Mutational robustness of hepatitis C virus intra-host variants

Summary form only given. Neutrality theory states that the majority of genotypic changes in evolution is selectively neutral, which has been confirmed in the analysis of neutral networks in genotype-phenotype models of RNA secondary structure and protein folding. These analyses have shown that collections of neutral genotypes, which are connected via single mutational steps, form extended networks that permeate large regions of genotype space. An interesting consequence is that a population does not move randomly over a neutral network but tends to concentrate in those regions of the network that are highly connected and, therefore, more robust to mutations. Here, we study experimentally obtained samples of intra-host HCV variants and show that the properties of their neutral networks show high mutational robustness. Intra-host sequences of the HCV E1/E2 region were extensively sampled from 16 chronically infected patients using a Next-Generation sequencing (NGS) technology. After error correction was performed, a network was created for each sample of sequences, where a node is a unique HCV sequence and two nodes are connected by an edge if their nucleotide distance is exactly one. Based on these networks, several neutrality measures were calculated. The average number of unique sequences obtained from each patient was 747.93, with an average of 6091.18 reads. Over all patients, the average frequency of the major is 0.2916. All sequences were used to generate one-step networks. There are some with a frequency higher than 5%, which were called big networks. In average a patient has 2.625 big networks, which account for 92.83% of all reads in a patient. Over all big networks, the average number of sequences and reads is 214.05 and 2157.78, respectively. For each sample, population neutrality was calculated as the spectral radius of the adjacency matrix. The average population neutrality over all samples was 5.6347 (ranging from 3.3685 to 9.5765 in all samples), whic- can be understood as the average number of point mutations that leave the phenotype unaltered. If the degree distribution is highly uniform over the network, network neutrality is equal to the average degree. In this set of samples the average network neutrality was 3.4110. When the population neutrality is greater than network neutrality, their difference precisely quantifies mutational robustness or the extent to which a population occupies the most connected areas of the neutral network. In this case, population neutrality was greater than network neutrality, yielding a mutational robustness of 0.6537 (ranging from 0.0964 to 1.6623 in all samples), and showing that under selection and mutation, the population evolved a mutational robustness that is 65.3% higher than if it were to spread randomly over the network. We took advantage of the capacity of NGS to produce a massive number of sequences, facilitating the accurate assessment of frequency of viral variants. If the population tends to concentrate in those regions of the network that are highly connected then the high centrality of a sequence should be associated with the high frequency in the population. Accordingly, we found that the correlation between the eigenvector centrality of the nodes in each network was highly and significantly associated with frequency, with an average Pearson correlation coefficient of 0.8311 (p = 0.0001). In conclusion, we have shown a strong mutational robustness of the one-step network of intra-host HCV variants, indicating that, even under conditions of similar fitness, the HCV population has a tendency to concentrate in areas of sequence space where mutations have a higher chance to be neutral and leave the phenotype unaltered.