7.7) at room temperature. Subsequently, cells were washed in a KH 2 PO 4 /K 2 HPO 4 buffer (ionic strength of 80 mM, pH 7.2) to remove residual chloroform and concentrated to an OD 600 nm of 1.5. The fusion proteins OBP 1-117 -EGFP, OBP 7-54 -EGFP, OBP 57-117 -EGFP, PVP 1-63 -EGFP, 201Q2-1 8-63 -EGFP and the wild type (wt) EGFP (negative control) were incubated at a concentration of 2 mM with the OM permeabilized P. aeruginosa PAO1 and S. Typhimurium LT2 cells, in a total volume of 200 ml for 5 min at room temperature. Unbound fusion proteins were removed with two consecutive washing steps in 5 mM HEPES/ NaOH buffer (pH 7.2). Cells were visualized using epifluorescence microscopy (Leica type DMLB DS 0.6.23./100 S/F, McBain instruments, Chatsworth, NJ, USA, equipped with a LED light source Fluo LED 4000, McBain Instruments and a GFP filter). At conditions where no binding occurred, epifluorescence was overlaid by phase contrast to visualize uncolored cells.
To quantify binding, cell suspensions (100 ml) were incubated with 2 mM fusion protein in a total volume of 200 ml. Subsequently, the labeled cells were washed twice with 5 mM HEPES/NaOH buffer (pH 7.2), transferred to microplate wells and placed in the Fluoroskan Ascent FL fluorescence reader (ThermoFischer Scientific). Fluorescence was measured at 520 nm, using an excitation wavelength of 480 nm. The OD 600 nm was measured in parallel and fluorescence was expressed per unit of OD 600 nm to take small differences in cell density into account.

Peptidoglycan binding affinity measured by surface plasmon resonance
Surface plasmon resonance (BIAcore X, BIAcore AB, Uppsala, Sweden) was used to obtain sensorgrams for binding of the fusion protein PVP 1-63 -EGFP to the surface of immobilized, autoclaved S. Typhimurium LT2 cells [10]. In a first step, the carboxymethylated surface of a BIAcore Pioneer C1 chip was coated with 0.5 mg/ml PVP 1-63 -EGFP (flow rate = 5 ml/min; 10 mM sodium acetate pH 5.0), using the amine coupling technique. Ethanolamine was added to block unbound positions on the chip. Secondly, autoclaved S. Typhimurium LT2 cells (density of 3610 10 cells/ml) resuspended in HBS-T buffer (10 mM HEPES, 150 mM NaCl, 3.4 mM EDTA, 0.005% Tween20, pH 7.8) were applied to the chip surface containing covalently bound PVP 1-63 -EGFP (total volume = 30 ml, flow rate = 3 ml/min). The obtained double layer served as the basis for the binding affinity measurements, performed at 25uC with 2 different concentrations (5 and 10 mM) of PVP 1-63 -EGFP for 3 min (association phase) and 12 min (dissociation phase) (flow rate = 10 ml/min). Each concentration was measured in three-fold. Regeneration of the coated chip in between different measurements takes place by several washing steps with increasing NaCl concentrations (2 M, 3 M, 4 M; 50 ml; 10 ml/min). Kinetic data are calculated based on the measurements using the BIA-evaluation software (version 3.0, BIAcore).

Quantification and characterization of muralytic activity
The PG lytic (or muralytic) activity of the purified proteins and their catalytic domains were quantified on P. aeruginosa PAO1, S.
Typhimurium LT2 and E. coli XL1-Blue cells with a chloroform/ Tris permeabilized OM [9]. In this assay, 30 ml of enzyme was added to 270 ml of OM permeabilized cells, resulting in a decrease of the optical density over time, as measured spectrophotometrically at 655 nm using a Microplate Reader 680 (Bio-Rad, CA, USA). A standardized calculation method to quantify the muralytic activities of lytic enzymes using this assay has been optimized and extensively described [11]. Obtained values for the negative control (30 ml of PBS pH 7.4) were subtracted from the sample values.
The influence of pH on the muralytic activity was assessed using OM permeabilized cells resuspended in an universal pH buffer (150 mM KCl, 10 mM KH 2 PO 4 , 10 mM Na-citrate and 10 mM H 3 BO 4 ) adjusted to different pH's between 3 and 10. To investigate the thermal stability of OBPgp279, 201Q2-1gp229 and PVP-SE1gp146 after heat treatment, 5 mM of each endolysin was incubated at 42uC and 50uC for different time-intervals (1, 2, 3, 4, 8 and 24 hours) in a Biometra T3000 Thermocycler (Göttingen, Germany) followed by a cooling step to room temperature. Proteins showing full activity at 50uC were heated up to 60, 70, 80, 90 and 100uC during 20, 40 and 60 min timeintervals and cooled down again. The residual muralytic activity of each sample relative to the activity of unheated reference sample at time 0 ( = 100% activity) was determined.

In vitro antibacterial activity assay
Mid-exponentially growing P. aeruginosa PAO1, P. aeruginosa Br667, E. coli XL1-Blue and S. Typhimurium LT2 cells (OD 600 nm = 0.6) were 100-fold diluted in 5 mM HEPES/NaOH (pH 7.2) to a final density of 10 6 colony forming units/ml. Each cell culture (100 ml) was incubated for 30 min at room temperature with 50 ml OBPgp279 (1.5 mM final concentration), PVP-SE1gp146 (5 mM) or 201Q2-1gp229 (3 mM) dialyzed against a PBS buffer (pH 7.4) together with 50 ml of 5 mM HEPES/NaOH (pH 7.2) buffer or 50 ml of EDTA (final concentration of 0.5 mM) dissolved in the same buffer. As negative control, 100 ml of cells were added to 50 ml of 5 mM HEPES/NaOH (pH 7.2) and 50 ml of a PBS buffer (pH 7.4). After incubation, cell suspensions were diluted in three-fold to 10 5 , 10 4 and 10 3 cells/ml and 100 ml of each dilution was plated out on LB medium. Colonies were counted after an overnight incubation at 37uC. The antibacterial activity was quantified as the relative inactivation in logarithmic units ( = log 10 (N 0 /N i ) with N 0 = number of untreated cells (in the negative control) and N i = number of treated cells counted after incubation).
The N-terminal domains of OBPgp279, 201Q2-1gp229 and PVP-SE1gp146 possess specific repeated motifs which are conserved in the confirmed PBDs of the modular endolysins KZ144 and EL188 [6]. These repeated motifs largely matched the consensus sequence (D-G-(Pho) 2 -G-K/N-G/N-T, Pho = hydrophobic amino acid) proposed previously [6]. Two such motifs (D- Figure S1). This conserved feature is seen in many proteins that interact with repeated structures such as PG [20]. For OBPgp279, seven additional amino acids (L-A-X-Pho-X-L-Y, X = same hydrophilic amino acid, Pho = hydrophobic amino acid) were detected in front of both repeated motifs with four amino acids in between, strengthening the idea that the PBDs of OBPgp279 were duplicated throughout evolution. The presence of these PG binding motifs inside the N-terminal domains strongly supported the predicted PG binding function. Furthermore, the repeated motif observation suggested the presence of a general sequence conserved among all PBDs of modular endolysins with a Gramnegative background.
N-terminal cell wall binding domains of OBPgp279, PVP-SE1gp146 and 201Q2-1gp229 Peptidoglycan binding capacity. To evaluate the PG binding capacity, we produced the fusion proteins OBP 1-117 -EGFP, PVP 1-63 -EGFP and 201Q2-1 8-63 -EGFP comprising the predicted binding domains of OBPgp279 (amino acids 3 to 117), PVP-SE1gp146 (amino acids 3 to 38) and 201Q2-1gp229 (amino acids 8 to 63), N-terminally fused to EGFP. For this purpose, OM permeabilized P. aeruginosa PAO1 and S. Typhimurium LT2 cells [9] were incubated with 2 mM fusion protein. Cells were washed to remove unbound protein and the retained fluorescence of PG bound protein was quantified ( Figure 2A). A negative wt EGFP control and a positive KZ 1-83 -EGFP [6] control were included for comparison. The results showed that the N-terminal domains of both OBPgp279 and PVP-SE1gp146 have PG binding activity, since the measured fluorescence levels under saturating concentrations were comparable with KZ 1-83 -EGFP [6]. This observation also indicated that the same number of binding sites was occupied by the fusion proteins. For 201Q2-1 8-63 -EGFP, no fluorescence was detected, possibly due to improper folding of the fusion protein during expression. Binding capacity of OBP 1-117 -EGFP and PVP 1-63 -EGFP was visually confirmed using epifluorescence microscopy ( Figure 2B). The cell wall of the targeted P. aeruginosa PAO1 and S. Typhimurium LT2 cells became fluorescent in less than 5 min after incubation with the fusion constructs OBP 1-117 -EGFP and PVP 1-63 -EGFP. However, a single subdomain (either OBP 7-54 or OBP 57-117 ) of OBPgp279 was not sufficient for cell wall binding since no fluorescence was retained with any of both fusion proteins (Figure 2A). This could be explained because of improper folding of the single subdomains in the absence of the other subdomain, or because both subdomains were needed simultaneously for effective PG binding.
Peptidoglycan binding affinity of PVP 1-63 -EGFP. For PVP 1-63 -EGFP, the interaction strength with the PG of S. Typhimurium LT2 ( = ligand) was analyzed using surface plasmon resonance analysis [10]. This technique made it possible to follow the interaction between the fusion proteins and immobilized Gram-negative cells in real time and to determine the binding affinity constant of each interaction. For this, autoclaved S. Typhimurium LT2 cells were immobilized on a PVP 1-63 -EGFP coated sensor chip. To this double layer, free PVP 1-63 -EGFP proteins were added allowing us to observe the interaction between the fusion proteins and the immobilized cells.
To calculate the apparent affinity constant K aff for binding of the fusion proteins to the Gram-negative cell wall, we used an indicative two-state model which assumes a 1:1 binding (AB) between fusion protein (A) and ligand (B) occurs followed by a conformational change in the complex (AB .. ): This proposed model enabled us to calculate the association (k a1 ), dissociation (k d1 ), forward (k a2 ) and backward (k d2 ) rate constants from which a K aff constant of 1.26610 6 M 21 (or a dissociation constant of 7.93610 27 M) was determined for fusion protein PVP 1-63 -EGFP (Table 1). The PG binding affinity constant of the PBD of PVP-SE1gp146 fell within the range of average SH2 domains and typical cell adhesion molecules [21].

Biochemical properties of OBPgp279, 201Q2-1gp229 and PVP-SE1gp146
Substrate specificity. To confirm the predicted catalytic activity and to determine the bacterial host substrate spectrum, each endolysin (1 mM final concentration) was incubated with three different OM permeabilized Gram-negative cell substrates (P. aeruginosa PAO1, S. Typhimurium LT2 and E. coli XL1-Blue) and four intact Gram-positive bacteria with different structural types of peptidoglycan (S. aureus subsp. aureus Rosenbach ATCC 6538, M. lysodeikticus ATCC 4698, B. subtilis PSB3 and L. lactis subsp. lactis) ( Figure S2). All three OM permeabilized Gramnegative cell substrates were efficiently lysed by the tested endolysins as shown by the decrease in OD, demonstrating their muralytic properties. In contrast, all tested Gram-positive bacteria survived the treatment even with an excess of OBPgp279, PVP-SE1gp146 and 201Q2-1gp229 ( Figure S2).
Quantification of muralytic activity. The muralytic activities of OBPgp279, 201Q2-1gp229 and PVP-SE1gp146 were first quantified on OM permeabilized P. aeruginosa PAO1 as described previously [11]. At the optimal pH for enzymatic activity of 7.2 which corresponded to the pH of the bacterial cytoplasm ( Figure  S3), lysis of P. aeruginosa PAO1 could be detected upon addition of 10 to 20 nM endolysin. Enzymatic activity further increased linearly dependent on the endolysin concentration. A saturated enzymatic activity was reached at concentrations of 50, 200 and 100 nM for OBPgp279, 201Q2-1gp229 and PVP-SE1gp146, respectively (data not shown). Based on the linear relation between lysis and concentration, muralytic activity was expressed in enzymatic units (Table 2) [11]. OBPgp279 possessed a 1.5 to 6 times higher muralytic activity value than PVP-SE1gp146 and 201Q2-1gp229, respectively, and a 5 to 10 times higher value compared to the previously reported modular endolysins EL188 and KZ144 [6]. As expected, muralytic activity values determined on OM permeabilized S. Typhimurium LT2 and E. coli XL1-Blue were consistent with those obtained for P. aeruginosa PAO1 (data not shown).
In addition, we tested the necessity of the peptidoglycan binding domain for proper muralytic activity. Therefore, the individual Cterminal catalytic domains of OBPgp279 (OBP 127-327 ), PVP-SE1gp146 (PVP 82-235 ) and 201Q2-1gp229 (201Q2-1 75-260 ) were cloned and purified to compare their activity with the obtained value for the full length endolysin. In the case of PVP 82-235 , no soluble protein could be obtained despite changes in expression parameters. Expression of both OBP 127-327 and 201Q2-1 75-260 yielded pure protein in large amounts for activity testing (Table  S1). In the absence of the PBD, the individual catalytic domains of OBPgp279 and 201Q2-1gp229 were still active, retaining 62 and 44% of the muralytic activity of the full length endolysins on OM permeabilized P. aeruginosa PAO1, respectively ( Table 2). For OBP 127-327 , the residual value was significantly higher than the full length modular endolysins 201Q2-1gp229, KZ144 and EL188. These results indicated that both catalytic domains could work independently from the PBD part and emphasized that the PBD also had an important contribution to the total enzymatic activity of the full length modular endolysins. Thermoresistance. All three endolysins could be stored at 4uC in elution buffer for several weeks and months without loss of activity. The residual muralytic activity of each endolysin on OM permeabilized P. aeruginosa PAO1 was determined after a heat treatment at 42uC and 50uC and a subsequent cooling step to room temperature. No reduction of enzymatic activity was observed after prolonged incubation of the enzymes at 42uC (data not shown). When heated at 50uC, the muralytic activity of OBPgp279 and 201Q2-1gp229 dropped below 10% of the initial activity after 2 and 1 h incubation, respectively ( Figure 3A). In contrast, PVP-SE1gp146 remained fully active, even after 24 h of incubation. At higher temperatures, PVP-SE1gp146 retained its maximal activity at 80uC incubation for 1 h and was only completely inactivated after 40 min at 100uC ( Figure 3B).

In vitro antibacterial activity with and without the OM permeabilizer EDTA
The antibacterial activity of OBPgp279, 201Q2-1gp229 and PVP-SE1gp146 was investigated on the Gram-negative bacteria P. aeruginosa PAO1, P. aeruginosa Br667, E. coli XL1-Blue and S. Typhimurium LT2, and compared with endolysins KZ144 and EL188 (Table 3A). To overcome the OM barrier, 0.5 mM EDTA was added (Table 3B). EDTA is extensively described as an effective OM permeabilizer [22,23] and was shown to promote the penetration of endolysin EL188 from P. aeruginosa infecting phage EL through the OM of clinical P. aeruginosa isolates in mM concentrations [24].
The addition of OM permeabilizer EDTA (Table 3B) caused a significant increase in antibacterial activity of OBPgp279 on all four strains. This observation was more pronounced for P. aeruginosa PAO1 (3 additional log units) and P. aeruginosa Br667 (2 additional log units) than for E. coli XL1-Blue and S. Typhimurium LT2 (both 0.70 additional log units). For endolysins PVP-SE1gp146 and 201Q2-1gp229, only P. aeruginosa PAO1 was efficiently reduced in presence of EDTA.
Remarkably, OBPgp279 reduced P. aeruginosa PAO1 and the multi-drug resistant P. aeruginosa Br667 with 1.1060.07 and 1.0860.08 log units, respectively, in the absence of OM destabilizing agents. Contrarily, E. coli XL1-Blue and S. Typhimurium LT2 showed no sensitivity to the endolysins action. Both PVP-SE1gp146 or 201Q2-1gp229 did not influence the two P. aeruginosa strains, neither did KZ144 and EL188 (Table 3A).

Modularity feature in endolysins of Gram-negative origin
Based on their structural and biochemical characteristics, OBPgp279, PVP-SE1gp146 and 201Q2-1gp229 are assumed to play an important role in the bacterial host lysis at the end of the phage replication cycle. In silico, these three endolysins were shown to share a modular structure with a PBD at the N-terminal side of the protein and a catalytic domain at the C-terminal side. For OBPgp279 and PVP-SE1gp146, the presence of the PBD was experimentally confirmed. This modular composition is rather rare among endolysins with Gram-negative background, as the majority of these endolysins are globular with only a catalytic domain [25]. Even though they share the same general function and modular composition, OBPgp279, PVP-SE1gp146 and 201Q2-1gp229 are unrelated at the sequence level. This observation emphasizes the relevance of the biochemical and antibacterial characterization of these enzymes in order to discover new and desirable traits for use as bio-control tools.
We may question why some endolysins of Gram-negative infecting phages have a PBD contained within their structure. For Gram-positive infecting phages, the presence of a PBD is speculated to be necessary to keep the enzyme bound to the cell wall debris. In this way, surrounding bacteria (which are potential candidates for a new infection) remain protected from the hydrolyzing activity of newly released endolysin [25]. In this respect, Loessner and coworkers [26] demonstrated that the Listeria monocytogenes phage endolysins Ply118 and Ply500 have a binding affinity (3- Not only endolysins of phiKZ-related phages feature a modular composition [6], but also members of other unrelated and smaller myoviruses as shown here for S. Enteritidis phage PVP-SE1 (''rV5like virus'') and as described in literature for P. aeruginosa phage QCTX (''P2-like virus'') [27] and Burkholderia cepacia phages Bcep781 and Bcep1 (''Bcep789-like viruses'') [28]. Modular endolysins are assumed to exert a higher enzymatic activity than non-modular or globular ones due to the presence of an additional   Cell cultures (initial cell density of 10 6 cells/ml) were incubated for 30 min with 1. PG binding domain which keeps the enzyme in close proximity to the PG substrate [25]. This assumption is confirmed here: the muralytic activity of the modular endolysins OBPgp279, PVP-SE1gp146 and 201Q2-1gp229 is 160, 104 and 30 times higher compared to the value for the globular phage T4 lysozyme [29].
For an unknown reason, some unrelated phages possess strong and fast-acting modular endolysins instead of less active, globular ones.

Biochemical activity of modular endolysins
Based on their biochemical activity, OBPgp279 and PVP-SE1gp146 degrade the PG layer of the tested Gram-negative species with the highest efficiencies compared to 201Q2-1gp229, EL188 and KZ144. In contrast to the catalytic transglycosylase specificity of KZ144 and EL188 [6,30], OBPgp279, PVP-SE1gp146 and 201Q2-1gp229 are predicted to have lysozymelike activity. Despite the similar catalytic function, a large variation in muralytic activity values is observed for the individual catalytic domains of OBPgp279 and 201Q2-1gp229, due to differences in the primary protein structure. This variation only partially explains the large differences in muralytic activity values of the full length endolysins, since the PBD part also influences the muralytic activity in a protein dependent manner.
Previously, Briers and co-workers [6] revealed the insensitivity of Gram-positive bacteria for the muralytic action of modular endolysins KZ144 and EL188. We found that also OBPgp279, PVP-SE1gp146 and 201Q2-1gp229 leave the Gram-positive bacteria S. aureus subsp. aureus Rosenbach ATCC 6538, M. lysodeikticus ATCC 4698, L. lactis subsp. lactis and B. subtilis PSB3 unharmed. An explanation must be found in the structure of their PG layer. Sensitive Gram-negative species share a common A1c chemotype of PG with a direct peptide bond between adjacent muropeptides, whereas insensitive S. aureus subsp. aureus Rosenbach ATCC 6538, M. lysodeikticus ATCC 4698 and L. lactis subsp. lactis have a PG of the A3a, A2a and A4a chemotype, respectively [31,32]. These last 3 chemotypes show resistance for the mode of action of muralytic enzymes, like HEWL, due to the absence of a direct inter-peptide cross-link [33]. Bacillus subtilis PSB3, however, possesses the A1c chemotype of its Gram-negative counterparts but 17.3% of the glucosamine sugars in its PG are N-deacetylated [34]. As previously shown for commercial HEWL, this species-specific modification explains for the resistance against PG lytic enzymes [33]. Therefore, a direct cross-linked A1c chemotype of PG without substantial N-deacetylation is an essential prerequisite for efficient hydrolysis by endolysins from Gram-negative infecting phages.

Thermoresistance of PVP-SE1gp146
An unexpected result is the observed thermoresistance of endolysin PVP-SE1gp146 at temperatures up to 90uC. To our knowledge, PVP-SE1gp146 is the first Gram-negative phage derived endolysin described so far to show thermoresistance. For the structural lysin gp36 of P. aeruginosa phage QKMV (KMV36C) only the catalytic domain-containing C-terminus was thermoresistant [35]. Heat-inactivated, unfolded KMV36C has the opportunity to refold back again into its (partially) active form after cooling and therefore retains its complete enzymatic activity. Recently, this mechanism was also suggested by Schmelcher and coworkers [36] to explain for the high thermostabilty of Grampositive Listeria monocytogenes phage endolysins HPL118 and HPL511. In their study, HPL118 and HPL511 retained 35% of activity after 30 min incubation at 90uC. The authors prove that the unfolding of HPL118 and HPL511 at higher temperatures is followed by a rapid refolding upon cooling. This proposed refolding mechanism may preserve the enzyme activity and could also explain the thermoresistance of PVP-SE1gp146. From this perspective, PVP-SE1gp146 could be an interesting candidate as antibacterial component in a hurdle approach for food preservation, in addition to other antibacterial additives like bacteriocin and nisin [37]. Because of the antibacterial and thermoresistant properties of PVP-SE1gp146, the use of chemical preservatives and the intensity of heat treatments could be reduced in such a way that quality foods with higher nutritional values are obtained.

Potential use of OBPgp279 as an enzybiotic
Despite the presence of an OM which shields the PG layer from the externally added endolysin, OBPgp279 is capable to reduce both P. aeruginosa PAO1 and the multi-drug resistant P. aeruginosa Br667, contrary to PVP-SE1gp146, 201Q2-1gp229, KZ144 and EL188. Therefore, OBPgp279 is a potential candidate for the use as enzybiotic to control multi-drug resistant opportunistic Pseudomonas infections in human and animals. In addition to its high muralytic activity, OBPgp279 is suggested to intrinsically destabilize the OM in the Pseudomonas cell wall so that the endolysin reaches its enzymatic target, which could explain the observed reductions. A similar characteristic was suggested for the Bacillus amyloliquefaciens phage IAM1521 endolysin Lys1521 [38]. In the case of this endolysin, the presence of positively charged amino acids in the C-terminal cell wall binding part that interact with the OM and destabilize it, explained the antibacterial property of the endolysin [39]. For OBPgp279, a similar mechanism is unlikely to induce the proposed OM destabilization as its C-terminal end does not contain more positively charged amino acids compared to PVP-SE1gp146 and 201Q2-1gp229, which do not affect the tested Pseudomonas strains. Furthermore, no obvious amphipathic helixes which are able to intercalate into the OM structure resulting in OM permeabilization [40] were detected at the N-and C-terminal ends of OBPgp279. Further structural analysis of the interaction between OBPgp279 and the Pseudomonas cell wall will be necessary to reveal the mechanism and determinants responsible for the observed Pseudomonas sensitivity.

Combinatorial endolysin/EDTA approach
To allow for a better passage of endolysins through the OM a combinatorial approach consisting of an endolysin with an OM permeabilizer was used. In this respect, Briers and coworkers [24] recently demonstrated the efficacy of a combination of an endolysin with Gram-negative background (i.e. endolysin EL188 of P. aeruginosa infecting phage EL) and the OM permeabilizer EDTA for antibacterial reduction of the Gram-negative pathogen P. aeruginosa. The positive effect of EDTA on antibacterial activity is caused by the powerful OM permeabilization capacity of EDTA through binding and withdrawal of the stabilizing divalent Mg 2+ and Ca 2+ cations present in the LPS layer of the Gram-negative OM [20]. As a result, the bacterial PG becomes more prone to the muralytic activity of the externally added endolysins. As with KZ144 and EL188, the three tested modular endolysins induce efficient reduction of the P. aeruginosa PAO1 when applied in combination with EDTA. In addition, the OBPgp279/EDTA combination also affects the multidrug-resistant P. aeruginosa Br667 strain [8], hinting at the applicability of this combinatorial strategy towards multidrug-resistant P. aeruginosa strains as alternative for antibiotic treatment. However, due to the low amount of phosphate groups per LPS molecule and the corresponding amount of stabilizing divalent cations in the OM of Enterobacteriaceae compared to that of Pseudomonas species [41], the sensitivity of E. coli and S. Typhimurium strains for the destabilizing EDTA action is much lower. A similar observation was made for endolysin EL188 combined with EDTA [24]. Consequently, the type and structure of the OM seems of crucial importance for the antibacterial efficiency of the endolysin/EDTA combinatorial approach. However, the elucidation of the unique biochemical and antibacterial characteristics of OBPgp279, PVP-SE1gp146 and 201Q2-1gp229 (modularity, thermoresistance, OM destabilizing activity) allows us to gain a better insight in the relationship between the structure and activity of endolysins, in general. In this way, we take an important first step in development of future antibacterial therapies based on modular Gram-negative endolysins.  Figure S3 pH optimization for enzymatic activity of OBPgp279 (diamonds), PVP-SE1gp146 (squares) and 201Q2-1gp229 (triangles). Muralytic activity is measured as the slope of the DOD 655 nm /min curve (Y-axis) on OM permeabilized P. aeruginosa PAO1 substrate and is shown for a pH range between 3 and 10 (X-axis). Final concentrations used here are 1 mM OBPgp279, 5 mM PVP-SE1gp146 and 3 mM 201Q2-1gp229, each dialyzed against a PBS buffer (pH 7.4). Averages and standard deviations of three repeated experiments are given. (TIF)

Table S1
Overview of all expression constructs used in this study. The specific insert, cloning vector, primers, expression/ purification conditions and expression yield are indicated for each construct. The restriction endonuclease recognition sites inside the primers are underlined. Extra nucleotides necessary for in frame cloning are indicated in bold. All constructs were expressed using the E. coli BL21(DE3)pLysS expression strain in LB medium upon induction with 1 mM IPTG. Proteins were purified on Ni 2+ -NTA columns using protein dependent imidazole concentrations given in the table. Expression yield is indicated in mg of obtained recombinant protein per liter expression culture as determined spectrophotometrically. (DOCX)