Comparative analysis of deep mutational scanning datasets in enteroviruses A and B identifies functional divergence and therapeutic targets

Figures

• 

  Fig. 1. Overview of viral systems and DMS datasets. a, Schematic of the enterovirus genome. UTR, untranslated region. b, Per-protein similarity at the AA level between EVA71 and CVB3 (n = 325 and n = 49, respectively). c,d, Phylogenetic trees of all known enterovirus species, illustrating sequence divergence across structural coding regions (c) and the 3D polymerase coding region. Nodes are shaded by bootstrap support levels (BS) shown in legend (d). The tip labels indicate enterovirus species. e, Heat maps of site MFEs across EVA71 and CVB3 proteins with line plots showing a five-residue sliding window average. f, Correlation of mean site MFE between EVA71 and CVB3 across proteins. The error bars indicate the standard error of the mean across all positions of each protein. The dashed line represents a 1:1 relationship between both axes. g, Genomic distribution of tolerated deletions unique to EVA71 or CVB3 or shared between the two viruses (common). h, Correlation of mean MFE based on the nature of the WT residue across AA physicochemical categories between the EVA71 and CVB3 DMS datasets. i, Correlation of mean MFE based on the nature of the mutation across AA physicochemical…

• 

  Fig. 2. Structural mapping of enterovirus A and B differential mutational tolerance. a, Site difference in MFE scores (MFE in CVB3 minus MFE in EVA71) across the aligned viral proteomes, averaged using a ten-residue sliding window. Positive values indicate higher mutational tolerance in CVB3, while negative values indicate higher tolerance in EVA71. A horizontal green line at y = 0 denotes similar tolerance across both viruses. b, Proportion of divergent sites for each viral protein. Divergent sites are defined as those with an absolute site MFE difference greater than the median plus two median absolute deviations (threshold, 0.86). Proportions are calculated relative to the total number of aligned residues per protein. c, Distribution of absolute site MFE difference for sites located inside versus outside active sites and capsid assembly interfaces. In each box plot, the lower and upper hinges correspond to the 25th and 75th percentiles, respectively, with the centre line indicating the median. The whiskers extend to the most extreme data points within 1.5× the interquartile range (IQR) from the hinges. Observations falling outside this range are treated as outliers and are plotted…

• 

  Fig. 3. Divergence in mutational tolerance at the EVA71 and CVB3 receptor-binding footprints. a, Distribution of absolute site MFE difference for sites located inside versus outside receptor footprints. In each box plot, the lower and upper hinges correspond to the 25th and 75th percentiles, respectively, with the centre line indicating the median. The whiskers extend to the most extreme data points within 1.5× the IQR from the hinges. Observations falling outside this range are treated as outliers and are plotted individually. Statistical significance was calculated using a two-sided Wilcoxon test. b, The total number of analysed sites alongside the count of the most divergent sites (top 1% by MFE difference), categorized by their location inside or outside the receptor-binding interface. Statistical significance was calculated using a two-sided Fisher’s exact test. c,d, Heat maps displaying the EVA71 (c) and CVB3 (d) MFE scores for divergent residues in receptor contact sites. The ‘X’ symbols indicate WT AA positions, and grey squares represent missing data. e–m, Surface representations of EVA71 in complex with its receptor SCARB2 (PDB: 6I2K) (e–g) and CVB3 with its receptors CAR (PDB:…

• 

  Fig. 4. Mapping of the divergent structural and functional regions in enterovirus 3A. a, Site difference in MFE scores (MFE in CVB3 minus MFE in EVA71) across the aligned 3A proteins, averaged using a ten-residue sliding window. b,c, AlphaFold3-predicted structures of the EVA71 (b) and CVB3 (c) 3A homodimers. Residues are coloured by site MFE difference according to the colour key. d,e, AlphaFold3-predicted structural models of the EVA71 (d) and CVB3 (e) 3A dimers in complex with the host factor GBF1. 3A residues are coloured by site MFE difference according to the colour key, and GBF1 is shown as a grey surface, with its 3A contact residues highlighted in yellow. EVA71 3A–GBF1 prediction: pTM = 0.60, ipTM = 0.20, ranking score = 0.45; CVB3 3A–GBF1 prediction: pTM = 0.59, ipTM = 0.19, ranking score = 0.44. f,g, Heat maps displaying MFE scores for the specific set of residues that are mutationally divergent and that constitute the EVA71 3A–GBF1 binding interface. Data are shown for EVA71 (f) and CVB3 (g). Each ‘X’ indicates the WT AA position. For all relevant panels, the predicted protein topology (cytoplasm, membrane, lumen), as determined by DeepTMHMM, is indicated. For all structural…

• 

  Fig. 5. Comparison of laboratory and natural selection pressures. a, Workflow for the comparison of natural diversity and DMS-derived MFEs. Sequence alignments were used to obtain a site-level diversity measure (Shannon entropy) and AA preferences. These metrics were compared with site MFEs and AA preferences derived from the DMS datasets. Site-level analyses were performed using Pearson correlation tests, and differential AA preferences between nature and laboratory were obtained using phydms. b, Correlation between the experimental DMS datasets and natural sequence diversity at the species and type levels, calculated separately for different functional regions (capsid, replication or whole proteome). All Pearson correlations shown are two-sided and statistically significant (P < 0.05). c, Differential AA preferences per protein between the experimental DMS datasets and natural sequence diversity for CVB3 and EVA71 at the species and type levels. The error bars indicate the standard error of the mean across all positions in each protein. Statistical significance was calculated using a two-sided Wilcoxon test across all positions in each viral proteome (species versus type, P < 2.2 ×…

• 

  Fig. 6. Identification of a conserved druggable pocket in the enterovirus 2C helicase. a, Workflow for identifying and prioritizing druggable pockets, combining computational druggable pocket screening with experimental determination of mutation tolerance from DMS studies. b, The total number of druggable pockets detected in the EVA71 and CVB3 proteomes using SiteMap. c, Box plots comparing the average MFEs for residues within conserved druggable pockets (fraction of shared residues >0.5, n ≥ 12 residues). In each box plot, the horizontal line indicates the median, the box shows the IQR, the whiskers extend to values within 1.5× the IQR and outliers are plotted individually. Lower MFE scores indicate higher mutational constraint. d, Surface representation of the AlphaFold3-predicted EVA71 2C helicase monomer. The identified pan-enterovirus druggable pocket (green) is shown in relation to the ATPase domain (pink), which includes the Walker A, B and C motifs. e, Surface representation of the AlphaFold3-predicted EVA71 2C helicase in its hexameric form, shown from the cytoplasmic, membrane and side views. f, Multi-species structural comparison of the 2C pocket. The surface of the…