Comparison of the live attenuated yellow fever vaccine 17D-204 strain to its virulent parental strain Asibi by deep sequencing

Abstract

Background: The first comparison of a live RNA viral vaccine strain to its wild-type parental strain by deep sequencing is presented using as a model the yellow fever virus (YFV) live vaccine strain 17D-204 and its wild-type parental strain, Asibi.

Methods: The YFV 17D-204 vaccine genome was compared to that of the parental strain Asibi by massively parallel methods. Variability was compared on multiple scales of the viral genomes. A modeled exploration of small-frequency variants was performed to reconstruct plausible regions of mutational plasticity.

Results: Overt quasispecies diversity is a feature of the parental strain, whereas the live vaccine strain lacks diversity according to multiple independent measurements. A lack of attenuating mutations in the Asibi population relative to that of 17D-204 was observed, demonstrating that the vaccine strain was derived by discrete mutation of Asibi and not by selection of genomes in the wild-type population.

Conclusions: Relative quasispecies structure is a plausible correlate of attenuation for live viral vaccines. Analyses such as these of attenuated viruses improve our understanding of the molecular basis of vaccine attenuation and provide critical information on the stability of live vaccines and the risk of reversion to virulence.

Keywords: 17D, Viral Population; Deep Sequencing; Flavivirus; Parallel Sequencing; Quasispecies; Vaccine; Variants; Yellow Fever Virus.

Figure 1.

Comparison of diversity indices for yellow fever virus Asibi and 17D-204 strains, at multiple scales. All statistical comparisons were performed using the Mann–Whitney U test; 2-tailed P values of <.05 were considered statistically significant. Downsampled Asibi read sets were averaged and floor-rounded by counts. A, Graph of complete Shannon entropy and Simpson's 1-D across both genomes, with consensus vaccine mutation positions for reference (black dots). B, Box plots of diversity indices computed for every nucleotide position along the entire genomes for Asibi and 17D-204 strains, and compared. C, Box plots of diversity indices for the vaccine mutations sites observed by Hahn et al [13], and compared. All box plots depict the median, first quartile, third quartile, with whiskers showing ± 1.5 times the interquartile distance. *P < .05. Abbreviation: NS, not significant.

Figure 2.

Quantile-quantile comparisons of error rate along the multiple genome scales for Asibi and 17D-204 strains. Downsampled Asibi read sets were averaged and floor-rounded by counts. Error rates were calculated as frequency of base calls at each site not corresponding to the consensus base call of the alignment. Curve shapes show that a greater number of high outlying errors is present in the Asibi strain alignment, compared with that of 17D-204. This pattern is evident for both the entire genome lengths and at the scale of specific genes. A, Full genome. B, Local comparisons along gene segments. Abbreviation: UTR, untranslated region.

Figure 3.

Intersection of variant identity between Asibi and 17D-204 strain quasispecies is unidirectional. Asibi frequencies are averaged for downsampled read sets. A, Bar plot showing sites of overt quasispecies identity at sites that are fixed in the 17D-204 strain. B, Box plot summary of intersection frequency for Asibi and 17D-204 strains, depicting the frequency of sites bearing the quasispecies identity of the opposing viral read set. By frequency, Asibi nucleotide identity intersects that of 17D-204 to a greater extent than the converse.

Figure 4.

Scatterplots of change in nucleotide frequency for paired-samples comparison. Plot depicts the set of mutations recovered from a binomial model, using the Asibi strain alignment as a control, and 17D-204 as a test case. Filled polygons represent mutants (152) that were recovered concordantly from all downsampled Asibi strain read sets, at P < .05. Mutations are classified, relative to the Asibi (control) strain genome, as coding for amino acid substitutions, as silent, and as silent from four-way degenerate sites. The pattern of significant frequency changes shows fixation in the 17D-204 genome relative to the parental strain Asibi.

Figure 5.

Paired-test model recovers patterns of local plasticity in the yellow fever virus (YFV) genome. A, Manhattan plot depicting all possible variant frequency changes in the modeled transition from YFV Asibi strain to 17D-204. Polygons represent significant frequency changes, expressed as log[1/P value], following Bonferroni correction. B, Line plot of a sliding-window mean of the most significant variant recovered from each position in the paired-test model. The graph permits visualization of mutationally diverse regions of the YFV genome under the selected model, in comparison to the expected mutation set (black points). C, Venn diagrams showing intersection of recovered variant populations with other models of YFV attenuation; this is separately performed for consensus sequences of parent and derivative strains for HeLa p6 virus and the French neurotropic vaccine [27, 28]. Abbreviations: FNV, French neurotropic vaccine; SNP, single-nucleotide polymorphism.