Isolation of bacteriophages
We developed a collection of bacteriophages that infect and lyse both Klebsiella pneumoniae and Klebsiella spp. isolates with the goal of determining whether these isolates would have activity against MDR and XDR clinical isolates. These phages were identified in the continental USA (Table S1) using our standard phage isolation techniques16 from various environmental sources. We identified and sequenced a total of 11 unique phages belonging to 4 different families (Table S1). Their genomes ranged in size from 37.9 Kbp to 177.7 Kbp. Each of these phages was closely related to previously characterized phages, having high levels of identity to phage sequences stored in the NCBI database (Table S2). The phages were found to represent four distinct phylogenetic clades, with Beam, KL35, Rec, QTY, and Turmeric making up one clade, APV, Ace, and LK1 making up another, and LK2 and Chai representing the third (Fig. S1). Phage LK3 was an outlier, as it was not genetically similar to any of the other phages included in this study (Fig. S1).
We characterized the morphologies of the phages via transmission electron microscopy (TEM) and identified a range of different morphologies (Fig. 1). For example, phages Ace, APV, Beam, QTY, Rec, Turmeric, LK1, and KL35 all had morphologies consistent with those of myoviruses (Panels A–F, H, K), while LK2 had a morphology consistent with podoviruses (Panel I), and others were consistent with siphoviruses (Panels G, J). These data confirm that we identified several different morphologies amongst the 11 different viruses we characterized.
TEM images of negatively stained phages with 2% of uranyl acetate (A–K). Scale bars are 200 nm represented on each panel.
Examination of host ranges
We evaluated the host ranges of the 11 phages against a group of 59 clinical isolates of K. pneumoniae and K. spp. to determine whether they were capable of lysing them. Of the 59 isolates tested, 20 of them were relatively susceptible to antibiotics, while 39 of them were known to possess Extended Spectrum Beta Lactamases (ESBLs) or other antibiotic resistance genes (Table S3). Of the 39 ESBL Klebsiella isolates, 4 were Carbapenem Resistant Enterobacteria (CRE), and of the other 20 Klebsiella isolates, 1 was a CRE. These isolates were derived from blood, urine, respiratory tracts, or other body sites from human subjects. Some of the 11 phages had little if any activity against the Klebsiella isolates, including phages LK3, LK2, and Chai, while others, including KL35, Tumeric, Rec, QTY, APV, Ace, and Beam had variable activity against the isolates (Fig. S2).
Experimental evolution
Building on a previous study that demonstrated the effectiveness of coevolutionary training for laboratory adapted E. coli isolates17, we hypothesized and tested that coevolutionary training might have an additional effect: broadening phage host ranges for clinical K. pneumoniae isolates alongside improving their effectiveness. We predicted that as phages engaged in an arms race with their hosts, they would either accumulate counter-defenses usable against multiple bacterial strains or evolve reduced specificity to overcome resistant hosts. The primary differences between the two protocols were the use of clinical isolates compared to laboratory-adapted isolates, the fact that Klebsiella isolates had mucoid phenotypes compared to non-mucoid E. coli, and the evolution process between host and phage took place for a full 30 days before the phages were recovered (Fig. 2A). Specimens were transferred into fresh media on a daily basis to prevent nutrient depletion, and specimens were titered every 3 days to evaluate whether the phages-maintained viability (Fig. S3). We then isolated the phages on day 30 and evaluated whether they had different host ranges than those that we started with and if they presented higher growth inhibition than the ancestral phages (Fig. 2B).
A Phages were coevolved with various clinical strains of K. pneumoniae with an MOI of 10 at 37 °C for 30 consecutive days and each day 1% of previous day phage and bacterial coevolved population was transferred to fresh media. The population dynamics of evolved phages and bacterial hosts were estimated based on plaque forming units (PFU) and colony forming units (CFU), respectively. Each phage and host was isolated and saved on every day 3 of coculturing. B After 30 days of coevolution study, the corresponding evolved phages were tested for kinetic and lytic efficiency by performing liquid growth curve and host range assays.
Host range expansion
We focused on two of our phages with relatively modest host ranges (Fig. S2), phages Ace and APV (belonging to the family Straboviridae) for our phage evolution/host range expansion experiments. We selected phages with modest host ranges to give them the opportunity for improvement. Prior to the expansion experiments, phage APV had some lytic activity against 27.12% of the isolates tested, while phage Ace had lytic activity against 42.37% of the tested isolates. After 30 days of the experimental evolution, in the spot titer tests, we found substantial differences in the lytic capacity of phage APV before and after the experiment. For example, different trained phage APV isolates varied in their lytic capacity, from having some lytic capacity against 16.95% of the isolates to 61.02% of the isolates. Overall, the host range of the co-evolved phages increased in four experiments and decreased in two experiments (Fig. 3). We found similar results for phage Ace, where the post-evolution phage isolates ranged in activity from 30.51% to 59.32% of the K. pneumoniae isolates, showing an expansion of host range in three experiments, a decrease in host range in two experiments, and remained the same in one experiment. There was host-range expansion identified against both ESBL isolates and non-ESBL isolates. These data strongly suggest that the 30-day process of phage-host evolution resulted in a set of phages with improved capabilities in lysing their K. pneumoniae hosts. Host range expansion was observed to K. pneumoniae clades for which the ancestral phages showed no lysis. Coevolved APV phages expanded their host ranges to four clades that were not lysed by the ancestral phage. Coevolved Ace phages expanded their host range to four clades that were not lysed by the ancestral phage (Fig S4). This host range expansion also included lysis of putative O-antigen and capsule serotypes not previously infected.
Phage lysate (5 µl) with 107 titer was spotted on a bacterial lawn. After overnight incubation, the plates were examined for lysis and no lysis. The dark blue boxes indicate complete lysis, medium to light blue boxes indicate medium to light lysis and boxes with no color represent no lysis. The isolates were grouped based on extended-spectrum beta-lactamases (ESBLs). CRE (Carbapenem Resistant Enterobacteriaceae) isolates are indicated with a star.
Evaluation of growth dynamics
Phage training involves co-incubating bacterial host and phage together over successive generations so that the phage adjusts to changes in the host to allow it to survive common host adaptations18. Such training experiments have previously been performed for E. coli, K. pneumoniae, P. aeruginosa, and L. monocytogenes19,20,21,22. While examination of phage activity using spot assays can provide an understanding of lytic potential, we believe the ability to inhibit growth longitudinally in broth media is a better approximation of the ability of phages to kill their hosts while overcoming their ability to develop resistance rapidly. We examined the ability of the evolved phages Ace and APV compared to the ancestral phages Ace and APV to determine whether the evolutionary process resulted in better adapted phages for longitudinal inhibition of K. pneumoniae host growth. We found that over the 72 h, the trained phages were superior to the ancestral phages in suppressing the growth of the K. pneumoniae isolates (Fig. 4). Indeed, in 10/12 examples, the trained phages demonstrated better suppression of K. pneumoniae growth longitudinally. We examined whether the suppression was statistically significant via area under the curve analysis (AUC) and found that in 7 of the 12 examples (58%), the longitudinal suppression of the K. pneumoniae isolates was significant (Fig. 5). These isolates were all highly susceptible to antibiotics and generally would not be the target of treatment with phages. To decipher whether the evolved versions of phages Ace and APV may have utility against antibiotic-resistant K. pneumoniae isolates, we would need to decipher whether those phages would be capable of suppressing MDR and XDR K. pneumoniae isolates.
The phage and bacteria were coevolved together for 30 consecutive days, the trained phages after 30 days were isolated and determined their kinetic efficiency on their corresponding coevolved hosts by monitoring bacterial growth for 72 h at 37 °C. Growth curves are shown as the average of 3 separate biological replicates for each microbe and/or microbe/phage pairing with standard deviation bars in each panel (A–L).
Experiments were performed at MOI of 1, and phages trained in KP117 were used for all experiments (APV.2 and Ace.2). Bacterial growth was monitored for 72 h by measuring the OD600 every 15 min in a microplate reader. Experiments were performed at 37 °C. Note: Growth curves for CRE isolates are indicated with a star. Growth curves are shown as the average of 3 separate biological replicates for each microbe and/or microbe/phage pairing with standard deviation bars in each panel (A–L).
Inhibition of MDR and XDR isolates
To determine whether evolved phages Ace and APV had greater activity against MDR and XDR K. pneumoniae isolates, we tested them against our collection of MDR and XDR isolates in broth media over 72 h. We found that in all cases, the trained phages were superior to their untrained counterparts in their ability to suppress growth of their hosts (Fig. 5). This trend of suppressing the growth of their hosts extends when evaluating a larger group of clinical K. pneumoniae isolates and even to CRE isolates (Fig. S6). We also evaluated whether the suppression of the K. pneumoniae hosts with these trained phages was statistically significant over the 72 h using AUC analysis and found that 21/28 (75%) of the isolates were indeed significantly suppressed (Figs. S7, S8). These data indicate that the process of evolving Klebsiella phages with their clinical hosts results in some phages with significantly improved suppression profiles of MDR and XDR isolates over at least 72 h.
Phage genetic changes
We examined the phages that were coevolved over the 30-day experiment to identify what differences may have occurred that led to the improved suppression profiles and the extended host ranges that led to the suppression of the MDR and XDR Klebsiella isolates. All trained Ace phages were similar on the whole genome level by average nucleotide identity level with less than 0.01% difference post 30-day evolution (Panel A in Figs. 6 and S9). APV phages had greater overall genome differential change at 0.02%, except APV.1 and APV.5 (Panel B in Figs. 6 and S9). Phage ACE.3 had a 165 nucleotide (nt) deletion in an intergenic region. All trained phages had at least one nonsynonymous mutation in the L-shaped tail fiber protein, which is responsible for phage recognition and binding to host cells. Indeed, the majority of the mutations observed were concentrated in the tail fiber and baseplate regions of both phages (Figs. 6 and S10), each of which may be involved in host recognition and binding. These data suggest that the coevolution process resulted in changes in phage recognition/binding, which may be responsible for the phenotypic changes in host suppression we observed.
Panel (A) represents phage Ace and Panel (B) represents phage APV. Ancestral phages were aligned with trained phage genomes and visualized using the Integrated Genome Browser. The Integrated Genome Browser was then used to generate mismatch graphs from the BAM files showing the number of mismatched nucleotides compared to the reference sequence.
