Translational adaptation of human viruses to the tissues they infect

Xavier Hernandez-Alias; Martin Schaefer; Luis Serrano

Although high-throughput sequencing of tRNAs has been only recently developed, cases of natural selection of codon usage towards the host have been previously proposed. For instance, codon usage of Parvovirus has been progressively adapted from dogs to cats after the host jump 43 . Influenzaviruses show a similar adaptation over time of viral isolation, deviating from the codon usage of avian hosts 44, 45 . However, whether these progressive changes in codon usage over time are directly driven by translational selection has remained elusive. With the advent of tissue-wide datasets of tRNAs and their translational efficiencies 2 , we can now compute the Supply-to-Demand Adaptation (SDA) of all viral proteomes in different tissues. From there, we then created a random forest model that predicts with high accuracy the viral tropism of proteins based on their profile of adaptation to human tissues. In consequence, the tRNA-based adaptation profile of a protein is descriptive of their viral tropism, indicating that translational selection could indeed drive tropism differences of codon usage. It is important to remark that viruses could still have a good SDA to non-target tissues with similar tRNA expression patterns that are not infected because they are not exposed to the virus.