CoRSeqV3-C: a novel HIV-1 subtype C specific V3 sequence based coreceptor usage prediction algorithm

Abstract

Background: The majority of HIV-1 subjects worldwide are infected with HIV-1 subtype C (C-HIV). Although C-HIV predominates in developing regions of the world such as Southern Africa and Central Asia, C-HIV is also spreading rapidly in countries with more developed economies and health care systems, whose populations are more likely to have access to wider treatment options, including the CCR5 antagonist maraviroc (MVC). The ability to reliably determine C-HIV coreceptor usage is therefore becoming increasingly more important. In silico V3 sequence based coreceptor usage prediction algorithms are a relatively rapid and cost effective method for determining HIV-1 coreceptor specificity. In this study, we elucidated the V3 sequence determinants of C-HIV coreceptor usage, and used this knowledge to develop and validate a novel, user friendly, and highly sensitive C-HIV specific coreceptor usage prediction algorithm.

Results: We characterized every phenotypically-verified C-HIV gp120 V3 sequence available in the Los Alamos HIV Database. Sequence analyses revealed that compared to R5 C-HIV V3 sequences, CXCR4-using C-HIV V3 sequences have significantly greater amino acid variability, increased net charge, increased amino acid length, increased frequency of insertions and substitutions within the GPGQ crown motif, and reduced frequency of glycosylation sites. Based on these findings, we developed a novel C-HIV specific coreceptor usage prediction algorithm (CoRSeqV3-C), which we show has superior sensitivity for determining CXCR4 usage by C-HIV strains compared to all other available algorithms and prediction rules, including Geno2pheno[coreceptor] and WebPSSMSINSI-C, which has been designed specifically for C-HIV.

Conclusions: CoRSeqV3-C is now openly available for public use at http://www.burnet.edu.au/coreceptor. Our results show that CoRSeqV3-C is the most sensitive V3 sequence based algorithm presently available for predicting CXCR4 usage of C-HIV strains, without compromising specificity. CoRSeqV3-C may be potentially useful for assisting clinicians to decide the best treatment options for patients with C-HIV infection, and will be helpful for basic studies of C-HIV pathogenesis.

Figure 1

V3 length and charge alterations associated with C-HIV coreceptor usage. (A, B), A comparison of V3 amino acid lengths segregating CXCR4-using from R5 C-HIV strains. (C, D), A comparison of V3 charge alterations segregating CXCR4-using from R5 C-HIV strains.

Figure 2

Entropy plot comparing V3 sequence variability between CXCR4-using and R5 C-HIV Envs. The potential N-linked glycosylation site, positions 1, 11, 25 and 35, the position of the two amino acid insertion, and the V3 GPGQ crown motif are highlighted. Boxes illustrate regions of specific interest, where there is variability within CXCR4-using V3 sequences but little or no variability within R5 V3 sequences.

Figure 3

A diagrammatic representation of the CoRSeq_V3-C algorithm. Query V3 sequences are aligned to the HXB2 V3 sequence and subjected to a sequential series of questions, which are ordered from top to bottom as indicated by arrows, and as described in detail in the Results section.