Staphylococcus aureus is a Gram-positive bacterium found on the skin and mucosa of humans and animals. It causes a wide range of human diseases, from endocarditis to pneumonia.¹ Methicillin-resistant S. aureus (MRSA) is a form of S. aureus that displays high levels of drug resistance towards β-lactam antibiotics, such as penicillins and cephalosporins. β-lactams act by inhibiting the activity of transpeptidase enzymes during peptidoglycan synthesis in bacterial cell walls.² β-lactams are one of the most commonly prescribed drug classes, and many are designated as “Critically Important” antimicrobials by the World Health Organization.³

A chief mediator of β-lactams resistance in MRSA is the SCCmec cassette, a mobile genetic element that carries the resistance gene mecA. MecA encodes for the penicillin-binding protein 2A (PBP2a), a transpeptidase that has a low affinity for β-lactams.⁴ This lower affinity permits PBP2a to participate in peptidoglycan synthesis even in the presence of β-lactams, ultimately resulting in cell survival. In addition to mecA, numerous MRSA strains carry β-lactamases, such as BlaZ, that degrade β-lactams, thus further contributing to drug resistance. MRSA infections pose a considerable public health risk as they are notoriously difficult to treat and are widespread in communities and hospital settings. In 2019, MRSA alone accounted for more than 100,000 deaths attributable to drug-resistant infections worldwide.⁵ Developing novel solutions to combat MRSA is a major focus in academia and industry.

Bacteriophages (phages) are viruses that infect and kill bacteria, posing one of the greatest existential threats to bacterial communities. Some estimates suggest that 40% of all bacterial mortality worldwide is caused by phage predation.⁶ Due to this lethality, phage therapy has been proposed as an attractive alternative for combating drug-resistant bacterial infections. In phage therapy, lytic phages are administered to kill the bacterial pathogen(s) causing an infection. Phages offer certain advantages over traditional antibiotics: they are highly specific for their hosts by reducing off-target killing; they self-amplify and evolve, enabling the rapid generation of new phage variants with improved activities; and they are generally regarded as safe, based on extremely rare cases of toxicity in animals and patients.⁷ Indeed, against staphylococcal infections, over a dozen promising case studies and clinical trials have been reported in the past decade.⁸ Despite these advances, routine use of phage therapy is still met with challenges.

Chief among these challenges is the inevitable rise of phage resistance as phage predation exerts a strong selective pressure on bacterial populations. According to one analysis that focused on phage therapy outcomes, resistance against phage evolved in 75 percent of human clinical cases in which the evolution of resistance was monitored.⁹ Mutations represent a chief pathway by which bacteria evolve resistance against phage. To date, the best characterized phage resistance mutations strains involve alterations of the cell surface that affect phage attachment. For most lytic phages, the first step of their infection involves attachment of the phage particle to a receptor molecule (or multiple molecules) on the cell surface; these receptors are often proteins or sugar moieties, which are recognized by phage proteins.^8,10,11 Mutations in these receptor molecules can prevent phage binding, leading to resistance. For example, in Escherichia coli, mutations in the cell-wall protein LamB confer resistance against lambda phage infection, while in S. aureus, mutations that modify the glycopolymers in the cell wall, such as wall teichoic acid (WTA), have been shown to limit the infection by phages.^11,12 Complicating the picture, studies have revealed that a plethora of additional host mechanisms, including dedicated anti-phage defense systems, can impact the evolution of resistance against phage.^13,14 These problems highlight the importance of developing phage treatment strategies that minimize or capitalize on the evolution of phage resistance.¹⁵

A unique aspect of phage therapy is the possibility to exploit evolutionary trade-offs to combat resistant pathogens. A genetic trade-off is defined as an evolved trait that confers a fitness advantage against a particular selective pressure at the expense of reduced fitness against an unselected pressure. Across many different species of bacteria, such fitness trade-offs are known to occur between phage resistance and antibiotic resistance (Figure 1A). For example, selection by a lytic phage PYO^Sa on the methicillin-sensitive S. aureus Newman selected for mutations in the femA aminoacyltransferase, leading to increased resistance against phage infection but reduced resistance towards antibiotics.¹⁶ These studies raise the question of whether such phage/antibiotic resistance trade-offs can be induced in actual MRSA strains, which would have significant therapeutic implications. Employing phages to drive the evolution of such beneficial trade-offs could serve as a means to reduce drug resistance. However, very little is known about how frequently these trade-offs occur in different bacterial species, or through what molecular mechanisms they work.

Figure 1.

Here we investigate how infection by staphylococcal phages causes three different MRSA strains to evolve sensitivity to different types of β-lactam antibiotics. Interestingly, despite being treated by the same phage, the loss of resistance phenotype is mediated by mutations distinct to each MRSA strain. In addition to resistance, phage infection also causes a reduction in hemolysis and clumping, which are key virulence phenotypes in MRSA. Finally, we isolated a mutant version of the phage that displays enhanced killing, improved resistance reduction, and reduced virulence phenotypes across all three MRSA strains.

Results

Identification of ΦStaph1N with activity against multiple MRSA strains

We subjected three MRSA strains – MRSA252 (USA200), MW2 (USA400), and LAC (USA300) – to a panel of staphylococcal phages using plaquing assays (Table S1, Figure S1). All three MRSA strains are pathogenic isolates that have been implicated in human infection and disease and have been used as representative examples for studying MRSA.^17–19 Of the phages tested, one phage, ΦStaph1N, was able to form distinct plaques on all three MRSA strains (Figure 1B). ΦStaph1N belongs to the myovirus family of lytic staphylococcal phages, with a genome size of ∼138 kB.^20,21

Despite its ability to infect all three MRSA strains, ΦStaph1N infection was often unable to effectively lyse MRSA cultures. For example, both MW2 and LAC displayed incomplete lysis in liquid culture at multiplicities of infection (MOIs) of 0.1 or lower (Figure S2). At an MOI of 0.1, we observed that all three MRSA strains lysed within 2-3 hours following phage infection but recovered high cell density after passaging one percent of the culture into fresh media. We therefore hypothesized that predation by ΦStaph1N selected for mutations conferred resistance against ΦStaph1N infection, allowing the culture to regrow. Indeed, ΦStaph1N was unable to form plaques on recovered MRSA cultures that survived in the initial ΦStaph1N infection (Figure 1B).

Next, we asked whether phage resistance against ΦStaph1N infection could cause a trade-off in drug resistance in MRSA. Since both phage and β-lactams interface with the bacterial cell wall, we hypothesized that phage resistance mutations could impact β-lactam activity. Parental strains of MRSA252, MW2, and LAC displayed higher minimal inhibitory concentrations (MICs) of ≥48 µg/mL against the β-lactams oxacillin (OXA), cefazolin (CEF), amoxicillin (AMX), and amoxicillin & clavulanic acid (AMX+CA), visually indicated by their ability to form lawns surrounding antibiotic strips (Figure 1C, Table S2). The MRSA strains were sensitive to vancomycin (VANC; MICs = 1.5 µg/mL), which inhibits cell wall synthesis through a different mechanism than β-lactams.

By contrast, all strains that survived ΦStaph1N infection displayed a reduction in resistance against OXA, CEF, and AMX+CA, with fold reductions in MIC between 10 and 1000-fold (Figure 1D). No changes in MIC were observed with VANC. These observations suggest that the loss of resistance might be specific to the mechanism of action of β-lactams. Furthermore, phage-treated strains remained resistant to AMX alone, likely caused by the activity of β-lactamases present in the three MRSA strains. Together, these results suggest that predation by ΦStaph1N causes three different MRSA strains to lose their resistance towards β-lactams.

Next, we asked how infection at lower MOIs with ΦStaph1N affected β-lactam resistance in MRSA. We infected the three MRSA strains with ΦStaph1N at MOIs ranging from 0.01 to 10^-5, isolated the surviving MRSA cells, and tested the survivors for their MIC against oxacillin; each condition was tested in three independent replicates (Table S2). For MRSA252, we observed a ∼3-order of magnitude fold reduction of MIC in the surviving cells at an MOI of 10^-5. With MW2, the reduction of MIC markedly decreased with lower phage levels, showing on average no reduction in MIC at MOIs of 10^-3 or lower (Table S2). For LAC, two of the replicates displayed an order of magnitude reduction of MIC at an MOI of 10^-4, while one replicate did show a strong decrease in MIC at the same MOI. These results suggest that for MRSA252, MOIs as low as 10^-5 of ΦStaph1N infection can still drive the loss of resistance, while for MW2 and LAC, higher MOIs of phage are needed to ensure the same outcome of reduced β-lactam resistance.

We then asked whether the reduction in β-lactam can be achieved with a phage from a different taxonomical group. The MRSA LAC strain is sensitive to infection by <ΙNM1ψ6, a lytic version of the temperate phage <ΙNM1, displaying plaquing comparable to ΦStaph1N (Table S1, Figure S1). <ΙNM1ψ6 belongs to the siphovirus family and was derived from the S. aureus Newman strain.^22–24 We therefore examined whether infection with <ΙNM1ψ6 could also reduce β-lactam resistance in LAC. We infected LAC with <ΙNM1ψ6 at an MOI of 0.1 and measured the MIC against oxacillin of the surviving cells. Here, recovered LAC cultures did not show a reduction in MICs (Figure S3). This suggests that <ΙNM1ψ6 is unable to reduce resistance in susceptible MRSA hosts, raising the possibility that different phages vary in their ability to change β-lactam resistance in MRSA.

Discovery of mutant in ΦStaph1N with enhanced activity against MRSA

A key advantage of phage therapy over conventional pharmaceutical antibiotics is that phages themselves can evolve. For ΦStaph1N, we noticed that while the phage was able to plaque all three MRSA strains, its plaque-forming efficiency was reduced on the MW2 and LAC strains (Figure 1B, Figure S1). ΦStaph1N plaques on MW2 and LAC strains were hazy and the overall efficiency of plaquing was approximately two orders of magnitude less than that with MRSA252. Unexpectedly, we consistently observed smaller, clear plaques arising in the larger, hazy plaques of LAC (Figure 2A). This led us to hypothesize that these clear plaques are caused by a mutant ΦStaph1N that plaques more efficiently on LAC. We therefore isolated phage clones from these single plaques and tested for their activity against MRSA. This mutant phage, which we called Evo2, is able to plaque on LAC and MW2 strains with higher efficiency, displaying comparable plaquing to MRSA252 (Figure 2A, Figure S1). In growth curve experiments, we further observed that Evo2 lyses MRSA cultures at lower MOIs compared to ΦStaph1N (Figure S2). Notably, Evo2 exhibits lytic activity against MW2 and LAC even at an MOI of 0.0001, a concentration at which ΦStaph1N does not show any detectable activity against the two strains.

Figure 2.

We sequenced the genome of Evo2 to determine the genetic mechanism driving this enhanced activity. We observed a single point mutation in ORF141 that induces a premature stop codon (Figure S4). Sequence analysis with HHpred predicts ORF141 to be a putative DNA binding protein with an HTH motif (PDB: 2LVS, E-value: 2.5e-9). We speculate that this protein is a transcriptional regulator that when inactivated by a nonsense mutation, increases ΦStaph1N infectivity. Future studies will center on determining the mechanism of this mutation and why Evo2 only evolved in the LAC strain. Given Evo2’s enhanced activity against MRSA, we then asked how predation by Evo2 impacted β-lactam resistance in our three MRSA strains. We infected MRSA252, MW2, and LAC with Evo2 at an MOI of 0.1 and measured the MICs against β-lactams after 48 hours of passaging. Broadly, similar to the ΦStaph1N infection, infection by Evo2 reduced the MICs of the three MRSA strains against OXA, CEF, and AMX+CA, while MICs against AMX alone and VAN did not change significantly (Figure 2B).

We then tested how different MOIs of Evo2 impacted β-lactam resistance in MRSA (Table S2). As with ΦStaph1N, we infected the three MRSA strains with Evo2 at varying MOIs from 10^-2 to 10^-5 and measured the MIC against oxacillin in the evolved MRSA; each condition was tested in three independent replicates. For most of the MRSA252 cultures, we did not see a significant difference in the fold reduction of MIC between ΦStaph1N and Evo2. Across all three strains, we observed that replicate cultures across different MOIs were unable to recover growth following Evo2 infection (Table S2). However, cultures of MW2 and LAC that did regrow displayed a loss of oxacillin resistance. Thus, overall Evo2 displays a higher infectivity against the three MRSA strains and a greater potency in reducing β-lactam resistance.

Genomic mutations in MRSA strains following phage infection

Following these MIC tests, we hypothesized that the phage-treated MRSA stains evolved mutations that increased their sensitivity to β-lactams. We sequenced the genomes of three clonal isolates (A-C) from each MRSA strain that underwent ΦStaph1N, Evo2, or a mock infection. We observed that each MRSA strain evolved distinct mutation profiles (Figure 3A). Irrespective of the strain and phage treatment, most mutations were predicted to result in substitutions in the protein product, followed by truncations (Figure S5). Cluster of Orthologous Genes (COG) variants associated with transcription, cell wall/membrane/envelope biogenesis, coenzyme transport and metabolism, and defense mechanisms, were the most commonly found categories. Mutations in annotated genes that appeared at least twice across the clonal replicates are summarized in Table 1. Information on all detected genetic variants is listed in Supplementary Table 1.

Figure 3.

Table 1.

In all MRSA strains, we found mutations that could account for the loss in β-lactam resistance. In MRSA252, both ΦStaph1N and Evo2 infection were selected for frameshift or nonsense mutations in the femA gene that would inactive the protein product. FemA is required for the synthesis of the pentaglycine branch on S. aureus Lipid II, the peptidoglycan precursor (Table 1). Deletions of femA have been shown to increase susceptibility to β-lactams even when PBP2a (encoded by mecA) is expressed.^37,38 These results thus provide a potential mechanism for how some MRSA252 cells lose β-lactam resistance after phage selection. At the same time, we noted the presence of other uncharacterized mutations in phage-treated MRSA252. For example, clone A of ΦStaph1N-treated cells carried 2 mutations: a frameshift in femA and a substitution mutation in an uncharacterized protein; meanwhile, clone B displayed a substitution mutation in pfkA, a predicted ATP-dependent 6-phosphofructokinase and mutation in an intergenic region; clone C showed a substitution mutation in a putative transport protein, called yueF (Figure 3B, Supplementary Table 1). The role of these mutations in mediating phage resistance or β-lactam sensitivity, if any, remains unknown.

In MW2, we found mutations in two transcriptional regulators, mgrA and sarA (Figure 3A, Table 1). Both mgrA and sarA belong to the family of MarR (multiple antibiotic resistance regulator)/SarA (staphylococcal accessory regulator A) proteins, which regulate drug resistance and virulence in S. aureus.^25–27 In ΦStaph1N-treated MW2, only mgrA was mutated, while in Evo2-treated MW2, clones also showed nonsense mutations in sarA. We also found mutations in metK and ytqA which are both predicted to be associated with S-adenosylmethione: metK synthesizes SAM, while ytqA belongs to the radical SAM enzyme family and is predicted to be involved in tRNA modification.^35,40 Phage-treated MW2 also displayed mutations in other genes, including tcaB and murE. Notably, each clonal replicate had multiple mutations in the genome, while by contrast untreated MW2 cells only displayed a deletion in an intergenic region that is not present in any of the phage-treated samples.

In the LAC strain, we observed a third, distinct mutational pattern (Figure 3A). Of note, we found nonsense mutations in arlR, which is part of the arlRS two-component signaling system (Table 1). The activity of ArlRS has been implicated in S. aureus virulence, pathogenicity, and oxacillin resistance.^30,31 Further, we observed substitution mutations in spoVG, which is a transcription factor regulating the expression genes involved in a variety of functions, including cell wall metabolism (Table 1).³² Indeed, SpoVG activates the expression of femA.³³ Studies have shown that SpoVG modulates β-lactam antibiotic resistance by modulating cell wall synthesis.³³ Similar to the other two MRSA strains, phage-treated clones of LAC showed multiple mutations in their genomes, although their association with β-lactam sensitivity remains unclear. Finally, we observed mutations in prsA and bioA that appeared in both the mock and phage treatment conditions, suggesting that these mutations do not arise due to phage selection.

Phage-treated MRSA strains display reduced virulence phenotypes

S. aureus strains encode a wide variety of virulence factors that facilitate bacterial infection, and previous studies have used MW2 and LAC strains as models of MRSA virulence.^27,28 In MW2 and LAC, we observed the evolution of nonsense mutations in known virulence regulators, such as mgrA, arlR, and sarA. We therefore hypothesized that phage-treated MRSA cells could also display altered virulence phenotypes in addition to reduced β-lactam resistance. We first tested the ability of MRSA strains that survived phage predation for their ability to form biofilms in a Crystal Violet assay. We found that Evo2 infection of MRSA252 resulted in a significant reduction in Crystal Violet absorption compared to the parental strain. However, we show no significant difference in Crystal Violet absorption between parental, mock- and phage-treated MW2 and LAC strains (Figure S6).

Next, we tested whether phage infection could affect the hemolysis of rabbit blood cells. Hemolysis is mediated by the secretion of toxins, such as alpha toxins, and plays an important role in MRSA infection.⁴³ Expression of these toxins is regulated by virulence pathways that comprise numerous transcription factors, including mgrA, arlR, and sarA. Parental MW2 and LAC colonies lysed rabbit blood cells on blood agar plates, producing distinct halos of clearance around the bacterial cells. As expected, for untreated LAC, the total area of hemolysis was on average 3-fold larger than that of untreated MW2 (∼210 mm² vs ∼70 mm², respectively); with MRSA252, by contrast, lysis was not detected (Figure 4A). Following treatment with ΦStaph1N, we observed that MW2 and LAC displayed a reduced area of hemolysis by 4 to 5-fold. With Evo2-treated cells, we found that in MW2 the fold reduction was comparable to that of ΦStaph1N-treated cells. However, for Evo2-treated LAC, loss of hemolysis was even more pronounced, with two of the replicates showing no detectable hemolysis.

Figure 2.

S. aureus is also able to bind to fibrinogen, resulting in the agglutination, or clumping, of bacteria. Clumping is thought to have several functions in the context of staphylococcal infections. Clumps are likely to be more resistant to clearance by the immune system, partly because they may be too large to be phagocytosed by neutrophils.²⁸ Similar to the expression of hemolysins, previous work has shown that clumping in MW2 and LAC is regulated by a network of virulence regulators. For MW2 and LAC, we observed that phage-treated cells displayed less clumping than the mock-treated or parental strain. For MW2, ΦStaph1N infection resulted in a modest reduction, while Evo2 infection resulted in a reduction of approximately 3-fold (Figure 4B). In LAC, we found that both ΦStaph1N and Evo2 treatment resulted in comparable reductions of clumping in surviving cells.

Discussion

Here we report the discovery that infection by the myovirus ΦStaph1N and its derivative Evo2 can drive MRSA populations to evolve favorable genetic trade-offs between phage and β-lactam resistance. MRSA strains that survived phage infection were resistant to phage; however, they displayed up to a 1000-fold decrease in their MICs against β-lactam antibiotics. Sequencing revealed that each of the three MRSA strains tested evolved distinct mutational profiles following phage infection. Of these mutations, a significant portion of the affected genes are involved in transcriptional regulation, antibiotic resistance, and cell maintenance. Concurring with the sequencing data, phage-treated MRSA strains displayed attenuated virulence phenotypes in the form of reduced hemolysis and cellular agglutination.

Our system highlights the multitude of evolutionary paths that different bacterial strains can undertake to evolve resistance against phage. Most staphylococcal phages are thought to adsorb to S. aureus by recognizing wall teichoic acid (WTA). Prior work has revealed that phage resistance in S. aureus can arise from the disruption of genes involved in the synthesis and modification of WTA, such as tagO or tarM.⁴⁴ Some MRSA strains also alter cell wall glycosylation through dedicated genes encoded on prophages.¹¹ Additionally, the presence of cell wall-bound Protein A has been shown to decrease phage absorption, likely by masking WTA.⁴⁵ However, our results suggest that the evolution of phage resistance can potentially impact networks of genes in the different MRSA strains, including metabolic genes and transcription factors. For example, both mgrA and arlRS regulate the expression of proteins associated with the cell wall, such as the Giant Staphylococcal Surface Protein (GSSP).²⁸ We hypothesize that mutations in these regulators might perturb the composition of proteins in the cell wall, which in turn can disrupt phage binding or prevent disrupt injection of the phage nucleic acid into the cytoplasm. Further studies are needed to dissect the effect of these gene networks on phage resistance and β-lactam resistance.

Our study also shows some of the challenges associated with determining phage/antibiotic resistance trade-offs. Not only do different bacterial strains evolve distinct mutations, but individual clones can accumulate multiple mutations in their genome. Such complex dynamics have been observed in other coevolving bacteria-phage systems before and can serve as useful models to study evolution. However, they make it difficult to predict broader phenotypic outcomes in the bacteria following phage infection. For example, phage resistance has also been shown to confer so-called “trade-ups,” where a non-selected trait is enhanced (e.g. cross-resistance between phage and antibiotic). In our study, we cannot rule out that our phage-treated MRSA has evolved (or will evolve) trade-ups. Although further research is required to test this possibility, the potential clinical benefit of reduced resistance and virulence in MRSA may outweigh the risks associated with trade-ups.

Despite evolving distinct mutational profiles, the three MRSA strains nonetheless displayed a convergence of phenotypes in the form of enhanced phage resistance, reduced β-lactam resistance, and attenuated virulence. We posit that this convergence is caused by the involvement of the cell wall in all three phenotypic outcomes. Phages must interface with the cell wall to adsorb, β-lactams target proteins associated with the maintenance of the cell wall, and many S. aureus virulence factors are embedded within the cell wall or need to get secreted through it. The integrity and composition of the cell wall can be determined by numerous genes, ranging from single proteins involved in cell wall synthesis, such as femA, to transcriptional regulators that control the expression of cell wall-associated proteins, such as mgrA. Thus, the mutational patterns observed in each MRSA strain might reflect unique genetic solutions that enable the bacteria to adapt to the phage predation, while also maximizing the fitness for that particular strain.

From a biomedical perspective, exploiting phage/antibiotic resistance trade-offs have been proposed as a means to combat resistance in bacterial pathogens.¹⁵ Not only could phage treatments reduce the bacterial load of an infection, but also potentially resensitize the bacterial population to antibiotics that it was previously resistant against. For example, in our study, we observed that the MIC of phage-treated MRSA against oxacillin dropped below 2 µg/mL, which is designated by the FDA as the oxacillin MIC susceptible breakpoint for S. aureus. We speculate that infection by ΦStaph1N, Evo2, or an equivalent phage can facilitate the re-deployment of β-lactams in the treatment against MRSA infections. Additionally, phage-treated MRSA might also lose virulence and become more susceptible to immune clearance, which can further improve treatment outcomes. Future studies will need to examine this hypothesis in animal infection models.

Acknowledgements

MT is supported by the SciMed GRS program at UW-Madison; CYM is supported by start-up funds from the Department of Bacteriology at UW-Madison. We are grateful to Dr. Wilmara Salgado-Pabón for providing us with MRSA strains and helpful suggestions. We thank the Marraffini and Hatoum-Aslan laboratories for providing us with bacteriophage samples. We also thank all members of the Mo lab for their scientific input.

Online Methods

Strains and culture conditions

The bacterial strains used in this study are listed in Table S1. Unless otherwise indicated, all MRSA strains (LAC, MRSA252, MW2) were grown in Brain Heart Infusion (BHI) media at 37 °C with shaking (235 RPM).

Plate-based plaque assay

Bacterial lawns were prepared by mixing 100 µL of an overnight culture with 5 mL of melted BHI agarose (top agar). The bacteria and top agar mixture were poured onto a solid BHI plate. The plate was dried for 10 minutes. 10-fold serial dilutions (10⁰-10^-7 unless otherwise noted) of phage were then spotted on the bacterial lawn. Plates were then incubated at 37 °C for 16 hours. Phage titer in plaque-forming units per µL (pfu/µL) was then calculated.

Phage infection assay

MRSA strains were plated onto BHI agar plates and grown overnight. Individual colonies from the parental strains (also referred to as P0) were picked. Single colonies were inoculated in a round bottle tube containing 5 mL BHI broth. The cultures were incubated at 37 °C, 235 RPM for 24 hours. The grown P0 cultures were then diluted 1:100 into fresh 5 mL BHI broth. At the early log phase (OD ∼0.3), the bacterial cultures were treated with phage at an MOI of 0.1, unless indicated otherwise. The treated bacterial cultures were incubated at 37 °C with shaking (235 RPM) for 24 hours. The cultures were then passaged 1:100 into fresh 5 mL BHI broth. This passage was then grown at 37 °C, 235 RPM, for another 24 hours. Surviving cultures were then used for both phenotypic assays and sequencing experiments. As a negative control, MRSA strains were passaged using the steps described above without phage treatment (mock).

MIC assay

Bacterial lawns were normalized to contain 1×10⁸ CFU/mL bacteria mixed with top agar for a total volume of 5 mL. The bacteria and top agar mixture is poured onto a solid BHI plate. The plate was dried for 10 minutes. MIC with increasing concentrations of antibiotics were placed on the semi-dried bacterial lawn and allowed to dry for 10 minutes. The plates are then incubated at 37 °C overnight. For analysis, the plates were imaged, and the MIC of the bacteria was determined. The MIC is determined at the edge of the inhibition ellipse intersects the side of the strip.

Rabbit blood hemolysis

Phage-treated or mock-treated cultures were diluted to an OD600 of 0.1, 5 µL of this dilution was spotted on rabbit blood TSB agar plates and incubated at 37 °C for 24 hours. The area of clearance was determined by the following formula: [π (diameter of clearance/2)²] - [π(diameter of bacterial spot/2)²]

Clumping assay

Clumping assays were performed as described previously.⁴⁶ In short, overnight cultures were diluted 1:100 and incubated at 37 ℃ until the cultures reached an OD₆₀₀ of 1.5. At this point, 1.5 mL of culture was washed two times and resuspended with PBS. Lyophilized human plasma was added for a final concentration of 1.25%. Resuspended cells were left to sit statically at room temperature. 100uL were taken from the top of the cell suspensions in 30-minute intervals and the OD₆₀₀ was measured.

Biofilm assay

Biofilm assay was performed using the crystal violet method as outlined.⁴⁷ In brief, overnight cultures grown in BHI at 37 °C were back diluted 1:10 into a 96-well bottomed microwell plate. The plates were incubated at 37 °C overnight. The contents in the plate were discarded and washed with PBS. Biofilm fixation was done with the addition of sodium acetate (2%). Crystal violet (0.1%) was used for staining followed by a final wash with PBS. Absorbance at 600 nm was read using a spectrophotometer.

DNA sequencing and genome assembly

Following published protocols, genomic DNA from bacteria and phage was isolated using phenol-chloroform extraction. Purified DNA was sent to Plasmidsaurus and SeqCenter for Nanopore and Illumina sequencing, respectively. Reference genomes for bacterial strains were assembled using Flye v2.9.3 with default settings for long-reads. This resulted in 1 singular contig assemblies for 252 (2902592bp, 125x) and MW2 (2820460, 600x), and 3 contigs for LAC (2907712, 645x). Phage assemblies for Evo2 and ΦStaph1N were done with the SPADES assembler v3.15.5.

Open-reading frames (ORFs) were called on the assembled bacterial genomes using Prodigal v.2.6.3,⁴⁸ resulting in a gene-feature file (GFF), and translated genes as .faa and .fna formats. We used BLASTp (Accessed June 4th, 2024), against the protein BLAST database swissprot_2023-06-28, with an expectation value cutoff of 0.001. The top hit for each ORF was used as the final functional annotation.

Mutation identification

A total of 27 genomic samples were collected. DNA was extracted and sent for long-read sequencing using Oxford Nanopore Technology (Plasmidsaurus, San Francisco, USA). Reads were filtered using filtlong v0.2.1 using default settings (https://github.com/rrwick/Filtlong) and with the -p flag 95 (keeping 95% of the best reads). Quality-filtered long reads were mapped against the respective genomes using minimap2 2.22-r1101⁴⁹ resulting in 1 alignment file output per sample (.sam file). The read mapping software Minimap2⁴⁹ was selected because of its suitability to map long reads. Samtools v1.20⁵⁰ was used to convert the .sam files into .bam files, sort the bam file, index the bam files, and generate a coverage table for each position along the alignment.

Reference files (fasta and GFF files) and the 53 “sorted.bam” alignments files were imported into Geneious. Variant calling was performed using Geneious Prime Geneious Prime ® 2024.0.2 (www.geneious.com), using the custom settings: 10% coverage, minimum 95% variant frequency threshold, and the option for “Analyze effect of variants on translation” checked. Variant results and genome annotations table files were exported as a tab-separated-table, and visualized using R v4.4.0, mostly with the tidyverse package. Sequencing data processing, quality filtering, and mapping were performed at the Center for High-Throughput Computing (https://chtc.cs.wisc.edu/). BLAStp was performed on usegalaxy.eu (Accessed June 4^th, 2024).

Supplemental data

Figure S1.

Figure S2.

Figure S3.

Figure S4.

Figure S5.

Figure S6.