Preclinical efficacy of a cell division protein candidate gonococcal vaccine identified by artificial intelligence

Abstract

Vaccines to curb the global spread of multidrug-resistant gonorrhea are urgently needed. Here, 26 vaccine candidates identified by an artificial intelligence-driven platform (Efficacy Discriminative Educated Network[EDEN]) were screened for efficacy in the mouse vaginal colonization model. Complement-dependent bactericidal activity of antisera and the EDEN protective scores both correlated positively with the reduction in overall bacterial colonization burden. NGO1549 (FtsN) and NGO0265, both involved in cell division, displayed the best activity and were selected for further development. Both antigens, when fused to create a chimeric protein, elicited bactericidal antibodies against a wide array of gonococcal isolates and significantly attenuated the duration and burden of gonococcal colonization of mouse vaginas. Protection was abrogated in mice that lacked complement C9, the last step in the formation of the membrane attack complex pore, suggesting complement-dependent bactericidal activity as a mechanistic correlate of protection of the vaccine. FtsN and NGO0265 represent promising vaccine candidates against gonorrhea.

Keywords: Neisseria gonorrhoeae; artificial intelligence; complement; vaccine.

Fig 1

Efficacy of 26 gonococcal vaccine antigens against N. gonorrhoeae strains MS11 (A) and H041 (WHO X) (B) in the mouse vaginal colonization model. Mice were immunized with combinations of individual antigens (15 µg of each antigen) indicated at the top of each column in the graphs (corresponding NGO numbers, indicated above each column); vaccines were adjuvanted with either 5 µg GLA-SE (groups 1–6 and group 11) or 5 µg GLA-SE plus Al(OH)₃ (groups 7–10). Mice were immunized intramuscularly at weeks 0, 3, and 6. Adjuvant control mice received GLA-SE alone. Two weeks post-dose 3, mice in the diestrus phase of the estrous cycle (n = 5/group) were challenged intravaginally with N. gonorrhoeae MS11 (2.6 × 10⁷ CFU) (A) or H041 (WHO X) (3.8 × 10⁷ CFU) (B). Vaginas were swabbed daily to enumerate gonococcal CFUs. In panels A and B, the top rows show times to clearance (Kaplan–Meier curves; groups compared by Mantel-Cox analysis), the middle rows show log₁₀ CFU versus time, and the bottom rows show analysis of AUC (means, ±95% confidence intervals were compared across groups by Mann Whitney’s non-parametric test).

Fig 2

Bactericidal activity of IgG in antisera from the 11 groups of mice immunized with 26 gonococcal antigens and correlation with efficacy in mice. (A) Complement-dependent bactericidal activity of IgG in sera from 11 groups of mice immunized with 26 gonococcal antigens against four strains of N. gonorrhoeae. Sera from 10 mice, immunized but not challenged with N. gonorrhoeae, were pooled, depleted of mouse IgM, and incubated with strains FA1090, MS11, F62 ΔlgtD, and H041. IgG- and IgM-depleted normal human serum at a final concentration of 28% (or 11% for MS11) was used as a source of complement. CFUs at 30 min relative to 0 min are expressed as percentage as shown on the Y-axis. (B) Survival of MS11 and H041 in serum bactericidal assays correlates inversely with the percent reduction in AUC in the mouse vaginal colonization model. The percent reduction in median AUCs of immunized mice versus adjuvant control mice was plotted against percent survival in serum bactericidal assays. Data were analyzed using linear regression (note that one datum point in the H041 graph lies beneath the X-axis and is superimposed on the “90” X-axis label). (C) Correlation between the survival of MS11 (red circles) and H041 (blue circles) in serum bactericidal assays and percent reduction in median AUCs using Z-score values for both parameters. Z-scores for each data point were calculated using the formula Z = (X − µ)/σ, where X is the raw (non-normalized) value, and µ and σ are the mean and standard deviations, respectively of each of the four sets of data (survival in bactericidal assays and percent reduction in median AUC for each strain). (D) Correlation between strain-specific EDEN prediction scores (group mean) and relative reduction in median AUCs (Z-scores) for MS11 (red circles) and H041 (blue circles).

Fig 3

Efficacy of NGO0265 and NGO1549 either singly or in combination against N. gonorrhoeae strains FA1090 and MS11. Female BALB/c mice were immunized with NGO1549 or NGO0265 (15 µg of each antigen, either alone or in combination) adjuvanted with GLA-SE (5 µg), intramuscularly at 0, 2, and 4 weeks. Adjuvant controls (Adj. cont.) received GLA-SE alone. Two weeks post-dose 3, mice were challenged with either strain FA1090 (3.6 × 10⁷ CFU/mouse; n = 5/group) (A) or MS11 (2.8 × 10⁷ CFU/mouse; n = 5/group) (B). Vaginas were swabbed daily to enumerate gonococcal CFUs. Kaplan–Meier curves showing time to clearance (graphs on left). Graphs in the middle column show log₁₀ CFU versus time. Comparisons were made by two-way ANOVA and comparisons between each vaccine group with the control group on each day by Dunnett’s multiple comparisons test. *P < 0.05; **P < 0.01; ***P < 0.001; and ****P < 0.0001. The colors of the asterisks correspond to the color of the line of the vaccine group. Graphs on the right show AUC (mean, ±95% confidence intervals). Comparisons across groups were made by one-way ANOVA using the non-parametric Kruskal-Wallis equality of populations rank test. The Kruskal-Wallis statistics were 10.98 (P = 0.0119) and 10.22 (P = 0.0168) for the graphs in panels A and B, respectively. AUC comparisons between each group and the adjuvant control group were made using Dunn’s multiple comparisons test.

Fig 4

Efficacy of NGO1549 and NGO0265 chimeric proteins adjuvanted with GLA-SE against N. gonorrhoeae in mice. Six-week-old female BALB/c mice were immunized with NGO0265, NGO1549 (FtsN), a combination (mixture) of NGO0265 and NGO1549, a chimera of NGO0265 (N-terminal) and NGO1549 (C-terminal) (CHIM_0265_1549), or a chimera with NGO1549 (N-terminal) and NGO0265 (C-terminal) (CHIM_1549_0265). All formulations were adjuvanted with GLA-SE (5 µg/dose). NGO0265 and NGO1549 (FtsN) were each used at 15 µg/dose when given individually or in combination; chimeric proteins were administered at 25 µg/dose. Adjuvant control mice received GLA-SE alone. Mice were immunized intramuscularly at 0, 3, and 6 weeks and challenged intravaginally with N. gonorrhoeae H041 (WHO X) (6 × 10⁷ CFU) 2 weeks post-dose 3 (n = 10 mice per group). Vaginas were swabbed daily to enumerate CFUs. Left graph: time to clearance (Kaplan–Meier curves, analyzed by Mantel-Cox log-rank test). P values for the NGO0265, NGO1549, NGO0265 plus NGO1549, CHIM_0265_1549, and CHIM_1549_0265 versus the adjuvant control group were 0.04, 0.0001, 0.0003, 0.0009, and <0.0001, respectively. Middle graph: log₁₀ CFU versus time. Right graph: AUC (mean, ±95% confidence intervals). Comparisons across groups were made by one-way ANOVA using the non-parametric Kruskal-Wallis equality of populations rank test. The Kruskal-Wallis statistic was 33.8 (P = 0.0001). AUC comparisons between each group and the adjuvant control group were made using Dunn’s multiple comparisons test.

Fig 5

Chimeric protein CHIM_0265_1549 adjuvanted with aluminum hydroxide (Alhydrogel) effectively clears gonococcal vaginal colonization in mice. Six-week-old female BALB/c mice were immunized with CHIM_0265_1549 (25 µg) adjuvanted with Al(OH)₃ (Alhydrogel) intramuscularly at weeks 0, 3, and 6. Two weeks post-dose 3, mice were challenged intravaginally with N. gonorrhoeae H041 (WHO X). Vaginas were swabbed daily to enumerate CFUs. Left graph: Kaplan–Meier curves showing time to clearance, analyzed by Mantel-Cox log-rank test. Middle graph: log₁₀ CFU versus time, analyzed by two-way ANOVA (with Šídák’s multiple comparisons test). *P < 0.05; **P < 0.01; and ****P < 0.0001. Right graph: AUC analysis (mean, ±95% confidence intervals). Comparisons were made with Mann-Whitney’s test.

Fig 6

GLA-SE and alum elicit T_H1 and T_H2-biased responses, respectively. (A) IgG subclasses elicited by CHIM_0265_1549 adjuvanted with either alum or GLA-SE. IgG subclasses in sera from mice immunized with CHIM_0265_1549/GLA SE (red line/solid red squares; n = 6) or CHIM_0265_1549/alum (blue line/blue triangles; n = 5) were measured by ELISA. X-axis, reciprocal serum dilution; Y-axis, absorbance (OD_405nm). (B) The left graph shows half-maximum (Half-max) titers (calculated dilution that yields half-maximal OD_405nm; median [95% CI]). The graph on the right shows IgG1:IgG2a ratios for each mouse, calculated using Half-max titers (median).

Fig 7

Complement-dependent bactericidal activity of IgG in immune sera elicited by NGO1549 (FtsN) and NGO0265 either as individual proteins, in combination (mixtures), or when expressed as chimeric proteins (NGO0265 is N-terminal in CHIM_0265_1549 and C-terminal in CHIM_1549_0265). Female BALB/c mice were immunized with the antigens and adjuvants (or adjuvant alone) as indicated in the legend. These sera were obtained from the same groups of mice immunized in Fig. 4 and 5, but which were not used in the challenge experiments. Immune sera in each group were pooled and depleted of mouse IgM by passage over anti-mouse IgM agarose. Pooled immune sera were added to each of the four test strains [FA1090, MS11, F62 ΔlgtD, and H041 (WHO X)] to a final concentration of 56% or 28% (solid and hatched bars, respectively), followed by human complement (11% for MS11 and 28% for the three other strains). Percent survival of bacteria at time 30 min relative to the start of the assay is indicated on the Y-axis.

Fig 8

Efficacy of CHIM_0265_1549 requires an intact terminal complement pathway. WT C57BL/6 mice or complement C9^−/− mice in a C57BL/6 background were immunized with CHIM_0265_1549 (25 µg/dose) formulated with GLA-SE (5 µg/dose), or GLA-SE alone intramuscularly at 0, 3, and 6 weeks. Two weeks post-dose 3, mice (n = 5 mice/group) were challenged intravaginally with either 2.6 × 10⁷ CFU N. gonorrhoeae MS11 (A) or 3.2 × 10⁷ CFU H041 (WHO X) (B) and vaginas were swabbed daily to enumerate CFUs. Top row: Kaplan–Meier curves for time to clearance. Comparisons across groups were made by Mantel-Cox log-rank test. Middle row: log₁₀ CFU versus time, analyzed by two-way ANOVA (with Šídák’s multiple comparisons test). *P < 0.05 and **P < 0.01. Bottom row: AUC analysis (mean, ±95% confidence intervals). Comparisons across groups were made by Mann-Whitney’s non-parametric test.

Fig 9

Bactericidal activity of anti-CHIM_0265_1549 antiserum elicited in C57BL/6 and C9^−/− mice against a panel of 50 wild-type N. gonorrhoeae isolates. Antisera obtained from C57BL/6 and C9^−/− mice immunized with CHIM_0265_1549 (Fig. 8) were pooled, depleted of mouse IgM, and tested for their ability to support killing of 50 wild-type gonococcal isolates in the presence of 20% human complement (IgG- and IgM-depleted normal human serum) as the complement source. (A) Bactericidal activity of antisera obtained from wild-type (solid blue bars) and C9^−/− mice (red hatched bars) used at a dilution of 1:5 (final anti-serum concentration 20%). The gonococcal strains are listed on the X-axis. Percent survival (CFUs at 30 min relative to 0 min) is shown on the Y-axis. Immune serum or complement alone did not lead to the killing of any strains (>100% survival, data not shown). (B) Bactericidal activity of anti-CHIM_0265_1549 anti-serum obtained from C9^−/− mice (red hatched bars) used at a 2:5 dilution (final anti-serum concentration 40%) against the nine gonococcal strains that showed greater than 50% survival in Fig. 9A. Percent survival using pooled sera obtained from mice given GLA-SE alone (adjuvant controls) is shown by the gray bars.