ViroFind: A novel target-enrichment deep-sequencing platform reveals a complex JC virus population in the brain of PML patients

Abstract

Deep nucleotide sequencing enables the unbiased, broad-spectrum detection of viruses in clinical samples without requiring an a priori hypothesis for the source of infection. However, its use in clinical research applications is limited by low cost-effectiveness given that most of the sequencing information from clinical samples is related to the human genome, which renders the analysis of viral genomes challenging. To overcome this limitation we developed ViroFind, an in-solution target-enrichment platform for virus detection and discovery in clinical samples. ViroFind comprises 165,433 viral probes that cover the genomes of 535 selected DNA and RNA viruses that infect humans or could cause zoonosis. The ViroFind probes are used in a hybridization reaction to enrich viral sequences and therefore enhance the detection of viral genomes via deep sequencing. We used ViroFind to detect and analyze all viral populations in the brain of 5 patients with progressive multifocal leukoencephalopathy (PML) and of 18 control subjects with no known neurological disease. Compared to direct deep sequencing, by using ViroFind we enriched viral sequences present in the clinical samples up to 127-fold. We discovered highly complex polyoma virus JC populations in the PML brain samples with a remarkable degree of genetic divergence among the JC virus variants of each PML brain sample. Specifically for the viral capsid protein VP1 gene, we identified 24 single nucleotide substitutions, 12 of which were associated with amino acid changes. The most frequent (4 of 5 samples, 80%) amino acid change was D66H, which is associated with enhanced tissue tropism, and hence likely a viral fitness advantage, compared to other variants. Lastly, we also detected sparse JC virus sequences in 10 of 18 (55.5%) of control samples and sparse human herpes virus 6B (HHV6B) sequences in the brain of 11 of 18 (61.1%) control subjects. In sum, ViroFind enabled the in-depth analysis of all viral genomes in PML and control brain samples and allowed us to demonstrate a high degree of JC virus genetic divergence in vivo that has been previously underappreciated. ViroFind can be used to investigate the structure of the virome with unprecedented depth in health and disease state.

Fig 1. Coverage plots of JC virus genome in brain samples from PML patients.

Coverage plots of JC virus genome in brain samples from 5 PML patients. The figure shows the number of times each nucleotide of the JC virus genome was read during the sequencing process (coverage). By using ViroFind we achieved high coverage of the JC virus genome, which is required for the reliable identification of all JC virus variants.

Fig 2. Multiple JC virus variants within the same PML brain sample.

Visual inspection of the JC virus 75 base-pair nucleotide reads reveals multiple single nucleotide polymorphisms (SNP). Each read is represented by a single horizontal arrow. SNPs in positions 3,122, 3,178 and 3,185 of the JC virus genome occur independently, which is consistent with the presence of at least three distinct viral variants in the same PML brain.

Fig 3. Genetic distance between the JC virus sequences of each brain sample and MAD-1 sequence.

The figure illustrates the genetic divergence between the JC virus sequences found in each PML brain and MAD-1 (red bar), the JC virus prototype sequence associated with PML. Each chord corresponds to a single nucleotide polymorphism. The presence of multiple JCV variants within each PML brain accounts for a remarkable degree of genetic variation and suggests that the genetic complexity of JC virus in the brain compartment has previously been underestimated.