Efficacy and safety assessment of two enterococci phages in an in vitro biofilm wound model

Chronic wounds affect thousands of people worldwide, causing pain and discomfort to patients and represent significant economical burdens to health care systems. The treatment of chronic wounds is very difficult and complex, particularly when wounds are colonized by bacterial biofilms which are highly tolerant to antibiotics. Enterococcus faecium and Enterococcus faecalis are within the most frequent bacteria present in chronic wounds. Bacteriophages (phages) have been proposed as an efficient and alternative against antibiotic-resistant infections, as those found in chronic wounds. We have isolated and characterized two novel enterococci phages, the siphovirus vB_EfaS-Zip (Zip) and the podovirus vB_EfaP-Max (Max) to be applied during wound treatment. Both phages demonstrated lytic behavior against E. faecalis and E. faecium. Genome analysis of both phages suggests the absence of genes associated with lysogeny. A phage cocktail containing both phages was tested against biofilms formed in wound simulated conditions at a multiplicity of infection of 1.0 and a 2.5 log CFU.mL−1 reduction in the bacterial load after at 3 h of treatment was observed. Phages were also tested in epithelial cells colonized by these bacterial species and a 3 log CFU.mL−1 reduction was observed using both phages. The high efficacy of these new isolated phages against multi-species biofilms, their stability at different temperatures and pH ranges, short latent periods and non-cytotoxicity to epithelial cells suggest their therapeutic use to control infectious biofilms present in chronic wounds.


Results and Discussion
Max and Zip are two newly isolated enterococci phages. Phages are ubiquitous in nature and they can be found wherever their host is present, therefore it is expected to isolate them in almost every environment 19 . In this work, phage vB_EfaS-Zip (Zip) infecting E. faecium and vB_EfaP-Max (Max) infecting E. faecalis were isolated using sewage water from wastewater treatments plants ETAR Braga (Frossos).
To better characterize the phages, their morphologies were observed by TEM and both revealed to belong to the Caudovirales order. According to the morphological evaluation 20 , E. faecalis phage Max belongs to the Siphoviridae family, having a non-contractile tail with 220 nm in length and 12 nm in width and a capsid with 58 nm in diameter (Fig. 1A). E. faecium phage Zip has a capsid with 46 nm in diameter and a short, non-contractile tail with 19 nm in length, and consequently belongs to the Podoviridae family (Fig. 1B).
To further characterize both phages' infection cycles, one-step growth curves were performed. Phage Max showed a latent period of 10 min with an average burst size of 38 PFU per infected cell (Fig. 1C). Although the burst size is smaller than other siphoviruses such as IME-EF1 (60 PFU per infected bacteria), the latent period of Max is shorter than the referred phage 21 . Phage Zip has a latent period of 25 min with average burst size of 52 PFU per infected cell (Fig. 1D). A morphological similar E. faecium phage, vB_EfaP_IME199, had a similar latent period, but a smaller burst size 22 .
It has been described that phages with short latent periods have more efficient replication cycles which are related with the success of in vitro studies 23 . Podoviruses are usually associated with short latent periods are higher burst sizes than more complex phages, such as myoviruses 23 .
Enterococci phages have a wide host lytic range among clinical isolates. The lytic spectra of both isolated phages was tested against a panel of enterococci strains to study their host range (Table 1). Even with different bacterial species as hosts, the host range of both phages is quite similar. E. faecalis phage Max can lyse 12 out of the 16 E. faecalis (75%) strains tested, and also 2 of the 13 E. faecium strains tested (15%). E. faecium phage Zip has a narrow host range in E. faecium lysing 23% (3 out of 13) and 69% of the E. faecalis strains tested (11 out of 16). Moreover, both phages did not show any lytic effect on the other Gram-positive species tested. Max and Zip are stable in wounds pH and temperatures. Physico-chemical factors, such as temperature and pH have been described to influence phage survival and persistence 25 . To assess the potential of both enterococci phages for therapy, stability tests were made exposing phages to different temperatures and pHs for 24 h to simulate putative product development steps and storage conditions. Temperature plays a fundamental role in phage attachment, genetic material ejection and phage multiplication 26 . A thermal stability test was carried out to determine the heat resistance of isolated phages at pH 7.0. Both phages were stable after 24 h, at −20 °C, 4 °C (control), 21 °C and 37 °C, showing a concentration of about 7 log CFU.mL −1 ( Fig. 2A,B). At 50 °C, both phages lost their titer in about 1 log CFU.mL −1 (p < 0.05). At 60 °C E. faecalis phage Max concentration decreased about 2 log CFU.mL −1 (p < 0.05) ( Fig. 2A), while phage Zip was completely inactivated at (Fig. 2B). The results are in agreement with those reported by Lee et al. 27 , where two different E. faecalis phages of the Siphoviridae family showed high tolerance to temperatures below 60 °C. In general, members of Siphoviridae family are considered to be stable at large temperature fluctuations 28  www.nature.com/scientificreports www.nature.com/scientificreports/ The acidity and alkalinity of the environment are also important factors influencing phage stability. Optimal pH was determined by testing the stability of phages at different pH values after 24 h of incubation at room temperature (21 °C) (Fig. 2C,D). Both phages were very stable in the pH range 5.0-11.0 and were completely inactivated at extreme pH values of 1.0, 2.0 and 13.0. For E. faecalis phage Max, the PFU counts decreased by 6 and 2 log CFU.mL −1 at pH 3.0 and 4.0, respectively (p < 0.05) (Fig. 2C). Regarding E. faecium phage Zip, viable phage counts were reduced in 1 log CFU.mL −1 at pH 4.0 and 12.0 and no phages were detected at pH 3.0 (p < 0.05) (Fig. 2D). Similarly to temperature stability results, Siphoviridae phages are described as the most resistant to adverse pH conditions 28 . The siphovirus phage Max showed some activity at pH 3.0, unlike the E. faecium podovirus Zip, inactivated at that pH value. The loss of titer at this acidic pH was already reported in the C. freundii phage LK1 29 .
Both phages studied in this work may be promising in combating biofilms in wounds, as they present stability in the pH range from 5-11, usually found in wounds environment. A pH value of about 4.7 is characteristic of the skin surface of adults and healthy children 30 . However, a study that analyzed 26 cases of second degree burns reported that infected wounds have an increased pH ranging from 6.5 to 9.0 31 . Moreover, concerning temperature, the skin surface temperature is about 33.2 °C 32 , but when there is an infection, the temperature in wounds can increase due to immune response, inflammatory cytokine-induced vasodilation and increased tissue metabolism 33 . Dini et al. analyzed the skin temperature in 18 patients affected by venous insufficiency and lower leg ulcers and described that wound injury temperature range was between 31 °C and 35 °C, and the perilesional skin temperature range was between 31 °C and 34 °C 34 .
Enterococcus phage genomes are free of lysogenic genes. Enterococcus sp. phage genomes were sequenced through a MiSeq illumina platform and de novo assembled with average coverage of 324x and 230x for Max and Zip, respectively.
Max genome is a linear dsDNA molecule of 40,975 bp with a GC content of 34.7%. It encodes 65 predicted proteins in both strands being most homologs of Max found in Enterococcus phage genomes ( Fig. 3A and Table S1). Nevertheless, only 26 proteins have assigned function to DNA replication (e.g. DNA polymerase, helicase), morphogenesis (e.g. major capsid protein, portal protein) and cell lysis (holin and endolysin). Max genome is highly syntenic with Enterococcus phage phiSHEF2 genome (MF678788) sharing 92% nucleotide identity and 58 of its genes. As reported for phage the similar myovirus IME-EFm1, it is expected that Max has a terminally non-redundant genome 35 . Regarding regulatory elements, we found six bacterial promoters and ten rho-independent terminators.
Zip genome has a length of 18,742 bp and GC content of 35.0%. It encodes 22 genes, of which only nine are annotated with function with relatively high identity (>84% average amino acid identity). No tRNAs or www.nature.com/scientificreports www.nature.com/scientificreports/ regulatory elements (promoters and terminators) were found. Annotated genes fall in three distinct functional modules: DNA packaging and structural proteins; DNA replication, recombination and modification; and cell lysis (Fig. 3B, Table S2). Comparative genomics show that Zip shares 77.4% overall nucleotide identity and all 22 genes with the Enterococcus phage vB_EfaP_IME199 (KT945995) ( Table S2). Due to the similarities between these two genomes, it is expected that Zip has short inverted terminal repeats 22 .
Overall, the characteristics of both enterococci genomes suggest that they are safe for use in further in vitro studies 36 . Antibacterial assays reveal that Max and Zip are active against enterococci biofilms. Although the microtiter well plate model is a common method to study biofilms, mainly due to its simplicity, low price and the fact that it allows multi-parameter analysis, it has the problem of not reflecting the environmental conditions present in the wound bed 37 .
In 2010, Werthén et al. developed an in vitro model, collagen wound model (CWM), to simulate wound injury in which bacterial cells do not attach to well-defined solid surfaces, but to a collagen matrix 38 . In the referred study, it was shown that the biofilms formed in this model are structurally similar to biofilms observed in in vivo conditions. For this reason, CWM was already used for the study of antimicrobial activity of antibiotics and silver containing wound-dressings 39,40 .
Herein, the efficacy of the selected phages against biofilms was tested in this model. After three hours of infection, E. faecalis phage Max reduced the number of viable cells in about 2 log CFU.mL −1 (p < 0.05) (Fig. 4A). However, at 6 h of infection, the reduction decreased to 1 log CFU.mL −1 . Concerning E. faecium phage Zip, it caused a reduction of approximately 1.5 log CFU.mL −1 between 3 and 6 h of infection (Fig. 4B). Curiously, after  www.nature.com/scientificreports www.nature.com/scientificreports/ 8 h the reduction increased to about 2 log CFU.mL −1 (p < 0.05). Although depolymerases have been described to be helpful on biofilm matrix degradation, it is currently accepted that the main function of these enzymes is the polysaccharides capsular degradation 41 . As enterococci phages Max and Zip, other phages that do not encode depolymerases have been suggested as good biofilm control agents 16,42 . In both cases at 24 h it was observed a similar number of biofilm cells between control and phage-treated biofilms. One possible explanation for these results might be related with the emergence of phage resistance. Phage resistance was already shown to occur in other biofilm/phage interaction studies, where the proliferation of phage-resistant variants was observed for the time-points after 6 h of single-phage treatment 43 . In the same study phages revealed to be the less effective 12 h after biofilm infection, and at 48 h post-infection, no statistical differences in the number of biofilm cells between control and phage-treated biofilms, were observed. Recently, it was suggested that the combined use of phages with mechanical débridement, or other antimicrobial agents, such as antibiotics might be very helpful clinically to overcome the development of phage ressitance 15 .
Herein, both phages presented the ability to infect and kill biofilm cells in the presence of medium simulating wound injuries. However, it should be noted that although better than the microtiter plate model, CWM also possesses limitations as it is a closed system, without the presence of host immune cells and physiological factors (e.g. hypoxia, presence/absence of exudate and drainage of the wound). Nevertheless, this is a model that better mimics wounds, and our results suggest that phages Max and Zip may be promising in combating wound biofilms in vivo.
The phage cocktail is active against dual-species biofilms in wound simulated conditions. The majority of the studies about chronic wounds detected more than one bacterial species on the injuries 8 . Consequently, to better simulate real biofilms, phage efficiency in mixed-Enterococcus strains biofilms was analyzed. The E. faecalis Efa1/E. faecium C410 consortia was studied. Competition between both species was observed as an adverse effect was detected on the E. faecium population. Mixed biofilms were composed by E. faecalis cells in about 8 log CFU.mL −1 , a value significantly greater in 3 log CFU.mL −1 than the number of E. faecium cells. The results show that the total cells of a multi-species biofilm are not necessarily the sum of the cells of each single species. Although several researchers reported that some specific strains of enterococci can inhibit the growth of other pathogenic bacteria such as Listeria monocytogenes, Clostridium tyrobutyricum and S. aureus, due to the production of bacteriocins, no study has described the interactions between two Enterococcus specieis 44,45 . Other factors, namely competition for nutrients and the production and accumulation of toxic metabolites during biofilm formation might have contributed to these results.
An advantage of the use of phages is that they can be mixed as cocktails to broaden collectively their antibacterial spectrum of activity 46 . In this work, a phage cocktail comprising Enterococcus sp phages Max and Zip was used to infect dual-species biofilms for a consortium composed by their host strains. It was decided not to apply the cocktail on single-species biofilms, due to the relevance of mixed-species biofilms. A statistically significant reduction (p < 0.05) was observed on cell concentration at 3, 6 and 8 h on treated biofilms with phage cocktail comparing to the controls (Fig. 5A). The phage cocktail was particularly efficient in reducing biofilms, where the cell concentration was reduced by approximately 2.5 log CFU.mL −1 after 3 h of infection (p < 0.05). The reduction remained after 6 and 8 h, although with less efficiency. As in mono-species biofilms, there is the possibility of biofilm cells might acquire resistance to phages after prolonged treatments, and consequently, after 24 h of phage infection the reduction was only of 1 log CFU.mL −1 (p < 0.05) (Fig. 5A). Besides the combined use of antibiotics and mechanical débridement, other possibility to avoid phage resistance relies on the use of phage cocktails comprising different phages targeting the same host, but with different receptors.
Analysis of the individual behavior of each strain shows greater reductions in the dominant strain (E. faecalis Efa1, Fig. 5B) than in E. faecium (Fig. 5C). Results suggest that the higher multiplicity of infection (MOI) applied in E. faecium had no positive influence on phage infection outcome. Similar observations occurred in E. coli, where phage JS09 was applied at different MOI and no differences were observed on phage efficiencies 47 .
The phage cocktail was more efficient in reducing biofilm cells in the CWM than any single phage administration in mono-species biofilms. Similar results were observed in an in vivo wound model where the therapeutic efficacy of Klebsiella pneumoniae phages was evaluated in comparison to the phage cocktail in the treatment of wound infection. Mice receiving phage cocktail treatment showed a decrease of 6 log CFU.mL −1 in bacterial load as compared to the control, while for single phage treatment a decrease of about 4.5 log CFU.mL −1 was observed 48 . Max and Zip are not cytotoxic to epithelial cells. Cytotoxicity assays were performed using E. faecalis phage Max and E. faecium phage Zip. These phages were tested in two different concentrations (10 8 and 10 7 PFU. mL −1 ) and in two different conditions -lysate and after PEG purification. Several assays are available to measure cell cytotoxicity, but no standard assay is described for determining the cytotoxic effect of phages 49 . In general, both phages were nontoxic to 3T3 cells in all conditions tested (viability of cells was about 100%) (Fig. 6A,B). Nevertheless, the lysate of phage Zip at 10 8 PFU.mL −1 showed some toxicity, causing a loss of about 30% of cells (Fig. 6A). It is possible that some components of the lysate, originating from the bacteria, have some cytotoxic effect on the cells, which disappears after PEG purification, demonstrating the phage itself is not toxic.
Other studies on phage cytotoxicity in mammalian cells have shown that phages have no effect on cell viability. For example, Merabishvili et al. concluded that a phage cocktail active against P. aeruginosa and S. aureus strains was not cytotoxic against human neonatal foreskin keratinocytes 50 . Although several in vivo studies reported the safety of phages, similar studies should be performed for every new isolated phage that could be used for therapeutic purposes.
www.nature.com/scientificreports www.nature.com/scientificreports/ Max and Zip are active against bacteria colonizing 3T3 cells. We also studied the behavior of phages against bacterial infections in animal cells by investigating the mechanisms by which bacteria and phages interact with these cells.
3T3 cells were infected with 10 8 CFU.mL −1 of each bacterium (E. faecalis Efa1 and E. faecium C410) and treated at 2 h post-infection with 10 7 PFU.mL −1 of the respective phage. The concentration of viable bacterial cells and the number of 3T3 cells were quantified at 6 and 24 h post-treatment (Fig. 7). Concerning the reduction of infection, both tested phages caused a decrease in viable bacterial cells (p < 0.05) after 6 h of treatment (Fig. 7A).
The reduction was of about 3 log CFU.mL −1 , which was higher than the reduction previously obtained on biofilms. In 3T3 cells, 6 h after the application of E. faecium phage Zip, it was possible to observe a greater concentration (1 log CFU.mL −1 , p < 0.05) of viable 3T3 cells than in the untreated control cells (Fig. 7B). In the other cases at 6 h of phage treatment, the concentration of 3T3 cells was about 5.5 log CFU.mL −1 , very similar between treated and untreated cells. Twenty-four hours after phage treatment, it was possible to observe that only phage-treated epithelial cells were still viable (Fig. 7B). Curiously, at this time point, phage efficacy in the reduction of bacterial cells was very reduced (between 0.2 and 1 log CFU.mL −1 , Fig. 7A), which demonstrates the long-lasting beneficial effect of phage treatment on mammalian cell viability. A possible explanation for mammalian cell survival might be related with the reduced virulence of phage resistant phenotypes. It has been suggested that the selective  www.nature.com/scientificreports www.nature.com/scientificreports/ pressure exerted by phages can lead to the proliferation of less virulent strains 51 . Moreover, in real infective conditions, the immune system can help phages to eliminate the surviving bacterial cells 52 .
The efficacy of phages against bacteria adhered to mammalian cells was observed by Mirzaei et al., when they studied the interaction between immortalized cell lines (HT-29 and Caco-2 intestinal epithelial cells) and four E. coli phages. In that study, phages showed a significant reduction between 1.5 and 4.5 log CFU.mL −1 of the bacterial content over a period of 8 h 53 . Although the data about the effects of direct addition of phages against bacteria in immortalized cell lines is scarce, phages efficacy on in vivo models suggests that this approach might be a valuable therapy.
This work showed that the two phages can infect bacteria that colonized epithelial cells, suggesting its therapeutic interest.

Conclusions
The properties demonstrated by the novel isolated phages, namely broad lytic spectra, stability at different temperatures and pH ranges, lack of lysogenic content on their genomes, efficacy against biofilms in simulated wound conditions and against colonized epithelial cells suggest their therapeutic use. Further studies should be performed to confirm the in vivo efficacy of single or combined application of the phages.

Methods
Bacterial strains and culture conditions. A total of 16 E. faecalis and 13 E. faecium clinical isolates from Hospital de Braga and collection strains were used for phages isolation (Table 1). To complete the analysis of the phages lytic spectra, one E. gallinarum strain and additional Gram-positive species from our collection were used, including three staphylococci strains and one representative of Listeria monocytogenes (Table 1). All strains were grown in Tryptic Soy Broth (TSB), Tryptic Soy Agar (TSA) or in TSA soft overlays (TSB with 0.6% agar) at 37 °C.
Phage isolation and production. Phage isolation was performed as previously described 54 . Wastewaters from ETAR Frossos, Braga, were centrifuged (8 500 × g, 4 °C, and 10 min) and the supernatant collected. Then, 50 mL of sewage supernatant were mixed with the same volume of double-strength TSB and 100 μL of each bacterial suspension grown overnight. The mixtures were further incubated for 24 h, 120 rpm, at 37 °C. The culture was centrifuged (10 000 × g, 4 °C, 10 min) and the supernatant filtered (0.22 μm). Spot assays were performed on bacterial lawns to check the presence of phages. Inhibition zones (clear or turbid), indicative of the presence of phages were further purified with toothpicks and paper to isolate all different phages. Plaque picking was repeated until single-plaque morphology was observed.
Each isolated phage was produced as previously described 55 . Briefly, a phage plaque of each phage was picked with a toothpick and stung in agar plates containing a bacterial lawn. A strip of paper was passed through the entire agar plates, and the plates were incubated overnight at 37 °C.
Phage particles were further purified as previously described by Sambrook et al. 56 . Sodium chloride (0.584 g/10 mL) was added to the phage lysate and the solution was incubated at 4 °C, 50-90 rpm, for 1 h. The solution was centrifuged (8 500 × g, 4 °C, and 10 min) and the supernatant was recovered. Then PEG 8000 (1 g/10 mL) was added and gently mixed until dissolved, and further incubated overnight at 4 °C, 50-90 rpm. The solution was then centrifuged (8 500 × g, 4 °C, and 10 min), and the pellet was recovered and eluted in SM buffer at 4 °C, 50-90 rpm, for 1 h. 25% (v/v) chloroform was added and mixed, and the solution centrifuged (3 500 × g, 4 °C, and 15 min). The top aqueous layer containing the purified phages was recovered, filtered (0.22 μm), and stored at 4 °C for later use.
Phage titration was performed according to the double agar overlay technique 57 . Briefly, 100 μL of serially diluted phage solution, 100 μL of host bacteria culture and 3 mL of soft agar were mixed and poured onto a TSA plate. After overnight incubation at 37 °C, the plaque forming units (PFUs) were determined. www.nature.com/scientificreports www.nature.com/scientificreports/ Electron microscopy. The morphology of phage particles was observed by TEM, as previously described 58 .
Briefly, a solution of phage lysate was centrifuged (1 h, 25 000 × g, 4 °C) and phage particles were collected. The pellet was washed twice in tap water using the same centrifugation conditions. Phages were further deposited on copper grids with carbon-coated Formvar films and stained with 2% uranyl acetate (pH 4.0). Phages were examined using a Jeol JEM 1400 transmission electron microscope.
Determination of the host range. The host range of phages Max and Zip was determined as previously described with some modifications 59 . A total of 33 strains were selected for determination of host range (Table 1). Bacterial lawns were made on TSA plates by adding 100 μL of exponential-phase cell cultures of each strain. Then, 10 μL of each phage solution at approximately 10 7 PFU.mL −1 were applied to the plate and incubated at 37 °C for 16-18 h to observe whether there were transparent regions on the plate. Thermal and pH stability. Thermal stability tests were carried essentially as previously described 60 . A phage titer of 10 8 PFU.mL −1 of each tested phage was incubated at −20 °C, 4 °C (as control), 21 °C (room temperature), 37 °C, 50 °C and 60 °C for 24 h. Similarly, the effect of pH was also evaluated using a universal pH buffer (150 mM potassium chloride, 10 mM potassium dihydrogen phosphate, 10 mM sodium citrate, 10 mM boric acid with pH adjusted to 1, 2, 3, 4, 5, 6, 7 (as control), 8, 9, 10, 11, 12 and 13, at  DNA isolation, genome sequencing and in silico analysis. Phage genomic DNA was isolated using the phenol-chloroform-isoamyl alcohol method as described elsewhere 58 . The DNA samples were used for library construction using the Illumina Nextera XT library preparation kit. The generated DNA fragments (DNA libraries) were sequenced in the lllumina MiSeq platform, using 250 bp paired-end sequencing reads. The raw sequence data underwent automatic initial treatment, namely adapters and low quality bases trimming. Demultiplexed reads were de novo assembled into a single contig using Geneious R9. The assembled genomes were scanned through MyRAST 61 , tRNAscan-SE 62 and BLASTP 63 to search for coding regions, tRNAs and find protein homologs, respectively. Proteins sequenced were analyzed with TMHMM 64 and SignalP 65 to predict transmembrane domains and signal peptide cleavage sites. Comparative genomics was performed with BLASTN or OrthoVenn 66 . Genomes were visualized with Easyfig. 67 . The complete genome sequence of phages Max and Zip have been deposited in the NCBI database under accession no. MK360024 and MK360025, respectively.
Biofilm formation using the collagen wound model. The CWM for biofilm formation was used as described by Werthén et al. 38 . Two mL of collagen from bovine Achilles tendon (50 μg.mL −1 suspended in 0.9% NaCl), were added to each well of 24-well plates (Orange Scientific) and the plates were incubated overnight at 4 °C. The coating solution was then gently removed and the wells were washed twice with 0.9% NaCl prior to biofilm formation.
Single-species biofilms were formed by mixing of bacterial suspensions (grown overnight at 37 °C and 120 rpm) and a mixture of TSB/SWF (1:1) culture medium (SWF: 50% Fetal Bovine Serum (FBS) and 50% physiological NaCl in 0.1% Pepton) at a approximate concentration of 2 × 10 7 CFU.mL −1 . For dual-species biofilms, 5 μL of each bacterial culture were added to each well. Plates were further incubated at 37 °C and 120 rpm for 48 h, with complete culture media renewal after 24 h. In negative control wells, only TSB/SWF (1:1) medium was added. After biofilm formation, wells were washed twice with 0.9% NaCl, to remove planktonic cells. Three experiments were performed in duplicate.

Biofilm infection.
After single-species biofilm formation, a phage solution with 2 × 10 8 PFU.mL −1 was added to each well. On dual-species biofilms a cocktail containing 1 × 10 8 PFU.mL −1 of each phage was added to each well. Then, the plates were incubated at 37 °C and samples were taken at 3, 6, 8 and 24 h to determine the number of cultivable bacteria. At each time point, biofilms formed in each well, were disrupted using an ultrasonic bath operating at 50 kHz for 30 min, and then the suspensions were collected and viable cells were determined by CFU counting. For counting of specific strains, CFUs were determined under selective conditions, by diluting the bacteria in the presence of the phage active against the opponent species. Three experiments were performed in duplicate.
The cytotoxicity of the two phages was determined according to the ISO 10993-5:2009, Annex A -Neutral Red Uptake (NRU) cytotoxicity test. Neutral red (NR) is a weak cationic dye used to identify viable cells in culture. This assay consists on the uptake of NR into lysosomes and its subsequent accumulation on living cells, indicating quantitative viability. If cells are damaged or dead, NR is no longer retained within the vacuole of cells. Thus, the NR viability assay quantifies the number of viable, uninjured cells after their exposure with test agents, through reduction in the absorbance 68 . Briefly, 3T3 cells were seeded into 96-well plates, at 1 × 10 5 cells.mL −1 in complete culture medium. Plates were maintained in the aforementioned culture conditions. After 24 h incubation, the culture medium was removed and 100 μL of treatment medium, composed of complete culture medium containing the appropriate concentrations of phages (10 8 PFU.mL −1 and 10 7 PFU.mL −1 ) (2019) 9:6643 | https://doi.org/10.1038/s41598-019-43115-8 www.nature.com/scientificreports www.nature.com/scientificreports/ was added. A positive control consisting of complete culture medium containing different concentrations (0.20, 0.15, 0.10 and 0.05 mg.mL −1 ) of sodium lauryl sulfate (SLS, ThermoFisher) was also added to the cells. A negative control was prepared with complete culture medium, and a blank with culture medium without 3T3 cells. The 96-well plates containing these preparations were incubated for 24 h in the conditions described previously.
In the third day, the culture medium was removed and the cells were washed with pre-warmed PBS. The stock solution of NR (3-amino-m-dimethylamino-2-methlphenazine hydrochloride) was prepared using 0.4% of NR stock in 100 mL of sterile water, which was then mixed with DMEM (to a final concentration of 1.25% (v/v)). Then 100 μL of this solution were added to each well and incubated at 37 °C and 5% of CO 2 , in a humidified atmosphere, for 3 h. The NR solution was removed and the cells were washed with pre-warmed PBS. Subsequently, 150 μL of NR desorb solution (1% acetic acid, 50% ethanol and 49% sterile water) were added to all wells and the plates were shaken for 10 min, to extract the NR from the cells and form a homogeneous solution. NR absorption was measured at an optical density of 540 nm (OD 540 ) in a spectrophotometer. To analyze cell viability, the NR absorbance of the sample was divided by the NR absorbance of the negative control. The assays were performed in triplicate, and considered valid only when the IC50 of SDS was within the confidence interval of 0.070-0.116 mg. mL −1 (95% confidence) and blank OD 540 was equal or greater than 0.3. Three experiments were performed in duplicate.
Efficacy of phages against bacteria colonizing 3T3 cells. Two culture media were used in this assay: Culture medium 1, DMEM supplemented with 10% (v/v) FBS and 1% (v/v) penicillin-streptomycin; and culture medium 2, DMEM supplemented with 10% (v/v) FBS. 3T3 cells were seeded into 96-well plates at 5 × 10 5 cells. mL −1 on culture medium 1, and plates were maintained at growing conditions for 24 h.
Bacterial suspensions grown overnight in TSB medium were centrifuged (11 000 × g, 1 min), washed twice with PBS, and the OD 600 nm adjusted to 0.08 with PBS. Bacterial cells were further centrifuged and re-suspended in culture medium 2. Then, 100 μL of the bacterial suspensions at a concentration of approximately 1 × 10 8 CFU. mL −1 were added to the 3T3 cells in the 96-well plates, previously washed with PBS. Negative control was prepared with culture medium 2. The 96-well plates containing these preparations were incubated for 2 h.
Phage treatment was performed on cells colonized by bacteria after 2 h. Briefly, 100 μL of phage solutions at 1 × 10 7 PFU.mL −1 were added to the cells, and 100 μL of culture medium 2 were added to negative and non-phage treated cells. The 96-well plates containing these preparations were incubated for 6 h or 24 h at growing conditions. Afterwards, the culture medium was removed and cells were washed twice with PBS. Then, 40 μL of prewarmed trypsin were added to all wells, and the plates were incubated for 15 min at growing conditions. To stall trypsin activity, 60 μL of culture medium 2 were added to each well. The CFU method was used for the determination of viable bacterial cell concentration. Mammalian cell death associated with bacterial adhesion was quantified with a Neubauer chamber and an inverted microscope. Three experiments were performed in duplicate.

Statistical analysis.
Statistical analysis was carried out by two-way repeated-measures analysis of variance (ANOVA) with Bonferroni post hoc tests. Differences between samples were considered statistically different for p-values lower than 0.05.

Data Availability
All data generated or analyzed during this study are included in this published article and its Supplementary Information files.