Hypermutability of genes in Homo sapiens due to the hosting of long mono-SSR

Simple sequence repeats (SSRs) are very common short repeatsin eukaryotic genomes. "Long" SSRs are considered "hypermutable"sequences because they exhibit a high rate of expansion andcontraction. Because they are potentially deleterious, longSSRs tend to be uncommon in coding sequences. However, severalgenes contain long SSRs in their exonic sequences. Here, weidentify 1,291 human genes that host a mononucleotide SSR longenough to be prone to expansion or contraction, being calledhypermutable hereafter. On the basis of Gene Ontology annotations,we show that only a restricted number of functions are overrepresentedamong those hypermutable genes including cell cycle and maintenanceof DNA integrity. Using a probabilistic model, we show thatgenes involved in these functions are expected to host longSSRs because they tend to be long and/or are biased in nucleotidecomposition. Finally, we show that for almost all functionswe observe fewer hypermutable sequences than expected undera neutral model. There are however interesting exceptions, forexample, genes involved in protein and RNA transport, as wellas meiosis and mismatch repair functions that have as many hypermutablegenes as expected under neutrality. Conversely, there are functions(e.g., collagen-related genes) where hypermutable genes aremore often avoided than in other functions. Our results showthat, even though several functions harbor unusually long SSRin their exons, long SSRs are deleterious sequences in almostall functions and are removed by purifying selection. The strengthof this purifying selection however greatly varies from functionto function. We discuss possible explanations for this intriguingresult.