Phage activity against Staphylococcus aureus is impaired in plasma and synovial fluid

S. aureus is a pathogen that frequently causes severe morbidity and phage therapy is being discussed as an alternative to antibiotics for the treatment of S. aureus infections. In this in vitro and animal study, we demonstrated that the activity of anti-staphylococcal phages is severely impaired in 0.5% plasma or synovial fluid. Despite phage replication in these matrices, lysis of the bacteria was slower than phage propagation, and no reduction of the bacterial population was observed. The inhibition of the phages associated with a reduction in phage adsorption, quantified to 99% at 10% plasma. S. aureus is known to bind multiple coagulation factors, resulting in the formation of aggregates and blood clots that might protect the bacterium from the phages. Here, we show that purified fibrinogen at a sub-physiological concentration of 0.4 mg/ml is sufficient to impair phage activity. In contrast, dissolution of the clots by tissue plasminogen activator (tPA) partially restored phage activity. Consistent with these in vitro findings, phage treatment did not reduce bacterial burdens in a neutropenic mouse S. aureus thigh infection model. In summary, phage treatment of S. aureus infections inside the body may be fundamentally challenging, and more investigation is needed prior to proceeding to in-human trials.


Phage activity against Staphylococcus aureus is impaired in plasma and synovial fluid
Michele Mutti 1,3 , David Sáez Moreno 1,3 , Marcela Restrepo-Córdoba 1 , Zehra Visram 1 , Grégory Resch 2 & Lorenzo Corsini 1* S. aureus is a pathogen that frequently causes severe morbidity and phage therapy is being discussed as an alternative to antibiotics for the treatment of S. aureus infections.In this in vitro and animal study, we demonstrated that the activity of anti-staphylococcal phages is severely impaired in 0.5% plasma or synovial fluid.Despite phage replication in these matrices, lysis of the bacteria was slower than phage propagation, and no reduction of the bacterial population was observed.The inhibition of the phages associated with a reduction in phage adsorption, quantified to 99% at 10% plasma.S. aureus is known to bind multiple coagulation factors, resulting in the formation of aggregates and blood clots that might protect the bacterium from the phages.Here, we show that purified fibrinogen at a sub-physiological concentration of 0.4 mg/ml is sufficient to impair phage activity.In contrast, dissolution of the clots by tissue plasminogen activator (tPA) partially restored phage activity.Consistent with these in vitro findings, phage treatment did not reduce bacterial burdens in a neutropenic mouse S. aureus thigh infection model.In summary, phage treatment of S. aureus infections inside the body may be fundamentally challenging, and more investigation is needed prior to proceeding to in-human trials.
Staphylococcus aureus (S. aureus) causes systemic infections 1 , ulcers 2 , atopic dermatitis 3 and implant associated infections 4 .Phage therapy, the therapeutic use of bacterial viruses called bacteriophages (phages), has been proposed for the treatment of S. aureus infections.
Phages are highly specific and replicate inside the host bacteria, ultimately releasing their progeny and killing the host cell.Staphylococcal phages are considered a potential therapeutic option, since they act independently of antibiotic resistance, can lyse bacteria in biofilms in vitro and can evolve towards increased virulence [5][6][7][8][9] .The use of S. aureus phages in animal models 10 and last-resort treatments of humans have had some effect 11 .However, several in vivo studies of staphylococcal phages suggest that their lytic activity might be reduced in presence of plasma, serum or synovial fluid.To our knowledge, the first to observe the inhibitory effect of serum on phages against E. coli and S. aureus lysis were Gratia et al. 12 .This was recently described in more detail for S. aureus 13,14 .Inhibition of phage activity on S. aureus in vivo was reported in a mouse bacteremia model 15 , where mice survived S. aureus experimental infection if the treatment was administered simultaneously, but not if the treatment was administered 30 min after the bacterial infection.This indicates that the adsorption of phages to S. aureus may be reduced in plasma 13,14 .Similarly, in a rabbit model of S. aureus abscesses, phages were only able to reduce the bacterial burden when administered simultaneously, and not 6, 12 or 24 h after the bacterial infection 16 .However, not many of the mechanics of the inhibition of S. aureus phage activity in plasma or synovial fluid have been described.
S. aureus has several virulence factors that use the human coagulation system to its advantage, by either inducing coagulation surrounding the bacteria (via coagulase A), or by binding fibrin and fibrinogen in a process called clumping (via clumping factor A and B, ClfA and ClfB, respectively).This greatly influences both the colonization of the site of infection and the evasion of the immune response 17 .In the context of periprosthetic joint infections (PJI), S. aureus also forms aggregates in synovial fluid 18,19 .This may enhance the stability of S. aureus biofilms 20 , its capacity to evade the immune system 21 and its tolerance to antibiotics 18,20 .

www.nature.com/scientificreports/
In this study, we tested wild type phages and phages evolved for increased virulence 9 in vitro in plasma and synovial fluid, and found that phage lytic activity is inhibited in these conditions.We show that masking of the phage receptor(s) on the surface of S. aureus cells by fibrinogen inhibits phage absorption.Furthermore, the thrombolytic drug tissue plasminogen activator (tPA) was found able to restore phage adsorption and lytic activity in vitro.These findings are consistent with the lack of efficacy of a dose of twice 7.5 × 10 9 plaque forming units (PFU) of S. aureus phages in an in vivo mouse thigh infection model.Overall, while phage therapy to treat S. aureus infections may have some potential based on in vitro results, plasma inhibition represents a limitation to their systemic efficacy in vivo and needs to be investigated further.

Inhibition of phage activity in human and guinea pig plasma
To better understand the extent of phage inhibition in plasma, we measured the antibacterial activity of phages in broth supplemented with human plasma.Multiple, previously described individual phages and cocktails were used, including the Podoviridae phage P66, two Herelleviridae phages (PM4 and PM93), as well as two phage cocktails (cPM398, composed of P66 and vB_SauH_2002, and cPM399, composed of PM4, PM93 and PM56), in a bacterial growth inhibition assay in brain-heart infusion (BHI) medium supplemented with increasing amounts of human plasma.
All tested phages and cocktails showed impairment of their lytic activity, starting already at 0.5% plasma (Fig. 1).While staphylococcal growth inhibition was observed over 48 h in growth medium without plasma, bacterial regrowth was observed in presence of 0.5% plasma.PM93 was less affected than other phages, as bacterial growth over 48 h was less pronounced at 0.5% plasma than for the other phages.The bacterial growth kinetics in presence of P66 were also very similar in absence and presence of 0.5% plasma, (initial lag phase before outgrowth is longer) than for reactions containing 5% or 10% plasma, where P66 appeared completely inactive.
As observed for human plasma, the activity of phage cocktails and single phages was also severely impaired in the presence of guinea pig plasma (Supplementary Fig. 1).

Phages are stable in plasma concentrations up to ~ 50%
We therefore tested the possibility that the loss of phage activity could be due to a direct interaction of the phages with the plasma or the anticoagulant.We incubated various described Herelleviridae phages with increasing concentrations of heparin or citrate plasma and tested their residual activity after 24 h.All phages tested showed < 1 log PFU/ml titer reduction after 24 h incubation at 37 °C in 50% heparin or citrate plasma (Fig. 2).At plasma concentrations > 50%, the phage titers dropped by up to 4 log PFU/ml.
These results suggest that direct interactions between plasma or anticoagulant and phages may play a role in their instability at high plasma concentrations, but that neither plasma nor anticoagulant (data not shown) reduce phage activity at concentrations below ~ 50%.Therefore, the reduced phage activity observed at a plasma concentration as low as 0.5-10% (Fig. 1) cannot be explained by a direct interaction between phages and plasma or anticoagulant.

Phage replication is not substantially impaired even though bacteria are not cleared
In phage propagation experiments, we used an input bacterial titer of 5 × 10 6 CFU/ml and PM93 at an initial MOI of 0.1.We measured the phage titer after 24 h incubation as a function of plasma concentration.Phage PM93 was selected since its activity was less impaired in plasma compared to the other phages (Fig. 1).Phage PM93 propagated in plasma in all the plasma concentrations tested, with a maximum increase of 4 logs at 7.5% Figure 1.48 h of growth inhibition assay in human heparin plasma (Red Cross, mix of two donors, blood groups A+/0+) The graphs show turbidity assays of bacterial suspensions at a starting bacterial concentration of 2.5 × 10 6 CFU/ml in presence or absence (growth control) of the different phages or phage cocktails at 2.5 × 10 8 PFU per ml (MOI = 100).The mixture was incubated for 48 h at 37 °C at different concentrations of human plasma as indicated on top of the graph.CFU colony forming units, MOI multiplicity of infection, OD optical density at 600 nm, PFU plaque forming units.plasma (Fig. 3a).The bacteria were able to grow despite efficient propagation of the phages in presence of ≥ 2.5% plasma (Fig. 3b).
These results suggest that in plasma, although phages can still propagate, there are bacterial cells that are surviving the phages.This may be due to a reduced adsorption rate, as suggested by previous studies 15 .

Phage adsorption is impaired in plasma
To test if the adsorption was impaired, we measured the efficiency of the center of infection (ECOI) in presence of plasma.In ECOI, one plaque corresponds to a bacterium successfully infected with a phage 22 .Due to the excess of bacteria relative to phages, ECOI is very sensitive to changes in low adsorption rates (while the classic phage adsorption assay 23 is more suitable to measure [log-scale] reductions in phage titer, indicating a high rate of adsorption).
As shown in Fig. 4, 10% plasma in BHI reduced the adsorption of the phages by 1.8-log infected CFU/ml compared to the same phages in BHI alone, suggesting that human plasma substantially affects the adsorption of the phage onto S. aureus cells.

Figure 2.
Stability of phages in human plasma.Phages were incubated for 24 h at 37 °C in BHI supplemented with human plasma as indicated, after which phage titer was determined by spot test on their common S. aureus host strain 124605 9 .The experiments were conducted as single replicates.A statistical evaluation (oneway ANOVA) that treats differences between the phages purely as experimental uncertainty yields p > 0.05 for all plasma concentrations equal or lower than 50% (vs.0%), p = 0.009 for 0% vs. 60% plasma and p < 0.05 for all values above 60% plasma.The vertical dotted line indicates 50% plasma concentration.The 0% datapoint corresponds to BHI not supplemented with plasma (positive control).www.nature.com/scientificreports/

Purified fibrinogen is sufficient to impair phage activity
Since phage adsorption was impaired in the presence of plasma, we tested whether this was related to the ability of S. aureus to form clumps, which is induced by fibrinogen 18 , and involves the formation of macroscopic biofilm-like structures.We first verified that purified fibrinogen in medium clumps S. aureus in absence (not shown) or presence of phages, starting at a concentration of 0.2 mg/ml (Fig. 5A).In absence of fibrinogen, the suspension exhibited turbidity and S. aureus clumping was not observed.
In the same conditions we also observed that the inhibition of the phage activity correlated with the concentration of purified fibrinogen supplemented BHI, as shown in Fig. 5B.All tested phages (Podoviridae P66 and highly diverse Herelleviridae BT3, PM4 and Remus/Romulus derivatives PM56/PM93) were completely inhibited by the addition of fibrinogen, starting at 0.4-1 mg/ml, which corresponds to 25% of its physiological  www.nature.com/scientificreports/concentration 24 .These findings suggest that fibrinogen is sufficient to induce clumping of S. aureus and to inhibit the phage activity in plasma.

S. aureus induces clotting and tPA lyses S. aureus-induced clots in plasma of different species
Fibrinogen alone can inhibit phage activity via clumping; however, in plasma S. aureus also induces coagulation, forming local clotting around the cells, which might also hinder the phages from adsorbing to the cell surface.It was previously shown by Kwiecinski et al. 25 that the thrombolytic drug tPA prevented S. aureus biofilm formation in presence of plasma by blocking its attachment.We tested whether fibrinolysis with tPA, might lyse the pre-formed clot and restore phage activity in vitro.
As shown in Supplementary Table 1 and Fig. 6, human tPA lysed S. aureus-induced clots in human heparin and citrate plasma, guinea pig citrate plasma and rabbit heparin plasma.Of note, S. aureus did not induce macroscopically visible clotting in murine citrate plasma, guinea pig heparin and rabbit citrate plasma.Moreover, no macroscopic clumping was observed in murine citrate or guinea pig plasma under these experimental conditions.

tPA partially restores phage activity in human plasma
Since tPA was able to efficiently lyse the clot formed by S. aureus ATCC 43300, we tested its effect on phage activity in human plasma.
A tPA concentration of 10 µg/ml was insufficient to restore phage activity at even 1% plasma (Fig. 7a).However, as shown in Fig. 7b, tPA at 64 µg/ml partially restored phage activity in up to 5% human citrate plasma.
Since fibrinolysis by tPA improved the phage activity, we tested whether inhibition of the coagulation, prior to clot formation, may have the same effect.Dabigatran (Pradaxa ® ) has inhibitory activity on the staphylococcal coagulase 26 , and therefore it may prevent the clotting induced by S. aureus.As shown in Supplementary Fig. 2, dabigatran alone, after incubation with plasma and bacteria, was not able to improve the phage activity.This underlines the observation that aggregation of the S. aureus cells (e.g. by fibrinogen-induced clumping) is sufficient to block phage activity.
These results strongly indicate that clotting (a process distinct from clumping) contributes to but is not required for the inactivation of phage activity in human plasma.

tPA restores phage activity in guinea pig plasma
As discussed above, human tPA was cross-reactive on S. aureus-induced clots in plasma from several mammalian species.Furthermore, only human tPA is commercially available.Therefore, we tested whether human tPA was also able to restore phage activity in guinea pig plasma.
As depicted in Fig. 8, tPA rescued phage activity in guinea pig heparin plasma, while with the highest guinea pig citrate plasma concentration there was staphylococcal growth after 12 h.The stronger effect of tPA in guinea pig plasma compared to human plasma could be explained by two factors: first, the fact that it contains lower concentrations of coagulation factors 27 .Second, in guinea pig plasma no clumping was observed (Supplementary Table 1).
These results suggest that both clumping and clotting contribute to phage inhibition in human plasma, while in guinea pig plasma clotting (and not clumping) seems to dominate the loss of phage activity.

tPA improves phage adsorption in guinea pig plasma
We tested whether phage activity could be recovered by tPA via recovery of adsorption with an ECOI assay as described above.tPA increased the infectivity of phages ten times (1 log), compared to the no tPA control (Fig. 9), confirming that tPA, in guinea plasma, increases the accessibility of the phage receptor on S. aureus cells.

Phage breeding in presence of plasma does not improve the phage activity
To improve the activity of phages in plasma, a combination of phages was subjected to directed evolution (what we called breeding), by passaging them with 24 different S. aureus strains and in presence of plasma for 20 rounds as described previously 9,28 .The single phages isolated after breeding did not show an improved activity in plasma in comparison to the ancestor phages (data not shown).These results imply that to be active in plasma, the phages might need a genetic change larger than what can be achieved in this lab-scale directed evolution assay.In the context of prosthetic joint infections, S. aureus is also able to form aggregates in the presence of synovial fluid.In previous studies 18 , it was shown that these aggregates in synovial fluid led to enhanced antibiotic resistance.To revert the formation of these macroscopic, biofilm-like structures, they dissolved the aggregates with tPA at 1 mg/ml (a concentration not achievable in vivo due to the risk of hemorrhage) thereby successfully restoring the effectiveness of antibiotics 18 .We tested whether phage activity in synovial fluid could also be restored with tPA.Macroscopic clumps were formed by incubating 3 × 10 7 CFU of S. aureus ATCC43300 at 37 °C for 24 h in 70% bovine synovial fluid, and then treating these with a combination of tPA and either phage cocktail cPM399 or a combination of vancomycin (VAN) and rifampicin (RIF) at tenfold MIC.As depicted in Fig. 10a, S. aureus formed large aggregates  in bovine synovial fluid after 24 h, consistent with the observation by Gilbertie et al. 18 in human, equine and porcine synovial fluid.
As depicted in Fig. 10b, none of the single treatment alone could dissolve the clots.However, tPA treatment alone led to turbid solutions with aggregates (as compared to clear solutions without tPA), indicating a partial de-aggregation of S. aureus clumps, consistent with previous findings 18 .The treatment with the combination of cPM-399 and tPA led to the dissolution of the aggregates and a clear solution, pointing at a strong synergy between tPA and the phage cocktail cPM-399.Interestingly, the combination of tPA and antibiotics did not induce dissolution of the aggregates.
The treatment with phage cocktail (cPM-399) also led to a statistically significant reduction of bacterial viability by 1.5 log CFU/ml versus control (Fig. 10c).Consistently with the turbidity observation depicted in Fig. 10b, tPA treatment alone did not reduce the CFU/ml.In contrast, combined treatment with tPA and cPM-399 led to a median reduction of 5 log CFU/ml.This demonstrates a strong synergy between tPA and the phage cocktail cPM-399.
We also measured the phage counts (PFU/ml) after the treatment and found that, consistent with the results depicted in Fig. 10a-c, the phages replicated approximately ten times more (1 log PFU/ml) in the wells treated with the combination of tPA + cPM-399 compared to the wells treated with PM-399 only (Fig. 10d).
These results are in agreement with the findings of Gilbertie et al. 18 , who showed that various antibiotics were not effective in killing the bacteria as long as aggregates could be observed macroscopically.
Whether tPA can be used to enhance the efficacy of phages to treat S. aureus infections in humans largely depends on the risk of hemorrhaging.In this respect, the approach shown here indicates a pathway to reduce the effective concentration of tPA to 64 µg/ml, if combined with phages, as compared to the combination with antibiotics (tPA at 1000 µg/ml) described previously 18 .

Phages were inactive in a mouse model of S. aureus thigh infection
Given that in vitro, phage activity was suppressed in presence of plasma, we evaluated whether this was relevant in vivo.
In the thigh, the C max was 1.8 × 10 6 PFU/ml after 30 min after i.p. administration of PM4, compared to a C max of 8.8 × 10 6 PFU/ml at the same time point after i.v.administration.The difference between administration routes was not statistically significant (Supplementary Fig. 3).In plasma, the C max was 1.8 × 10 8 PFU/ml 30 min after i.p. administration of PM4, compared to a C max of 1.2 × 10 9 PFU/ml at the same time point after i.v.administration.Also in this case, the difference between administration routes was not statistically significant.
The phage distributed effectively across the different tissues, although the titers in the thigh were lower compared to the other organs, where the titers decreased over time from 10 4 to 10 7 PFU/g after 30 min, to 10 3 -10 5 PFU/g at 6 h (Supplementary Fig. 3).
Next, we assessed the efficacy of PM4 and one of its ancestor wild type phages, 812, in a thigh model in neutropenic mice (to avoid confounding effects with the immune system), as adapted from Wicha et al. 29 using vancomycin as a control (N = 5 per group).Two doses of 7.5 × 10 9 PFU or 60 mg/kg vancomycin were administered at 2 h and 8 h after infection, respectively, and mice were sacrificed 24 h after infection (16 h after the second dose).As shown in Fig. 11, and despite presence of phages demonstrated in the PK study, neither PM4 nor 812 significantly reduced the bacterial burden in the thigh, compared to the control group.Vancomycin reduced the bacterial burden by about 4 log CFU/g tissue compared to vehicle treatment.
These results suggest that the phage-host interaction is affected also in vivo, most probably due to the same phenomenon observed in vitro: the phages are not able to either reach the bacteria in the infected animals, or, more likely, the bacteria are protected from the phages in the environment of the blood and interstitial tissues.This latter interpretation fits well with the data shown previously, indicating that S. aureus binds elements of the coagulation system, blocking adsorption of the phages.

Discussion
In this study, we analyzed the activity of staphylococcal phages in presence of physiological fluids (plasma and synovial fluid).We show that both plasma and synovial fluid of various mammals, including humans, impair the activity of any phage tested, of taxonomic families as diverse as Podoviridae and Herelleviridae.We thereby substantially extend the findings described previously 13,14 .Although the phages could not prevent the growth of bacteria in presence of these physiological fluids, we found that phages were propagating under these conditions, and were stable at plasma concentrations below 50%.This suggests that the impairment of bacterial growth inhibition by the phages in vitro is not due to a stability issue, but rather to an altered phage-host interaction.
Shinde et al. hypothesized that this inhibition might be due to the masking effect of the coagulase activity, and/or masking by the anti-teichoic acid IgG antibodies 13 .However, our data indicate that reduced phage-host absorption via a combination of clotting (e.g. via coagulase A) and clumping (e.g. via ClfA and ClfB) most likely reduces phage activity.Given the complex and interconnected mode of action of these virulence factors, we evaluated the role of the two pathways separately.
First, we tested the effect of fibrinogen-dependent clumping, driven by ClfA and ClfB, on the phage activity, by adding human purified fibrinogen to the medium.Fibrinogen-dependent clumping leads to the formation of biofilm-like aggregates, which play a pivotal role in the pathogenesis, and decrease the susceptibility of S. aureus to antibiotics 17 .Purified fibrinogen added to medium was sufficient to inhibit phage activity, and to form macroscopically visible clumps.Interestingly, the clumping activity of S. aureus was observed in human plasma, but not in guinea pig or murine plasma.Despite the lack of clumping activity in the latter species, phage activity was also inhibited.Therefore, we tested if coagulation could also play a role.We assessed this by testing both the destruction of the clot by tPA, and by preventing its formation with dabigatran.The addition of tPA to human plasma improved phage activity, particularly at low concentration of plasma.Moreover, the supplementation with tPA increased phage adsorption by one order of magnitude, supposedly promoting phage propagation and reducing growth of the bacterial population.In contrast, the addition of dabigatran did not improve phage activity in plasma.These apparently contradictory results may be explained considering that tPA activates plasmin, which in turn not only degrades fibrin in clots (coagulase A -dependent), but also degrades the fibrinogen that mediates clump cohesion (ClfA-and ClfB-dependent) 30,31 .tPA therefore induces the lysis of both clots and clumps.
This suggests that while the clumping activity in human plasma is sufficient to impair the phages, the coagulation is not, and both effects may be additive or synergistic.In guinea pig plasma, we did not observe any macroscopic clumping, suggesting that clotting plays a more dominant role in this species, and it explains why tPA is very effective in guinea pig plasma.
We also tried to breed the phages in plasma, as previously described 9,28 , to improve virulence and kinetics of propagation in medium.Despite the high number of phages used and 20 rounds of passaging, we did not observe any improvement.This indicates that the genetic changes required in the phages to overcome the adsorption inhibition could not be achieved in this setup.
We demonstrate that the inhibitory effect on phages is not limited to plasma but extends to synovial fluid (which does not contain any anticoagulants).This can be explained by the presence of fibrinogen 17 and other elements of the coagulation system in this fluid 32,33 .However, the activity of phages was impaired only when more than 50% synovial fluid was present, consistent with the lower concentration of coagulation proteins compared to plasma 32,33 .
To dissolve aggregates of S. aureus in synovial fluid, Gilbertie et al. 18 tested a dose of 1 mg/ml of tPA to restore antibiotic activity, a concentration that would incur a strong risk of hemorrhaging in vivo 34 .Our data indicate that the synergy between tPA and phages would allow a substantial reduction in the effective concentration of tPA (from 1 mg/ml to 64 μg/ml) to dissolve the macroscopic S. aureus clumps forming in synovial fluid and killing the bacteria.
To validate our conclusions in vivo, we tested phage activity in a thigh infection model in neutropenic mice preceded by a PK study to assay the bioavailability and distribution of PM4 phages administered i.v. or i.p.. Consistent with the hypothesis that phage adsorption is inhibited in physiological fluids containing elements of the coagulation system, no CFU reduction was observed between phage treatment and control, despite good bioavailability of phages in all tissues tested.
In a previous study by Prakaz et al. 35 , a 4-phage-cocktail including 2002 (ancestor of PM4 used in our study) was effective in reducing bacterial densities in pneumonia in rats.This suggests that in certain environments, such as the lungs, phages may not be inhibited.
Our experiments also highlight the importance of testing phage activity in vitro under conditions that resemble the physiological state as much as possible.Moreover, the study of staphylococcal aggregates in body fluids such as plasma and synovial fluid, along with the study of biofilms, is an imperative to further develop phage therapy for infections such as PJIs, chronic wound infections, osteomyelitis, soft tissue abscesses, and endocarditis.www.nature.com/scientificreports/

Phage activity kinetics on planktonic bacteria
Phage activity was measured by reading the OD620 or OD590 every 5 min for several hours at 37 °C with either Tecan F NANO+ or its equivalent Tecan F machine Before each reading the plate was shaken for 5 s with 3 mm amplitude.In all the experiments, BRANDplates ® , 96-well, pureGrade™ S, PS, transparent, flatbottom plates (#781662, BRAND) were used.

Determination of the efficiency of center of infection (ECOI)
Adsorption of phages was assessed by determining the efficiency of center of infection (ECOI) 22 .Briefly, bacteria were pre-incubated with either BHI or BHI supplemented with plasma for 30 min at 37 °C, and then phages (MOI of 0.01) were added.After 5 min, free phages were removed by centrifugation at 11,000g and washing with BHI adjusted to pH 3.After pelleting again at 11,000g, cells were spotted on top agar prepared with uninfected bacteria, and incubated overnight at 37 °C.The day after plaques were enumerated, and the adsorption was calculated by assuming that a plaque corresponds to a single infected cell.

Phage stability testing in presence of plasma
Phages were incubated in BHI supplemented with plasma for 24 h at 37 °C, and then they were spotted on a bacterial lawn of their host, S. aureus strain 124605, and the plate was incubated overnight at 37 °C.The day after plaques were enumerated, and the relative phage concentration was calculated.

Phage activity on pre-formed aggregates in synovial fluid and plasma and treatment with tPA
The effect of tPA on phages in synovial fluid was assessed with a protocol similar to 18 , as follows.700 ml of bovine synovial fluid was inoculated with 300 ml of S. aureus strain ATCC 43300 at OD600 = 0.2 in BHI, and the mixture was incubated for 24 h at 37 °C with shaking at 120 RPM for the formation of clumps.The supernatant was removed, the following reagents in order: 700 ml of synovial fluid, 100 ml of tPA (final concentration 64 mg/ ml) or DPBS, and PM-399 (in BHI, final concentration of 2 × 10 8 PFU/ml) or BHI, or rifampicin/vancomycin (rifampicin 0.23 mg/ml; vancomycin 33.3 mg/ml, both at a concentration tenfold higher their MIC).The sample was incubated for 24 h at 37 °C and 120 RPM.After the treatment, clumps were thoroughly resuspended by pipetting, serially diluted and plated to assess CFU and PFU counts.To detect the CFU, the sample was diluted 1:10 in PBS (100 mM KH 2 PO 4 and 50 mM NaCl) pH 3.0 for 1 min to inactivate the phages, and then serially diluted in 1:10 in PBS pH 7.0, and spotted on BHI agar.To detect the PFU, the sample was filtered with MultiScreenHTS-GV 0.22 mm (#MSGVS2210, Merk, Germany), diluted in PBS pH 7.0 and spotted on a lawn of S. aureus strain ATCC 43300.

Breeding of phages in plasma
Phage breeding was conducted based on the principles of the Appleman´s Protocol 39 , by using previously described phages 9,28 .Two different phage cocktails were bred on a subpanel of 24 S. aureus strains (Suppl.Table 2) in a 96-well microtiter plate.One phage cocktail was based on the Remus/Romulus (Remus, Romulus, PMP56 and PM93) and the other was based on the phage K-like phages (812, 2002, BT3,PM4, PM9 and PM32).Before the infection, the S. aureus strains were pre-incubated 30 min at 37 °C, in BHI supplemented with 10% of human plasma, to allow the formation of the clumps and clots.For the first round of breeding, single phages with similar concentrations were mixed and then serial tenfold diluted in PBS.Starting concentrations of 2 × 10 7 PFU/ml of phages were mixed with bacterial overnight cultures diluted to ~ 1 × 10 8 CFU/ml.After overnight incubation at 37 °C, clear lysates and the first dilution of turbid lysates were pooled, sterile filtrated and used to infect the next round of breeding.This was repeated for up to twenty rounds.Single individual phages were isolated from the bred mixture lysate by re-streaking on host strain at least four times.

Pharmacokinetic study in NMRI mice and thigh model of infection
Phage PM4 and 812 were purified by tangential flow filtration (TFF) prior to administration to mice.
For the PK study, 18 female NMRI mice (29 g) received 0.2 ml of PM4 (2.6 × 10 10 PFU) via i.p. or i.v.injection, followed by euthanasia with Ketamine/Xylacin (0.1 ml narcotic mix/20 g) after 30 min, 2 h or 6 h, in groups of 3 randomly selected mice.The organs (liver, lungs, thighs, kidneys, and spleen) were explanted and homogenized in saline.Blood was withdrawn by cardiac puncture and collected in heparin tubes, and plasma was prepared by centrifuging the blood for 5 min at 4500g.All the organ samples were centrifuged at 20,000g for 5 min and sterile-filtered by a 0.22 µm filter plate.In order to assess the phage tires, the samples were serially diluted in saline and 2 µl drops were spotted on a lawn of S. aureus strain 124605, the plates were incubated at 37 °C overnight and plaques were enumerated the day after.
For the thigh model of infection, 25 female NMRI mice were made neutropenic by administering cyclophosphamide, at 150 mg/kg and 100 mg/kg four and one day before the bacterial challenge, respectively.The animals were then challenged by intramuscular injection of 1 × 10 5 CFU/thigh of S. aureus strain B409.The animals were randomly divided in 5 treatment groups: early control (animals sacrificed at the onset of the first treatment), late control (animals dosed with buffer only), phage PM4 (7.5 × 10 9 PFU/dose), phage 812 (7.5 × 10 9 PFU/dose), and vancomycin (60 mg/kg/day).Treatments were provided 2 h and 8 h after bacterial challenge, a summary of treatment groups is provided in Supplementary Table 3. Phages were administered i.p. and vancomycin i.v.
Early control group were sacrificed right before the onset of treatment (time 2 h).The remaining mice were sacrificed 24 h after infection and thighs were collected for the assessment of bacterial burden.Thighs were homogenized in saline and used for determination of bacterial CFU/ml in MH agar with sheep blood.For determination of phage PFU/ml samples were centrifuged at 20,000g for 5 min and sterile-filtered by a 0.22 µm filter

Figure 3 .
Figure 3. (a) PM93 phage propagation in plasma.The graph shows phage titer after 24 h incubation as a function of plasma concentration.(b) 24 h of growth inhibition assay by PM93 in human plasma.The graphs show OD600 of bacterial suspensions at a starting bacterial concentration of 5 × 10 6 CFU/ml, in presence or absence (growth control) of PM93 (MOI of 0.1).The mixture was incubated for 24 h at 37 °C at different concentrations of human citrate plasma as indicated on top of the panel.

Figure 4 .
Figure 4. ECOI of Herelleviridae phage PM93, at MOI of 0.01 in human heparin plasma.The log 10 of the number of infected cells (cell suspension at 5 × 10 8 CFU/ml infected with 5 × 10 6 PFU/ml) is plotted against the different conditions tested in the assay: control without bacteria, BHI, BHI + 10% plasma.LLOQ lower limit of quantification.

Figure 5 .
Figure 5. (A) S. aureus aggregates in presence of fibrinogen.Macroscopic clumps formed after 24 h in wells treated with phage P66, with increasing fibrinogen concentration as indicated.(B) 24 h of growth inhibition assay in presence of fibrinogen.The graphs show OD600 of bacterial suspensions (5 × 10 6 CFU/ml) in presence or absence (growth control) of the different bacteriophages (MOI = 10) as indicated on the left of the panel.The mixture was incubated in BHI for 24 h at 37 °C at different concentrations of fibrinogen as indicated on top of the panel.Physiological concentration range according to Asselta et al. 24 .

Figure 6 .
Figure 6.Example of tube clotting test.400 µl of human 25% heparin plasma (diluted with saline) were clotted by adding 10 µl of S. aureus strain ATCC 43300 at OD = 0.5 in suspension and incubated at 37 °C overnight.The clot was then treated by adding 225 µl of 25% (final concentration) of plasma diluted with saline and 64.5 µg/ml tPA (final concentration) or no tPA (control).Where the clot was dissolved, the material collects at the bottom of the tube (clot lysed), while it sticks to the top (unlysed clot) upon treatment with saline.

Figure 7 .
Figure 7. 24 h of growth inhibition assay of PM-398 in human citrate plasma supplemented with tPA.The graphs show optical density measurements at 600 nm (OD600) of bacterial suspensions at 5 × 10 6 CFU/ml in presence or absence of cPM 398 (phage cocktail of P66 + 2002, MOI 0.1) as indicated on the left and on top of the panel.The mixture was incubated for 24 h at 37 °C at varying concentrations of plasma and with different concentration of tPA as indicated in the graph [(a) 10 μg/ml, (b) 64 μg/ml, or PBS].

Figure 8 .
Figure 8. 24 h growth inhibition assay of PM-399 (MOI 100) in guinea pig plasma in presence or absence of tPA and plasma.The different concentrations of citrate or heparin plasma used are indicated on top and left of the panel.The graphs show optical density measurements at 600 nm (OD600) of suspensions of S. aureus ATCC43300 at a starting bacterial concentration of 5 × 10 6 CFU/ml, for 24 h at 37 °C.

Figure 9 .
Figure 9. ECOI of Herelleviridae phage PM93 at MOI 0.01 in guinea pig heparin plasma.The log 10 of the number of infected cells (cell suspension at 5 × 10 8 CFU/ml infected with 5 × 10 6 PFU/ml) is plotted against the different conditions tested in the assay: BHI, BHI + 10% plasma and BHI + 10% plasma + tPA at 64 μg/ml.LLOQ lower limit of quantification.

Figure 10 .
Figure 10.Effect of tPA in synovial fluid in combination with phages or antibiotics.(a) Clumps of S. aureus strain ATCC 43300 in bovine synovial fluid after 24 h of incubation (in triplicate).(b) Same samples of (a) 24 h after the treatment.(c) CFU/ml counts of samples in (b), p-values were calculated with a two-tailed one-way ANOVA.(d) PFU/ml counts of samples in (b), p-value was evaluated with an unpaired t-test.LLOQ lower limit of quantification, VAN vancomycin, RIF rifampicin. https://doi.org/10.1038/s41598-023-45405-8

Figure 11 .
Figure 11.S. aureus log CFU retrieved per thigh in vivo after phage treatment per treatment group.Neutropenic mice (N = 5 per group) were challenged with S. aureus strain B409 to both thighs, and treated with phages, antibiotics, or buffer only (late control).The early control group was sacrificed at the onset of the first dosing (2 h).p-value was calculated with one-way Brown-Forsythe and Welch ANOVA (corrected by Games-Howell test).