Rapid and Integrated Bacterial Evolution Analysis unveils gene mutations and clinical risk of Klebsiella pneumoniae

Abstract

Bacteria continually evolve. Previous studies have evaluated bacterial evolution in retrospect, but this approach is based on only speculation. Cohort studies are reliable but require a long duration. Additionally, identifying which genetic mutations that have emerged during bacterial evolution possess functions of interest to researchers is an exceptionally challenging task. Here, we establish a Rapid and Integrated Bacterial Evolution Analysis (RIBEA) based on serial passaging experiments using hypermutable strains, whole-genome and transposon-directed sequencing, and in vivo evaluations to monitor bacterial evolution in a cohort for one month. RIBEA reveals bacterial factors contributing to serum and antimicrobial resistance by identifying gene mutations that occurred during evolution in the major respiratory pathogen Klebsiella pneumoniae. RIBEA also enables the evaluation of the risk for the progression and the development of invasive ability from the lung to blood and antimicrobial resistance. Our results demonstrate that RIBEA enables the observation of bacterial evolution and the prediction and identification of clinically relevant high-risk bacterial strains, clarifying the associated pathogenicity and the development of antimicrobial resistance at genetic mutation level.

Fig. 1. Associations among HMV, serum susceptibility, and gene mutation frequency in the Kp clinical isolates.

a Influence of bacteraemia caused by Kp infection on 60-day mortality. Multivariate logistic regression analysis (two-sided) demonstrated that Kp bacteraemia was associated with increased 60-day mortality independent of these confounders. No multiple-comparison adjustments were applied. A p-value of < 0.05 was considered statistically significant. b Numbers of HMV-Kp and non-HMV-Kp isolates derived from clinical specimens. The percentage indicates the prevalence of HMV-Kp isolates. c Susceptibility of human serum. The values indicate the minimum inhibitory concentration (MIC) of human serum (%, vol/vol in MHBII). d Distribution of serum MICs. The y-axis shows the prevalence of isolates among the HMV and non-HMV groups. The dotted red line indicates the breakpoint of serum resistance. No significant difference in serum resistance was observed between the HMV and non-HMV groups by two-sided Fisher’s exact test (p = 0.5574). e Frequency of gene mutation in the Kp clinical isolates determined via the rifampicin assay. We defined mutators as low (< 5 × 10⁻⁸), moderate (from 5 × 10⁻⁹ to 10⁻⁸), high (from 10⁻⁸ to 10⁻⁷), and hyper (> 10⁻⁷) according to their mutation. A hypermutator was identified from a clinical respiratory sample (non-HMV isolate SMKP590; mutation frequency: 4.43 × 10⁻⁶). The geometric means were indicated, and no significant differences in gene mutation frequency were detected among each isolation site by the two-sided Kruskal‒Wallis test (p = 0.2519). f Comparison of the mutation frequencies of the HMV-Kp (n = 26) and non-HMV-Kp (n = 241) clinical isolates. The value of the hypermutator non-HMV strain was removed to evaluate the majority. The geometric means and geometric standard deviations are given, and a two-sided Student’s t test was used for the statistical analysis. g Core genome SNP analysis of respiratory Kp clinical isolates. The 84 Kp clinical isolates identified by the MALDI Biotyper were 75 Kp strains, eight Klebsiella quasipneumoniae strains, and one Klebsiella variicola strain, as determined by average nucleotide identity (ANI) analysis. HMV (string test-positive) isolates were classified into several STs (ST23, ST39, ST65, ST86, ST218, ST268, ST458, ST893, and some novel STs) and are shown in the red square. We included the carbapenem-resistant Kp strains of ST258 from the NCBI database (shown in orange) as references.

Fig. 2. Serial passaging experiments and TraDIS analysis in human serum using a non-HMV-Kp SMKP838.

a, b Serial passaging experiments with the parent (wild-type, WT) and mutS mutant (ΔmutS) strains in the presence of human serum. The red circles indicate the wells in which the strains/mutants grew. The geometric means and geometric standard deviations from three independent experiments are given. c Venn diagram of the gene mutations accumulated in SMKP838ΔmutS clones (#1 to #3) during the serial passaging experiments in the presence of human serum after 20 days. The accumulated gene mutations are listed in Supplemental Dataset 1. d Schematic of the TraDIS analysis. e Volcano plots of SMKP838 genomes determined via TraDIS analysis in the presence of 40 mg/L SPA (blue) and 4% (green) and 8% (red) human serum. The x-axis shows the change in abundance of each SMKP838 gene compared with that in the control (a log2-fold change). The y-axis shows the p values of the detected SMKP838 genes compared with the control. f The number of genes detected via TraDIS with significantly increased or decreased detection in human serum compared with the control (FDRp < 0.05). g Venn diagram representing the integration of a total of 140 mutant genes identified in the serial passaging experiment in the presence of human serum in (c) and the genes associated with serum resistance derived from TraDIS analysis performed in the presence of 4% (620 genes) and 8% serum (794 genes) in (f). Putative genes involved in serum resistance when mutations enhance their function in blue, and genes involved in serum resistance when their function is disrupted/decreased are marked in red (less or more than a 2-fold difference in detection in the presence of human serum vs. without human serum in TraDIS, FDRp < 0.05, respectively). h Serum susceptibility of the SMKP838 pORTMAGE mutants. These mutants possessed gene mutation(s) identified in the serial passage experiments in the presence of serum, as shown in (b). PORTserumA (LOCUS_14270: p.Ala293Thr + ramA: p.Tyr34His), PORTserumB (LOCUS_21770: p.Leu113Pro + ramA: p.Tyr34His and a spontaneous mutation: glnD:p.Gly841Glu), PORTserumC (LOCUS_21770: p.Leu113Pro + ramA: p.Tyr34His), PORTserumD (LOCUS_39850: p.Val134Ala + ramA: p.Tyr34His), PORTserumE (ramA: p.Tyr34His and a spontaneous mutation: LOCUS_21770: p.Leu181Pro), and PORTserumF (ramA: p.Cys78Arg).

Fig. 3. Impact of gene mutation frequency on the development of serum resistance in non-HMV-Kp clinical isolates.

a Serial passaging experiment in the presence of human serum. Fifteen serum-susceptible non-HMV-Kp clinical isolates (with serum MICs ranging from 8 to 16%) from among the hyper- (n = 1, SMKP590), high- (n = 9; including one K. quasipneumoniae strain), and low-mutator strains (n = 9) were inoculated in 96-well plates containing serial dilutions of human serum (from 4 to 68%) in MHBII for culture at 37 °C for 24 h and subcultured for 20 days. Significantly more high-mutator isolates acquired serum resistance than low-mutator isolates, as determined by two-sided Fisher’s exact test (p = 0.015). b Serum susceptibility and accumulated gene mutations in the hypermutable isolate, SMKP590, during the serial passaging experiment in (a). We determined the serum MICs (red circles) and number of gene mutations (orange, grey, and white represent nonsynonymous mutations, synonymous mutations, and gene mutations in noncoding regions, respectively) of SMKP590 mutants obtained during serial passaging in the presence of human serum. c Number of novel and accumulated gene mutations in SMKP590 during the serial passaging experiments in (a). The novel and accumulated (continuously detected) gene mutations were counted and compared with those of the day before. The mutated genes are shown in the Supplemental Data (Dataset 3). d Numbers of novel nonsynonymous and other gene mutations that occurred during the serial passaging experiments with SMKP590 in (a). Other genes contained synonymous mutations and gene mutations in noncoding regions. e Integrated analysis of serum resistance in SMKP590. We integrated the gene sets for serum resistance identified via TraDIS (Fig. 2) with the gene mutations accumulated in SMKP590 during the serial passaging experiment in the presence of human serum, as shown in (b). Putative genes involved in serum resistance are shown in blue and red (< 2-fold or >2-fold difference in the presence of human serum vs. without human serum via TraDIS, FDRp < 0.05, respectively; Supplemental Data, Dataset 4). f We listed the serum resistance-associated gene mutations of SMKP590 that occurred during the serial passage experiment. Blue and red represent decreased or increased genes, respectively, as shown in (e). The underlined genes are known to be associated with serum resistance^,. D, Day.

Fig. 4. Influence of the gene mutations on the pathogenesis in non-HMV-Kp-infected mice.

a, b Intrabronchial infection mouse model and bacterial viability assessment. An intrabronchial infection model was established using non-HMV-Kp SMKP838 (WT) and its mutS-deletion mutant (ΔmutS) (5 × 10⁶ CFUs). Female BALB/c mice (10–12 weeks old), with or without immunosuppression, received 250/125 mg/kg cyclophosphamide monohydrate intraperitoneally prior to infection (n = 5 biologically independent experiments). Lung (a) and blood (b) bacterial counts were determined 48 hours post-infection. c Ciprofloxacin treatment and resistance analysis. Ciprofloxacin treatment was administered to infected immunosuppressed mice via intraperitoneal injection. d, e Bacterial counts in the lungs (d) and blood (e) were assessed with or without ciprofloxacin treatment (n = 6 biologically independent experiments). Parent strains (day 0) and ciprofloxacin-susceptible (CIP^S) or ciprofloxacin-resistant (CIP^R) mutants derived from 10, 19, or 20 days of serial passaging (Supplementary Figs. 4a, 5a). f, g Serum and ciprofloxacin susceptibility. Serum MICs were determined for SMKP838 and its mutants after 32 h of infection. Fifty colonies of each mutant were evaluated. h Ciprofloxacin MICs were measured in SMKP838 and its mutants post-ciprofloxacin treatment. i, j Lethality and histological examination. (i) Lethality assessment of SMKP838 WT (day 0), ΔmutS (day 0), and ΔmutS-serum^R (day 20) mutants evolved under serum exposure (serum MIC: 72%, Supplementary Fig. 2c). Mice counts (biologically independent experiments): WT (n = 14), ΔmutS (n = 6), ΔmutS-serum^R (n = 16), PORTserumA (n = 20). j Bacterial counts in the lungs and blood 32 hours post-infection (n = 6 biologically independent experiments). k Histological analysis (H&E staining) of infected immunosuppressed mouse tissues (lungs, liver, kidneys) 24 h post-infection. Severe bacterial infiltration (clumps) was noted in alveoli (red arrows), interstitium (yellow arrows), capillaries (B), and glomeruli (C) (white arrows). Histological scores (n = 3 biologically independent experiments) are shown on the right. Geometric means and standard deviations from biological independent experiments are shown (a–e, i, j, k), and two-sided one-way ANOVA with Dunnett’s test was used vs. WT (day 0). Log-rank test was used in (i).

Fig. 5. Influence of bacterial evolution on the outcomes of internationally spreading high-risk multidrug-resistant non-HMV-Kp strains.

We used the non-HMV-Kp strain BIDMC1 as a representative for the internationally spreading high-risk clone ST258. a Antimicrobial susceptibility of BIDMC1. S and R indicate susceptible and resistant, respectively. b Gene mutation frequency of BIDMC1 and the mutS nonsense mutable mutant (BIDMC1 MutS_Tyr37Stop). A rifampicin assay was performed to determine gene mutation frequency. The floating bars represent the maximum and minimum values, and the lines represent the geometric means (n = 3 biologically independent experiments). A two-sided Student’s t test was used for the statistical analysis. c Serial passaging experiments with BIDMC1 and the mutS mutant in the presence of human serum. BIDMC1 (wild-type, WT) and BIDMC1 MutS_Tyr37Stop were used (the accumulated mutations are listed in Supplemental Data Dataset 6). The geometric means and geometric standard deviations for biologically independent triplicate experiments are given (d–g). Bacterial growth of the BIDMC1-derived mutants during the serial passaging experiment in (c). We examined three clones of each of the WT and MutS_Tyr37Stop strains. Bacterial growth was evaluated as turbidity (OD600) after 16 h of cultivation in MHBII in (d). Bacterial growth of the mutants derived after 20 days of the serial passaging experiment in the presence of human serum was also evaluated by counting the viable bacterial number as colony formation units (CFU) at 0, 1, 3, 6, and 16 h in (e–g). The means and standard deviations for biologically independent triplicate experiments are given. A two-sided one-way ANOVA test was used for statistical analysis with multiple comparisons vs WT Day 0 (* indicates p < 0.05). The significant dh, Number of accumulated gene mutations of BIDMC_1 and BIDMC_1 MutS_Tyr37Stop in (a). i Survival rates of immunosuppressed mice intrabronchially infected with BIDMC_1, MutS_Tyr37Stop, and serum-resistant MutS_Tyr37Stop (each n = 6 biologically independent experiments). The mouse infection model was the same as that described in Fig. 4c. The log-rank test was used for statistical analysis.

Fig. 6. Schematic of the RIBEA method and identification of the clinical risk of evolved bacteria.

a Scheme of the RIBEA method developed in this study. We integrated serial passaging experiments to monitor the rapid bacterial evolution by using hypermutable strains in selective environments, such as different sites of infection and in the presence of antimicrobial agents; whole genome sequencing (WGS) to identify accumulated gene mutations during bacterial evolution; transposon-directed insertion sequencing (TraDIS) analysis to identify potential bacterial factors that contribute to survival in the selective environments (identification of resistance genes); and an in vivo model to evaluate the pathogenesis of the evolved bacteria and determine the potential clinical risk. RIBEA can be completed within approximately one month. b The mechanism by which pathogenicity develops in evolved bacteria. In this study, we revealed the pathogenesis mechanism and the potential clinical risk of the evolved non-HMV-Kp, a principal bacterial pathogen. Non-HMV-Kp does not exhibit typical clinical symptoms upon infection and can be eradicated by innate immune defences in immunocompetent hosts. In immunosuppressed (and/or immunodeficient) hosts, non-HMV-Kp infection can cause lower respiratory tract infections. Non-HMV-Kp, which has the potential for bacterial evolution (hyper- and high gene mutation frequencies), increases its clinical impact by spreading from the primary site (lung) to the blood with or without the development of antimicrobial (abxA) resistance, which leads to a decrease in the treatment efficacy of antimicrobial agents. Therefore, via RIBEA, we revealed the potential and the current clinical risks presented by bacteria that have a high frequency of gene mutations during infection.