ABSTRACT
Klebsiella pneumoniae is a Gram-negative bacterial pathogen that causes a range of infections, including pneumonias, urinary tract infections, and septicemia, in otherwise healthy and immunocompromised patients. K. pneumoniae has become an increasing concern due to the rise and spread of antibiotic-resistant and hypervirulent strains. However, its virulence determinants remain understudied. To identify novel K. pneumoniae virulence factors needed to cause pneumonia, a high-throughput screen was performed with an arrayed library of over 13,000 K. pneumoniae transposon insertion mutants in the lungs of wild-type (WT) and neutropenic mice using transposon sequencing (Tn-seq). Insertions in 166 genes resulted in K. pneumoniae mutants that were significantly less fit in the lungs of WT mice than in those of neutropenic mice. Of these, mutants with insertions in 51 genes still had significant defects in neutropenic mice, while mutants with insertions in 52 genes recovered significantly. In vitro screens using a minilibrary of K. pneumoniae transposon mutants identified putative functions for a subset of these genes, including in capsule content and resistance to reactive oxygen and nitrogen species. Lung infections in mice confirmed roles in K. pneumoniae virulence for the ΔdedA, ΔdsbC, ΔgntR, Δwzm-wzt, ΔyaaA, and ΔycgE mutants, all of which were defective in either capsule content or growth in reactive oxygen or nitrogen species. The fitness of the ΔdedA, ΔdsbC, ΔgntR, ΔyaaA, and ΔycgE mutants was higher in neutropenic mouse lungs, indicating that these genes encode proteins that protect K. pneumoniae against neutrophil-related effector functions.
INTRODUCTION
The Gram-negative encapsulated bacterium Klebsiella pneumoniae is both a commensal in the gastrointestinal tract (1–3) and an opportunistic pathogen. It causes a range of infections, including pneumonia, urinary tract infection, and bacteremia (4–8), and can disseminate systemically to secondary sites (9–11). K. pneumoniae causes approximately 10% of hospital-acquired infections, of which 11.8% are pneumonias (5). This pathogen is a steadily increasing threat due to a rise in antibiotic-resistant strains, a subset of which is resistant to all available antibiotics (12–14).
Two of the most at-risk populations for K. pneumoniae infections are people with diabetes or malignancies (15–20), conditions that generate multifactorial immunodeficiencies (21–23). Frequent features of these immunodeficiencies include neutropenia or impaired neutrophil (polymorphonuclear cell [PMN]) effector functions (21, 22, 24). In mouse studies, some clinical strains of K. pneumoniae are controlled in part by PMN population size (25) or PMN effector functions (26–32). PMN effector functions that protect against K. pneumoniae include phagocytosis (33–36) and the production of reactive oxygen species (ROS) (37) and reactive nitrogen species (RNS) (38).
Despite the range and severity of infections caused by K. pneumoniae, only a few of its virulence factors have been well-characterized, including capsule, lipopolysaccharide (LPS), type I and III fimbriae, and siderophores (39). Additional K. pneumoniae factors required for infection in the lungs were identified in previous high-throughput screens that independently screened approximately 25,000 and 4,800 K. pneumoniae transposon (Tn) insertion mutants in vivo in the lungs of mice (40, 41). Collectively, these studies found and characterized several K. pneumoniae genes required for lung infection (40–43). However, neither of these screens was comprehensive or performed on immunocompromised mice.
Since immunosuppressed hosts are highly susceptible to K. pneumoniae infection (16), we screened for the growth of Tn mutants in the lungs of wild-type (WT) and immunosuppressed mice. This study utilized an arrayed library of K. pneumoniae Tn insertion mutants in a transposon sequencing (Tn-seq), high-throughput screening technique (44, 45). After identifying putative K. pneumoniae virulence determinants required for growth in the lungs of immunocompetent and/or neutropenic mice, a combinatorial pooling approach was used to identify and retrieve mutants (46). A total of 62 K. pneumoniae Tn insertion mutants were retrieved and used in follow-up assays to characterize growth under various stress conditions. Selected single-gene deletion mutants were generated to validate the Tn-seq results and probe for the roles of these genes, including protection of K. pneumoniae from PMNs, during lung infection.
RESULTS
Identification of K. pneumoniae genes required for infection of the lungs of WT and neutropenic mice using Tn-seq.A high-throughput genetic approach, Tn-seq (45, 47), was used to identify the virulence determinants required for K. pneumoniae replication in lungs during pneumonic infection in both immunocompetent and immunocompromised hosts. A schematic of the experimental setup of the screen is presented in Fig. S1 in the supplemental material. An arrayed library of 13,056 K. pneumoniae transposon (Tn) insertion mutants generated in a spectinomycin-resistant K. pneumoniae ATCC 43816 derivative (MKP220) was pooled, and 2 × 104 CFU was intranasally delivered into Swiss Webster mice (41, 48, 49). This outbred mouse strain is more genetically diverse than classic inbred strains of mice, so attenuated mutants are more likely to be defective in a variety of genetic backgrounds. Because K. pneumoniae ATCC 43816 infection is controlled in part by neutrophils (polymorphonuclear cells [PMNs]) (17, 18, 25–28), this library was also used to infect neutropenic mice treated with an anti-Ly6G antibody and depleted of Ly6G+ PMNs (α-Ly6G mice).
In each of six experiments with WT mice and four experiments with α-Ly6G mice, the frequency of Tn mutants in each input library was highly reproducible, with a mean Spearman rho value of 0.953 (Fig. 1A). On average, 11,975 and 11,861 mutants with insertions in 3,722 and 3,691 genes, which are approximately 71% of the genes in ATCC 43816 (50), were present in the inocula used to infect cohorts of WT and α-Ly6G mice, respectively (Fig. 1B and C). At 33 h postinfection (hpi), lungs were collected and analyzed for the bacterial load and the PMN population. To limit confounding variables within cohorts, such as large differences in bacterial growth or PMN populations, only mice with 1 × 108 to 2 × 1010 CFU/g lung (an increase of 104 to 106 K. pneumoniae compared to the inoculation dose) and over 40% PMNs in the WT mouse lungs and less than 15% in α-Ly6G mouse lungs were analyzed. Thirty-eight WT mice and 20 α-Ly6G mice fit within these cutoffs (Fig. 1D and E). More than 109 CFU/g lung was recovered, on average, from both cohorts (Fig. 1D), and PMNs (Gr1hi CD11b+) comprised approximately 60% of the live cells in the lungs of WT mice (Fig. 1E). PMN depletion with an anti-Ly6G (1A8) antibody (51) effectively diminished the population of PMNs to approximately 7% of the lung cells (Fig. 1E). On average, 27% of the mutants (3,276 insertions in 1,593 genes) colonized the lungs of WT mice, while the α-Ly6G mice had a significantly less restrictive bottleneck that allowed 43% of the mutants to colonize (5,147 insertions in 2,060 genes) (Fig. 1F and G), indicating that PMNs control the initial seeding and/or growth of K. pneumoniae in the lungs.
High-throughput screen to identify Klebsiella pneumoniae genes needed for lung infection of healthy and neutropenic hosts. (A) Reproducibility between inputs was assessed by calculating Spearman’s correlation efficient, which averaged 0.953 and which ranged from 0.9374 to 0.9657, with P being <0.001. (B and C) Number of individual insertion mutants (B) and genes (C) represented in the inputs and outputs of WT and anti-Ly6G-treated (α-Ly6G) mice. (D and E) WT and α-Ly6G mice were intranasally infected with 2 × 104 CFU of a library containing ∼13,000 Tn insertion mutants and sacrificed at 33 hpi, and their lungs were collected for analysis. The number of CFU per gram of lung tissue (D) and the polymorphonuclear cell (PMN) population (the percentage of Gr1hi CD11b+ cells within the live cell population) (E) were quantified by flow cytometry. Each dot represents a mouse, and the lines indicate the geometric mean (D) or the mean (E). The dotted lines indicate the cutoffs for mice included in further analyses. (F and G) Histograms of the log2-transformed nCI values of insertion mutants by gene in WT (F) and α-Ly6G (G) mice. (H and I) Genes were considered to have statistically significant contributions in WT (H) or α-Ly6G (I) mice if the Tn mutants of that gene had a WT-MAX99% of <0.27 (H) or a PMN-MAX99% of <0.43 (I), respectively, as indicated by dotted lines. Data either are representative of those from 4 to 6 independent experiments (A) or were pooled from 4 to 6 independent experiments (B to I). Statistical significance was determined by one-way analysis of variance on log-transformed values with Tukey’s posttest (B and C) or Student’s t test (D and E) on either log-transformed (D) or nontransformed (E) values. ns, not significant; ****, P < 0.0001.
Given the bottlenecks, insertion mutants were clustered at the gene level to improve the statistical power. For each gene, the mean-normalized competitive index (mean-nCI) was calculated and log2 transformed for each gene following infection of WT and α-Ly6G mice (WT-mean-nCI and PMN-mean-nCI, respectively; see Materials and Methods), which resulted in unimodal distributions with means of 0.27 and 0.43, respectively (Fig. 1F and G). The unimodal distributions suggest that analysis of 38 WT mice and 20 α-Ly6G mice overcame the bottleneck in the lung at the gene level, whereas a bimodal distribution would indicate that the bottleneck was not overcome (see Materials and Methods). To identify genes that were significantly underrepresented in the output populations, a WT-MAX99% or a PMN-MAX99% metric was calculated (see Materials and Methods). These metrics account for the standard deviation (SD) of each mean-nCI, the bottlenecks, and a confidence level of 99% obtained from a Student’s t-distribution value of the 99% confidence level (t0.99; 2.43 for 38 WT mice and 2.54 for 20 α-Ly6G mice). The representation of genes with a MAX99% value below the WT mouse bottleneck of 0.27 and the α-Ly6G mouse bottleneck of 0.43 was considered significantly lower (Fig. 1H and I; Table S1). In total, 166 genes in WT mice and 194 genes in α-Ly6G mice had values below these cutoffs. Of the mutants defective for growth in WT mice, the growth of mutants with insertions in 51 of these genes also appeared to be defective in α-Ly6G mice, while the growth of mutants with insertions in 52 genes appeared to be restored in α-Ly6G mice, suggesting a role for the latter genes in protecting K. pneumoniae against PMNs or a PMN-modified environment (Table S2; Fig. S2A). Predictions for the remaining mutants could not be made because fitness defects were intermediate in WT mice and/or in α-Ly6G mice. Several previously identified genes that express factors critical for virulence were identified in this intermediate-fitness population, including those related to capsule (e.g., manC), LPS (e.g., pagP and wzzE), and siderophores (e.g., fepG) (Table S2) (52–56).
The predicted protein products of genes with defects in WT mice were matched to clusters of orthologous groups (COGs) (Fig. S2B and C) to get a broad overview of the putative functions of the genes necessary for K. pneumoniae virulence in the lungs of WT mice. Many of these genes encode proteins with predicted functions in metabolism, such as in the transport, biosynthesis, and metabolism of amino acids, carbohydrates and inorganic ions (COGs E, G, and P, respectively), energy production and conversion (COG C), or transcription and translation (COG J). The functional categories of the 52 genes needed to infect WT mice that were no longer necessary once PMNs were depleted were also identified. Many genes were predicted to be involved in carbohydrate metabolism (COG G), including genes needed for synthesizing capsule, such as manC (52), or inorganic ion transport and metabolism (COG P). All 3 replication, recombination, and repair genes (COG L) were required only in the presence of PMNs, suggesting that these proteins function in response to stresses caused by PMNs.
Identification of genes with defects under nutrient-limited growth conditions and in oxidative and nitrosative stress resistance using mini-Tn-seq screens.Using a combinatorial pooling and deep sequencing approach (46) to locate specific mutants, 62 insertion mutants were retrieved (Tables S3 and S4) from the K. pneumoniae library of 13,056 Tn insertion mutants and combined to generate a mini-Tn-seq library. This minilibrary was comprised of 42 insertion mutants with putative virulence defects in WT mice, 15 with intermediate fitness defects in WT mice, and 5 with no fitness defects in either WT or α-Ly6G mice serving as controls. Of the mutants with putative virulence defects in WT mice, 13 appeared to be defective in α-Ly6G mice and 16 were fully virulent (Table S2 and S3). The strains with intermediate defects in WT mice were included to assess the conservativeness of our fitness defect cutoffs (Tables S3 and S4). Ten of these mutants had defects in metabolic functions (Table S4) and have been discussed elsewhere (57).
The minilibrary was screened for growth and survival defects in nutrient-limited environments and under oxidative or nitrosative stress (Fig. S3), all of which are conditions encountered during lung infection (57–61). PMNs employ a number of methods to restrict pathogens, including the release of ROS and RNS (58, 59, 61), which can damage and kill invading bacteria (60). Growth under nutrient-restricted conditions was assessed by inoculating M9 minimal medium containing glucose (M9-glucose) with the K. pneumoniae mini-Tn-seq library and permitting growth for 5 h (Fig. S3A; Table S3). The Tn::dsbA, Tn::glpR, Tn::yajL, and Tn::ycgE mutants had growth defects compared to the growth of the neutral mutants (Table S3A). Three of these mutants, the Tn::dsbA, Tn::yajL, and Tn::ycgE mutants, were defective in WT and α-Ly6G mice in the Tn-seq screen (Table S2). No mutants had defects in growth compared to the growth of WT K. pneumoniae when grown alone under similar conditions (unpublished data), suggesting that the observed growth defects were a result of interstrain competition.
Transition metals, such as iron, copper, and zinc, are obtained from the environment and essential for K. pneumoniae growth but are limited by the host, including through PMN-mediated means (62–68). Swiss Webster mice were intraperitoneally injected with either phosphate-buffered saline (PBS; mock treatment) or deferoxamine (DFO), an iron chelator that Klebsiella can bind to and use to gather iron (69, 70). Mice were intranasally infected with 2 × 104 CFU of WT K. pneumoniae, and their lungs were harvested at 33 hpi and plated for bacterial burden (Fig. 2A). Mice receiving DFO had a higher bacterial burden than mock-treated mice (Fig. 2A), suggesting that under these conditions K. pneumoniae requires iron for virulence. The pooled mini-Tn-seq library was grown for 5 h in low-salt Luria-Bertani medium (L) or L containing 0.13 mM 2,2′-bipyridyl (DIP), a high-affinity chelator of iron and other divalent metals that, unlike DFO, cannot be utilized by K. pneumoniae to acquire iron (Fig. S3B; Table S3). Only the Tn::appC mutant had a significant decrease in growth under these conditions compared to the growth of the neutral strains. This phenotype was confirmed in 1:1 competition with WT K. pneumoniae (Fig. 2B; Table S3), suggesting that AppC assists in iron acquisition during growth.
Multiple gene hits in Tn-seq are utilized by K. pneumoniae to counteract neutrophil effector functions. (A) Mock-treated (circles) and DFO-injected (triangles) mice were intranasally infected with 2 × 104 CFU of WT K. pneumoniae and sacrificed at 33 hpi. Their lungs were collected for analysis, and the number of CFU per gram of lung tissue is presented. Each symbol represents a mouse, and lines indicate the geometric mean. (B) The K. pneumoniae Tn::appC mutant and neutral insertion mutants (MKP330, MKP332, and MKP354) were grown in L until they were in mid-log phase. The Tn::appC mutant or the neutral insertion mutants were pooled 1:1 with WT K. pneumoniae grown under the same conditions and used to inoculate L with 0.125 mM DIP. The cultures were grown for 5 h and then plated on L and L-KAN plates. The competitive index (CI) was calculated using (number of CFU of mutant-treated cultures/number of CFU WT-treated cultures)/(number of CFU of mutant-untreated cultures/number of CFU WT-untreated cultures) and normalized (norm) to the neutral WT growth. (C) WT (circles) and gp91phox−/− (squares) C57BL/6 mice were retropharyngeally infected with 1.5 × 103 CFU of WT K. pneumoniae and sacrificed at 45 hpi. The bacteria in the lungs were quantified as described in the legend to panel A. (D and E) The indicated K. pneumoniae strains were grown, pooled as described in the legend to panel B, incubated in M9-glucose with 1 mM H2O2 (D) and M9-glucose with 1 mM DETA-NONOate (E) for 1 h, and then plated. CI was calculated as described in the legend to panel B. (A to E) Data were pooled from 3 independent experiments with at least 2 mice per cohort (A, C) or in technical duplicate (B, D, and E). Statistical significance was determined by Student’s t test on log-transformed values (A, C) or one-way analysis of variance on nontransformed values with Dunnett’s posttest (B, D, and E). ns, not significant (P > 0.05); *, P < 0.05; ****, P < 0.0001.
When retropharyngeally infected with WT K. pneumoniae, gp91phox−/− mice unable to produce NADPH-dependent ROS (29, 71) experienced higher lung bacterial burdens than the C57BL/6 mouse controls (Fig. 2C). Both cohorts of mice had comparable PMN populations (WT mice, 43% ± 8%; gp91phox−/− mice, 36% ± 4%), a finding consistent with previous observations in studies with Escherichia coli (72). These results indicate that ROS production by phagocytes inhibits K. pneumoniae lung infection either directly or indirectly through downstream ROS-dependent neutrophilic functions (73, 74). The K. pneumoniae minilibrary was screened for genes important for resisting ROS, specifically, H2O2 stress, by incubating the pooled strains in M9-glucose alone or M9-glucose with 1 mM H2O2 for 1 h (see Materials and Methods) (58, 59). Overall, the H2O2-treated samples were recovered at 40% ± 5% (P < 0.05) of the levels for the untreated samples. The Tn::gntR and Tn::yaaA mutants had survival rates significantly below the rate for the WT in H2O2 (Fig. S3C). In a 1:1 competition with WT K. pneumoniae, these mutants were also defective in survival in H2O2 (Fig. 2D), suggesting that GntR and YaaA protect K. pneumoniae against H2O2. The disruption of several genes increased K. pneumoniae fitness in H2O2 in the minilibrary (Fig. S3C), and 1:1 competition between Tn::ycgE and WT K. pneumoniae confirmed this finding (Fig. 2D).
To determine if any K. pneumoniae mutants had increased sensitivity to nitrosative stress, the minilibrary was grown in M9-glucose in the presence of 1 mM diethylenetriamine (DETA)-NONOate for 1 h. Under these conditions, 30% ± 8% (P < 0.001) of the K. pneumoniae population was killed. The Tn::ybhU mutant was sensitive to nitric oxide (NO) compared to the control strains in the mini-Tn-seq screen and 1:1 competition (Fig. 2E; Fig. S3D), while the Tn::gntR and Tn::dedA mutants were 10 and 5% more resistant than the control strains in the mini-Tn-seq screen, respectively (Fig. S3D). However, when grown in 1:1 competition, the Tn::gntR mutant was impaired in RNS resistance and the Tn::dedA mutant trended toward impairment (P = 0.07) (Fig. 2E). Combined, these results indicate that multiple Tn mutants were defective in vitro against one or more host immune effector functions, consistent with the virulence defects seen in the Tn-seq screen.
K. pneumoniae requires dedA, dsbC, gntR, wzm, yaaA, and ycgE for virulence during single-strain lung infections.To determine if the K. pneumoniae mutants with defects in the in vitro mini-Tn-seq screens (Fig. 2B and D to E; Fig. S3) were impaired in colonizing lungs in single-strain infections, mice were intranasally infected with 2 × 103 CFU of WT K. pneumoniae or the Tn:appC, Tn::dedA, Tn::dsbA, Tn::gntR, Tn::yaaA, Tn::ycgE, or Tn::yhbU mutant strain, and the bacterial burden in the lungs was measured at 45 hpi (Fig. 3A). The Tn::dedA and Tn::dsbA mutants had a drastic defect, with a greater than 104-fold decrease in the bacterial load compared to that of WT K. pneumoniae being seen, while the levels of the Tn::gntR, Tn::ycgE, and Tn::yhbU mutants were approximately 100-fold lower (Fig. 3A). In contrast, the results for the Tn::appC and Tn::yaaA mutants were not statistically significantly distinguishable from those for the WT, although the Tn::yaaA mutant population recovered was about 8-fold smaller than the WT population (Fig. 3A).
K. pneumoniae (Kp) requires dedA, dsbC, gntR, wzm, yaaA, and ycgE in the lungs for virulence during single-strain infections. (A to D) WT (circles) and α-Ly6G (triangles) mice were intranasally infected with 1 × 103 CFU of the indicated Tn mutant (A) or 2 × 103 CFU of the indicated strains (B to D) and sacrificed at 45 hpi (A and B) or 42 hpi (C and D). The lungs were collected for analysis, and the number of CFU per gram of lung tissue (A to C) and the PMN population (the percentage of Gr1hi CD11b+ cells within the live cell population), as quantified by flow cytometry (D), are presented. Each dot represents one mouse, and open dots indicate the limit of detection where no CFU was recovered. Bars indicate the geometric mean (A to C) or the mean (D). Each strain was tested in cohorts of at least 2 mice in at least 2 independent experiments, and the data were pooled. Statistical significance was determined by one-way analysis of variance on log-transformed values (A to C) with Dunnett’s (A and B) or Fisher’s (C) least-significant-difference posttests. ns, not significant (P > 0.05); *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.
In-frame deletion mutants of dedA (VK055_RS24345), gntR (VK055_RS18575), yaaA (VK055_RS12895), ycgE (VK055_RS08755), and yhbU (VK055_RS25280) were generated using bacteriophage lambda Red recombination (75) to evaluate their virulence in single-strain lung infections and, thus, confirm the importance of these genes in K. pneumoniae growth. In addition, mutants with in-frame deletions of dsbC (VK055_RS21055) and wzm (VK055_RS25470) were generated because these Tn mutants were not retrievable from the arrayed combinatorial pool but were identified to be significant hits and were of interest (Tables S1 and S2). We were unable to construct an in-frame deletion of dsbA (VK055_RS16690) using these methods. Mice were infected with 2 × 103 CFU, and the bacterial load in the lungs was measured at 45 hpi (Fig. 3B). Two mutants, the ΔdedA and Δwzm mutants, had bacterial burdens 6 orders of magnitude lower than those of the WT, and four mutants, the ΔdsbC, ΔgntR, ΔyaaA, and ΔycgE mutants, had bacterial burdens 100-fold lower than those of the WT. The ΔyhbU strain was as virulent as the WT and was not evaluated further.
Tn-seq data in α-Ly6G mice predicted that the dsbC, wzm, and ycgE deletion strains would remain attenuated in α-Ly6G mice compared to their virulence in WT mice, whereas the virulence of the dedA, gntR, and yaaA deletion strains would be recovered (Tables S1 and S2). To evaluate whether any of these mutants colonized better in the absence of PMNs, additional cohorts of WT and α-Ly6G mice were intranasally infected with 2 × 103 CFU of the ΔdedA, ΔdsbC, Δwzm, ΔgntR, ΔyaaA, or ΔycgE mutant or WT K. pneumoniae. At 42 hpi, lungs were collected and analyzed for the bacterial load and the PMN population (Fig. 3C and D). PMN depletion using an anti-Ly6G antibody (the 1A8 antibody) effectively diminished the population of PMNs to approximately 2% of the lung cells (Fig. 3D). Under these conditions, bacterial levels did not differ between WT and neutropenic mice, despite a large influx of Gr1hi CD11b+ cells in the former (Fig. 3D), which is consistent with previous findings (Fig. 1A and B). In α-Ly6G mice, the ΔdedA, ΔdsbC, ΔgntR, ΔyaaA, and ΔycgE mutants had increased bacterial burdens (Fig. 3C) compared to those in the WT mice, suggesting that K. pneumoniae utilizes DedA, DsbC, GntR, YaaA, and YcgE to protect against PMNs. The Δwzm strain remained severely attenuated even in the absence of PMNs (Fig. 3C and D), indicating that wzm or other genes in the operon are required for virulence, regardless of host PMN status.
Mutants are not differentially restricted for growth under low-iron conditions in vitro.Based on homology, the predicted functions of the selected genes are as follows: dedA encodes a putative inner membrane protein; dsbC and wzm encode proteins with predicted roles in periplasmic protein folding and transporting O antigen, respectively; ycgE encodes a MerR-like transcriptional regulator; gntR encodes a negative transcriptional regulator of gluconate uptake; and yaaA encodes a peroxide stress protein (76). To further understand and characterize the mechanisms behind the sensitivity of each of the six mutants to either PMNs or the overall host lung environment, the six deletion mutants and WT K. pneumoniae were evaluated in the in vitro functional assays used to screen the Tn insertion minilibrary mutants (Fig. S3).
To evaluate if dedA, dsbC, gntR, wzm, yaaA, and ycgE are critical for growth under low-iron conditions, each strain was grown in L broth in the presence or absence of DIP for 16 h (Fig. 4A). While all strains had reduced growth in 0.25 mM DIP, the growth of none of the six tested mutants was impaired compared to that of the WT, and only the control strain, the ΔaroA mutant (57), failed to grow (Fig. 4A). These findings suggest that these genes are not involved in iron scavenging and/or acquisition, which is consistent with the mini-Tn-seq results (Fig. S3B).
K. pneumoniae requires dsbC, wzm, and ycgE for mucoviscosity and wzm for capsule production. (A to E) WT K. pneumoniae and the indicated mutants were grown overnight in L. (A) L (circles) or L with 0.25 mM DIP (triangles) was inoculated with overnight cultures, and the OD630 was measured for 16 h. (B) The total bacterial density (OD600) of each strain was measured, the strains were centrifuged at 1,000 × g for 5 min, and the OD600 of the supernatant was measured. Mucoviscosity was calculated as (OD600 supernatant)/(OD600 total), and the value was normalized to the value for the WT. (C and D) The glucuronic acid content following capsule precipitation was determined by a colorimetric reaction (OD520) and interpolated from a standard curve for glucuronic acid. (E) RNA was isolated, DNase treated, and reverse transcribed into cDNA, and wzt mRNA was quantified and normalized relative to the level of 16S mRNA expression for each strain. Data are representative of those from 3 independent experiments (A) or were pooled from at least 2 (B) or 3 (C to E) independent experiments and are presented as the mean ± SEM (B to E). Significance was determined by one-way analysis of variance on the OD630 at 16 h with Dunnett’s posttest (A), with nontransformed values with Dunnett’s posttest (B to D), or with nontransformed values with Welch’s t test (E). ns, not significant (P > 0.05); *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.
K. pneumoniae requires dsbC, wzm, and ycgE for general mucoviscosity and wzm for capsule production in vitro.Capsule is a major virulence factor of K. pneumoniae (36, 39, 41, 77) that impairs PMN-mediated phagocytosis (33, 78, 79). To evaluate whether dedA, dsbC, gntR, wzm, yaaA, and ycgE play a role in capsule production, the mucoviscosity of the six deletion mutants was measured after growth in L for 16 h (40, 78). The ΔdsbC, Δwzm, and ΔycgE strains had less mucoviscosity than WT K. pneumoniae (Fig. 4B) but more mucoviscosity than an acapsular mutant, the ΔcpsB mutant. This suggests that these strains have impaired, but not completely defective, production and/or release of capsular components. Since mucoviscosity is not a direct reflection of capsule production, the K. pneumoniae capsule content was quantified by measuring glucuronic acid, a polysaccharide component of the Klebsiella capsule (80). Only the Δwzm mutant was defective in capsule content, with capsule levels being comparable to those in the ΔcpsB mutant (Fig. 4C). However, the glucuronic acid content of a Δwzm strain complemented with pWzm failed to reach WT levels (Fig. 4D). Analysis by quantitative PCR suggested that the wzm deletion had potentially polar effects on the downstream ABC transporter component wzt (Fig. 4E). Therefore, the Δwzm mutant was complemented with a plasmid containing both O-antigen ABC transporter-encoding genes, wzm and wzt, and the glucuronic acid content of the resulting strain, the Δwzm/pWzmWzt strain, was found to be restored to WT levels (Fig. 4C), indicating that the defect lies in wzt or wzm and wzt and that the in-frame deletion alters wzt. Henceforth this mutant is referred to as the wzm-wzt mutant.
gntR and yaaA contribute to ROS resistance and dedA and ycgE contribute to NO resistance in vitro.To determine if any of the six K. pneumoniae in-frame deletion mutants exhibited increased sensitivity to oxidative stress, WT K. pneumoniae and the ΔdedA, ΔdsbC, ΔgntR, Δwzm-wzt, ΔyaaA, and ΔycgE mutants were evaluated for survival in M9-glucose in the presence of 1 mM H2O2 for 1 h (58, 59). Under these conditions, 1% of the WT population survived, whereas <0.01% of the ΔgntR mutant population and <0.1% of the ΔyaaA mutant population survived (Fig. 5A), consistent with the findings for the Tn insertion mutants (Fig. 2D). The survival of the ΔgntR mutant after H2O2 exposure was restored to WT levels upon complementation (Fig. 5B). Furthermore, the expression levels of gntK and edd, two genes negatively regulated by GntR in E. coli (81–83), were increased in a ΔgntR mutant background (Fig. 5C and D). Combined, these results indicate that the activities of GntR, a transcriptional regulator of gluconate metabolism, and YaaA (76, 84), a peroxide stress protein, protect K. pneumoniae against H2O2.
K. pneumoniae requires gntR and yaaA for ROS resistance in vitro. (A to D) WT K. pneumoniae and the indicated mutants were grown overnight in L. (A and B) The strains were then recultured until log phase, incubated in M9-glucose with or without 1 mM ROS for 1 h, and plated to determine survival, and the results were normalized to those for the M9-glucose-only condition for each strain. (C and D) RNA was isolated from log-phase cultures of the indicated strains, DNase treated, and reverse transcribed into cDNA, and gntK (C) or edd (D) mRNA was quantified and normalized relative to the level of 16S mRNA expression for each strain. (A to D) Data were pooled from at least 3 independent experiments and are presented as the mean + SEM. Statistical significance was determined by one-way analysis of variance on log-transformed values with Dunnett’s posttest (A and B) or Welch’s t test on nontransformed values (C and D). *, P < 0.05; **, P < 0.01; ***, P < 0.001.
To determine if any of the six K. pneumoniae in-frame deletion mutants exhibited increased sensitivity to nitrosative stress, the mutants were grown for 1 h in M9-glucose with or without 1 mM DETA-NONOate, a slow-releasing nitric oxide donor. Under these conditions, roughly 70% of the WT population survived. Meanwhile, two mutants, the ΔdedA and ΔycgE mutants, were more sensitive to nitrosative stress than the WT (Fig. 6A). The survival of the ΔycgE mutant after RNS exposure was restored to WT levels upon complementation (Fig. 6B). The expression levels of three genes repressed by YcgE in E. coli (85, 86) were increased in a ΔycgE mutant (Fig. 6C to E), indicating the YcgE acts as a transcriptional regulator.
K. pneumoniae requires dedA and ycgE for RNS resistance in vitro. (A to E) WT K. pneumoniae and the indicated mutants were grown overnight in L. (A and B) Strains were cultured until log phase and then incubated in M9-glucose with or without 1 mM DETA-NONOate for 1 h and plated for survival, and the survival rate was normalized to the rate for the M9-glucose-only condition for each strain. (C to E) RNA was isolated from log-phase cultures of the indicated strains, DNase treated, and reverse transcribed into cDNA, and the relative copy number of each target mRNA was quantified and normalized relative to the level of 16S mRNA expression for each strain. Data are representative of those from at least 3 independent experiments (A and B) or were pooled from at least 3 independent experiments (C to E) and are presented as the mean ± SD (A and B) or the mean + SEM (C to E). Statistical significance was determined by one-way analysis of variance with Dunnett’s posttest (A and B), Welch’s t test (C and D), or Student's t test (E). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.
DISCUSSION
This study provides evidence for the requirement of 6 operons containing the genes dedA, dsbC, gntR, wzm-wzt, yaaA, and ycgE for K. pneumoniae survival and/or growth in lungs. Most of these—dedA, dsbC, gntR, yaaA, and ycgE—facilitate K. pneumoniae growth in the presence of PMNs. In contrast, wzm-wzt is necessary in both the presence and the absence of PMNs. This study also identifies putative functional roles for these proteins in growth and protection against various facets of the immune response or stress conditions. Notably, the ΔdsbC, Δwzm-wzt, and ΔycgE mutants were defective in mucoviscosity, with the Δwzm-wzt mutant also exhibiting a significant decrease in glucuronic acid content, likely reflecting a decrease in the production or excretion of capsular material from the surface. These decreases in capsule production or stability may be due to roles that the periplasmic disulfide isomerase, encoded by dsbC, and the O-antigen transporter protein, encoded by wzm-wzt, play in LPS assembly. Recent work has also identified outer membrane components, including LPS biosynthesis genes, to be important for capsule production by K. pneumoniae ATCC 43816 (87). Additionally, capsule polysaccharide interacts with the O antigen on LPS molecules in some K. pneumoniae strains, as well as E. coli K1 (88, 89). Capsular polysaccharide is a well-studied, but highly variant, K. pneumoniae virulence factor that increases resistance to opsonophagocytic uptake by phagocytic cells (90, 91). Increased capsule expression is correlated with an increase in K. pneumoniae virulence in vivo (92, 93), suggesting that these mutants may be attenuated due to a defect in capsule integrity. The emergence of hypermucoviscous K. pneumoniae strains able to infect immunocompetent populations, in addition to immunodeficient populations (90), highlights the importance of identifying critical capsular virulence factors for therapeutic targeting.
Additionally, multiple genes, including dedA, gntR, yaaA, and ycgE, played roles in K. pneumoniae ROS and/or RNS resistance, and the majority of these had not previously been implicated in K. pneumoniae virulence. In fact, to our knowledge, only yaaA has been described as protecting against ROS in other bacteria. In E. coli, yaaA protects against H2O2 by reducing iron levels and preventing damage by the Fenton reaction (84). Both GntR and YcgE are canonical repressors (GntR family and MerR family, respectively); GntR regulates genes involved in gluconate metabolism in E. coli (83), while YcgE has been shown to regulate genes involved in biofilm formation in E. coli (86). While the specific roles that GntR and YcgE play in ROS and RNS resistance, respectively, are unknown, upregulation of their downstream targets was observed in the K. pneumoniae ΔgntR and ΔycgE mutants, indicating that these or other targets play a role in the increased ROS and RNS sensitivity of these mutants and warrant further study. Although capsule and ROS/RNS assays are not a direct measurement of PMN functions in response to K. pneumoniae, these assays putatively identify how each K. pneumoniae gene or operon is involved in neutralizing or counteracting specific PMN effector functions. Future work aims to further elucidate the interplay between PMN phagocytosis, PMN-derived ROS killing, and the function of these genes.
The retrieval of mutants from our arrayed Tn-seq library using combinatorial pooling was invaluable to this study, as it provided a rapid means of isolating mutants to be screened in a high-throughput manner. However, the use of this minilibrary of K. pneumoniae Tn insertion mutants comes with caveats. First, it is unknown if the observed phenotypes were due to the disrupted gene or downstream genes in the same operon. Therefore, we followed up on the retrieved mutants by generating single gene deletions for each mutant of interest with the bacteriophage lambda Red system (75). Second, while most insertion mutants evaluated had intragenic Tn insertions, one retrieved mutant, the appC mutant, had an insertion immediately upstream of the start codon, where we postulated that it interfered with expression. In fact, the Tn::appC mutant in our in vitro minilibrary assays was sensitive to growth under conditions of low iron availability, suggesting that upstream insertions did interfere with gene expression. On the other hand, the Tn::appC mutant remained virulent in single-strain WT mouse infections. This lack of a phenotype with this specific upstream insertion mutant does not mean that the corresponding gene does not play a role under those conditions. Additionally, yhbU was a hit in the Tn-seq screen, but the growth of the mutant with a yhbU in-frame deletion was not attenuated in single-strain WT mouse infections. This suggests that the Tn insertion impacted the surrounding genes, which was likely responsible for the Tn-seq phenotype and which underscores the importance of generating single-gene deletions for each mutant of interest.
In addition to revealing potential roles for these 6 operons in K. pneumoniae physiology and virulence, whi8ch may be added to the roles of 10 genes revealed in a previously published study (57), over 100 additional genes that are potentially involved in K. pneumoniae infection of the lungs of immunocompetent and/or immunodeficient mice were identified in the Tn-seq screen. Further validation and then studies into this trove should increase the number of potential targets for novel therapies, as well as generate insights into bacterial requirements in immunocompromised and immunocompetent lung tissues. However, further validation of these putative candidates is required because several false-positive results were detected in our analysis of our initial screening of 38 WT mice and 20 neutropenic mice. In fact, one challenge of high-throughput genomic screens of bacterial insertion mutants under selection conditions is bottlenecks, which initially occurred during the seeding event in the lungs and which severely restricted the size of the bacterial population in this study (94). Due to bottlenecks, not every bacterium in the 2 × 104-mutant inoculum gives rise to a clonal population in the lung, regardless of the growth ability or virulence of the mutant. This bottleneck is at its highest and most restrictive in an immunocompetent host, where immune cells can eliminate much of the bacterial population before seeding can occur (95). A less restrictive bottleneck was observed in the neutropenic mice, indicating that PMNs control the initial levels of successful seeding events and/or the early replication of K. pneumoniae in the lung.
The bottleneck during the initial seeding of K. pneumoniae in the lung impacts the statistical power when analyzing and identifying potential hits (96). Methods of analysis to overcome this limitation include performing statistical tests, such as paired t tests or rank tests, for each gene, comparing the findings before and after selection without adjusting for multiple comparisons (45, 97) to determine significance. However, these methods are more effective with smaller libraries or libraries with less stringent bottlenecks, as they are sensitive to background noise and fail to address the potential for false-positive results arising from many statistical comparisons. To correct for the large number of comparisons made with these tests, only insertion mutants never recovered in any mice would be considered significantly defective and the corresponding genes would be considered important. While nonparametric statistical tests, such as the Mann-Whitney test, are less likely to generate false-positive results due to background noise, they can miss candidate genes of interest because they lack power (96), resulting in the disqualification of many genes. We chose our method of data analysis, detailed in Materials and Methods, for several reasons. First, compiling our library by gene rather than Tn insertion enabled us to increase the number of times that genes were sampled and, thus, our confidence in our analysis. If nCI had been calculated for each insertion rather than each gene, the bottleneck would not have been overcome by screening only 38 WT mice and 20 neutropenic mice (i.e., Fig. 1F and G would have had bimodal populations). We calculated the standard deviation (SD) of the nCI for each individual gene rather than the population average and combined each gene’s SD with the corresponding nCI to create MAX99% values, which were used to determine significance. By including the SD for each individual gene, the individual variation for each gene was considered. This eliminates some false-positive results arising from genes with low frequencies among the strains in the input that would otherwise be competent to colonize. These genes often have higher SDs because the strains carrying these genes have a low chance of seeding the lung due to their low input numbers; however, when they do seed, they are equally competent to colonize. Finally, by setting the cutoff for significance based on t0.99, we accounted for the size of the experimental population.
A similar approach has been published in another study (47) using a cutoff such as 2 + SDnCI of all genes from the mean, where SDnCI of all genes refers to the SD of the nCI of all genes considered collectively. When applied to our data set, both our method and this previously published method gave significance to a preponderance of genes whose input levels were, in general, less than the average input level of the library. However, we noted that the latter method enriched for many more insertion mutants that were not well represented in the input. This enrichment for mutants that are not well represented in the inputs could occur for several reasons. For instance, these mutants may be defective in their ability to grow both in vitro compared to the ability of the other mutants in the library and in vivo in the mouse lung or just on L plates. With fewer representatives in the input library, these strains may not overcome the bottleneck and therefore are not recovered for stochastic reasons.
Despite its caveats, this approach taken to analyze our Tn-seq data followed by proper validation yielded several novel K. pneumoniae genes whose contribution to virulence in the lungs was confirmed in vivo here and elsewhere (57). However, several genes expected to be necessary for K. pneumoniae virulence in the lungs, based on previously published studies (40, 55), were not identified in our analysis. These included genes related to the synthesis of the siderophore enterobactin, entE and entF, and a gene previously identified to be critical for virulence in the lungs, aroE (40, 98). These three genes were well represented in our inputs, raising a concern that, despite optimization, our analysis was biased toward identifying genes with insertions that were present at a low frequency in the input. This suggests that further refinement in the analysis to account for variations in the location of the Tn insertions for each gene present in the input and the bottleneck may be informative.
In summary, using high-throughput approaches, several K. pneumoniae virulence determinants needed for lung infection and, in some cases, protection against PMNs were discovered and confirmed, including dedA, dsbC, gntR, wzm-wzt, yaaA, and ycgE. In addition, putative functions for these genes were identified under a variety of conditions, including growth under conditions of low metal availability, production of exopolysaccharide, and resistance to ROS and RNS. Future work will aim to further characterize the roles of these genes in establishing infections in lungs and in other tissues.
MATERIALS AND METHODS
Generation of K. pneumoniae strains.All strains used in this study are listed in Tables S1 and S5 in the supplemental material. A K. pneumoniae WT strain (ATCC 43816; MKP203) was made spectinomycin resistant (SPTr) by introducing the mini-Tn10 Tn carrying an SPTr cassette through the conjugation of K. pneumoniae WT with a diaminopimelic acid (DAP)-auxotrophic E. coli strain, MFDpir/pDL10987 (MKP214) (45). K. pneumoniae SPTr derivatives of the WT strain were selected for on L-agar plates containing spectinomycin (SPT; 50 μg/ml). Several K. pneumoniae SPTr strains were verified to be as fit as the parental K. pneumoniae WT strain, based on in vitro growth in L and in vivo infection of the lungs (data not shown). One strain, MKP220, was used as the WT K. pneumoniae SPTr strain in all subsequent experiments.
K. pneumoniae ΔcpsB was generated in MKP220 using bacteriophage lambda Red recombination as previously described (75) with primers MKP133, MKP134, MKP135, and MKP136 (Table S5). Briefly, primers MKP133 and MKP134 had homology to 57 bp upstream and the start codon of cpsB and 57 bp downstream and the stop codon of cpsB, respectively, and were used to amplify an apramycin resistance cassette. K. pneumoniae strains were confirmed to have the ΔcpsB deletion by PCR amplification and sequencing with primers MKP135 and MKP136. In-frame deletion mutants K. pneumoniae ΔdedA, ΔgntR, ΔyaaA, ΔycgE, and ΔyhbU were generated in MKP220 and the ΔdsbC and Δwzm mutants were generated in MKP203 using bacteriophage lambda Red recombination (75) with the appropriate primers (Table S6).
K. pneumoniae insertion mutants were created by introducing a kanamycin-resistant (KANr) Himar1 mariner Tn into MKP220 (WT SPTr) through conjugation with the DAP-auxotrophic E. coli MFDpir/pSC189 strain (47). K. pneumoniae Tn mutants were selected on L-agar plates containing both kanamycin (KAN; 100 μg/ml) and SPT (50 μg/ml). Individual colonies were picked using a Genetix QP Expression automated colony and array picker and inoculated into 384-well plates (Corning) containing L with KAN (100 μg/ml) and 15% glycerol. The resulting arrayed library, consisting of the colonies in 34 plates, was incubated at 37°C overnight and stored at −80°C until further use. A total of 13,056 K. pneumoniae insertion mutants were generated and arrayed from 3 independent rounds of 3 separate conjugations.
Analysis of arrayed K. pneumoniae library.Based on principles previously published (46), Tn mutants of interest were located in the array using a combinatorial pooling method, for which two new scripts, which are available upon request, were developed. The method increased the throughput using a different robotic system, and the new scripts enabled insertion sites to be correlated with clone location. To identify the location of individual Tn insertions, the arrayed library was thawed and duplicated into 384-well plates containing L-KAN using a Genetix QP Expression robot. The plates were subsequently incubated at 37°C overnight without shaking. A Tecan Freedom Evo 200 liquid-handling robot system was used to generate 24 predetermined combinatorial libraries, where each contained a unique subset of these 13,056 clones. Genomic DNA (gDNA) was then isolated from these 24 pools, processed, barcoded (Table S7), and deep sequenced as described below for Tn-seq experiments in WT and neutropenic mice. The resulting reads from each pool were mapped to the K. pneumoniae reference genome with GenBank accession no. NZ_CP009208.1 using the Bowtie2 program. The resulting SAM files for these 24 pools of combinatorial libraries were used as an input for a custom-developed Perl script that correlates insertion sites with individual clones. The location of each insertion mutant identified in the pools was then mapped to specific wells based on the pools in which the mutants were present. Specific insertions were preliminarily assigned to about 35% of the wells.
To confirm this initial assignment, K. pneumoniae insertion mutants of interest were then retrieved and sequenced after PCR amplification of the region adjacent to the Tn, where one primer complemented one end of the Tn (MKP030 or MKP079) and the other insertion mutant-specific primer complemented a site 600 to 800 bp downstream of the expected insertion site (Table S6). Any resulting PCR products of the correct size were then sequenced to verify the site of the Tn insertion. Using this method, the sequences of approximately 80% of the insertion mutants were confirmed.
Mouse infections.Mouse infections were carried out as previously described (40, 41, 48, 49, 57) with the following modifications. For infections with the arrayed library of 13,056 K. pneumoniae insertion mutants, the arrayed strains were duplicated into L-KAN-containing plates, incubated overnight at 37°C without shaking, pooled, diluted to a final concentration in 15% glycerol, aliquoted, and stored at −80°C until use. Prior to infection, an aliquot was thawed, plated on 150-mm L-KAN agar plates to obtain at least 1 × 105 colonies, and grown overnight at 37°C. The resulting colonies were collected, pooled, and used to start an overnight culture in L that was grown at 37°C with aeration. On the following morning, the culture was serially diluted in sterile PBS, and the infectious dose was quantified on L-agar plates. For single-strain infections, K. pneumoniae strains were cultured overnight at 37°C with aeration in 2 ml L. From the overnight cultures, the bacteria were serially diluted in sterile PBS and the infectious dose was quantified on L-agar plates.
Seven- to 12-week-old female Swiss Webster mice (from NCI for the initial large-scale Tn-seq screen and from Taconic for all remaining experiments) or 7- to 12-week-old gp91phox−/− mice (The Jackson Laboratory) with age- and sex-matched C57BL/6 mice (The Jackson Laboratory) were anesthetized with 3% isoflurane and infected intranasally (90) or retropharyngeally (99), respectively, with a 50-μl bolus containing the number of CFU of K. pneumoniae indicated in the figure legends. The mice were depleted of PMNs as previously described (51, 57) by intraperitoneal injection of 50 μg of an anti-Ly6G antibody (0.5 mg/ml 1A8 antibody; Fisher Scientific) 16 h prior to infection. Mock-depleted mice received 100 μl of PBS (Corning). Mice were administered 500 μl deferoxamine mesylate salt (DFO; 2.5 mg/ml; Sigma-Aldrich) 16 h prior to infection by intraperitoneal injection. Mock-injected mice received 500 μl of PBS (Corning). At the times after infection indicated below, mice were sacrificed by CO2 asphyxiation. Lungs were harvested into sterile PBS, weighed, and homogenized by pushing tissue through a 70-μm-mesh-size cell strainer. Homogenates were plated for determination of the number of CFU per gram of lung tissue. In the experiments with lung homogenates, lung homogenates from each mouse were plated on 150-mm L-agar plates and at least 1 × 105 CFU per sample was collected for analysis by deep sequencing. All mice were handled in accordance with protocols approved by the Institutional Animal Care and Use Committee of Tufts University.
Flow cytometry analysis.Lung homogenates were stained with anti-Gr1 (Ly6G/Ly6C) phycoerythrin-Cy7 (eBioscience) and anti-CD11b fluorescein isothiocyanate (eBioscience), quantified on an LSRII flow cytometer (BD), and analyzed with FlowJo (version 10.0) software by first gating on the live population, based on forward and side scatter, and then quantifying the Gr1hi CD11b+ population within the live cell population.
Analysis of Tn-seq in WT and α-Ly6G Swiss Webster mice.A schematic of the experimental setup of the Tn-seq screen is presented in Fig. S1. After infecting the mice and collecting the bacteria from the lungs at 33 hpi, the input bacteria and bacteria recovered from the lungs were lysed separately to obtain gDNA using a DNeasy blood and tissue kit (Qiagen). The gDNA was processed as previously described (45). Briefly, the gDNA was sheared and C tails were added to the 3′ ends. PCR was used to amplify the fragments containing the Tn and the region immediately downstream using a primer on the 3′ end of the Tn (MKP030) and a C-tail-complementary primer (MKP028) (Table S6). A nested PCR was then performed using a Tn-specific primer (MKP036) and a primer with a different barcode for each sample (primers BC33 to BC92; Table S6). The samples were multiplexed using comparable molar concentrations and then deep sequenced on an Illumina HiSeq 2500 sequencer in two lanes using a single-read 50-base format and High Output V3 chemistry. The frequency of each insertion mutant recovered in each sample was determined. Insertions were aggregated by gene and annotated based on the K. pneumoniae ATCC 43816 genome (GenBank accession no. NZ_CP009208.1) (50), using a custom Tn-seq pipeline that is housed on the TUCF Genomics Core’s Galaxy server (100).
To obtain the normalized competitive index (nCI) of each gene (i.e., for insertion mutants aggregated by gene), the number of reads for all insertion mutants corresponding to each gene were combined and normalized to the total number of reads for each sample. These values were further normalized to the size of that gene to estimate and account for the expected number of insertions per gene. The nCI for each gene was compared to the nCI of the rest of the genome to obtain a standardized read frequency for each gene in the library, where standardized read frequency = (number of reads with insertions in gene x/total number of reads in the sample)/(size of gene x/size of the genome), where the sizes of the genes and genomes are in base pairs. The size of the genome is 5,374,834 bp (101). The nCI of each gene in each WT and α-Ly6G mouse (WT-nCI and PMN-nCI, respectively) was calculated using the equation nCI = outputstandardized read frequency of insertion mutants in gene x/inputstandardized read frequency of insertion mutants in gene x. Next, the mean nCI of each gene in WT and α-Ly6G mice was calculated by averaging the value within each cohort to obtain WT-mean-nCI and PMN-mean-nCI, respectively. On average, the mean nCI of all genes was 0.27 and 0.43 in WT and α-Ly6G mice, respectively, reflecting the bottleneck of the population (where a value of 1.0 is expected in the absence of a bottleneck).
To assess whether individual genes were recovered at a frequency beyond a normal range, we used a metric that accounted for the mean nCI of each gene and its standard deviation (SDnCI) and the bottlenecks in WT and α-Ly6G mice in combination with a Student’s t-distribution value of t0.99, where t0.99 represents a 99% confidence level. The t0.99 values for WT and α-Ly6G mice were 2.43 and 2.54, respectively, and were derived from a Student’s t-distribution table based on the number (n) of mice used per condition (n − 1, where n = 38 for WT mice and n = 20 for PMN-depleted mice). These metrics were coined WT-MAX99% and PMN-MAX99%, respectively, and were calculated as WT-MAX99% = WT-mean-nCIof gene x + t0.99·SDnCI of gene x and PMN-MAX99% = PMN-mean-nCIof gene x + t0.99·SDnCI of gene x for WT and α-Ly6G mice, respectively. A gene was considered to have been recovered below the normal range in WT and α-Ly6G mouse lungs if the WT-MAX99% or PMN-MAX99% was less than the bottleneck in those mice (i.e., 0.27 for WT mice and 0.43 for α-Ly6G mice). This value considers both the standard deviation for each gene and the stochastic selection imposed by the bottleneck on the bacterial population. Using this metric, 166 and 194 genes were considered significant in WT and α-Ly6G mice, respectively.
Quantification of mucoviscosity.Mucoviscosity was quantified as previously described, with modifications (40). The K. pneumoniae strains indicated above were grown overnight in L at 37°C with aeration. For each culture, 1 ml of overnight culture was centrifuged at 13,000 × g for 4 min. The resulting supernatant was discarded, pellets were resuspended in 1 ml PBS, and the optical density at 600 nm (OD600) of each sample (OD600 total) was measured. The samples were centrifuged at 1,000 × g for 5 min, and the OD600 of the resulting supernatant (OD600 supernatant) was measured. Mucoviscosity was calculated as OD600 supernatant/OD600 total and normalized to the value for the K. pneumoniae WT for each experiment.
Quantification of glucuronic acid.Glucuronic acid was quantified as previously described, with modifications (80). The K. pneumoniae strains indicated above were grown overnight in L at 37°C with aeration. For each culture, 5 × 109 CFU in 1 ml H2O was mixed with 200 μl of 1% Zwittergent 3-12 (Sigma) in 100 mM citric acid (pH 2.0; VWR), and the mixture was incubated at 55°C for 20 min. After centrifugation, 250 μl of supernatant was precipitated with absolute ethanol (final concentration, 80%). The pellet was dissolved in 200 μl of H2O and 1.2 ml of 12.5 mM boric acid (Fisher) in concentrated H2SO4 (Fisher). The mixture was vortexed and boiled at 95°C for 5 min and cooled. Next, 20 μl of 15% 3-hydroxydiphenol (Sigma-Aldrich) was added, and the OD520 was measured and compared to that on a standard curve of d-glucuronic acid (Sigma).
Growth curves in M9-glucose and low levels of metal.The K. pneumoniae strains indicated above were grown overnight in L at 37°C with aeration. The strains were pooled to create a minilibrary. The minilibrary was designed so that 50% was composed equally of mutants with insertions at neutral sites (MKP330, MKP332, and MKP354) and the other 50% was split by the remaining 59 insertion mutants listed in Table S3. The cultures were washed in M9-glucose, plated to obtain gDNA from the inocula, and diluted 1:200 into 4 ml of M9-glucose, L, or L containing 0.125 mM 2,2′-bipyridyl (DIP; Sigma-Aldrich), a chelator of transition metals (70, 102). The diluted samples were incubated for 5 h at 37°C with aeration and then plated in serial dilutions. The resulting colonies from the inocula and samples were collected separately (at least 1 × 105 colonies/sample), processed for gDNA, and sequenced as described above. Growth in M9-glucose and L with DIP (L-DIP) was evaluated by comparing the frequency of each mutant in the inoculum (FreqInoc) and after growth (for M9-glucose, the frequency of each mutant in M9-glucose [FreqM9-glucose]/FreqInoc; for L, the frequency of each mutant in L [FreqL]/FreqInoc; and for L-DIP, the frequency of each mutant in L-DIP [FreqL-DIP]/FreqL).
To test Tn insertion mutants of interest for growth in low levels of metal in 1:1 competition assays with K. pneumoniae WT SPTr (MKP220), the strains were grown separately overnight in L at 37°C, mixed at a 1:1 ratio based on the OD600, and then diluted 1:200 into 4 ml L or L plus 0.25 mM DIP. These samples were incubated at 37°C for 5 h with shaking and then plated in serial dilutions on L and L-KAN plates to quantify the bacteria. Growth in low levels of metal was assessed by comparing the competitive index (CI) in L and that in L-DIP using the equation growth in low metal = (number of KANr CFU in L-DIP/number of kanamycin-susceptible [KANs] CFU in L-DIP)/(number of KANr CFU in L/number of KANs CFU in L) and normalizing the value to the value for the neutral Tn insertion mutant (MKP330) under conditions of metal restriction.
For in-frame deletion mutants, the strains indicated above were grown overnight as described above and 100 μl of 1 × 107 CFU/ml of each strain was placed in a 96-well plate in L with or without 0.25 mM 2,2′-bipyrydyl (DIP; Sigma). The cultures were incubated at 37°C for 16 h with shaking, and the OD630 was measured every 15 min in a plate reader (BioTek Synergy HT).
H2O2 and NO resistance assays.The K. pneumoniae strains indicated above were grown overnight in L at 37°C with aeration. The strains were diluted 1:50 in L, incubated for 2 h at 37°C with aeration, washed twice with M9-glucose, and resuspended in an equivalent volume of M9-glucose. The resulting pools (200 μl) were added to 5 ml M9-glucose, M9-glucose with 1 mM H2O2 (CVS), or 1 mM DETA-NONOate (Cayman Chemical), incubated at 37°C with aeration for 1 h, and plated on L-agar plates.
For minilibrary and 1:1 competition assays, overnight cultures of each strain were treated as described above, resuspended in an equivalent volume of M9-glucose, and then pooled. Samples were treated with the compounds indicated above for 1 h as described above, plated on L and L-KAN plates in serial dilutions, and then quantified to determine the CI, where CI = number of CFU of KANr insertion mutant/number of CFU of KANs K. pneumoniae WT. More than 1 × 104 colonies per sample were collected, processed for gDNA, and deep sequenced as described above. The H2O2, HOCl, and NO resistance of each insertion mutant was calculated using the following equations: frequency of H2O2 resistance = frequency of resistance with H2O2 [
Statistical analysis.With the exception of the Tn-seq analysis, the statistical significance for all mouse and in vitro experiments was assessed using GraphPad Prism (version 7.01) software (GraphPad Software, La Jolla, CA). The analytical tests used are described in the figure legends. A P value of <0.05 was considered significant.
ACKNOWLEDGMENTS
We thank Wai-Leung Ng for advice and use of his robotics system and Erin Green, Lamyaa Shaban, Giang Nguyen, Alyssa Fasciano, Jessie Hem, Ruby Pima-Mimbela, Kim Davis, Edward Geisinger, and Kimberly Walker for many helpful discussions, advice, help with experiments, and/or critical reading of the manuscript.
R.J.S. and M.K.P. were supported by NIH, NIAID, grants 2T32AI007077 and T32AI007422 (to M.K.P.). J.M. and A.L.M. were supported by NIH, NIAID, grants AI113166 and AI107055. A.L.M. was supported by NIGMS grant 1K12GM133314.
R.J.S., M.K.P., A.L.M., and J.M. designed the research and wrote the paper. R.J.S., M.K.P., A.L.M., and C.H.M. performed the research and analyzed the data. A.K.T. and D.W.L. contributed analytic tools.
FOOTNOTES
- Received 16 January 2020.
- Accepted 17 January 2020.
- Accepted manuscript posted online 27 January 2020.
Supplemental material is available online only.
- Copyright © 2020 American Society for Microbiology.
REFERENCES
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵
- 90.↵
- 91.↵
- 92.↵
- 93.↵
- 94.↵
- 95.↵
- 96.↵
- 97.↵
- 98.↵
- 99.↵
- 100.↵
- 101.↵
- 102.↵
