Our recent work described successful vaccination strategies using the ΔyscN vaccine strains that protected female BALB/c mice from pneumonic plague initiated by aerosolized Y. pestis CO92 (encapsulated) and Y. pestis C12 (nonencapsulated) strains (34, 39). Here, we performed follow-on studies to compare the efficacy of these vaccines in male BALB/c mice. Groups of male BALB/c mice were given an initial (“prime”) vaccine injection with a combination of CO92 ΔyscN and C12 ΔyscN (“Combo ΔyscN”), followed three weeks later by a “boost” injection with either Combo ΔyscN or a subunit vaccine. The subunit vaccine was rF1-V, mixed with Alhydrogel, and CpG as an additional adjuvant ( Table 1 , experiment A). Sham immunization was an injection of 0.2 mL PBS for both prime and boost, as in all subsequent experiments. In descriptions of these vaccine regimens, in text and figure legends, “P” indicates the prime injection and “B” indicates the boost injection.

Four weeks post-boost, male mice were challenged with aerosolized Y. pestis C12 (calculated inhaled dose 1.1x10⁵ CFU). 10 mice in each group were followed for survival, and 6 were euthanized at 3 dpi (days post-infection) for measurement of bacterial burden. All male mice succumbed to infection by 6 dpi ( Figure 1A ). In previous studies using the same vaccines in female mice (34), survival was 90% with the P:Combo ΔyscN B:rF1-V strategy, and 0% in females given P:Combo ΔyscN B:Combo ΔyscN or sham (P: PBS B: PBS). The increased survival in females, despite a higher calculated exposure dose (7.9x10⁵ CFU) in the female experiment, led us to further characterize sex differences in this infection model.

We performed an LD₅₀ estimation using 7–9-week-old naïve male BALB/c mice. As shown in Supplementary Figure 1 , the LD₅₀ value for Y. pestis C12 was between 1.6x10⁴ and 3.2x10⁵ inhaled CFU. Given that our calculated LD₅₀ value for female BALB/c mice is approximately 7.7x10⁴ CFU (22), the overall virulence of Y. pestis C12 in male mice does not appear to differ from female mice and is likely not the source of the differences in vaccine efficacy.

In each group above, 6 mice were euthanized for sample collection at 3 dpi. We find 3 dpi to be a timepoint where Y. pestis replication in susceptible mice is near peak in both lungs and secondary tissues and sham-vaccinated mice are still available for a direct comparison of samples (34, 48, 49). Blood and organ homogenates were assessed by using serial dilutions plated on SBA to enumerate CFU. All animals but one had detectable bacteria in the lungs, with no significant difference between vaccinated groups and sham ( Figure 1B ). However, in the spleen, those given P:Combo ΔyscN B:rF1-V had significantly less bacteria detected ( Figure 1C ), indicating that systemic spread of Y. pestis was reduced despite no protection from lethal disease. Whereas 100% of the control mice had bacteria in their blood, only 50% of the mice receiving heterologous vaccines were bacteremic ( Figure 1D ). No significant reductions in bacterial burden were seen in male mice given two doses of Combo ΔyscN ( Figures 1B–D ).

Although the ΔyscN strains are considerably attenuated (in subcutaneous challenge they have an LD₅₀ over 10⁶-fold greater than parent CO92 or C12 strains) (39), the attenuation is entirely due to the deletion of the yscN gene and they do not meet the requirements for excluded status. Hence, these vaccine strains must be handled under Biosafety Level 3 (BSL3) conditions and retain federal select-agent designation and the associated regulatory burden. Therefore, we decided to test another strategy to create a vaccine strain that is both attenuated and nonencapsulated. The attenuated strain used to create new mutants was the established CO92 pgm- pPst- mutant, which is excluded from the select agent list based on established criteria (https://www.selectagents.gov/sat/exclusions/hhs.htm) (42). In addition, based on historical data, encapsulated LAVs induce a robust anti-F1 response that appears to dominate the protective humoral arm of the immune response, likely redirecting antibody responses from other potentially protective targets such as Yops or V antigen. To alleviate possible interference of F1 in the induction of the immune response to other antigens, a caf1 deletion mutant on CO92 pgm- pPst- background (Δcaf1) was constructed. Furthermore, a yopD mutation was shown to upregulate expression of various T3SS associated genes, including effector Yops and V (45). To further promote the induction of an immune response to more diverse protective antigens that may be naturally present at below immune-activation threshold levels, a yopD mutation (ΔyopD/Δcaf1) was also constructed in the pgm- pPst- strain.

To test the protective efficacy of these strains as LAVs, BALB/c mice were immunized via heterologous prime-boost vaccine regimens. For half of the immunized animals, the prime was with a LAV (either Δcaf1 or ΔyopD/Δcaf1), followed by boost with rF1-V. For the other half, the prime was with rF1-V, and the boost was with Δcaf1 or ΔyopD/Δcaf1 LAV ( Table 1 , experiments B and C). Equal numbers of age-matched males and females were used. Males and females were exposed simultaneously to the same aerosolized bacteria; however, the males were estimated to have inhaled about 15-20% more bacteria, based on differences in body mass and estimated lung volumes (50, 51).

Only 10% of female and 0% male mice immunized with P:Δcaf1 B:rF1-V were protected against aerosol challenge with ~6-7 LD₅₀ of strain C12. However, when the order was reversed and mice were immunized with P:rF1-V B:Δcaf1, there was significantly higher survival for both sexes. Furthermore, there was significantly higher survival in females than males given this vaccine regimen (100% for females, 50% for males) ( Figure 2A ). The importance of giving the rF1-V vaccine first was also seen using ΔyopD/Δcaf1 LAV. 100% of females and 90% of males survived aerosol challenge when immunized with P:rF1-V B:ΔyopD/Δcaf1, compared to 20% and 0% when the order was reversed ( Figure 2B ). Males and females had similarly high survival in this experiment, likely related to the lower estimated inhaled dose in this iteration.

Mice were euthanized at 3 dpi and bacterial burdens in lungs, spleen and blood were quantified as described above. Among mice immunized with Δcaf1 regimens, those that received P:rF1-V B:Δcaf1, which was 100% protective in females and 50% in males, had the lowest bacterial burden in the lungs and spleen, and those that received P:Δcaf1 B:rF1-V (0-10% protective) also had significantly lower bacterial burden than the PBS sham-vaccinated group ( Figures 3A, B ). The fraction of mice with bacteremia ( Figure 3C ) reflected the bacterial burden seen in the organs, with all males receiving the P:Δcaf1 B:rF1-V (0% protective) being bacteremic, in addition to having bacterial burden in the lungs similar to the PBS group.

Among mice given ΔyopD/Δcaf1 vaccine regimens, a similar pattern was seen. Those that received P:rF1-V B:ΔyopD/Δcaf1 (100% protective in females, 90% in males) had significantly lower bacterial burden in the lungs and spleen than the PBS group. Those that received P:ΔyopD/Δcaf1 B:rF1-V (0-20% protective) had significantly lower burden in the spleen, but not in the lungs due to the very high bacterial burden in males, which again were 0% protected by the vaccine ( Figures 3D, E ). All vaccinated females, and 7/8 vaccinated males, had no detectable bacteremia ( Figure 3F ), indicating that even the unsuccessful vaccine regimens enabled some control of systemic bacterial spread from the lungs.

To learn more about the state of infection in male and female mice, we measured the expression of a panel of cytokines in lung and spleen homogenates at 3 dpi. Thirty cytokines were analyzed for lungs, and 27 cytokines for spleens. In this experiment, additional mice were included that had been challenged alongside the mice in Table 1 (experiments B and C) but received an immunization strategy that protected 0% of both males and females (data not shown). This unsuccessful strategy consisted of one injection of ΔyopD/Δcaf1 or Δcaf1 LAV, and one injection of approximately 10⁸ radiation-inactivated C12 strain CFU mixed with supernatant containing LcrV protein adsorbed to Alhydrogel. Including these samples, in addition to samples from the vaccine groups shown in Figures 2 , 3 , increased the sample size to 36 males and 33 females.

Most cytokines had a clear bimodal distribution, with higher levels in samples with greater bacterial burden. For lungs, this bimodal pattern could be captured by categorizing the samples as high-burden (> 10⁶ CFU/g) or low-burden (< 10³ CFU/g). For spleens, high-burden was defined as > 8x10³ CFU/g and low-burden as < 2x10³ CFU/g. Table 2 shows the levels of all cytokines in lung samples, stratified by sex and bacterial burden (two columns on the left). Interestingly, among lung samples with low bacterial burden, males had a pattern of lower cytokine levels than females, particularly IL-5 and IL-13 (over 5-fold higher in females). For 9/30 cytokines, the average level in females was over twice that in males. This pattern was not seen in lung samples with high bacterial burden.

Fold differences between high-burden and low-burden samples is shown in the right-hand columns of Table 2 . For both males and females, most cytokines were positively correlated with high bacterial burden, many over 10-fold, especially canonical proinflammatory cytokines such as IL-6, IL-1β, and IFN-γ, and others like G-CSF, GM-CSF and GRO-α (CXCL1) that attract and stimulate innate immune cells. A total of 21/30 cytokines were significantly overexpressed (p < 0.05) in lungs with high bacterial burden in both males and females. Exceptions to this pattern were ENA-78 (CXCL5) and IL-2 (correlated with bacterial burden in females only); IL-5 and IL-13 (correlated with burden in males only); and RANTES and Eotaxin which had high expression (over 300 pg/mL) in both low- and high-burden samples. The only cytokines significantly lower in organs with high bacterial burden were Eotaxin, IL-28, and IL-23, with this relationship seen for IL-28 and IL-23 in both sexes, and Eotaxin in males only.

Linear regression analyses, using CFU/g to predict the level of each cytokine in the lung tissue, were calculated after natural log (ln) transformation, for the male and female populations. Examples of regression lines are in Supplementary Figure 2 . The regression coefficients (slope and intercept) were compared between male and female populations to see which cytokines had significant sex differences in the relationship between cytokine and bacterial burden. Slope and intercept, along with sex differences (female minus male) and p-values of the sex differences, are shown in Supplementary Table 3 . For IL-5, IL-13, and GRO-α (CXCL1) the slope and intercept are both significantly different (p < 0.05). IL-5 and IL-13 are the only two cytokines where levels decreased with higher bacterial burden in females but increased in males. Also, IP-10 (CXCL10), MIP-1α (CCL3), MIP-1β (CCL4), IL-27, and LIF had p < 0.05 between males and females for one of the regression coefficients and p < 0.10 for the other.

For comparison across vaccine groups, we removed from analysis the 23 cytokines whose expression in lungs were positively correlated with bacterial burden. Of the remaining 7 cytokines, some trends were of interest, although no comparisons (n = 3 or 4) reached statistical significance. In mice vaccinated with P:rF1-V B:Δcaf1 LAV, females exhibited higher IL-4, IL-5, IL-13 and RANTES levels on average than the corresponding males ( Supplementary Figure 3A ). In mice vaccinated with P:rF1-V B:ΔyopD/Δcaf1 LAV, females had more IL-4, IL-13, IL-2, and Eotaxin, and less ENA-78 ( Supplementary Figure 3B ). In combination with the above data showing that females had less pronounced cytokine overproduction in high-bacterial burden samples compared to low-burden samples, this suggests that females had a better ability to control damaging inflammation in the lungs.

Because Th2 cytokines (IL-4, IL-5, IL-13) (52) were higher in females in the low bacterial-burden group than males ( Table 2 ), we looked for evidence that females had elevated tissue eosinophils. Additionally, we looked for evidence of elevated tissue neutrophils, having seen that low-burden females also had higher chemokines secreted by neutrophils (GRO-α, IP-10, MIP-1α, MIP-1β) (53). Lung homogenates were quantified for eosinophil peroxidase (EPO) and myeloid peroxidase (MPO), which are released prolifically by eosinophils and neutrophils respectively during antibacterial response (52, 54). Among the low-burden group, females and males had similar bacterial burdens ( Figure 4A ) and MPO levels ( Figure 4B ). EPO was twice as high in females, with a geometric mean of 37.5 compared to 18.8 in males, although the difference was not statistically significant (p = 0.107) in a nonparametric test ( Figure 4C ). Among the high-burden group, bacterial burden was significantly higher in males ( Figure 4D ), but males and females had similar MPO and EPO levels ( Figures 4E, F ).

The levels of all cytokines in spleen samples at 3 dpi were stratified by sex and bacterial burden ( Table 3 , left columns). Unlike in lung samples, the spleen samples with low bacterial burden had similar cytokine levels in males and females, with the only differences greater than 1.5-fold being MCP-3 (CCL7), GRO-α (CXCL1), and G-CSF (higher in males) and MIP-1α (CCL3), MIP-1β (CCL4), IP-10 (CXCL10), and M-CSF (higher in females). This observation was surprising since 10/11 females in the low bacterial burden group (< 2,000 CFU/g) had no detectable bacteria in the spleen, compared to only 4/11 males ( Figures 3B, E ).

Fold differences between high-burden and low-burden samples are shown in the right-hand columns of Table 3 . Even more so than in lungs, cytokines were overexpressed in spleens with high bacterial burdens, with this overexpression being significant (p < 0.05) for both sexes in 21 of 27 cytokines.

Linear regressions of ln-transformed data were also performed for cytokine levels in spleen samples. In comparing male and female regression coefficients ( Supplementary Table 4 ), females had higher intercepts for most cytokines, with several being significantly higher. The only cytokines with a significantly different slope of the regression line were ENA-78 (CXCL5) and IP-10 (CXCL10), where overexpression in high-burden samples was more pronounced in males. These results differ from the cytokine results in lung samples. In lungs, there was a pattern of females having higher cytokine levels at low bacterial burden but males having more upregulation at high bacterial burden, while in spleens both sexes had similar upregulation from low to high.

We have previously demonstrated the power of layering medical countermeasures in a pneumonic plague model (34). In those studies, we demonstrated that suboptimal vaccination regimens combined with suboptimal (delayed) antibiotic treatments can lead to statistically significant synergy and a dramatic increase in survival rates over vaccine or antibiotic alone. Here we sought to corroborate our previous work by vaccinating with the less protective regimens used in the current studies, and then administering streptomycin to see if this could augment the suboptimal level of protection observed. In this experiment we gave mice suboptimal vaccine regimes, challenged them with aerosolized Y. pestis C12 four weeks post-boost, and then attempted to rescue the challenged mice with streptomycin treatment starting 60 h post-challenge ( Table 1 , experiment D). The positive-control vaccination sequence (P:rF1-V B:Δcaf1), even without streptomycin, conferred 90% protection in female mice and 60% protection in male mice. Suboptimal vaccine regimes alone (P:Δcaf1 B:rF1-V or P:ΔyopD/Δcaf1 B:rF1-V) protected 30%-40% of female mice, and streptomycin alone only delayed the time to mortality. However, when layered together these suboptimal countermeasures protected 70%-90% of the female mice, significantly more than streptomycin alone ( Figure 7A ). The layering of medical countermeasures resulted in statistically significant synergy in female mice immunized with P:Δcaf1 B:rF1-V (p = 0.0017) or nearly significant synergy in female mice immunized with P:ΔyopD/Δcaf1 B:rF1-V (p = 0.0575), using the Bliss independence model to calculate synergy scores (62). However, in males, when streptomycin was given to mice immunized with a suboptimal vaccine, still only 12%-30% of males were protected. In males the layering of suboptimal vaccine and streptomycin provided no significant benefit beyond streptomycin or vaccine alone, with no evidence of Bliss synergy ( Figure 7B ).

Y. pestis is a re-emerging threat due to the existence of natural plague foci in many countries, its potential as a biothreat agent, rapid progression to severe disease, and acquisition of antibiotic resistance. Furthermore, although most wild-type strains produce a robust outer capsule composed of polymeric F1, natural isolates deficient in capsule production have been identified (63–65) and studies confirmed that capsule production is not essential for virulence (22, 34, 66, 67). The most established vaccine platforms rely on anti-F1 and anti-LcrV immune responses to control infection. An immune response against F1 will not protect against capsule-deficient strains and thus, protection against these strains would only be attributed to anti-LcrV response. There is a great need for effective medical countermeasures that include novel vaccine formulations and schedules that provide robust protection against both capsule-producing and capsule-deficient strains.

To this end we developed two novel LAVs (pgm- pPst- Δcaf1 and pgm- pPst- ΔyopD/Δcaf1) with the aim of enriching for protective antigens exclusive of the F1 protein. Importantly, these results provide experimental data in mice that can be used to develop or refine next-generation vaccination strategies. We acknowledge the inherent difficulty in manufacturing and delivering heterologous vaccines, including but not limited to: requirement of separate manufacturing streams, the disparate requirements for ensuring quality control to produce protein subunit and LAV vaccines, and even separate distribution and delivery requirements. While heterologous vaccination strategies are clearly more complex than traditional homologous approaches, this work highlights their potential utility to combat isolates of bacteria with varying virulence attributes, which is of particular importance in context of antimicrobial resistance, engineered threats, and potential emerging or re-emerging bacterial pathogens.

A second objective of this work was to corroborate and further characterize the sex differences in vaccine protective efficacy after intranasal challenge with Y. pestis reported by Bowen et al. (47) in both BALB/c and C57BL/6 mice. These differences were an important, albeit hard to explain, observation and clearly must be considered when using a mouse model to evaluate medical countermeasures against a biodefense pathogen.

We tested the novel LAVs using heterologous vaccination strategies combining a dose of LAV with a dose of rF1-V subunit vaccine, 3-4 weeks apart. We saw that the order of administration of the two doses was critical: priming with rF1-V subunit, followed by either Δcaf1 or ΔyopD/Δcaf1 LAV boost, was more protective than LAV prime followed by rF1-V boost for both males and females. Priming with the subunit vaccine offers clear advantage in this model of pneumonic plague with the vaccines described in this study. This vaccine strategy may provide a more robust stimulation of memory B cells with a more effective booster immune response due to the LAV producing native rather than recombinant proteins. Priming with either LAV and boosting with rF1-V led to no more than 20% protection, differing from our recent study where priming with Combo ΔyscN LAV and boosting with rF1-V conferred 90% protection in female mice (34). One explanation could be that the ΔyscN strains express a wider array of Y. pestis proteins due to their retention of the pgm locus and pPst plasmid (11, 48). In contrast, Δcaf1 and ΔyopD/Δcaf1 are on a pgm- pPst- background. The ΔyscN strains do have greatly impaired virulence (37–39) as they are unable to secrete Yops through the T3SS. This sequestration of the Yops might also assist the development of immune memory because of the anti-inflammatory, anti-phagocytic and otherwise immune suppressive functions of Yops (68).

This study builds on other studies on heterologous prime-boost immunization, against Y. pestis (69) or other pathogens (70–73), showing major differences in immunity depending on which formulation is administered first. Our studies found that female mice were more protected than males by these heterologous vaccines, whether they included pgm- pPst- ΔyopD/Δcaf1, pgm- pPst- Δcaf1, or CO92 ΔyscN + C12 ΔyscN as the LAV. Especially interesting was that in addition to enhanced vaccine protection, female mice were more responsive to a post-challenge antibiotic treatment designed to be ineffective for sham-vaccinated mice, but rescue those with partial vaccine protection ( Figure 7 ). These data demonstrate that the disadvantage observed in the vaccinated male mice is substantial enough that it cannot be augmented with antibiotic treatment in the same manner as females.

We saw more male survival with the ΔyopD/Δcaf1 vaccine (90%) than the Δcaf1 vaccine (50%). However, the Δcaf1 vaccine experiment used more than double the challenge dose, so we would not consider this evidence that the ΔyopD/Δcaf1 vaccine is more effective, especially given the serology and ELISpot data. Difference in challenge doses may also account for the slight increase in bacterial dissemination from the lungs, as seen with more Δcaf1 vaccinated mice having bacteremia than corresponding ΔyopD/Δcaf1 vaccinated mice.

Bacterial burden in lungs was low (compared to sham immunization) in both males and females receiving rF1-V prime/LAV boost vaccines, and was similarly low in the females receiving LAV prime/rF1-V boost vaccines that were only 10-20% protective. Males receiving LAV prime/rF1-V boost vaccines, however, had high bacterial burden in lungs, like sham-vaccinated mice. However, even these males had less bacteria in the spleen than sham-vaccinated mice. This could be due to vaccine-associated mitigation or delay of systemic dissemination of bacteria at 3 dpi, without controlling bacterial replication in the lungs. Supporting this, for the mice where bacterial replication in the lungs was uncontrolled despite having high anti-rF1-V antibody titers ( Figure 6A ), bacteria in the spleen were partially controlled ( Figure 6B ).

When tissue cytokine levels were measured in C12-challenged mice, we saw that most of the cytokines, unsurprisingly, were present at higher levels in samples with more bacteria. Among the exceptions were IL-5 and IL-13, which were positively correlated to bacterial burden in males, but negatively correlated to bacteria in females ( Supplementary Figure 2 , Supplementary Tables 3, 4 ). For another group of cytokines (GRO-α, IP-10, MIP-1α, MIP-1β, IL-27, and LIF), both males and females had positive correlation between cytokine and bacterial burden, but one or both regression coefficients were different between sexes. These sex differences were largely due to females having higher cytokine levels than males in lungs with low bacterial burden. Variation within the population of low-burden samples is interesting, because this is a population that has been challenged, has controlled the bacterial growth usually seen in the preinflammatory phase, but still might succumb depending on the strength of the immune response. IL-5 and IL-13 are associated with Th2 immunity, and GRO-α, IP-10, MIP-1α, and MIP-1β are known to be induced by pattern recognition receptor (PRR) recognition of bacterial pathogen-associated molecular patterns (PAMPs), and produced by neutrophils, being chemotactic for various inflammatory cells (53, 74–76). Supporting the IL-5 and IL-13 findings, we saw that there was a near-significant enhancement (p = 0.077) of the eosinophil protein EPO in female low-burden lung samples compared to males, which was not seen for neutrophil protein MPO. Among high-CFU mice, although males had significantly higher bacterial burden, there was no sex difference for MPO or EPO.

The assay best predicting survival in the present study was serological titer against the antigens used in subunit vaccine, as has been seen before (77–79). All vaccinated groups had considerable antibody titers against rF1-V and rV antigens ( Tables 4 , 5 ), including those with no protection from mortality or bacterial growth in lungs, but the highest titers were in the groups with the best survival (females given rF1-V prime and LAV boost). Survival was also correlated with titer against TS CO92 whole-cell antigen, though not with titer against TS C12, even though the challenge was with C12. This suggests that the antigens exposed on irradiated whole-cell Y. pestis might not reflect the antigens expressed in vivo (34). The relatively low antibody titers against whole-cell antigen preparations contrast with substantial ex vivo re-stimulation of splenocytes, suggesting that in these preparations there are unspecified molecules that play an important role in cellular immunity. Some of that response is possibly attributed to endotoxin that is present on the LAV and the irradiated samples used for restimulation assays, although Y. pestis lipopolysaccharides when grown at 37°C are known to be less stimulatory than endotoxin from other bacteria (80–82).

In comparing the two LAVs, mice vaccinated with P:rF1-V B:Δcaf1 had greater antibody titers against both TS CO92 and TS C12 than mice vaccinated with P:rF1-V B:ΔyopD/Δcaf1 ( Table 6 ). Female mice vaccinated with P:rF1-V B:Δcaf1 also had higher anti-rF1-V antibody titers than those given P:rF1-V B:ΔyopD/Δcaf1 ( Table 4 ). Although the two LAVs were similarly protective, the ΔyopD/Δcaf1 strain might be less immunogenic if it replicates less than Δcaf1 in vivo and/or is cleared faster from the body, which was suggested by earlier work showing fitness impairment in yopD mutants (46, 83).

We have sought to reduce the Th2 polarization of vaccine-induced immune responses, and make them more balanced between Th1 and Th2, by adding CpG to subunit vaccine formulations (22). This objective is based on the expectation that a robust Th1 response is needed to protect against pneumonic plague (47, 84). In the present study, IgG1 titer measurements were consistently higher than IgG2a, partially due to more sensitive assay conditions, but rF1-V prime/LAV boost led to a more balanced response, with more IgG2a antibodies (linked to Th1) compared to the reverse strategy with LAV prime/rF1-V boost. Similar results were obtained in an earlier study where subunit prime with LAV boost led to more Th1-skewed response, and LAV prime with subunit boost led to more Th2, although in that study all mice survived intranasal challenge (69). We also noted that although the P:Δcaf1 B:rF1-V regimen induced higher anti-rF1-V titers than P:ΔyopD/Δcaf1 B:rF1-V, the IgG2a response was more pronounced in the latter, with a higher IgG2a/IgG1 ratio in both males and females. The relatively similar IgG1/IgG2a ratios in male and female mice given P:rF1-V B:ΔyopD/Δcaf1 may relate to the similarly high protection in males and females, whereas in mice vaccinated with P:rF1-V B:Δcaf1, only 50% protection was achieved in males. A way to further increase the Th1/Th2 ratio could be by administering the vaccine intranasally, which has been seen to produce Th1/Th2 balanced immunity to Y. pestis, compared to Th2 bias from subcutaneous immunization (85). Based on other pathogens this could better establish multifunctional and long-lasting CD4 cells in the lungs (86, 87).

Like antibody titers, the number of B cells reacting to ex vivo stimulation with Y. pestis antigens as detected by ELISpot was higher in mice vaccinated with Δcaf1 regimens than corresponding groups vaccinated with ΔyopD/Δcaf1, especially against whole cell antigens. These trends were not seen in T cell ELISpot assays. However, there was some role for cellular immunity in the differential protection between vaccine groups, as seen by differences in the amounts of cytokines secreted upon restimulation.

In splenocytes from vaccinated mice stimulated ex vivo with rF1-V, the females immunized with rF1-V prime and LAV boost had less production of cytokines compared to other vaccinated groups ( Supplementary Table 7 ). This includes IL-18, IFN-γ, IL-6, MIP-1α, MIP-1β, MCP-3, and IP-10, a suite of cytokines that in vivo would be highly inflammatory and attract phagocytes to the tissue (88). Meanwhile, when splenocytes were stimulated ex vivo with TS C12, females immunized with P:rF1-V B:Δcaf1 also had a more limited expression of cytokines compared to the other Δcaf1 vaccinated groups, with several cytokines underexpressed compared to the PBS group. In females given P:rF1-V B:ΔyopD/Δcaf1, this pattern was not as noticeable; instead, both males and females had lower expression of several cytokines compared to those given P:ΔyopD/Δcaf1 B:rF1-V ( Supplementary Table 10 ). This suggests that more protective vaccines led to less of the damaging systemic inflammation, similar to what we have observed previously in tissue extracts (34). Female mice four weeks post-boost also had a higher percentage of CD44+ splenic CD4 and CD8 T cells than the corresponding males, suggesting more memory cells. A recent study used three common mouse strains and found strain-specific sex differences in immune cell populations at steady state, with BALB/c females notably having more splenic CD8+ T cells, regulatory CD4+ T cells, plasmacytoid DCs, and fewer NK cells than males (89).

Sex differences in intrinsic Y. pestis resistance have rarely been observed, with most studies showing similar susceptibility between sexes. Several research groups have combined males and females in pneumonic plague experiments without noting any difference in outcomes, using BALB/c mice (78), C57BL/6 mice (49, 90), rats (91), and NHPs (92). In some studies females even fare worse, as assessed by mortality in rats (90) or other markers, such as earlier time to bacteremia in Swiss Webster mice (93). One group looked at Brown Norway rats and saw no difference in mortality but reported minor differences in vital signs as monitored by telemetry (94). In an intravenous infection model using KIM5 and KIM27 strains, naïve males were more susceptible (95–97). But in murine aerosol and intranasal Y. pestis infections, clear examples of naïve males being more susceptible to challenge than naïve females were not identified; rather, immunized male mice may be less protected than immunized females.

One study used four strains of mice and found that males vaccinated with rF1 and rV subunits had no impairment in antibody titers, but slightly increased mortality when challenged with the GB strain of Y. pestis. The increased mortality was observed in CBA and CB6F1 mouse strains but not in BALB/c or C57BL/6 (98). A more recent study showed surprising differences between males and females vaccinated with rF1-V + Alhydrogel, with vaccinated female BALB/c and C57BL/6 mice being significantly more resistant to challenge than vaccinated males, despite being similar for all immune factors measured in the study (47). Likewise in our current study, we also saw females more likely to be protected by vaccination, most notably vaccination with rF1-V prime and Δcaf1 LAV boost (100% protection in females and 50% in males in Experiment B, 90% and 60% in Experiment D), and the combination of LAV prime, rF1-V boost, and post-challenge streptomycin (70-90% protection in females and 12-30% in males).

Many sex differences in immune responses to vaccines have been described, whether in mice or human patients. This difference is most often linked to the fact that females have higher antibody levels, both baseline and antigen-specific antibodies induced by vaccines, but also the large number of immune genes and microRNAs on the X chromosome (99). Enhanced humoral immunity protects against infections, but also predisposes females to autoimmune disorders (100). Sex hormones, especially estradiol, influence factors like cytokine secretion, phagocytosis, antigen presentation, and endothelial permeability, with effects varying across different systems (101, 102). Androgens appear to be protective against overactive Th2 immunity in the lungs, as shown in allergy models where female mice had higher levels of eosinophils and associated airway inflammation (103, 104). In this study we saw significantly elevated IL-5 and IL-13 in the lungs of females, compared to males with similar bacterial burden, and EPO in lung tissue was also elevated in females, approaching significance.

The tendency for females to have elevated humoral immune responses does not extend to innate immune factors. Even though experimental injection of endotoxin in women induces more cytokines than in men (105), and female cells express more innate pathogen sensors like TLR2, TLR4 and TLR7 (106, 107), females tend to have innate regulatory mechanisms producing relatively restrained complement activation, controlled proinflammatory cytokine secretion (“cytokine storm”), and better survival during sepsis (56, 57, 108). In one study, male mice had more inflammation and tissue damage in both bacterial and sterile peritonitis, due to differences in resident immune cell populations (107). On the other hand, restrained innate immunity makes female mice more susceptible in some bacterial infection models (109, 110), and women seemingly more susceptible to diseases like cystic fibrosis, listeriosis, and extrapulmonary tuberculosis (111–113).

In pneumonic plague, the phenomenon of the pre-inflammatory phase of infection means that vaccination needs to activate immunity and neutralize bacteria before they can use their immune evasion factors to replicate undetected. An influx of neutrophils is beneficial at this stage, restricting the bacterial growth that would induce a greater, more damaging neutrophil influx later (54, 114). Macrophages and neutrophils derived from female subjects have enhanced phagocytosis in various experimental systems (107, 115, 116), and enhanced bacterial killing by alveolar macrophages (117). Finally, female hormones are well established as beneficial in clearing bacteria from lungs and recovering lung function after injury (118–120).

Although sex differences in immunity do not appear to produce differences in pneumonic plague survival for naïve animals, we hypothesize that in immunized mice with high opsonizing antibody titer against Y. pestis, a synergistic effect could give females more efficient phagocytosis, bacterial killing, and/or cytokine production by phagocytes. This would then lead to earlier influx of neutrophils, minimizing both bacterial spread and the damage of the pro-inflammatory phase. In our highly susceptible mouse model, immunized males had incomplete protection from pneumonic plague, and less beneficial effect from post-challenge antibiotics. This went along with males having lower serum antibody titers, fewer CD44+ T cells, and different cytokine expression patterns, including less IL-5, IL-13, and eosinophil activity in male lungs. Meanwhile, males and females were similar for other factors like IFN-γ secretion by ex vivo restimulated T cells, and neutrophil activity in lungs. C3a and C5b-9 levels were higher in males, indicative of higher complement activation in males as seen in previously published data (56, 57). When comparing vaccinated mice to those receiving PBS, vaccinated females had greater reduction in C3a and C5b-9 than vaccinated males did, which may be attributed to lower bacterial load as well as a different antibody response against various epitopes. This may mean that complement activation is enhanced in males, specifically the functional response through third-order avidity interactions with such factors as C1q.

Males and females in our study could all control bacterial spread to some extent, although males who received nonprotective vaccine regimens differed from females by having uncontrolled bacterial replication in the lungs. Further work is needed to establish whether sex differences are seen with other established vaccines and in other models besides mice. It was surprising to us that these sex differences could not be overcome by providing suboptimal post-exposure antibiotic therapy to vaccinated males. We will continue to explore this concept in future experiments.

The immune protection given by our heterologous prime-boost vaccine is likely a combination of enhanced antibody titer against antigens on the bacterial surface; enhanced titer against antigens expressed upon replication; and immune priming that lets innate immune cells respond rapidly without being counteracted by the bacterial Yops that impair phagocyte function so effectively in naïve animals. In addition, the relatively higher IgG2a/IgG1 antibody ratio signifies a stronger T cell response in addition to humoral immunity. Finally, in a heterologous prime-boost incorporating a live attenuated bacterial strain, giving the subunit first would be recommended for safety reasons, as preexisting immunity from the subunit prime should further limit the replication of the LAV (69). Based on survival data, serology, and preferred safety profile, the best vaccine regimen from this study appears to be giving rF1-V with CpG and Alhydrogel as a prime, and pgm- pPst-/Δcaf1 LAV as a boost.

While the sex-related differences reported here are likely species-specific, more work is warranted to further elucidate the sex differences observed in mice vaccinated against plague as they may serve as a unique model to better understand correlates of protection. Working with male mice, particularly under high-containment conditions can be challenging for personnel, and aggression between cagemates can result in other scientific parameters that can confound data (i.e. physical wounding or physiological/immunological differences associated with stress), as recently reviewed by Biltz et al. (121) and Takahashi et al. (122). Social stress differs from other stressors in its effects on the immune system (123), and stress can be beneficial in establishing immune memory and priming innate immunity (124–127) though debilitating long-term (128). In addition to differences between dominant and subordinate mice, subordinate mice are variable in their susceptibility or resilience to a given social stressor (129), and mice wounded during aggressive encounters show immune phenotypes not seen in those that experience stress without wounding (130). These factors lead to greater variance among males than females and are not easily addressed (131). These issues with male mice are significant enough, in our opinion, to justify their exclusion in early exploratory research. However, the results of this study illustrate the importance of testing medical countermeasures using both male and female animals during later-stage research and in advanced development studies before making conclusions that would apply to the entire population.

The virulent nonencapsulated Y. pestis strain C12 was derived from the clinical isolate CO92 (35) through site-directed mutagenesis of the caf1 gene (36). C12 was used in all challenge experiments. C12 and the encapsulated strain CO92, both inactivated by γ-radiation, were used as antigen stimuli in in vitro experiments. CO92 ΔyscN and C12 ΔyscN were derived from an in-frame deletion of the gene in the CO92 background, as previously described (39). CO92 pgm- pPst- is a highly attenuated strain made by curing the pPst plasmid via serial passage at 4°C, and additional passage on Congo Red agar leading to a deletion of the unstable pgm locus (39, 41). CO92 pgm- pPst- was used as the basis for the novel Δcaf1 and ΔyopD/Δcaf1 mutants described below.

Bacteria were plated on 5% sheep blood agar (SBA) plates, or tryptose blood agar (TBA) slants, and grown in liquid culture in heart infusion broth supplemented with 0.2% xylose (HIBX). For vaccine injections, bacteria were resuspended in 10 mM potassium phosphate, pH 7.3–7.4 (“KPhos”). SBA was from Remel Inc. (Lenexa, KS); TBA and HIB were from BD Difco (Sparks, MD).

Primers and plasmids used to construct the novel mutants described herein are listed in Supplementary Tables 1, 2 . For construction of the caf1 deletion, flanking regions of the gene were amplified using the primers listed to create a 919 bp upstream fragment and a 973 bp downstream fragment. Overlapping PCR was used to splice the two amplicons together to generate a fragment lacking the coding sequence of caf1. The Y. pestis fragment contained SphI and SacI restriction sites engineered into the primers and were used to ligate the fragment into the Y. pestis suicide vector pCVD442 (132) to generate the plasmid pCVD422-Δcaf1. To generate the Y. pestis ΔyopD mutant, FastCloning method was utilized (133). The ΔyopD upstream deletion fragment was 1,053 bp and downstream fragment was 1,038 bps and cloned into pCVD422 plasmid to generate pCVD422-ΔyopD. The resulting plasmids were verified by PCR screening and Sanger sequencing.

The pCVD422-Δcaf1 was electroporated into electrocompetent Y. pestis pgm- pPst- (41) using the methods previously described (134) to generate a cointegrate strain. Briefly, cointegrates were selected on LB Lennox agar plates containing 50 μg ampicillin/mL. The cointegrate strains were grown overnight in HIB and plated on LB agar plates containing 5% sucrose to select for allelic exchange recombinants. Those colonies that grew in the presence of sucrose were screened for loss of the caf1 gene by PCR with the primer pairs listed in Supplementary Table 2 . To construct the double mutant strain, the pCVD422-ΔyopD plasmid was moved into the electrocompetent derived Δcaf1 Y. pestis mutant. The allelic exchange mutant was obtained as described above. The ΔyopD mutation in this background was confirmed by PCR with the screening primers listed in Supplementary Table 2 .

The mutants constructed contain several highly stable mutations resulting in attenuation and in some cases exclusion from the US Federal select-agent regulations. The pgm deletion was the result of a loss of an approximately 102-kb deletion of the pgm locus as described by Jenkins et al. (41). The pPst deletion was the result of curing the entire plasmid from the strain as described by Welkos et al. (135). While the original caf1 mutant (strain C12 used for challenge experiments) was created by site directed mutagenesis in the fully virulent CO92 strain (36), the caf1 and yopD deletion mutants constructed in the pgm- pPst- parental background in this current study resulted from in-frame deletions of the caf1 and yopD genes. The caf1 deletion retained only the first three nucleotides and last three nucleotides of the coding region and the yopD deletion retained only the first 6 nucleotides and last three nucleotides of the coding region. Thus, the mutations created in the respective vaccine strains are very stable and the fact that these strains contain multiple significant mutations make spontaneous reversion highly unlikely if not impossible. If these LAV strains are selected for advanced development, additional phenotypic characterizations and whole-genome sequencing will be carried out.

BALB/c mice (males and females) were purchased from Charles River Laboratories (Frederick, MD), and were 7-9 weeks of age at the time of initial vaccination. Mice were housed in groups of up to 10 in ventilated filter-top cages connected to a common water supply. In some cases, male mice were housed in smaller groups to limit aggressive behavior. Euthanasia was carried out in accordance with IACUC approved early endpoint euthanasia criteria.

All vaccine formulations are described in Table 1 . In this study, all immunizations followed a prime-boost strategy, with two doses of vaccine or sham administered 3-4 weeks apart. Vaccine injections were either LAV (CO92 ΔyscN, C12 ΔyscN, CO92 pgm- pPst- Δcaf1, CO92 pgm- pPst- ΔyopD/Δcaf1), or subunit (recombinant F1-V fusion protein [rF1-V] mixed with adjuvant). rF1-V was, formulated under CGMP regulations and stored at -80°C in single-use aliquots (136, 137). Subunit vaccine doses included 2 µg of rF1-V in PBS, mixed with 250 µg Alhydrogel (InvivoGen, San Diego, CA), and 5 µg CpG (ODN 2006, InvivoGen) as additional adjuvant (22). The subunit vaccinations were delivered in 0.1 mL subcutaneous injections. Since all rF1-V subunit formulations were mixed with CpG and Alhydrogel, for the sake of brevity we abbreviate “rF1-V+CpG+Alhydrogel” as “rF1-V”.

For preparation of LAV strains, flasks of HIBX (supplemented with 2.5mM CaCl₂) were inoculated with a suspension of colonies from a freshly inoculated SBA plate and incubated for 24 h at 28-30°C with shaking (200 rpm). The next day, bacterial concentration was determined by OD₆₀₀, followed by diluting to an OD₆₀₀ of 0.1 into fresh medium and growth for 6 h, then harvesting and resuspending in KPhos to 5.0x10⁷ cells/mL, for a target of approximately 10⁷ cells in each subcutaneously injected dose of 0.2 mL (39). Colony-forming units (CFU) were confirmed by diluting and plating on SBA plates.

Approximately four weeks after vaccine boost, mice were exposed to challenge doses of aerosolized virulent Y. pestis C12. For aerosol doses, colonies from a freshly inoculated TBA slant were suspended in HIBX to an OD₆₂₀ of 0.01 and incubated for approximately 24 h at 28-30°C with shaking (150 rpm). The day of challenge, bacteria were harvested by centrifugation and resuspended in HIB (no supplemental xylose) to a concentration estimated to achieve the desired inhaled dose of bacteria. Exposure to aerosolized bacteria was accomplished as previously described (39). Briefly, mice were temporarily housed in wire mesh cages that were placed in a whole-body aerosol chamber, located within a class 3 biological safety cabinet in a biosafety level-3 laboratory. Aerosols were created by a three-jet Collison nebulizer and delivered to mice for 10 min. Samples of the aerosol were collected in an all-glass impinger, then plated on SBA plates to calculate the bacterial concentration and the estimated inhaled dose (138). When males and females were challenged in the same whole-body aerosolization cohort ( Tables 1 , experiments B and D), age-matched males had higher estimated inhaled doses due to greater minute volumes due to larger body mass (50, 51).

Bacterial burdens in mouse tissue were determined at 3 days post-infection (dpi). After terminal blood collection and euthanasia, spleens and lungs were excised and homogenized in 15 mL Precision tissue grinders (Covidien, Republic of Ireland) in a volume of 1 mL KPhos, while blood was collected in serum separator tubes (365967, BD, San Jose, CA) by terminal collection from axillary vessels. Organ homogenates were diluted in KPhos and plated on SBA plates (100 µl/plate), then CFUs counted. Fresh whole blood was also spread on SBA plates for counting; mice with any CFUs detected in 10 µl blood were considered bacteremic. Limits of detection were 5 CFU/organ and 100 CFU/mL blood. The remainder of the blood was separated, and serum was frozen. All samples from infected mice were sterilized via irradiation with approximately 21 kGy of γ-radiation, and lack of viable bacteria was confirmed by plating on SBA plates before storing samples at -80°C for later use in immunological assays.

Streptomycin at 20 mg/kg was delivered via intraperitoneal injection every 6 h for five days. As previously described (34), the antibiotic treatment was initiated at approximately 60 h after exposure to aerosolized Y. pestis C12. This resulted in a purposefully sub-optimal antibiotic regimen that could be used to identify synergistic effects between vaccination strategies and antibiotic treatments. Streptomycin was manufactured by XGen Pharmaceuticals (Big Flats, NY) and resuspended in water for injection (Corning Inc., Corning, NY).

Sera from terminal blood collection were measured for Immunoglobulin G (IgG) antibody levels by semi-quantitative endpoint ELISA in 96-well Immulon 2HB plates (Thermo Fisher, Rochester, NY). Plates were coated overnight with antigens at 4°C as previously described (22, 39). Coating antigen solutions were either rF1-V (2 µg/mL), rV (2 µg/mL), or inactivated Y. pestis CO92 or C12 whole cells (10 µg/mL). Recombinant F1-V was stored as described above and V protein (rV) was obtained from BEI Resources (Manassas, VA); aliquots were stored at -80°C until ready to use. Y. pestis cells used as antigen had been 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. These antigens were designated TS (temperature-shifted). Two-fold dilutions of serum in PBS/0.05% Tween 20 were made in triplicate and incubated for 30 min at 37°C, then washed and signal detected as previously described (22). 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.

Groups of mice were euthanized 25-28 days after the vaccine boost, and splenocytes were isolated using minimal disruption as described previously (39). Briefly, spleens were excised from mice (n = 5 mice per group) and disaggregated in RPMI 1640 medium (Gibco, Grand Island, NY). Red blood cells in the spleen homogenate were lysed with Ammonium-Chloride-Potassium (ACK) Lysing Buffer (Gibco) after the extract was diluted with RPMI 1640 medium and cells pelleted by centrifugation at 335 x g for 10 min. Splenocytes were resuspended and counted using a TC20 Cell Counter (Bio-Rad, Hercules, CA) then all samples were diluted in CTL-Test medium (IFN-γ ELISpot) and/or RPMI complete medium (Luminex and B-cell ELISpot) 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).

For analysis of cellular responses, splenocytes were incubated in CTL-Test medium (ImmunoSpot, Shaker Heights, OH) and stimulated in vitro with rF1-V (25 µg/mL) protein, rV (25 µg/mL) protein, inactivated Y. pestis TS CO92 (5 µg/mL), or inactivated Y. pestis TS C12 (5 µg/mL) and incubated at 37°C/5% CO₂. Negative controls were cells stimulated with medium only, while positive controls were stimulated with 100 ng/mL phorbol 12-myristate 13-acetate (PMA) and 500 ng/mL ionomycin. Before cells were plated, plates were coated with capture anti-mouse interferon-gamma (IFN-γ) monoclonal antibody for T-cell ELISpot. After approximately 24 h incubation, ELISpot plates were washed, developed, and spots detected and quantified as previously described (22).

For B cell ELISpot, splenocytes were first stimulated for 5 days with 1x B-Poly S Polyclonal B cell Stimulator (ImmunoSpot) in complete medium. B cell ELISpot plates were coated with rF1-V (15 µg/mL) protein, rV (15 µg/mL) protein, inactivated Y. pestis TS CO92 (5 µg/mL), inactivated Y. pestis TS C12 (5 µg/mL), or positive control wells coated with Anti-Igk and Anti-Igλ Capture Ab (ImmunoSpot). Antigen-treated plates were washed and splenocytes added and incubated for approximately 8 h, then washed, developed and spots detected as per manufacturer’s instructions.

For mice pre-challenge, splenocytes were isolated and counted as described above, then incubated in RPMI complete medium with antigens rF1-V (25 µg/mL), rV (25 µg/mL), inactivated Y. pestis TS CO92 (5 µg/mL), or inactivated Y. pestis TS C12 (5 µg/mL) and incubated at 37°C/5% CO₂. Negative and positive controls were cells stimulated with medium only, or PMA and ionomycin as described above. 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.

For mice post-challenge, lung and spleen homogenates at 3 dpi (see above) were frozen at -80°C and irradiated, then later evaluated for measurement 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. In each experiment, any cytokine whose standard curve had low R² value (< 0.95) was removed from analysis. Also, cytokines with insufficient positive samples (more than 2/3 of values at or below the lower limit of detection) were removed from analysis. Cytokine levels are expressed in pg/mL.

Lung homogenates at 3 dpi were measured for the proteins eosinophil peroxidase (EPO) and myeloperoxidase (MPO). Homogenates were diluted 1:50 for EPO and 1:500 and 1:5,000 for MPO measurement in the respective assay diluents. Serum samples at 3 dpi were measured for complement proteins C3a and C5b-9, after dilutions of 1:15 and 1:150 (C3a) and 1:50 (C5b-9) in assay diluents. Kits used were: MPO (HK210-02, Hycult, Uden, Netherlands), EPO (CSB-E15830m, Cusabio, Wuhan, China), C3a (NBP2-70037, Novus/Bio-Techne, Minneapolis, MN), and C5b-9 (LS-F22262, LSBio, Seattle, WA).

For mice pre-challenge, splenocytes were isolated and cryopreserved in freezing medium, then transferred to liquid nitrogen vapor storage. After thawing and counting, cells were labeled with viability dye (L34970 and L34964, Thermo Fisher), washed with FACS buffer, then incubated with 1:200 dilution of Mouse FcBlock (553142, BD Pharmingen, San Diego, CA) in FACS buffer. Cells were stained at 0.5-1.0 x 10⁶ cells/well in a 96-well round-bottom plate. Staining panel used for T cell subsets was CD3e (clone 145-2C11), CD19 (clone 1D3), CD4 (clone GK1.5), CD8a (clone 53-6.7), CD44 (clone IM7), and CD62L (clone MEL-14). Staining panel used for DC subsets was CD3e (clone 145-2C11), CD19 (clone 1D3), CD11c (clone N418), CD11b (clone M1/70), CD8a (clone 53-6.7), CD80 (clone 16-10A1), and CD86 (clone GL1). Samples were fixed in PBS with 2% formaldehyde (Thermo Fisher, Rockford, IL), and run on a FACSCanto II, then analyzed in FlowJo v10.8 (FlowJo, Ashland, OR). FACS buffer was PBS with 1% bovine serum albumin (SH30574, HyClone); freezing medium was 90% fetal bovine serum (SH30396, HyClone) and 10% DMSO (D2660, Sigma).

Survival curves of vaccinated and control mice were estimated with the Kaplan-Meier method. Potential synergistic effects of vaccine and antibiotic on survival were analyzed using the Bliss synergy (62). Synergy score values reflect the ratio of median TTD comparing vaccinated and antibiotic treated groups to groups vaccinated without antibiotic and were based on a log-logistic or log-normal parametric survival model.

Correlations were analyzed using the Spearman nonparametric test. Significant differences between groups were compared using the Mann-Whitney test for pairwise comparisons, with relevant comparisons (between males and females in the same vaccine group, or between vaccine groups within the same sex) shown in figures. For linear regressions using bacterial burden to predict cytokine levels, data were ln transformed and slope and intercept calculated for each cytokine, then the method of least squares was used to calculate p-values of the difference between females and males for slope or intercept. For ELISA antibody titers, tables show the geometric mean and standard error (GM and GSE) as previously described (34, 39). Values below the limit of detection (LOD) were replaced by LOD/SQRT 2. Statistical analyses were performed using SAS version 9.4, and figures were generated using Prism version 9.4 (GraphPad Software, Boston, MA), except as indicated. Tables were generated using Microsoft Excel. Conditional formatting in tables indicates the range from low (red) to high (green).