All experimental protocols and procedures of the study were reviewed and approved by the Institutional Review Committee on Animals Ethics of the Institute of Primate Research (ref no: IRC/02/14) ethical committee on the use of laboratory animals for biomedical science in accordance with the international guidelines on animal care, handling and use for biomedical research. The experiments are reported in accordance with ARRIVE guidelines.¹⁹

The study included 30 BALB/c mice of mixed sexes aged 6 to 8 weeks which were sourced from the Institute of Primate Research rodent house, Karen-Nairobi, Kenya, and maintained at their rodent facility. The mice were fed on antibiotic-free food rations (obtained from Unga Feeds, Nairobi, Kenya) and water was provided ad libitum.

S. aureus bacterial strains were isolated from 20 sewage samples collected from informal settlements, sewage treatment plants and abattoirs within Nairobi County, Kenya. Specimens from all the samples were streaked on selective Mannitol salt agar (Liofilchem^®, Roseto degli Abruzzi, Italy) supplemented with 4 µg of ciprofloxacin [Liofilchem^®, Roseto degli Abruzzi, Italy]. Single colonies were amplified in nutrient agar (HiMedia, Mumbai, India). Staphylococcus strains were identified using microscopy, physiological tests and an Analytical Profile Index of Staphylococcus system [bioMérieux, Marcy l’Etoile, France]. We screened isolated S. aureus bacteria for antimicrobial resistance to the following drugs: ceftazidime 30 µg, oxacillin 1 µg, vancomycin 30 µg, netilmicin 30 µg, gentamicin 10 µg, erythromycin 15 µg, trimethroprim-sulfamethoxazole 25 µg and cefuroxime 30 µg (Liofilchem^®, Roseto degli Abruzzi, Italy), according to the Clinical Laboratory Standard Institute protocol.²⁰ Isolates were considered to be MDRSA, if they were non-susceptible to more than one class of antibiotics. Colonies identified as MDRSA were re-suspended in 50% glycerol-nutrient broth and stored at -20 ^°C until use.

We analysed half a litre of sewage samples collected from 10 sites (Nairobi City sewage treatment plant, informal settlements and an abattoir) within Nairobi County, Kenya. Twenty specific S. aureus lytic phages were isolated from these samples using the modified method of Anany et al.²¹ Briefly, equal volumes of ultra-filtered sewage samples and nutrient broth, plus 1 mL of 18-hour-old MDRSA culture were mixed and incubated overnight at 37 ^°C while shaking at 120 rpm (LAB-LINE^® INCUBATOR-SHAKER, Waltham, Massachusetts, United States). The following day, the culture was centrifuged at 10 000 g for 10 minutes (Fisher Centrific^® Centrifuge, Waltham, Massachusetts, United States), and the supernatant was filtered through a 0.22 µm filtration unit (Cambridge, Massachusetts, United States), then screened for the presence of phages using a double-layer plaque assay. Individual plaques were selected and sub-cultured in 2 mL of nutrient broth containing a sensitive bacterial host (10⁶ CFU/mL). The isolated phages were then tested against the previously-isolated MDRSA by spot assay. The virulence of the phages was determined by using an efficiency of plating analysis,²² whereby the strain with numerous plaques on MDRSA lawn was considered the most virulent. The virulent phage was selected and used as a therapeutic agent against MDRSA in mice.

Thirty mice were randomly assigned to one of three groups: the MDRSA infection group (n = 20), the non-infection group (n = 5) and the phage infection group (n = 5). Twenty mice were infected with 10⁸ CFU/mL of MDRSA isolate intravenously by means of the tail vein and then sub-divided into four sub-groups: MDRSA with clindamycin (Clindar, Indi Pharma, Mumbai, India) (8 mg/kg body weight) treatment (n = 5); MDRSA with phage (10⁸ PFU/mL) treatment (n = 5); MDRSA with combination treatment (clindamycin [8 mg/kg body weight] and phage [10⁸ PFU/mL]) (n = 5); and MDRSA with no treatment as the control (n = 5) (Figure 1). Presence of infection was determined by observing the physical appearance of the mice and assessing the bacteraemia level. The bacteraemia count was assessed by culturing blood samples from each mouse by using the pour plate method. Fifty microlitres (50 µL) of blood was obtained from each mouse from the tail vein and mixed with 950 µL sterile normal saline. This mixture (blood-normal saline) was added to molten nutrient broth, allowed to cool, then incubated overnight at 37 ^°C. The bacterial colony count was recorded per millilitre for each mouse. A similar procedure was used to determine the bacteraemia for each mouse after treatment was initiated. Treatment with clindamycin, phage or combination was administered at 72 hours post-MDRSA infection when the bacterial levels in the blood began to rise. Each mouse in the treatment groups received a single dose of the mentioned therapeutic agents. The experiments were repeated three times; thus, the total number of mice used in the study was 90.

Upon necropsy, lungs were obtained aseptically and homogenised for bacterial culture as follows: 500 µL of homogenised tissue was diluted with normal saline at a ratio of 1:20 (homogenised lung tissue to saline). The diluted, homogenised tissue was then plated on 7.5% sodium chloride nutrient agar to select for MDRSA and incubated at 37 ^°C for 18–20 hours.

Phage safety and MDRSA pathogenicity were determined by euthanising the mice when they exhibited poor physical appearance and breathing difficulties in order to alleviate pain and suffering. Surviving mice were euthanised at 10 days post-infection. Lung samples were collected in 10% formaldehyde for histological analysis. Levels of inflammation were scored using categories as described by Schünemann et al.²³ Briefly, the categories worst, worse, bad, good, better and best were assigned based on a scale of 5 to 0. The ‘worst’ category was assigned for severely inflammed septa, numerous collapsed alveoli, large pockets of pneumonia, mucus-congested alveoli, presence of perivascular fibrosis blood vessels and lack of ventilation (score = 5). The ‘worse’ category included all conditions in the worst category but with poor ventilation (score = 4). The ‘bad’ category was assigned for minor inflammation of septa, few pockets of pneumonia and moderate ventilation (score = 3). The ‘good’ category was assigned for minor inflammation of septa, few pockets of mucus-congested alveoli and good ventilation (score = 2). The ‘better’ category was assigned for no inflammed septa and a few pockets of mucus-congested alveoli (score = 1). The ‘best’/normal category was assigned for no inflammation and good ventilation (score = 0).

Bacterial and phage counts were represented as mean ± standard error of the mean. The statistical significance of differences between groups was determined by one-way analysis of variance, followed by Tukey’s multiple comparison test, using Graph Pad Prism 5.0.1 (Graph pad software, San Diego, California, United States). A p-value of less than 0.05 was considered statistically significant.

The environmentally-isolated S. aureus bacterium was resistant to β-lactam antibiotics (ceftazidime, cefuroxime and oxacillin). In addition, it was non-susceptible to glycopeptide (vancomycin), aminoglycoside (netilmicin and gentamicin) and macrolide (erythromycin) antibiotics. However, it was susceptible to nucleic acid synthesis inhibitors (co-trimoxazole [trimethroprim-sulfamethoxazole]) (Figure 2). Thus, the bacterial isolate was considered to be multidrug-resistant, as it was non-susceptible to more than one class of antibiotics.

Spot assays were used to identify 10 S. aureus lytic phages which produced large clear plaques on the MDRSA lawn (Figure 3) while other isolates failed. Of the 10 phage isolates, one created a larger patch on the MDRSA lawn. Its efficiency of plating analysis showed that it had more plaques than other phages.

Two groups of mice had 100% survivorship (n = 5), namely, the non-infected/non-treated group and the phage-infected mice. Only three mice survived in the MDRSA-infected group before treatment (72 hours post-infection) commenced. All the mice that were treated with either clindamycin, phage or the combination therapy (clindamycin plus phage) survived at day 7 post-infection and each group had three mice. Mice that had phage infection (phage control group) were more active than mice in other treatment groups (clindamycin, phage and combination). Only one mouse in the MDRSA-infected control group survived at day 7 post-infection and it was weak (Table 1).

The number of viable MDRSA in mouse lung homogenates was significantly lower in mice treated with phage only (log₁₀ CFU/gm = 0.5 ± 0.2) compared with those in the MDRSA-infected, non-treated control group (log₁₀ CFU/gm = 8.0 ± 0.2) (p < 0.0001) (Figure 4). Compared with the MDRSA-infected, non-treated group, statistically-significantly differences were observed in the viable bacterial counts in the MDRSA-infected, clindamycin-treated (log₁₀ CFU/gm = 4.4 ± 0.2; p < 0.05) and MDRSA-infected, combination-treated groups (log₁₀ CFU/gm = 4.0 ± 0.2; p < 0.001). The differences in counts between the phage-treated group and groups treated with clindamycin and the combination therapy were also statistically significant (p < 0.0001).

The lung tissues of mice in the phage-infected group were relatively normal, with minor focal congestion compared with tissue from mice in the MDRSA-infected non-treated group, which had deflated alveoli congested with mucus, lymphocyte-infiltrated septa and pockets of serous fluid. The lung tissue of MDRSA-infected, non-treated mice were similar to lung tissue from clindamycin-treated mice. However, lung tissue from the clindamycin-treated group (n = 3) overall was graded as bad, since they had less congestion, fewer pockets of serous fluid and deflated alveoli compared with the non-treated MDRSA infected mice. Mice that received the combination therapy (n = 3) showed inflammation of the septa and blood vessel walls, moderately deflated alveoli and several serous fluid pockets compared with the non-infected (n = 5) and phage-infected mice (n = 5) (p < 0.0001) (Figure 5).

Our study results were generated in mice. While our findings provide vital information on the antimicrobial resistance profile in Kenya, they cannot be translated directly to humans. However, comparative animal models can be utilised to ascertain the safety and therapeutic potential of environmentally-available phages against infections caused by multidrug-resistant bacteria.

In conclusion, we report for the first time that a single dose of MDRSA-specific lytic phage is efficacious against haematogenous pneumonia caused by multidrug-resistant S. aureus in mice.