CRISPR-Cas9 allows precise gene editing of a virus that kills many types of Staphylococcus bacteria

We established a two-step protocol that includes HR and subsequent CS, as illustrated in Fig. 1A. In order to circumvent the pervasive restriction barriers often found in many S. aureus isolates, we opted to use S. aureus RN4220. This laboratory strain, obtained through extensive chemical mutagenesis of strain 8325-4, has had its resident prophages and restriction-modification systems removed. This modification allows for the transformation of RN4220 with E. coli-derived plasmid DNA (51).

Phage K, characterized by its broad host range, was selected as our engineering scaffold. Its host range primarily includes S. aureus and extends to other Staphylococcus species (52), making it a versatile candidate for both diagnostic and therapeutic purposes. This potential is evidenced by successful applications of Twortvirinae phages in therapeutic contexts, which have demonstrated their efficacy against a variety of clinically relevant bacterial strains (53–60).

As a proof of concept for our engineering pipeline, we constructed a bioluminescent reporter phage, K::nluc, by inserting the nanoluciferase (nluc) gene into the genome of phage K. Utilizing existing data on genome structure and transcriptomic profiles, we identified a region associated with the major capsid protein (cps) as a favorable locus for payload insertion. This region has been shown to be highly expressed in phage K (61) and has also served as a common integration site for payload insertions in other phages (21, 31, 42). To mediate expression from a strong, endogenous promoter, we integrated nluc—along with a ribosomal binding site—immediately downstream of the cps coding sequence. Nanoluciferase (NLuc), a bioluminescent luciferase derived from Oplophorus gracilirostris (NanoLuc, Promega), has been used successfully in prior phage engineering studies and enables sensitive detection due to its small size (516 bp) and strong luminescent output (42, 62).

The first step of the pipeline involves propagating phage K in the presence of the homology donor plasmid (pEDIT_nluc). This plasmid carries the payload gene, along with up- and downstream flanking homology arms, which guide the sequence-specific integration of the payload.

The lysate of phage K, produced on RN4220 pEDIT_nluc, comprises a heterogeneous population, predominantly of wild-type phages, interspersed with a minority of recombinants. This mixed lysate was subsequently propagated on another RN4220 host, one equipped with a CS plasmid (pSELECT_CPS). This plasmid encodes an episomal CRISPR-Cas system, serving as the principal mechanism of selection to isolate the desired recombinant phages.

The pSELECT_CPS plasmid encodes the S. pyogenes cas9 gene, along with tracrRNA and crRNA elements. The crRNA element carries a repeat-spacer1-repeat-spacer2-repeat (RS1RS2R) sequence, with each spacer targeting one homology arm flanking the nluc integration site. We designed the protospacer-adjacent motifs (PAMs) in the editing template (pEDIT_nluc) to contain silent PAM mutations. This design enables K::nluc replication in the presence of pSELECT_CPS (Fig. 1B). Infection of RN4220 pSELECT_CPS with wild-type phage K resulted in complete phage restriction (Fig. 1C). When the same host strain was infected with a lysate containing a mixed population of recombinant and wild-type phages following HR, escape mutants were obtained at a frequency of approximately 10−4 (Fig. 1C). We then performed PCR amplification of the insertion site and found that three out of four plaques we picked yielded PCR products of the expected size (1,326 bp, Fig. 1D), indicating successful payload integration. Further confirmation was obtained through Sanger sequencing of the PCR products with the correct band size, which revealed a correct insertion of the payload at the intended locus.

To assess whether reporter integration affected plaque morphology, we quantitatively compared plaque sizes of K::nluc and wild-type phage K on PSK using ImageJ-based analysis of plaque assay plates. This revealed a small but statistically significant reduction in plaque diameter for K::nluc (Fig. 1E). We then proceeded to assess the infectivity of K::nluc across a diverse set of 71 different S. aureus strains using efficiency of plaquing (EOP) assays (Fig. S1). This set included the phage propagation host PSK, common laboratory strains, and clinical isolates with varying degrees of vancomycin resistance. Out of 71 strains tested, 51 showed plaque formation (limit of detection: 100 PFU/mL) when treated with K::nluc, indicating successful infection by the phage. This assessment is crucial as it provides insight into the potential range of application for K::nluc in detecting different S. aureus strains.

Next, we established the minimum inhibitory concentration (MIC) of vancomycin for all strains, utilizing the cutoffs as defined by reference 63: vancomycin-susceptible (VSSA) strains exhibited an MIC ≤ 2 µg mL⁻¹, vancomycin intermediate-resistant (VISA) strains showed an MIC between 4 and 8 µg mL⁻¹, and vancomycin-resistant (VRSA) strains had an MIC ≥ 16 µg mL⁻¹ (Fig. S2). We selected one phage K-susceptible isolate from each category (PSK [VSSA], LI6 [VISA], and VRSA7 [VRSA]), and investigated the kinetics of phage K::nluc-induced bioluminescence emission.

In all three K::nluc infections (PSK, LI6, and VRSA7), we observed similar signal intensities and kinetics of bioluminescence generation. There was a swift and consistent increase until it reached a plateau, with a peak fold-change of approximately 1 × 10⁶ relative light units (RLU) above the background luminescence, achieved roughly after 3 h (Fig. 2A).

To further assess the sensitivity of our reporter phage system, we examined the minimum dose response of bioluminescence emitted by PSK, LI6, and VRSA7 upon infection with a fixed concentration of K::nluc (5 × 10⁷ PFU/mL) (Fig. 2B). Although the minimum dose response of LI6 (2,617 CFU/mL) was higher than that of PSK (1,093 CFU/mL) and VRSA7 (707 CFU/mL), all values fall within the same order of magnitude. These results are in line with detection thresholds reported for other reporter phage systems (42, 62, 64).

Subsequently, we quantified the range of bioluminescence detection of K::nluc across all 71 S. aureus strains and compared it with the previously determined EOP (Fig. 2C). We successfully detected bioluminescence above background levels in all tested strains (71/71), encompassing even those 20 strains that had previously exhibited no evident plaque formation. To further assess the specificity of K::nluc, we also tested its activity against 17 non-Staphylococcus species and observed no bioluminescent signal indicative of active infection (Fig. S4).

Staphylococcus bacteremia accounts for a significant proportion of bloodstream infections (65, 66). Considering the complexity of blood as a biological matrix compared to standard laboratory culture conditions, we tested our reporter phage's functionality in whole human blood. It is essential to evaluate this, as the in vitro infection kinetics of K::nluc with pathogenic bacterial strains might differ in more complex biological settings (67–69). We optimized the assay conditions using whole human blood spiked with the PSK strain, as depicted in Fig. S3. The most rapid and robust signal response was achieved under the following conditions: the spiked blood was diluted fivefold in growth medium, and the cells were incubated at 37°C for 1 h to stimulate host cell metabolism before reporter phage infection. We observed higher bioluminescent signals when using a citrate-based anticoagulant (Na3-citrate, citric acid, glucose, and potassium sorbate) compared to Li-Heparin, both of which are common storage solutions in clinical practice.

Leveraging these optimized conditions, we determined the detection limits using the previously selected S. aureus strains PSK, LI6, and VRSA7 (Fig. 3A). These strains were detectable at concentrations as low as 2,151 CFU/mL (PSK), 136 CFU/mL (LI6), and 1,270 CFU/mL (VRSA7). While there are minor differences in the detection thresholds across media, these values remain within 1 order of magnitude of those obtained in growth medium (Fig. 2B) and demonstrate robust detection in a clinically relevant matrix.

Bovine raw milk, another complex matrix, has also been reported to negatively impact bacteriophage proliferation (70–72). To investigate this, we conducted a similar experiment, replacing blood with unpasteurized, raw bovine milk. Contrary to expectations, treating bovine raw milk infected with S. aureus strains PSK, LI6, and VRSA7 with K::nluc resulted in bioluminescence detectable at minimal dose responses of 55 CFU/mL (PSK), 191 CFU/mL (LI6), and 514 CFU/mL (VRSA7). These results indicate that our assay performs reliably even in a matrix previously shown to inhibit phage activity, suggesting that the optimized protocol is robust to environmental complexity.

The MIC of the 71 S. aureus strains used in this study was determined under the standard culture conditions as outlined in reference 63. Briefly, 96-well plates were prepared, each well containing 250 µL of Miller-Hinton broth supplemented with a range of vancomycin concentrations (from 0.0625 to 64 µg/mL, with a twofold increase between subsequent concentrations). Each well was inoculated with 2 µL of 1:1,000 diluted bacterial cultures (grown overnight at 37°C) for each of the 71 strains. After an incubation period of 18 h at 37°C, turbidity was measured photometrically (OD₆₀₀). Wells with an OD₆₀₀ > 0.1 were considered turbid. The strains were then classified as VSSA (MIC ≤ 2 µg/mL), VISA (MIC = 4 to 8 µg/mL), or VRSA (MIC ≥ 16 µg/mL) based on the MIC values.

S. aureus PSK (ATCC 19685) was used as propagation and S. aureus RN4220 (DSM 26309) as the engineering host of K and K::nluc. E. coli XL1-blue MRF' (Stratagene) was used for plasmid amplification prior to RN4220 transformation. RN4220 and XL1-blue cultures were grown overnight (O/N) at 37°C in brain heart infusion (BHI) broth (Biolife Italiana) and Luria-Bertani/lysogeny broth (LB) medium (3 M sodium chloride, 10 g/L tryptone, and 5 g/L yeast extract, pH 7.2), respectively. The selection of 71 laboratory strains and clinical isolates (Fig. S2) were grown on BHI Broth with corresponding antibiotic supplements.

Phage K was propagated on S. aureus PSK using the soft-agar overlay method with BHI as bottom agar and LC agar (LB supplemented with 10 mM CaCl₂, 10 mM MgSO₄, and 10 g/L glucose) as top agar. Overlays were incubated O/N at 37°C and phages extracted from plates with semi-confluent lysis using 5 mL SM buffer (4°C, 2 h, constant agitation). Lysates were sterile-filtered (0.22 µm pores). Phage particles were precipitated O/N at 4°C using polyethylene glycol (7% PEG8000 and 1 M NaCl) and purified using stepped CsCl gradient ultracentrifugation. The obtained phage suspension was dialyzed twice against 1,000× excess of SM buffer. Purified samples were then stored long-term at 4°C. The titer was determined using the soft-agar overlay method.

XL1-blue electrocompetent cells were electroporated at 2.5 kV, 200 Ω, 25 µF, incubated for 1 h with SOC recovery medium ( 2% [wt/vol] Tryptone, 0.5% [wt/vol] yeast extract, 10 mM NaCl, 2.5 mM KCl, 10 mM MgCl₂, and 20 mM glucose) at 37°C, and plated on selective agar to isolate successful transformants. RN4220 electrocompetent cells were electroporated at 1.8 kV, 600 Ω, 10 µF with500 to 1,000 ng of amplified plasmid DNA, incubated for 1 h at 37°C in B2 recovery medium (10 g/L casein hydrolysate, 5 g/L d-glucose, 1 g/L potassium phosphate dibasic, 25 g/L NaCl, and 25 g/L yeast extract) and plated on selective medium to isolate successful transformants.

The plasmid pLEB579 (kindly gifted by T. Takala, University of Helsinki, Finland) is a shuttle vector shown to have high transformation efficiencies in both E. coli and S. aureus and was therefore used as a backbone for both the editing template (pEDIT_nluc) and the CRISPR-Cas9-counterselection system (pSELECT_CPS). We used a previously reported, Streptococcus pyogenes-derived Cas9 (SpyCas9)-based CRISPR system (21) and exchanged the two spacers in the repeat-spacer1-repeat-spacer2-repeat (RS1RS2R) region. This was done to allow targeted restriction of wildtype phage K at two distinct loci (8,304 and 148 bp up- and downstream of the intended insertion site, respectively) designed to contain a PAM-disrupting synonymous mutation in the successful recombinants. pEDIT_nluc was constructed by integrating an nluc gene (optimized for S. aureus codon usage, avoidance of Rho-independent termination, and an added upstream ribosome binding site: GAGGAGGTAAATATAT), flanked by 400 bp (upstream) and 300 bp (downstream) homology arms, corresponding to the intended K insertion site, into the linearized pLEB579 backbone. Two silent point mutations were included to disrupt the two PAMs adjacent to the target DNA (Fig. 1B). All synthetic sequences were acquired as GeneArt String DNA Fragments (Thermo Fisher), albeit the RS1RS2R sequences, which were ordered as GeneArt Gene Synthesis (Thermo Fisher) (Table S2). Assembly of all constructs was performed using isothermal Gibson Assembly (22) (NEBuilder HiFi DNA Assembly Master Mix) and subsequently transformed into E. coli XL1-blue MRF' for plasmid amplification. Plasmids and primers are listed in (Table S1).

pEDIT_nluc and pSELECT_CPS were transformed into S. aureus RN4220 to acquire the strains required for recombination and counterselection, respectively. RN4220 pEDIT_nluc was infected with K via soft-agar overlay and a high titer lysate obtained as described previously (see Bacteriophage propagation section). Plates showing semi-confluent lysis were washed with SM buffer and 10-fold dilutions of the resulting lysate were used to perform soft-agar overlays on RN4220 pSELECT_CPS. Individual plaques were picked from the plates showing the fewest (non-zero) plaques, resuspended in SM buffer, and clonally isolated by three rounds of plaque-purification. PCR amplification using primers flanking the insert site (Table S1) was performed and products showing a size indicative of the intended insertion were purified and Sanger sequenced (Microsynth AG, Balgach, Switzerland) to validate the correct genomic sequence. Validated phage lysates were purified using ultracentrifugation on a cesium-chloride gradient and subsequent dialysis as described in Bacteriophage propagation section.

BHI soft agar (5 mL) was melted and cooled to 47°C. The molten soft-agar was inoculated with 200 µL bacterial culture of adequate turbidity (OD₆₀₀> 1) and 10 µL of phage suspension, briefly mixed by agitation, and spread evenly onto BHI agar plates with. Plates were let dry for 15 min at room temperature (RT) and subsequently inverted and incubated at 37°C for 12–18 h.

BHI soft agar (5 mL) was melted and cooled to 47°C. The molten soft-agar was inoculated with 200 µL of bacterial culture, briefly mixed by vortexing, and spread evenly on BHI agar plates. Plates were dried at RT for 15 min; droplets of phage suspension were then placed carefully on the dried soft agar. Plates were dried for 15 min, inverted, and incubated at 37°C for 16 h.

EOP of bacteriophage suspensions was determined by performing spot-on-lawn assays on bacterial strains of interest. Tenfold dilutions of the phage suspension were prepared up to a maximum dilution of 1 × 10⁻⁸. Spot-on-lawn assays were performed as described in the previous section using the series of bacteriophage dilutions and O/N cultures of the corresponding host strains. The EOP is given by the number of plaque-forming units (PFU) after infection of a specific strain with K::nluc. Values are relative to the host with the highest measurement (LS20 for bioluminescence, LS14 for EOP).

To assess whether reporter gene insertion affected phage fitness, we compared plaque sizes between wild-type phage K and K::nluc. Both phages were plated at similar titers on three independent S. aureus PSK cultures using 1/2 BHI agar plates and LC soft agar. Plates were incubated overnight at 37°C and scanned at 1,000 dpi. Image analysis was performed in Fiji/ImageJ using a custom macro. All images were processed with identical brightness settings and subjected to automatic thresholding. Plaques with an area larger than 10 pixels and circularity above 0.75 were included. Manual cleanup was used to remove overlapping or irregular plaques. Mean plaque areas were compared using an unpaired two-tailed t-test in GraphPad Prism.

Stationary phase bacterial cultures were diluted to OD₆₀₀ = 0.01, inoculated with 5 × 10⁷ PFU/mL K::nluc, and incubated at 37°C (180 rpm agitation). Bioluminescence measurements were taken by combining 25 µL of the sample solution with an equal volume of prepared buffer-reconstituted nluc substrate as detailed by the manufacturer (NanoGlo Luciferase Assay System; Promega). Measurements were taken every 20 min (225 min total) in Nunc F96 MicroWell 446 plates using a GloMax navigator luminometer (Promega) with 5 s integration time and 2 s delay. To account for background luminescence from residual Nanoluciferase in the purified phage preparation, all K::nluc infection measurements were first background corrected by subtracting the signal obtained from a phage-only control (K::nluc in BHI medium). To calculate fold-change values, background-corrected RLU values were subsequently divided by the corresponding RLU values obtained from a wild-type phage K control infection on the same strain. This approach yields a background-corrected fold-change of RLU.

To determine the minimum concentration of cells giving a significant bioluminescence signal above the background, stationary phase bacterial cultures were diluted to a range of concentrations in ten-fold increments (OD₆₀₀ = 10⁻¹ to 10⁻¹⁰). Measurements were taken 3 h post-infection (5 × 10⁷ PFU/mL). The minimum concentration of cells that could be detected was established as the smallest measurement exceeding a signal threshold, defined as the mean plus three times the standard deviation of the background signal. The theoretical minimal concentration was ascertained at the point where the established signal threshold intersects with a linear regression of the data, considering only RLU values surpassing the threshold.

Stationary phase bacterial cultures were diluted to OD₆₀₀ = 0.01, inoculated with 5 × 10⁷ PFU/mL K::nluc, and incubated at 37°C with 180 rpm agitation. Bioluminescence was measured at 3 h post-infection and background corrected by subtracting the signal obtained from K::nluc incubated in BHI alone. For each strain, the mean RLU value was determined from three biological replicates. Background-corrected RLU values were then normalized to the highest measured response (strain LS20) to allow for comparison across strains.

Minimal dose response of K::nluc infection on one representative each of VSSA, VISA, and VRSA was conducted in triplicate as done with regular growth medium described above, albeit with some modifications. First, spiked whole human blood or bovine raw milk was mixed 1:5 with BHI growth medium and incubated for 1 h at 37°C with agitation (180 rpm) prior to infection with 5 × 10⁷ PFU/mL K::nluc. Bioluminescence was measured after 3 h. The whole human blood samples were stored in anticoagulant solutions containing either 1.89 mg/mL Na3-citrate, 0.69 mg/mL citric acid, 2.1 mg/mL glucose, and 0.03 mg/mL potassium sorbate (BD Vacutainer (REF 367756), Becton, Dickinson and Company, NJ, USA) or Li-Heparin (17 IU/mL) (BD Vacutainer (REF 368886), Becton, Dickinson and Company).

CLC Genomics Workbench version 20.0.4 was used for sequence analyses such as primer and string design as well as evaluation of Sanger sequencing results. Plotting was done using GraphPad Prism version 10.0.0 for Windows. OpenAI's ChatGPT 4 (87) was used as a tool to assist with formatting and editing the manuscript. This involved iterative refinements to ensure clarity and conciseness of the content presented.