Acinetobacter baumannii adaptation to the host pH microenvironment is mediated by allelic variation in a single residue of BauA protein

Abstract

Acinetobacter baumannii has been listed as one of the most critical pathogens in nosocomial infections; however, the key genes and mechanisms to adapt to the host microenvironment lack in-depth understanding. In this study, a total of 76 isolates (from 8 to 12 isolates per patient, spanning 128 to 188 days) were longitudinally collected from eight patients to investigate the within-host evolution of A. baumannii. A total of 70 within-host mutations were identified, 80% of which were nonsynonymous, indicating the important role of positive selection. Several evolutionary strategies of A. baumannii to increase its potential to adapt to the host microenvironment were identified, including hypermutation and recombination. Six genes were mutated in isolates from two or more patients, including two TonB-dependent receptor genes (bauA and BJAB07104_RS00665). In particular, the siderophore receptor gene bauA was mutated in multiple isolates from four patients with three MLST types, and all mutations were at amino acid 391 in ligand-binding sites. With 391T or 391A, BauA was more strongly bound to siderophores, which promoted the iron-absorption activity of A. baumannii at acidic or neutral pH, respectively. Through the A/T mutation at site 391 of BauA, A. baumannii displayed two reversible phases to adapt to distinct pH microenvironments. In conclusion, we demonstrated the comprehensive within-host evolutionary dynamics of A. baumannii, and discovered a key mutation of BauA site 391 as a genetic switch to adapt to different pH values, which may represent a model in the pathogen evolutionary adaption of the host microenvironment.

Introduction

Acinetobacter baumannii is an important opportunistic pathogen that can cause various infections, such as ventilator-associated pneumonia, bloodstream infections, skin and soft-tissue infections, wound infections, and urinary tract infections (1). During the last three decades, A. baumannii has become a leading cause of nosocomial infections (2), primarily due to the rapid spread of multidrug-resistant (MDR) A. baumannii strains in hospitals (3). MDR A. baumannii has been listed by the World Health Organization as one of the most critical pathogens in hospitals; in particular, few treatment options are available for patients infected with A. baumannii strains, which poses a serious problem for healthcare facilities.

Acinetobacter baumannii can persist in tissues, which leads to chronic infection. It is believed that two attributes, drug resistance and adaptation to the host environment, have enabled A. baumannii to thrive in patients for a long periods (1). In addition to antibiotics, A. baumannii must also overcome various challenges in the host, such as a lack of nutrient sources, constant changes in the microenvironment (e.g. pH), and immune response clearance. Bacteria have developed many strategies to adapt to diverse and ever-changing conditions. These mechanisms include multidimensional modifications within their genomes, such as point mutations and recombination.

Whole-genome sequencing (WGS) provides a powerful tool for pathogen genome analysis (4–6). It has widely been used to investigate the epidemiology of MDR A. baumannii infection outbreaks in hospitals (7) and academic burn centers (8), population and genome dynamics during colonization and infection (9, 10), the dissemination and variation of MDR A. baumannii strains (11), and the development of drug resistance (12, 13). However, due to the lack of isolates from chronically infected patients, the evolutionary dynamics and genetic basis for prolonged survival of A. baumannii in the host environment have rarely been studied.

In this study, we investigated the clinical course of eight inpatients with MDR A. baumannii chronic infections in the Fifth Medical Center of PLA General Hospital in Beijing, China. A total of 76 isolates sampled from the sputum of each patient were longitudinally collected. Through WGS and comparative analysis of all available strains, we sought to uncover the evolutionary dynamics of A. baumannii and discover the key adaptive genes and mechanisms in chronic persistent infection.

Results

Longitudinal collection and microevolutionary analysis of the isolates from patients

To develop a comprehensive genomic view of A. baumannii evolution within the host, we selected and collected isolates sampled from chronic inpatients in China. The selection criteria for the patients were as follows: availability of numerous longitudinally sampled isolates (>8 isolates); relatively similar isolated time intervals; and long isolate time spans (>120 days). Finally, we longitudinally collected 76 A. baumannii isolates sampled from eight patients (Table S1) who had long intensive care unit (ICU) stays and multiple clinical symptoms, including pneumonia. The number of isolates per patient was in the range of 8–12 during a time span of 128–188 days between the first and the last isolation (Fig. 1A, Table S2). The resistance profile of the strains was determined using standard antimicrobial susceptibility tests (Table S3). The results showed that all of the isolates, except for the strain P7_T05, were resistant to more than three classes of antibiotics, including commonly used drugs such as cefoperazone/sulbactam (SCP), ciprofloxacin (CIP), and meropenem (MEM); these 75 strains were considered MDR A. baumannii.

To identify genomic changes that accumulate during longitudinal infection, the whole genomes of 76 isolates were sequenced by an Illumina HiSeq 2000 platform to >200-fold coverage. A core phylogeny tree was constructed using kSNP3 software (14). A total of 58,558 single-nucleotide polymorphisms (SNPs) were identified, which were used to build a phylogenetic tree. All strains were categorized into five major clades (clades A–E), except for the strain P7_T05 (Fig. 1B, Table S2). In addition, we performed multilocus sequence typing (MLST) analysis for these 76 strains in accordance with the Oxford MLST typing system (15). Four MLSTs were found, except in the strain P7_T05 (ST819), which was in agreement with the clade analysis results based on whole-genome SNPs (Fig. 1B).

Then, we carefully compared the genomes of the isolates from each patient to infer the microevolution of MDR A. baumannii in each patient. The three temporary acquisition strains (P7_T05, P8_T06, and P8_T07) were excluded from the following analysis. Positions differing from the first isolate (T01) in each patient were considered mutation sites, and the SNPs within each patient were identified. A total of 78 mutation sites were found in these eight patients, and all of the mutations were annotated in accordance with the reference genome BJAB07104 (accession number: NC_021726). All of the mutations were verified by PCR amplification and Sanger sequencing. Eight mutations in the isolates from patient P1 were clustered in the 44 bp genome sequence from site 1653441 to site 1653484 (Fig. S1), which was identified as a recombination event using Gubbins (16). The recombination event was identified in the gene BJAB07104_RS08095, which encodes a putative secretion-activating protein (Fig. S1). Although recombination has already been reported as a common phenotype in A. baumannii (17), to the best of our knowledge this is the first time that it has been described in the evolution within the host. The remaining 70 SNPs included 56 nonsynonymous mutations (NS), eight synonymous mutations (S), and six mutations in noncoding regions (NC). Hence, there was a high proportion of NS relative to S during within-host evolution [NS/(NS + S), 0.91 ± 0.09]. The mutations were significantly more common in the coding regions (Table S4, P = 0.02), and there were significantly more nonsynonymous than synonymous mutations (P < 0.01), indicating the dominance of positive selection during the within-host evolution of A. baumannii.

The rate of diversification was calculated for each patient using the mean pairwise SNP distance (18); a large variation was found among the isolates from the eight patients (8.8 ± 7.1 SNPs genome⁻¹ y⁻¹; Fig. 1C). In particular, P4, P6, P7, and P8 were infected by a common clade D A. baumannii strain; however, the rate of diversification in patient P7 (1.3 SNPs genome⁻¹ y⁻¹) was much lower than that in the other three patients (P4: 6.8; P6: 6.5; P8: 4.7 SNPs genome⁻¹ y⁻¹). The results showed that the host microenvironment may be an important factor contributing to the diversification rate of A. baumannii. In one patient (patient P2), the diversification rate (27.5 SNPs genome⁻¹ y⁻¹) was significantly higher than that in the isolates from other patients (Fig. 1C, P < 0.01, Grubbs’ test for outliers). Then, we found that all of the isolates from patient P2 contained a single nonsynonymous mutation P118S in the DNA repair relative protein MutS (Fig. S2). Mutation site 118 is at a conserved position in MutS domain I in MutS (Fig. S3), and its defects are known to cause excess mutations (19). The results suggested the presence of hypermutation in the isolates from patient P2. Although hypermutation is a common phenotype in many pathogens, such as Pseudomonas aeruginosa in cystic fibrosis patients, and it is hypothesized to accelerate the evolution of antibiotic resistance, it has not previously been described in A. baumannii.

The within-host mutation events of the isogenetic strains across multiple patients

Microevolutionary analysis of pathogen genomes (20) was applied to investigate transmission. As shown in the results of the phylogenetic analysis (Fig. 1B), there were 33 strains from four patients, and they all belonged to the same clade D. All these four patients were in a respiratory ward, and there was an overlap in their hospitalization time, which might indicate putative transmissions or the infection of a common ancestry strain in the hospital environment (5) (Fig. 1D). By tracking the mutations detected in the longitudinally collected genomes from each of the patients, we constructed the complete landscapes of the sequence of mutation events from the first to the last isolate in different patients (Fig. 1E). The clade D strains had distinct mutation genes, sites, and dominant clone types among the four patients (Fig. 1E). In patient P4, the strains evolved into a diverse community, including multiple clone types with different mutations. In patient P6, two dominant clone types appeared simultaneously in the strains, which may reflect two distinct host microenvironments in different infection sites. In contrast, different dominant clone types appeared successively in the strains from P8 to adapt to the changes in the host microenvironment. In the strains from P7, the initial strain was continuously the dominant clone type. All these findings further confirm the crucial role of the host microenvironment in the evolutionary within-patient dynamics of A. baumannii.

Parallel mutated genes involved in the within-host evolution of A. baumannii

To observe the within-host evolution of A. baumannii at the gene levels, we sought to identify genes that showed parallel evolution in different hosts. Finally, a total of six genes (bauA, wzc, ntrC, AdeJ, BJAB07104_RS00665, and BJAB07104_RS16200) were mutated in multiple patients or MLST types (Fig. 2A). These six genes were more frequently mutated than what would be expected under random conditions (P < 0.01), which suggests that parallel mutations in these six genes are the result of positive selection undergoing adaptive evolution. The gene wzc (also referred as ptk) was identified with three nonsynonymous mutations (ST368 in P1: A540G, ST381 in P3: L567F, G667D; Fig. 2B), encodes a putative protein tyrosine kinase, and is required for capsule formation (21). All of the mutations were located in a conserved region of Wzc (pfam10609: ParA domain; Fig. 2B). The results showed that none of the three mutations (G667D, L567F and A540G) affected the amount of the capsule (Fig. S4A), but two mutations (G667D and A540G) significantly changed the capsule diameters (Fig. S4B). The mutation G667D, which was also identified in a previous report (22), significantly increased the capsule diameters (P < 0.01). In contrast, the mutation A540G significantly decreased the capsule diameters (P < 0.01). Capsule formation is important for immune escape of pathogens during infection and is a major virulence factor in A. baumannii infection (23, 24). These adaptive mutations in the Wzc protein might bring some advantage for the strain to escape the host immune response. AdeJ is a part of a three-component RND efflux system AdeIJK in A. baumannii, which contributes to the resistance to multiple antibiotics (25). Two mutations were identified in the adeJ gene, and both of them resulted in a premature stop codon and gene inactivation (ST218 in P2:W13*, ST195 in P4:Q68*). These two mutations were only found in a single isolate and were not identified in the subsequent isolates in two patients, indicating that these two mutations served for temporary adaptation to a special host environment. A transcriptional regulator gene ntrC was mutated in two clades (ST218 in P2: L228F, ST195 in P6: L187I). NtrC is a part of a two-component system, which has been reported to be an important regulator during stress response (26). Thus, the nonsynonymous mutation in NtrC may modulate the expression of regulated genes, which is helpful for the survival of A. baumannii in the host. The gene BJAB07104_RS16200 that codes a protein with a domain of LPS biosynthesis glycosyl-transferase was mutated in multiple isolates from two patients (ST368 in P1:S8R, ST195 in P8:A159T). In particular, two genes (BJAB07104_RS00665, bauA) encoding the domain of TonB-dependent receptor (pfam00593: TonB_dep_Rec) were mutated in the isolates from different patients. The gene BJAB07104_RS00665, which encodes the putative receptor for an unknown substrate, was mutated in the isolates of patients P2 and P4 (ST218 in P2:D111A, ST195 in P4: synonymous mutation in site 643; Fig. 2B). The gene bauA was mutated in the isolates from four patients with three MLST types (ST368 in P1 and P5: T391A, ST218 in P2: A391T, ST381 in P3: A391T; Fig. 3A), and it encodes an acinetobactin (or preacinetobactin) outer membrane receptor protein (27). The mutation (A391T) has also been identified in A. baumannii isolates in previous research (22). They were all mutated at the same amino acid position (site 391) in a conserved domain of TonB_dep_Rec (pfam00593; Fig. 3A). Interestingly, the mutations both A391T and T391A were found in multiple isolates from four patients (P1, P2, P3, and P5; Fig. 3A). The results showed that site 391 of BauA was involved in the within-host evolution, which may be a key mutation site for the host adaptation of A. baumannii.

BauA-391 mutation alters the pH-related iron-absorption activity of A. baumannii

BauA is a multiple transmembrane protein receptor for preacinetobactin (and acinetobactin), which is a major siderophore utilized by the human pathogen A. baumannii (28). In addition, 2D and 3D topology models of the BauA protein showed that site 391 is located in the ligand-binding sites of BauA (364–392 amino acids, Fig. 3B), which is in the fifth loop of the extracellular region (Fig. S5). These data suggested that BauA's T391A or A391T mutation may affect the iron acquisition activity of A. baumannii. Acinetobacter baumannii is capable of growing over a wide pH range, with most sites of infection being acidic (29); thus, the activity of siderophore-mediated iron acquisition (Fig. 3C) was identified at pH from 5.8 to 7.2. Interestingly, the levels of ⁵⁵Fe³⁺ uptake of the clinical isolates from P2 or P5 were significantly different in a neutral (pH 7.2) or acidic environment (pH 5.8 or 6.5; Fig. 3D, P < 0.01). The six clinical isolates (from P2-T01 to P2-T06) with BauA-391A acquired more ⁵⁵Fe³⁺ at pH 7.2; in contrast, the other six clinical isolates (from P2-T07 to P2-T12) with BauA-391T acquired significantly more ⁵⁵Fe³⁺ at pH 5.8 or 6.5 (Figure 3D, left panel). Similar results were also detected in the 10 clinical isolates from P5 (P5-T01 to P5-T10, Fig. 3D, right panel). Then, we further confirmed the results with the same genetic strains. The single amino acid mutant of the BauA 391 site (A391T, named 391T) or the deleted mutant of the bauA gene (named ΔbauA) of strain P5_T08 (named 391A) was generated using a suicide plasmid recombination system (30), and the activity of iron acquisition was detected and compared (Fig. 3E). The levels of ⁵⁵Fe³⁺ uptake of the strain with 391A were significantly higher than those in the strain with 391T at pH 7.2 (P < 0.01); in contrast, they were significantly lower at an acidic pH (at pH 5.8 or 6.5, P < 0.01). The activity of ΔbauA was significantly reduced compared with the two mutants at both neutral and acidic pH. The results demonstrated that the mutation between 391A and 391T in BauA promoted the iron-absorption activity of A. baumannii at different pH values to adapt to an iron-deficient environment.

The mutations 391A and 391T enhance BauA's binding ability to siderophore at acidic or neutral pH

To uncover the molecular basis of the activity of BauA site 391, two types of BauA protein with 391T and 391A were expressed and purified, and the interactions between BauA and siderophore-Fe³⁺ (2:1) at different pH values were identified by isothermal titration calorimetry (ITC; Fig. 4A). The results showed that the combination of BauA and siderophore-Fe³⁺ was greatly influenced by pH. For example, at pH 7.2, the interaction between BauA and the siderophore-Fe³⁺ complex was an exothermic reaction; in contrast, the binding reaction was endothermic at pH 5.8 or 6.5 (Fig. 4A). More importantly, the mutations between 391A and 391T remarkably enhanced the affinity of BauA and siderophore-Fe³⁺ at acidic or neutral pH, respectively. The binding affinity of BauA-391A was higher than that of BauA-391T at pH 7.2 (Kd: 2.47E-6 vs. 2.67E-5) and lower than that at acidic pH (Kd: 2.14E-7 or 4.06E-6 vs. 3.78E-5 or 1.30E-5 at pH 5.8 or 6.5). These results revealed that through the opposite mutations at BauA-391, A. baumannii was able to increase its own iron levels in response to the challenges of pH change. When environmental pH changed from neutral to acidic, the binding affinity between siderophore-Fe³⁺ and BauA-391A gradually increased, thereby improving the iron-absorption capacity and intracellular iron concentration (Fig. 4B). In contrast, when pH values changed inversely, through a point mutation from 391A to 391T, A. baumannii maintained strong siderophore-Fe³⁺ binding ability and acquired more exogenous iron ions.

BauA-391T/A variants are dominant in distinct pH microenvironments

To address the character of BauA-391T/A variation, we investigated the sequences of protein BauA from different strains of A. baumannii in the NCBI database (Fig. S6). The BauA proteins of A. baumannii are divided into three clades based on sequence similarity (Fig. S6A and S6E), and the BauA proteins of each clade are encoded by the strains of different MLST types (Fig. S6B–D). Among the three clades, clade I is with the largest number of strains and protein sequences (Fig. S6A and S6E). Many MLST types (e.g. ST208, ST195) that have recently been reported to be prevalent in MDR A. baumannii belong to clade I (Fig. S6C) (31–33). The BauA proteins of 75 MDR strains in this study belong to clade I, and the sequence homology with clade II or III is 69 to 70% or 58 to 59%, respectively (Fig. S6F). Especially, site 391 is the only high-frequency mutation site in BauA proteins of clade I (Fig. S6G). Among the 12,133 sequences in clade I, the frequencies of two amino acid residues (T and A) at BauA 391 site are 56.5 and 40.2%, respectively (Fig. S6H). The high proportions of these two amino acids in nature further emphasize the importance of this mutation site in the environmental persistence of the clade I strains of A. baumannii.

To better understand the physiological significance of the mutation between 391A and 391T, the growth and infection levels of the two mutants were examined at different pH values. The growth curves of A. baumannii with different mutations were compared in an iron-excess or iron-deficient LB medium. In an iron-excess LB medium, there were no significantly differences between the three strains in all the three pH values (Fig. S7A). In an iron-deficient LB medium, the strain with BauA-391A grew significantly faster at pH 7.2; in contrast, the strain with BauA-391T grew significantly better at pH 5.8 or 6.5 (P < 0.01, Fig. 5A). ΔbauA was associated with significantly slower growth compared with the two mutants at neutral and acidic pH (P < 0.01). Similar results were observed in the iron-deficient M9 medium (Fig. S7B). Then, the competitive advantage of the strains with BauA-391A or 391T at different pH values was compared in a competitive growth experiment (Fig. 5B). With the successive passage of the mixed strains (1:1), the percentages of BauA-391A mutant gradually increased at pH 7.2 and progressively decreased at pH 5.8 or 6.5 (Fig. 5C). Then, the two mutants of A. baumannii were used to infect human lung epithelial cells A549. Plating gentamicin (Gm)-treated cell lysates on an LB medium were used to count the number of intracellular bacteria (Fig. 5D). The number of intracellular bacteria with BauA-391A was significantly higher than that with BauA-391T at pH 7.2 (P < 0.01), and significantly lower than that with BauA-391T at pH 5.8 or 6.5 (P < 0.01). Similar results were observed in human macrophage cells THP-1 (Fig. S8). Then, the infection with an equal mixture of the two mutants at different pH values was compared in a competitive infection experiment (Fig. 5E). The percentages of intracellular BauA-391T mutants increased from 50 to 71.3% at pH 5.8 (66.9% at pH 6.5); in contrast, they evidently decreased from 50 to 33.8% at pH 7.2 (Fig. 5F). These results indicated that through the A/T mutation at BauA site 391, A. baumannii displayed two reversible phases to adapt to distinct pH microenvironments (Fig. 5G). Based on its competitive advantage in iron-deficient environments, such as infected human organs or cells, the BauA-391A mutant was dominant in acidic pH environments, while the 391T mutant was dominant in neutral pH environments. The dynamic equilibrium between two BauA 391 mutants and the microenvironmental pH promotes the persistent survival and infection of A. baumannii.

Discussion

Multidrug-resistant A. baumannii has become one of the most critical resistant pathogens worldwide. However, there is limited information about its evolution among patients. In this study, 76 isolates (8–12 isolates per patient, spanning 128–188 days) were longitudinally collected from eight patients to investigate the dynamics and mechanisms of within-host evolution using WGS. Longitudinal and high-density continuous sampling helped us to obtain many intermediate isolates, which provided high-resolution insight into the degree of heterogeneity between isolates. Furthermore, 33 isolates were from transmission between four patients, which provides a rare opportunity to compare the evolutionary dynamics of the isogenetic strains across multiple patients. Surprisingly, there was significant variation in the diversification rates, evolutionary routes, and dominant clone types in the strains from the four patients, which revealed the key role of host factors in the within-patient evolution of A. baumannii. In addition, we found a high ratio of nonsynonymous relative to synonymous mutations during the within-host evolution of A. baumannii in the lungs of these patients (0.91), which is remarkably higher than the reported ratio at the population level of A. baumannii (0.18) (10). The mutations were significantly more frequent in protein-coding regions, and multiple genes were identified with parallel mutations in the isolates from different patients. These results strongly suggest that the effect of positive selection is predominant for the within-host evolution of A. baumannii.

In this study, we identified multiple within-host evolutionary mechanisms of A. baumannii that increase the evolutionary potential to adapt to host environment. An important mechanism used by pathogens is hypermutation, which accelerates the rate of adaptation by increasing the mutation rate. In this study, all of the 12 isolates from patient P2 harbored a mutation P118S in the MutS protein; these isolates were hypermutators, with a 5.1-fold increased mutation rate (P2 vs. other seven patients: 27.5 vs. 5.4 SNPs genome⁻¹ y⁻¹), which greatly increased the evolutionary potential to adapt to the novel host environment. Recombination is another factor that can lead to faster diversification of the genome than point mutations alone. In the isolates from patient P1, eight mutations were simultaneously induced by a recombination event, which suggests that recombination is also a strategy to increase the supply of genetic novelty for A. baumannii.

In the host, the pathogens such as A. baumannii must overcome various challenges, including limited nutrient sources. Hosts have evolved elaborate mechanisms to sequester nutrients, such as iron, a process that is referred to as nutritional immunity (1). Acinetobacter baumannii responds to iron starvation by modifying gene expression for genes involved in various processes such as respiration, biofilm formation, and motility, highlighting the importance of iron levels to A. baumannii survival and infection (34, 35). In this study, we observed two nutrition-acquisition-related TonB-dependent receptor genes (bauA and BJAB07104_RS00665) mutations in multiple isolates from different patients, discovering that “nutritional immunity” is an important factor in the within-host evolutionary adaptation of A. baumannii. The siderophore receptor gene bauA was mutated in multiple isolates from four patients. All of these mutations were at amino acid site 391 in the ligand-binding sites of BauA, with the mutation between A and T. Infected body sites can have a significantly lower pH than healthy tissues (29). Acinetobacter baumannii also needs to adapt to different pH values during chronic infection, especially in the acidic microenvironment. This observation is consistent with the discovery of the function of BauA in this study. With two different amino acids (T or A) at site 391, the strains of A. baumannii have distinct activities of siderophore-dependent iron acquisition, and reveal growth advantage at acidic or neutral pH, respectively. A comparable mechanism is that of phase variation, which in general refers to a reversible switch between an “all-or-none” (on/off) expressing phase, resulting in variation in the level of expression of one or more proteins between individual cells of a clonal population (36). Similar to the “phase variation,” A. baumannii with BauA-391A or BauA-391T also reveals two reversible phases to adapt to acidic and neutral environments. Different from the “phase variation,” the genetic basis of acidic and neutral phases of A. baumannii is the allelic variation in a single residue of BauA protein. It is worth noting that the BauA proteins of A. baumannii are divided into three clades based on sequence similarity (Fig. S6). The findings of BauA-391A or -391T in this study were limited to the strains belonging to clade I.

The respiratory tract is the main site of infection and colonization of A. baumannii in patients. Airway surface liquid (ASL) is an important microenvironment for A. baumannii infection. Many studies reported that the ASL pH varied during pathogen infection in patients (37). One of the studies showed that the respiratory mucus pH was associated with the onset of pneumonia symptoms in the ICU patients (38). In the patients who developed initial colonization and then pneumonia, the pH did not change after colonization, but became acidic with the development of pneumonia. The pH returned to basal levels after recovery from pneumonia. In this study, multiply patients, such as P2 and P5, had recurrent episodes of pneumonia with intermittent fever, and the pneumonia or fever improved with antibiotics. The temperature curves of patients P2 and P5 over time are shown in Fig. S9. These disease presentations suggest that the ASL pH of patients P2 and P5 might vary with the reduction, disappearance, and reappearance of pneumonia symptoms. To adapt to the changing ASL pH microenvironments, A. baumannii displayed two reversible phases through the opposite mutation at BauA site 391. At different pH microenvironments, the proportions of the isolates from the two phases were distinct, as shown in Fig. 5G. However, the clinical strains used in this study were monoclonal, were isolated and randomly collected from the sputum cultures of patients, and thus were mainly representative isolates of the dominant phase. The isolation and culture process used to produce clinical strains have reduced their complexity compared with that of the bacterial community at the site of infection. Therefore, it is necessary to conduct further studies on uncultured clinical infection samples using metagenomes in the future.

In conclusion, our genomic analysis of the longitudinally collected isolates from eight patients revealed the great variation in the rate of diversification and the multiple evolutionary strategies such as hypermutation and recombination, increasing the understanding of the within-host genetic dynamics of A. baumannii. In addition, we found multiple adaptive genes and mutation sites related to within-host evolution, which could be targeted to further study the mechanism of A. baumannii adaption to host. Importantly, we discovered that an adaptive site of BauA site 391 was an important regulator for A. baumannii to adapt to different pH levels, which depicts a model for bacteria adaptation to a constantly changing environment.

Materials and methods

Plasmids, bacteria strains, cells, and reagents

The expression vector pET-28a and pMo130-TelR plasmids were stored in our laboratory. A549 and THP1 cells were also from our laboratory. A HisTrap Fast Flow column (5 mL) was purchased from GE Healthcare (Beijing, China). Taq DNA polymerase and T4 DNA ligase were purchased from New England Biolab. PCR primers were synthesized by GENEWIZ (China). The bacteria genomic DNA extraction kits and plasmid kits were purchased from Qiagen. ⁵⁵FeCl₃ solution was from Perkin Elmer. Acinetobactin and preacinetobactin were synthesized by Sks Chem Company. All other chemicals and reagents were obtained from commercial sources and were of the highest purity available.

Patients and bacterial isolates

The patients attending the Fifth Medical Center of PLA General Hospital were closely followed; their sputum samples were tested regularly for the presence of A. baumannii, and the identified isolates were stored frozen for further study. Isolation and identification of A. baumannii were carried out as previously described (39). A retrospective collection of clinical information about the patients was conducted from the hospital medical records; the collected data included age, admission date, inpatient ward, diagnoses, medications, clinical course, and isolation date.

Antimicrobial susceptibility tests

Clinical strains were grown on blood agar plates, and the minimum inhibitory concentrations of antibiotics were determined for all isolates using BD Phoenix-100 Automated Microbiology System. The antibiotics included amoxicillin/clavulanic acid, amikacin, aztreonam, ceftazidime, chloromycetin (C), CIP, ceftriaxone, cefuroxime, cefazolin, ertapenem, cefepime, fosfomycin, cefoxitin, gentamicin, imipenem, levofloxacin, MEM, minocycline, colistin, nitrofurantoin, moxifloxacin, tobramycin, norfloxacin, ampicillin/sulbactam, SCP, trimethoprim/sulfamethoxazole, tetracycline, piperacillin/tazobactam, and tigecycline. The results were interpreted using Clinical and Laboratory Standards Institute breakpoints.

Whole-genome sequencing

Overnight cultures were grown in LB broth, and genomic DNA was extracted using Wizard Genomic DNA Wizard Genomic DNA Purification Kit (Promega, WI, USA). Paired-end libraries (500 bp fragments) were constructed for all of the strains and were sequenced using an Illumina HiSeq-2000 platform to obtain a 100-bp paired-end read. An average of 9,691,933 reads (range, 5,557,016–17,200,862 reads) were obtained for each of the genomic libraries. In general, we sequenced the isolates with an average genomic coverage of 245-fold (range, 140- to 435-fold). The Illumina short reads were assembled using VELVET (40) and annotated using NCBI's PGAAP pipeline (http://www.ncbi.nlm.nih.gov/genome/annotation_prok/). All of the genomes are available at NCBI under BioProject PRJNA355052.

Bioinformatics analysis

Single-nucleotide polymorphisms between all of the isolated strains were identified by kSNP3 (14) software using the default parameters. A core phylogenetic tree was inferred using SNPs in at least 80% of the genomes and was constructed using the maximum parsimony method in kSNP3 software. The Illumina sequence reads were aligned to the A. baumannii reference genome BJAB07104 (39) (accession number: NC_021726; genome size of 3.9 Mb) using bwa v0.7.10 (41) software to identify SNPs between strains within each patient. The sites with a mapping quality of <30 and <5 paired-end reads were removed using SAMtools v0.1.19 (42) and homemade Perl script. The remaining SNPs were used to construct the phylogenetic tree via the maximum-likelihood method by RAxML v8.1.6 (43) software (1,000 bootstrap replicates), with the GTR model of nucleotide substitution with the gamma model of rate heterogeneity. Recombination events were investigated using Gubbins (16). Prediction of the hydrophobic transmembrane regions in BauA was done with PRED-TMBB (http://biophysics.biol.uoa.gr/PRED-TMBB/). The webserver was used to find the topology of the loops in addition to localizing the transmembrane strands (15). All the 78 within-host mutations were verified by PCR amplification and Sanger sequencing.

Expression and purification of BauA proteins

To express 6×His fusion proteins, the gene bauA (391A and 391T) was cloned into pET-28a and expressed in Escherichia coli BL21(DE3). Bacterial culture was grown to an OD₆₀₀ of 0.6 at 37°C in Luria–Bertani (LB) broth; IPTG was added to 0.1 mM, and then cultivated with bacteria with shaking for 20 h at 20°C. Bacteria were collected and crushed by an ultrasonic cell disruptor. Supernatant proteins were purified by Ni-NTA Sepharose in accordance with the manufacturer's protocol and were identified by SDS-PAGE. Finally, target proteins were saved in dialysis MES or HEPES buffer (pH 5.8, 6.5, or 7.2) and stored at −80°C.

Construction of the mutants of A. baumannii

To construct the single amino acid mutant of BauA site 391 (A391T, named 391T) or the bauA gene deleted mutant (named ΔbauA) of A. baumannii strain P5_T08, we used a suicide plasmid recombination system (30). Specifically, 1,000 bp homologous regions that were cloned from A. baumannii by PCR and SacB gene were inserted into plasmid pMo130-TelR to construct a recombination vector. With the work of sucrose pressure, the target gene bauA was knocked out or replaced. The mutants were obtained and identified by PCR method and Sanger sequencing.

Capsule detection and diameter measurement

The number of capsules was detected using the same method in a previous report (44). The membrane extracts standardized by Bradford assay according to protein concentration were separated on 10% SDS-PAGE. The total proteins were visualized using Coomassie blue stain, and cis-diol sugars were visualized using periodic acid-Schiff stain. Bacterial capsules were detected using the India ink method (22). Briefly, bacterial colonies and India ink were mixed in 0.85% NaCl solution, and the bacteria were observed and photographed under a light microscope (1,000×). The capsule diameters were measured for each isolate using image J software (https://imagej.nih.gov/ij/).

In vitro iron uptake assays

In vitro iron uptake assays were performed as previously described (45). Briefly, acinetobactin and preacinetobactin were synthesized as previously described (46) and dissolved in MOPS buffer. Acinetobactin or preacinetobactin was incubated with ⁵⁵FeCl₃ at 37°C for 30 min; 50 µL phosphate sodium buffer was added to remove free iron, and the supernatant was carefully taken after centrifuge. The bacteria were cultured overnight and were collected via centrifugation at 10,000 g. The bacterial deposits were washed with cold MOPS three times and were resuspended in MOPS buffer containing 2 g L⁻¹ glucose to an OD₆₀₀ of 0.6. The bacteria were preheated at 37°C for 5 min and mixed with acinetobactin or preacinetobactin-⁵⁵FeCl₃ complex. Next, 100 µL aliquots were taken out at 30 min, and we immediately added 1 mL EDTA solution to stop the reaction. The bacteria were centrifuged and washed with 0.85% NaCl, and then 100 µL lysis solution was added for 5 min at 80°C. Finally, the bacterial deposits were incubated with scintillation solution for 12 h in dark and then detected with scintillation analyzer (Perkin Elmer).

Isothermal titration calorimetry

Isothermal titration calorimetry experiments were carried out on an Affinity ITC instrument at 25°C. The BauA-391A and BauA-391T protein concentrations were 20 µM and were dissolved at pH 5.8, 6.5, or 7.2 buffers. Acinetobactin or preacinetobactin was dissolved in the corresponding buffer as proteins with a 200 µM concentration. All binding experiments consisted of the first injection of 2 mL and a 60 s delay. The subsequent 20 injections were of 8 mL at a 200 s interval. The ITC data were analyzed using Launch NanoAnalyze. ITC experiments were done in duplicate.

Bacterial growth and infection of cells at different pH values

The 391A and 391T mutant strains were grown overnight and diluted to a final OD₆₀₀ of 0.01 in M9 or LB medium with 0, 50, or 200 µM DIP adjusted to pH 5.8, 6.5, or 7.2. The bacteria were cultured in 96-well plates with shaking at 37°C, and the OD₆₀₀ value was monitored hourly for 12 h. All of the experiments were performed in triplicate. In the infection experiment, bacterial strains were cultured overnight at 37°C and suspended in PBS; after the cells grew to 90%, we removed the culture medium and replaced it with serum-free DMEM medium (pH 5.8, 6.5, or 7.2). Bacterial cells with a multiplicity of infection of 50 were added to A549 cells or THP-1 cells and incubated at 37°C for 4 h. Then, the infected monolayers were washed three times with PBS. To determine intracellular bacterial persistence, the monolayers were then treated with 2 mL of gentamicin (10 mg mL⁻¹ in PBS) for 30 min, the shortest time point that resulted in the death of all bacteria added to the monolayers. The cells were lysed with 0.1% Triton X in PBS; appropriate lysate dilutions were plated on LB agar containing no antibiotics; and the numbers of colony-forming units (CFUs) were recorded after overnight (10 to 12 h) incubation at 37°C. Duplicate assays were done at least three times using fresh samples each time. The data were statistically analyzed using the Student’s t test. P-values <0.05 were considered statistically significant.

The competitive growth and infection experiment

The 391A and 391T mutant strains were grown overnight, equally mixed (1:1), and diluted to a final OD₆₀₀ of 0.01 in LB medium containing 200 µM DIP adjusted to pH 5.8, 6.5, or 7.2. The bacteria were cultured at 37°C for 12 h, and they were divided into two parts. The first part was used to extract genomic DNA. The gene bauA was PCR-amplified and sequenced using an Mi-seq platform to detect the percentages of BauA-391T and BauA-391A. The second part was diluted to a final OD₆₀₀ of 0.01, and four successive passages, sampling and sequencing were performed. In the competitive infection experiment, equal mixture of two mutants BauA-391A and BauA-391T was added to A549 cells at pH 5.8, 6.5, or 7.2, and incubated at 37°C for 4 h. The infected cells were lysed after gentamicin treatment; gene bauA was PCR-amplified and sequenced to detect the percentages of BauA-391T and BauA-391A.

Statistics

Data were described with means and standard deviation. Dichotomous variables were described with percentages. The quantitative data were compared using the paired Student’s t test. P-values <0.05 were considered statistically significant, and all P-values were two-tailed. We conducted all statistical analyses using GraphPad Prism version 6.01.

Ethics Approval Statement

The study was reviewed and approved by the Fifth Medical Center of PLA General Hospital Research Ethics Committee. Microbiology isolation and identification were routine work in our hospital. Only strains that have been routinely collected for diagnosis were studied. No extra sampling from the patients was performed. The Fifth Medical Center of PLA General Hospital Research Ethics Committee waived the need for written informed consent from the participants due to the deidentified secondary data analyzed in this study.

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Funding

This work was supported by National Natural Science Foundation of China (82072251), National Key Basic Research Program (973) of China (2015CB554202), and Beijing Nova Program (Z181100006218111).

Author Contributions

H.W. and T.L. designed the study. C.B., F.Q., and B.L. provided the isolates and patient information. N.N. and X.L. extracted DNA of the isolates. X.L. conducted the genome assembly, analysis, and data deposition. L.Z., Z.L., D.L., H.G., X.Y., and Y.H. participated in data analysis and discussion. F.C., T.L., and N.N. performed the experiment about BauA and Wzc. The manuscript was written by T.L. and X.L., and modified by D.L. and H.W. All authors read and approved the final manuscript.

Data Availability

This Whole Genome Shotgun project has been deposited in GenBank under BioProject PRJNA355052. The genome sequences and annotation were available at the NCBI (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA355052).

References

Author notes