Effects of thymic selection of the T-cell repertoire on HLA class I-associated control of HIV infection

Abstract

Without therapy, most people infected with human immunodeficiency virus (HIV) ultimately progress to AIDS. Rare individuals ('elite controllers') maintain very low levels of HIV RNA without therapy, thereby making disease progression and transmission unlikely. Certain HLA class I alleles are markedly enriched in elite controllers, with the highest association observed for HLA-B57 (ref. 1). Because HLA molecules present viral peptides that activate CD8(+) T cells, an immune-mediated mechanism is probably responsible for superior control of HIV. Here we describe how the peptide-binding characteristics of HLA-B57 molecules affect thymic development such that, compared to other HLA-restricted T cells, a larger fraction of the naive repertoire of B57-restricted clones recognizes a viral epitope, and these T cells are more cross-reactive to mutants of targeted epitopes. Our calculations predict that such a T-cell repertoire imposes strong immune pressure on immunodominant HIV epitopes and emergent mutants, thereby promoting efficient control of the virus. Supporting these predictions, in a large cohort of HLA-typed individuals, our experiments show that the relative ability of HLA-B alleles to control HIV correlates with their peptide-binding characteristics that affect thymic development. Our results provide a conceptual framework that unifies diverse empirical observations, and have implications for vaccination strategies.

Figure 1. Thymic selection against fewer self-peptides leads to a more cross-reactive T cell repertoire

(a) Histograms for the frequency with which T cells recognize viral peptides (i.e., not seen in the thymus) via only a small number (0, 1, 2, 3) of important contacts are shown for three T cell repertoires that developed with different numbers of self peptide-HLA complexes in the thymus. Important contacts were determined by making single point mutations. If the TCR-peptide-HLA interaction is sufficiently strong, no single point mutation can abrogate recognition, resulting in 0 important contacts. A higher frequency of occurrence of a small number of important contacts implies a more cross-reactive T cell repertoire because only mutations at these contacts are likely to abrogate recognition. The frequency with which T cells recognize viral peptides via many significant contacts (greater than 4) is larger for T cell repertoires restricted by HLA alleles that present more self peptides in the thymus (not shown). (b) The probability that a TCR binds to viral peptides with a certain interaction strength is shown for three T cell repertoires (as in (a)). A particular TCR recognizes a viral peptide when the binding strength exceeds the recognition threshold (dotted black line). Members of a T cell repertoire selected against fewer self peptides are more likely to recognize a viral peptide. The model we used describes qualitative trends robustly^, (Methods), but is not meant to be quantitatively accurate.

Figure 2. Model of host-pathogen interactions shows superior viral control by cross-reactive CD8⁺ T cell repertoires

(a) Dynamical model: the virus mutates, infects limited target CD4⁺ T cells, and is cleared. Infected CD4⁺ T cells produce more free virus, and die. Infected cells present viral peptides in complex with HLA molecules (until peptides unbind from HLA). Activated CD8⁺ T cells produced by recognition of viral epitopes on APCs proliferate and differentiate into effector CTLs. CTLs kill infected cells bearing cognate peptide-HLA complex, and turn into memory cells which are activated upon reexposure to antigen (b) Simulated HIV viral loads versus time for different cross-reactivities (CR) of the CD8⁺ T cell repertoire. Black curve: highly cross-reactive case. Red curve: low cross-reactivity. Each curve is averaged over 500 simulations (each simulation represents a person). The model shows a reduced setpoint viral load for persons with a more cross-reactive T cell repertoire. Other models of host-pathogen dynamics show similar effects of T cell cross-reactivity (Fig. S7–S8). (c) Virus diversity and immune pressure for representative persons (i.e., representative simulations) with high cross-reactivity (left panels) and low cross-reactivity (right panels) of CD8⁺ T cell repertoires. Top panels show the relative population sizes of two dominant viral strains: the infecting strain (black), and an emerging, less fit strain (green) (other less populous viral strains not shown). For persons with a more cross-reactive T cell repertoire, the emergent mutant strain only begins to dominate the infecting strain after 175 days, whereas for low cross-reactivity the mutant increases to nearly 100% of the viral population within 100 days after infection. The lower panels show the relative immune pressure, defined as the rate of killing of an infected cell (third term, right-hand-side, Equation (4), Methods), imposed on each viral strain by different CD8⁺ T cell clones. Each curve represents the relative immune pressure exerted on that viral strain by a particular T cell clone. For persons with a more cross reactive T cell repertoire, multiple T cell clones are exerting immune pressure on both the infecting and emergent strains. For persons with a low-cross-reactivity T cell repertoire, the emergent strain is not recognized, and thus escapes.

Figure 3

HLA-B alleles associated with greater ability to control HIV correlate with smaller peptide binding propensities. The odds ratio (OR) for an allele is defined as: $\frac{p_w/p_{wo}}{c_w/c_{wo}}$, where p_w and p_wo are the numbers of individuals in the progressor cohort with and without this HLA, respectively; c_w and c_wo are the numbers of individuals in the controller cohort with and without this HLA, respectively. This definition implies that the OR measures likelihood of an allele being correlated with progressors versus controllers, with an OR value greater than one implying association with the progressor cohort. The fraction of peptides derived from human proteome that bind to a given HLA allele was determined using the most accurate predictive algorithms (Methods, Table S1). Compared to experimental data, the predictive algorithms for peptide binding by HLA-B*3501 are less accurate than algorithms for the other three alleles (Fig. S17, Table S1); the number reported here for HLA-B*3501 using the most accurate algorithm underestimates the binding fraction. The errorbars reflect 95% confidence interval for ORs. The dotted line corresponds to equal odds for an allele being associated with progressors and controllers.