Engineered reporter phages for detection of Escherichia coli ... - Nature.com
Abstract
The rapid detection and species-level differentiation of bacterial pathogens facilitates antibiotic stewardship and improves disease management. Here, we develop a rapid bacteriophage-based diagnostic assay to detect the most prevalent pathogens causing urinary tract infections: Escherichia coli, Enterococcus spp., and Klebsiella spp. For each uropathogen, two virulent phages were genetically engineered to express a nanoluciferase reporter gene upon host infection. Using 206 patient urine samples, reporter phage-induced bioluminescence was quantified to identify bacteriuria and the assay was benchmarked against conventional urinalysis. Overall, E. coli, Enterococcus spp., and Klebsiella spp. were each detected with high sensitivity (68%, 78%, 87%), specificity (99%, 99%, 99%), and accuracy (90%, 94%, 98%) at a resolution of ≥103 CFU/ml within 5 h. We further demonstrate how bioluminescence in urine can be used to predict phage antibacterial activity, demonstrating the future potential of reporter phages as companion diagnostics that guide patient-phage matching prior to therapeutic phage application.
Introduction
Urinary tract infections (UTIs) are among the most common community-acquired and healthcare-associated bacterial infections in all age groups and represent a major public health problem worldwide1. The societal and economic burden of UTIs is substantial, as evidenced by its global prevalence of around 150 million cases per year and an estimated direct healthcare expenditure of over 6 billion dollars1. Furthermore, UTIs aggravate the antimicrobial resistance crisis as they constitute a leading cause of antibiotic prescription, second only to respiratory tract infections2. While uncomplicated UTIs generally present minimal morbidity, the presence of certain predisposing medical conditions (e.g., metabolic disorders, immunosuppression, and obstructive uropathy) can lead to the development of life-threatening complications such as pyelonephritis and urosepsis3,4. The etiology of UTI is complex, often polymicrobial, and can involve a plethora of bacterial species and certain fungi. Gram-negative bacteria, predominantly from the Enterobacteriaceae family, including Escherichia coli and Klebsiella spp., cause the majority of UTI cases, closely followed by Enterococcus faecalis as the major Gram-positive uropathogen5,6. Due to this diverse etiology, rapid and accurate pathogen identification is of paramount importance to guarantee a timely and effective therapy and to limit the preventive prescription of broad-range antibiotics.
The gold standard for diagnosing UTIs is the presence of symptoms combined with a positive urine culture7. Diagnostic accuracy can be further improved by integrating mass spectrometry identification and/or 16s rRNA sequencing, when required7. Despite providing quantitative results with species-level resolution, culture-based assays require a minimum of 18–30 h to return clinically actionable results7. Long turnaround times propel the empirical prescription of antibiotics, which, in turn, aggravates the global spread of antimicrobial resistance. To tackle this issue, multiple immunological and molecular diagnostic techniques have been developed to enable uropathogen detection in point-of-care (POC) settings7. However, they either present poor specificity and/or sensitivity (e.g., urine dipsticks), require considerable laboratory resources (e.g., flow cytometry), or fail to discriminate between dead and viable cells (e.g., multiplex PCR)7. Most of these methods cannot identify the causative agent and some are not suitable for Gram-positive pathogen detection (e.g., dipstick nitrite test). Therefore, an alternative POC diagnostic, which combines the desired 4 S properties (simplicity, speed, sensitivity, and specificity), could provide substantial value to UTI management beyond what is currently available.
Bacteriophages (phages) are viral predators of bacteria that bind and infect their host cells with genus-, species-, or strain-level specificity. The clinical use of phages as pathogen-specific antimicrobials (phage therapy) is a promising strategy to tackle multi drug-resistant (MDR) infections8 and to manage microbial dysbiosis9. By leveraging their specific interaction with bacterial cells, phages have also been used to develop an array of diagnostic tools for rapid and accurate pathogen identification (reviewed elsewhere10). For example, phages can be genetically engineered to encode a heterologous reporter gene—a concept known as "reporter phage". Upon infection of the host bacterium, phage-driven reporter gene expression is detected through enzymatic substrate conversion, indicating the presence of viable host cells. Nanoluciferase (NLuc)—an engineered luciferase derived from deep-sea shrimp—is one of the most promising genes for reporter phage engineering. NLuc is a small and highly stable enzyme that produces strong, ATP-independent, glow-type luminescence upon substrate addition11. NLuc reporter phages can reliably detect bacterial contaminations at very low concentrations12,13,14.
In this study, we isolate suitable phage backbones, develop the required engineering technology to construct reporter phages, and devise a simple and accurate NLuc reporter phage-based POC diagnostic tool. Compared to the gold standard, this assay demonstrates rapid and specific detection of the three most prevalent uropathogens from 206 fresh clinical urine specimens. Besides supporting clinical decision-making, we use patient isolates to demonstrate the potential application of reporter phages as companion diagnostics to identify responder patients that could benefit from personalized therapeutic phage application.
Results
Isolation and characterization of phages against E. coli, Klebsiella, and Enterococcus
To validate UTI etiology, we collected 663 uropathogen isolates from 442 incidents of either UTI (227 incidents) or asymptomatic bacteriuria (215 incidents) in Zurich, Switzerland (the Zurich Uropathogen Collection Figs. 1a, S1, Table S1). We identified 30 different bacterial species, with E. coli (31%), Enterococcus spp. (20%), and Klebsiella spp. (18%) as the three most prevalent uropathogens, which is consistent with previous etiological studies on UTIs5. Strictly lytic phages against each of the top three uropathogens were isolated from different wastewater sources across Switzerland (Fig. 1a) by sequential phage propagation on multiple strains of the respective host species, which provided selective pressure towards isolating phages with a broad host range15. To ensure activity on uropathogenic strains, sequential isolation was performed using UTI isolates from the Zurich Uropathogen Collection. Prior to performing plaque assays, a short enrichment step in synthetic human urine medium (SHU, pH 6) was included for each host to select for phage activity in high urea-containing environments16.
a Workflow depicts the sequential multiple host method for the isolation of phages with broad host-range. Wastewater samples from different wastewater treatment plants across Switzerland were mixed with the first clinical isolate (Host A) of the respective target species selected from the Zurich Uropathogen Collection and briefly enriched in synthetic human urine medium (SHU) before performing plaque assays. Emerging plaques were pooled, extracted, and propagated on 4 subsequent hosts (B, P, C, D) before individual phages were purified on a propagation Host P. b Summary of the genomic and morphological characterization of six selected phage scaffolds infecting E. coli (phages E2 and E4), Enterococcus spp. (phages EfS3 and EfS7) and Klebsiella spp. (phages K1 and K4), respectively. Some elements of the figure were created using biorender.com. Source data are provided as a Source Data file.
By applying this stringent selection process, we isolated phages E2 and E4 (targeting E. coli), K1 and K4 (targeting Klebsiella spp.), and EfS3 and EfS7 (targeting Enterococcus spp.). All phages were de novo sequenced, visualized by electron microscopy, and their plaquing host ranges were determined on more than 50 urological isolates. A summary of phage characteristics is presented in Fig. 1b. E2, E4, K1, and K4 are members of the Tevenvirinae subfamily within the Straboviridae family. Accordingly, these phages feature circularly permuted dsDNA genomes of about 170 kbp that are injected through a characteristic contractile tail. Phage EfS3 is a member of the Herelleviridae family with a contractile tail and a genome size of 152,305 bp, while EfS7 has an elongated head and a non-contractile tail and belongs to the Saphexavirus genus with a 56,144 bp unit genome. The combined plaquing host range for each phage pair was broad, covering 71% (E. coli), 60% (K. pneumoniae and K. oxytoca), and 77% (E. faecalis) of the tested urological isolates (see in vitro assessment).
Engineering of nanoluciferase reporter phages
Due to their broad host ranges, these six phages were used as scaffolds to engineer NLuc reporter phages for E. coli, Klebsiella spp., and Enterococcus spp. detection. The nluc sequence was integrated alongside a ribosomal binding site (RBS) downstream of phage structural genes (Fig. 2a). Because the structural gene cassette is strongly expressed from endogenous phage promoters, this strategy confers a high level of reporter gene expression. Homologous recombination-based and CRISPR-Cas9-assisted phage engineering has previously been achieved for Tevenvirinae17,18,19 and was adopted for the construction of phages E2::nluc, E4::nluc, K1::nluc, and K4::nluc (see methods section for more detail). In contrast, no established technology was available for genetic engineering of virulent Enterococcus phages, and we therefore developed a two-step protocol as shown in Fig. 2b–d (workflow shown in Fig. S2a).
a Schematic representation of the nanoluciferase gene (nluc) insertion sites in the genomes of all reporter phages. For E. coli and Enterococcus phages, codon optimized nluc (516 bp + RBS) was incorporated immediately downstream of the major capsid gene (cps), whereas for Klebsiella phages, nluc was inserted downstream of the prohead assembly protein (gp167 for K1 and gp168 for K4). b–d Engineering method for virulent Enterococcus phages combining homologous recombination with CRISPR-Cas9-assisted counterselection. b Left: Schematic representation of the pSelect CRISPR locus used to restrict wildtype phage genomes by targeting two sequences flanking the nluc integration site. Right: Spot-on-the-lawn assays of 10-fold phage dilutions show plaquing efficiency on hosts with (+) or without (scr, -) CRISPR targeting. scr = non-targeting crRNA; - = no pSelect. c Schematic representation of the editing plasmids (pEditEfS3, pEditEfS7) used to incorporate mutated protospacer-adjacent motifs (PAMs) and nluc gene sequences downstream of the major capsid protein (cps) regions of EfS3 and EfS7 by homologous recombination and subsequent CRISPR escape of EfSs::nluc and EfS7::nluc candidates (counterselection) shown by spot-on-the-lawn assays of serial dilutions of phages after recombination. d 10 individual plaques were isolated from one replicate of data shown in Fig. S2b, c. Plaques were validated genetically and functionally using PCRs with primers P19/20 and P21/22 and RLU determination from individual plaques, respectively. HAL = homology arm left; HAR = homology arm right; put. cds = putative coding sequence; mtp = major tail protein; RLU = relative light unit. Source data are provided as a Source Data file.
In the first step, Enterococcus phages were propagated in the presence of an editing template vector (pEdit) containing the nluc gene and flanking regions of homology that direct sequence-specific integration (pEDITEfS3 and pEDITEfS7, respectively). Resulting phage lysates contain a mixed population with few recombinants and a majority of wildtype phages. In the second step, these lysates were subjected to counterselection on a host encoding an episomal CRISPR-Cas system (pSelect) designed to cleave wildtype phage genomes without targeting recombinants, which enables selective enrichment of recombinant phages. The plasmids pSelectEfS3 and pSelectEfS7 contain the S. pyogenes cas9 gene, tracrRNA, and a crRNA element with two spacers (S1 and S2) (Fig. 2b), which target the two regions flanking the nluc insertion site of EfS3 and EfS7, respectively. Recombinant genomes are protected from counterselection through the integration of silent mutations within the protospacer adjacent motifs (PAMs) on pEdit (see Fig. 2c). This CRISPR-Cas9 system completely restricted wildtype EfS3 and EfS7 plaque formation on E. faecalis (Figs. 2b and S2b, c) and enabled rapid identification of recombinant, CRISPR-resistant phage candidates (Fig. 2c, d), all of which featured the expected genotype and bioluminescence (Fig. 2d). EfS3::nluc and EfS7::nluc were plaque purified alongside the E. coli and Klebsiella reporter phages and sequenced to confirm reporter gene integration.
In vitro functional assessment of reporter phages
To guide the development of our reporter phage urine assay, we used in vitro experiments to determine key functional characteristics of all engineered phages, including infection kinetics, detection limits, detection range, and potential cross-reactivities. All phage lysates were purified by polyethylene-glycol precipitation and caesium chloride (CsCl) gradient ultracentrifugation to remove residual NLuc enzyme present within raw phage lysates.
Initial phenotypic characterization of our reporter phages revealed no significant alterations in plaque morphology (Fig. 3a) or host-cell lysis (Fig. 3b) when compared to their parental wildtype counterparts, suggesting that nluc gene insertion is not associated with major fitness costs. To determine the time required to reach a stable endpoint bioluminescence signal, we performed kinetic infection experiments. To this end, individual reporter phages were monitored during infection of their production hosts by measuring bioluminescence in relative light units (RLU) over 4–5 h. As shown in Fig. 3c, all reporter phages presented a steep rise in luminescence intensity almost immediately after the onset of infection with a stable plateau of RLU fold change (FC) above 105 reached between 1.5 and 3 h post infection (p.i.) for E. coli and Klebsiella and a plateau with RLU FC 104 between 2 and 4 h for Enterococcus. While these kinetic profiles are promising, they can exhibit variations depending on the infected strain and on the sample matrix. For example, urine composition and pH can vary significantly between patients20. Hence, to ensure measurement of stably saturated signals in patient urine, all bioluminescence assays were measured at 3 h p.i. for E. coli and Klebsiella or at 4 h p.i. for Enterococcus. To determine the detection limit of our assay, reporter phage infections were performed on serial dilutions of host bacteria. The detection limit was defined as the minimum number of colony-forming units (CFUs) required to produce a signal above the background threshold (i.e., average background RLUs plus three-fold standard deviations (SD)). Without any pre-enrichment of bacteria, detection limits below 100 CFU/ml were observed for all reporter phages (Fig. 3d).
a Plaque phenotypes of wildtype and recombinant reporter phages were compared using the soft-agar overlay method. b Infection kinetics of engineered vs. wildtype phages were compared by quantifying the optical density (OD600) of infected bacterial cultures over time (liquid infection assay). Data are mean ± SD (n = 3). c The kinetics of reporter phage-mediated NLuc signal production were determined using bioluminescence time course assays. Fold-change (FC) RLU equals RLU (sample) divided by RLU (phage only control). Data and mean±standard error of the mean (SEM) from biological triplicates. d Assay sensitivity was determined in vitro for each reporter phage by infecting serial host cell dilutions and quantifying luminescence at 3 h (E. coli and Klebsiella phages) or 4 h (Enterococcus phages) post infection. Detection limits (vertical dotted lines) were calculated as the cell number required to produce a signal that is three standard deviations (3σ) above the background luminescence (phage in medium, indicated as horizontal dotted lines). Datapoints for linear regression are from biological triplicates. All infections were performed on the phage propagation hosts. p.i. post infection; RLU relative light unit. Source data are provided as a Source Data file.
The bioluminescence detection range of our reporter phages was tested against a panel of 154 randomly selected urological isolates (52 E. coli, 52 E. faecalis, and 50 Klebsiella spp.) and compared to phage plaquing ability, as determined by spot-on-the-lawn assays (Fig. 4a). While bioluminescence demonstrates the ability of a phage to inject genetic material into the bacterial cytoplasm and express luciferase, the criterion for plaquing is more stringent, as the phage must complete a full infection cycle and lyse the bacterial cell21. Each individual phage exhibited a broad bioluminescence detection range of 81% (E2::nluc and E4::nluc), 71% (EfS3::nluc), 48% (EfS7::nluc), and 84% (K1::nluc), 80% (K4::nluc). Despite significant overlaps, complementation of the detection range was observed for all phage pairs. Accordingly, the combined range of detection for E. coli, E. faecalis, and Klebsiella spp. could be improved to 83%, 96%, and 90%, respectively. All phages presented significantly broader bioluminescence detection ranges compared to their plaquing ability (Fig. 4b), which could be due to bacterial intracellular defense systems that prevent the completion of a full infection cycle in some strains22.
a Plaquing host ranges and bioluminescence detection ranges were determined in growth medium on 52 urological E. coli and E. faecalis isolates and 50 urological Klebsiella spp. isolates from the Zurich Uropathogen Collection (listed in Table S2). Strains were infected with reporter phages and luminescence quantified at 3 h (E. coli and Klebsiella spp.) or 4 h (Enterococcus spp.) post infection. Heat-map shows fold change RLU (FC-RLU). Threshold (TH): ≥ 100 FC-RLU (E. coli and Klebsiella spp.) and 50 FC-RLU (Enterococcus spp.). Clinical isolates that allow for plaque formation are indicated with a green dot. Plaquing host ranges were determined from single spot on the lawn infection assays and bioluminescence detection ranges were from biological quadruplicates (E. coli and K. pneumoniae) or biological triplicates (E. faecalis). b Summary of individual and combined host- and detection ranges. Each reporter phage was able to transduce bioluminescence into strains that were not permissive to plaque formation, effectively leading to an increased detection range compared to the plaquing host range. c Heat maps showing reporter phage-induced luminescence in human urine spiked with 6 selected strains of each bacterial target species. d, e Dose response assay with Enterococcus phages EfS3::nluc and EfS7::nluc in human urine d or blood e spiked with serial dilutions of the clinical isolate E. faecalis Ef24. After 1/10 dilution of the spiked urine/blood with buffered media and a 1 h enrichment, reactions were mixed with 2 × 107 PFU/ml reporter phages and FC-RLU was determined at 2 h and 4 h p.i. FC RLU equals RLU (sample) divided by RLU (phage only control). All data in c–e is shown as mean (±SEM) from biological triplicates. Individual datapoints are shown as black dots. Source data are provided as a Source Data file.
Low diversity within strain collections can lead to an over- or underestimation of the phage host range. We addressed the clonal diversity of our host range panel (i) by analyzing the signal intensity distributions in response to reporter phage infection, and (ii) by comparing antibiotic resistance profiles within the E. coli strain panel. Luminescence intensities were distributed over several orders of magnitude (Fig. S3f), suggesting that the selected strains differ significantly in phage-susceptibility. Analysis of antibiotic susceptibility patterns from E. coli revealed the presence of at least 24 different resistance profiles in 52 strains (Fig. S3g). Taken together, these data suggest a high level of clonal diversity across tested isolates.
Since bacterial UTIs are of diverse etiology and frequently polymicrobial in nature, it is important to ensure that the reporter phages exhibit high specificity toward their target bacterial species. While false-positives are best assessed in a large patient study (as shown in Fig. 5), an initial cross-reactivity bioluminescence assay was performed in vitro using 67 non-target bacterial isolates, including representative strains from different uropathogenic species. No cross-reactivity was observed for E2::nluc and E4::nluc; however, K1::nluc and K4::nluc did cross-react with one clinical isolate of Raoultella ornithinolytica (Fig. S3d). The genus Raoultella was formerly a cluster within the genus Klebsiella, but was recently reclassified based upon advanced phylogenetic analysis23. Nevertheless, the cross-reactivity observed can only be explained by the Raoultella isolate presenting functional cell surface receptors for phages K1 and K4. The Enterococcus reporter phages EfS3::nluc and EfS7::nluc revealed genus-specific infectivity, however, EfS3::nluc produced bioluminescence (but no plaques) on 2/12 tested S. aureus strains and a single Pediococcus acidilactici isolate (Fig. S3b, c). In addition, we demonstrated that phage EfS3::nluc is able to infect 9 out of 14 tested E. faecium strains, whereas both the plaquing- and detection range of phage EfS7::nluc did not include E. faecium (Fig. S3a). Generally, species-level specificity is less critical for enterococci because E. faecium is very rarely detected in patient urine24.
Workflow of the urine diagnostic assay is shown in a. Fresh patient urine was collected and bacterial metabolic activity activated by a short 1 h enrichment in growth media followed by addition of reporter phages. Luminescence was quantified at 3 h (E. coli, Klebsiella spp.) and 4 h (Enterococcus spp.) post infection. b Luminescence quantification (RLU) was correlated to bacterial CFU/ml counts from differential/selective plating of 206 patient urine samples on UriSelect and KFS agar plates. In parallel, all results were confirmed by the diagnostic laboratory of the Institute of Medical Microbiology (IMM). Urine samples were derived from two different hospitals (Site 1 and Site 2). Thresholds (TH) for luminescence were ≥ 100 FC-RLU (E. coli, green and Klebsiella spp., purple) and 7.5 FC-RLU (Enterococcus spp., blue). The test was considered positive if at least one phage per target pathogen had activity >TH. False negatives (FN, red) are samples with no luminescence but bacterial growth ( ≥ 103 CFU/ml), false positives (FP, yellow) exhibit luminescence but no bacterial growth. c Summary of the test performance showing calculated sensitivity, specificity, accuracy for all target pathogens at both sampling sites. Some elements of the figure were created using biorender.com. Strains from the field evaluation are listed in Table S3. Source data are provided as a Source Data file.
To assess infection and reporter protein expression under relevant conditions, we tested each reporter phage in pooled human urine spiked with six different clinical isolates. Infectivity and bioluminescence generation was maintained in 34/36 patient isolate-phage combinations, suggesting urine-dependent matrix effects do not constitute a major hurdle for assay development (Fig. 4c). Besides UTIs, enterococci are a frequent cause of Gram-positive bacteraemia and sepsis25. Therefore, we tested our Enterococcus reporter phages in whole human blood spiked with E. faecalis strain Ef24 and compared signal intensity and kinetics against a comparable experiment with spiked human urine (Fig. 4d, e). Although bioluminescence signal generation was slightly delayed compared to urine, both EfS3::nluc and EfS7::nluc enabled robust detection of 104 CFU/ml within 4 h in whole blood (Fig. 4e). Enterococcus detection in blood was confirmed with an additional set of five clinical isolates (Fig. S3e).
Reporter phages enable rapid and sensitive detection of E. coli, Klebsiella, and Enterococcus in patient urine
Culture-based detection of 105 CFU/ml in urine is the gold standard (GS) for diagnosis of asymptomatic bacteriuria and UTI, although lower thresholds (103–104 CFU/ml) have been suggested, particularly for catheter-associated UTIs (CAUTIs)24,26. Based on our in vitro reporter phage characterization, we developed a protocol for rapid and direct detection of E. coli, Klebsiella spp., and Enterococcus spp. in patient urine and tested its performance in a large-scale patient evaluation (workflow: Fig. 5a). To this end, 206 fresh urine samples were collected from two university hospitals in Switzerland (site 1 and site 2) over a period of five months and subjected to reporter phage assays. For benchmarking, all samples were independently analyzed by the Institute of Medical Microbiology (IMM, University of Zurich, Switzerland). In brief, this involved culture-based identification combined with mass spectrometry, where required (i.e., MALDI-ToF-MS)27,28. In addition, all samples were analyzed in house by selective and differential plating to complement IMM results.
Prior to phage addition, urine specimens were enriched in growth medium for 1 h to activate the metabolism of stressed bacteria and to reduce possible matrix effects. Individual reporter phages were added to the enriched urine and luminescence was quantified at 3 h p.i (E. coli, Klebsiella spp.) or 4 h p.i. (Enterococcus spp.). A heatmap was generated to compare the reporter phage assay results to selective plating (Fig. 5b). E. coli was identified in 59 out of the 206 urine samples through conventional detection (i.e., selective plating), whereas combined results from individual E2::nluc and E4::nluc assays led to 40 true positive identifications of E. coli (68% sensitivity). In contrast, the sensitivity for E. faecalis detection by EfS3::nluc and EfS7::nluc and Klebsiella spp. detection by K1::nluc and K4::nluc was considerably higher at 78% (35/45) and 87% (26/30), respectively. Besides K. pneumoniae and K. oxytoca, K4::nluc also identified one urine sample (specimen 152) with the atypical Klebsiella subspecies K. aerogenes (previously Enterobacter aerogenes)29. Interestingly, while all 15 Klebsiella-containing samples from Site 1 were detected, the assay failed to identify 4 out of 12 positive samples from Site 2, possibly due to site-specific difference in circulating clones. In total 5 false positives (2.4%) were observed across 206 samples and considering all three target pathogens: Specimens 6, 18, and 37 were likely due to sample cross-contamination; specimen 101 was a result of cross-reactivity of EfS3::nluc with S. xylosus; sample 146 yielded a false-positive result due to different data evaluation criteria between the reporter phage assay and the GS analysis: Specifically, we detected Enterococcus by reporter phage bioluminescence and in-house plating, however, GS analysis returned a negative result because cell numbers remained below the diagnostic threshold of 103 CFU/ml.
In sum, across 206 patient specimens, the reporter phage assay yielded an overall diagnostic sensitivity of 68%, 78%, and 87%, a specificity of 99%, 99%, and 99%, and an accuracy of 90%, 94%, and 98% for E. coli, Enterococcus spp., and Klebsiella spp., respectively (summarized in Fig. 5c). The species distribution of strains identified during the field evaluation was comparable to the Zurich Uropathogen Collection, with E. coli, Klebsiella spp., and Enterococcus spp. within the top five most prevalent uropathogens (Fig. S4).
Reporter phages as companion diagnostics for phage-patient matching
Therapeutic phage candidates are typically selected based on a broad host range within the bacterial target species. However, individual phages rarely form plaques on more than 70% of patient isolates (see plaquing of phages in Fig. 4a). These differences in antibacterial efficacy are mostly due to the presence of phage defense mechanisms encoded within the host genomes, inevitably resulting in treatment failure22. To increase the likelihood of therapeutic success, a rapid strategy is required to identify suitable phages for each individual patient. Such a companion diagnostic assay for phage-patient matching should predict the antibacterial effect against specific patient isolates under relevant conditions (e.g., directly in urine or blood)30.
Beyond their application for detection and differentiation, the speed and simplicity of reporter phage assays could potentially be harnessed to develop companion diagnostics for phage therapy. However, while reporter phage-induced bioluminescence is a reliable indicator of successful genome delivery, it is unclear whether it can predict killing of patient-derived strains. To address this question, we isolated 50 E. coli, 22 Klebsiella spp., and 36 E. faecalis strains from the field evaluation and quantified bioluminescence in response to reporter phage challenge under defined conditions (107 CFU/ml, 109 PFU/ml, SHU). These RLU values were compared with results obtained from turbidity reduction assays in SHU as a measure of antibacterial activity (Fig. S5a). To this end, growth curves were monitored over 12 h in the presence or absence of wildtype phage and the relative reduction in area under the curve (AUC ratio = AUCphage/AUCcontrol) was quantified. Using these standardized conditions, a RLU vs. AUC ratio dot plot was generated for each pathogen (Fig. S5b), and calculations for precision, sensitivity and F1-score were used to determine RLU and AUC ratio cut-offs that yield the best overall prediction of antibacterial efficacy for individual phage-host pairs (Fig. S5c). The F1-score is the harmonic mean of precision and sensitivity and can be used to optimize trade-offs between these two metrics. Based on these parameter optimizations, true positives were gated at an AUC ratio <0.2 and at RLU values > 105 (E. coli), >104 (Enterococcus spp.), or >106 (Klebsiella spp.).
These gating parameters (Fig. S5) were determined in a defined growth medium (SHU) with a defined cell concentration (107 CFU/ml), which does not reflect a clinical scenario where the patient's urine composition and bacterial cell numbers are unknown. To test how well treatment success can be predicted based only on the raw luminescence values obtained from patient urine, we applied these gating criteria to the luminescence data from the field evaluation (workflow: Fig. 6a). The corresponding dot plots comparing RLU to the AUC ratio are shown for all three pathogens in Fig. 6b–d (right). A selection of activity plots for correctly identified responder patients (true positives) with their matching therapeutic phages are shown in Fig. 6b–d (left). The positive predictive value (i.e., precision) for the companion diagnostic test in patient urine was calculated at 78% (E. coli), 93% (Enterococcus spp.), and 47% (Klebsiella spp.). It can be concluded that RLU is an excellent predictor of antibacterial efficacy for Enterococcus and a good predictor for E. coli, suggesting that reporter phage data could be employed for therapeutic phage selection. Due to many false-positive predictions, companion diagnostic assay development is more challenging for Klebsiella spp. Overall, our companion diagnostic tends to exclude some patients that may benefit from therapy (low sensitivity), which is due to rather stringent selection criteria and/or low bacterial burden in patient urine resulting in weak signals (=false negatives). Nevertheless, our data suggests that rapid reporter phage-based screening could be implemented as an inclusion criterion for phage therapy trials in the future, depending on the pathogen.
a Process of phage-patient matching based on reporter phage-mediated luminescence in urine samples. b E. coli, c Enterococcus, and d Klebsiella strains from the field evaluation were isolated and tested in vitro against each wildtype phage using turbidity reduction assays over 12 h in SHU. Activity was quantified as the ratio of the area under the curve (AUC ratio: AUCHOST+PHAGE/AUCHOST) where a value of one corresponds to no activity and a value of zero corresponds to maximal activity. For each phage–host pair, the AUC ratio was plotted against the raw luminescence value from the field evaluation and the thresholds for optimal activity prediction were applied to these dotplots. Optimal RLU and AUC ratio thresholds values were determined as detailed in Fig. S5. For each pathogen, true positives (TP) are highlighted in yellow. Turbidity-reduction curves of selected correctly identified phage-patient matches are displayed on the left to demonstrate host killing. All in vitro data is shown as mean ± SD (n = 3). TN true negatives, FP false positives, FN false negatives. Some elements of the figure were created using biorender.com. Source data are provided as a Source Data file.
Discussion
By enabling rapid microbial identification, POC diagnostic platforms (e.g., droplet microfluidics, electrochemical biosensors) have the potential to assist antimicrobial stewardship and improve patient care7. For UTIs, the adaptation of these emerging technologies to direct testing in urine remains challenging, largely owing to their uncontrollable response to complex urine matrices of varying pH and ionic strength7. Contrarily, while less complicated technologies (e.g., immunoassays and urine dipsticks) are matrix-insensitive and amenable to POC testing, they often lack adequate accuracy and resolution7. In this study, we resolved these two technological dilemmas by developing a reporter phage-based assay. Using E. coli, Klebsiella spp., and Enterococcus spp. as targets, we demonstrated the feasibility of this platform for identifying the predominant uropathogens from unprocessed patient urine at a resolution of ≥103 CFU/ml within 5 h (Fig. 5). The assay demonstrated high specificity with only few cross-reactivities observed, such as the Klebsiella reporter phages detecting Raoultella, a close phylogenetic relative of Klebsiella23 and EfS3::nluc detecting Staphylococcus and Pediococcus. However, these false-positives should not impact patient care as Raoultella is treated analogous to other Enterobacterales infections (including Klebsiella) and staphylococci such as S. aureus are rarely found in urine (13/663 urological isolates, Fig. S1). Further efforts are clearly necessary to improve the sensitivity for all three assays (68% for E. coli; 78% for Enterococcus spp., and 85% for Klebsiella spp.), which directly stems from the limited host range of natural phages. This could be addressed by engineering additional reporter phages with complementary host ranges, or by expanding the detection spectrum of the current phages through genetic engineering of phage receptor binding proteins31,32,33,34.
Rapid POC identification and species-level differentiation of bacterial pathogens could greatly assist empiric therapy with antibiotics, and seems particularly useful for treating common infections, such as uncomplicated UTIs. Through early pathogen identification, the treatment regimen can be adjusted ensuring a minimal risk of treatment failure. For example, the increasing prevalence of extended-spectrum beta-lactamase producing Enterobacterales35 advocates for administering nitrofurantoin, trimethoprim and sulfamethoxazole, or fosfomycin instead of other antibiotics (such as aminopenicillins or cephalosporins) upon POC Enterobacterales identification36. These other first-line antibiotics could therefore be reserved for treating other bacterial infections, for example, amoxicillin remains the treatment of choice for E. faecalis infections given over 99% of isolates remain susceptible to this antibiotic37. At the same time, identification of Enterococcus contraindicates the use of cephalosporins due to the intrinsic resistance of enterococci to this class of antibiotics38. Moreover, our findings suggest that reporter phages can successfully detect pathogens in whole-blood within less than 6 h, opening promising avenues for septicemia diagnosis. This further supports the adaptability of reporter phage-based diagnostics to diverse and complicated biomatrices.
Despite their transformative potential in bacterial diagnostics, the adoption of reporter phage assays in the clinical realm is still in its infancy. Challenging bottlenecks include the requirements for instrumentation, system automation, workflow integration, and more importantly, establishment of reporter phage libraries covering the clonal diversity of ind...
Comments
Post a Comment