Localization and conservation of beneficial and non-beneficial OLP in HIV-1 clade B and C cohorts. Total HIV-1-specific T cell responses were assessed in a cohort of 223 chronically HIV clade B infected, untreated individuals in Lima, Peru (graphs A-C) and in 631 chronically HIV clade C infected, untreated individuals in Durban, South Africa (graphs D-F) using peptide test sets of 410 18 mer overlapping peptides (OLP) spanning the consensus B and C sequences, respectively [2,31]. For each OLP, the protective ratio (PR, defined as "the ratio of the log median viral load in OLP non-responders divided by log median viral load in OLP responders") was determined. Each symbol represents an individual OLP, grouped either by (A, D) proteins or (B, E) protein-subunits for OLP located in Gag, Pol and Env (p-values in A, D based on ANOVA, in B, E on Mann-Whitney by pariwise comparing the different protein subunits, red lines indicating median PR values). In (C and F), the OLP-specific entropy (a measure of the viral diversity in the region the OLP spans) is compared to the OLP-specific PR and shows an inverse association between the sequence conservation and PR (Spearman rank).

Increased breadth of responses to beneficial OLP results in gradually reduced viral loads and is independent of cohort and HLA-B27, -57, -B58, -B81 and -B63. (A) The number of responses to beneficial OLP in the clade C cohort in Durban was determined for each individual and compared to viral loads. An increased breadth of responses to the 22 beneficial OLP was associated with reduced viral loads (ANOVA, p < 0.0001). (B) This association remained equally stable after removing all individuals expressing known beneficial HLA allele (HLA-B27, -B57, -B5801, -B63, -B81) from the analysis (ANOVA, p < 0.0001). (C) The set of 26 beneficial and 17 non-beneficial OLP identified in the clade B infected cohort in Lima, Peru was tested in a second clade B infected cohort in Barcelona. HIV controllers showed a significantly higher focus of responses on the 22 beneficial OLP (61% of all responses to the 43 OLP) while non-controllers reacted predominantly with the non-beneficial OLP (only 29% of all responses targeting beneficial OLP). The Barcelona cohort did not included subject expressing any HLA allele previously associated with relative control of HIV-1 (p = 0.0011, Mann Whitney).

Responses to beneficial OLP are of higher functional avidity and suppress in vitro viral replication more effectively. (A) Responses to the four beneficial OLP located in HIV-1 clade B Gag p24 were retested using a peptide set of 10 mers overlapping by 9 residues. A total of 21 responses in HIV-1 controllers and 24 responses in HIV-1 non-controllers were titrated and the SD50% compared between the two groups, showing a significantly higher functional avidity in the controllers (p = 0.0051, Mann Whitney). (B) Responses to 17 different optimally defined CTL epitopes located in beneficial, neutral and non-beneficial OLP were titrated in samples from 78 HIV infected individuals with variable viral load and disease status. The median SD50% (ng/ml) was defined for each epitope and compared to the OLP-specific protective ratio (Spearman Rank test, p = 0.0020). (C) Ten individuals who mounted responses to well-defined optimal CTL epitopes located in beneficial as well as in non-beneficial clade B OLP were identified and their responses titrated. The SD50% for responses detected in the same individual were compared (Wilcoxon matched pairs test, p = 0.0039). (D) In-vitro viral replication inhibition assays [48] were performed using a Nef modified and Raltegravir resistant test virus and purified CTL effector populations from the same individual targeting beneficial and non-beneficial OLP. One representative experiment of three assays conducted in different individuals is show. Levels of Gag p24 were determined after 4 days of co-culture of effector cells and auologous CD4 T cells used as target cells. Target cells were stimulated 3 days prior with dual-specific anti-CD3/8 mAb and infected at a MOI of 0.1. The negative control contained wells with target cells only ("no CD8").

Genome distribution, entropy and RT localization of OLP with significant impact on viral loads in HIV-1 clade B and C infection: The distribution of OLP with significantly elevated or reduced PR across the viral proteome is shown in A) for clade B infection (cut-off uncorrected p-value of p < 0.05) and in B) for clade C infection (cut-off q < 0.2). The entropy of beneficial and non-beneficial clade B OLP is compared in C) while in D), the entropy of beneficial OLP in HIV clade C Gag is compared to the remainder of Gag OLP (p-values based on Mann Whitney, red lines indicating median sequence entropies). In E and F, protein structures for HIV-1 reverse transcriptase (Protein databank structure ID 3IG1) were loaded into the Los Alamos HIV Database "protein feature accent" tool http://www.hiv.lanl.gov/content/sequence/PROTVIS/html/protvis.html and locations of beneficial RT OLP identified in clade B (Table 1) and in clade C (Table 2) marked by red highlights.

Responses to OLP identified in multi-variate analysis are associated with reduced viral loads: Response patterns and HLA class I genetics in the clade B cohort in Lima and clade C cohort in Durban were subjected to FASS multivariat analysis [41-43]. Viral loads in individuals mounting zero vs. at least one response to beneficial OLP identified by the FASS multi-variate analysis were compared for (A) the Lima clade B cohort and the (B) Durban clade C cohort. The larger data set for the clade C cohort allowed for a further stratification of the responder group by increasing numbers of targeted OLP emerging from the FASS analysis (C). A gradually declining median viral load in relation to an increasing breadth of these responses was seen (ANOVA, p < 0.0001).