Immune Correlates of Disease Progression in Linked HIV-1 Infection

Abstract

Genetic and immunologic analyses of epidemiologically-linked HIV transmission enable insights into the impact of immune responses on clinical outcomes. Human vaccine trials and animal studies of HIV-1 infection have suggested immune correlates of protection; however, their role in natural infection in terms of protection from disease progression is mostly unknown. Four HIV-1⁺ Cameroonian individuals, three of them epidemiologically-linked in a polygamous heterosexual relationship and one incidence-matched case, were studied over 15 years for heterologous and cross-neutralizing antibody responses, antibody binding, IgA/IgG levels, antibody-dependent cellular cytotoxicity (ADCC) against cells expressing wild-type or CD4-bound Env, viral evolution, Env epitopes, and host factors including HLA-I alleles. Despite viral infection with related strains, the members of the transmission cluster experienced contrasting clinical outcomes including cases of rapid progression and long-term non-progression in the absence of strongly protective HLA-I or CCR5Δ32 alleles. Slower progression and higher CD4/CD8 ratios were associated with enhanced IgG antibody binding to native Env and stronger V1V2 antibody binding responses in the presence of viruses with residue K169 in V2. ADCC against cells expressing Env in the CD4-bound conformation in combination with low Env-specific IgA/IgG ratios correlated with better clinical outcome. This data set highlights for the first time that V1V2-directed antibody responses and ADCC against cells expressing open, CD4-exposed Env, in the presence of low plasma IgA/IgG ratios, can correlate with clinical outcome in natural infection. These parameters are comparable to the major correlates of protection, identified post-hoc in the RV144 vaccine trial; thus, they may also modulate the rate of clinical progression once infected. The findings illustrate the potential of immune correlate analysis in natural infection to guide vaccine development.

Keywords: ADCC; BEAST; IgA/IgG ratio; V1V2 antibody binding; epidemiologically-linked infection; human immunodeficiency virus (HIV); protective immune parameters and host factors; viral signature K169.

Figure 1

Clinical parameters of the four Cameroonian HIV+ study participants. Table summarizing longitudinal study time points, dates of sampling, the age of participants, accomplished viral sequencing (seq) and selected clinical parameters. The results of incidence testing (Inc) are shaded in gray (#f1, #f2, #m) or pale red (#f3) according to sampling start at a stage of chronic infection (>6 months) or recent infection (<6 months), respectively. In addition, the dates of diagnosis (Diag) and the estimated dates of infection (EDI) based on BEAST analyses (Figure 2) are indicated. The time point when superinfection (SI) was detected in #f3 is labeled (gray). Time points at which the participants received antiretroviral treatment (ART) are highlighted in orange with the ART start date indicated. The clinical classification as long-term non-progressor (LTNP, #f2) is based on CD4 counts >500 cells/μL in the absence of AIDS-defining symptoms for >8 years without ART. Slow progression (#f3) is defined as the absence of AIDS-defining symptoms without ART for >8 years. The classification as progressor (#m) is based on CD4 counts dropping below 300 cells/μL and ART initiation within 4 years after diagnosis; rapid progression (#f1) is defined by CD4 counts repeatedly dropping below 200 cells/μL within 4 years after diagnosis and estimated date of infection.

Figure 2

Phylogenetic analysis, transmission events, and clinical parameters in four Cameroonian HIV⁺ individuals. (A) Timeline of essential study time points (spheres). Orange pill icons indicate time points of antiretroviral treatment initiation. TM and SI indicate the occurrence of transmission or superinfection events, respectively. (B) Temporal estimation of epidemiologically-linked transmission events and dates of infection using BEAST time-calibrated MCC trees (564 functional env sequences, 2002–2017, HIV region 6225–7817 according to HXB2 numbering). The branches of the trees are color-coded according to each individual. A color gradient along the branches and a lightning symbol (same color gradient) indicate historical transmission events among individuals. Transmission events are numbered consecutively for each recipient. The median estimated dates of infection (including 95% highest posterior density) are labeled and highlighted in the tree with a star. (C) Bump chart illustrating longitudinal CD4/CD8 ratios in the four study participants along the timeline on the X-axis modeled based on available data points. The magnitude of CD4/CD8 ratios is indicated on the Y-axis, sorted in descending order. Pill icons and brown lines indicate long-term antiretroviral treatment. BEAST, Bayesian evolutionary analysis by sampling trees; LTNP, long-term non-progressor; MCC, maximum clade credibility; URF, unique recombinant form. Inset, Schematic overview of transmission events and clinical classifications (2002–2017) of the four participants. Virus symbols indicate genetically related strains of subtypes URF (blue-purple) and CRF02_AG (red).

Figure 3

Env genetic analysis in four Cameroonian HIV⁺ individuals. (A) Neighbor-joining phylogenetic trees (Kimura two-parameter model) displaying env sequences (HIV region 6225-7817 according to HXB2 numbering). Taxa are color-coded according to the participant (Figure 2). For each study participant, all sequences from multiple time points are displayed, i.e., ≥20 env sequences per time point (Figure 1). Bootstrap values are indicated in gray for the major branches of the participants' CRF02_AG and unique recombinant form (URF) sequences. The bar indicates a genetic distance of 5%. Reference sequences (LANL Database) are shown in black. CRF02_AG sequences of participant #f3 are separated into two populations: pre-superinfection (red icon; time points 1-9) and post superinfection (red-gray icon; time points ≥10). (B) Schematic representation of env recombination patterns in the three viral populations of the four study participants (left) as determined in Simplot bootscan analyses (right). The horizontal bar indicates the entire env genomic region; F2 parts are displayed in turquoise, CRF02_AG parts are displayed in dark red, and dashed gray indicates sequences of a genetically distant (>5%) CRF02_AG variant. Simplot Bootscan analyses were performed using representative query sequences for each virus population (indicated on top of each plot) against subtype CRF02_AG (red), F2 (turquoise) and B (light brown and green, outlier) reference sequences (boxed). The window width, step size, and bootstrap replicates were set to 200 bp, 20 bp, and 100, respectively. The Y-axis indicates bootstrap support; the X-axis indicates the env region. Recurring breakpoints supported by bootstrap values >70% are indicated as vertical black lines. The full range of genetic distances in env between viral populations is indicated (middle). (C) Env epitope analysis of six representative single genome amplified (SGA) sequences (middle), isolated from the four HIV+ individuals. At the bottom, a sequence logo analysis is shown for the entire set of studied Env sequences; asterisks indicate 100% conserved sites. Reference sequences of clade B (HXB2) and CRF02_AG (250) are shown on top. N-glycosylation sites critical for broadly neutralizing antibodies (bnAbs) are colored in olive green: N88 (gp120/gp41 interphase bnAb 35O22), N156 and N160 (V2 glycan bnAbs, e.g., PG9/PG16), N234 and N276 (gp120/gp41 interphase bnAb 8ANC195), N301 and N332/N334 (V3 glycan bnAbs, e.g., PGT121/PGT128) and N611 and N637 (gp120/gp41 interphase bnAb PGT151). Sites of immune pressure in the RV144 vaccine trial (K169, V172, and I181 mm (mismatch) in V2) are highlighted in light blue. Possible sites of resistance to CD4 binding site bnAbs are highlighted in orange. Dots indicate the presence of the above-listed amino acid residue. Black entries (or light blue in the case of I181 mm) indicate divergence from the above-listed amino acid.

Figure 4

Heterologous neutralization in study participants. Heterologous neutralization responses of longitudinal plasma samples (time points in brackets; S: detected superinfection) against six reference pseudoviruses (Env subtypes and viral strains listed); murine leukemia virus (MLV) was used as negative control (n.c.). Mean plasma dilutions (dil) needed for 50% neutralization (IC₅₀) were calculated from two independent experiments in duplicates and are illustrated in a heat map. Breadth, potency, and combined breadth-potency (BP) scores were calculated for each sample separately and in relation to mean IC₅₀ values of the studied pseudoviruses.

Figure 5

IgA/IgG levels, cell-associated Env binding and antibody-dependent cellular cytotoxicity (ADCC). (A,B) Longitudinal ADCC analysis using Nef- and Vpu-deficient (N-U-) (A) and wild-type (WT) (B) HIV-1 ADA infected CEM-NKR cells in plasma samples of the four study participants. (C,D) Longitudinal analysis of cell-associated Env binding (cell staining) using HIV-1 ADA infected CEM-NKR cells. Infection experiments were performed both with N-U- (C) and WT virus (D). (E,F) Longitudinal analysis of Env-specific (E) and total (F) IgA/IgG ratios. (G,H) Longitudinal analysis of Env-specific (G) and total (H) IgA levels. (I,J) Longitudinal analysis of Env-specific (I) and total (J) IgG levels. On the X-axis, participant IDs are listed chronologically by time point. T and S in superscript indicate time points where participants were on antiretroviral treatment or at superinfection (SI), respectively. Gray arrows indicate selected patterns of changes upon superinfection or initiation of antiretroviral treatment. MAb A32 was used as a positive control (P.C.) and mock-infected cells and/or HIV negative plasma as a negative control (N.C.). Means of two or three independent experiments performed in duplicates (IgA and IgG quantitations) or triplicates (ADCC and cell stainings) are shown. Also, floating bars (min to max) are shown in (A,B,E,F) and standard deviations in (C,D). All statistically significant differences between study participants are indicated as well as selected statistical comparisons between two-time points (One-way ANOVA, ^*P < 0.05; ^**P < 0.005; ^***P < 0.0005; ^****P < 0.0001; ns, not significant).

Figure 6

ADCC and IgA/IgG patterns. (A) Scatter plot analysis of Env-specific plasma IgA/IgG ratios (top) and ADCC (N-U-)(bottom) against HIV-1 ADA-infected CEM-NKR cells. Mean and standard deviation for the five longitudinal values per participant are shown. All statistically significant differences among participants are indicated (one-way ANOVA, ^*P < 0.05; ^**P < 0.005). (B,C) Correlation analysis between ADCC (N-U-) and Env-specific plasma IgA/IgG ratios using all data points (B) and means per participant and parameter (C). Correlation coefficients r (with 95% confidence intervals if applicable) and P-values are indicated using non-parametric Spearman rank. Asterisks indicate statistically significant correlations in a two-tailed (B) or one-tailed Spearman rank test (means) (C) (^*P < 0.05). ADCC, antibody-dependent cellular cytotoxicity; ART, antiretroviral treatment; N-U-, Nef- and Vpu-deficient.

Figure 7

Antibody binding profiles of study participants. (A) Heatmap of longitudinal total IgG antibody binding levels, determined by a multiplex bead assay (Luminex) using plasma diluted 1:200 against a set of eleven Env antigens. Antigen #10 is trimeric (3x), and #11 is pentameric (5x). BSA coated beads and plasma from an uninfected, healthy Cameroonian individual were used as negative controls. ^T: antiretroviral treatment; ^S: superinfection. (B) Quantitative scatter plot analysis of total IgG levels against gp120, V3, C5, and V1V2. For gp120 and V1V2, the average binding levels of two gp120 proteins or six scaffolded V1V2 proteins (see A) are shown, respectively. The dots represent the different time points per participant for the corresponding antigen; SI: superinfection. All statistically significant differences (one-way ANOVA) in overall binding levels among individuals are indicated, as well as statistical analyses comparing the binding of every antigen against V1V2 within each individual. IgG subclass composition (IgG1-4) of plasma binding to the four selected Env antigens (averaged longitudinal values) are shown as pie charts according to the indicated color-code. (C) Cell-surface staining of CEM-NKR cells infected with a virus encoding ADA wild-type (WT, dark gray, top) or Nef-, Vpu-deficient (N-U-, light gray, bottom) virus. Staining of native Env was done using longitudinal plasma samples (time points in brackets) and is displayed in stacked bar graphs with means of at least two repetitive experiments. Statistical analysis was performed for the N-U- data set using one-way ANOVA. ^*P < 0.05; ^**P < 0.005; ^***P < 0.0005; ^****P < 0.0001; ns, non-significant.

Figure 8

IgG apparent affinities, ELISA plasma binding levels, and binding correlation analysis. (A) Heatmap of half-maximum binding concentrations (EC50, μg/mL) determined by ELISA using plasma-purified IgG (range of 0.2–500 μg/mL) from longitudinal time points (shown in parentheses). A set of six Env antigens was used: trimeric SOSIP/BG505, gp120/JRFL, cyclic V3/ZM109 peptide, constrained V1V2/ZM109-1FD6 scaffold, gp120 core/JRFL, and MPER gp41/con B peptide. T and S in superscript indicate time points where participants were on antiretroviral treatment or at superinfection, respectively. (B) Plasma samples diluted 1:100 were tested against six different V1V2 constructs: V1V2-gp70 fusion proteins A244 (clade AE), CN54 (clade B/C), and CaseA2 (clade B), V1V2-1FD6 fusion proteins ZM109 and CAP45 (both clade C), as well as trimeric V1V2-2J9C fusion protein ZM53 (clade C). Antigens gp120 MN (clade B) and gp41 MN (clade B), as well as anti-V2 mAb 697-D, were used as positive controls; BSA, plasma of a healthy HIV⁻ Cameroonian individual and anti-parvovirus B19 mAb 1418 were used as negative controls. A heat map of binding levels is shown at the bottom. The V1V2 binding responses are summarized for each time point sample to a V1V2 cross-reactive binding score, shown on top. V1V2 cross-reactive binding scores were statistically compared between participants using one-way ANOVA; asterisks indicate statistically significant differences according to the boxed P-value scheme. Bi-colored diamonds as symbols indicate the time point when superinfection (S in superscript) was detected in #f3. (C–E) Correlation analysis between Luminex V1V2, V3, C5, gp120, and cell-associated Env binding levels. Linear regression analysis was performed for the following data sets: all individual data points of the four participants (C), binding means per participant and antigen (D) and means per participant/antigen from time points pre-ART and pre-superinfection (E). Correlograms are shown, sized and color-coded according to the correlation coefficient (r). Asterisks indicate all correlations that reached statistical significance in a two-tailed Spearman rank test (C) or one-tailed Spearman rank test for the means (D,E) according to the provided P-value scheme (on the left). Red background indicates the presence of a positive correlation pattern; the blue background indicates the tendency of an overall inverse or absence of correlation pattern. ART, antiretroviral treatment.

Figure 9

Correlation analysis of the V1V2 binding response and epitope site-scanning. (A) Correlation analysis of V1V2 binding levels with selected immune responses and the presence of residue K169 and mismatch (mm) at I181 in contemporaneous viral sequences (according to the sieve analysis in the RV144 vaccine trial). In the correlogram, circles are sized and color-coded according to linear regression coefficients (r) between the indicated parameters. Asterisks indicate all statistically significant correlations (Spearman rank, ^*P<0.05; ^**P<0.01; ^***P<0.005). The two most significant correlations with V1V2 binding (avg) are depicted as scatter plots on the left (95% confidence intervals in brackets). Sieve (V2) on the x-axis is scaled according to the presence of K169 (score of 1), a mismatch at I181 (score of 1), or both (score of 2) in contemporaneous viruses of participants. ART: antiretroviral treatment; avg: average; Env, Env-specific; naïve, ART-naïve; SI, superinfection; WT, wild-type infection; N-U-, Nef- and Vpu-deficient virus infection. (B) Sequence logo analysis of the immunodominant V1V2 region (Env amino acid region 169-181 according to HXB2 numbering) from all functional viral sequences of the four study participants. The presence of residue K169 or mismatch of I181 is boxed and highlighted in light blue. (C) Site-scanning analysis of all non-conserved amino acid sites within the immunodominant V1V2 region of longitudinal viral sequences from the four study participants. Spearman rank correlations were performed between longitudinal V1V2 binding (avg) responses and the presence of the indicated amino acid residue(s) in the contemporaneous viruses. Correlation coefficients r, 95% confidence intervals (CI), significance P, and multiple-comparisons-adjusted P-values (Benjamini Hochberg and Bonferroni method) are displayed for each correlation; statistically significant results (P < 0.05) are highlighted in light blue.

Figure 10

Clinical correlation analysis with viral, immunologic, and host parameters. (A) Chord diagram illustrating the network of linear correlations among eight major viral, immunologic, and clinical factors. Chords are color-coded according to the magnitude of the correlation coefficient (r); chord width inversely corresponds to the P-value. Two-tailed Spearman rank tests were performed using data points at time points pre-ART and pre-superinfection. Asterisks indicate all statistically significant correlations within chords (^*P < 0.05). Colored spheres highlight statistically significant linear correlations with CD4/CD8. (B,C) Radar plots comparing the inter-individual patterns of ten selected viral, immunologic, host, and clinical factors (means from data points pre-ART and pre-superinfection). (B) Overlay of radar plots from the four participants. Scaling of the radial axes is indicated and the direction goes from low protective values in the center to high protective values at the rim of the octagons (inv: inverse order). (C) Individual radar plots are ordered from left to right, according to lower disease progression and higher CD4/CD8 ratios. Mean CD4/CD8 ratios from time points pre-ART and pre-superinfection are classified as: 0–0.25 (–); >0.25–0.5 (+); >0.5–0.75 (++); >0.75 (+ + +). Env, Env-specific; HLA, protective HLA scoring; WT, wild-type infection; Neut, Neutralization breadth-potency score; N-U-, Nef- and Vpu-deficient virus infection; PVL, plasma viral load; V1V2, averaged binding to scaffolded V1V2 antigens.