Induction of Protective Antiplague Immune Responses by Self-Adjuvanting Bionanoparticles Derived from Engineered Yersinia pestis

A Yersinia pestis mutant synthesizing an adjuvant form of lipid A (monophosphoryl lipid A, MPLA) displayed increased biogenesis of bacterial outer membrane vesicles (OMVs). To enhance the immunogenicity of the OMVs, we constructed an Asd-based balanced-lethal host-vector system that oversynthesized the LcrV antigen of Y. pestis, raised the amounts of LcrV enclosed in OMVs by the type II secretion system, and eliminated harmful factors like plasminogen activator (Pla) and murine toxin from the OMVs.

ABSTRACT

A Yersinia pestis mutant synthesizing an adjuvant form of lipid A (monophosphoryl lipid A, MPLA) displayed increased biogenesis of bacterial outer membrane vesicles (OMVs). To enhance the immunogenicity of the OMVs, we constructed an Asd-based balanced-lethal host-vector system that oversynthesized the LcrV antigen of Y. pestis, raised the amounts of LcrV enclosed in OMVs by the type II secretion system, and eliminated harmful factors like plasminogen activator (Pla) and murine toxin from the OMVs. Vaccination with OMVs containing MPLA and increased amounts of LcrV with diminished toxicity afforded complete protection in mice against subcutaneous challenge with 8 × 10⁵ CFU (80,000 50% lethal dose [LD₅₀]) and intranasal challenge with 5 × 10³ CFU (50 LD₅₀) of virulent Y. pestis. This protection was significantly superior to that resulting from vaccination with LcrV/alhydrogel or rF1-V/alhydrogel. At week 4 postimmunization, the OMV-immunized mice showed more robust titers of antibodies against LcrV, Y. pestis whole-cell lysate (YPL), and F1 antigen and more balanced IgG1:IgG2a/IgG2b-derived Th1 and Th2 responses than LcrV-immunized mice. Moreover, potent adaptive and innate immune responses were stimulated in the OMV-immunized mice. Our findings demonstrate that self-adjuvanting Y. pestis OMVs provide a novel plague vaccine candidate and that the rational design of OMVs could serve as a robust approach for vaccine development.

INTRODUCTION

Plague caused by Yersinia pestis, a Gram-negative bacterium, has been one of the most feared infectious diseases in human history, resulting in over 200 million deaths (1, 2). Pneumonic plague is highly contagious and easily transmitted from person to person through airborne droplets, resulting in the rapid onset of disease and a near 100% mortality rate if left untreated for more than 24 h postexposure (3, 4). Owing to these deadly features, Y. pestis is considered a potential biological weapon and is classified as a tier 1 select agent by the CDC (5). Although plague causes only a few thousand cases worldwide annually (6), this disease remains a public health concern due to frequent sporadic outbreaks in different regions (7–9). Moreover, there are increasing concerns about the occurrence of multiple antibiotic-resistant Y. pestis strains (10, 11). For long-term prophylaxis, vaccination is considered an efficient strategy. Currently, no licensed plague vaccines are recommended by the WHO.

Outer membrane vesicles (OMVs) are nanosized lipid particles released by a diverse range of Gram-negative bacteria that are enriched in protein, polysaccharide, and lipid components, including a plethora of potent immunogens (12). By retaining the composition of the antigenic pathogen surface, OMVs elicit innate immunity as well as prime humoral and cell-mediated immune responses (13). A licensed OMV vaccine against Neisseria meningitidis has been proven safe and protective in humans (14). Recently, the human commensal gut bacterium Bacteroides thetaiotaomicron was engineered to deliver LcrV or F1 antigen by its OMVs. Intranasal vaccination with those OMVs elicited substantive and specific immune and antibody responses to LcrV or F1 in both serum and respiratory tract, but no protection data were provided in this study (15). We hypothesized that OMVs directly from Y. pestis, comprising a broad spectrum of immunogens, would provide the theoretical advantage of simultaneously priming immunity against many antigens instead of just one or two antigens (i.e., LcrV or F1-LcrV). Y. pestis has been shown to release native OMVs under physiological conditions (16, 17), but the use of OMVs from Y. pestis as a vaccine has never been attempted.

Y. pestis is capable of synthesizing stimulatory hexa-acylated lipid A due to the presence of LpxP (palmitoleoyltransferase) in the flea (26°C). In mammals, in contrast, the bacterium produces tetra-acylated lipid A at a typical mammalian temperature of 37°C due to the absence of LpxL (lauroyltransferase) (18), thereby allowing the bacterium to evade host innate immune surveillance because tetra-acylated lipid A is not recognized by Toll-like receptor 4 (TLR4) (19). Previously, we generated a Y. pestis mutant that produced hexa-acylated lipid A independently of temperature by incorporating the Escherichia coli lpxL gene into the Y. pestis chromosome and then adding to the mutant strain the Francisella tularensis lpxE gene, which encodes lipid A 1-phosphatase. This enzyme removed the 1-phosphate of hexa-acylated lipid A to predominantly yield 1-dephosphorylated hexa-acylated lipid A (monophosphoryl lipid A, MPLA) (20). MPLA, an endotoxin derivative, exhibits potent adjuvant activity but is 100- to 10,000-fold less toxic than the native lipid A (biphosphoryl lipid A) (21) and has been approved by U.S. and European authorities as a vaccine adjuvant in humans (22). Based on these results, we attempted to investigate the effects on OMV biogenesis in Y. pestis of remodeling lipid A with the addition of a fatty acid chain or the subtraction of a phosphate and to evaluate the protective immunity of the self-adjuvanting OMV against lethal challenges with Y. pestis.

RESULTS

Lipid A 1-dephosphorylation of Y. pestis affects bacterial morphology and increases OMV biogenesis.

Previously, Y. pestis KIM6+ isogenic mutants χ10015 (ΔlpxP::P_lpxLlpxL) and χ10027 (ΔlpxP::P_lpxLlpxL ΔlacZ::P_lpplpxE) (Table 1) produced conventional hexa-acylated lipid A and 1-dephosphorylated hexa-acylated lipid A at 28°C and 37°C, respectively (20). χ10027 was more susceptible to polymyxin B than KIM6+ and χ10015 (20), suggesting that lipid A 1-dephosphorylation might influence bacterial membrane stability and morphology. Thus, we employed transmission electron microscopy (TEM) to visualize all three strains when they were cultured in heart infusion broth (HIB) at 28°C for 14 h and then incubated at 37°C for 4 h. The morphologies of KIM6+ and χ10015 were observed as a mixture of coccus and bacillus shapes (Fig. 1A and B), while the χ10027 strain had been completely altered into cocci (Fig. 1C). The χ10027 strain had a higher percentage of cell wall bulges that were localized to the bacterial surface than the other two strains (Fig. 1).

TABLE 1.

Strains and plasmids used in this study

Strain or plasmid	Genotype or relevant characteristics	Reference or source
E. coli strains
χ6212	F– λ– ϕ80 Δ(lacZYA-argF) endA1 recA1 hsdR17 deoR thi-1 glnV44 gyrA96 relA1 ΔasdA4	69
χ7213	thi-1 thr-1 leuB6 fhuA21 lacY1 glnV44 ΔasdA4 recA1 RP4 2-Tc::Mu [λpir]; Kmr	70
Y. pestis strains
KIM6+ (pCD1Ap)	pCD1Ap, pMT1, pPCP1, Pgm+	Wild-type strain
KIM6+	pCD1- pMT1, pPCP1, Pgm+	20
χ10015	ΔlpxP::PlpxLlpxL	20
χ10027	ΔlpxP::PlpxLlpxL ΔlacZ::PlpplpxE	20
YPS1	Δasd12 KIM6+	This study
YPS2	Δasd12 χ10015	This study
YPS3	Δasd12 χ10027	This study
YPS4	Δasd12 Δymt50 KIM6+	This study
YPS5	Δasd12 Δ ymt50 χ10015	This study
YPS6	Δasd12Δ ymt50 χ10027	This study
YPS7	Δasd12 Δymt50 KIM6+ pPCP1-	This study
YPS8	Δasd12 Δymt50 χ10015 pPCP1-	This study
YPS9	Δasd12 Δymt50 χ10027 pPCP1-	This study
Plasmids
pRE112	Suicide vector, Cmr, mob- (RP4)R6K ori, sacB	71
pYA3342	Asd+; pBR ori	69
pYA3493	Asd+; β-lactamase signal sequence-based periplasmic secretion, pBR ori	69
pYA4373	cat-sacB cassette in sites of PstI and SacI pUC18	72
pSMV12	Full-length Y. pestis lcrV cloned into pYA3342	This study
pSMV13	Full-length Y. pestis lcrV cloned into pYA3493	This study
pSMV25	Flanking regions of Δasd of Y. pestis cloned into XmaI and KpnI sites of pRE112	This study
pSMV26	Replication origin of pPCP1 cloned into pYA4373	This study

Fig 1.

Comparison of morphological alterations in Y. pestis strains by TEM imaging. Samples of strains Y. pestis KIM6+ (A), χ10015 (ΔlpxP:: P_lpxLlpxL) (B), and χ10027 (ΔlpxP:: P_lpxLlpxL ΔlacZ:: P_lpplpxE) (C) were prepared by conventional staining with 1% aqueous uranyl acetate as described in the Materials and Methods. The results are representative of three repeated experiments.

To determine the effect of lipid A remodeling on Y. pestis OMV biogenesis, we initially confirmed that OMV biogenesis occurred in each Y. pestis strain cultured at 28°C for 14 h and then incubated at 37°C for 4 h. The results showed that KIM6+, χ10015, and χ10027 all produced OMVs, but the sizes of the OMVs from χ10027 were much smaller than those from Y. pestis KIM6+ and χ10015 (Fig. 2A and Fig. S1 in the supplemental material). Proteomic analysis by mass spectrometry showed that 293 proteins were detectable in OMVs from all three strains (Table S1) and included 12.3% outer membrane proteins, 4.1% periplasmic proteins, and 83.6% cytoplasmic proteins (Fig. 2B). OMVs from all three strains contained specific major outer membrane proteins, such as Pla, Ymt, Ail (OmpX), OmpA, F1, and Psn (Table S1). The total protein amounts and lipid contents in OMVs from χ10027 were increased ∼16- and ∼150-fold compared to those from KIM6+ and χ10015, respectively (Fig. 2C). OMVs from KIM6+ and χ10015 showed comparable total protein amounts and lipid contents (Fig. 2C). The amounts of several outer membrane proteins (Psn, OmpA, Pla, and F1) were comparable among OMVs isolated from KIM6+, χ10015, and χ10027 (Fig. 2D), but the total protein amounts in OMVs from χ10027 were clearly higher than those from KIM6+ and χ10015 (Fig. 2E). Thus, the results suggested that lipid A 1-dephosphorylation in χ10027 increased OMV biogenesis, while lipid A acylation in χ10015 did not.

Fig 2.

Analysis of Y. pestis outer membrane vesicles (OMVs). (A) TEM of OMVs purified from Y. pestis KIM6+ culture supernatants. (B) Subcellular distribution of proteins present in Y. pestis KIM6+ OMVs as a percentage of the total proteins identified by mass spectrometry listed in Table S2. (C) Amounts of protein and relative lipid contents in OMVs purified from different Y. pestis strains (Y. pestis KIM6+, χ10015 [ΔlpxP::P_lpxLlpxL] and χ10027 [ΔlpxP::P_lpxLlpxL ΔlacZ::P_lpplpxE]). All the values were normalized according to the total bacterial number (×10¹¹ CFU). (D) Whole-cell lysates or OMVs isolated from Y. pestis KIM6+, χ10015, and χ10027 were examined for the presence of the outer membrane proteins Psn, OmpA, Pla, and Caf1 (F1) by immunoblotting. (E) Whole-protein profiles of OMVs from different Y. pestis strains as shown in an SDS-PAGE gel. Rabbit polyclonal Psn and Pla antibodies (lab stock), rabbit polyclonal OmpA antibody (LS‑C369146, LSBio), and mouse monoclonal F1 antibody (YPF19, Santa Cruz Biotechnology). The results are representative of three experiments. Statistical significance: ns, no significance; ****, P < 0.0001.

A balanced-lethal system for oversynthesizing LcrV antigen in Y. pestis.

The three above-described strains harboring the virulence plasmid pCD1 are select agents and must be studied in a biosafety level 3 (BSL3) lab. Growing large cultures of these bacteria in BSL3 for OMV isolation is inconvenient and prohibited. A suite of virulence effectors termed Yops (YopE, YopJ, YopH, YopM and YopT) are encoded on the virulence plasmid pCD1 (∼70 kb) and act to suppress innate immunity to favor Y. pestis infection upon translocation into host mammalian cells by the type III secretion system (T3SS) (23). Several studies have illustrated that OMVs from different bacterial pathogens can package multiple virulence factors, including virulence effectors of the T3SS and toxins, and deliver them into host cells, causing immune suppression and cytotoxicity (24, 25). As a vaccine, OMVs derived from pCD1+ Y. pestis that package Yops may result in potential immune suppression. To avoid these concerns, we used pCD1-deficient Y. pestis strains to produce OMVs in a BSL2 lab. However, OMVs from pCD1-deficient Y. pestis lack the indispensable protective antigen LcrV, which is encoded on the pCD1 plasmid (26). To overcome this deficiency, we constructed a balanced-lethal system (27) to introduce an asd mutation into each Y. pestis strain to generate YPS1, YPS2, and YPS3, respectively (Table 1), which can adopt an Asd+ plasmid for the oversynthesis of LcrV.

Two Asd+ plasmids were constructed: pSMV12 (V), containing the native lcrV gene of Y. pestis, and pSMV13 (Bla-V), containing the N-terminal β-lactamase signal sequence (bla SS) fused with Y. pestis lcrV to facilitate LcrV secretion into the periplasm by the type II secretion system (T2SS) (Fig. 3A and Table 1). Subsequently, both plasmids were introduced individually into the YPS1, YPS2, and YPS3 strains to compare the amounts of LcrV in the bacterial cell fractions, including the whole-cell lysate, cytoplasm, periplasm, and OMV fractions. The results showed that all mutant strains harboring the Bla-V plasmid secreted more LcrV into the periplasmic fractions than those harboring the V plasmid, indicating that the β-lactamase secretion signal peptide can facilitate LcrV secretion into the periplasmic space in Y. pestis mutants (Fig. 3B). The amounts of LcrV in the cytoplasm and whole-cell lysates of each strain harboring the V or Bla-V plasmid were comparable (Fig. 3B). Moreover, OMVs isolated from all strains harboring Bla-V enclosed larger amounts of LcrV than those harboring the V plasmid (Fig. 3C). Therefore, we chose the pSMV13 (Bla-V) plasmid for the following studies.

Fig 3.

Subcellular location analyses of the oversynthesis of LcrV antigen in Y. pestis mutants. (A) Maps of the Asd+ plasmids pSMV12 (harboring the native lcrV gene of Y. pestis) and pSMV13 (harboring the N-terminal β-lactamase signal sequence [bla SS] and lcrV fusion to facilitate LcrV secretion by the T2SS). (B) Comparison of LcrV amounts in different cell fractions. The total cell lysates and subcellular fractions, including the cytoplasmic and periplasmic fractions, were prepared from YPS1, YPS2, and YPS3 strains individually harboring pYA3342 (an empty plasmid), pSMV12, or pSMV13 (Table 1). The cells were grown in HIB at 28°C for 14 h and then incubated at 37°C for 4 h, as described in the supplemental methods. Fractions with 25-μl volumes from cultures grown to an OD₆₀₀ of 0.8 were evaluated by immunoblotting with LcrV-specific polyclonal rabbit antibody. GroEL was used as a cytoplasmic marker for fractionation. (C) Comparison of the LcrV amounts in the OMV fractions isolated from YPS1, YPS2, and YPS3 strains individually harboring pSMV12 or pSMV13 (Table 1). OMVs were isolated from bacterial cultures as described in the Materials and Methods. Five-microliter volumes of OMVs normalized according to the bacterial numbers were evaluated by immunoblotting with LcrV-specific polyclonal rabbit antibody. (D) Amounts of protein and relative lipid contents in OMVs purified from YPS1, YPS2, and YPS3 strains individually harboring pSMV13. All the values were normalized according to the total bacterial numbers (×10¹¹ CFU). (E) Comparison of Psn, LcrV, and F1 synthesis in the OMV fractions isolated from YPS1, YPS2, and YPS3 strains individually harboring pSMV13. (F) Whole-protein profiles of OMVs from YPS1, YPS2, and YPS3 strains individually harboring pSMV13 were examined by SDS-PAGE gels. The results are representative of three experiments. Statistical significance: ns, no significance; ****, P < 0.0001.

Elimination of potential virulence factors from Y. pestis OMVs.

Y. pestis harbors two additional plasmids, pPCP1 (9.6 kb), encoding the plasminogen activator (Pla) (28), and pMT1 (102 kb), encoding murine toxin (Ymt) and the protective antigen F1 (29). Pla is necessary for Y. pestis dissemination and the inhibition of immune cell recruitment (30) and induces fibrinolysis (31). Murine toxin, which is encoded by ymt, is highly toxic in mice and rats but is less toxic in larger animals (32). Pla and Ymt are clearly present in Y. pestis OMVs (Fig. 2D and Table S1). To eliminate the potential adverse effects of Pla and Ymt on hosts, we cured the pPCP1 plasmid and deleted the ymt gene from strains YPS1, YPS2, and YPS3 individually by using sequential steps to generate mutant strains designated YPS7, YPS8, and YPS9, respectively (Table 1 and Fig. S2). Then, we individually introduced the Bla-V plasmid into the YPS7, YPS8, and YPS9 strains and compared the OMV production levels of these mutant strains. The results showed that YPS9(Bla-V) with 1-phosphorylated lipid A still generated higher numbers of OMVs (Fig. 3D) and enclosed substantially higher levels of LcrV and Psn antigens than YPS7(Bla-V) or YPS8(Bla-V). The amounts of F1 antigen in OMVs derived from each strain were comparable (Fig. 3E). Additionally, the total protein amounts in OMVs from YPS9(Bla-V) were clearly higher than those in OMVs from the other two strains (Fig. 3F).

We also compared the stimulation and cytotoxicity of OMVs from YPS7(Bla-V), YPS8(Bla-V), or YPS9(Bla-V) cultured with different cell lines in vitro and found that OMVs from YPS9(Bla-V) could activate the TLR4-mediated NF-κB signaling pathway but showed less stimulatory activity than OMVs from the other two strains (Fig. S3A). OMVs from YPS7(Bla-V) or YPS8(Bla-V) at a concentration of 25 μg/ml generated significantly larger amounts of tumor necrosis factor alpha (TNF-α) and cytotoxicity in RAW 264.7 cells than OMVs from YPS9(Bla-V) at the same concentration, but a low concentration (10 μg/ml) of all three types of OMVs produced decreased amounts of TNF-α without any difference and showed diminished cytotoxicity in cells (Fig. S3B). Given the above results, we used the YPS9(Bla-V) strain to produce the greatest amounts of OMVs decorated with MPLA, which present low toxicity and enclose large amounts of protective antigens for vaccine evaluation.

Immunization with self-adjuvanting OMVs affords complete protection against Y. pestis challenge.

Groups of mice (n = 10) were intramuscularly (i.m.) immunized with OMVs purified from YPS9(Bla-V), LcrV/alhydrogel, rF1-V (BEI)/alhydrogel, or PBS/alhydrogel (sham) and boosted at 21 days after the prime immunization (Fig. 4A). None of the vaccinations affected body weight increases in mice (Fig. S4A) or caused observable health issues. The measurement of serum antibody responses showed that the total anti-LcrV IgG titers were primed at higher levels in LcrV or rF1-V-immunized mice than in OMV-immunized mice at week 2 postvaccination, but were boosted to the same levels in both immunized groups by week 4 postvaccination (Fig. 4B). The anti-YPL IgG titers (Y. pestis whole-cell lysate) in OMV-immunized animals at weeks 2 and 4 postvaccination were substantially higher than those in LcrV- or rF1-V-immunized animals (Fig. 4B). Higher anti-F1 IgG titers were primed at week 2 postvaccination and boosted at week 4 postvaccination in the OMV-immunized mice than in the rF1-V-immunized mice (Fig. 4B). Additionally, significant anti-LcrV, YPL, and F1 IgM titers were primed in OMV-immunized mice at week 2 postvaccination, which were substantially increased at week 4 postvaccination after the booster. The LcrV, YPL, and F1 IgM titers in the LcrV- or rF1-V-immunized mice were not substantially different, but they were significantly lower than those in the OMV-immunized mice at weeks 2 and 4 postvaccination (Fig. S4B).

Fig 4.

Total IgG titers in LcrV-, rF1-V-, or OMV-immunized mice and the survival of mice challenged by virulent Y. pestis. (A) Immunization scheme used for the mouse study. (B) LcrV, YPL (Y. pestis whole-cell lysate), and F1-specific total IgG titers. (C) Immunized and PBS (sham) groups of Swiss-Webster mice (10 mice per group, equal numbers of males and females) were subcutaneously challenged with 8 × 10⁵ CFU of Y. pestis KIM6+(pCD1Ap) (8 × 10⁴ LD₅₀). (D) Immunized and PBS (sham) groups of Swiss-Webster mice (10 mice per group, equal numbers of males and females) were intranasally challenged with a low dose (L: 5 × 10³ CFU, 50 LD₅₀) or a high dose (H: 5 × 10⁴ CFU, 500 LD₅₀) of Y. pestis KIM6+(pCD1Ap). Statistical significance: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

On day 42 after the initial vaccination, the mice were challenged by the subcutaneous (s.c.) or intranasal (i.n.) route to mimic bubonic or pneumonic plague, respectively. All OMV-immunized mice survived s.c. challenge with 8 × 10⁵ CFU (8 × 10⁴ 50% lethal dose [LD₅₀]) of Y. pestis KIM6+(pCD1Ap), while 80% of the LcrV-immunized or rF1-V-immunized mice survived the same challenge (Fig. 4C). The OMV vaccination afforded 100% and 50% protection in mice against i.n. challenge with a median dose of 5 × 10³ CFU (50 LD₅₀) and a high dose of 5 × 10⁴ CFU (500 LD₅₀) of Y. pestis KIM6+(pCD1Ap), respectively. The LcrV or rF1-V vaccination conferred the same decreased protection (Fig. 4D), and none of the sham mice survived both challenges (Fig. 4C and D).

Vaccination with self-adjuvanting OMVs elicited Th1/Th2-balanced immune responses.

In mice, IgG1 is associated with a Th2-like response, while a Th1 response is associated with the production of IgG2a, IgG2b, and IgG3 antibodies (33). Therefore, we analyzed the IgG subtypes produced in response to each antigen to distinguish between Th1/Th2 immune responses in immunized mice. The anti-LcrV IgG1 titers were high and showed similar profiles as the anti-LcrV total IgG titers in LcrV-, rF1-V-, and OMV-immunized mice (Fig. 4B and Fig. 5A). Figure 5A showed that LcrV-, rF1-V-, and OMV-immunized animals were primed with moderate titers of anti-LcrV IgG2a and IgG2b and did not show a substantial difference at week 2 postvaccination. After the booster, the titers of anti-LcrV IgG2a and IgG2b were slightly increased in the LcrV- or rF1-V-immunized groups at week 4 after initial vaccination, but rapidly increased in the OMV-immunized groups, and they were significantly higher than those in the LcrV-immunized groups. At week 4 postvaccination, the ratios of anti-LcrV IgG1/IgG2a and IgG1/IgG2b were more than one (∼1.5) in LcrV- or rF1-V-immunized mice, respectively, while the ratios of anti-LcrV IgG1/IgG2a and IgG1/IgG2b were 1.0 and 0.99 in OMV-immunized mice, respectively (Fig. 5A). As shown in Fig. 5B, the OMV-immunized mice were primed with high titers of anti-YPL IgG1, IgG2a, and IgG2b at week 2 postvaccination, which were substantially increased at week 4 postvaccination after the booster. However, the titers of anti-YPL IgG1, IgG2a, and IgG2b remained steady in the LcrV- or rF1-V-immunized groups at weeks 2 and 4 postvaccination. At week 4 postvaccination, the ratios of anti-YPL IgG1/IgG2a and IgG1/IgG2b were higher than one (∼1.6) in the LcrV- and rF1-V-immunized groups, respectively, whereas the ratios of anti-YPL IgG1/IgG2a and IgG1/IgG2b were 1.0 and 1.1 in the OMV-immunized groups, respectively (Fig. 5B). Additionally, the OMV-immunized mice were primed with high titers of anti-F1 IgG1, IgG2a, and IgG2b at week 2 postvaccination, which were substantially increased at week 4 postvaccination after the booster (Fig. 5C). The ratios of anti-F1 IgG1/IgG2a and IgG1/IgG2b were 1.2 and 1.0 at week 4 after initial vaccination, respectively. In the rF1-V-immunized group, the ratios of anti-F1 IgG1/IgG2a and IgG1/IgG2b were around 2.3 at weeks 2 and 4 after initial vaccination, respectively. Collectively, the OMV-immunized mice generated more broad antibody responses against multiple antigens and more balanced Th1/Th2 responses than the LcrV-immunized mice. Our data showed that protection and antibody responses were no different between rF1-V-immunized and LcrV-immunized groups, so we excluded the rF1-V-immunized group for further studies.

Fig 5.

Antibody isotypes in immunized mouse sera collected at days 14 and 28 after prime and booster immunization. (A) Anti-LcrV IgG1, IgG2a, and IgG2b. (B) Anti-YPL IgG1, IgG2a, and IgG2b. (C) Anti-F1 IgG1, IgG2a, and IgG2b. The statistical significance among the groups at day 14 and day 28 was analyzed by two-way multivariate ANOVA with a Tukey post hoc test: *, P < 0.05; **, P < 0.01; ***, P < 0.001, ****, P < 0.0001.

Vaccination with self-adjuvanting OMVs induced potent cellular immune responses.

After 72 h of in vitro induction with LcrV or phosphate-buffered saline (PBS) (as control), lung lymphocytes from OMV-immunized mice showed substantial increases in both the CD4 and CD8 T cell populations (Fig. 6A and B). Lung CD4⁺ T cells from OMV-immunized mice displayed significantly higher production of gamma interferon (IFN-γ), interleukin-2 (IL-2), and TNF-α than those from LcrV-immunized and sham mice (Fig. 6A). Lung CD8⁺ T cells from OMV-immunized mice stimulated with LcrV protein showed higher production of TNF-α than those from LcrV-immunized and sham mice, but did not show increased production of IFN-γ, IL-2, and IL-17 in comparison to that in sham- or LcrV-immunized animals (Fig. 6B). Both lung CD4⁺ and CD8⁺ T cells from LcrV-immunized mice demonstrated higher production of IL-4 than those from OMV-immunized mice after in vitro stimulation with LcrV (Fig. 6A and B).

Fig 6.

Analysis of antigen-specific T cells obtained from lungs and associated cytokine responses. On day 42 after the initial immunization, lymphocytes were aseptically isolated from mice and stimulated in vitro with 20 μg/ml purified recombinant LcrV protein for 72 h to detect specific CD4⁺ and CD8⁺ T cells encoding IFN-γ, IL-2, IL-4, IL-17, and TNF-α. Sham mouse lung cells were considered controls. (A) CD4⁺ T cell numbers in lungs and CD4⁺ IFN-γ⁺, CD4⁺ IL-2⁺, CD4⁺ IL-4⁺, CD4⁺ IL-17⁺, and CD4⁺ TNF-α⁺ cell numbers. (B) CD8⁺ T cell numbers in lungs and CD8⁺ IFN-γ⁺, CD8⁺ IL-2⁺, CD8⁺ IL-4⁺, CD8⁺ IL-17⁺, and CD8⁺ TNF-α⁺ cell numbers. Each symbol represents a data point obtained from an individual mouse, with horizontal mean value bars ± SD. Statistical significance: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Similarly, splenocytes from both OMV- and LcrV-immunized mice also showed increased production of CD4⁺ and CD8⁺ T cells in comparison to those from sham mice after in vitro stimulation with LcrV (Fig. S5). Spleen CD4⁺ T cells from OMV-immunized mice demonstrated significantly higher production of IL-2 and IL-17 than those from LcrV-immunized and sham mice. Significantly higher production of IFN-γ and IL-4 was observed in spleen CD4⁺ T cells from both OMV- and LcrV-immunized mice in comparison to those from sham mice (Fig. S5A). Spleen CD8⁺ T cells from OMV-immunized mice showed higher production of TNF-α than those from LcrV-immunized and sham mice (Fig. S5B). However, both spleen CD4⁺ and CD8⁺ T cells from LcrV-immunized mice also produced higher levels of IL-4 than those from OMV-immunized mice (Fig. S5). These results suggested that OMV vaccination elicited more potent LcrV-specific cellular immune responses in mice than LcrV vaccination.

In vivo responses after Y. pestis pulmonary challenge.

Furthermore, we specifically monitored bacterial burdens in different tissues, variations of different cells in lung and bronchoalveolar lavage fluid (BALF), and cytokine production in BALF on day 2 after pulmonary Y. pestis challenge to determine the correlation between animal survival and host responses. On day 2 postinfection, the sham mice were found to have strikingly increased Y. pestis titers (mean 7.8 log₁₀ CFU/g tissue) in lung and moderate bacterial titers in liver (mean 3.8 log₁₀ CFU/g tissue) and spleen (mean 2.0 log₁₀ CFU/g tissue). In the LcrV-immunized mice, the bacterial titers reached moderate levels (mean 3.6 log₁₀ CFU/g tissue) in the lungs, but the bacteria could not disseminate into the liver and spleen (Fig. 7A). No Y. pestis titers were observed in the lungs, livers and spleens of OMV-immunized mice (Fig. 7A).

Fig 7.

In vivo responses after Y. pestis pulmonary challenge. Sham-, LcrV-, or OMV-immunized Swiss-Webster mice (3 mice per group) were infected i.n. with 3 × 10³ CFU of Y. pestis KIM6+(pCD1Ap). The groups of immunized mice infected with PBS served as negative controls. On day 2 postchallenge, different tissues (lungs, livers, and spleens) and bronchoalveolar lavage fluid (BALF) were collected from the euthanized mice. (A) Bacterial burden was evaluated in the lungs, livers, and spleens. (B) CD4⁺CD44⁺ cell numbers in the lungs of mice with or without infection were analyzed. (C) Alveolar macrophages in the BALF of mice with or without infection. (D) Neutrophils in the BALF of mice with or without infection. (E) Alveolar macrophages in the lungs of mice with or without infection. (F) Neutrophils in the lungs of mice with or without infection. Statistical significance: ns, no significance; *, P < 0.05; **, P < 0.01; ***, P < 0.001, ****, P < 0.0001.

Upon comparison of immunized mice with or without infection, significant increases in CD4⁺CD44⁺ cells were observed in the lungs of LcrV- or OMV-immunized mice after infection (Fig. 7B). Moreover, the number of CD4⁺CD44⁺ cells in the lungs of OMV-immunized mice was significantly higher than that in the lungs of sham or LcrV-immunized mice at day 2 postinfection (Fig. 7B). There were no substantial differences in CD4⁺CD44⁺ cell numbers in sham mice preinfection and postinfection (Fig. 7B). Slight decreases in CD8⁺CD44⁺ cells but no significant differences were observed in the lungs of sham, LcrV- and OMV-immunized mice preinfection and postinfection (Fig. S6A). In the BALF, the numbers of alveolar macrophages (AMϕ) in sham mice with or without Y. pestis infection were comparable (Fig. 7C), but the numbers of neutrophils were dramatically elevated in sham mice after Y. pestis infection in comparison to those in noninfected mice (Fig. 7D). In contrast, the numbers of AMϕ were significantly increased in LcrV- or OMV- immunized mice on day 2 postinfection in comparison to those in immunized mice without infection (Fig. 7C), while the numbers of neutrophils did not show substantial differences in LcrV- or OMV-immunized mice preinfection and postinfection (Fig. 7D). In lung tissues, no obvious alterations in AMϕ numbers were observed in mice with or without Y. pestis infection (Fig. 7E), but the numbers of lung neutrophils were dramatically increased in sham mice compared with LcrV- or OMV-immunized mice on day 2 postinfection (Fig. 7F). Slight increases in monocytes but no substantial differences were observed in sham, LcrV- or OMV-immunized mice preinfection and postinfection (Fig. S6B and C). Additionally, dramatically increased levels of proinflammatory cytokines (IL-1α, IL-1β, IL-6, IL-17, and IFN-γ) and chemokines (G-CSF, KC, and MIP-1α) associated with the recruitment of neutrophils were secreted into the BALF of sham mice on day 2 postinfection in comparison to the levels in sham mice without infection. However, there were no differences in these cytokines and chemokines in the BALF from LcrV- or OMV-immunized mice between preinfection and postinfection (Fig. 8). These data showed that LcrV or OMV vaccination rapidly activated CD4⁺ T memory cells, increased the number of AMϕ in BALF, and reduced neutrophil recruitment after Y. pestis pulmonary infection, which effectively controlled Y. pestis dissemination and cytokine storms that typically lead to the rapid death of mice.

Fig 8.

Comparison of cytokine and chemokine levels in the BALF from mice with and without pulmonary Y. pestis challenge. Sham-, LcrV-, or OMV-immunized Swiss-Webster mice (3 mice per group) were infected i.n. with 3 × 10³ CFU of Y. pestis KIM6+(pCD1Ap). The groups of immunized mice infected with PBS served as negative controls. On day 2 postchallenge, BALF from each euthanized mouse was collected at 48 h postinfection, filtered through a 0.22-μm syringe filter and checked for sterility before transfer to the BSL2 lab for analysis. A Bio-Plex Pro mouse cytokine assay kit (Bio-Plex) was used to detect the cytokines and chemokines, such as IL-1α, IL-1β, IL-6, IL-17, IFN-γ, G-CSF, KC, and MIP-1α, in the BALF collected from mice according to the manufacturer’s instructions. The statistical significance among the groups was analyzed by two-way multivariate ANOVA with a Tukey post hoc test. ****, P < 0.0001. Abbreviations: IFN, interferon; G-CSF, granulocyte colony-stimulating factor; KC, keratinocyte chemoattractant; MIP-1-α, macrophage inflammatory protein 1-alpha.

DISCUSSION

Generally, the removal of phosphate groups decreases the overall negative charge of a bacterium, thus reducing the electrostatic interactions of the phosphates in lipid A with cationic antimicrobial peptides. Decreased negative charge typically also decreases susceptibility to polymyxin B, which is a cationic antimicrobial peptide that binds negatively charged phosphate groups in lipid A units in LPS on the bacterial membrane and inserts its hydrophobic tail into the outer membranes of bacteria, causing membrane damage and bacterial killing (34). The removal of 1-phosphate from the conventional biphosphorylated lipid A in E. coli (35) and Salmonella (36) decreased their susceptibility to polymyxin B, but the opposite was observed in Y. pestis (20). The possible reasons for this are as follows: (i) Y. pestis masks the phosphate groups with 4-amino-4-deoxy-l-arabinose (l-Ara4N) to reduce the negative charge at its surface using different regulatory strategies than those used by Salmonella (37); (ii) Y. pestis naturally lacks O-antigen (38) because bacteria with the full O-antigen are more resistant to polymyxin B than O-antigen isogenic mutants (39, 40); (iii) lipid A 1-dephosphorylation in Y. pestis may cause cation displacement in the outer membrane (OM), resulting in reduced OM integrity and increased OM permeability, thereby changing Y. pestis morphology by increasing the OM curvature (Fig. 1C) and OMV formation (Fig. 2C). The replacement of acylated fatty acid chains (palmitoleate, C16) in Y. pestis KIM6+ with laurate (C12) in the χ10015 (ΔlpxP::P_lpxLlpxL) strain did not substantially affect bacterial morphology or OMV production (Fig. 1 and 2C). However, lipid A alteration via the constitutive expression of pagL in Salmonella enterica serovar Typhimurium to remove the β-hydroxymyristoyl group at position 3 in lipid A significantly increased vesiculation and induced OMV production (41). Thus, alterations in lipid A acylation at different positions in Kdo2 lipid IV_A may produce different outcomes during bacterial membrane vesiculation. Further investigations are needed to dissect this process.

Vaccination with OMVs derived from a wild-type Y. pestis strain containing very small amounts of LcrV provided very limited protection against plague (unpublished data). We therefore adapted an Asd+-based balanced-lethal Salmonella system (42, 43) into our Y. pestis system that was successful in overcoming this limitation by oversynthesizing LcrV (Fig. 3B). Our data demonstrate that the localization of the LcrV protein secreted into the Y. pestis periplasm by the T2SS led to the enclosure of large amounts of LcrV by OMVs (Fig. 3C). This strategy would thus be applicable to the delivery of antigens from other pathogens.

In addition to the production of high titers of IgGs against LcrV, YPL, and F1 antigen (Fig. 4B) that can synergize with cellular immune responses to defend against Y. pestis infection (44, 45), vaccination with OMVs also elicited significantly increased titers of IgM against LcrV, YPL, and F1 in mice compared to vaccination with LcrV (Fig. S4B). IgM has been demonstrated to play a protective role in extracellular and intracellular bacterial infections (46, 47) and to facilitate the removal of foreign pathogens due to its efficient agglutination (48). In mice, B-1a cells spontaneously maintain steady-state levels of natural IgM, while B-1b cells secrete IgM in response to pathogen encounters or heterologous antigens (49, 50). Recently, Levy et al. showed that the capsular F1 antigen of Y. pestis was recognized by B1b cells and generated high levels of anti-F1 IgM, which played a significant role in responses to plague challenge (51). We speculate that high levels of IgM induced by vaccination with self-adjuvanting OMVs containing capsular F1, LcrV, or other antigens from Y. pestis may produce better protection against plague than vaccination with LcrV antigen. Further investigations are needed to fully understand the role of IgM secreted from B1b cells in preventing Y. pestis infection.

Previous studies showed that recombinant, bacterially derived OMVs induced a more balanced Th1/Th2 response (52, 53). Both LcrV and OMV vaccination elicited the production of significant levels of IgG against LcrV and YPL in mice (Fig. 4B), but OMV vaccination induced a more balanced Th1/Th2 immune response than LcrV vaccination (Fig. 5A and B). Consistent with the antibody responses, both lung and spleen CD4⁺ T cells from OMV-immunized mice produced higher levels of Th1 cytokines (IFN-γ, IL-2, IL-17, or TNF-α) and significantly smaller amounts of Th2 cytokines (IL-4) than those from LcrV-immunized mice after LcrV stimulation in vitro (Fig. 6A and Fig. S4A). Studies have shown that protection against plague is known to require humoral immunity and cell-mediated immunity induced by IFN-γ and TNF-α (54, 55). The induction of potent Th1 cell responses by self-adjuvanting OMV vaccination might be one of the primary reasons it offers better protection against lethal infection by Y. pestis than LcrV vaccination (Fig. 4C and D). However, Fig. 4D showed the i.m. immunization of OMVs conferred 50% survival against i.n. challenge with a high dose of 5 × 10⁴ CFU (500 LD₅₀) of Y. pestis. One possible reason is that i.m. immunization with OMVs may not induce the potent mucosal immunity in the lung that is required to overwhelm a high dose of Y. pestis. The current paradigm is that parenteral vaccination rarely induces immune responses in mucosal tissues (56), although the concept is not universally applicable. One study has shown that i.n. immunization with acellular pertussis vaccine induced potent mucosal immunity in mice, thereby preventing Bordetella pertussis infection (57). We will thus evaluate the protective immunity of i.n. immunization with OMVs in the next studies.

The disease progression of primary pneumonic plague in several animal models is biphasic and consists of a preinflammatory and a proinflammatory phase (58, 59). The early preinflammatory phase of the disease (initial 36 h postinfection) is characterized by rapid Y. pestis replication in the lungs of mice but an absence of measurable host immune responses or obvious disease symptoms (59, 60). In contrast, the proinflammatory phase (48 h postinfection) is characterized by continuous increases in bacterial titers and dramatic increases in the levels of cytokines (IL-1α, IL-1β, IL-6, IFN-γ, and IL-17) and chemokines (KC, G-CSF, and MIP-1α) accompanied by massive neutrophil influx in the lungs and alveolar spaces, resulting in acute lethal pneumonia (59, 60). Our data showed that the responses in the sham group mice on day 2 postinfection (Fig. 7) were consistent with previous observations (61). In contrast, both LcrV and OMV vaccination subverted the progression of Y. pestis pulmonary infection in mice, resulting in low or absent bacterial titers in the lungs, spleens, and livers (Fig. 7A) and significant increases in CD4⁺CD44⁺ memory T cells (Fig. 7B) and AMϕ in BALF (Fig. 7C), which were not observed in sham mice. Our results suggest that the presence of memory CD4⁺ T cells, along with high titers of specific anti-Y. pestis antibodies, might activate AMϕ and enhance their phagocytosis, leading to the rapid elimination of inhaled Y. pestis.

In Y. pestis pulmonary infection, the massive recruitment of mature and immature neutrophils in response to an increasing bacterial burden leads to highly necrotic, lethal pneumonia (62). This phenomenon occurred in sham mice on day 2 postinfection and was characterized by dramatic increases in neutrophils in the BALF and lung (Fig. 7D and F) and large amounts of proinflammatory cytokines and chemokines in the BALF and sera (Fig. 8 and Fig. S5). However, the recruitment of neutrophils (Fig. 7D and F) and the production of proinflammatory cytokines and chemokines (Fig. 8 and Fig. S5) in both OMV- and LcrV-immunized mice were well controlled. Increasingly, evidence has shown that “trained immunity” mediated by innate immune cells primed by encounters with certain pathogens or molecular patterns associated with pathogens (PAMPs) could achieve broad protection (63, 64). We speculate that OMV or LcrV vaccination might endow macrophages, neutrophils, and other innate cells in the lung with high expression rates of activation markers that allow these cells to form an organized and protective inflammatory response to Y. pestis infection. Therefore, it is worthwhile to investigate whether the potent “trained immunity” induced by self-adjuvanting OMVs derived from Y. pestis engineered with an array of PAMPs plays an important protective role against Y. pestis infection.

Our studies showed that protective immunity elicited by self-adjuvanting OMVs derived from engineered Y. pestis was greater than that elicited by LcrV/alhydrogel, suggesting that OMVs could be utilized as antigen carriers for delivering antigens and adjuvants as part of a promising and effective novel plague vaccine.

MATERIALS AND METHODS

Bacterial strains, plasmids, culture conditions, and molecular operations.

All bacterial strains and plasmids used in this study are listed in Table 1. All bacterial cultures and molecular procedures are described in the supplemental methods.

OMV isolation.

OMVs were isolated from Y. pestis strains as previously described with minor modifications (41). Briefly, the strains were grown at 28°C in heart infusion broth (HIB) (Difco) for 14 h and then incubated at 37°C for 4 h. The bacterial cultures were supplemented with EDTA (pH 8.0) at 100 mM and kept on ice for 1 h. Then, the bacterial cells were pelleted by centrifugation at 10,000 × g at 4°C for 20 min. The culture supernatant was filtered using a 0.45-μm pore membrane (Millipore) to remove the residual bacterial cells and concentrated with a 100-kDa filter using a Vivaflow 200 system (Sartorius). The OMVs were harvested by ultracentrifugation (120,000 × g) for 2 h at 4°C. The vesicle pellet was washed and resuspended in 0.1× sterilized PBS (pH 7.4) and the ultracentrifugation step was repeated. The final vesicle pellet was resuspended in 0.1× sterilized PBS, filtered with a 0.22-μm pore membrane (Millipore) and stored at –80°C for subsequent experiments. The bacteria and OMV were viewed by transmission electron microscopy (see supplemental methods).

OMV analysis.

A Bradford assay was performed as described previously for quantifying the total protein abundance associated with OMVs (65). The relative lipid contents of the OMVs were determined via an FM4-64 fluorescence dye binding assay measured by a SpectraMax iD3 Multi-Mode microplate reader (Molecular Devices) as described in previous reports (66). The values of the protein amounts and lipid contents were normalized according to the total bacterial number (×10¹¹ CFU). The major outer membrane proteins present in the OMV preparations were detected by immunoblotting. Proteomic analysis of OMVs is included in the supplemental material.

Animal experiments.

All animal studies were conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals and approved by the Institutional Animal Care and Use Committee at Albany Medical College (IACUC protocol number 18-02004). Mice (6-week-old male and female Swiss Webster mice) were purchased from Charles River Laboratories (Wilmington, MA) and acclimated for 1 week after arrival. The groups of mice were intramuscularly (i.m.) immunized with 400 μg OMVs (roughly containing 10 μg LcrV and 5 μg F1) in 100 μl PBS buffer, 100 μl of a mixture containing 10 μg LcrV/alhydrogel (as a positive control), or 100 μl PBS/alhydrogel (as a negative control). Booster vaccinations were then administered 3 weeks after the initial vaccination. Blood was collected via submandibular veins every 2 weeks to harvest sera for antibody analysis. At 42 days after the initial vaccination, animals were challenged s.c. with Y. pestis KIM6+(pCD1Ap) in 100 μl PBS by the front neck injection to mimic bubonic plague. For mimicking pneumonic plague, animals were anesthetized with a 1:5 xylazine/ketamine mixture and were challenged i.n. with virulent Y. pestis in 40 μl PBS. The LD₅₀ values of Y. pestis KIM6+(pCD1Ap) administered by s.c. and i.n. challenge in mice were 10 CFU and 100 CFU, respectively (67). All infected animals were observed over a 15-day period. For the determination of the bacterial burden, the animals were euthanized with an overdose of sodium pentobarbital. Lungs, livers, and spleens were removed at the indicated times and homogenized in ice-cold PBS (pH 7.4) using a bullet blender (Bullet Blender Blue; NY, USA) at power 7 for 2 min. Serial dilutions of each organ homogenate were plated on HIB agar, and each count was confirmed with duplicate plates with a minimum of 2 dilutions to determine the titers of bacteria per gram of tissue. The experiments were performed twice and the data were combined for analysis.

Measurement of antibody responses and cytokines.

An enzyme-linked immunosorbent assay (ELISA) was used to assay antibody titers against LcrV, F1, or Y. pestis whole-cell lysates (YPL) in serum as described in our previous report (68). A mouse multiplex cytokine assay kit (Bio-Plex) was used according to the manufacturer’s instructions to detect the cytokines and chemokines in the BALF and sera collected from the mice.

Analysis of cellular immune responses.

Lungs and spleens were obtained aseptically from euthanized animals and dissociated with 70-μm strainers to obtain single cells. The red blood cell (RBC)-lysed individual cell populations (2 × 10⁶) were seeded in 12-well cell culture plates and stimulated in vitro for 72 h with 20 μg/ml rLcrV. Four hours before the collection of the cells, the culture medium in each well was supplemented with brefeldin-A and a monensin cocktail (1:1 ratio) to block Golgi-mediated cytokine secretion. For the flow cytometric analysis of the T-cell populations and their corresponding cytokines, the induced cells were harvested and resuspended in fluorescence-activated cell sorting (FACS) staining buffer containing CD16/32 antibodies (1:200) for 10 min on ice. The T-cell-specific markers were stained using anti-mouse antibodies as in Table S3 according to the manufacturer’s protocol. The samples were acquired on BD flow cytometers (LSRII) and were analyzed using FlowJo v.10.

Cells from the BALF and lungs of mice were resuspended in 30 μl of FACS staining solution containing Fc block (CD16/32) at a 1:100 dilution and incubated at room temperature for 15 min to block macrophage Fc receptors. The cell suspensions were then pelleted at 650 × g for 5 min at 4°C. The cells were resuspended and incubated for 30 min at 4°C with the fluorescently labeled antibodies (Table S3) in flow cytometry buffer (1% bovine serum albumin [BSA] in PBS) for the staining of cell-surface markers. The stained cells were analyzed based on fluorescence staining patterns to identify the alveolar macrophages (Siglec-F⁺ F4/80⁺ CD11b^mid/low+ CD11c^high+ Ly6G⁻), monocytes (CD11b^high+ CD11c^low+ Ly6G⁻), and neutrophils (CD45⁺ Ly6G⁺).

Statistical analysis.

Each experiment included a significant number (minimum of 3) of biological replicates, with 2 to 3 replicates performed in a synchronized fashion to establish reproducibility. The statistical analyses of the data among the groups were performed with one-way analysis of variance (ANOVA)/univariate or two-way ANOVA with Tukey post hoc tests. The log-rank (Mantel-Cox) test was used for the survival analysis. All data were analyzed using GraphPad PRISM 8.0 software. The data are represented as the mean ± standard deviation with ns, no significance; *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001.

Data and material availability.

All data needed to evaluate the conclusions in the paper are present in the paper and/or the supplemental material. Additional data related to this paper are available upon request to the authors.