A STING-adjuvanted outer membrane vesicle nanoparticle vaccine vs. Klebsiella pneumoniae elicits broad capsule-type cross-protection

Abstract

Klebsiella pneumoniae (Kp) is a leading cause of hospital- and community-acquired infections and is increasingly resistant to last-line antibiotics, positioning it as a World Health Organization–designated priority pathogen. Vaccine development has been hindered by the extensive diversity of Kp exopolysaccharide capsules. We developed a cellular nanoparticle (CNP) platform in which bacterial outer membrane vesicles (OMVs) are stably coated onto a STING (stimulator of interferon genes)-adjuvanted nanoparticle core (CNP-KpSTING). The resulting ∼100-nm nanoparticles are homogenous and optimized for recruitment and activation of antigen-presenting cells in draining lymph nodes. CNP-KpSTING vaccination elicited robust antibody responses, enhanced neutrophil-mediated bacterial clearance, and conferred cross-protection against heterologous Kp capsule types in vitro and in lethal pneumonia models. Passive transfer of postvaccination sera reduced bacterial burden and inflammatory cytokines in the lungs of infected mice. These findings establish CNP-KpSTING as a versatile platform that addresses key antigenic barriers inherent to Kp and support OMV-CNPs as a promising foundation for broadly protective vaccines against multidrug-resistant gram-negative pathogens.

Significance statement.

Multidrug-resistant Klebsiella pneumoniae is a major cause of severe pneumonia and sepsis, yet no licensed vaccine currently exists. By stabilizing bacterial outer membrane vesicles on an adjuvanted nanoparticle core, we achieve durable, cross-reactive immunity and high-level protection against lethal pneumonia, foreshadowing a practical path to broad-coverage vaccines for vulnerable populations.

Introduction

Infectious diseases continue to impose a substantial global burden, both in terms of mortality and healthcare infrastructure. Klebsiella pneumoniae (Kp) is a prominent contributor to this burden, responsible for a wide range of illnesses, including pneumonia, urinary tract infections, and bloodstream infections (1, 2). Kp is particularly problematic among intensive care patients and immunocompromised individuals, and the organism is emerging as a leading cause of neonatal sepsis and pneumonia in low- and middle-income countries (3). Kp accounts for 6–17% of hospital-acquired infections worldwide and is increasingly detected in community settings (2, 4), with mortality rates reaching up to 50% in some reports (5). The World Health Organization (WHO) has designated Kp as a high-priority pathogen requiring urgent therapeutic and preventive innovations.

Antimicrobial resistance (AMR) significantly exacerbates this challenge. In 2019, an estimated 3.5 million deaths were attributable to bacterial pathogens associated with AMR—including Kp—with future projections warning of worsening death tolls and economic costs (6). Kp exemplifies this crisis through the global spread of multidrug-resistant (MDR) lineages, particularly carbapenem-resistant and extended-spectrum cephalosporin-resistant clones in healthcare settings (7–10). Approximately one-third of Kp isolates globally are carbapenem resistant, with disproportionately high mortality (10–13). Compounding this is the rise of hypervirulent Kp, particularly ST23 lineages now acquiring carbapenemase genes, which enable severe invasive infections even in previously healthy individuals (14). These trends diminish therapeutic options and underscore the need for preventative strategies. To date, no licensed vaccine against Kp exists (15).

Kp capsules—polysaccharide layers with diverse repeating sugar units—define >80 capsule types (K types), some of which are disproportionately enriched among clinical isolates (16). Capsules are major Kp virulence factors that shield the pathogen from phagocytosis and dampen host immune responses (17–19). Hypervirulent Kp strains often overproduce capsule, contributing to their hypermucoviscous phenotype and invasive potential (20, 21). Most prior vaccine strategies have focused on administering purified polysaccharides or proteins, including capsules, lipopolysaccharides (LPSs), membrane proteins, siderophores, or their receptors (22–27). However, strain-to-strain variability of surface antigens, coupled with capsule-mediated masking, presents a major obstacle to single-antigen vaccines and limits their breadth of protection (15, 28, 29).

Outer membrane vesicles (OMVs) offer a compelling alternative. Gram-negative bacteria naturally release OMVs enriched in surface proteins and lipids, presenting a multivalent antigen repertoire reflective of the parent bacterium (30–33), and thus less vulnerable to single-epitope escape or capsule masking. However, native OMVs are often unstable and polydisperse (34–36), presenting a major translational barrier. Here, we address these bottlenecks by coating Kp OMVs onto an adjuvanted nanoparticle core, creating cellular nanoparticles (CNPs) of ∼100-nm-size optimal for antigen-presenting cell uptake. In choice of adjuvant, we focused on the STING (stimulation of interferon genes) pathway, which senses cytosolic nucleic acids to promote type I interferon responses (37) and has been previously leveraged to enhance cancer vaccine immunogenicity (38). We show that our resulting vaccine candidate, CNP-KpSTING, rapidly activates antigen-presenting cells, induces robust Kp-specific immunoglobulin (Ig)G/IgA/IgM, enhances neutrophil-mediated opsonophagocytic clearance, and protects against lethal pneumonia. Notably, sera raised against a single K-type cross-recognize and functionally target multiple capsule types, addressing long-standing constraints in Kp vaccine breadth.

Results

CNP-KpSTING production yields stable, homogeneous ∼100-nm cellular nanoparticles

We generated the STING-adjuvanted core by self-assembly of manganese ions, cyclic di-adenosine monophosphate (CDA), and 1,2-distearoyl-sn-glycero-3-phosphoethanolamine (DSPE). OMVs from hypervirulent K1 strain NTUH-K2044 (39) were coated onto the core by sonication (Figure 1A). Successful coating was verified by reduced size heterogeneity (polydispersity index) and shifted zeta potential by dynamic light scattering (Figure S1A and B). The resulting ∼100-nm CNP population (Figure 1B) falls within the optimal size range for antigen-presenting cell uptake (40–42) and remained stable for >1 week at room temperature or 37 °C (Figure S1C and D). CNP-KpSTING did not induce cytotoxicity in human peripheral blood mononuclear cells, as assessed by lactate dehydrogenase (LDH) release at 4 or 24 h (Figure S1E). Silver staining confirmed consistent OMV protein profiles between Kp NTUH-K2044 production batches (Figure S1F). These results demonstrate precise, reproducible coating and stability of CNP-KpSTING.

Figure 1.

Assembly and characterization of CNP-KpSTING nanoparticles. A) CNP-KpSTING are produced by first generating a STING-adjuvant core through the self-assembly of manganese ions, CDA and DSPE-H11, followed by sonication with Kp OMVs at a 1:1 weight ratio. B) Dynamic light scattering shows the representative size distribution of uncoated Kp OMVs compared with fully formed CNP-KpSTING, demonstrating reduced polydispersity and formation of a stable, homogeneous ∼100-nm nanoparticle population.

CNP-KpSTING vaccination recruits and activates antigen-presenting cells in draining lymph nodes

To evaluate the early immunological effects of the vaccine, C57BL/6 mice were vaccinated subcutaneously with 0.1 µg of CNP-KpSTING or, as a control, human red blood cell membrane-coated STING nanoparticles (RBC-STING), which retain the STING-adjuvant core but lack bacterial antigens. Draining inguinal lymph nodes were harvested 24 h later and analyzed by flow cytometry (gating strategy in Figure S2A–J). Mice receiving CNP-KpSTING exhibited a marked increase in the absolute numbers of B cells (CD19⁺), macrophages (F4/80⁺), and dendritic cells (CD11c⁺) compared with the RBC-STING control (Figure 2A and C). The proportion of live cells within the dendritic cells and macrophage subsets was also significantly higher following CNP-KpSTING vaccination (Figure 2D). Importantly, dendritic cells in CNP-KpSTING–vaccinated mice expressed elevated levels of the co-stimulatory molecules CD80, CD86, and CD40, indicative of enhanced activation and antigen-presenting capacity (Figure 2B, E, and F). These findings demonstrate that CNP-KpSTING nanoparticles not only recruit antigen-presenting cells to the draining lymph nodes but also promote their activation within 24 h. The coordinated expansion and maturation of B cells, macrophages, and dendritic cells suggest that CNP-KpSTING establishes a local immunological environment conducive to antigen uptake, processing, and the priming of adaptive immune responses.

Figure 2.

CNP-KpSTING vaccination enhances recruitment and activation of immune cells in draining lymph nodes. C57BL/6 mice were vaccinated subcutaneously with 0.1 µg of CNP-KpSTING or RBC-STING on the flank. After 24 h, inguinal lymph nodes were harvested for single-cell isolation and flow cytometry analysis. A) Cells were stained with specific antibodies and analyzed by flow cytometry; gates were established using single-stained, unstained, and fluorescence-minus-one controls. B cells were identified as live CD19⁺, macrophages as live F4/80⁺, and dendritic cells as live CD11c⁺. B) Dendritic cell (DC) activation was assessed by expression of co-stimulatory markers CD80, CD86, and CD40. C) Total number of B cells, DCs, and macrophages. D) Percentage of each immune subset relative to the total lymph node population. E) Total cell counts for the indicated subsets shown in (C). F) Percentage of dendritic cells expressing activation markers. Each dot represents an individual mouse. ns, not significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. MP, macrophages.

CNP-KpSTING vaccination protects mice from lethal Kp lung infection

To establish lethal challenge models, nonimmunized C57BL/6 mice were infected intratracheally with either Kp NTUH-K2044 (K1) or ATCC 43816 (K2). Both strains were highly virulent, with LD₈₀ values of 2 × 10³ colony-forming unit (CFU) and 5 × 10³ CFU, respectively (Figure 3A and B). We next evaluated vaccine efficacy. Mice were immunized subcutaneously with three weekly doses of 1 µg CNP-KpSTING (derived from either the K1 or K2 strain) or with RBC-STING as a control (Figure 3C). No overt signs of distress or illness were observed in vaccinated mice, consistent with prior safety data for a Pseudomonas aeruginosa–STING vaccine (43). On day 28, mice were challenged intratracheally with an LD₈₀ inoculum of Kp, and clinical symptoms and weight were monitored twice daily. When infected with 2 × 10³ CFU of K1 bacteria, both CNP-KpSTING vaccine groups exhibited significantly higher survival rates (70–80%) compared with RBC-STING–vaccinated mice (10%; Figure 3D). Notably, mice vaccinated with CNP-KpSTING derived from the K2 strain exhibited similar protection against K1 challenge, demonstrating cross-protection between capsule types. Reciprocal experiments using 5 × 10³ CFU of the K2 strain confirmed bidirectional cross-protection, with significantly improved survival in both CNP-KpSTING–vaccinated groups relative to RBC-STING controls (Figure 3E). Assessment of serum antibodies demonstrated robust induction of anti-Kp IgG titers throughout the vaccination series in both CNP-KpSTING groups, whereas RBC-STING controls remained at baseline (Figure 3F and G). Comparable IgG titers were observed against both K1 and K2 strains, consistent with the observed cross-protective phenotype (Figure 3F and G). Together, these findings show that CNP-KpSTING vaccination is well-tolerated, elicits strong IgG responses, and confers protection against otherwise lethal pneumonia caused by hypervirulent Kp, including across distinct capsule types.

Figure 3.

CNP-KpSTING vaccination protects mice from lethal Kp lung infection and induces robust IgG responses. C57BL/6 mice were challenged intratracheally by Kp and monitored for survival. A) Lethality curves of nonimmunized mice infected with NTUH-K2044 (K1), establishing an LD₈₀ of 2 × 10³ CFU. B) Lethality curves of nonimmunized mice infected with ATCC 43816 (K2), establishing an LD₈₀ of 5 × 10³ CFU. C) Vaccination regimen: mice received three weekly subcutaneous doses of 1 µg CNP-KpSTING (derived from K1 or K2) or RBC-STING, followed by intratracheal challenge on day 28. Clinical signs and body weight were monitored twice daily. n = 10 mice per group. D and E) Kaplan–Meier survival curves of vaccinated mice challenged with Kp NTUH-K2044 (D) Kp ATCC 43816 (E). F) Anti-NTUH-K2044 IgG titers corresponding to (D). G) Anti-ATCC 43816 IgG titers corresponding to (E). Titers were measured by ELISA on serum obtained days 0, 7, 14, and 28. D and E) Statistical analysis by Kaplan–Meier log-rank test. F and G) Two-way ANOVA with Bonferroni correction for multiple comparisons. ns, not significant. Each dot represents an individual mouse. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. DPI, day postinfection.

CNP-KpSTING vaccination in rabbits induces robust antibody responses that mediate protection in passive-transfer experiments

To evaluate the immunogenicity of CNP-KpSTING in a second species, two New Zealand White rabbits were immunized with four subcutaneous doses totaling 1 mg over 8 weeks. Serum collected at day 0 (prevaxx) and day 56 (postvaxx) was analyzed for antibody responses (Figure S3A). Postvaxx sera demonstrated markedly elevated titers of IgG (∼4 log₁₀ increase), IgA (∼2 log₁₀ increase), and IgM (∼3 log₁₀ increase) against heat-killed Kp NTUH-K2044 (K1), confirming induction of robust humoral immunity (Figure 4A). To evaluate functional antibody activity, an opsonophagocytic killing (OPK) assay was performed using freshly isolated human neutrophils. Postvaxx sera significantly enhanced neutrophil-mediated killing of NTUH-K2044 compared with prevaxx sera, normalized to the neutrophil-only control (Figure 4B). No significant differences in OPK activity or enzyme-linked immunosorbent assay (ELISA) titers were observed at intermediate time points, indicating that the antibody response plateaued by day 35 in both rabbits (Figure S3B–E). Specificity testing confirmed that vaccine-elicited antibodies selectively targeted Kp, as no enhancement of neutrophil killing was observed against other Gram-negative bacteria (Escherichia coli, P. aeruginosa, Acinetobacter baumannii), indicating Kp specificity (Figure S4A). Consistent with this, postvaxx IgG titers against P. aeruginosa and A. baumannii remained at baseline, while E. coli titers showed only a modest increase (Figure S4B). To assess in vivo protective efficacy, 200 µL of pre- or postvaxx rabbit sera was transferred intravenously (retro-orbitally) into C57BL/6 mice, followed 24 h later by intratracheal challenge with 2 × 10³ CFU NTUH-K2044 (Figure 4C). Mice were euthanized 48 h postinfection for bronchoalveolar lavage (BAL) and lung collection. Postvaxx serum transfer significantly reduced bacterial burden in both BAL fluid and lung tissue (Figure 4D) and was associated with lowered BAL concentrations of interleukin-16 (IL-6) and tumor necrosis factor α (TNFα), consistent with reduced inflammation and improved bacterial clearance.

Figure 4.

CNP-KpSTING vaccination in rabbits elicits robust, Kp-specific antibodies that enhance neutrophil killing and confer protection by passive transfer. New Zealand White rabbits were vaccinated with four subcutaneous doses totaling 1 mg CNP-KpSTING and bled on day 0 (prevaxx) and day 56 (postvaxx). (A) IgG, IgA, and IgM titers against heat-killed Kp NTUH-K2044 by ELISA. B) OPK assay: fold-change in CFU recovery for NTUH-K2044 using human neutrophils, normalized to the (+) neutrophil/(−) serum condition (n replicates = 6; n human donors = 4). C) Passive-transfer design: C57BL/6 mice received 200 µL pre- or postvaxx rabbit serum intravenously, and then were challenged intratracheally 24 h later with 2 × 10³ CFU NTUH-K2044; mice were euthanized 48 h postinfection for BAL and lung harvest (n = 10 per group; two pooled experiments). D) CFU enumeration from BAL fluid and lung homogenates. E) BAL cytokines IL-6 and TNFα measured by ELISA. Each dot represents an individual mouse. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

CNP-KpSTING–elicited rabbit antibodies cross-recognize diverse Kp capsule types

Given that CNP-KpSTING vaccination conferred cross-protection between K1- and K2-derived formulations in mice, we next investigated whether vaccine-induced antibodies could recognize additional capsule types commonly found in clinical isolates (16). Capsular heterogeneity has long posed a major obstacle to the development of broadly protective Kp vaccines. Serum from CNP-KpSTING–vaccinated rabbits cross-reacted broadly, with IgG titers against all capsule types tested comparable to those measured against the homologous NTUH-K2044 (K1) vaccination strain (Figure 5A). Functional OPK assays further demonstrated that postvaxx sera significantly enhanced neutrophil-mediated killing across all capsule types tested (Figure 5B). Recovered CFUs were reduced by 50 to 90%, indicating robust—albeit variable—killing activity across strains. Importantly, opsonic enhancement was observed even against Kp isolates with inherent complement sensitivity, confirming that vaccine-induced antibodies provided additional protective benefit beyond innate mechanisms. However, the degree of bacterial killing varied by strain and did not correlate directly with capsule type or antibody binding strength. Notably, significant variability was seen even within the K2 capsule family, and no consistent relationship emerged between IgG titers and neutrophil clearance activity. These findings suggest that opsonic efficacy may be influenced by factors such as capsule expression level, structural differences, or other strain-specific outer membrane components (Figure S5). The capacity of CNP-KpSTING to generate cross-reactive yet functionally diverse antibody responses supports its potential as a broadly protective vaccine platform against Kp.

Figure 5.

CNP-KpSTING-induced antibodies cross-recognize diverse Kp capsule types. A) IgG titers by ELISA against heat-killed strains representing multiple capsule types, measured in sera from rabbits vaccinated with NTUH-K2044-derived CNP-KpSTING (K1). B) OPK assays showing fold-change in recovered CFUs of diverse capsule-type strains after incubation with pre- or postvaxx rabbit sera. *q < 0.05.

Discussion

Kp poses a formidable public health threat by combining extensive capsule diversity, hypervirulent lineages, and MDR, making both treatment and prevention exceptionally challenging. Despite its clinical importance, no licensed Kp vaccine exists, and prior candidates have been limited by narrow coverage and poor durability of protection (23, 28, 44–47). In this study, we show that coating Kp-derived OMVs onto a STING-adjuvanted nanoparticle core (CNP-KpSTING) creates a stable, immunogenic formulation capable of inducing broad antibody responses, conferring cross-protection in stringent murine pneumonia models, and eliciting rabbit sera that opsonize diverse capsule types. These findings suggest that nanoparticle stabilization of OMVs can address several of the historical roadblocks in Kp vaccine development by presenting a native, multivalent antigen repertoire in a potent immunostimulatory context.

Capsule variability has long undermined vaccine strategies, with more than 80K types defined serologically (16) and over 140 distinct K loci identified genomically (48). By incorporating complete OMVs into the KpSTING formulation, we adopt a multivalent approach that minimizes the risk of serotype-restricted efficacy, as evidenced by the observed cross-reactivity of vaccine-elicited antibodies across multiple capsule types. Both the K1- and K2-derived formulations showed strong cross-protection in mice, and rabbit antisera recognized all tested capsule types. While we cannot exclude the possibility that some capsule types may escape recognition, this limitation could be addressed by producing and combining OMVs from additional strains—analogous to the stepwise model in pneumococcal conjugate vaccines (49).

A major advantage of this platform is the ease and scalability of OMV isolation, which is far less labor intensive and costly than mass production of individual purified antigens. Additionally, the bacterial capsule likely does not obscure all surface antigens, suggesting that noncapsular epitopes contribute meaningfully to the immune response. The multivalent presentation may also reduce the likelihood of immune escape, as targeting multiple antigens complicates bacterial evasion. Further studies are warranted to define the dominant antigens, their abundance, and their relative contribution to protection. Prior studies have shown that mice inoculated with capsule-deficient variants can resist subsequent challenge with capsulated strains (19), underscoring the potential of noncapsular targets.

OMVs contain a rich mixture of outer membrane proteins, LPSs, and other immunogenic components, and defining which of these are responsible for protection will guide rational formulation and inform safety assessments. In particular, the potential for LPS-mediated toxicity must be carefully addressed. Nonetheless, our findings represent an important step forward, demonstrating protective efficacy in rigorous intratracheal challenge models. Such a vaccine could reduce the burden of MDR Kp infection, including hospital-associated pneumonia and sepsis, while also decreasing healthcare costs associated with prolonged hospitalization (50). Moreover, a preventive Kp vaccine could reduce antibiotic use, thereby slowing the emergence of resistance to first- and last-line agents.

Adjuvant selection plays a pivotal role in achieving robust immune activation. Here, we employed a STING-adjuvanted core based on its established ability to promote dendritic cell activation and enhance both systemic and mucosal immunity (51–53). However, the nanoparticle production pipeline is modular and not inherently dependent on STING. We have previously shown that poly(lactic-co-glycolic acid) (PLGA), a Food and Drug Administration–approved inert polymer, can serve as an effective core for OMV stabilization. In scenarios where STING activation is unnecessary or where an alternate adjuvant—such as alum—is preferred, OMV-coated PLGA nanoparticles may be used in combination with the desired adjuvant. Importantly, the coating mechanism itself appears integral to efficacy: prior work showed that uncoated A. baumannii OMVs failed to elicit comparable immune activation relative to coated counterparts (54). Optimization of dose and schedule is ongoing, but data from P. aeruginosa STING vaccination suggested efficacy can be achieved with as little as 0.01 µg (43). A regimen involving an initial dose followed by a single booster at one month—timed with germinal center maturation—may strike an optimal balance between immunogenicity, cost, and patient compliance.

Beyond systemic protection, we observed robust IgA induction following CNP-KpSTING vaccination, supporting its potential for mucosal immunity. Since Kp is a common cause of urinary tract infections, mucosal IgA may have a secondary benefit to reduce colonization and disease in the urinary tract (55, 56). The gut also represents a critical reservoir for Kp, with intestinal carriage increasingly recognized as a risk factor for extraintestinal infections and hospital outbreaks (57–59). Investigating whether CNP-KpSTING vaccination reduces or eliminates gastrointestinal carriage could further expand its clinical utility. This OMV-nanoparticle platform, given its scalable production, antigenic breadth, and efficacy in prior preclinical studies with A. baumannii (54) and P. aeruginosa (43), could provide a versatile foundation for next-generation antimicrobial vaccines against a variety of gram-negative pathogenic threats.

Materials and methods

The materials and methods, describing in detail the reagents, bacterial strains, and culture condition; OMV isolation; RBC-membrane derivation; CNP-KpSTING preparation and characterization; rabbit immunization; opsonophagocytic assays; murine models for infection, active and passive immunization; ELISA; flow cytometry; and statistical analysis are provided in SI Appendix including Table S1.

Ethics declarations

Animal experiments followed all ethical regulations for animal research and were carried out in accordance with the rules and regulations of the Institutional Animal Care and Use Committee, which was approved by the UC San Diego IRB protocol S00227M. The blood collection protocol was approved by the UC San Diego Human Research Protection Program under IRB protocol #131002. All blood donors signed a written informed consent form.

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Competing Interest

The reported technology is covered by patent applications related to a stealth-mode spin-off company. E.B., V.N., and L.Z. are equity holders in this company.

Funding

This work was supported by a Swiss National Science Foundation Postdoctoral Fellowship (H.B.), an A.P. Giannini Foundation Postdoctoral (E.B.), and an NIH grant R01AI176554 (V.N.).

Author Contributions

Hervé Besançon (Conceptualization, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing - original draft, Writing—review & editing), Daniel Sun (Investigation, Writing—review & editing), Kalisa Kang (Investigation, Writing—review & editing), Samuel Kim (Investigation, Writing—review & editing), Samira Dahesh (Investigation, Writing—review & editing), Sydney Morrill (Investigation, Writing—review & editing), Shawn M. Hannah (Investigation, Writing—review & editing), Anh T. P. Ngo (Investigation, Writing—review & editing), Liangfang Zhang (Conceptualization, Funding acquisition, Investigation, Methodology, Writing—review & editing), Victor Nizet (Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Writing - original draft, Writing—review & editing), and Elisabet Bjånes (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Validation, Visualization, Writing - original draft, Writing—review & editing)

Data Availability

All study data are included in the article and Supplementary material.