Phylogenetic analysis of HIV-1 transmission: pol gene sequences are insufficient to clarify true relationships between patient isolates

Introduction HIV-1 has a high replication rate of about 109 viruses produced each day [1,2]. Because of the relatively high error rate of the HIV-1 reverse transcriptase (RT) of about 1 : 1000 to 1 : 10 000 errors per incorporated base [3,4], there is a high degree of variability between different virus isolates at the genomic level [5]. However, if two isolates have a common source, the degree of variation should be dependent on the time elapsed since the transmission event; at the time of transmission, the virus isolates should have been virtually indistinguishable. This fact has been shown for the env and the gag genes [6,7]. For that reason phylogenetic analyses are useful for the investigation of transmission events. Recent studies investigating transmission events Several studies have shown the usefulness of phylogenetic methods to confirm known or to determine unknown relationships among HIV isolates [8–21]. In nearly all published studies a region in the env gene containing the V3 loop was used for comparison alone [9,18,19] or in combination with the p17 region located in the gag gene [10–12,15,20,21] or the RT region located in the pol gene [13,17]. In one study only the p17 region was analysed [14], in another p17 and RT [8], and in yet another all three regions [16]. For: The pol gene on its own is sufficient for phylogenetic analysis Genotypic HIV-1 resistance testing has become widely accepted as a tool for monitoring antiretroviral therapy and has been implemented into current therapy guidelines [22,23]. Therefore, many laboratories should have a database containing pol gene sequences encompassing the protease (PR) and RT region. A recently published study investigating the possibility of phylogenetic analysis including the PR/RT region alone to reconstruct transmission events came to the conclusion that the genetic information derived from this gene is sufficient [24]. The authors were able to confirm all significant clusters [bootstrap value (BV) ≥ 99] in the PR/RT region when using env and gag genes for verification. Against: The pol gene on its own does not allow phylogenetic analysis During a forensic investigation into alleged HIV-1 transmission events, we tested if a phylogenetic analysis using only the information obtained by pol gene sequences would be sufficient to provide evidence in support of a suspected transmission event, or if information from at least two different genes must be used to achieve this goal. For the phylogenetic analysis of the PR/RT region, two patient samples (02-30434 and 02-35144) linked through HIV-1 transmission were used, together with a set of 20 patient samples selected out of our HIV-1 resistance database containing nearly 3200 pol sequences for having subtype HIV-1B to get a comparable local group of sequence patterns. In addition, 16 subtype B GenBank isolates were included in this analysis (AF042101, AF075719, AF286365, K02007, K02013, K03455, M17449, M17451, M26727, M38429, M38431, U12055, U21135, U43141, U63632, U69593). The following isolates were considered as related: M17449 and AF075719 are isolate MN and a clone derived from this isolate named MNTQ, respectively [25]; K03455, U12055, and K02019 are the isolates HXB2 (HTLV-IIIb reference strain), LW12.3 isolated from a labworker infected by HTLV-IIIb [26], and BRU (formerly LAV-1) which is similar to HTLV-IIIb [27]. The same GenBank isolates were also used in the env analysis together with the two suspected transmission cases and six local patient samples from routine subtyping, four of whom had also clustered in the pol analysis. Viral RNA isolation and generation of sequence data for the RT and PR gene regions of all samples were performed as described elsewhere [28,29]. For analysing the C2V3 region of the env gene, an in-house PCR was performed using primers published recently [30]. Both primers were flanked at their 5′-end with either the M13 universal or reverse primer sequence, respectively, as described elsewhere [31]. The sequencing products were analysed on an ABI 377-96 sequencer (Applied Biosystems). A consensus sequence of the PR/RT region was generated using the HIV-1 Genotyping system software version 2.5, and the SeqMan module of the LaserGene software [32] was used for generating the C2V3 consensus sequence. For the alignment of the PR/RT sequences, all isolates were trimmed to a common range including the complete PR and amino acids 1–293 of the RT corresponding to position 2253–3430 in the HXB2 genome. The C2V3 sequences were trimmed to encompass positions 6889–7407 of HXB2. Sequence alignments were performed using the MegAlign module of the LaserGene software [32] and the Clustal W algorithm [33]. Phylogenetic tree analyses were conducted using the Mega 2 software [34]. Bootstrap tests with 10 000 replications based on the Neighbor-Joining method [35] with the Kimura 2-parameter substitution module [36] were performed to construct the phylogenetic trees. A BV > 70 was considered as supporting the evidence for transmission events [37]. The sequence identity between the two patient samples linked through HIV-1 transmission and another two pairs of patient control samples compared to all subtype B isolates in the C2V3 region were analysed for statistically significant differences using the t test. The phylogenetic analysis of the PR/RT region showed several clusters of isolates with significant BV (Fig. 1). The two samples representing the suspected transmission event (02-30434/02-35144) clustered with BV of 100. In addition, the GenBank isolates MN/MNTQ, and HXB2/LW12.3/BRU clustered with BV of 99, and 89–96, respectively. Two pairs of patient samples out of the local control group (R013/R024 and R004/R016) clustered both with a BV of 99.Fig. 1.: Phylogenetic tree of the PR/RT region. GenBank Isolates are indicated by their subtype followed by accession number, local control group patient isolates by ‘R’ and consecutive numbering, and the two isolates with a known transmission event by laboratory numbers (02-30434/02-35144). Trees were generated by a bootstrap test with 10 000 replications based on Neighbor-Joining method with Kimura two-parameter substitution module. Bootstrap values are shown on the nodes.As shown in the phylogenetic analysis of the C2V3 region, some of the samples that already clustered in the pol analysis formed clusters in the env phylogenetic tree, too (Fig. 2). These were 02-30434/02-35144 with BV of 100; the GenBank isolates MN/MNTQ (BV of 100), and HXB-2/LW12.3/BRU (BV of 93–99). The four patient samples of the control group that clustered in pairs in the pol analysis still clustered; R004/R016 with a BV below the limit of significant evidence for relatedness (BV of 30), and R013/R024 with a BV of 94, supporting a strong relation between these two samples.Fig. 2.: Phylogenetic tree of the C2V3 region. GenBank isolates are indicated by their subtype followed by accession number, local control group patient isolates by ‘R’ and consecutive numbering, and the two isolates with a known transmission event by laboratory numbers (02-30434/02-35144). Trees were generated by a bootstrap test with 10 000 replications based on Neighbor-Joining method with Kimura two-parameter substitution module. Bootstrap values are shown on the nodes.To confirm these findings further, a statistical analysis was performed comparing the C2V3 sequence identity between all subtype B isolates used in the phylogenetic analyses with the C2V3 sequence identity of the patient sample pairs 02-30434/02-35144 (confirmed relation), R013/R024 (possible relation indicated by BV), and R004/R016 (unlikely relation indicated by BV) using the t test. The mean C2V3 sequence identity of all subtype B isolates was 88.4% compared to 98.0% for 02-30434/02-35144, 89.1% for R013/R024, and 88.4% for R004/R016. The C2V3 sequence identities of 02-30434/02-35144, and R013/R024 were significantly different from the complete subtype B group (P < 0.001 and P = 0.022, respectively), whereas the C2V3 sequence identity of R004/R016 was not (P = 0.996). Because all statistical parameters supported a true relation between the patient samples R013/R024, both patients were interrogated independently. Unaware of each other also being questioned, they admitted to having had a sexual affair about 3 years ago, and one of them confirmed that from this time point he was HIV-1 positive. Which is the best way to reconstruct transmission events? In most of the studies investigating possible transmission events of HIV-1 [8–21], at least two regions from different genes of HIV-1 were analysed to strengthen the validity of the results. However, in four studies only one gene region was used to investigate molecular relationships: C2V3 on its own in three [9,18,19], and P17 only in one case [14]. Because of the widespread use of genotypic drug resistance testing, a huge amount of sequence data should be available for the PR/RT region located in the pol gene. Therefore, pol has become an attractive target for phylogenetic studies: One recent study investigated possible transmission events initially using this gene [24] and could confirm all significant clusters (BV ≥ 99), using the env and gag genes for verification. However, we cannot support this finding when analysing our data: While the expected relationships between the GenBank isolates were supported by the phylogenetic analysis of the PR/RT region (Fig. 1), three clusters of patient samples could be identified; of these, only one had been expected, while the others had been selected randomly out of our sequence database comprising about 3200 pol sequences. When trying to verify these results by analysing the C2V3 region, the same clusters of GenBank samples as described for the PR/RT analysis were found (Fig. 2). Again, the patient samples 02-30434/02-35144 clustered as expected. For one of the other two pairs (R04/R016) the phylogenetic analysis of the second region did not support a strong relationship on the basis of a BV of 30. In contrast, the patient samples R013/R024 still clustered with a significant BV of 94. Additional statistical analysis of the sequence homology using the t test further confirmed this phylogenetic tree construction. An independent interrogation of both patients that clustered unexpectedly was carried out to review the background of how they became infected with HIV-1 and revealed a sexual relationship about 3 years ago. By his own account, one of the two patients probably acquired HIV infection during this time. Conclusions In our opinion the PR/RT region cannot be used on its own to prove true relationships between unknown patient isolates. The discovery of a previously unknown sexual transmission clearly demonstrates the ability of phylogenetic analysis including both the PR/RT and the C2V3 regions to detect and confirm HIV-1 transmission events. However, based on PR/RT analysis alone, another transmission event would have been assumed erroneously. If analysis by a single gene region is to be used, we would advise use of the C2V3 region preferentially.