Covariance of charged amino acids at positions 322 and 440 of HIV-1 Env contributes to coreceptor specificity of subtype B viruses, and can be used to improve the performance of V3 sequence-based coreceptor usage prediction algorithms

Abstract

The ability to determine coreceptor usage of patient-derived human immunodeficiency virus type 1 (HIV-1) strains is clinically important, particularly for the administration of the CCR5 antagonist maraviroc. The envelope glycoprotein (Env) determinants of coreceptor specificity lie primarily within the gp120 V3 loop region, although other Env determinants have been shown to influence gp120-coreceptor interactions. Here, we determined whether conserved amino acid alterations outside the V3 loop that contribute to coreceptor usage exist, and whether these alterations improve the performance of V3 sequence-based coreceptor usage prediction algorithms. We demonstrate a significant covariant association between charged amino acids at position 322 in V3 and position 440 in the C4 Env region that contributes to the specificity of HIV-1 subtype B strains for CCR5 or CXCR4. Specifically, positively charged Lys/Arg at position 322 and negatively charged Asp/Glu at position 440 occurred more frequently in CXCR4-using viruses, whereas negatively charged Asp/Glu at position 322 and positively charged Arg at position 440 occurred more frequently in R5 strains. In the context of CD4-bound gp120, structural models suggest that covariation of amino acids at Env positions 322 and 440 has the potential to alter electrostatic interactions that are formed between gp120 and charged amino acids in the CCR5 N-terminus. We further demonstrate that inclusion of a "440 rule" can improve the sensitivity of several V3 sequence-based genotypic algorithms for predicting coreceptor usage of subtype B HIV-1 strains, without compromising specificity, and significantly improves the AUROC of the geno2pheno algorithm when set to its recommended false positive rate of 5.75%. Together, our results provide further mechanistic insights into the intra-molecular interactions within Env that contribute to coreceptor specificity of subtype B HIV-1 strains, and demonstrate that incorporation of Env determinants outside V3 can improve the reliability of coreceptor usage prediction algorithms.

Figure 1. Correlation between charged amino acids at Env positions 322/440 in phenotypically characterized B-HIV strains.

(A) Values represent the percentage of Envs that have the indicated type of amino acid (+, positively charged; -, negatively charged; o, neutral) at position 440 out of all Envs with the indicated type of amino acid at position 322. Subscript values represent the percentage of R5 or CXCR4-using Envs that have the indicated combination of amino acid types at positions 322/440. (B) Values represent the percentage of R5 or CXCR4-using Envs that have the combination of indicated amino acids at Env positions 322/440. Amino acids are colored according to charge; positively charged (+, Arg and Lys) in blue, negatively charged (-, Asp and Glu) in red and neutral (o) in white. 322/440 amino acid combinations are colored according to charge; +322/+440 in dark blue, −322/−440 in dark red, +322/−440 or −322/+440 in purple, +322/o440 or o322/+440 in light blue, −322/o440 or o322/−440 in light red and o322/o440 in white. P-values were calculated by chi-square analysis.

Figure 2. Nucleotide codon usage of R5 and CXCR4-using B-HIV Envs at position 440.

The size of each node indicates the relative frequency of the indicated nucleotide codon. A black line connects codons separated by one nucleotide substitution. Codons encoding positively charged amino acids at colored blue, codons encoding negatively charged amino acids are colored red and codons encoding neutral amino acids are colored white. Nucleotide codons at position 440 were aligned and analysed using CLC Main Workbench version 6.5. 75 CXCR4-using (46 R5X4 and 29 X4) and 293 R5 sequences were analyzed.

Figure 3. Three-dimensional model of the CD4-bound gp120 containing V3 and C4 docked to a CCR5 N-terminus peptide.

(A) The gp120 molecule is shown in ribbon representation, the V3 is shown in grey and the C4 region is shown in blue. The CCR5 N-terminus peptide is shown in ribbon representation (red) with wire mesh van der Waals surface representation, and the amino acids are shown in stick representation. Asp11 in the CCR5 N-terminus and Glu322 and Arg440 in gp120 are shown in stick representation and colored according to charge; positively charged residues are colored blue and negatively charged residues colored red. (B, C) Close up of gp120 and the CCR5 N-terminus peptide interface is represented as described above. Notably, CCR5 N-terminus mesh surface representation and amino acid stick representation was removed, and mesh surface van der Waals surface representation was added to Asp11 in the CCR5 N-terminus, Arg440 (B) and Lys 440 (C). (D) Three-dimensional model of CD4-bound gp120 containing V3 and C4 docked to the CCR5 N-terminus (to aid visual distinction between gp120 and the CCR5 N-terminus peptide) with (E) original Glu322/Arg440 genotype or (F) mutated Arg322/Glu440 genotype, shown in surface charge representation colored according to charge. The number of hydrogen bonds (N_HB) and salt bridges (N_SB) predicted to form at the interface between gp120 and the CCR5 N-terminus was calculated using PISA software.

Figure 4. Analysis of covariance of charged amino acids at Env positions 322/440 in non-B subtype HIV-1 strains.

Values represent the percentage of Envs within each subtype that have the indicated type of amino acid (+, positively charged; -, negatively charged; o, neutral) at position 440 out of all Envs with the indicated type of amino acid at position 322. Subscript values represent the percentage of Envs within each subtype that have the indicated combination of amino acid types at positions 322/440. Amino acids are colored according to charge, as described in the legend for Figure 1. P-values were calculated by chi-square analysis (ns, not significant; N/A, not applicable).

Figure 5. Correlation between charged amino acids at Env positions 322/440 and coreceptor usage for phenotypically characterized C-HIV, D-HIV and AE-HIV strains.

Values represent the percentage of R5 or CXCR4-using Envs within C-HIV (A), D-HIV (B) and AE-HIV (C) subtypes that have the indicated type of amino acid (+, positively charged; −, negatively charged; o, neutral) at position 440 out of all Envs with the indicated type of amino acid at position 322. Subscript values represent the percentage of R5 or CXCR4-using Envs within each subtype that have the indicated combination of amino acid types at positions 322/440. Amino acids are colored according to charge, as described in the legend for Figure 1.