Interaction of multidrug-resistant hypervirulent Klebsiella pneumoniae with components of human innate host defense

ABSTRACT

Klebsiella pneumoniae is an important cause of healthcare-associated infections worldwide. This opportunistic pathogen, known as classical K. pneumoniae (cKp), typically causes infections in individuals with comorbidities. cKp is often multidrug resistant, and treatment options are limited. By comparison, hypervirulent K. pneumoniae (hvKp) can cause infections in healthy individuals outside of the healthcare setting. Notably, there has been emergence of strains containing both multidrug resistance and hypervirulence genotypes (MDR hvKp). Whether these strains can circumvent killing by components of the innate immune system remains incompletely determined. Here, we compared the ability of selected hvKp (ST23 and ST86) and MDR hvKp (ST11 and ST147) clinical isolates to survive in human blood and serum and tested phagocytic killing of the microbes by human neutrophils. On average, the hvKp isolates tested had greater survival in blood and serum compared with MDR hvKp isolates. Compared with MDR hvKp isolates, the hvKp isolates had less surface-bound serum complement following culture in normal serum. Consistent with these findings, the percentage of neutrophils with phagocytosed hvKp isolates was limited (<5%), whereas >67% of neutrophils contained ingested MDR hvKp. Phagocytosis of the MDR hvKp isolates was accompanied by significant bacterial killing (P < 0.05). The inability of neutrophils to ingest and kill these hvKp isolates was, in part, overcome by addition of rabbit antiserum specific for these clinical isolates. The results provide insight into host defense against emerging MDR hvKp and are an initial step toward assessing the potential of a vaccine or immunotherapy approach for treatment of infections.

IMPORTANCE

Klebsiella pneumoniae strains with a combination of multidrug resistance and hypervirulence genotypes (MDR hvKp) have emerged as a cause of human infections. The ability of these microbes to avoid killing by the innate immune system remains to be tested fully. To that end, we compared the ability of a global collection of hvKp and MDR hvKp clinical isolates to survive in human blood and resist phagocytic killing by human neutrophils. The two MDR hvKp clinical isolates tested (ST11 and ST147) were killed in human blood and by human neutrophils in vitro, whereas phagocytic killing of hvKp clinical isolates (ST23 and ST86) required specific antisera. Although the data were varied and often isolate specific, they are an important first step toward gaining an enhanced understanding of host defense against MDR hvKp.

INTRODUCTION

Klebsiella pneumoniae is widely known as a gut commensal microbe and among the leading causes of healthcare-associated infections worldwide (1, 2). For example, Diekema et al. reported that K. pneumoniae ranked as the third most common pathogen isolated from patients with bloodstream infections (2). This opportunistic pathogen, referred to as classical K. pneumoniae (cKp), typically causes infections in individuals with significant comorbidities or underlying susceptibilities (3). In addition to bacteremia, cKp is a frequent cause of urinary tract infection, device-related infection, intraabdominal infection, and pneumonia (3, 4). The relatively high number of infections caused by cKp is compounded by multidrug resistance, most notably carbapenem resistance (1). Treatment options are often limited, and mortality for individuals with infections caused by carbapenem-resistant K. pneumoniae is relatively high (~42%) (5). Multiple studies conducted within the past decade provide support to the idea that vaccine or immunotherapy approaches for treatment of carbapenem-resistant K. pneumoniae are feasible (6 – 10).

A separate group of strains known as hypervirulent K. pneumoniae (hvKp) can cause infections in healthy individuals outside of the healthcare setting (3, 4). Historically, antibiotic resistance in hvKp strains has been limited (11). These pathogens gained notoriety in the 1980s and 1990s for causing pyogenic liver abscesses and accompanying metastatic spread in diabetic individuals (5, 12, 13). K. pneumoniae serotypes K1 and K2 were the most abundant serotypes recovered from these patients, and they remain prominent causes of hvKp infections at present (12, 14, 15). The K. pneumoniae hypervirulence phenotype is associated with the presence of a virulence plasmid, which contains genes encoding molecules involved in siderophore and capsule production (4, 16). Enhanced capsule production is also associated with hypervirulence (17). A hypermucoviscous colony phenotype was traditionally used to designate K. pneumoniae clinical isolates as hvKp (18, 19). Hypermucoviscosity is associated with rmpA or rmpA2 operons (rmpADC or rmpA2D2) and linked specifically to rmpD (20 – 25). Subsequent work by Yang et al. reported similar findings with the rmpA/A2 operons located on plasmids, which is typical of hvKp strains (20, 25, 26). Aerobactin, a siderophore encoded by iucA, has also been shown by multiple studies to contribute to the hypervirulence phenotype of hvKp (27 – 29). Notably, Russo et al. demonstrated recently that iucA, rmpA, rmpA2, iroB, or peg-344, which are present on a virulence plasmid, can be used to identify hvKp with greater than 95% accuracy (30).

Inasmuch as carbapenem resistance and hypervirulence are each conferred mostly by plasmid-encoded molecules, it is perhaps not unexpected that there has been emergence of strains with a combination of multidrug resistance and hypervirulence genotypes and/or phenotypes (MDR hvKp) (31 – 35). The two most widely recognized mechanisms by which MDR hvKp strains emerge involve horizontal gene transfer (36). First, hvKp strains acquire antibiotic resistance and/or resistance plasmids. These microbes have been termed type I MDR hvKp strains (36). Alternatively, multidrug-resistant strains, including those that are carbapenem resistant, acquire the pLVPK virulence plasmid. These strains have been termed type II MDR hvKp (36). For instance, Gu et al. reported deaths of five patients following an outbreak of carbapenem-resistant hvKp caused by an ST11 strain that had acquired a virulence plasmid (32). More recently, Turton et al. identified an hvKp isolate (ST23) that encoded carbapenemase (bla _OXA-48) and a cKp isolate (ST147) that contained bla _NDM-1 and rmpA/rmpA2 (33). Reports of MDR hvKp-associated infections have increased noticeably over the past few years, but our understanding of the interaction of these pathogens with the host innate immune system is incomplete. Such knowledge is optimal for the development of new prophylactic approaches and/or therapeutic measures directed to prevent or treat severe infections. As a step toward this end, we evaluated the ability of selected MDR hvKp clinical isolates to survive during culture in human blood and serum and during phagocytic interaction with human polymorphonuclear neutrophils (PMNs).

RESULTS

Genomic characteristics of hvKp and MDR hvKp clinical isolates

We first generated complete genome sequences of four selected hvKp and MDR hvKp clinical isolates of different sequence types (STs) (ST23/KL1, ST86/KL2, ST11/KL64, and ST147/KL64). The isolates were classified as hvKp or MDR hvKp based on molecular characteristics described above. A virulence plasmid of 204.7–276.1 Kb was identified in each genome (Fig. 1). Virulence plasmids from isolates 53370, 56575, and 53374 showed high sequence similarity to that of the K1 strain NTUH-K2044 (pK2044; >99% blast identity and >95% query coverage), and each contains iro, iuc, rmpADC, and/or rmpA2 (37). In contrast, the virulence plasmid from isolate 56661 (p56661-276.1) lacked iro and rmpADC, harbored truncated rmpA2, and contained a conjugation transfer module (tra gene cluster) (Fig. 1). To gain additional insight into virulence gene diversity among these four STs, we performed genome sequencing with 15 additional clinical isolates (Table S1; Table 1). Among the 19 isolates, 11 contain genes encoding extended-spectrum β-lactamases (ESBLs) (CTX-M-14, CTX-M-15, CTX-M-55, and CTX-M-64), and 13 contain at least one or more genes encoding carbapenemase (KPC-2, KPC-3, NDM-1, OXA-48, and VIM-1). Sixteen isolates harbored at least one virulence gene and most commonly, iuc and iro (Table 1). The rmpA operon (rmpADC) was intact in eight isolates, whereas the rmpA2 operon was intact only in one isolate (53374) (Fig. 1).

Fig 1.

Linear alignment of virulence plasmids. Genes are represented by arrows and colored based on gene function classification. Light blue shading represents regions of homology. Red arrows, virulence gene; green arrows, conjugative transfer genes; blue arrows, replicons; yellow arrows, insertion sequences; gray arrows, other genes.

TABLE 1.

K. pneumoniae clinical isolates ^a

Isolate	MLST	CPS	Location	String test	Virulence genes	ESBL, Carbr	Description
53370	ST23	KL1 (K1)	Taiwan	+	iuc, iro, rmpADC, ΔrmpA2, ybt, clb		hvKp
56598	ST23	KL1 (K1)	South Korea	+	iuc, iro, rmpADC, ΔrmpA2, ybt(ΔybtT), clb	CTX-M-15, NDM-1	MDR-hvKp
56545	ST23	KL1 (K1)	Malaysia	+	iuc, iro, rmpADC, ΔrmpA2, ybt	CTX-M-15, NDM-1	MDR-hvKp
44338	ST23	KL1 (K1)	Columbia	–	iuc, iro, rmpDC(ΔrmpA), ΔrmpA2, ybt, clb,	KPC-2	MDR-hvKp
56575	ST86	KL2 (K2)	Germany	+	iuc,iro, rmpADC, ΔrmpA2		hvKp
56639	ST86	KL2 (K2)	Thailand	+	iuc, iro, rmpADC		hvKp
56622	ST86	KL2 (K2)	Colombia	+	iuc, iro, rmpADC, ybt		hvKp
33654	ST86	KL2 (K2)	Mexico	+	iuc, iro, rmpDC(ΔrmpA)		hvKp
33637	ST86	KL2 (K2)	Mexico	+	iuc, iro, rmpDC(ΔrmpA)		hvKp
53374	ST11	KL64	Taiwan	+	iuc, iro, rmpDC(ΔrmpA), rmpA2, ybt	CTX-M-14, OXA-48	MDR-hvKp
50673	ST11	KL64	ukn	+	iuc, rmpAC(ΔrmpD), ybt	CTX-M-65, KPC-2	MDR-hvKp
50700	ST11	KL64	ukn	+	iuc, rmpDC(ΔrmpA), ΔrmpA2, ybt	CTX-M-65, KPC-2	MDR-hvKp
51578	ST11	KL47	ukn	+	iuc, iro(ΔiroC), rmpDC(ΔrmpA), ΔrmpA2, ybt (Δirp2)	CTX-M-65, KPC-2	MDR-hvKp
49088	ST11	KL47	Taiwan	+	iuc(ΔiucB,ΔiutA), iro(ΔiroC), rmpADC, ΔrmpA2, ybt	CTX-M-65, KPC-2	MDR-hvKp
56661	ST147	KL64	Russia	–	iuc,ΔrmpA2	CTX-M-55, OXA-48, KPC-3	MDR-hvKp
56669	ST147	KL64	Italy	–	iuc, rmpADC, ΔrmpA2	CTX-M-15, NDM-1	MDR-hvKp
46596	ST147	KL64	India	–		CTX-M-15, NDM-1	cKp
46564	ST147	KL64	Greece	–		KPC-2, VIM-1	cKp
49190	ST147	KL20	Turkey	–		CTX-M-15, OXA-48	cKp

^a CPS, capsule polysaccharide; MLST, multilocus sequence type.

There were loss-of-function mutations identified in capsule-encoding genes of multiple K. pneumoniae clinical isolates, including frameshift mutation of wzi in 56622, a stop codon mutation of wcoT and deletion of wzx, wzy, wcoV, KL64_09, wcoU, and wcsF in 50673, deletion of wcoU, wcoT, and wcsF in 50700, deletion of wcqC, wcuT, rmlD, and rmlC in 51578, frameshift mutation of wzx, wzy, wcoU, and wcoT in 46596, a stop codon mutation of rmlB, and deletion of rmlC in 49088, frameshift mutation of wcoU and wcoT in 46564, and a stop codon mutation of wzy in 49190 (Table S1; Fig. S1). We also found an IS5 insertion upstream of wzi in 56661. The potential impact of other nonsynonymous point mutations identified in these genes remains unclear (Table S1). Taken together, the data suggest that capsule synthesis is altered in some of the clinical isolates, which, in turn, could increase susceptibility of the pathogen to components of innate host defense.

Differential survival of hvKp and MDR hvKp clinical isolates in human blood and serum

To better understand the ability of the human innate immune system to protect against emerging MDR hvKp strains, we tested the ability of the four selected hvKp or MDR hvKp clinical isolates (ST23/KL1, ST86/KL2, ST11/KL64, and ST147/KL64) to survival in human blood and serum (Table 1; Fig. 2A and B). Each of these clinical isolates harbors iuc and rmpA or rmpA2 operons on a virulence plasmid, consistent with an hvKp genotype (Table 1; Fig. 1) (30). Notably, MDR hvKp isolates 53374 (ST11) and 56661 (ST147) had significantly reduced survival over time in heparinized human blood, whereas survival of traditional hvKp isolates 53370 (ST23) and 56575 (ST86) was not reduced significantly (Fig. 2A). Moreover, there was a significant difference in survival between the ST11 and ST86 isolates at each of the times tested (Fig. 2A). Consistent with the findings in human blood, the ST11 isolate (53374) had significantly less survival in normal human serum (NHS) compared with the other three clinical isolates tested (Fig. 2B). As a first step toward determining whether the results are specific to the isolates tested or common to the multilocus sequence type (MLST or ST), we extended these experiments to include the other 15 clinical isolates whose genomes were sequenced as described above (Fig. 2C). These 19 selected clinical isolates had varied capacity to survive in NHS, although there were some notable trends among the STs. For example, there was 100% or greater (growth) survival in 25% NHS for three of the ST23 or ST86 isolates tested, whereas each of the ST11 and ST147 isolates were killed at least in part under these culture conditions. These findings are consistent with the results in human blood (compare Fig. 2A and C). Two isolates, 46564 and 49190, were killed completely in 25% NHS, and each lacked a virulence plasmid and had loss-of-function mutations in capsule encoding genes (Table 1; Table S2; Fig. 2C). On the other hand, there were striking within-ST differences in survival. For instance, survival of ST23 isolates 56598 and 56545 was 121.4% ± 32.5% and 2.4% ± 1.2% after culture for 1 h in 83% NHS (Fig. 2C). Although it is tempting to speculate about the contribution of the virulence plasmid to survival in assays with NHS, this correlation was not consistent among all isolates (Table 1; Fig. 2C). Indeed, results with wild-type and plasmid-cured isogenic mutant K1 (ST23) and K2 (ST65) strains provide support to the notion that there is varied contribution of the virulence plasmid to bacterial survival in NHS (Fig. 2D). For example, there was no significant difference in serum susceptibility between wild-type and isogenic plasmid-cured ST23 strains (survival was 85.1% ± 26.3% for the wild-type strain versus 105% ± 24.7% for the plasmid-cured strain) (Fig. 2D). On the other hand, survival of the isogenic plasmid-cured ST65 strain was significantly less than that of the wild-type strain at the highest serum concentration tested (46.3% ± 4.3% for the wild-type strain versus 16.2% ± 5% for the isogenic plasmid-cured mutant strain, P < 0.5) (Fig. 2D). Taken together, these findings indicate that there is differential ability of K. pneumoniae clinical isolates—some closely related by molecular typing methods—to survive in human blood and serum.

Fig 2.

Survival of hvKp clinical isolates in human blood and NHS. Survival of hvKp clinical isolates in heparinized human blood (A) or NHS (B, C, and D) was as described in Materials and Methods. For panels A and B, results are the mean ± standard deviation of the indicated number of separate experiments. For panel A, ^# P < 0.05 for 56575 versus 53374; *P < 0.05 versus 0 min. For panel B, ^#P < 0.05 for the indicated comparisons; *P < 0.05 versus 0 min. Data were analyzed by using a mixed-effects analysis or repeated-measures analysis of variance (ANOVA) and Dunnett’s (30 min and 60 min versus 0 min) or Tukey’s (all four isolates compared at each time point) posttest. Asterisks (symbol color associated) indicate comparison to 0 min. Hashtag color indicates comparison to the symbol of same color. Symbols in panel C)are associated with clinical isolates in Table 1. MLST is color coded. Black symbols, ST23; green symbols, ST86; red symbols, ST11; and blue symbols, ST147. Results are expressed as the mean percent survival of three separate experiments. (D) Survival of wild-type (53370/ST23, 53322/ST65) and isogenic plasmid-cured mutant strains (55731/ST23, 55727/ST675) was compared as described in Materials and Methods. Results are the mean ± standard deviation of three separate experiments. ^# P < 0.05 for 53322 versus 55727 in 83% NHS; *P < 0.05 versus 0 min. Data were analyzed by using a repeated-measures ANOVA and Dunnett’s (all versus 0 min) or Bonferroni’s posttest (selected pairs).

Surface deposition of serum complement components is varied among the STs tested

We next used flow cytometry to determine whether differential survival of selected ST11, ST23, ST86, and ST147 clinical isolates in human blood and NHS correlates with differences in the deposition of serum complement proteins on the bacterial surface (Fig. 3A through C). Following incubation in NHS, there was deposition of C3/C3b/iC3b (herein simplified as C3) and C5b-9 (components of the membrane attack complex) on the surface of each clinical isolate tested (Fig. 3A). However, there were significant differences in complement protein surface binding among these isolates following incubation in 10% or 50% NHS (Fig. 3B and C). For example, in 50% NHS, there was significantly more C3 and C5b-9 on the surface of isolates 53374 (ST11) and 56661 (ST147) compared with isolates 53370 (ST23) and 56575 (ST86) (Fig. 3C).

Fig 3.

Deposition of serum complement onto the surface of hvKp clinical isolates. Binding of C3 or C5b-9 onto the surface of bacteria was determined by flow cytometry. (A) Representative flow cytometry histograms from a single experiment following incubation with 50% NHS. Quantitation of surface-bound C3 or C5b-9 following incubation in 10% (B) or 50% (C) NHS for 20 min at 37°C. Each symbol in panels B and C indicates a separate human serum donor. *P < 0.05 as determined by using a one-way or repeated-measures one-way ANOVA and Tukey’s posttest to correct for multiple comparisons. Symbols †, #, ‡, and § indicate P < 0.05 versus isotype control antibody as determined by a Mann-Whitney test (B) or paired t-test (C). αC3, anti-C3/C3b/iC3b mAb; αC5b-9, anti-C5b-9 + C5b-8 mAbs (Abcam) antibody.

Phagocytosis and killing of hvKp and MDR hvKp clinical isolates by human PMNs

We demonstrated previously that capsule polysaccharide (CPS) of ST258 strains inhibits PMN phagocytosis (6). However, K. pneumoniae CPS is varied among different STs, and the ability of human PMNs to phagocytose ST23, ST86, ST11, or ST147 is unknown or incompletely characterized. To that end, we measured phagocytosis of isolates 53370, 53374, 56575, and 56661 by human neutrophils (Fig. 4A). After 30 min of phagocytic interaction, 67% ± 13% of PMNs ingested 53374, and 82% ± 8.1% ingested 56661 (Fig. 4A). There were on average 1.8 (53374) and 3.1 (56661) ingested bacteria per PMN (Fig. 4B). By comparison, only 4.5% ± 4.1% or 0.5% ± 1.0% of PMNs contained ingested 53370 or 56575 (Fig. 4A). Thus, isolates 53370 (ST23) and 56575 (ST86) were largely resistant to phagocytosis by human PMNs; such results are consistent with the complement deposition data shown in Fig. 3C.

Fig 4.

Phagocytosis of hvKp clinical isolates by human PMNs. (A) hvKp isolates were incubated with human PMNs (bacteria:PMN ratio ~5:1) in 10% autologous NHS for 15 or 30 min as indicated. Percentage of human PMNs containing surface bound + ingested (left panel) or ingested (right panel) bacteria was determined by fluorescence microscopy. (B) Association index (surface bound + ingested) or phagocytic index (ingested only) for data from panel A. Results are the mean ± standard deviation of three or four separate experiments. *P < 0.05 for the indicated comparisons as determined by using a repeated-measures one-way ANOVA and Tukey’s posttest.

Phagocytosis triggers production of superoxide by an NADPH-dependent oxidase, which assembles at the phagosome membrane. Superoxide is converted rapidly to hydrogen peroxide and other reactive oxygen species (ROS) within the phagocytic vacuole. In accordance with the PMN phagocytosis data, serum opsonized 53374 and 56661 elicited production of intracellular PMN ROS and were killed by human PMNs (Fig. 5A and B). In contrast, neither 53370 nor 56575, whether serum opsonized or unopsonized, caused production of PMN ROS. Moreover, there was limited killing of these clinical isolates by human PMNs (Fig. 5B). By comparison, the plasmid-cured mutant ST23 strain (55731) had significantly reduced survival in the PMN killing assays compared with the 53370 wild-type strain (94% ± 9.4% survival for the wild-type strains versus 62 %± 16% for the isogenic plasmid-cured mutant strain, P < 0.5) (Fig. 5C). Results for the ST65 wild-type and plasmid-cured mutant strain pair were similar (Fig. 5C). These findings indicate that the virulence plasmid can contribute to the ability of hvKp strains to circumvent killing by human PMNs, although a specific mechanism remains to be determined.

Fig 5.

PMN activation and concomitant phagocytic killing of hvKp clinical isolates. (A) Production of PMN reactive oxygen species during phagocytic interaction with the indicated clinical isolates. Zymosan and opsonized zymosan (OPZ) were used as positive controls. Data are the mean of at least four separate experiments. (B) Survival of the indicated hvKp clinical isolates during phagocytic interaction with human PMNs. Results are the mean ± standard deviation of five separate experiments. Data were analyzed by using a repeated-measures ANOVA combined with a Dunnett’s or Tukey’s posttest. *P < 0.05 versus 0 min; ^# P < 0.05 versus 53370, 56575, and 56661 at 30 and 60 min; ^† P < 0.05 versus 53374 and 56661 at 60 min; ^‡ P < 0.05 versus 53370, 56575, and 56661 at 60 min. (C) Survival of wild-type (53370/ST23 and 53322/ST65) and isogenic plasmid-cured mutant strains (55731/ST23 and 55727/ST675) during interaction with human PMNs was determined as described in Materials and Methods. Results are the mean ± standard deviation of three separate experiments. ^# P < 0.05 for the comparison of each wild-type and plasmid-cured strain pair at 60 min. *P < 0.05 versus 0 min. Data were analyzed by using a repeated-measures ANOVA and Dunnett’s (all versus 0 min) or Bonferroni’s posttest (selected pairs). (D) Relative titer of naturally occurring human serum antibodies specific for hvKp clinical isolates. Binding of IgA, IgG, or IgM to hvKp clinical isolates was determined in the presence (+NHS) or absence (−NHS) of 10% NHS by flow cytometry as described in Materials and Methods. Each symbol indicates data from a separate human serum donor.

To gain additional insight into the observed differences in PMN phagocytosis and killing of these clinical isolates, we measured levels of naturally occurring K. pneumoniae-specific antibody in sera obtained from seven healthy individuals (Fig. 5C). Indeed, sera from these individuals contained IgG, IgM, or IgA that bound each of the isolates tested (Fig. 5C). Collectively, these results provide support to the idea that healthy individuals have naturally occurring antibody specific for prominent hvKp and MDR hvKp lineages, albeit antibody levels were varied among individuals. Whether these antibodies protect against severe infection in vivo in humans remains to be determined.

Ability of antibody specific for hvKp and MDR hvKp clinical isolates to promote PMN bactericidal activity

We next tested the ability of rabbit polyclonal antisera specific for 53370, 53374, 56575, and 56661, named αhvKp here, to promote phagocytosis and killing of these isolates by human PMNs (Fig. 6 and 7). This rabbit antiserum was generated against a mixture of heat-killed bacteria and resulted in titers as great as 1:100,000, depending on the isolate tested (Fig. 6A and B). Addition of αhvKp to assays with human PMNs increased phagocytosis significantly for each of the isolates tested, although uptake remained limited for 56575 by comparison (Fig. 7A). For example, at 30 min, neutrophil phagocytosis of 53370 increased from 11% ± 6.4% (with preimmune serum) to 61% ± 11% following addition of αhvKp (P < 0.05) (Fig. 7A). By comparison, phagocytosis of 56575 was only 20% ± 4.5% after addition of αhvKp at the same time point (30 min) (Fig. 7A).

Fig 6.

αhvKp titer assessed by flow cytometry. Antisera specific for isolates 53370, 53374, 56575, and 56661 (αhvKp, 1:100 dilution) were generated in rabbits as described in Materials and Methods. (A) Isolate-specific titers of αhvKp as determined by flow cytometry. Data shown are from a single representative experiment. Preimmune rabbit sera (Pre) were used as a baseline. (B) Quantitation of antibody titers. Results are the mean ± standard deviation of at least three separate experiments.

Fig 7.

Isolate-specific ability of αhvKp to promote PMN phagocytosis and bactericidal activity. Antisera specific for isolates 53370, 53374, 56575, and 56661 (αhvKp, 1:100 dilution) was generated in rabbits as described in Materials and Methods. (A) Phagocytosis was determined as described in Materials and Methods. Results are the mean ± standard deviation of three to six separate experiments. Data were analyzed by using a one-way ANOVA and Tukey’s posttest. (B) Ability of αhvKp to promote killing of the indicated clinical isolate by human PMNs. Bactericidal activity was determined by plating for viable colony-forming units as described in Materials and Methods. Results are the mean ± standard deviation of four separate experiments. Data were analyzed by using a repeated-measures ANOVA and Tukey’s posttest.

In the absence of PMNs, αhvKp failed to promote significant killing of these isolates in 10% NHS, although there was a trend for reduced survival (Fig. 7B, open bars). Consistent with data shown in Fig. 5B, isolates 53374 and 56661 were killed significantly by addition of PMNs in these assays (Fig. 7B). Inasmuch as there was significant PMN-mediated killing of these two isolates in 10% NHS (survival after 1 h was reduced to 15.6% ± 10.6% and 7.9% ± 8.9% for 53374 and 56661), addition of αhvKp failed to further reduce this limited survival significantly (Fig. 7B). In contrast, PMN-mediated killing of 53370 and 56575 required—or was optimal in the presence of—αhvKp. For instance, survival of 53370 was 83.7% ± 11.2% with added preimmune sera compared with 11.5% ± 7.6% survival with added αhvKp (P = 0.0002) (Fig. 7B). The observed limited phagocytosis of 56575 in the presence of αhvKp was reflected by comparatively limited PMN bactericidal activity (compare Fig. 7A and B). Although αhvKp had limited capacity to promote PMN-mediated killing of isolate 56575 (ST86), it is noteworthy that this rabbit antiserum was not optimized for opsonophagocytic killing. Indeed, recent studies have shown that K. pneumoniae CPS can block binding of antibodies directed against O-antigen (38), and thus, it is possible that such a phenomenon (i.e., blocking antibody access to noncapsule antigens) contributed to the limited ability of αhvKp to promote PMN phagocytosis and killing of 56575 in our studies. Nonetheless, these findings in aggregate provide further support to the idea that a vaccine approach for prevention of K. pneumoniae infection may be feasible.

DISCUSSION

The interaction K. pneumoniae with the host innate immune system has been studied extensively (reviewed in references 39 – 41). It is widely acknowledged that CPS and lipopolysaccharide promote K. pneumoniae survival during infection. These molecules work in concert to protect the microbe from killing by serum complement and phagocytic leukocytes. In the context of hvKp, early studies linked magA (now known as wzy, which encodes capsule polymerase) to capsule production and the ability of K1 K. pneumoniae (hvKp) to cause pyogenic liver abscesses (42, 43). Lin et al. and Yeh et al. demonstrated that serotype K1 and K2 clinical isolates—hvKp by current definitions—were more resistant to neutrophil phagocytosis compared with non-K1 or non-K2 isolates (44, 45). Several research groups have shown recently that a vaccine approach can protect mice against severe hvKp infection (46 – 48). For example, Diago-Navarro et al. reported the ability of monoclonal antibodies specific for serotype K1 CPS to promote macrophage phagocytosis and protect mice from death following hvKp infection (48). More recently, Feldman et al. showed that a bioconjugate vaccine protected mice against severe pulmonary infection caused by hvKp (NTUH K2044 and ATCC 43816, serotypes K1 and K2) (47). Our results with the ST23 clinical isolate 53370 and αhvKp antisera are consistent with these previous studies (Fig. 7A and B).

By comparison, less is known about the interaction of MDR hvKp with components of innate host defense and whether a vaccine approach can protect against severe infection. In 2018, Gu et al. reported fatal infections caused by a carbapenem-resistant and hypervirulent ST11 strain (an MDR hvKp strain) (32). The authors evaluated human neutrophil bactericidal activity toward these MDR hvKp clinical isolates and showed that hvKp and MDR hvKP ST11 clinical isolates had comparable survival that was significantly greater than that of cKp ST11 isolates (32). Our finding that MDR hvKp isolates, including an ST11 isolate, are phagocytosed and killed by human neutrophils is seemingly at variance with that of Gu et al. (32). However, the phagocytosis assay systems and clinical isolates differ between the two studies. For example, Gu et al. used a phagocytosis assay with neutrophils in suspension, whereas we used a synchronized phagocytosis assay for the current study. Both assays are appropriate but measure phagocytosis in different contexts. The synchronized phagocytosis assay utilizes neutrophils that are primed by contact adherence; thus, phagocytosis is more efficient. This assay in general mimics neutrophil phagocytosis in tissues.

Although the ST11 (53374) and ST147 (56661) isolates that we tested contained a virulence plasmid, they were killed more readily than the ST23 (53370) and ST86 (56575) isolates (Fig. 4B and 7B). The three of four selected hvKp and MDR hvKp isolates (53370, 56575, and 53374) evaluated in our study harbored iuc and rmpADC or rmpA2 operons, which are predictive of strains with a hypervirulence phenotype (30). Isolate 56661 contained intact iuc, but rmpA2 was truncated (Table 1). This mutation might have contributed to the observed limited survival during exposure to components of human host defense. It is also possible that the molecular machinery needed to confer a combination of hypervirulence and antibiotic resistance phenotypes imparts a fitness cost to the microbe (39). Indeed, such a phenomenon has been proposed as the basis for the relatively limited number of high-level (VanA-type resistance) vancomycin-resistant S. aureus isolates that have been recovered from patients over several decades (49). A fitness cost might explain why the convergence of multidrug resistance and hypervirulence in K. pneumoniae is not more widespread. Based on clinical phenotypes and in vitro data, there is likely a spectrum of K. pneumoniae virulence potential in which in the recently emerged MDR hvKp fall somewhere between cKp and hvKp. How an intermediate phenotype fits with the current classification scheme (cKp versus hvKp) merits further discussion. Although many of the clinical isolates tested here meet molecular criteria for classification as hvKp and/or MDR hvKp, whether these microbes have a hypervirulence phenotype in vivo remains to be tested directly. Thus, an important next step is to correlate the in vitro phenotypes reported here with mouse virulence studies in vivo. Such future work is needed to better understand the interaction of these emerging pathogens with the host and thus facilitate development of an optimized vaccine approaches.

MATERIALS AND METHODS

Clinical isolates and culture conditions

K. pneumoniae isolates were selected based on genotype, including presence or absence of virulence genes and/or antibiotic resistance genes, clinical history and relevance, and capsule type (Table 1). Clinical isolates were classified as hvKp or MDR hvKp based on characteristics described previously (18, 30). Bacteria from frozen stocks were cultured overnight in Luria-Bertani (LB) broth. The string test for hypermucoviscosity was performed as described (18, 19). Overnight cultures were diluted 1:100 in LB and incubated further to an OD₆₀₀ of ~0.75. Bacteria were pelleted by centrifugation, washed in Dulbecco’s phosphate-buffered saline (DPBS), and resuspended in DPBS at ~10⁸ CFU/mL. Bacteria were then diluted to 10⁷ CFU/mL in RPMI 1640 medium buffered with 10 mM Hepes, pH 7.2 (RPMI/H), prior to use in assays with PMNs.

Preparation of serum and neutrophils

NHS was prepared with a standard method, which entailed clotting for 30 min at 37°C followed by centrifugation. NHS was used on the day of preparation (assays with PMNs) or frozen and thawed one time before use. Human PMNs (or neutrophils) were prepared using a standard method (50). Purity (99% ± 0.4% granulocytes, n = 20 most recent PMN preparations) and viability (99.7% ± 0.1% exclude propidium iodide, n = 20 most recent PMN preparations) of PMNs in each preparation were determined by flow cytometry (BD FACSCelesta, Becton, Dickinson & Company).

Complement deposition

Deposition of serum complement components on the surface of K. pneumoniae clinical isolates was determined by flow cytometry as described previously (51). Bacteria were pelleted by centrifugation at 2,400 × g for 5 min at ambient temperature, washed with Dulbecco's phophate-buffered saline (DPBS) and incubated in 0.7-mL blocking buffer (5% normal goat serum in DPBS) for 45 min on ice. Bacteria were pelleted by centrifugation and resuspended in 1-mL DPBS. A 100-µL aliquot of resuspended cells was dispensed into new 1.5-mL tubes and labeled with mouse mAbs coupled to fluorescein isothiocyanate (FITC). To detect surface-bound C3 or C5b-9, bacteria were incubated with anti-C3/C3b/iC3b (Cedarlane, clone 7C12) or anti-C5b-9 + C5b-8 mAbs (Abcam, clone aE11) in DPBS followed by FITC-conjugated donkey anti-mouse IgG (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA). FITC-conjugated mouse IgG1 or IgG2a was used as an isotype control antibody (eBioscience, ThermoFisher Scientific, Waltham, MA, USA). Antibody labeling was done on ice for 30 min. A 0.8-mL aliquot of wash buffer (2% goat serum in DPBS) was added to each tube, and cells were centrifuged at 4,300 × g for 4 min at 4°C. Cell pellets were resuspended in 500 µL of wash buffer, and bacteria were analyzed by flow cytometry (BD FACSCelesta cell analyzer, Becton, Dickinson & Company). Fifty-thousand events were collected with BD FACSDiva software v8.0.1.1 (Becton, Dickinson & Company), and data were analyzed by using BD FlowJo v10.8.0 (Becton, Dickinson & Company).

Determination of αhvKp titer by flow cytometry

Rabbit polyclonal antiserum, herein named αhvKp, was generated against a mixture of heat-killed 53370, 53374, 56575, and 56661 that was prepared at mid-log and stationary phases of growth (custom antibody, Pacific Immunology, Ramona, CA, USA). To test antibody titers, bacteria were cultured to mid-exponential phase of growth (OD₆₀₀ = ~0.75), washed once in DPBS, and resuspended in 5% normal goat serum. Following a 30-min incubation on ice, bacteria were pelleted by centrifugation and resuspended in 1-mL DPBS. A 100-µL aliquot of bacteria in DPBS was combined with 100 µL of αhvKp or rabbit preimmune serum (Pre) in a 1.5-mL tube to yield final serum dilutions of 1:4,000 to 1:100,000. The mixture was incubated for 1 h on ice, after which samples were diluted with 800 µL wash buffer. Bacteria were then pelleted by centrifugation and resuspended in 100 µL wash buffer containing goat anti-rabbit antibody coupled to FITC (1:500 final dilution). Samples were stained in the dark on ice for 45 min, diluted with 400 µL wash buffer, and analyzed by flow cytometry as described above.

Assay for intracellular (intraphagosomal) PMN ROS

Opaque (white) 96-well microplates were coated with 20% pooled human serum for 30 min at 37°C. Serum was aspirated, and plates were washed twice with PBS. Bacteria were cultured to midexponential phase of growth (OD₆₀₀ = ~0.75), washed once in PBS, and resuspended at 5 × 10⁷ CFU/mL in RPMI/H. Alternatively, bacteria were washed once in PBS and opsonized in 10% NHS for 30 min at 37°C. After a second wash in PBS, bacteria were suspended in RPMI/H as described above, and 100 µL (5 × 10⁶ CFU) was added to microplate wells according to experimental setup. Catalase (2 µL of 200,000 U/mL catalase stock solution) and superoxide dismutase (SOD, 2 µL of 5,000 U/mL SOD stock solution) were added to each well to eliminate extracellular superoxide and hydrogen peroxide. Human PMNs were suspended in RPMI/H at 10⁷ cells/mL, and luminol (1:1,000 dilution of a 50 mM stock solution in 0.1 M NaOH) was added to a final concentration of 50 µM. PMNs (10⁶ in 100 µL) were added to microplate wells containing bacteria or buffer alone (~200 µL final volume), and plates were centrifuged at 1,400 rpm for 5 min to synchronize phagocytosis. Production of PMN ROS was determined by luminol-enhanced chemiluminescence at 37°C (reads every 2 min) with a BioTek SynergyMX microplate luminometer (BioTek Instruments, LLC).

PMN phagocytosis assays

Synchronized phagocytosis of K. pneumoniae clinical isolates was measured using a published method (52) but with modifications designed to optimize uptake and detection of hvKp (e.g., using αhvKp as a detection antibody). As a first step, 12-mm microscope coverslips were soaked in nitric acid for 48 h, washed/stored in ethanol, and flame sterilized immediately before use. Coverslips were added to tissue culture plates and coated with 60µL 100% NHS for 1 h at 37°C and then washed twice in PBS. Bacteria were prepared as described above, pelleted by centrifugation, resuspended in cold RPMI/H at 2.5 × 10⁷ CFU/mL, and kept chilled on ice until used. PMNs (5 × 10⁵ in 300 µL RPMI/H) were added to wells of 24-well tissue culture plates containing serum-coated glass coverslips. PMNs were allowed to adhere to serum-coated coverslips for 15 min at ambient temperature, and then plates were chilled on ice. Autologous human serum (44 µL; 10% final concentration) was added to each well, along with an ~1:100 dilution of αhvKp or rabbit Pre as needed. A 100-µL aliquot of chilled bacteria in RPMI/H (2.5 × 10⁶ CFU: final bacteria-to-PMN ratio was 5:1) was added to each well, and plates were centrifuged at 450 × g for 7 min at 4°C to synchronize phagocytosis. Assay plates were then transferred to a humidified CO₂ incubator for 15 or 30 min. At the desired time, assay media were aspirated, wells were washed with cold DPBS, and samples were fixed with 4% paraformaldehyde on ice for 10 min and then additionally at ambient temperature for 10 min. Wells were washed three times with DPBS, and then samples were incubated with 300 µL blocking buffer (5% normal goat serum in DPBS) for 30–60 min at ambient temperature. A 1:100 dilution of αhvKp was added as primary antibody in a 300-µL aliquot of blocking buffer. After 30 min at ambient temperature, samples were washed three times in DPBS containing 1.25% normal goat serum (wash buffer). Samples were then stained with 10 µg/mL goat anti-rabbit secondary antibody coupled to AlexaFluor594 (Life Technologies) in 300-µL blocking buffer for 30 min in the dark. Following three rinses with wash buffer, PMNs were permeabilized with 0.3% Triton X-100 in DPBS for 10 min in the dark at ambient temperature. Samples were then washed three times and blocked for 20 min in the dark at ambient temperature with 300 µL blocking buffer. Permeabilized PMNs were then stained sequentially with αhvKp primary antibody and 2 µg/mL goat anti-rabbit secondary antibody coupled to AlexaFluor488 (Life Technologies) using the method described above. After three washes, coverslip samples were mounted onto glass microscope slides using 1 drop of ProLong Gold antifade reagent (Life Technologies). Fifty PMNs from five separate fields of view were scored for each experiment by using a Zeiss Axioscope 5 fluorescence microscope (Zeiss). Total PMN-associated bacteria, including surface bound and ingested, were stained with both AlexaFluor594 AlexaFluor488 (red and green fluorescence), whereas those stained only with AlexaFluor488 were scored as ingested/phagocytosed.

Bactericidal activity assays

To evaluate survival in human blood, hvKp clinical isolates were cultured as described above and 5 × 10⁵ bacteria in 50µL DPBS were combined with 550 µL of heparinized human blood (prepared fresh and kept at ambient temperature until use) in a 1.5-mL microfuge tube. Tubes were rotated gently (horizontal instead of end over end) at 37°C for 0, 30, 60, and 120 min. At the desired time point, an aliquot of assay mixture was diluted in sterile saline and plated on LB agar. Plates were incubated overnight at 37°C, and bacterial colonies were enumerated the following day.

Serum bactericidal activity was determined using a published method (51), except tubes were rotated slowly to mix assay contents instead of by using agitation. Bacteria were cultured as described above, and 0.75–1 × 10⁵ CFU (in 50 µL RPMI/H) was combined in microfuge tubes with NHS diluted to the desired concentration in RPMI/H (final volume of 300 µL). Assay tubes were rotated slowly at 37°C for 1 h, at which point an aliquot of the assay mixture was plated on LB agar. Plates were incubated overnight at 37°C, and bacterial colonies were enumerated the following day. Percent survival was determined with the following equation: percent survival = CFU/mL_{60 min +NHS} / CFU/mL_{60 min -NHS} × 100.

PMN bactericidal activity was determined with a published method (6, 52) but revised to be consistent with conditions in the PMN phagocytosis assays. Bacteria (2.5 × 10⁶ CFU in 100 µL) or assay buffer (100 µL RPMI/H) and 100% autologous NHS (44 µL, 10% final concentration) were added to 24-well culture plates (precoated with 20% NHS at 37°C for 1 h) containing 100 µL RPMI/H (444 µL final volume). Assay plates were allowed to mix for 5 min at ambient temperature, after which human PMNs (5 × 10⁵ cells in 100 µL) were added to the designated wells. Alternatively, αhvKp or Pre was added to assays at ~1:100 final dilution and allowed to diffuse/bind for 5 min prior to addition of PMNs. Plates were then centrifuged at 450 × g for 6 min to synchronize phagocytosis, which was followed by incubation at 37°C for 60 min. Phagocytosis was terminated by addition of saponin (0.1% final concentration) and chilling on ice for 15 min. Aliquots of assay mixtures were diluted in sterile saline and plated on LB agar. Percent survival was determined with the following equation: percent survival = CFU/mL _+PMN /CFU/mL _-PMN × 100.

Flow cytometric analyses of antibody titers in human sera

Human sera and bacteria were prepared as described above. A 200-µL aliquot of bacteria at midexponential phase of growth were washed in 1 mL DPBS and resuspended in 1-mL blocking buffer (2% bovine serum albumin in DPBS) for 1 h on ice. Bacteria were pelleted by centrifugation and resuspended in 1- mL DPBS. A 100-µL aliquot of bacterial suspension was combined with 100 µL of 10% NHS in DPBS (1:20 final dilution). Bacteria-NHS mixtures were chilled on ice for 30 min. Wash buffer (800µL; 0.8% bovine serum albumin in DPBS) was added to each tube, and bacteria were pelleted by centrifugation. Samples were resuspended in 100µL DPBS containing a 1:200 dilution of secondary antibody (goat anti-human IgG-FITC, goat anti-human IgM-FITC, or goat anti-human IgM-FITC). Bacteria were labeled for 30 min on ice, washed once in 800 µL wash buffer, and resuspended in 1 mL wash buffer. Surface-bound antibody was determined by flow cytometry as described above for analysis of surface-bound complement.

Whole genome sequencing and genome analyses

K. pneumoniae isolates were subjected to genome sequencing using Illumina HiSeq (Illumina, Inc.), and raw reads were assembled de novo using SPAdes v3.14 (53). We also performed nanopore sequencing with isolates 53370, 56575, 53374, and 56661 by using the MinION platform (Oxford Nanopore Technologies). A hybrid assembly was conducted with Unicycler v0.49 using the combination of Illuminia Hiseq and MinION sequence reads (54). Virulence genes (iro, iut, rmpADC, and rmpA2) and capsule types were identified by using Kleborate (https://github.com/klebgenomics/Kleborate). Capsule gene mutations were identified by aligning genome sequences to the corresponding capsule reference sequences (accession no. AB924547 for KL1, AB371296 for KL2, AB371289 for KL20, AB924584 for KL47, and AB924600 for KL64) with Snippy (https://github.com/tseemann/snippy). Virulence plasmid comparisons were visualized by Easyfig (https://github.com/mjsull/Easyfig). All genome sequence data were deposited in GenBank under the bioproject accession number PRJNA549322.

Plasmid curing

The curing of virulence plasmids from isolates 53370 and 53322 was conducted using a CRISPR-Cas9 mediated pCasCure platform, as described previously (55). In brief, a 20-bp gRNA target (n20) (TGTCTATACGCAGAACCACA) was cloned into pCasCure-Rif plasmid. The plasmid vector was then transferred into 53370 and 53322 via electroporation, followed by selection on rifampicin plates (100 µg/mL). Single colonies of electrotransformants were then subcultured in LB broth with 0.5% arabinose to induce Cas9 expression for plasmid curing. The plasmid-cured strains, designated 55731 (plasmid-cured 53370) and 55727 (plasmid-cured 53322), were treated with 5% sucrose on LB agar to remove the pCasCure vector. The success of plasmid curing was examined by a multiplex PCR for the plasmid virulence genes rmpA, rmpA2, iutA, and iroN (56), followed by whole genome sequencing to confirm the elimination of plasmids and the absence of off-target mutations.

Statistical analyses

Statistical analyses were performed using Prism 9.5.1 software (GraphPad Software, La Jolla, CA, USA). Normality was assessed by visualization with a Q-Q plot. A Mann-Whitney test or paired t-test was used to compare data sets with two samples. Data with wild-type and isogenic plasmid-cured strains were evaluated by using a repeated-measures ANOVA (for paired samples using human blood donors) and Bonferroni posttest to compare the selected strain pairs at each time point. Data with three or more samples were analyzed by using a one-way ANOVA and Tukey’s or Dunnett’s posttest or a repeated-measures ANOVA or mixed-effects analysis and Tukey’s or Dunnett’s posttest.

ETHICS APPROVAL

Venous blood or heparinized venous blood was obtained from healthy volunteers in accordance with a protocol (01IN055) approved by the Institutional Review Board for Human Subjects at the National Institutes of Health. All volunteers gave written informed consent prior to participation in the study.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/mbio.01949-23.

Figure S1. mbio.01949-23-s0001.pdf.

Loss-of-function mutations in the cps region of K. pneumoniae clinical isolates listed in Table S1.

Supplemental text and table. mbio.01949-23-s0002.docx.

Legend for Figure S1 and Table S1.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.