Genetic Signatures from Adaptation of Bacteria to Lytic Phage Identify Potential Agents To Aid Phage Killing of Multidrug-Resistant Acinetobacter baumannii

ABSTRACT

With the increasing morbidity and mortality rates associated with multidrug-resistant bacteria, interest in bacteriophage therapy has been revived. However, bacterial resistance to phage infection threatens the usefulness of phage therapy, especially its inclusion in modern medicine. Multidrug-resistant Acinetobacter baumannii is a top-priority pathogen requiring urgent intervention and new therapeutic approaches, such as phage therapy. Here, we experimentally adapted A. baumannii WHG40004 to its lytic phage P21 and thereafter isolated a phage-resistant bacterial mutant, named Ev5-WHG. We then aimed to identify potential agents to aid phage killing of Ev5-WHG by analyzing its genome and that of the wild-type strain. The enriched Gene Ontology (GO) analysis based on genetic alterations in minor alleles and mutations showed that pathways such as zinc ion transport and cell membrane synthesis could play certain roles in phage resistance. Remarkably, the combination of zinc acetate and P21 showed increased bactericidal effect on Ev5-WHG. Significantly also, we showed that P21 completely prevented the growth of wild-type WHG40004 in the presence of antibiotics (meropenem and imipenem). The results from this study indicate that the analysis of phage resistance signatures during adaptation of bacteria to a lytic phage can inform the choice of agents to work cooperatively with phage to limit and/or reverse resistance. This approach could be important for guiding future successful phage therapy.

IMPORTANCE Bacteriophages have proven very useful as alternative therapeutic agents in combating multidrug-resistant bacterial infections; however, bacterial resistance to phages threatens their use. In this study, we showed a new strategy of leveraging genetic signatures that accompany phage resistance in bacteria to predict agents that can be used with lytic phages to combat multidrug-resistant Acinetobacter baumannii. Significantly, this approach was helpful in suggesting the use of zinc acetate to reduce resistance in phage-resistant bacteria, as well as the use of phage with antibiotics meropenem and imipenem to prevent resistance in a wild-type strain of multidrug-resistant A. baumannii. The approach of this study will be helpful for improving the outcome of phage therapy and in overcoming antimicrobial resistance.

INTRODUCTION

The continuous emergence of resistance to antibiotics remains a major problem in health care, restricting treatments and resulting in high mortality due to multidrug-resistant (MDR) bacteria. One such MDR bacterium is the nosocomial carbapenem-resistant Acinetobacter baumannii (CRAB), which is a leading cause of bloodstream infections and ventilator-associated pneumonia (1, 2). This Gram-negative bacterium is considered priority 1 in the list of World Health Organization (WHO) priority pathogens for research and development of new antibiotics (http://www.who.int/mediacentre/news/releases/2017/bacteria-antibiotics-needed/en/) and an urgent public health threat by the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/drugresistance/biggest-threats.html). Thus, only a few therapeutic options remain to treat infections caused by this pathogen.

Generally, interest in the use of bacteriophages (viruses that infect bacteria) to treat bacterial infections has been rekindled. Bacteriophages (phages) offer a number of advantages as alternative therapeutics to MDR bacterial infections, and many reports have shown the effectiveness of phage therapy (3–5). However, one obvious limitation to phage therapy is bacterial resistance to phage infection (6–8). Bacteria are able to resist phage infections through varied mechanisms (9, 10). Phages, being biological agents, are also able to evolve counterdefense mechanisms against phage-resistant mutants, including adaptation to modified receptors (11–13).

Bacteria can resist phage infection via spontaneous mutations (14–17). Spontaneous mutation, being the source of all genetic polymorphism (18), can reveal how natural selection drives the adaptation between phages and their bacterial hosts. Also, since these acquired alterations vary across bacterial isolates, it may be important to individually relate different phage-bacterial models and understand the dynamics that exist therein. Importantly, this understanding has the potential to inform ways to either prevent phage resistance or reverse it after occurring.

In order to overcome resistance during phage therapy, the use of phages in combination with antibiotics has been proposed (19). This is importantly feasible in that, when bacteria become resistant to phage, they sometimes suffer fitness costs which resensitize them to antibiotics to which they were previously resistant (20). Therefore, reusing these antibiotics could improve the outcome of therapy (21). However, this approach would entail using the right antibiotics to complement phage during therapy. Some studies had previously used traditional analysis of point mutations in phage-resistant mutants to predict the right antibiotics to use (20, 22). In the current study, we attempted to base our prediction on genetic polymorphisms of minor alleles in the genome of phage-resistant mutants.

By studying the resistance mechanism of Staphylococcus aureus to a lytic phage during an in vitro experimental adaptation, we have previously shown success in predicting agents that, when used in combination with lytic phages, can enhance the killing of phage-resistant mutants (13). In the current study, we adapted wild-type A. baumannii WHG40004 to its lytic phage P21 over an extended time of 10 days. We then analyzed the alterations in the genome of the resulting phage-resistant mutant and uncovered genetic variations of minor alleles in functional genes with key roles in important pathways. By leveraging this analysis, we then predicted agents to aid phage killing of resistant bacterial mutants, as well as preventing resistance in wild-type bacteria.

RESULTS

Dynamics of WHG40004 and P21 interaction in experimental adaptation.

In the long-term (10-day) experimental adaptation, a number of interesting results were obtained. Expectedly, WHG40004 readily evolved resistance to P21 as confirmed by spot assay on agar plates at each serial transfer. The resulting bacteria at the end of the fifth transfer, here called Ev5-WHG, were confirmed phage resistant by spot assay (Fig. 1A). Although both the test bacterial liquid cultures (with phage) and phage-free control bacterial cultures maintained substantial densities throughout the transfer, the density of the former was a little lower than the phage-free control. Interestingly, from the estimation of the viable phage densities, we discovered a significant decline in the density of P21 and an eventual loss after the third serial transfer (Fig. 1B). We infer that WHG40004 quickly becomes resistant to P21 during culture in liquid medium. This further indicates that P21 could only moderately reduce the growth rate of WHG40004 in the liquid culture compared to that of the phage-free control.

FIG 1.

Bacterial adaptational response to phage and phage resistance mechanisms. (A) Spot assay confirmation of resistance in Ev5-WHG. (B) Change in bacterial and phage densities during serial transfer. (C) Cell surface appearance visualized by SEM. (D) Phage adsorption assay on WHG40004 and Ev5-WHG. (E) Growth of wild-type and phage-resistant mutant A. baumannii in the absence of P21. (F) Comparison of growth rates of WHG40004 and Ev5-WHG. Differences in bacterial growth between Ev5-WHG and WHG40004 were assessed by Student’s t test (P < 0.05). Data are expressed as mean ± standard error of the mean (n = 2). ns, not significant.

Next, we sought to steer P21 to evolve a counterdefense against Ev5-WHG, a phenomenon that had been previously successful in some phage-bacterial models (13, 17, 23). Surprisingly, P21 lost its ability to evolve infectivity of the Ev5-WHG mutant, following 20 consecutive serial passages. Even by complementation of the Ev5-WHG mutant with its wild-type strain, P21 was not able to rescue phage susceptibility. We infer that the long-term adaptational interactions between WHG40004 and P21 contributed to the inability of the phage to regain infectivity of Ev5-WHG, unlike in the previous studies where resistant mutants were selected after only a short interaction.

Interestingly, scanning electron microscopy (SEM) image analysis revealed that the cell surface appearance of the Ev5-WHG mutant was rough and appeared thickened compared to a smooth appearance in wild-type bacteria (Fig. 1C). Also, phage adsorption assays showed that P21 had impaired attachment to Ev5-WHG (Fig. 1D). While P21 displayed adsorption rates of 70.02% (±2.34), 71.53% (±1.55), and 75.01 (±0.66) at 10 min, 20 min, and 30 min, respectively, of interaction with wild-type WHG40004, it showed significantly lower adsorption rates of 3.66% (±0.16), 3.41% (±0.14), and 3.66% (±0.16) to mutant Ev5-WHG, respectively. These results implied that the change in cell surface appearance could have probably inhibited phage adsorption, thus affecting phage attachment to the cell surface of Ev5-WHG.

Further, to test whether phage resistance incurred a fitness cost on the growth of the Ev5-WHG mutant, we compared its growth rate in the absence of phage with the wild-type strain (Fig. 1E and F). We noticed that the two strains had similar rates of growth. In other words, there was no trade-off in the growth of the Ev5-WHG mutant following its resistance to P21. Also, MICs of antibiotics (colistin, polymyxin B, meropenem, and imipenem) were 2.5 μg/mL, 2.5 μg/mL, 32 μg/mL, and 64 μg/mL, respectively, for wild-type WHG40004 and 1.25 μg/mL, 1.25 μg/mL, 32 μg/mL, and 128 μg/mL, respectively, for the Ev5-WHG mutant. This indicates that A. baumannii WHG40004 is carbapenem/multidrug resistant and that the evolution of phage resistance in Ev5-WHG was not accompanied by a resensitization to these antibiotics.

Genomic analysis reveals multiple multidrug resistance genes and several polymorphic sites.

The genome of a single (doubly purified) colony isolated from the whole bacterial population confirmed phage resistant was sequenced, alongside that of the wild-type bacteria. Following high-throughput sequencing data analysis, both wild-type and phage-adapted bacteria were found to have the same length of 4,066,117 bp with average sequencing depths of 1,580 and 1,062, respectively. Overall, a total of 3,825 coding sequences (CDSs), 74 tRNAs, and 18 rRNAs were annotated. Notably, there were 27 multidrug-resistance-related genes on the genomes of both bacteria (see Table S1 in the supplemental material).

The results of pairwise whole-genome alignment revealed no indel on the genome of Ev5-WHG but 22 single nucleotide mutations on 13 genes (Table S2). The average sequencing error rate of next-generation sequencing (NGS) data was estimated to be about 0.003 using Jellyfish software. Interestingly, analysis of minor alleles, the less common alleles for single nucleotide sites in genomes, revealed more genetic polymorphisms in the genomes of A. baumannii used in this study. There were 851 polymorphic sites with a minor allele frequency (MAF) larger than 0.06 in the genome of wild-type bacteria, among which there were 129 sites with a MAF larger than 0.1. Also, there were 1,898 polymorphic sites with a MAF greater than 0.06 on the genome of the Ev5-WHG mutant, among which there were 784 sites with a MAF greater than 0.1 (Fig. 2A). Compared with the wild-type bacteria, the proportion of MAFs above 0.1 was significantly higher in the genome of the adapted bacteria (Fig. 2B). However, based on statistical analysis, there was no correlation between the MAF value of each polymorphic site and its sequencing depth (Fig. 2C and D).

FIG 2.

Genetic polymorphisms in WHG40004-P21 adaptational system. (A) The genome of Ev5-WHG revealed 1,898 polymorphism sites. (B) The proportion of MAFs above 0.1 was higher in wild-type WHG40004. (C) Correlation between the MAF value of each polymorphism site and its sequencing depth in WHG40004 determined by Pearson correlation analysis (P < 0.05). (D) Correlation between the MAF value of each polymorphism site and its sequencing depth in Ev5-WHG determined by Pearson correlation analysis (P < 0.05).

Further, we performed comparative genomic analysis to determine the location of polymorphic sites and single nucleotide mutations. Polymorphic sites with a MAF larger than 0.06 were found in 574 genes in the genome of Ev5-WHG. Further, 4 genes had both mutations and polymorphic sites (Table S3), and only one gene, coding for d-gluconic acid reductase B, had 5 mutations. It is worth noting that the genetic polymorphisms were not uniformly distributed across the genome. There were three regions with a discriminably large number of genetic polymorphisms in the adapted bacterial genome (Fig. 3). These regions contain genes involved in multiple components of the cell envelope (outer membrane, cell wall, and cell membrane).

FIG 3.

Comparative genomics analysis reveals a large number of polymorphism sites and single nucleotide mutations in Ev5-WHG.

GO enrichment analysis reveals pathways with roles in phage resistance and suggests probable agents to aid phage killing of Ev5-WHG.

We executed Gene Ontology (GO) enrichment analysis to determine which biological process, molecular function, or cellular component had been altered by the genes with mutations or genetic polymorphisms (Tables S4 and S5). We obtained five modules related to biological processes and cellular components from the enrichment analysis (Fig. 4). We infer that these processes played certain roles in the resistance of Ev5-WHG to P21. Notably, genes with genetic polymorphisms or mutations were significantly enriched in the “cell outer membrane-related” GO unit, which is consistent with the result of SEM analysis. Importantly, 14 processes involved in the pathways for synthesis of components of both cell wall and cell outer membrane were shown to have been altered in the cell outer membrane-related GO unit (Fig. 4). Thus, we predicted that antibiotics colistin and polymyxin B, which act on bacterial cell outer membrane, and antibiotics meropenem and imipenem, which act on the cell wall (24), may be probable agents capable of aiding phage killing of Ev5-WHG. Significantly also, a biological process related to zinc ion transport was enriched. Thus, we further predicted that zinc acetate may be able to enhance the killing of Ev5-WHG when used in combination with P21.

FIG 4.

Gene Ontology enrichment analysis reveals five modules related to biological processes and cellular components. The width of the connector lines represents the number of shared genes between two nodes.

Combination of zinc acetate and P21 enhanced the killing of Ev5-WHG.

To test our predictions, zinc acetate was used together with phage P21 to evaluate its potential in reducing phage resistance in Ev5-WHG. The results showed that the combination of zinc acetate and P21 enhanced the killing of Ev5-WHG (Fig. 5A). Further, we observed the growth of wild-type WHG40004 in the presence of zinc acetate and a combination of zinc acetate and P21; the results showed that zinc acetate has no significant effects on the bacterial growth but that a combination of zinc acetate and P21 did reduce the bacterial growth (Fig. 5B). Importantly, in the presence of P21, zinc acetate delayed resistance in wild-type bacteria for about 6 h (Fig. 5C).

FIG 5.

(A) Combination of zinc acetate and P21 enhanced the killing of resistant mutant Ev5-WHG. (B and C) Zinc acetate limits phage resistance in wild-type WHG40004 for 6 h (4 to 10 h). Differences were assessed by Student’s t test (P < 0.05). Data are expressed as mean ± standard error of the mean (n = 3). *, P < 0.1; **, P < 0.01; ****, P < 0.001.

Antibiotics could not limit phage resistance in mutant Ev5-WHG.

The predicted use of P21 and antibiotics colistin and polymyxin B did not aid phage killing of mutant Ev5-WHG (Fig. S1A and B). Since the GO domain “cell outer membrane” covers multiple components of the cell envelope, we also attempted to use the antibiotics meropenem and imipenem, which target the bacterial cell wall, with P21 to see their influence on phage killing of mutant Ev5-WHG. However, P21 still could not limit the growth of mutant Ev5-WHG in the presence of sublethal doses (one-half and one-fourth MICs) of these antibiotics (Fig. S1C and D). In other words, its resistance to phage P21 could not be reversed even in the presence of these antibiotics.

Since the desired situation is actually to avoid phage resistance, we thought to use these antibiotics against wild-type WHG40004 to see if they could delay or prevent phage resistance. Remarkably, it was observed that bacterial growth in the presence of P21 and meropenem or imipenem was completely prevented (Fig. 6A and B). However, antibiotics colistin and polymyxin B did not show any significant difference from the single use of P21 (Fig. 6C and D).

FIG 6.

(A and B) Meropenem (A) and imipenem (B) in combination with P21 or with P21 and zinc result in complete clearance of the wild-type WHG40004 strain. (C and D) Colistin (C) and polymyxin B (D) could not limit the growth of wild-type WHG40004 when used to complement phage P21.

Moreover, we attempted to evaluate if there is a differential effect of using both zinc and antibiotics on phage killing of Ev5-WHG and WHG40004. Interestingly, the combined use of zinc and antibiotics (meropenem and imipenem) prevented phage resistance in Ev5-WHG (Fig. 7A and B). Also, zinc and antibiotics (colistin and polymyxin B) with P21 showed a greater antibacterial effect than their single use (Fig. 7C and D). Similarly, the combined use of zinc and antibiotics prevented the growth of wild-type WHG40004 (Fig. 6A to D).

FIG 7.

(A and B) Combination of P21, zinc, and antibiotics (meropenem and imipenem) prevents the growth of Ev5-WHG. (C and D) Combination of P21, zinc, and antibiotics (colistin and polymyxin B) showed a greater antibacterial effect than single use.

DISCUSSION

As more and more multidrug-resistant bacterial strains emerge, it is feared that we may be coming to the end of the antibiotic era. This calls for possible solutions, including minimizing the emergence of resistance to antibiotics or finding entirely different approaches to combating infections caused by bacteria. Phage therapy has shown promising efficiency in controlling MDR bacteria; however, bacterial resistance to phage threatens its longevity. Understanding the varied phage resistance mechanisms and what ways remain to successfully combat MDR bacterial infections is thus very important.

In a previous study, we showed for the first time that genetic polymorphism can alter the susceptibility of Staphylococcus aureus AB91118 to its lytic phage (13). By experimentally adapting MDR A. baumannii to its lytic phage P21, we demonstrate in the current study that genetic polymorphisms also exist in the adaptational system of A. baumannii to its lytic phage. We utilized the same bioinformatics pipeline to predict an effective agent that, when used with lytic phage, can enhance the killing of a phage-resistant mutant. Of note is that, though phage P21 is a lytic phage of the WHG40004 strain of A. baumannii used in this study (Fig. 1A), it possesses only a suppressing effect, as the bacteria maintained some growth in liquid culture (see Fig. S2 in the supplemental material). This occurrence, which has also been reported in soil bacteria Pseudomonas putida F1 and Bacillus subtilis 168 with phage PEf1 and coliphage T4 (25), may indicate that the bacteria easily generate resistance to the phage.

Following multiple rounds of adaptation of A. baumannii WHG40004 to its lytic phage P21, the resulting phage-adapted bacteria had 18 mutations and some genetic polymorphisms in multiple components of the cell envelope. This finding is consistent with the result from scanning electron microscopy, which also showed changes in the cell surface appearance. Besides, phage adsorption experiments showed that the adsorption of phage P21 to the adapted bacteria was lost. We therefore infer that phage resistance might have proceeded via loss of phage adsorption to Ev5-WHG. Previous studies have shown that change in the phage-binding receptor on the bacterial surface is mostly responsible for phage resistance in bacteria (26–28). This could explain why it was not possible for P21 to evolve infectivity of Ev5-WHG, even after several rounds of adaptation.

Further, genomic analysis revealed multiple multidrug-resistant phenotypes in both wild-type and adapted bacteria, making it difficult to select effective antibiotics against the bacteria, especially the adapted bacteria which now have both multidrug resistance and phage resistance phenotypes. Therefore, we proceeded to understand the genetic polymorphisms that have occurred from the selection for phage resistance in Ev5-WHG. We associated the mutated genes or minor alleles with biological process, molecular function, or cellular component through the GO enrichment. We thought that the enrichment results could explain the potential biological mechanisms for the phage resistance. Notably, we observed alterations in the zinc ion transport and in pathways related to the bacterial cell outer membrane, which helped us to predict the use of zinc acetate and antibiotics colistin and polymyxin B.

Thus, by predicting that complementation of P21 with zinc acetate could enhance the killing of the Ev5-WHG mutant based on the observed genetic alteration in the zinc ion transport, an increased bactericidal effect on the phage-resistant bacterial mutant was observed. Zinc is an important element for microorganisms (29, 30). Some previous studies have shown that zinc ions possess antimicrobial activities against bacteria mainly via two mechanisms: direct interaction with microbial membranes resulting in membrane destabilization and increased permeability and also interaction with nucleic acids and deactivation of enzymes of the respiratory system (31–34). The antimicrobial effect of zinc ions had been previously tested on two bacterial species: Escherichia coli (Gram negative) and Staphylococcus aureus (Gram positive) (32). Thus, we infer that the zinc acetate destabilized the membrane of Ev5-WHG for the adsorption of P21, resulting in the phage killing of the bacteria.

Ontology enrichment analysis of the genes in Ev5-WHG with unique minor alleles different from wild-type WHG40004 revealed 14 pathways related to the bacterial cell outer membrane (Fig. 4). Among them, some processes, such as the peptidoglycan metabolic process and cellular protein modification process, are involved in the pathways for synthesis of components of the cell wall and cell outer membrane. Others relate to the catabolic process of some important molecules, such as ribonucleosides, drugs, cellular nitrogen, etc. These results implied that these pathways may play certain roles in the phage resistance of Ev5-WHG. Therefore, we inferred that antibiotics which target both bacterial cell outer membrane and cell wall may affect the generation of resistance. The results of using antibiotics colistin and polymyxin B (targeting cell outer membrane) and meropenem and imipenem (targeting synthesis of cell wall) showed that these antibiotics failed as probable agents to aid phage in killing the Ev5-WHG mutant (Fig. S1A to D). However, antibiotics meropenem and imipenem were able to prevent the generation of resistance to phage in wild-type WHG40004 (Fig. 6A and B), while colistin and polymyxin B could not (Fig. 6C and D). Colistin and polymyxin B act on the outer cell membrane (35, 36), so there could be a possibility that they could not go inside the bacteria to interact with certain pathways related to phage resistance. In contrast, a key factor in the efficacy of carbapenems is their ability to bind to multiple different penicillin-binding proteins (PBPs) (37); so, there is a possibility that meropenem and imipenem could go inside the bacteria and interact with some pathways to prevent the wild-type bacteria from generating phage resistance. But, since the mutant Ev5-WHG was already resistant to the phage and its cell wall had changed, the combination of P21 and meropenem/imipenem could not enhance the killing of Ev5-WHG.

On the other hand, zinc ions have been shown to inhibit the growth of carbapenem-resistant A. baumannii by producing reactive oxygen species (ROS) which elevate membrane lipid peroxidation that causes membrane leakage of reducing sugars, proteins, and DNA and reduces cell viability (38). This can therefore explain why the combination of zinc acetate and P21 enhanced the killing of Ev5-WHG. It can further explain why the combined use of zinc acetate and antibiotics could prevent and limit phage resistance in Ev5-WHG (Fig. 7). Zinc acetate probably destabilized the cell membrane for the adsorption of these antibiotics and P21, which was not possible when only antibiotics and P21 were used.

Furthermore, it is of more interest to prevent resistance from occurring than to find means to reverse it afterward. Thus, the results that the combined use of P21 and meropenem or imipenem with sublethal doses of the MIC could prevent the growth of WHG40004 and the combined use of P21 and zinc could delay resistance generation and partially reduce resistance of Ev5-WHG are important, showing that it is possible to find the right combinations for guiding better phage therapy by using genetic signatures. Since phages and antibiotics use different modes of action, the likelihood of bacteria evolving simultaneous resistance to both agents is low. However, the results of the combined use of P21 and colistin or polymyxin showed that not every phage and antibiotic combination will work. A solution to this is to find the right combination. The experimental adaptation and the genomic analysis used in the current study may help to find the right combination.

Our investigation in this study was based on changes in optical density (at 600 nm) (OD₆₀₀) to measure cell density. Although we attempted to perform plaque assays, clear spots on Ev5-WHG plates could not be detected after adding P21 with zinc acetate. This is not unexpected, as the addition of zinc acetate and P21 could only reduce the growth rate of Ev5-WHG (Fig. 5A). Also, zinc acetate only partially reduced phage resistance. Measuring bacterial density, which we also used in our previous study (13), allows for gauging the extent to which resistance was reduced, as well as the period for which the reduction occurred. Overall, the scope of the current study was to use a bioinformatic pipeline based on minor alleles to identify potential agents that can be used with phage to limit phage resistance. More studies are needed in the future to provide insights into the molecular mechanisms behind why zinc acetate could reduce phage resistance in the resistant mutant.

Summarily in this study, we showed a strategy of using genetic signatures of genomic alterations arising from evolution of phage resistance to predict agents that, when used in combination with phage, can (i) enhance the killing of phage-resistant bacterial mutants (in this case zinc acetate was used) and (ii) limit resistance in bacteria during phage treatment by choice of the right antibiotics to use together with phage. If these findings could be proven in more phage-bacterial interactions and models, it would be importantly helpful in improving the outcome of phage therapy and make a change in our fights in overcoming microbial resistance.

MATERIALS AND METHODS

Bacterial strain, phage type, and culture conditions.

Carbapenem-resistant Acinetobacter baumannii strain WHG40004 and its phage P21 (39) were obtained from culture stocks of the Diagnostic Microbiology Unit of the Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), People’s Republic of China. Bacteria were previously isolated from health care-associated carbapenem-resistant A. baumannii infections in the intensive care unit of a hospital. Bacteria were routinely grown in lysogeny broth (LB) medium and incubated at 180 rpm/37°C. For routine phage amplification, purified phage was added to mid-log growing culture of bacteria and phage buffer was added. Plaque purification and enumeration of phage titers were done using the double-layer overlay method.

In vitro bacterial adaptation to phage.

Bacterial adaptational response to phage (5 serial transfers over 10 days) was experimentally performed as previously described (40). Briefly, a 2-mL microcosm of LB medium was inoculated with 5 μL fresh lysate of P21 and 20 μL overnight culture of WHG40004 from a single ancestral A. baumannii colony, at a multiplicity of infection (MOI) of 1, and incubated at 37°C/180 rpm. In the control treatment, phage was not added. After every 48 h, a 20-μL sample of culture containing both bacteria and phage was transferred to a new microcosm with fresh LB medium, up to 5 transfers. At each transfer, a spot assay was performed to determine phage resistance of the evolved bacteria. Also, the total phage and bacterial biomass remaining from each transfer was tracked by performing measurement of bacterial CFU and phage PFU, on the same day as transfer.

At the end of the fifth transfer, the surviving bacteria were streaked onto an LB agar plate and incubated at 37°C overnight. A single colony was picked from the plate and restreaked to obtain a double-purified isolate. The purified isolate was then inoculated into 10 mL of LB medium and incubated at 37°C overnight; this was taken as the phage-evolved bacterial mutant. Aliquots of this mutant were preserved in 50% glycerol and stored at −80°C for subsequent use. Phage resistance of the mutant was confirmed by cross streaking against the ancestral phage.

Experimental phage counteradaptation to phage-resistant mutant.

At the start of the experiments, stationary-phase phage-resistant mutants were aliquoted and preserved at −80°C, and one aliquot was used for each round of adaptation. Experimental counteradaptation of P21 to the phage-resistant bacterial mutant was done through two methods. In one method, ancestral phage was propagated on phage-resistant bacteria for 20 consecutive serial passages as previously described (41, 42). In the second method, propagation of P21 on resistant bacteria was complemented with the wild-type strain.

Bacterial growth assay.

Bacterial growth was determined in a 96-well assay plate by mixing phage and bacteria at an MOI of 1 and incubating them at 37°C. Turbidity was measured by a spectrophotometer (BioTek, USA) at 600 nm at regular intervals up to 10 h. The OD₆₀₀ values as a function of time were plotted to generate the growth curve. OD₆₀₀ values were also used to calculate bacterial growth rate (ΔOD₆₀₀/Δt) (43).

Phage adsorption assay.

Adsorption of P21 to bacteria was performed by mixing overnight bacterial cultures and phages from fresh lysates at an MOI of 0.01 (44). Suspensions were collected at 10, 20, and 30 min and centrifuged at 4,000 × g for 3 min. Dilutions of the supernatant were plated for quantification of free phage particles.

SEM for bacterial morphology observation.

Two milliliters of bacterial suspension (∼10⁶ CFU/mL) in LB was put onto clean cover slides in 6-well plates and incubated for 4 h at 37°C, followed by washing with phosphate-buffered saline (PBS). The bacteria were then fixed with 2.5% glutaraldehyde for another 4 h, washed with PBS, and dehydrated by washing with increasing concentrations of ethanol (30%, 50%, 70%, 85%, 90%) for 5 min each. The slides were finally washed with 100% ethanol for 15 min. The slides were examined with a scanning electron microscope (HR-SEM, SU8010) at 3.0-kV accelerating voltage.

Bacterial genomic DNA extraction and sequencing.

A. baumannii cells were collected from 10 mL overnight culture by centrifuging at 12,000 rpm for 2 min, followed by washing with TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8). The cell lysate was then treated with lysis buffer (40 mM Tris, 1 mM EDTA, 1% SDS, 20 mM sodium acetate [NaAc], pH 8) at 50°C for 1 h, RNase A (10 mg/mL, Thermo Fisher Scientific) at 37°C for 1 h to remove RNA, and then with 0.5 mg/mL proteinase K at 56°C for 1 h. Finally, bacterial genomic DNA was extracted using the phenol-chloroform method as previously described (45). The DNA library was constructed using the FS DNA library prep kit (ABclonal Science, Inc.). The quality and quantity of the DNA obtained were determined using a NanoDrop spectrophotometer (ND-2000; Thermo Fisher Scientific, Waltham, MA, USA). Genome high-throughput sequencing was performed with an Illumina NovaSeq 6000 sequencer, and 150-bp paired-end reads were generated.

Genome analysis. (i) Genome de novo assembly.

To achieve a high-quality bacterial genomic sequence, both third-generation sequencing (TGS) and next-generation sequencing (NGS) data were used for the de novo assembly. First, the initial draft genome was assembled from the TGS data using NextDenovo software (version: 2.1-beta, https://github.com/Nextomics/NextDenovo) with the ‘choice -g 4mb -d 30’ option. The recommended calculated value from the internal plugin seq_stat based on the TGS data was used as the seed cutoff value. Given a higher sequencing error rate of long-read sequencing than of short-read sequencing (46), NGS data were used to correct the draft genome assembled from TGS data using NextPolish software (version: 1.0.5) (47).

(ii) Whole-genome variation analyses of bacterial genomes.

Annotation of bacterial genes and pairwise whole-genome alignments were performed using Prokka and MAFFT (version: 7.471) software, respectively (48). Further, variations (mutations, insertions, and deletions) between wild-type and phage-adapted bacterial genomes were identified by an in-house script (https://github.com/MisgaXiong/Sripts_for_Paper/tree/master/A.baumanii_mutation).

(iii) Genetic polymorphism analyses of bacterial genomes.

Jellyfish software (version: 2.3.0) was used to calculate the sequencing error rate of NGS data (49). Reads of the clean NGS data (>Q30) of both wild-type and phage-adapted bacteria were aligned to the wild-type bacterial genome using BWA software (50). Besides mutation analysis, minor alleles (the less common allele for a single nucleotide site in the genome) were also analyzed. VarScan software (version: 2.3.9) was used to calculate the minor allele frequency (MAF) (51). Each site with a high likelihood of genetic polymorphisms (MAF value greater than 0.06 and sequencing depth greater than 100) was confirmed with IGV software (version: 2.7.2) (52).

(iv) Prediction of functional change in phage-adapted bacteria.

All minor alleles with a MAF value greater than 0.06, alongside mutated genes in the genome of adapted bacteria, were selected for Gene Ontology (GO) analysis using the R package clusterProfiler (R language version: 3.6.1; clusterProfiler version: 3.14.3) (https://cran.r-project.org/bin/windows/base/old/3.6.1/, clusterProfiler). The network analysis of GO enrichment results was performed with the CluGO plugin of Cytoscape software (Cytoscape version: 3.7.2) (53, 54).

Combined use of zinc acetate and phage to aid killing of resistant bacteria.

The experiment consisted of five different experimental groups: group 1, bacteria and phage at an MOI of 100, plus zinc acetate (Zn²⁺) added to a final concentration of 2 mM; group 2, bacteria and phage added at an MOI of 100; group 3, bacteria and Zn²⁺ added to a final concentration of 2 mM; group 4, only bacteria; and group 5, LB blank control. Bacterial growth in each group was continuously monitored at OD₆₀₀ using a spectrophotometer (BioTek, USA), up to 14 h.

Combined use of antibiotics and phage to aid killing of resistant bacteria.

First, MICs of antibiotics were determined. Stock solutions of colistin (Beijing Solarbio Science and Technology Co., Ltd., China; CAS 1264-72-8), polymyxin B (Beijing Coolaber Science and Technology Co., Ltd., CAS 1405-20-5), meropenem (Aladdin Biochemical Technology, Shanghai, China; CAS 119478-56-7), and imipenem (Yuanye Bio-Technology, Shanghai, China; CAS 64221-86-9) were prepared to a final concentration of 1.28 mg/mL. MICs were determined in a 96-well assay plate with a final bacterial concentration of ∼10⁵ CFU/mL, using the microbroth-dilution protocol (55). After incubation for 20 h at 37°C, MIC was determined as the antibiotic concentration at which no visible growth of bacteria was observed.

Bacterial growth was assessed to investigate the effects of using antibiotics with phage P21. The experiment, performed in a 96-well assay plate, included bacteria only (growth control), bacteria and antibiotics, bacteria and phage, and LB medium (blank control), as well as a combination of bacteria, phage, and antibiotics. Sublethal concentrations of antibiotics, one-fourth the MIC and one-half the MIC, were used, and phage and bacteria were added at an MOI of 1. Also, bacterial growth was assessed in the presence of combinations of zinc (final concentration of 2 mM), phage (MOI of 100), and sublethal concentrations of antibiotics. Plates were incubated at 37°C, and bacterial growth was tracked by checking the optical density at OD₆₀₀ using a microplate reader (BioTek, USA). The experiments were performed at two different times in triplicate.

Statistical analysis.

Data are presented as mean ± standard deviation (SD). All statistical analyses were conducted with GraphPad Prism (version 8.0.1) software (GraphPad Inc., San Diego, CA) and the R program language (version 3.6.3) (https://cran.r-project.org/bin/windows/base/old/3.6.3/). Differences in bacterial growth between Ev5-WHG and WHG40004 were assessed by Student’s t test. Correlations between the MAF value of each polymorphism site and sequencing depths in bacterial genomes were determined by Pearson correlation analysis. A P value of 0.05 or less was considered statistically significant.

Data availability.

The sequenced genomics data of bacteria and phage reported in this study have been deposited in the Genome Warehouse in the National Genomics Data Center https://ngdc.cncb.ac.cn/search/?dbId=&q=PRJCA007903 (56), Beijing Institute of Genomics (China National Center for Bioinformation), Chinese Academy of Sciences. The BioProject accession number is PRJCA007903. The accession number of raw sequencing data is CRA005814. The accession numbers of the submitted phage and bacterial genomes are GWHBHAN01000000 and GWHBHAO01000000, respectively.