Triple-acting Lytic Enzyme Treatment of Drug-Resistant and Intracellular Staphylococcus aureus

Multi-drug resistant bacteria are a persistent problem in modern health care, food safety and animal health. There is a need for new antimicrobials to replace over used conventional antibiotics. Here we describe engineered triple-acting staphylolytic peptidoglycan hydrolases wherein three unique antimicrobial activities from two parental proteins are combined into a single fusion protein. This effectively reduces the incidence of resistant strain development. The fusion protein reduced colonization by Staphylococcus aureus in a rat nasal colonization model, surpassing the efficacy of either parental protein. Modification of a triple-acting lytic construct with a protein transduction domain significantly enhanced both biofilm eradication and the ability to kill intracellular S. aureus as demonstrated in cultured mammary epithelial cells and in a mouse model of staphylococcal mastitis. Interestingly, the protein transduction domain was not necessary for reducing the intracellular pathogens in cultured osteoblasts or in two mouse models of osteomyelitis, highlighting the vagaries of exactly how protein transduction domains facilitate protein uptake. Bacterial cell wall degrading enzyme antimicrobials can be engineered to enhance their value as potent therapeutics.

lysostaphin fragment fused to the 3' end of pSB1101 was amplified with primers LysoSalIF and LysoXhoIR, digested with XhoI and SalI, and introduced into pSB0301 linearized at the XhoI site, generating pSB1101.
Triple fusion L-K encoded by expression vector pSB1801 was created in several steps.
The lysostaphin gene from plasmid p5301 was truncated by PCR-amplifying the M23 peptidase domain with the primers LysoAD155XhoIR and LysoAA1Ndel F and introducing this PCR product into Ndel + Xhol-digested pET21a, thereby generating pSB1701. A second intermediate construct was produced by amplification of the lysostaphin SH3b domain from plasmid template p5301 with the primers LysoSH3b SallF and pET21a StyIR, digesting with SalI + StyI, and introducing the amplified fragment into Xhol+ Styl-digested pSB0301, thereby generating plasmid pSB1001. The final triple fusion L-K construct was generated by introducing the PCR product generated by amplification of the template pSB1001 with the primers LysKaaSalF and pET21aStyIR (harboring the two lytic domains of LysK) into Xhol + Styl-digested pSB1701, thereby generating pSB1801.
Fusion of PGH to PTD sequences was accomplished by reverse translation of the 12 PTD sequences (Table 1) into nucleotide sequences (with an E. coli codon usage bias), followed by commercial synthesis (Genscript, Piscataway, NJ). The individual sequences were inserted at the XhoI site of pET21a such that each PTD coding sequence was in frame with the His6 coding sequences of the vector. The four PGH sequences (encoding mature lysostaphin, lysK, and triple fusions K-L and L-K) were introduced into these PTD-His6 encoding vectors via standard procedures following restriction enzyme digests of the parental PGH vectors (described above) at unique sites (XbaI or NdeI and XhoI) to generate DNA fragments harboring the entire PGH coding region, and ligation of these fragments into similarly digested pET21a vectors harboring the 12 individual PTD sequences.
Protein expression and purification. Recombinant PGHs were expressed in E. coli BL21 (DE3) (Novagen) grown to mid-log phase at 37°C in modified LB broth (15g tryptone, 8g yeast extract, 5g NaCl per liter, pH 7.8) with shaking, induced with 1mM IPTG, and expressed for 20 h at 10ºC. Cells were disrupted by sonication and purified via nickel affinity chromatography (Ni-NTA-agarose), per the manufacturer's instructions (Qiagen, Carlsbad CA) to >95% purity as described previously 3 . For PGHs used in animal and cell culture experiments, endotoxin was removed via Triton X-114 washes of protein on Ni-NTA columns 18 . Representative random samples were evaluated for endotoxin content via the Limulus amoebocyte lysate assay (LAL QCL-1000, Lonza, Walkersville, MD) and shown to be <5 Units/ml (e.g. for nasal colonization <5 Units/~40 mg protein) endotoxin after purification.
Preliminary characterization of PGH activity. SDS-PAGE, zymogram, turbidity reduction and plate lysis assays were performed as described previously 2 with minor modifications as follows. For SDS-PAGE and zymogram analysis, 1 µg of the purified PGH proteins and Kaleidoscope protein standards (Invitrogen, Carlsbad, CA) were analyzed by 15% SDS-PAGE. SDS-PAGE and zymogram gels were prepared and electrophoresed in parallel. Zymograms incorporated embedded cells equivalent to 300 ml of mid-logarithmic phase (OD600nm 0.4-0.6) S. aureus Newman into the SDS-PAGE matrix. The SDS-PAGE gels were stained with Coomassie blue using standard protocols, and zymograms were washed twice in excess water for 30 min to remove SDS and incubated for <1 h at room temperature in water until cleared zones developed.
For the plate lysis assays, purified enzymes were serially diluted in saline lysis buffer (SLB; 150 mM NaCl 10 mM Tris buffer, pH 7.5) with 15% glycerol, to yield concentrations of 100, 10, 1, and 0.1 pmoles/10 μl. S. aureus Newman was cultivated in tryptic soy broth (TSB) to an OD600 nm = 0.4 -0.6. The bacterial cells were harvested and diluted to yield a suspension with an OD600 nm of 0.1. Tryptic soy agar (TSA) plates were flooded with 3 ml of the bacterial cell suspension. Excess culture was removed, and the plates air dried at room temperature for ~30 min in a laminar flow hood. 10 μl of each PGH dilution was then spotted onto the air-dried lawn, allowed to air dry, and the culture plates incubated overnight at 37°C. The following day, plates were evaluated visually and photographed.
Resistance development assays. MIC repeated exposure method. MIC determinations were performed with S. aureus strain Newman. 100 µl of S. aureus strain Newman culture that survived at ½ the MIC (the first well with visible growth) for each round was inoculated into 5 ml TSB and cultivated to mid-log phase growth. This culture was used as the inoculum for the next round of MIC exposure (overnight growth), and the cycle was repeated for 10 rounds. Cells recovered from the well that represented the ½ MIC concentration on round 10 were confirmed to be S. aureus by PCR 13 . Plate lysis method. Bacterial cells were scraped from a sub-lethal (not fully cleared) spot from a plate lysis assay (described above), and these 'exposed' cells were used to inoculate 5 ml of TSB and grown for 4-6 hours to generate a new culture and lawn for subsequent exposure. Bacteria were passaged in this assay for up to 10 consecutive days, at which time the S. aureus were tested in MIC assays. Plate lysis repeated exposure experiments were performed in duplicate.

Determination of PGH minimum inhibitory concentrations (MIC).
The MIC of each protein for multiple S. aureus strains was determined as previously described 10 with the following modifications. Enzymes were serially diluted two fold across a 96 well plate such that after dilution 50 µl of the enzyme solution in 300 mM NaCl, 50 mM NaH2PO4, 30% glycerol pH 7.5, remained in each well. To each well was added 50 µl of TSB and 100 µl of S. aureus Newman in TSB (diluted to 5 x 10 6 cells/ml). The plates were incubated 20 h at 37ºC and read with a 96 well plate reader. MIC determination values were reported as the median of >4 replicates (Supplementary Table 1).

Confirmation of three lytic PGH enzyme activities in triple fusion constructs.
LysK is a 495 amino acid protein with a C-terminal SH3b CBD (SH3-5, Pfam 8 PF08460) and two lytic domains (Fig. 1A). The N-terminal lytic domain is a cysteine, histidinedependent amidohydrolase/peptidase (CHAP) domain 19 (Pfam PF05257), and the second (internal) lytic domain has been classified as an amidase-2 domain (PFAM PF01510). Both lytic domains are active and show specific cleavage sites on purified S. aureus PG 2,7 . The CHAP endopeptidase activity cleaves between the D-alanine of the stem peptide and the first glycine of the pentaglycine cross-bridge peptide, and the amidase-2 domain harbors an N-acetylmuramoyl-L-alanine amidase activity that cleaves between the N-acetylmuramic acid of the polysaccharide strand and L-alanine of the stem peptide (Supplementary Fig. 2). Mature lysostaphin is a 246 amino acid protein with a single enzymatic domain (Fig. 1A), an N-terminal M23 glycyl-glycine endopeptidase (Pfam PF01551) that cleaves between the second and third, or third and fourth, glycines of the S. aureus pentaglycine cross bridge (Supplementary Fig. 2 Fig. 3) for 1 h at room temperature, and washed twice with 100 µl SLB. Adherent bacteria were fixed with 200 µl 95% ethanol-5% glacial acetic acid for 20 min, washed once with water, and stained with 100 µl 0.4% crystal violet for 15 min at room temperature. Excess stain was removed with 3 water washes. Bound stain was resolubilized in 33% acetic acid, and 20 µl was transferred to a 96 well plate containing 180-µl water. The OD590nm was determined in a Spectra Max plate reader.
Percent reduction represents the difference in absorbance between biofilms exposed to buffer only and experimental wells that were exposed to the PGHs. Rat model of nasal colonization. The rat nasal colonization model is an adaptation of the mouse nasal colonization model that was described previously 20 . Wistar rats (5 to 6 wks of age) were obtained from Charles River Laboratories (Wilmington, MA) and given food and water ad libitum. The animals were housed two per cage in a modified barrier facility under viral antibody-free conditions. The rats were given drinking water containing streptomycin sulfate (0.5 g/liter; Sigma) 1 day prior to bacterial inoculation and for the course of the experiment. The drinking water and cages were changed three times per week. Rats were restrained briefly for inoculation in a DecapiCone bag (Braintree Scientific, Braintree, MA) and inoculated intranasally with 10 µl of 10 7 CFU S. aureus strain ALR (Sm-resistant) on day 0. The rats were treated intranasally twice a day (6-7 h apart) on days 5, 6, and 7 with 200 µg of either lysostaphin (AMBI, Tarrytown, NY) or recombinant PGHs in 20 µl buffer. Control rats received buffer alone.
On day 10 the rats were euthanized; the area around the nasal region was wiped with 70% isopropyl alcohol, and the nasal tissue was excised and homogenized on ice in 600 µl TSB. Quantitative cultures of the homogenates were prepared by plating duplicate 100 µl aliquots on TSA plates containing 0.5 mg/ml Sm (to recover the S. aureus strain used for inoculation) and one blood agar plate (to evaluate the total nasal flora). The number of S. aureus colonies from quantitative plate counts was used to estimate the CFU/nose for each rat. The results from a three independent experiments were combined. For rat nasal colonization experiments, significant (P <0.05) differences between the median values of quantitative culture results for different rat groups were compared to the control group by the Mann-Whitney test (InStat; GraphPad Software).

Ex-vivo calvaria infections with S. aureus.
Calvaria were isolated from neonatal mice and infected with UAMS-1 strain of S. aureus (1x10 6 CFU/calvaria) for 2 h at 37° C in osteoblast growth medium in the absence of antibiotics. Following 3X washes with PBS, calvaria were exposed to 50 µg/ml of gentamicin for 15 min to kill extracellular bacteria and further cultured in gentamicin containing osteoblast growth media. After 24 h, calvaria were treated with PGHs (5 µg/ml) or gentamicin only for 2 h. Lysis buffer containing 0.05% trypsin and 0.25% Triton X-100 was used to homogenize the calvaria, the homogenates were quantitatively plated on LB agar plates, and the CFU were Statistical analysis. The variables were analyzed as two-factor mixed models with Treatment as the factor and Mouse as random Blocks. This model ignores within mouse correlations. Log(TNFα) was used. The variance grouping technique was used for log CFU and log TNFα to correct for variance heterogeneity. Means comparisons were done with Sidak adjusted p-values so that the experiment-wise error was 0.05.