Los Alamos National Laboratory, Los Alamos, NM 87545
Recently there has been intensive activity in the area of discovering new HIV-1 CRFs, although many of the new CRFs have not yet been published (Table 1). In addition, there have been suggestions of new sub-subtypes and potentially a new subtype (10, 11). The classification of new HIV-1 sequences follows the proposed HIV nomenclature guidelines (13, 14). When classifying new sequences, the HIV-1 subtyping reference set is often used. This set has not been updated since 2001, and thus it is time to update this set.
The criteria for updating the reference set were:
The alignments were based on the 2004 web and compendium alignments, which were constructed using HMMER (with a model calculated on the 2003 alignment) (2). The alignment was further improved using SynchAligns (http://hiv-web.lanl.gov/content/sequence/SYNCH_ALIGNS/SynchAligns.html) to add additional sequences, GeneCutter (http://hiv-web.lanl.gov/content/sequence/GENE_CUTTER /cutter.html) to correct reading frames, and manual editing using Se-Al (12).
As in the original subtype reference proposals (1, 6, 9), four sequences per subtype were chosen so that the reference set remains small while allowing the diversity of each subtype to be roughly the same as for all available sequences (similar to an effective population size). In addition, four taxa is the smallest informative unit in an unrooted tree.
Subtype A: Sub-subtype A1 is well established and somewhat more diverse than some other subtypes. A1 is updated with one more recent sequence, and the oldest and divergent reference sequence U455 has been omitted. Sub-subtype A2 is less well established, and as only three full genome sequences have been described there is no choice as to which sequences to include in this sub-subtype. Sub-subtype A3 (10) has not been included at this time because it is less well established and also does not cluster separately from A1 throughout the genome. Similarly, the proposed sub-subtypes A4 and A5 are not included because they have yet not been published (11). Finally, the cluster of A1 sequences from the former Soviet Union countries is well defined in all genomic regions of the HIV genome, but at this time it is not assigned a separate sub-subtype to avoid confusion with other potential sub-subtype candidates.
Subtypes B and D: Subtypes B and D are closer to each other than other subtypes. In most genomic regions they behave as sub-subtypes. Subtype D has a large sample from Uganda (27 out of 51 available sequences), which makes it look as if D may be divided into sub-subtypes. But this is not true; it is simply an effect of dense sampling of an Ugandan population (3). The D reference sequences have been updated to reflect the modern diversity of D. Similarly, subtype B has been updated with more recently sampled sequences. Also, since subtype B is involved in several epidemics in Asia, a reference sequence from this region is included in the set.
Subtype C: This subtype is the most well-described subtype as measured by available full-length genomes, and many countries are represented. There is no sign of subdivision within this subtype, although there is limited geographic clustering of subtype sequences from Asia versus those form India. One more recent sequence from South Africa has been added to the current reference set.
Subtypes F, G, H, J and K: Few full genomes exist, and what is available has been used. Sub-subtype F2 now has four sequence representatives, but otherwise there are no changes since the 2001 reference set.
The U sequences in the 2001 reference set have been omitted. It is possible that these are representatives of a new subtype, but so are all U sequences. Importantly, the U sequences are not a homogeneous group, but rather a collection of unclassified or at the time of submission unclassifiable sequences.
Circulating Recombinant Forms. In previous reference alignments each of the CRFs were described by four representatives. With the large increase of reported CRFs the reference alignment would increase to an extent that it would cause problems in some analyses if four sequences were included for each CRF. Thus, we have limited the CRF section to one sequence per CRF. Except for the E part of CRF01 (and other CRFs that contain subtype E), all subtypes that build up the CRFs are already part of the subtype section in the alignment. The sequence selected for each subtype is now intended to show how it is composed of the included subtypes. If more subtype E sequences are needed in an analysis, one can either refer to the 2001 reference selection or retrieve all E sequences from the HIV sequence database search interface (http://hiv-web.lanl.gov/components/sequence/HIV/combined_search_s_tree/search.html).
Groups O and N: Group N now has three full genome sequences available, and all are included in the reference set. At this time it is unclear whether group O should be divided into subtypes because only 22 full genomes are available which do not describe the full spectrum of group O diversity that is suggested through analysis of partial genome sequences (15). Four sequences, with one change compared to the 2001 set, are included in the references set.
CPZ sequences are included for outgrouping purposes of HIV clusters, and two from each of the Pan troglodytes subspecies troglodytes and schweinfurthii are included in the alignments. The selection is not meant to be representative for the larger PLV evolution. For that purpose we refer to the complete PLV alignment, which has representatives for all major lineages in the PLV tree. See discussion in the PLV section of the HIV Sequence Compendium 2003 (8).
Given enough sequence information, the phylogenetic clades that define HIV-1 groups and subtypes can be reconstructed from any part of the HIV-1 genome. As a rule of thumb, enough sequence information to reconstruct the subtype clades is achieved when the alignment is at least 300-500 characters long. In some regions fewer characters are needed, e.g., env V3 region, while other regions under slower evolutionary change, such as pol RT, need more characters to give reliable results (5, 6). Also, essentially all phylogenetic reconstruction methods are capable to infer the subtype clades (5). Beyond subtyping, however, for more critical phylogenetic analyses of transmission patterns more characters than the minimum above and more advanced reconstruction methods such as maximum likelihood should be used (4, 5, 7).
As part of the revision of the reference alignments presented in the 2005 version, many trees were constructed. These trees were created using enhanced and parallelized versions of Gary Olsen's fastDNAml maximum likelihood tree fitting (RevML) and site rate estimation codes (RevRates). This code was written by Tanmoy Bhattacharya of LANL, and fits a general time reversible model (4).
The trees were created as follows: A candidate tree topology was created assuming uniform site rates and an initial random estimate of nucleotide frequencies and transition rates. RevML proceeds in a heuristic and piecewise way, starting from a small set of sequences and building up the tree topology and branch lengths while making placement decisions that maximize the tree likelihood score. The resulting tree then constrains per-site rate optimization of tree likelihood as a function of global estimates of baseline nucleotide frequency and transition rates. These estimates are fit using the conjugate gradient algorithm in the RevRates program. A second RevML run was then performed using these estimates and in turn another rate estimation procedure refined from the second tree. A final tree was estimated using the twice-refined global and local site rates. Each of the trees in the refinement procedure was independently estimated from the global and site local rate parameters.
Trees were inferred from each gene (env, gag, nef, pol, rev, tat, vif, vpr and vpu) on alignments with all non-recombinant sequences, only group M sequences and all sequences on both globally gap stripped and non-gap stripped data. As expected, alignments with fewer than 400 characters generated trees with some problems. In general, the problems consisted of a lack in resolution among sub-subtypes (mixing of A1 and A2, mixing of F1 and F2 and sometimes K, mixing of B and D). In addition, placement of CPZ sequences was not consistent among short genes. Thus, this reiterates the fact that too short an alignment will not give good tree reconstructions. Trees based on full-length env, pol and gag showed full subtype and sub-subtype resolution (Figure 1). The subtype classifications were clear whether only group M sequences or all sequences were used. Gap stripping already short alignments made the results even worse, while on long alignments it had no effect on subtype associations.
1. Carr, J. K., B. Foley, T. Leitner, M. O. Salminen, B. Korber, and F. McCutchan. 1999. Reference sequences representing the principal genetic diversity of HIV-1 in the pandemic, p. III-10-19. In B. Korber, C. Kuiken, B. Foley, B. Hahn, F. McCutchan, J. Mellors, and J. Sodroski (ed.), Human Retroviruses and AIDS 1998. Theoretical Biology and Biophysics, Los Alamos National Laboratory, Los Alamos, NM.
2. Eddy, S. 1995. HMMER Hidden Markov Models of Protein and DNA Sequence, 1.8 ed. Washington University School of Medicine, St. Louis, MO.
3. Harris, M., D. Serwadda, N. Sewankambo, B. Kim, G. Kigozi, N. Kiwanuka, J. Phillips, F. Wabwire, M. Meehen, T. Lutalo, J. Lane, R. Merling, R. Gray, M. Wawer, D. Birx, M. Robb, and F. McCutchan. 2002. Among 46 near full length HIV type 1 genome sequences from Rakai district, Uganda, subtype D and AD recombinants predominate. AIDS Research and Human Retroviruses 18:1281-1290.
4. Korber, B., M. Muldoon, J. Theiler, F. Gao, R. Gupta, A. Lapedes, B. H. Hahn, S. Wolinsky, and T. Bhattacharya. 2000. Timing the ancestor of the HIV-1 pandemic strains. Science 288:1789-1796.
5. Kuiken, C. L., and T. Leitner. 2001. HIV-1 subtyping, p. 27-53. In A. Rodrigues and G. Learn (ed.), Computational analysis of HIV molecular sequences. Kluwer Academic Publishers.
6. Leitner, T. 1997. Genetic subtypes of HIV-1, p. III-28-40. In G. Myers, B. Korber, B. Foley, K.-T. Jeang, J. W. Mellors, and S. Wain-Hobson (ed.), Human Retroviruses and AIDS 1996: a compilation and analysis of nucleic acid and amino acid sequences. Los Alamos National Laboratory, Los Alamos, NM.
7. Leitner, T., D. Escanilla, C. Franzén, M. Uhlén, and J. Albert. 1996. Accurate reconstruction of a known HIV-1 transmission history by phylogenetic tree analysis. Proceedings of the National Academy of Sciences USA 93:10864-10869.
8. Leitner, T., B. Foley, B. Hahn, P. Marx, F. McCutchan, J. Mellors, S. Wolinsky, and B. Korber. 2004. HIV Sequence Compendium. Theoretical Biology and Biophysics, Los Alamos National Laboratory, Los Alamos, NM.
9. Leitner, T., B. Korber, D. Robertson, F. Gao, and B. Hahn. 1998. Updated proposal of reference sequences of HIV-1 genetic subtypes. In B. Korber, B. Foley, T. Leitner, J. W. Mellors, F. McCutchan, B. Hahn, G. Myers, and C. Kuiken (ed.), Human Retroviruses and AIDS 1997: a compilation and analysis of nucleic acid and amino acid sequences. Los Alamos National Laboratory, Los Alamos, NM.
10. Meloni, S., B. Kim, J. Sankale, D. Hamel, S. Tovanabutra, S. Mboup, F. McCutchan, and P. Kanki. 2004. Distinct human immunodeficiency virus type 1 subtype A virus circulating in West Africa: sub-subtype A3. Journal of Virology 78:12438-12445.
11. Peeters, M. 2005. Personal communication.
12. Rambaut, A. 1996-2002. Sequence Alignment (Se-Al) Program, 2.0a11 ed. Department of Zoology, University of Oxford, Oxford.
13. Robertson, D. L., J. P. Anderson, J. A. Bradac, J. K. Carr, B. Foley, F. Gao, B. H. Hahn, C. Kuiken, G. H. Learn, T. Leitner, F. McCutchan, S. Osmanov, M. Peeters, D. Pieniazek, M. Salminen, S. Wolinsky, and B. Korber. 2000. HIV-1 nomenclature proposal. Science 288:55.
14. Robertson, D. L., J. P. Anderson, J. A. Bradac, J. K. Carr, R. K. Funkhouser, F. Gao, B. H. Hahn, C. Kuiken, G. H. Learn, T. Leitner, F. McCutchan, S. Osmanov, M. Peeters, D. Pieniazek, M. Salminen, S. Wolinsky, and B. Korber. 2000. HIV-1 nomenclature proposal: a reference guide to HIV-1 classification. In B. Korber and et al (ed.), Human Retroviruses and AIDS 1999: a compilation and analysis of nucleic acid and amino acid sequences. Los Alamos National Laboratory, Los Alamos, NM.
15. Yamaguchi, J., P. Bodelle, L. Kaptue, L. Zekeng, L. Gurtler, S. Devare, and C. Brennan. 2003. Near full-length genomes of 15 HIV type 1 group O isolates. AIDS Research and Human Retroviruses 19:979-988.