In this study, we assessed the feasibility of using 454 pyrosequencing to study quasispecies of the whole genome of dengue virus. We performed two sequencing runs with templates prepared from different sized PCR products. In the first run, we amplified the dengue genome as five fragments, with approximate sizes of 2.5 kb. The 2.5 kb product size gave a good yield and required only a few reactions to meet the amount needed for template preparation. However, there was an issue when performing DNA fragmentation using a nebulizer on small 2.5 kb DNA fragments. The standard protocol for library preparation did not work well. The sizes of the randomly fragmented DNA templates were still bigger than those recommended for sequencing. DNA fragments of 2.5 kb were too short for the standard protocol. Therefore, the nitrogen gas pressure in the nebulization step was changed to 45 psi for 1 minute 30 seconds. This modified protocol generated random DNA fragments with sizes in the recommended range, but the high Nitrogen gas pressure caused a large amount of DNA fragment loss, as previously reported . Another issue with the 2.5 kb DNA fragments was the fluctuation of coverage depth. Technical issues, such as DNA fragmentation of small PCR products and human errors when pooling many fragments together, could contribute to the fluctuation in coverage. To alleviate the problem, we reduced the number of templates from five to two. As a result, the coverage of data 2 fluctuated less than data 1 (Figure 2).
We found some reads where one part could be mapped to one serotype and the other part to another serotype, suggesting recombination. Natural recombination was reported in DENV1 strains ; however, only evidence for intra-serotype recombination among dengue viruses has been reported, not for inter-serotype recombination . In our study, we used dengue virus prototype strains from cell cultures. It was unlikely that inter-serotype recombination could have occurred. We hypothesize that this was an artifact and could lead to misinterpretation of sequencing pools of clinical samples. These sequences could have occurred from the self-ligation of small nebulized DNA fragments during the step of adding adaptor A/B. According to the manual provided by Roche Applied Science, the adaptor will ligate to a blunt ended DNA fragment. Therefore, blunt end DNA templates themselves may become ligated to each other before the adaptor ligates to both of their ends.
Recently, many viral quasispecies studies used next-generation sequencing (NGS) for study of viral diversity, for example HIV, hepatitis B virus, hepatitis C virus, and influenza H5N1 virus. However, most studies were done on specific genes or regions, not the whole genome. For the HIV studies, 454 pyrosequencing was used to identify rare variation at low frequency, usually focusing on drug resistance mutations [9, 17]. Only specific regions or genes were used for sequencing with NGS using Amplicon sequencing  and Shotgun sequencing with DNA fragment size ~1.5 kb . The study of hepatitis B virus also use 454 pyrosequencing to detect drug resistant mutations in specific genes, using amplicon sequencing with a multiplex identifier sequence . Samples for the study were collected from HBV patients' plasma. Moreover, pyrosequencing was used to identify intra-host hepatitis C variation using amplicon sequencing on the hypervariable region . For the influenza H5N1 study, the whole genome of influenza H5N1 was sequencing by GS FLX, with the goal of developing a diagnostic system for H5N1. For the influenza genome, a single stranded shotgun library protocol was used for library construction . Those studies used PCR amplicons as templates; however, amplifying the whole genome to get PCR products suitable for 454 sequencing could be tedious, even for a small genome such as dengue virus. This study demonstrated that NGS could be used for studying quasispecies of dengue virus and validated the sensitivity of detection using simulated quasispecies.