The E. coli strains RS218 (O18:K1:H7) (Achtman et al., 1983), ATCC 23506 (O10:K5(L):H4), and E69 (O9:K30) (Orskov et al., 1977) were used for mouse infection, capsule isolation and serum sensitivity assays. E. coli BL21(DE3) was used for protein expression and purification. Bacteria were grown in LB (10 g tryptone, 5 g yeast extract, 10 g NaCl per liter) broth in 37°C shakers unless otherwise noted. The concentration of viable bacteria was determined by colony counts using LB agar (1.3% w/v) plates.

The E. coli strain EV36 (Vimr and Troy, 1985) was used to propagate the coliphages K1E, K1F, and K1H (Bull et al., 2010). E. coli ATCC 23506 and E69 were used for propagation, respectively, of the coliphages K1-5 (Scholl et al., 2001, 2004) and K30 (Whitfield and Lam, 1986). Phages were grown and purified as previously described (Scholl et al., 2001; Leiman et al., 2007). Briefly, the coliphages were added to bacterial cultures at an OD₆₀₀ = 0.25–0.4 at a multiplicity of infection of 4, followed by incubation at 37°C with aeration until the culture cleared. Phages were precipitated with 0.5 M NaCl and 10% PEG 8000, and then purified by equilibrium CsCl gradient centrifugation in SM buffer (50 mM Tris–HCl, 100 mM NaCl, 8 mM MgSO₄, pH 7.5) supplemented with CsCl to a density of 1.5 g/ml. After dialysis into SM buffer using 12–14 kDa MWCO dialysis membranes (Spectrum), phage titers were determined by plaque counts using a LB soft agar (0.65%) overlay.

Phage genomic DNA was extracted as described for phage λ (Sambrook et al., 1989), and then was used as template to amplify the depolymerase genes K1E, K1F, K1H, K5 or those for the putative enzymes K30 gp41 and K30 gp42. Gene information and the PCR primers are listed in Supplementary Table S1. PCR products were then assembled into NdeI- and EcoRI-digested pET28b (EMD Biosciences Inc.) using the Gibson Assembly Master Mix (NEB Inc.).

After overexpression of their genes, capsule depolymerases were purified essentially as previously described (Leggate et al., 2002). Briefly, the pET28b derivatives were transformed into E. coli BL21(DE3). Expression of the recombinant His-tagged depolymerase genes was induced at A₆₀₀ = 0.6 with 0.5 mM isopropyl-β-D-thiogalactopyranoside (IPTG) followed by overnight growth at 20°C. Cells were lysed by sonication in lysis buffer (50 mM Na₂HPO₄, 300 mM NaCl, 10 mM imidazole, pH 7.5) and the depolymerases were purified using HisPur Ni-NTA resin (Thermo Fisher Scientific Inc.) according to the user guide. After dialysis into phosphate-buffered saline (PBS) buffer (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, pH 7.5) using 3.5 kDa MWCO dialysis membranes (Spectrum), the depolymerases were used directly in all experiments. Protein concentrations were estimated by A₂₈₀, using a Nanodrop ND-1000.

Purified proteins were analyzed by SDS-PAGE and size exclusion chromatography. SDS-PAGE was performed using a 10% resolving gel with a 4% stacking gel. Proteins were denatured at 100°C for 5 min. After electrophoresis, proteins were stained with Coomassie brilliant blue. Size exclusion chromatography was performed on an AKTA FPLC (GE Healthcare) at 4°C. Depolymerases, in 25 mM sodium phosphate, 150 mM NaCl, pH 7.5, were applied to a Superose 6 10/300 GL column (GE Healthcare). Elution was at a flow rate of 0.4 ml/min and proteins were detected at A₂₈₀. Molecular weights were estimated using the high molecular weight gel filtration calibration kit (GE Healthcare).

Mouse work conformed to NIH guidelines and the University of Texas IACUC protocol approval (AUP-2015-00035). 4–6 weeks old female NIH Swiss outbred mice (Envigo Inc.) weighing 20–25 g were used. For infections, 1 – 4 × 10⁸ CFU (colony forming units) of bacteria in up to 100 μl were injected into the left thigh (Smith and Huggins, 1982; Bull et al., 2012; Ponnusamy et al., 2016). Doses ranged from 1.2 to 3.5 × 10⁸ CFU for E. coli RS218, 1.7 – 3.7 × 10⁸ CFU for E. coli ATCC 23506, and 1.0–3.7 × 10⁸ CFU for E. coli E69. The lower end of the dose range may be near a threshold that enables viability, e.g., infection by 1.2 × 10⁸ CFU of E. coli ATCC 23506 allowed 2 of 3 control mice to survive, whereas doses above 1.7 × 10⁸ CFU were routinely inimical to survival.

Treatment was performed by injecting a depolymerase appropriate for the capsule type into the right thigh within 0.5 h after the bacterial injection, i.e., K1E, K1F, or K1H depolymerase with E. coli RS218, K5 depolymerase with E. coli ATCC 23506, K30 gp41 or K30 gp42 with E. coli E69. Different doses were obtained by dilution of the stock depolymerase into PBS to yield 100 μl for an injection. The effective doses were first determined by giving 3 mice each dose (0, 2, 5, or 20 μg). Optimal doses were then used with more mice to allow statistical analysis of therapeutic efficacy. Mice were monitored at least twice daily for 5 days, and moribund mice were euthanized. The numbers of surviving mice at Day 5 were plotted, and Fisher’s Exact Test was used to evaluate the therapeutic efficacy of a depolymerase. Using SPSS software, Kaplan–Meier survival curves (Rich et al., 2010) were plotted to show the cumulative probability of survival over the 5-day period, using the Log Rank test or generalized Wilcoxon test for statistics.

To assess potential acute toxicity from the depolymerase, 3–5 mice were injected with 100 μg of depolymerase (in 100 μl PBS) or 100 μl PBS alone in the right thigh, in the absence of bacterial infection. Mice were monitored for 5 days for survival; behavior and daily body weights were measured. Statistics of the body weight gains over 5 days were performed by mixed ANOVA with repeated measures using SPSS software.

Escherichia coli RS218, ATCC 23506 and E69 were grown overnight at 37°C in defined medium (10 g Casamino Acids, 10 g glucose, 12.5 g Na₂HPO₄⋅2H₂0, 0.9 g KCl, and 0.6 g MgSO₄⋅7H₂0 per liter). Isolation of K1, K5, or K30 type capsule used extraction with pyridine acetate as previously described (Pelkonen et al., 1988). Capsules were dissolved in sterile water and stored at 4°C. The capsule concentrations were quantified by the phenol-sulfuric acid method (Dubois et al., 1956), using glucose to generate standard curves.

Degradation of capsules was monitored by gel electrophoresis followed by alcian blue staining (Møller et al., 1993; Pan et al., 2013). 10–20 μg of capsule was mixed with serial dilutions of corresponding depolymerase and incubated at 37°C for 1 h (K1E, K1F, or K1H depolymerase with K1 capsule; K5 depolymerase with K5 capsule; K30 gp41 or K30 gp42 with K30 capsule). Reactions were loaded on 12% TBE-PAGE (Tris-Boric acid-EDTA polyacrylamide gel electrophoresis) essentially as previously described (Pelkonen et al., 1988). XC (xylene cyanol), BPB (bromophenol blue) and PR (phenol red) were used together with all blue protein standards (Bio-Rad Inc.) as tracking dyes and molecular weight markers.

Quantitative analyses of depolymerase activity were performed by incubating 30–45 μg of capsule with serial dilutions of depolymerase, or by incubating 10 μg/ml depolymerase with serial dilutions of capsule, both for 30 min at 37°C, followed by determination of reducing sugar using the modified dinitrosalicylic acid (DNSA) reagent (Miller, 1959). Glucose served as the standard. The quantity of reducing sugar released against enzyme concentration was plotted, and enzyme specific activity is expressed as nmol glucose equivalents released per min per mg protein (McCallum et al., 1989). Hanes–Woolf plots (a/v against a, where a is the capsule concentration and v is the reaction velocity) allowed the determination of kinetic parameters, assuming K1 capsule at a molecular weight of 54 kDa (Hallenbeck et al., 1987; Leggate et al., 2002).

The assay was adapted from previous work (Podschun et al., 1993; Fang et al., 2004; Mushtaq et al., 2004). Briefly, 4–6 × 10⁷ CFU/ml of log phase bacteria were incubated with or without depolymerase (100 μg/ml) for 1.5 (K1, K5) or 2 h (K30) at 37°C. Mixtures were diluted and 4–6 × 10⁴ cells were incubated with 75% of human serum (Sigma–Aldrich Inc.), heat inactivated serum (56°C, 30 min) or PBS for 1.5 (K5) or 2 h (K1, K30) at 37°C and then plated to determine CFU. Assays were repeated at least three times, and Student’s t-test with an appropriately adjusted degree of freedom was used to evaluate the enzyme’s effect in sensitizing cells to serum killing by the survival ratio X(serum)/X(PBS), where X is CFU/ml. The null hypothesis is that one treatment is the same as the other, or X1(serum)/X1(PBS) = X2(serum)/X2(PBS), i.e., log [X1(serum)] – log[X1(PBS)] – {log [X2(serum)] – log[X2(PBS)]} = 0.

Depolymerases from phages K1E, K1F, K1H, and K5 were purified from expression plasmids of previously identified genes (Petter and Vimr, 1993; Long et al., 1995; Clarke et al., 2000; Machida et al., 2000; Muhlenhoff et al., 2003; Scholl et al., 2004). K30 depolymerase was described as a trimer of a heterodimer of 90 and 52 kDa proteins (McCallum et al., 1989). Inspection of the subsequently deposited and annotated K30 genome sequence (Genbank NC_015719), which is largely syntenic with the well-characterized T7 genome, revealed only two likely candidate genes for the proteins: the K30 gene 42 product is a putative lipase/acylhydrolase, and gene 41 is a putative tailspike. Expected sizes of both proteins correspond to subunits of the depolymerase originally characterized biochemically. Both genes were therefore cloned into expression plasmids and all the His-tagged proteins were purified, yielding approximately 10 mg of K5 or K30 gp41 per liter culture, 20 – 30 mg per liter of K1E, K1F, or K1H, and 40 mg per liter of K30 gp42. We therefore expressed and tested a total of six proteins, including four depolymerases and two putative depolymerases.

The affinity-purified K1 and K5 depolymerases migrated on SDS-PAGE in accordance with their expected sizes after proteolysis (K1E 76 kDa, K1F 103 kDa, K1H 93 kDa and K5 52 kDa) (Figure 1). Sizes of the two K30 proteins were estimated to be 97 and 57 kDa, suggesting that they are not post-translationally cleaved. By densitometry analysis all purified proteins appeared >90% pure in SDS-PAGE.

Capsule depolymerases were tested in a mouse thigh model of infection. Without treatment, infection was usually lethal, whereas most mice were rescued by treatment when the enzyme dose was 20 μg per mouse, i.e., 0.8–1 mg/kg weight (Figures 2, 3). The exception was the K1E enzyme, which rescued only 3 of 32 mice at a dose of 20 μg per mouse (Figures 2A, 3A). Preliminary trials of the three K1 enzymes at lower doses suggested the effective doses of K1F and K1H were between 2 μg (both partially rescuing) and 5 μg (both rescuing 3 of 3 mice) per mouse (Figure 2A).

For K5, the effective dose was between 2 and 20 μg per mouse (Figures 2B, 3B). Of the two putative K30 depolymerases, only K30 gp41 rescued mice and then only at the higher dose tested (20 μg per mouse) (Figures 2C, 3C). A mixture of both K30 gp41 and K30 gp42 yielded the same survival outcome as K30 gp41 alone (Figure 2C), although the small sample size limits a statistical resolution. K30 gp41 appears somewhat less effective than K5 and two of the K1 enzymes.

To evaluate potential acute toxicity from enzyme injection, mice received 100 μg of depolymerase in the right thigh and were monitored for survival, behavior and body weight gains for 5 days. All the mice survived and appeared healthy without any behavior change observed. Statistics by ANOVA indicates no significant difference in body weight gains of the treated mice compared to that of the control mice receiving PBS injection (Supplementary Figure S1). These indicate no or little toxicity from enzyme injection.

The depolymerases were generally effective for the in vivo treatment, but different enzymes also showed differences in efficacy, especially K1E, which was a relatively poor therapeutic agent. K30 gp42 had no effect and likely lacks depolymerase activity. Therefore, in vitro assays were conducted to directly assess enzyme activities.

Activities of the different K1 enzymes were compared using a gel assay to monitor capsule degradation. Apparent complete degradation (by visual inspection) of 10–20 μg of K1 capsule in 1 h was achieved by 4–8 μg/ml of K1E or K1F depolymerase, and by 8–16 μg/ml of the K1H enzyme (Figure 4A). Two μg/ml of K1E or K1F enzyme completely degraded the capsule within 3 h, but K1H was again significantly less effective (data not shown). Thus, both K1E and K1F perform better than K1H during in vitro capsule degradation, in contrast to in vivo therapeutic efficiencies.

K1E depolymerase activity was quantified by assaying reducing sugar release from capsule (Figure 4D and Supplementary Table S2). Assays with 30 μg of K1 capsule and varying K1 enzyme doses showed similar kinetics of the three K1 enzymes, with similar specific activities at lower enzyme doses (K1E 80 – 240, K1F 80 – 360, K1H 70 – 260 nmol glucose equivalents released per min per mg protein) (Figure 4D). Assays with 10 μg/ml of depolymerase and varying K1 capsule concentrations showed slightly better binding affinity for K1E [K_M = 4.16 μM, similar to previous reports (Long et al., 1995; Leggate et al., 2002)] than K1F and K1H, and higher catalytic efficiency than K1H (Supplementary Table S2).

Similar assays were performed with K5 and K30. 10–20 μg of K5 capsule was degraded in 1 h by 4–8 μg/ml K5 depolymerase (Figure 4B). Of the two putative K30 depolymerases, 64 μg/ml K30 gp41 was required to degrade 20 μg K30 capsule in 1 h (Figure 4C). K30 gp42 did not detectably degrade capsule, and combining K30 gp42 with K30 gp41 in different molar ratios provided no increase in reactivity (data not shown) though the two proteins appeared to bind when mixed (Supplementary Figure S2). Quantitative assays confirmed the high activity of K5 depolymerase (260–850 nmol glucose equivalents released per min per mg protein), low activity of K30 gp41 (35–60 nmol per min per mg protein) (Figure 4D), and no activity of K30 gp42 (not shown).

Depolymerases can strip capsules and expose the underlying bacterium to immune attack such as complement-mediated killing (Roberts, 1996; Finlay and McFadden, 2006). We therefore tested our purified proteins using in vitro serum sensitivity assays. In the absence of serum, none of the depolymerases affected bacterial survival. Serum alone had a small effect in killing (Figure 5), while heat-inactivated serum slightly increased bacterial numbers (Supplementary Figure S3). For K1 bacteria, enzyme plus serum decreased bacterial survival by at least an order of magnitude (Figure 5A), with >10⁴ killing if the incubation time was extended to 3 h (not shown). As in the capsule degradation assays, K1E exhibited similar activity to K1F depolymerase, and both were superior to K1H in sensitizing bacteria to serum (Figure 5A).

K30 gp41 depolymerase, and especially K5 depolymerase, also sensitized bacteria to serum (Figures 5B,C). K30 gp42 alone had no significant effect on bacterial survival (with or without serum), and failed to provide any synergistic effect when combined with K30 gp41 (Figure 5C).

Most depolymerases are homotrimers, and one possible explanation for the discrepancy between the poor therapeutic performance of K1E in mice but good in vitro activity is that the purified proteins did not correctly trimerize. Analytical size exclusion chromatography of the depolymerases indicated that the K1E enzyme was mostly present as an ∼18-mer with only a trace of an apparent trimer (Figure 6A), while the K1F and K1H enzymes, and those from K5 and K30, were mostly trimeric (Figures 6B–F). Multimerization of K1E may limit its in vivo distribution following intramuscular injection, as further discussed below.