Background: Haplotyping of HIV-1 populations is an essential step to better understand the evolutionary dynamics of the virus. With the advent of next-generation sequencing (NGS), haplotyping of viral populations has become feasible. Since HIV-1 is highly heterogeneous, several statistical methods have been devised to deal with error-prone NGS data, however, they often do not capture the population correctly. In order to correct for errors, the use of PrimerIDs (primer identifiers) has been proposed. Here, we used PrimerIDs to systematically estimate different enzymatic error rates and to comprehensively study the feasibility of PrimerIDs.
Methods: Plasmids containing full-length genomes of 5 HIV-1 clones were separately amplified in bacteria and then transfected into 293T cells. Generated infectious HIV-1 particles were pooled, DNase treated, and a fragment of the pol gene was reverse transcribed with SuperScript III reverse transcriptase (RT) and primers containing random 10-mers. Reverse transcription was performed in six independent replicates. Subsequently, nested PCR was performed using Platinum Taq DNA Polymerase followed by adapter ligation and sequencing with Illumina MiSeq.
Results: From an average number of 1.1 million reads, we called consensus sequences for PrimerIDs, each supported by at least 10 sequencing reads, to yield on average 11,000 consensus sequences per replicate. From these consensus sequences, we could call all mutant bases from the five reference viruses. We estimated a RT error rate of 6.23e-4 (95% CI: [6.13e-4, 6.32e-4]). We inferred the recombination rate of the RT to be 3.44e-5 (95% CI: [2.26e-5, 4.92e-5]). The PCR substitution rate of 1.18e-4 (95% CI: [1.14e-4, 1.22e-4]) was determined from those mutants having arisen in the first cycle of the PCR. We calculated the total number of transcribed RNAs to be on the order of 60,000 from the observed collision rate of 2%. We observed no sequence-specific bias in PrimerID frequencies, the same RT efficiencies as compared to commonly used short, specific RT primers, and no effects of primerIDs on the estimated distribution of the five viruses in the mix.
Conclusions: PrimerIDs allow for determining error rates in RT-PCR-NGS protocols and are applicable to study HIV-1 heterogeneity when attention is paid to collision rates. Given these advantages, the protocol is still labor- and cost-intensive and does not significantly improve on the variance of frequency estimates.