Ecological analysis of antigen-specific CTL repertoires defines the relationship between naive and immune T-cell populations

Abstract

Ecology is typically thought of as the study of interactions organisms have with each other and their environment and is focused on the distribution and abundance of organisms both within and between environments. On a molecular level, the capacity to probe analogous questions in the field of T-cell immunology is imperative as we acquire substantial datasets both on epitope-specific T-cell populations through high-resolution analyses of T-cell receptor (TCR) use and on global T-cell populations analyzed via high-throughput DNA sequencing. Here, we present the innovative application of existing statistical measures (used typically in the field of ecology), together with unique statistical analyses, to comprehensively assess how the naïve epitope-specific CD8(+) cytotoxic T lymphocyte (CTL) repertoire translates to that found following an influenza-virus-specific immune response. Such interrogation of our extensive, cumulated TCR CDR3β sequence datasets, derived from both naïve and immune CD8(+) T-cell populations specific for four different influenza-derived epitopes (D(b)NP(366), influenza nucleoprotein amino acid residues 366-374; D(b)PA(224), influenza acid polymerase amino acid residues 224-233; D(b)PB1-F2(62), influenza polymerase B 1 reading frame 2 amino acid residues 62-70; K(b)NS2(114), and influenza nonstructural protein 2 amino acid residues 114-121), demonstrates that epitope-specific TCR use in an antiviral immune response is the consequence of a complex interplay between the intrinsic characteristics of the naïve cytotoxic T lymphocyte precursor pool and extrinsic (likely antigen driven) influences, the contribution of which varies in an epitope-specific fashion.

Fig. 1.

Selective clonotype expansion in immune epitope-specific responses. (A) Clonotypic diversity within naïve (N, closed symbols) and immune (I, open symbols) mice was measured using Simpson’s diversity index (*P < 0.01 using Student’s unpaired t test). (B) Evenness of clonotype distribution in naïve and immune mice was determined by ranking clonotypes from largest to smallest for each mouse and the cumulative proportion of total clonotypes was plotted against the cumulative percentage of the response (percentage of the total sequences). Shown are the means ± SD of data obtained from 6, 6, 5, and 4 naïve mice (closed symbols) and 11, 13, 19, and 2 immune mice (open symbols), for D^bNP₃₆₆, D^bPA₂₂₄, D^bPB1-F2₆₂, and K^bNS2₁₁₄, respectively. (C) Statistical comparison of clonotype distribution between naïve (N, closed symbols) and immune (I, open symbols) populations was performed using the Gini coefficient (26) (*P < 0.001, comparing naïve to immune sets).

Fig. 2.

Clonotype sharing versus abundance in virus-specific CTL responses. (A) PTIC is the proportion of the clonotypic response from individual immune mice that is shared by at least 33% of mice sampled. Shown is mean ± SD for sequence data described in Fig. 1. Statistical analyses were performed using the Mann–Whitney test. NS, not significant, P > 0.05. (B) Relative abundance of shared clonotypes was determined by plotting, for each mouse, the proportion of clonotypes shared in ∼25% (first column) or ∼40% (second column) of mice (actual percentage of mice is noted within each plot and is determined by the number of mice analyzed) against the proportion of the total response those clonotypes contribute (proportion sequences shared).

Fig. 3.

Nucleotide flexibility within and between naïve individuals. Model sharing indices were generated using two methods of rarefaction to test how the measures of within- and between-mouse sharing compared with a random distribution. (A) All experimentally observed nucleotide sequences from naïve animals were randomly assigned to one of six theoretical “mice” (without replacement) and the within-mouse and between-mice sharing measures were calculated for this artificially produced dataset, with 1,000 total simulations. (B) For each epitope specificity, sequences were randomly distributed among theoretical mice, by randomly drawing from the pool of sequences, with each sequence having an equal probability of being chosen in each draw (i.e., sampling with replacement). Within-mouse and between-mice sharing is then calculated (as described) for the theoretical set (dots; a representative 100 of a total 1,000 simulations) or actual data (triangles).