HIV drug resistance surveillance using pooled pyrosequencing

Abstract

Background: Surveillance for HIV transmitted drug resistance (TDR) is performed using HIV genotype results from individual specimens. Pyrosequencing, through its massive parallel sequencing ability, can analyze large numbers of specimens simultaneously. Instead of using pyrosequencing conventionally, to sequence a population of viruses within an individual, we interrogated a single combined pool of surveillance specimens to demonstrate that it is possible to determine TDR rates in HIV protease from a population of individuals.

Methodology/principal findings: The protease region from 96 treatment naïve, HIV+ serum specimens was genotyped using standard Sanger sequencing method. The 462 bp protease amplicons from these specimens were pooled in equimolar concentrations and re-sequenced using the GS FLX Titanium system. The nucleotide (NT) and amino acid (AA) differences from the reference sequence, along with TDR mutations, detected by each method were compared. In the protease sequence, there were 212 nucleotide and 81 AA differences found using conventional sequencing and 345 nucleotide and 168 AA differences using pyrosequencing. All nucleotide and amino acid polymorphisms found at frequencies >/=5% in pyrosequencing were detected using both methods with the rates of variation highly correlated. Using Sanger sequencing, two TDR mutations, M46L and I84V, were each detected as mixtures at a frequency of 1.04% (1/96). These same TDR mutations were detected by pyrosequencing with a prevalence of 0.29% and 0.34% respectively. Phylogenetic analysis established that the detected low frequency mutations arose from the same single specimens that were found to contain TDR mutations by Sanger sequencing. Multiple clinical protease DR mutations present at higher frequencies were concordantly identified using both methods.

Conclusions/significance: We show that pyrosequencing pooled surveillance specimens can cost-competitively detect protease TDR mutations when compared with conventional methods. With few modifications, the method described here can be used to determine population rates of TDR in both protease and reverse transcriptase. Furthermore, this pooled pyrosequencing technique may be generalizable to other infectious agents where a survey of DR rates is required.

Figure 1. Concordance of variations detection by pyrosequencing and Sanger sequencing.

The concordance rates were calculated as percentage of pyrosequencing detected sequence variations that were also observed in Sanger sequencing. Frequency ranges are categorized based on those detected by pyrosequencing.

Figure 2. Bidirectional sequence electropherograms for the two TDR mutations detected by conventional Sanger sequencing.

Two TDR mutations, M46L and I84V, were detected by conventional Sanger sequencing, each in one of the 96 examined specimens. Electropherograms demonstrate that both mutations existed as a component of a mixture.

Figure 3. Phylogenetic analysis on pyrosequencing reads with SDRMs detected in bulk, Sanger sequencing.

All positive pyrosequencing reads containing M46L (a) and I84V (b) were analyzed with Sanger sequences from the 96 specimens using Neighbour-Joining (K-2-P) with 100 bootstraps. The pyrosequencing reads with corresponding SDRM are shown in blue with the 96 Sanger sequences shown in red.

Figure 4. Phylogenetic analysis of pyrosequencing reads with SDRMs not detected by bulk Sanger sequencing.

All positive pyrosequencing reads containing M46I (a) and F53L (b) were analyzed with Sanger sequences from the 96 specimens using Neighbour-Joining (K-2-P) with 100 bootstraps. The pyrosequencing reads with corresponding SDRM are shown in blue with the 96 Sanger sequences shown in red. Clusters of pyrosequencing reads are indicated with a green circle.

Figure 5. Consistent and comparable frequency readouts for minor protease DRMs by the two approaches.

Eighteen minor DR mutations (IAS-USA 2008) were detected by either pyrosequencing or Sanger sequencing among the 96 specimens. Individual mutations are plotted against thee frequency detected by each Chart shows the frequency at which the individual mutations were detected by each method.

Figure 6. Phylogenetic analysis on pyrosequencing reads with minor DRMs and Sanger sequences from all 96 pooled specimens.

All positive pyrosequencing reads containing L33V (a) and L33I (b) were analyzed with Sanger sequences from the 96 specimens using Neighbour-Joining (K-2-P) with 100 bootstraps. The pyrosequencing reads with corresponding SDRM are shown in blue with the 96 Sanger sequences shown in red. Clusters of pyrosequencing reads are indicated with a green circle.