Stability evolution as a major mechanism of human protein adaptation in response to viruses

Abstract

Pathogens were a major driver of genetic adaptation during human evolution. Viruses in particular were a dominant driver of adaptation in the thousands of proteins that physically interact with viruses (VIPs for Virus-Interacting Proteins). This however poses a conundrum. The best understood cases of virus-driven adaptation in specialized immune antiviral factors or in host viral receptors are numerically vastly insufficient to explain abundant adaptations in VIPs. What adaptive mechanisms can then at least partly close this gap? VIPs tend to be broadly conserved proteins with conserved host native molecular functions. Because many amino acid changes in a protein can alter its stability --the balance between the folded and unfolded forms of a protein-- without destroying conserved native activities, here we ask if stability evolution was an important mechanism of virus-driven human protein adaptation. Using predictions of protein stability changes based on Alphafold 2 structures and validated by multiple lines of evidence, we find that amino acid changes that altered stability experienced highly elevated adaptative evolution in VIPs, compared to changes with a weaker impact on stability. We further find that RNA viruses, rather DNA viruses, predominantly drove strong adaptation through stability changes in VIPs. We also observe that stability in immune antiviral VIPs preferentially evolved under directional selection. Conversely, stability in proviral VIPs needed by viruses evolved under compensatory evolution following viral epidemics. Together, these results suggest that stability evolution, and thus functional host protein abundance evolution, was a prominent mechanism of host protein adaptation during viral epidemics.

Mega-scale experimental analysis of protein folding stability in biology and protein design

“…Protein sequences vary by more than ten orders of magnitude in thermodynamic folding stability (the ratio of unfolded to folded molecules at equilibrium) (1,2). Even single point mutations that alter stability can have profound effects on health and disease (3)(4)(5), pharmaceutical development (6)(7)(8), and protein evolution (9)(10)(11)(12)(13). Thousands of point mutants have been individually studied over decades to quantify the determinants of stability (14,15), but these studies highlight a challenge: similar mutations can have widely varying effects in different protein contexts, and these subtleties remain difficult to predict despite substantial effort (16,17).…”

Quantifying Adaptive Evolution of the Human Immune Cell Landscape

“…To measure a significant increase in adaptation, we build a null distribution accounting for 1,000 control datasets matching the size of the analysed immune genes subset for each cell population, following a bootstrapping approach [31,32] . The procedure allowed us to match, resample and compare genes with similar overall average values for many confounding factors that could affect the adaptation rate and mislead the comparison (see Methods for further information) [31,33,34] . We evaluated each cell population estimation in accordance with its corresponding null empirical distribution.…”

“…We measured the long-term rate of protein adaptation using ABC-MK, an Approximate Bayesian Computation extension of the MK-test[22,23,34]. Such a rate is usually estimated as the proportion of adaptive non-synonymous substitution and can be derived from classical MK-test by the quantity α [14] where D S and D N are synonymous and non-synonymous fixed differences, and P S and P N are synonymous and non-synonymous polymorphic sites.…”

“…This list contains Ensembl v99 human genes with orthologs in a minimum of 251 out of 261 mammalian genome assemblies. As exposed in Di et al [34] , the assemblies were extracted from NCBI Genome and had a minimum N50 contig size of 30kb to reduce the number of possible truncated genes. The orthologs for mammals were defined based on the best-reciprocal hits using Blat [79] alignments, resulting in 13,495 orthologs.…”

“…Quantifying adaptation using ABC-MK. We measured the long-term rate of protein adaptation using ABC-MK, an Approximate Bayesian Computation extension of the MKtest [22,23,34] . Such a rate is usually estimated as the proportion of adaptive non-synonymous substitution and can be derived from classical MK-test by the quantity α [14] (1)…”