|
Background:
Whilst it is evident that the principal determinants for reduced Enfuvirtide
(ENF) susceptibility reside in HIV-1 gp41, the possibility remains that
other determinants within baseline (BL) HIV-1 envelope amino acid sequences
may also influence ENF susceptibility for virus from Fusion Inhibitor naïve
patients. We explored such relationships by further mining the TORO 1 and
TORO 2 clinical trial databases using univariate and multivariate
statistical methods.
Methods:
Paired BL genotype and phenotype data were analyzed for 377 R5 tropic
recombinants from patients receiving ENF and an optimized background
regimen. HIV-1 envelope (complete gp160) amino acid sequences and ENF
susceptibility were generated using the GeneSeqTM and PhenoSenseTM
HIV Entry assays. The relationship of genotype (JRCSF reference) and
phenotype at each individual gp160 position was explored using analysis of
variance (ANOVA) models. The joint relationship between the entire
gp160 sequence and ENF susceptibility was studied using cluster analysis and
regression tree modeling.
Results:
The fold change in IC50 was approximately log-normally distributed with a
geometric mean (GM) of 1.58 (range 0.04–37.71) and a standard deviation of
2.64. ANOVA identified “important” amino acids (p-value<0.05) in both gp41
(e.g., G3GT, N42S and V69L) and gp120 (e.g., T50I and L444M) that were all
associated with >1.5 fold change in susceptibility from the GM. To explore
differences among gp41 genotypes, cluster analysis was used to identify two
clusters of 365 (Cluster 1) and 12 (Cluster 2) recombinants. Recombinants
in Cluster 2 were identified as non-B subtypes. Cluster 2 (GM=0.78, range
0.04–4.44) showed a higher ENF susceptibility (p-value=0.01) than Cluster 1
(GM=1.61, range 0.05–37.71); Cluster 1 had a GM and range similar to the
complete dataset. Recombinants in Cluster 2 had higher frequencies of the
polymorphisms N42S, E151A, N305D and T130T, which were associated with
increased ENF susceptibility. Additionally, the clusters were compared
using Week 24 efficacy data; patients in Cluster 2 experienced a larger drop
in VL than patients in Cluster 1. To further explore the genotype and
phenotype relationship, binary regression tree models were constructed. The
optimized tree contained five terminal nodes. Tree branch points occurred
at gp41 positions 3ins (insertion at 3), 24 and 42, which also appeared in
the “important” amino acid set obtained by ANOVA.
Conclusions:
Statistical analyses have revealed an association between polymorphic sites
in HIV Env and baseline variability of ENF susceptibility in HIV-1 R5 tropic
recombinants. Cluster analysis of gp41 viral sequences identified
combinations of polymorphisms that were associated with higher ENF
susceptibility in the 12 non-B R5 recombinants and a regression tree model
illustrated amino acid interactions associated with ENF susceptibility.
The significance of these associations are under study. |
