Two mutations in NS2B are responsible for attenuation of the yellow fever virus (YFV) vaccine strain 17D

Abstract

Vaccines have done more to improve the health of humankind over the past century than almost any other technology. Among vaccines, the live-attenuated yellow fever (YF) vaccine (17D) is highly effective, providing long-lasting immunity against yellow fever virus (YFV) infection with a single dose. Developed in the 1930s through extensive serial passage of the virulent YFV-Asibi strain through mouse and chicken embryonic tissue, 17D acquired several mutations that render it attenuated in humans and non-human primates. Over the past century, 17D has become a widely studied immunogen and has also been developed into a vaccine platform for other pathogens. Despite this, most studies of 17D have focused exclusively on the host, without clearly defining the virus-intrinsic features of attenuation. Consequently, the genetic determinants of 17D attenuation remain unknown and are assumed to be multigenic. Here, we leverage the hamster host, which recapitulates many important features of human YF disease, to understand the genetic basis of 17D attenuation. We developed a YFV reverse genetics system and generated hamster-adapted Asibi/17D chimeric viruses, discovering that viruses containing 17D-derived mutations in the viral gene NS2B were significantly attenuated in the hamster. Further analysis revealed that the two non-synonymous mutations in NS2B that distinguish 17D from Asibi, I37L and I109L, act cooperatively to mediate attenuation, with both mutations required to fully prevent key features of YF disease including liver injury and coagulopathy. These findings establish NS2B as an important and unexpected determinant of YFV-17D attenuation in vivo. In addition to the implications of these findings for improving the efficacy and safety of the 17D vaccine platform, this discovery also provides a new direction for understanding more generalized principles and mechanisms of durable vaccine-induced immunity.

Fig 1. HA-Asibi but not 17D causes disease in hamsters.

Hamsters were infected by intraperitoneal injection with 1 × 10⁵ focus-forming units (FFU) of HA-Asibi (yellow, n = 15), 17D (blue, n = 3), or mock (black, n = 3). Data represents a composite of 6 independent experiments. (A) Survival analysis, (B) weight change as a percentage of weight at the time of inoculation, and (C) viremia at 3 days post inoculation (dpi). Error bars show the mean $\pm$ standard error of mean (SEM). In C, the dashed line shows the limit of detection; statistical significance determined via unpaired two-tailed t test with Welch’s correction for unequal variance (****: p < 0.0001).

Fig 2. Recombinant HA-YFVs generated via CPER recapitulate disease phenotypes observed with biological isolates.

(A) Schematic of the YFV genome showing amino acid differences in the polyprotein (yellow rectangle) between Asibi and 17D-204 (gray) along with mutations acquired during hamster adaptation of Asibi (purple with E-D155A shown in red). Overlapping DNA fragments amplified from YFV cDNA-containing plasmids are shown below, assembled in a circular polymerase extension reaction (CPER) with a linker sequence containing the CMV promoter and enhancer, bovine growth hormone polyadenylation (bGHpA) signal, and the self-cleaving hepatitis delta virus ribozyme (HDVr). The resulting circular CPER product was transfected into BHK-21 cells to rescue infectious recombinant YFV. The nomenclature for recombinant hamster-adapted YFV variants is shown in the box. (B) Heatmap showing the frequency of Asibi/17D mutations across biological and recombinant HA-YFVs. Viral genomes with an Asibi backbone were mapped to the Asibi reference (AY640589) and those with a 17D backbone were mapped to the 17D-204 reference (MN708488). Columns represent independently rescued virus stocks, and rows indicate individual amino acid differences between Asibi and 17D, grouped by CPER fragment, as well as all hamster adaptation mutations. Variant frequencies are color-coded: yellow indicates >50% Asibi variant, blue indicates >50% 17D variant. Frequencies of hamster adaptation mutations are shown using a red color scale. Grey boxes represent positions with low sequencing coverage (<100 reads). (C) Virus production from BHK-21 cells transfected with CPER product, quantified by focus-forming assay (FFU/mL: focus-forming units per mL of supernatant). (D) Growth kinetics of recombinant YFVs on Vero cells inoculated with passage-0 stocks (i.e., filtered and titered BHK-21 supernatants) at MOI of 0.1, quantified by focus-forming assay (E-G) 5-7-week-old female hamsters were inoculated intraperitoneally with 1 × 10⁵ FFU of rHA1-Asibi (orange, n = 17), rHA7-Asibi (brown, n = 15), rHA1-17D (light blue, n = 5), or rHA7-17D (dark blue, n = 3). Control groups from Fig 1 are included for comparison. Data represents a composite of 9 independent experiments. (E) Survival analysis. (F) Body weight at 6 dpi, shown as a percentage of weight at the time of inoculation. (G) Viremia measured at 3 dpi, expressed as YFV genome copies per mL of serum (log₁₀). Dashed lines represent the lower limits of detection. Error bars show mean $\pm$ SEM. Statistical significance determined via one-way ANOVA with multiple comparisons. In F, comparisons were made to the mock-infected group. In G, rHA1-Asibi and rHA7-Asibi were compared to HA-Asibi, while rHA1-17D and rHA7-17D were compared to 17D. (**: p < 0.01; ****: p < 0.0001; ns: not significant).

Fig 3. Attenuation determinant(s) of 17D map to Fragment 2a containing non-structural genes.

(A) Heatmap showing the frequency of Asibi/17D mutations across chimeric rHA1-YFVs in which individual CPER fragments were swapped between Asibi and 17D backbones; see description in Fig 2B for additional details. (B) Body weight of hamsters inoculated with 1 × 10⁵ FFU of each virus at 6 dpi, shown as a percentage of of starting weight. Data from animals inoculated with the recombinant parent virus are shown with bolded outlines. (C) Viremia measured at 3 dpi, expressed as genome copies per mL of serum (log₁₀). Dashed line in C represents the lower limit of detection. Error bars show mean $\pm$ SEM. Statistical significance was determined by one-way ANOVA with multiple comparisons, with Asibi-backbone chimeras compared to rHA1-Asibi, and 17D-backbone chimeras compared to rHA1-17D. (**: p < 0.01; ***: p < 0.001; ****: p < 0.0001; ns: not significant.).

Fig 4. Attenuation determinant(s) of 17D map to NS2B.

(A) Heatmap showing the frequency of Asibi/17D mutations across chimeric rHA1-YFVs in which individual genes (E, NS1, NS2A, and NS2B) were swapped between Asibi and 17D backbones; see description in Fig 2B for additional details. (B) Percent body weight of hamsters inoculated with 1 × 10⁵ FFU of each virus at 6 dpi, shown as a percentage of starting weight. Data from animals inoculated with the recombinant parent virus are shown with bolded outlines. (C) Viremia measured at 3 dpi, expressed as genome copies per mL of serum (log₁₀). Dashed line in C represents the lower limit of detection. Error bars show Data represented as mean $\pm$ SEM. Statistical significance was determined by one-way ANOVA with multiple comparisons, with Asibi-backbone chimeras compared to rHA1-Asibi, and 17D-backbone chimeras compared to rHA1-17D. (*: p < 0.05; **: p < 0.01; ****: p < 0.0001; ns: not significant.).

Fig 5. The NS2B-I109L mutation attenuates rHA1-Asibi.

(A) Heatmap showing the frequency of Asibi/17D mutations across rHA1-Asibi chimeras in which single or double 17D-NS2B mutations were introduced; see description in Fig 2B for additional details. (B) Percent body weight of hamsters inoculated with 1 × 10⁵ FFU of each virus at 6 dpi, shown as a percentage of starting weight. Data from animals inoculated with the recombinant parent virus are shown with bolded outlines. (C) Viremia measured at 3 dpi, expressed as genome copies per mL of serum (log₁₀). Dashed line in C represents the lower limit of detection. Error bars show Data represented as mean $\pm$ SEM. Statistical significance was determined by one-way ANOVA with multiple comparisons, with Asibi-backbone chimeras compared to rHA1-Asibi. (*: p < 0.05; **: p < 0.01; ns: not significant.).

Fig 6. The NS2B mutations I37L and I109L cooperate to attenuate rHA7-Asibi.

(A) Heatmap showing the frequency of Asibi/17D mutations across rHA7-Asibi chimeras in which single or double 17D-derived NS2B mutations were introduced into rHA7-Asibi; see description in Fig 2B for additional details. (B) Percent body weight of hamsters inoculated with 1 × 10⁵ FFU of each virus at 5 dpi, shown as a percentage of starting weight. Data from animals inoculated with the recombinant parent virus are shown with bolded outlines. (C) Viremia measured at 3 dpi, expressed as genome copies per mL of serum (log₁₀). Dashed line in C represents the lower limit of detection. Error bars show Data represented as mean $\pm$ SEM. (D) Liver viral load, (E) alanine aminotransferase (ALT), and (F) prothrombin time (PT) measured at 5 dpi. Dashed lines represent the lower and upper limits of measurement range; the gray shaded area represents the reference range for hamsters, calculated from pooled data of mock-infected animals (see Methods for details). Error bars show Data represented as mean $\pm$ SEM. Statistical significance was assessed by one-way ANOVA with multiple comparisons: Asibi-backbone chimeras were compared to rHA7-Asibi in panels B-C, and to HA-Asibi in panels D-F; in B and C, the two rHA7-17D viruses were compared using unpaired t-test (*: p < 0.05; **: p < 0.01; ***: p < 0.001; ****: p < 0.0001; ns: not significant).