Coordinate linkage of HIV evolution reveals regions of immunological vulnerability

Abstract

Cellular immune control of HIV is mediated, in part, by induction of single amino acid mutations that reduce viral fitness, but compensatory mutations limit this effect. Here, we sought to determine if higher order constraints on viral evolution exist, because some coordinately linked combinations of mutations may hurt viability. Immune targeting of multiple sites in such a multidimensionally conserved region might render the virus particularly vulnerable, because viable escape pathways would be greatly restricted. We analyzed available HIV sequences using a method from physics to reveal distinct groups of amino acids whose mutations are collectively coordinated ("HIV sectors"). From the standpoint of mutations at individual sites, one such group in Gag is as conserved as other collectively coevolving groups of sites in Gag. However, it exhibits higher order conservation indicating constraints on the viability of viral strains with multiple mutations. Mapping amino acids from this group onto protein structures shows that combined mutations likely destabilize multiprotein structural interactions critical for viral function. Persons who durably control HIV without medications preferentially target the sector in Gag predicted to be most vulnerable. By sequencing circulating viruses from these individuals, we find that individual mutations occur with similar frequency in this sector as in other targeted Gag sectors. However, multiple mutations within this sector are very rare, indicating previously unrecognized multidimensional constraints on HIV evolution. Targeting such regions with higher order evolutionary constraints provides a novel approach to immunogen design for a vaccine against HIV and other rapidly mutating viruses.

Fig. 1.

Defining collectively coevolving groups of sites (sectors) in Gag. (A, Upper) Number of eigenvalues (ordinate) of the correlation matrix, C, for Gag (defined in the main text) with a given magnitude (abscissa) is shown. In total, 1,600 sequences of Gag polyproteins (500 residues long) were used. Note that a relatively large number of sequences of HIV proteins are available. (Lower) Distribution of eigenvalues obtained from 1,000 randomly generated matrices of the same size as C for Gag (described in the main text and SI Appendix 2). (B) Analyzing the eigenvectors corresponding to eigenvalues larger than the highest eigenvalue for random matrices yields collectively coevolving groups of sites or “sectors” (SI Appendix 3–7). Sites within a sector are grouped together along the rows and columns so that groups of collectively coevolving sites are vivid as squares along the diagonal of a heat map representing the values of the correlations (obtained from the “cleaned” correlation matrix as described in SI Appendix 2). (C) Mean frequency of the dominant amino acid at single sites within each of the five Gag sectors. (D–F) Threshold value defining a significant correlation was chosen to be such that correlations with magnitudes greater than this value arise with vanishing probability in the randomized matrices (P < 0.02). Changing this threshold value does not change qualitative results (SI Appendix, Fig. S11). (D) Percentage of significant negative (C_ij < −0.03) pair correlations within each sector. (E) Percentage of significant positive (C_ij > 0.1) pair correlations within each sector. (F) Percentage of significant three-site negative (C_ijk < −0.01) correlations within each sector (method in SI Appendix 17). (G) Percentage of significant three-site positive (C_ijk > 0.1) correlations within each sector.

Fig. 2.

Protein structures reveal the origin of collective correlations. (A) Sector 3 sites represented on the structure of the p24 monomer (PDB ID code 3GV2) are shown as purple spheres. (B) Sector 3 sites represented on the structure of the p24 hexamer (PDB code 3GV2) and the structure of the interface between hexamers (PDB code 2KOD). Sites at interfaces between two p24 proteins belonging to two adjacent hexamers are shown in green, and sites at interfaces between two p24 molecules within a hexamer are shown in red. The few remaining sites in sector 3 that are not part of these interfaces are shown in purple. (C) Sector 1 sites are shown in cyan on the structure of the p24 hexamer.

Fig. 3.

Concentration of sites from the dominant epitopes presented by elite controllers in each sector and comparison of mutation patterns in sectors 1 and 3 in viruses derived from elite controllers. (A) Epitope is defined as dominant if it is the most targeted (in Gag) by individuals with the corresponding HLA allele. Concentration is the number of epitope sites in the sector divided by the total number of residues in the sector. The P value of association with each sector is in SI Appendix 10. Only sector 3 is significantly enriched with sites from the dominant epitopes of HLA molecules associated with control. Sectors 4 and 5 are not targeted at multiple points because they contain only 10 and 14 sites, respectively. Sectors 1 and 3 are targeted at multiple points. (B) Comparison of the number of unique viral strains obtained from a cohort of elite controllers that contains different numbers of mutations in sectors 1 and 3.