For use in immunological profiling, the wild-type Y. pestis CO92 strain (65) was grown on 5% sheep blood agar (SBA) plates for approximately 48 h at 28-30°C. Liquid culture of Y. pestis CO92 used as an antigen in immune assays was prepared in heart infusion broth (HIB) medium supplemented with 0.2% xylose (HIBX) and grown at 28-30°C for 21 h followed by switch to 37°C for an additional 3 h to upregulate the presentation of potential antigens. The bacteria were then harvested via centrifugation (10 minutes at 13,000 x g) and inactivated by exposure to approximately 21 kGy of gamma-radiation. This antigen preparation was designated CO92 TS (temperature shifted) (56). Bacteriological media were obtained from Thermo Fisher-Remel (Rockville, MD).

NLPs were prepared as previously described (57–60). Briefly, lipids (DOPC and DGS-NTA-Ni) in chloroform were dried into a thin film using N₂ and resuspended in PBS with 60 mM sodium cholate. After adding E4.22k scaffold protein (murine sequence), the assembly mixture was dialyzed overnight against PBS. Assembly mixtures were then purified by size exclusion chromatography (SEC), and fractions corresponding to pure NLPs were pooled and concentrated. NLPs were subsequently filter-sterilized, formulated with 0.1M trehalose, aliquoted, and lyophilized. To assemble NLP-based vaccines, lyophilized NLPs were rehydrated and mixed with His-tagged recombinant F1 or V.

Recombinant F1 and V proteins were expressed in cell-free reactions overnight at 25°C using the DASbox Mini Bioreaction System (Eppendorf, Hamburg, Germany) with proprietary E. coli lysates and proprietary culture media. F1 was expressed as a TwinStrep-TEV-F1-His construct and purified using three methods sequentially: 1) affinity chromatography using StrepTrap column with Tris buffers containing EDTA, TEV cleavage to remove the TwinStrep tag, 2) cation exchange using HiTrap Capto SP ImpRes using Tris buffers and a 1 M sodium chloride gradient, and 3) affinity chromatography using a HisTrap Excel column with Tris buffers containing the reducing agent TCEP. The expressed V protein contained only a histidine tag and utilized a two-step method: 1) affinity chromatography using a HisTrap Excel column with Tris buffers containing TCEP and 2) size-exclusion chromatography on a 26/60 Sephacryl column using an isocratic elution in 1X PBS. For both proteins, desired fractions from the final purification step were pooled, concentrated, then quantified using the BCA assay. Purity of the final products was assessed by SDS-PAGE ( Supplementary Figure 1 ). Antigen endotoxin levels were assessed using limulus amebocyte lysate Endosafe-PTS cartridges from Charles River Laboratories (Wilmington, MA). Endotoxin levels for F1 and V were measured to be 13.7 EU/mg and 30.0 EU/mg, respectively. With antigen doses of 0.5 μg (see below), animals received 0.0068 EU/dose (F1), 0.015 EU/dose (V), or 0.022 EU/dose (F1 + V). These levels are significantly lower than cited thresholds of 0.1 EU/20 g mouse (www.fda.gov). All columns used for purification were purchased from Cytiva (Malborough, MA).

The Y. pestis F1 amino acid sequence (NCBI Accession WP_002216410.1) featured a C-terminal 6X His-tag and an R124N mutation to decrease cleavage during the self-assembly reaction. TEV cleavage is denoted by a period:

HMSAWSHPQFEKGGGSGGGSGGSAWSHPQFEKGSGENLYFQ.GMADLTASTTATATLVEPARITLTYKEGAPITIMDNGNIDTELLVGTLTLGGYKTGTTSTSVNFTDAAGDPMYLTFTSQDGNNHQFTTKVIGKDSRDFDISPKVNGENLVGDDVVLATGSQDFFVNSIGSKGGKLAAGKYTDAVTVTVSNQHHHHHH.

The Y. pestis LcrV amino acid sequence (NCBI Accession WP_002212981.1) featured a C-terminal 6X His-tag:

HMIRAYEQNPQHFIEDLEKVRVEQLTGHGSSVLEELVQLVKDKNIDISIKYDPRKDSEVFANRVITDDIELLKKILAYFLPEDAILKGGHYDNQLQNGIKRVKEFLESSPNTQWELRAFMAVMHFSLTADRIDDDILKVIVDSMNHHGDARSKLREELAELTAELKIYSVIQAEINKHLSSSGTINIHDKSINLMDKNLYGYTDEEIFKASAEYKILEKMPQTTIQVDGSEKKIVSIKDFLGSENKRTGALGNLKNSYSYNKDNNELSHFATTCSDKSRPLNDLVSQKTTQLSDITSRFNSAIEALNRFIQKYDSVMQRLLDDTSGKHHHHHH.

Conjugation of antigens to NLPs was verified by SEC using Cytiva Superose 6 Increase (5 x 150 mm) columns. SEC was used to verify that all His-tagged antigen was bound to the NLP at the chosen antigen:NLP ratios, with no unconjugated antigen observed. Dynamic Light Scattering was conducted in PBS using a Malvern Zetasizer Nano ZSP (Westborough, MA).

Recombinant F1V fusion protein was used as a positive control and to assess vaccine noninferiority. F1V was formulated under GMP regulations and stored at -80°C in single-use aliquots prior to use (66, 67).

Toll-like receptor 9 (TLR9) oligonucleotide (ODN) CpG ODN 2006 (CpG2006) was purchased from InvivoGen (San Diego, CA) and reconstituted in accordance with manufacturer recommendations. Alhydrogel (Alu) was sourced from InvivoGen.

Vaccines were formulated immediately prior to vaccination. For NLP formulations, lyophilized NLPs were rehydrated in water and mixed with F1 and/or V antigens (0.5 μg) at antigen-to-NLP ratios where no unconjugated antigens are observed. Formulations were then mixed with CpG and Alu such that the final formulations contained 10 μg of CpG and approximately 250 μg of Alu per dose.

Female BALB/c mice (n = 10 per experimental group) were obtained from Charles River (Frederick, MD) and were 7–9 weeks of age at time of vaccination. The vaccines described consisted of 0.5 μg of antigens (F1V, F1, or V), 250 μg of Alhydrogel and 5 μg of CpG. Mice were vaccinated twice subcutaneously, 21–28 days apart, and exposed four weeks later by the aerosol route to a lethal dose of Y. pestis CO92. Two independent studies were performed, with details shown in Figure 1 and Table 1 . In Experiments A and B, sera and spleens were collected from a cohort of mice (n = 4/group in A, n = 6/group in B) the day prior to challenge. In addition, in Experiment A lung, sera, and spleens were collected from a cohort of mice (n = 3/group) three days following challenge. The remaining mice (n = 9 or 10/group) were monitored for signs of infection for 21 days following challenge. Euthanasia of moribund animals was conducted in accordance with approved early endpoint intervention criteria.

The animal research was conducted under an Institutional Animal Care and Use Committee (IACUC) approved protocol in compliance with the Animal Welfare Act, Public Health Service Policy on Humane Care and Use of Laboratory Animals, and other federal statutes and regulations relating to animals and experiments involving animals. USAMRIID is accredited by the AAALAC International and adheres to the principles stated in The Guide for the Care and Use of Laboratory Animals (National Research Council, 2011). Mice were checked daily for food and water and at least daily for assessment of clinical impact of the Y. pestis infection. Whenever possible, euthanasia of moribund animals was conducted in accordance with approved early endpoint intervention criteria. Mice were evaluated daily after exposure to aerosolized Y. pestis; scores of 0–2 represented normal mice, scores of 3–7 indicated significant clinical manifestations and these mice warranted multiple clinical assessments per day, and final scores of 8 or greater indicated severe clinical manifestations and mice were euthanized immediately. When mice met pre-determined euthanasia criteria and they were not in the sampling cohorts, mice were euthanized by CO₂ exposure (flow rate 6–11 ft³/h) or by barbiturate overdose through intraperitoneal injection (approximately 0.15 mL for 20 g of body weight) of Euthasol^® euthanasia solution (or equivalent) and then death was confirmed by cervical dislocation.

Mice were exposed to aerosolized challenge doses of virulent Y. pestis CO92 strain. Bacterial preparation for aerosol challenge involved growing bacteria on tryptose blood agar (TBA) base slants for approximately 48 h at 28-30°C. Bacteria from TBA slants were then suspended in HIBX to an initial OD₆₂₀ of approximately 0.01 and incubated for approximately 24 h at 28–30°C. Then, the cultures were harvested by centrifugation (10 min at 13,000 x g) and suspended in HIB medium (no xylose) to the concentration yielding the number of LD₅₀ doses indicated in the text and figure legend. A solution of 10 mM potassium phosphate, pH 7.3–7.4 (KPhos) was used to dilute bacterial inocula. Challenge doses were determined by serial dilutions in KPhos buffer and plating on SBA.

Mice were transferred to wire mesh cages and were placed in a whole-body aerosol chamber within a class three biological safety cabinet located inside a BSL-3 laboratory. Mice were exposed to aerosols of Y. pestis CO92 created by a three-jet collison nebulizer. The LD₅₀ of Y. pestis CO92 in BALB/c mice is approximately 6.8x10⁴ inhaled CFU. Samples were collected from the all-glass impinger (AGI) vessel and analyzed by performing CFU calculations to determine the inhaled dose of Y. pestis (68, 69). The actual inhaled dose was calculated to be approximately 5.3x10⁵ CFU (approximately 8.0 LD₅₀s).

The tissues collected from necropsied mice, 3 days following challenge, included lung, spleen, and blood. For sample collections, mice were deeply anesthetized with approximately 0.3 mL/20 g of body weight with a mixture of ketamine (10 mg/mL)-acepromazine (1 mg/mL)-xylazine (2 mg/mL), underwent a terminal blood collection via the axillary vessels, and then were euthanized by cervical dislocation prior to organ harvesting. The organs were weighed, homogenized in 1 mL KPhos per sample with disposable PRECISION™ homogenizers (Covidien, Dublin, Ireland), and the CFU in the homogenate were quantified by plating on SBA plates. Undiluted homogenate and ten-fold dilutions in KPhos were plated in duplicate. The limit of detection (LOD) was approximately 100 CFU/mL blood or 5 CFU/organ. After CFU determinations, samples were radiation-inactivated, sterility checked and stored at -80°C for immunological analyses.

Sera from 27 days post vaccine boost and 3 days post-challenge blood collection were measured for Immunoglobulin G (total IgG along with IgG1 and IgG2a subclasses) antibody levels by semi-quantitative endpoint ELISA in 96-well Immulon 2HB plates (Thermo Fisher, Rochester, NY). Plates were incubated overnight with antigens at 4°C to coat wells. Coating antigen solutions were either pure proteins F1V, F1 or V at 2 µg/mL, or inactivated Y. pestis CO92 TS at 10 µg/mL in 0.1 M carbonate buffer, pH 9.5. Wells were washed five times with 1X PBS including 0.05% Tween 20. Two-fold dilutions of serum in PBS/0.05% Tween 20 were made in triplicate and incubated for 60 min at 37°C, then washed five times, blocked with 1% Casein in PBS (Pierce) for 30 min at 37°C, and signal detected as previously described (35). Detection antibodies were horseradish peroxidase-conjugated goat anti-IgG, -IgG1, or -IgG2a (Southern Biotechnology) at dilution of 1:5,000. After 30 min at 37°C with detection antibodies, plates were washed five times, incubated with 3,3’,5,5’-tetramethylbenzidine substrate (Pierce) for 20 min at 37°C, and the reaction stopped with 2 N sulfuric acid.

A Biotek ELx808 plate reader was used to quantify the amount of bound antibody by colorimetric measurement at 450 nm. Results are reported as the geometric mean (GM) and geometric standard error (GSE) of the reciprocal of the highest dilution giving a mean OD of at least 0.1 ± 1 SD at 450 nm (570 nm used as reference wavelength), then triplicates averaged. The limit of detection was 50, with values ≤ 50 considered negative. To assess the avidity of serum antibodies, we performed the ELISA as described above, in parallel conditions with or without an additional 15 min incubation in 6 M urea at room temperature, immediately after incubation with samples. We chose the lowest dilution that had an OD₄₅₀ below 1.5 and calculated the avidity index based on the values at this dilution, as (OD with urea/OD with wash buffer) x 100%.

Groups of mice were euthanized 27 days after the vaccine boost, and splenocytes were isolated using minimal disruption as described previously (54). Briefly, spleens were excised from mice (n = 4 [Experiment A] and n = 6 [Experiment B] mice per group) and disaggregated in RPMI 1640 medium (Gibco, Grand Island, NY). Red blood cells in the splenocyte preparation were lysed with Ammonium-Chloride-Potassium (ACK) Lysing Buffer (Gibco) the extract was then diluted with RPMI 1640 medium and cells pelleted by centrifugation at 335 x g for 10 min. Splenocytes were resuspended in CTL-Test medium (ImmunoSpot, Shaker Heights, OH) and counted using a TC20 Cell Counter (Bio-Rad, Hercules, CA), then all samples were diluted in RPMI complete medium to normalize cell concentrations. Complete medium was RPMI-1640 including 10% fetal bovine serum (HyClone, Logan, UT), 100 U/ml penicillin-streptomycin (Gibco), 1x MEM Non-essential Amino Acid Solution (Sigma, St. Louis, MO), 1 mM sodium pyruvate (Sigma), and 55 µM β-mercaptoethanol (Gibco). ELISpot kits from ImmunoSpot were used (Mouse IgG1/IgG2a Double-Color ELISPOT Assay) and manufacturer’s instructions followed, as described here in brief.

For B cell ELISpot, splenocytes were first stimulated for 6 days with 1x B-Poly S Polyclonal B cell Stimulator (ImmunoSpot) in complete medium. Stimulation took place at 37°C/5% CO₂, with 5x10⁶ cells/mL in 48-well tissue culture plates at a total volume of 1.6 mL for each sample. Plates provided with the kit were coated with F1V (10 µg/mL), V (10 µg/mL), or F1 (10 µg/mL) protein or positive control wells coated with Anti-Igκ and Anti-Igλ Capture Ab (ImmunoSpot). Antigen-treated plates were washed, splenocytes were added at 5x10⁵ cells/well and incubated for approximately 8 h, then washed, developed and spots detected as per manufacturer’s instructions. Spot forming units (SFU), indicative of antigen specific antibody secreting cells, were normalized to the 1,000 SFU based on the number of total antibody secreting B cells. All ELISpot assays were performed with each sample in two duplicate wells.

For cytokine determination in vaccinated mice pre-challenge, splenocytes were isolated and counted 27 days after the vaccine boost. Cells were incubated in RPMI complete medium with antigens F1V (25 µg/mL), F1 (25 µg/mL), or V (25 µg/mL) at 37°C/5% CO₂. Splenocytes stimulated with medium only were included as negative controls, while positive controls were stimulated with 100 ng/mL phorbol 12-myristate 13-acetate (PMA) and 500 ng/mL ionomycin. After approximately 48 h incubation, plates were centrifuged at 1,200 x g for 10 min, then supernatants were harvested for evaluation of cytokine expression, as described below.

For mice post-challenge, lungs and spleens were homogenized at 3 days post-infection as described in section 2.6; homogenates were frozen at -80°C and irradiated prior to quantification of cytokines and other proteins in the tissue. All samples were thawed from cryopreservation, then centrifuged at 10,000 x g for 10 min to minimize debris.

Supernatants and homogenates were measured for cytokine expression levels using the Cytokine & Chemokine 36-Plex Mouse ProcartaPlex panel (Thermo Fisher) and MagPix instrument per manufacturer’s instructions. Cytokine results that were above or below the highest or lowest values of the standard curve were imputed to the value of the highest or lowest standard. In addition, any cytokine whose standard curve had low R² value (< 0.95) or a bead count of < 35 was removed from analysis. Cytokine levels are expressed as the fold change relative to PBS and in pg/mL.

Splenocytes from mice 27 days post-vaccine boost were isolated and cryopreserved in freezing medium, then transferred to liquid nitrogen vapor storage. After thawing and counting, cells were labeled with viability dye (LIVE/DEAD Fixable Green or Violet, Thermo Fisher), washed with FACS buffer, incubated with 1:200 dilution of Mouse Fc Block (5534142, BD), then stained with fluorescently conjugated antibodies and fixed in 2% formaldehyde (Thermo Fisher, Rockford, IL). The surface markers used for defining subsets were CD3e (antibody clone 145-2C11), CD19 (clone 1D3), CD4 (clone GK1.5), CD8a (clone 53-6.7), CD62L (clone MEL-14), CD11c (clone N418), CD11b (clone M1/70), CD40 (clone 3/23), and CD80 (clone 16-10A1). FACS buffer was 1% bovine serum albumin (SH30574, HyClone) in PBS, and freezing medium was 10% DMSO (D2660, Sigma) and 90% fetal bovine serum (SH30396, HyClone). Samples were run on a FACSCanto II (BD) and analyzed in FlowJo v10.8.

The mouse survival rates at selected time points were compared by Fisher exact test and the time to death or euthanasia (TTD) were analyzed by Log-rank test for the comparison against PBS group. For ELISA, pairwise treatment groups were compared by negative binomial generalized linear mixed model. Negative binomial regression was used for the flow cytometry data analysis. Data (ELISpot and Luminex) were log10 transformed prior to analysis. For ELISpot and Luminex cytokine results, pairwise treatment groups were compared by linear mixed effects model. The multiplicity was adjusted by Tukey’s method. Analysis was implemented in SAS version 9.4 (SAS Institute Inc., Cary, NC).

NLPs were prepared through a well-established self-assembly reaction. Briefly, recombinantly expressed apolipoprotein scaffold (22 kDa N-terminal domain of ApoE) is solubilized in a surfactant solution with both bilayer-forming lipids (DOPC) and a lipid with a nitrilotriacetic acid (NTA) modified head group chelating nickel. As the surfactant is removed via dialysis, the components spontaneously self-assemble into NLPs. The inclusion of the nickel lipid creates functionalized NLPs that are capable of binding histidine-tagged proteins ( Figure 2A ). Conjugation of his-tagged proteins to the NLPs can give varied results in terms of ideal protein:NLP molar ratios, total protein binding capacity, and the size of the particles after conjugation (with some proteins tending to induce the formation of large aggregates); therefore, conjugation of the F1 and V antigens to the NLP was assessed separately using size exclusion chromatography (SEC) and dynamic light scattering (DLS). Screens were performed in the absence of the adjuvants used in the final vaccine formulations.

NLPs elute as a single peak with a retention time (t_r) of 3 min, providing a baseline for comparison of NLPs with protein conjugates. Larger species, such as protein:NLP conjugates, will elute earlier (i.e., closer to the column void volume at 2 min) and any smaller species, such as monomeric unconjugated protein, will elute after the NLP (~ 4 min). Initial conjugation screens were conducted at increasing antigen:NLP ratios, ranging from 0:1 to 16:1 (data not shown). SEC was used to identify conjugation ratios that showed complete antigen binding (e.g., no unconjugated antigen) while minimizing larger species (i.e., elution in the void volume). These screens identified that F1 and V ratios per NLP of 4:1 and 8:1 (respectively) demonstrated successful conjugation to the NLP ( Figures 2B, C ). For the F1:NLP ratio of 4:1, addition of the protein to the NLP causes a shift to shorter retention times, indicating the successful conjugation and formation of slightly larger F1:NLP particles, without any indication of free (or unconjugated) F1 protein ( Figure 2B ). Conjugation of V to the NLP at an 8:1 ratio was also successful, with all V protein conjugated to the NLPs. Some larger multimeric NLP species with V are observed (~2 min), suggesting some degree of interparticle multimerization ( Figure 2C ).

The formulations were further characterized by DLS to evaluate average size of the NLPs and the NLP conjugates ( Figure 2D ). The diameter of the particles, based on number mean, was analyzed for the NLP alone and the F1:NLP at a 4:1 ratio and V:NLP at an 8:1 ratio. The average size of the NLPs without conjugated protein was 15.1 ± 1.7 nm. The addition of F1 protein resulted in an increase in average particle size to 26.8 ± 1.7 nm. The addition of V resulted in a mix of larger NLP particles (~37 nm) and some multimeric species (> 85nm), with an average size of 53.4 ± 28.6 nm. These results mirror the trends seen by SEC, and together demonstrate that the NLPs can be readily formulated with the F1 and V antigens at tailored ratios for use in downstream vaccine studies.

To date, F1V admixed with the adjuvants Alu and CpG2006 remains the benchmark subunit vaccine for Y. pestis in our laboratory (35, 36, 39, 70). Unfortunately, the chimeric nature of the fusion protein may potentially induce immune responses to novel linear or conformational epitopes that are absent from native protein structures and hence diminish the protective efficacy of the vaccine. To lessen this possibility, we evaluated individual F1 and/or V antigens conjugated to the Ni-chelating NLP-based vaccine delivery platform and compared it to the F1V formulation. Mice were vaccinated with two doses of the vaccines, four weeks apart and the cytokine response was assessed in splenocytes approximately four weeks after the last vaccine dose, against F1V, V, and F1 proteins (Table 2 and Supplementary Table 1). As expected, the benchmark F1V vaccine induced a strong anti-F1V, V, and F1 cytokine response, based on the fold change differences relative to sham (PBS) vaccinated mice. NLP formulations, much like F1V vaccine, required addition of Alhydrogel (Alu) to induce a similar cytokine response. F1:V:NLP splenocytes had little cytokine response upon restimulation, but F1:V:NLP + Alu splenocytes produced a strong response, with many cytokines upregulated at similar or higher levels than the results from the benchmark F1V vaccine. For both vaccines, the pattern of which cytokines were most induced in F1V-restimulated splenocytes was very similar, with exceptions being IL-17A and Eotaxin which were significantly upregulated in the F1V group but not F1:V:NLP + Alu.

Results from restimulation with individual V and F1 proteins were especially encouraging ( Tables 2B, C ). Splenocytes from F1:V:NLP + Alu vaccinated mice secreted more than twice the IL-5, IL-13, IL-2, IFN-γ, IL-4, IL-3, GM-CSF, IL-22, IL-10, and LIF than splenocytes from F1V vaccinated mice when restimulated with F1 alone. Similar results were seen for restimulation with V alone, although less pronounced because splenocytes from F1V vaccinated mice already had a stronger response to V than F1 restimulation. These data suggest that the NLP formulation with V and F1 included as separate components is more effective than F1V fusion protein at presenting these two antigens to the immune system. As would be expected, the cytokine response of the NLP formulations was antigen-specific, with a very limited recall response induced by V restimulation of splenocytes from F1 vaccinated mice or vice versa.

A similar cytokine profile was recapitulated in mice that received prime and boost doses 21 days apart rather than 28 ( Table 3 , Supplementary Table 2 ). Again, mice receiving NLP vaccine formulations had a strong splenocyte cytokine response against the Y. pestis antigen(s) in the vaccine, with the strength of this response dependent on the presence of Alu. Several cytokines that were not significantly induced above the sham (PBS) group in Experiment A were seen here to be significantly induced in F1V or F1:V:NLP + Alu groups, including IL-27, GRO-α and G-CSF (induced in F1:V:NLP + Alu) and IL-12p70, IL-18, and LIF (upregulated in both). In addition, unlike Experiment A, Eotaxin did not have enhanced production in the F1V group, and IL-17A was significantly induced in both groups rather than only in F1V. However, the apparently increased sensitivity of this experiment went along with more false positives in which splenocytes from mice vaccinated with F1 alone were restimulated by V antigen, or vice versa.

It was notable and encouraging that both the benchmark F1V vaccine and F1:V:NLP, when formulated with Alhydrogel and CpG, produced highly similar recall responses in splenocytes. Any notable differences between the two vaccines were only seen in Experiment A or B but not both, with the exception that F1:V:NLP + Alu and V:NLP + Alu splenocytes produced at least twice as much IL-10 than F1V splenocytes in both experiments. IL-10 is an anti-inflammatory cytokine that plays a role in maintaining tissue homeostasis and limiting tissue damage. Otherwise, splenocytes from both groups produced especially high levels of Th2 cytokines (IL-5, IL-13, IL-4), and elevated levels of several others including the Th1 cytokine IFN-γ; the T cell stimulant IL-2; and proinflammatory cytokines like GM-CSF and IL-6 that are mostly produced by myeloid cells.

To assess innate and cell-mediated immune responses induced by vaccination, we used flow cytometry to phenotype the purified splenocytes approximately four weeks after the second vaccine dose. CD62L (L-Selectin) was used to distinguish naïve and activated T cell phenotypes. On T lymphocytes, CD62L is needed for naïve cells to home to lymphoid organs, and its loss is indicative of activation, cell division and acquiring a memory or effector phenotype (78–80) ( Figure 7A ). Upregulation of costimulatory molecules CD80 and CD40 was used to distinguish activated CD11b⁺ dendritic cells (DCs) associated with antigen presentation to CD4⁺ T cells (81, 82) ( Figure 8A ). Mice vaccinated with F1V + Alu showed no reduction in CD62L on T cells relative to the PBS group, while F1:V:NLP + Alu vaccinated mice had a significant decrease in CD62L on both CD8⁺ ( Figure 7B ) and CD4+ ( Figure 7C ) T cells. In addition, among the NLP vaccine groups, F1:NLP + Alu and V:NLP + Alu showed a loss of CD62L on T cells compared to the corresponding groups without Alu, as well as the loss of CD62L compared to F1V + Alu, indicating that the combination of NLP and Alu was highly immunogenic, although the loss of CD62L was only significant in one comparison ( Figures 7B, C ). Likewise, mice vaccinated with F1V + Alu showed no upregulation of CD80 and CD40 on DCs, while F1:V:NLP + Alu vaccinated mice had a significant increase compared to PBS and F1V + Alu ( Figure 8B ). In this analysis of bulk splenocytes it appeared that more CD4 T cells, CD8 T cells, and dendritic cells had an activated phenotype following F1:V:NLP + Alu vaccination; whereas in mice vaccinated with F1V + Alu, the relative proportions of naïve and activated T cells and DCs was similar to the PBS group.

The F1V vaccine is highly efficacious at protecting against encapsulated Y. pestis and is thus an appropriate comparator to the NLP formulations. Four weeks following the last dose of the vaccine, mice were exposed to aerosolized Y. pestis CO92 (approximately 8 LD₅₀) and monitored for 21 days. All sham (PBS) vaccinated mice succumbed to infection or were euthanized in accordance with early endpoint euthanasia criteria 4 days following challenge, and the V:NLP vaccine conferred no significant protection relative to the sham group ( Figure 9 ). The F1:NLP vaccine was able to confer 22% protection while all F1:V:NLP vaccinated mice succumbed to infection within nine days following challenge. While none of the vaccines without Alu resulted in significant protection, addition of Alu to the vaccine formulations drastically increased vaccine efficacy conferring 56% protection to both F1:NLP + Alu and V:NLP + Alu vaccinated mice (p = 0.029 compared to negative control mice). Mice vaccinated with either the F1V + Alu or the F1:V:NLP + Alu were fully protected for the duration of the study (p < 0.0001 compared to negative control mice) ( Figure 9 ).

Bacterial burden was evaluated in vaccinated mice, three days post challenge, to assess the impact of bacterial growth in the lungs and dissemination to the blood and spleen ( Figure 10 , Supplementary Table 6 ). Low bacterial burden was observed in mice vaccinated with F1V + Alu and F1:V:NLP + Alu. The bacterial burdens observed in the cohorts receiving these two most-protective vaccines were statistically indistinguishable from each other when comparing bacterial burden in lungs ( Figure 10A ) and spleens ( Figure 10B ) (p = 0.94 and p = 0.19, respectively for lungs or spleens), and there was no detectable bacteria in blood in any mouse in either cohort ( Figure 10C ). Surprisingly, the F1:NLP vaccinated mice had no detectable CFUs in the lungs; however, those same mice had a detectable bacterial burden in the spleen, indicative of bacterial dissemination ( Figure 10 ). The F1:NLP + Alu vaccinated mice had a moderate to high bacterial burden in the lungs, but no detectable CFUs in the spleen. This is in contrast with the two most protective vaccines, F1V + Alu and F1:V:NLP + Alu, which had low to no detectable CFUs in the lungs and spleen, respectively. The addition of Alu had varying impacts on the lung bacterial burden depending upon the vaccine antigens being evaluated. In the case of the F1:V:NLP, the addition of the Alu resulted in at least a 6-log reduction of bacteria in the lungs (p < 0.0001). As depicted in Figure 10B , the addition of Alu to the vaccine formulations decreased bacterial dissemination to the spleens in two experimental vaccines F1:NLP (p < 0.0001) and F1:V:NLP (p = 0.029). The addition of Alu did enhance the bacterial clearance associated with the V:NLP vaccine in the lungs (p = 0.017) but did not decrease dissemination to or clearance from the spleen in a statistically significant manner (p = 0.076). Mice vaccinated with V:NLP, F1:V:NLP, and V:NLP + Alu also had recoverable bacteria in the blood ( Figure 10C ), even though V:NLP + Alu conferred 56% protection ( Figure 9 ).

The ability of vaccines to mitigate cytokine storm is indicative of protective efficacy following infection. Accordingly, we analyzed the cytokine profile in the lungs and spleens of mice three days following challenge. Results are expressed as fold changes ( Table 4 ) and concentrations (pg/mL; Supplementary Table 7 ). Lower fold changes representing a more controlled cytokine response following challenge were consistent with reduced bacterial burden in those organs ( Figure 10 ). Specifically, vaccine formulations that resulted in lower levels of IL-6, IL-22, IFN-γ, IL-17A G-CSF, GM-CSF, TNF-α, CCL3, IL-1β, CXCL1, IL-1α, CCL4, IL-12p70, and IL-27 in the lung afforded protection from challenge. A high bacterial burden in the lungs of mice vaccinated with V:NLP, V:NLP + Alu and F1:V:NLP was associated with a similar uncontrolled cytokine response in the lungs, with similar cytokine levels compared to the sham vaccinated mice. The F1:NLP vaccinated mice which had no detectable CFUs in the lungs also had cytokine levels that were drastically lower relative to the control group, while the F1:NLP + Alu vaccinated mice which had a moderate amount of CFUs in the lungs also had a less pronounced reduction in cytokine levels. Similar to the cytokine response in the lungs, lower bacterial burden in the spleens of mice vaccinated with the F1 inclusive vaccines was associated with a decreased cytokine response relative to sham vaccinated mice ( Table 4 , Supplementary Table 7 ). The two most protective vaccines, F1V + Alu and F1:V:NLP + Alu, had the lowest induction of cytokines in both lungs and spleens, with F1:V:NLP + Alu vaccinated mice showing the lowest levels of cytokine induction.

To further correlate the protective efficacy of the vaccines to the immune response we evaluated total IgG antibody titers 3 days post challenge directed against irradiated whole cell CO92 TS. The two most protective vaccine formulations, F1V + Alu and F1:V:NLP + Alu, also had the highest total IgG antibody response ( Table 5 , Supplementary Table 8 ). These formulations both produced strong IgG1 but not IgG2a endpoint titers and were similar in terms of avidity ( Figure 4 ). Surprisingly the F1:NLP + Alu formulation also produced a robust total IgG and IgG1 response yet was only partially protective. Furthermore, while the V:NLP + Alu vaccinated group achieved the same level of partial protection (56% survival) as F1:NLP + Alu, it had negligible anti-CO92 TS titers. None of the vaccine formulations induced an IgG2a response against whole cell Y. pestis CO92 TS ( Table 5 ).

Consistent with previous data, inclusion of Alu was critical for the induction of a robust antibody response. For total IgG and IgG1, F1:NLP + Alu and F1:V:NLP + Alu had significantly higher anti-CO92 TS titers than F1:NLP and F1:V:NLP without Alu. The majority of the anti-CO92 TS titer can be attributed to antibodies against the F1 capsule; however, V:NLP + Alu did induce low but detectable anti-CO92 TS IgG1 titers (100-400), while any antibodies induced by V:NLP was undetectable ( Table 5 ).