Response to ‘Single phylogenetic reconstruction method is insufficient to clarify relationships between patient isolates in HIV-1 transmission case’ by Jenwitheesuk and Liu

Abstract

Jenwitheesuk and Liu criticized the procedure we used for phylogenetic tree construction to analyse true relationships between patient HIV-1 isolates [1]. Although they agreed with our statement that the HIV-1 pol gene (protease and first 239 amino acids of the reverse transcriptase) is insufficient to clarify true relationships, they recommended using at least two different methods for phylogenetic analysis, as done in other studies [2–6]. In addition, they claimed that our conclusion would have been different if other methods were used. The only study published about phylogenetic analysis in forensic investigations of HIV-1 transmission cases in Germany used only one phylogenetic method, and the results were accepted by the German court [7,8]. We therefore performed the same procedure to analyse our data [1]. However, this does not mean that we insist on using only one phylogenetic method for analysing transmission events if there is sufficient evidence for the false classification of samples. As suggested by Jenwitheesuk and Liu, we reanalysed our data using the following different phylogenetic methods with different seed numbers: neighbour-joining, minimum evolution, and unweighted pair group method using arithmetic averages with both the Kimura-2-parameter and the Jukes–Kantor substitution module, and maximum parsimony implemented in the Mega 2 software [9]. In addition, we used DNAML from the PHYLIP package version 3.62 [10] for a maximum likelihood analysis. We did not find any significant differences in the phylogenetic trees when using these different methods for phylogenetic analysis of the pol and env (C2V3 region) genes. The two sample pairs linked by the confirmed transmission event clustered in all trees with significant bootstrap values, and the sample pair with the unlikely transmission event clustered only in the pol gene analysis with significant bootstrap values. The analysis of the C2V3 region confirmed the initial results [1]. Figure 1 shows the phylogenetic tree of the C2V3 region using the maximum likelihood analysis. The sample pair R004/R016 clustered with a bootstrap value of 53, which does not support a likely transmission. None of the other C2V3 trees grouped this sample pair with significant bootstrap values; in the maximum parsimony and unweighted pair group method using arithmetic averages analysis this sample pair did not cluster at all (data not shown).Fig. 1: Phylogenetic tree of the C2V3 region. GenBank isolates are indicated by their subtype followed by accession number, local control group patient isolates are indicated by ‘R’ and consecutive numbering, and the two isolates 02-30434/02-35144 with a known transmission event (GenBank accession numbers AY878685–AY878692) are shown. Trees were generated by a bootstrap test with 100 replications based on the maximum likelihood method implemented in the PHYLIP package (DNAML). Bootstrap values are shown on the nodes.The additional analysis of our data using several phylogenetic methods presented here does not alter our initial results. Furthermore, several published studies using more than one method showed that analysing possible transmission events with different phylogenetic methods resulted in the same classification of significant clusters [2–6]. In summary, we do not agree that using a single phylogenetic method is insufficient for the reconstruction of HIV-1 transmission events. However, we accept that using two methods instead of only one might yield phylogenetic analyses with greater significance. Data deposition: The sequences reported in this response and in the paper discussed here have been deposited in the GenBank database (accession nos. AY878662–AY878692).