Escherichia coli strains (ECOR10, ECOR32, ECOR57, and ECOR63) were obtained from the E. coli reference collection (ECOR; Ochman and Selander, 1984), and were routinely cultured in lysogeny broth (LB) at 37°C with shaking at 120 RPM.

Escherichia coli phages SU10 (Podoviridae), SU32 (Siphoviridae), SU57 (Siphoviridae), and SU63 (Myoviridae) were isolated from the Käppala Wastewater treatment works, Lidingö, Stockholm county, Sweden, and have been characterized elsewhere (Khan Mirzaei et al., 2014; Khan Mirzaei and Nilsson, 2015). Phages were routinely cultured in LB with appropriate hosts (ECOR10, ECOR32, ECOR57, and ECOR63) prior to purification in PEG-8000 and CsCl gradient centrifugation as previously described (Villafane, 2009). Phages were enumerated using the agar overlay method as previously described (Khan Mirzaei and Nilsson, 2015). Extended-spectrum beta-lactamase producing E. coli strains, here denoted ESBL-32 (strain 07RAFM-ECO-32) and ESBL-198 (strain 07RAFM-ECO-198), were grown in RPMI-1640 complete medium (Invitrogen, Sweden) supplemented with 10%(v/v) heat inactivated fetal calf serum (FCS; Invitrogen), 1%(v/v), penicillin–streptomycin (PEST; Thermo Scientific, Logan, UT, USA), 2%(v/v) L-glutamine and 4%(v/v) HEPES (HyClone Laboratories, Inc, South Logan, UT, USA). ESBL-198 was used for the infection of SU32, SU57, and SU63 while ESBL-32 was used due to the limited host range of SU10.

Bacterial debris was prepared from a 1.5 L overnight culture, centrifuged at 11000 × g for 20 min and re-suspended in 20 mL of phosphate buffered saline (PBS) then sonicated with 80% power for 15 min in a VCX130 Vibra-Cell (Sonics & Materials Inc., Newtown, CT, USA) to break up bacterial cells. Bacterial debris samples were subjected to CsCl purification as previously described (Villafane, 2009). In order to confirm the efficacy of debris removal, samples were taken from the same point in the ultracentrifuge tube as phage stocks.

HT-29 (HTB-38) and Caco-2 (HTB-37) were cultured at 37°C in a 5% CO₂ atmosphere in McCoy’s 5A (modified) medium (ATCC or HyClone Laboratories Inc.) and in Dulbecco’s modified Eagle’s medium (DMEM; HyClone Laboratories, Inc.), respectively. Both media were supplemented with 10%(v/v) FCS and 1%(v/v) PEST. To maintain exponential growth phase, cells were treated with trypsin-EDTA (Invitrogen) and sub-cultured before reaching confluence (Haileselassie et al., 2013).

In order to provide stimulation, HT-29 and Caco-2 cells were trypsinated, washed in fresh medium and seeded at a final concentration of 2 × 10⁵ cells/well for cytokine analysis in 48 well plates (Costar, Cambridge, UK) and 5 × 10⁴ cells/well in flat-bottomed 96-well plates (Sarstedt Inc., Newton, NC, USA) and grown over night at 37°C in a 5% CO₂ atmosphere. Following incubation, both Caco-2 and HT-29 cells were stimulated as follows: (1) addition of phages alone (10⁹ PFU/well), (2) addition of bacteria alone (10⁵ CFU/well), (3) simultaneous addition of phages and the host bacteria (10⁹ PFU/well and 10⁵ CFU/well), and (4) complete media only. Cells were incubated for 8 h for cytokine analysis and 72 h for 3-(4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide (MTT) analysis at 37°C in 5% CO₂. Supernatants from IEC stimulations were collected and frozen at -20°C until cytokine analyses and the cells were subjected to MTT analysis.

To profile the response of stimulated HT-29 epithelial cells, 38 cytokines were simultaneously analyzed using a Human cytokine array kit (Cat no. ARY005, R&D Systems, Abingdon, UK) according to the manufacturer’s instructions. Qualitative analysis was performed by comparing the density of the detected spots to that of the reference spots included in the assay.

Post-stimulation, 100 μL/well of their respective complete medium containing 9.1% 12 mM MTT solution (Vybrant^® MTT Cell Proliferation Assay Kit (Thermo Fisher Scientific, Sweden) was added into the PBMC and IEC (Caco-2 and HT-29) culture wells and the plates were incubated at 37°C for 3 h. MTT solution containing complete media was also added in an empty well as blank control. The supernatant was then removed, and the purple formazan crystals formed were solubilized with 50 μL/well of DMSO (Sigma-Aldrich, MO, USA). The absorbance was determined using an automatic microplate spectrophotometer at 540 nm. The OD values for each stimulus were adjusted for the contribution of the blank control (Peng et al., 2007).

Purified phage preparations and bacterial debris were checked once for endotoxin content using a Pierce Limulus amebocyte lysate (LAL) Chromogenic Endotoxin Quantitation Kit (Thermo Fisher, Sweden) according to the manufacturer’s instructions.

A one way ANOVA was used for the analysis of normally distributed whole data sets, 2-tailed t-tests for pairwise comparisons within normalized data sets, Kruskal–Wallis tests for non-normally distributed whole data sets or Mann–Whitney for pairwise comparisons between non-normalized data. Statistical analysis was performed in the R Statistical software package (R Core Team, 2015). Individual statistical tests used are identified in the figure legends or text where necessary.

When incubated with epithelial cells and hosts, phages were able to significantly reduce the bacterial content over a period of 8 h (Figure 1A; P < 0.0001 by ANOVA). This reduction in bacterial content was shown to be the result of productive phage replication by an increase in phage titer over the same period (Figure 1B; P < 0.0001 by ANOVA). It should also be noted that all phages, were reduced in number following 8 h incubation with epithelial cells when compared to the initial inoculum (from 5 × 10⁵ to the lowest 3 × 10⁵; P < 0.05 by two tailed t-test). For SU32 this reduction was smaller (from 5 × 10⁵ to the lowest 4 × 10⁵; P > 0.05 by two tailed t-test). No significant impact on the efficacy and stability of phages was observed between the two cell lines.

When phages were incubated with HT-29 cells in the absence of a suitable host, cytokines IL-8, CXCL-1/GROα and Macrophage migration inhibitory factor (MIF) were induced (Table 1). In the presence of a bacterial host (ESBL-198), a more intense pattern of cytokine activation was observed for phages SU32, SU57, and SU63. An additional cytokine, IL-1ra/IL-1F3 was activated in the presence of ESBL-198 only.

Due to the host range of SU10, a second ESBL containing E. coli (ESBL-32) was chosen for comparative analysis. This combination of phage and host induced an additional pair of cytokines, IL-18/IL-1F4 and IL-1ra/IL-1F3 in the presence of HT-29 cells (Table 1). These additional cytokines were also present in the ESBL-32 control.

When assessed at the highest phage concentration (10⁹ PFU/well) all samples were found to induce IL-6, IL-10, and TNF-α (Figure 2) at varying levels depending on the phage/cytokine combination. When compared to the medium (negative) control, IL-6 was significantly induced by all phages (P < 0.002 by Mann–Whitney) with the exception of SU32. Both IL-10 and TNF-α were significantly induced by SU57 and SU63 (P < 0.05 by Mann–Whitney) but not by SU10 or SU32 when compared to the negative control. When compared to the positive (LPS) control, all phages were significantly less immunostimulatory (P < 0.05 by Mann–Whitney) with the exception of SU63. TNF-α induction levels for SU57 were insignificant when compared to LPS. When comparing the ratio of TNF-α to IL-10, phage treated samples were higher than the medium control (P > 0.05 by Mann–Whitney). At lower phage concentrations (10⁷ and 10⁵ PFU/well) no significant induction of cytokines were observed (data not shown). IFN-γ was induced by all four phages but only in PBMCs obtained from two donors. The levels of IL-2, IL-4, and IL-17 were below the limit of detection for all conditions tested. In addition, no response to the purified bacterial debris when incubated with PBMC was observed (data not shown).

In a long term assessment of the cytotoxic effects of individual phages on IECs and PBMCs, phages at a concentration of 10⁹ PFU/well were not lethal to the HT-29 and Caco-2 cells after 96 h of incubation (P < 0.05 by two tailed t-test) and no significant differences between the cell lines were observed (Figure 3). Phage stimulated IECs produced a significantly (P < 0.05 by two tailed t-test) higher OD response compared to the negative control. Phages were unable to induce a significant cytotoxic effect on PBMCs.

Phage preparations at the highest concentrations (10⁹ PFU/mL) showed marked differences in endotoxin content compared to the bacterial debris control.

In order to be a viable adjuvant or alternative to conventional antibiotic therapy, whole phage therapeutics requires the establishment of a productive infection lifecycle. Following administration, this would require the migration of phages to the site of bacterial infection and the ability to persist in the presence of, or evade, the immune system while maintaining an infectious dose (Abedon and Thomas-Abedon, 2010; Abedon, 2011; Bull and Gill, 2014). ESBL-32 was chosen as a specific host for SU10 for the current investigation in order to produce comparable levels of activity between the four phages (Khan Mirzaei and Nilsson, 2015). Despite the maintenance of high killing efficacy and reproductive ability in the presence of IECs and suitable bacterial hosts (Figure 1), the number of phages reduced with a different level in the absence of bacterial hosts compared to the initial inoculum which could be due to difference in their stability under the experimental conditions. However, this low level of clearance would not explain the disparity between the activities of different phages, i.e., approximately 2–3 log₁₀ difference between SU10 and the other phages and may result from the unique infection dynamics of each phage (Bull et al., 2014). Although not considered in this study, the in situ environmental conditions (e.g., gut pH) may potentially have a detrimental effect on the stability and efficacy of the phages.

The current study shows the varying degrees of immunogenicity exhibited by four distinct phages of the Caudovirales order when stimulating human PBMCs and IECs. IL-18 activation is of current interest as a vaccine adjuvant due to the ability to activate T-cells in the presence of different bacteria and cell types (Miettinen et al., 1998; Sugimoto et al., 2004). It is generally believed that phages are safe for human use, however, the lack of appropriate clinical trials and a lack of regulatory approval have restricted routine clinical use to specialist centers around the world (Skurnik and Strauch, 2006; Sarker et al., 2012). However, the current investigation suggests that both SU57 and SU63 at high concentrations (10⁹ PFU/well) can significantly induce IL-6 and TNF-α from PBMCs compared to the negative (medium) control. SU57 and SU63 were also able to significantly induce the anti-inflammatory cytokine IL-10 compared to the negative (medium) control. In addition to this, SU10 also activates IL-6 release with a significant level from PBMCs compared to the negative (medium) control. Despite a lack of toxicity in both HT-29 and Caco-2 cells, phage exposure inhibited cell proliferation via an unknown mechanism although there is some suggestion that the interaction between phages and epithelial cells may result in the rapid release of reactive oxygen species (Gorski et al., 2012).

In order to satisfy regulatory requirements for non-topical clinical application, particularly intravenous injection, it is necessary to quantify endotoxin levels in the final product (Merabishvili et al., 2009). This is routinely accomplished with the “gold standard” method of an LAL assay. However, a number of studies suggest that the use of LAL assays results in ambiguous results when applied on phages (Merabishvili et al., 2009; Cooper et al., 2014) and was again highlighted in the current study. This discrepancy between LAL results and immune response studies may be the result of phages/LAL cross reactivity. Although supplemental methods, including in vivo assessments exist, these methods may not provide the required level of sensitivity necessary for pharmaceutical quality control and also raise ethical concerns over routine animal usage. An ELISA assay to profile immune responses to formulations may provide a sensitive alternative to LAL assessment and address ethical concerns over routine animal usage as has been suggested by Hodyra-Stefaniak et al. (2015) and should be the focus of future work.