Author:

Abstract:

Currently, only a limited number of full genome sequences from clinical human cytomegalovirus (HCMV) strains are available. Additional sequences of clinical isolates are necessary to better apprehend the genetic diversity and coding capacity of clinical HCMV strains. In this study, we developed an amplification, sequencing and analysis pipeline to resolve complete genomes of clinical isolates by using the high-throughput capacity of next-generation sequencing. These methods were validated and subsequently used to analyze the genetic diversity of clinical HCMV strains. After minimal passaging of clinical isolates on human fibroblasts, viral DNA was purified from intracellular virions and amplified via whole genome amplification. Sequence data were generated using a combination of 454 and Illumina sequencing. This combined dataset was used to assemble de novo contigs. After mapping of these contigs on HCMV reference sequences obtained from GenBank, a hybrid reference was built consisting of de novo contigs connected by pieces of the appropriate reference sequence. 454 and Illumina reads were mapped on this hybrid reference and the consensus sequence was manually corrected by visualizing the assembly. Sanger sequencing was used to resolve remaining uncertainties in the mapped sequences. We evaluated these methods using three different approaches. In a first approach, we resequenced the reference strain Merlin. Only two mutations were detected in the 236 kb consensus sequence; a silent mutation in UL32 and one point mutation in the IRL. In addition, one clinical sample was divided into two aliquots, which were independently cell culture passaged, sequenced and assembled. Both consensus sequences were completely identical, except for the lengths of three homopolymer regions. To further evaluate the influence of cell culture adaptation, the sequence of one strain (4941) was obtained both from urine and from a passaged isolate. Consensus sequences were identical apart from one point mutation in UL30 and lengths of three homopolymer regions. These results indicate that the presented workflow had a minimal impact on consensus sequences of clinical strains. Therefore, we applied these methods to identify the full genome consensus sequence for 25 clinical isolates. Four strains did not contain any obvious gene-disrupting mutations in the consensus sequence. Strain 4941 was mutated in RL5A, UL9 and UL150 but these mutations were independent of in vitro passaging as they were also present in the original urine sample. Other gene-disrupting mutations were found in multiple clinical strains isolated from different patients in this study or were shared with previously published HCMV strains, suggesting they could be present in clinical isolates. These included mutations in RL5A, RL6, UL1, UL9, UL111A, UL150 and US9. Although the function of most of these genes during infection is not yet elucidated, UL111A encodes viral interleukin-10, which may play a role in immune regulation. The observed existence of HCMV variants in natural settings might have clinical significance.