A hierarchy of selection pressures determines the organization of the T cell receptor repertoire

Abstract

We systematically examine the receptor repertoire in T cell subsets in young, adult, and LCMV-infected mice. Somatic recombination generates diversity, resulting in the limited overlap between nucleotide sequences of different repertoires even within the same individual. However, statistical features of the repertoire, quantified by the V gene and CDR3 k-mer frequency distributions, are highly conserved. A hierarchy of immunological processes drives the evolution of this structure. Intra-thymic divergence of CD4+ and CD8+ lineages imposes subtle but dominant differences observed across repertoires of all subpopulations in both young and adult mice. Differentiation from naive through memory to effector phenotype imposes an additional gradient of repertoire diversification, which is further influenced by age in a complex and lineage-dependent manner. The distinct repertoire of CD4+ regulatory T cells is more similar to naive cells in young mice and to effectors in adults. Finally, we describe divergent (naive and memory) and convergent (CD8+ effector) evolution of the repertoire following acute infection with LCMV. This study presents a quantitative framework that captures the structure of the repertoire in terms of its fundamental statistical properties and describes how this structure evolves as individual T cells differentiate, migrate and mature in response to antigen exposure.

Keywords: CDR3AA motifs; LCMV; TCR repertoire; aging; epitope-specific repertoire.

Figure 1

Clonal expansion and diversity of the TCRβ repertoire in different subsets of young and adult mice. (A) Summary of T cell compartments and pipeline for cell isolation and TCR repertoire sequencing and analysis. (B) The TCRs in each repertoire were ranked according to frequency, and the proportion within each decile is illustrated (low abundance sequences in white, ranging to high abundance sequences in dark red). The percentage of the distribution represented by the top decile is shown in white text. (C) The sequence abundance distribution in each compartment. The plots show the proportion of the repertoire (y-axis) made up of TCR sequences observed once, twice, etc. (x-axis). Repertoires from young mice are shown with red dots, repertoires from older mice in blue dots, and synthetic repertoires in green. (D) Simpson and Shannon scores for subsampled repertoires of equal size (1000 CDR3NTs) from each compartment and mouse. Colors same as panel (C) Mean is shown as black lines (n=3). (E) PCA of the Renyi diversities of order 0, 0.25, 0.5, 1, 2, 4.

Figure 2

Different sub-populations of T cells in young individuals have different V gene distributions. (A) Cosine similarity was calculated between the V gene distributions for all pairs of repertories in young (left) or adult (right) mice and displayed as a heatmap. Hierarchical clustering dendrograms show the assigned organization at each plot, colored by CD4+ and CD8+ groups (grey and red branches respectively) and labels by compartment (text and symbol). Tissues are marked in symbols shape (SP = triangles, BM = circles). (B) PCA of the Vβ usage of CD4+ and CD8+ compartments in young (upper) and adult (lower) panels. Each color represents one compartment from one mouse (e.g., CD8+ Effectors, BM, mouse 1). See legend for symbols and color code. The Vβ genes with the highest influence (loading) are marked with arrows. (C) Cosine similarity between Vβ gene usage distributions between individuals (circles) or within individuals (between spleen and bone marrow, triangles). The inter-individual variability was calculated separately for spleen and bone-marrow. Each point is the cosine value calculated between two different mice and tissues (SP-SP, SP-BM, BM-BM). T cells compartments (colored dots) are divided into CD4+ (left) and CD8+ (right) from young or adult mice. Mean is shown by horizontal grey lines.

Figure 3

Nucleotide sequence sharing patterns differ between T cell sub-compartments. (A) Each circus plot represents a single mouse CD4+ or CD8+ compartment (upper and lower panel, respectively). Circus sharing levels illustrate the number of clones shared between two compartments (band widths), and the proportion of shared clones attributed to each compartment (circus arcs). Only sequences shared by at least two compartments were included in the analysis. (B) Pairwise cosine similarity of the CDR3βNT sequences from representative young and adult mouse repertoires. Correlation levels are represented by color (high=light blue, low= dark blue). In color and text, hierarchical clustering dendrograms for all T cell compartments are plotted to the left of each heatmap (CD4+=circle, CD8+= triangles). (C) The similarity matrices shown as heatmaps in B are represented in two dimensions by NMDS. (D) Cosine similarity between CDR3NTβ chains across (triangles) and within individuals (between spleen and bone marrow, circles). T cells compartments (colored dots) are divided to CD4+ (upper) and CD8+ (lower) from young (left) and adult (right) mice repertories. Mean is shown by horizontal black lines. (E) The surface phenotype of Foxp3+ Tregs. The plot shows the percentage of Foxp3 positive cells (Treg) which have the phenotype: CD44- CD62L+ (naive-like), CD44+CD62L+ (central memory -like) or and CD44+CD62L- (effector-like). Mean is shown by horizontal black lines. Each data point represents one mouse. Significant differences between age groups or intra and inter individuals are denoted by asterisks (P-values: *<0.05, **< 0.01, ***< 0.001, with FDR correction t-test).

Figure 4

Differential frequency of amino acid motifs in TCR repertoires from different subpopulations. The most abundant amino acid triplets are selected by the mean frequency of each motif across all compartments and mice as described in materials and methods (A) PCA analysis of CDR3βAA triplet frequency distributions for CD4+(left) or CD8+(right) from young (upper) or adult (lower) mice (e.g., CD8+ effectors, BM, mouse 1). (B) Cosine similarity between the frequency distributions of the 350 most abundant CDR3βAA amino acid triplets between individuals (circles) or within individuals (between spleen and bone marrow, triangles). T cells compartments (colored dots) are divided into CD4+ (left) or CD8+ (right) from young or adult mice. Mean is shown by horizontal black lines. (C) Differentially expressed triplets in Treg and CD4+ effector from young and adult mice. Each dot represents a single triplet most 350 abundant or all 8000 triplets in red or black dots, respectively). P-value (t-test) was calculated for each triplet across six samples (three mice and 2 tissues) of CD4+ Treg and CD4+ effector cells. The y-axis shows FDR-adjusted p-values. The x-axis shows the log 2-fold-change, calculated between Treg and CD4+ effector mean triplets or motifs frequency across compartments (6 samples in each). Significance thresholds are marked in blue lines: (1) at y=1.3 (equivalent to p-value of 0.05) and x= ± 1 (denoting a total fold-change of 2). Representative triplets above both thresholds are labeled with red text and dots. (D) Significantly expressed triplets are found in various positions along the CDR3AA sequences. Triplets overexpressed in CD4+ Treg are frequently located in position 4 of the CDR3AAs (–9). Triplets overexpressed in CD4 effector can be located mainly in position 2-3 or further along the CDR3AA sequences. The color represents the log10 frequency of each aligned triplet.

Figure 5

Hierarchical impact of different immunological processes on repertoire diversification. The TCR repertoire is considered as evolving in four dimensions, captured by the diagram in (A). (B)The Davies-Bouldin (DB) index applied to Vβ frequencies, capturing the average separability (ratio of the within-cluster variance to the separation between cluster centroids (lower score means better clustering)) of clusters of different repertoires from their nearest counterpart. The reference distribution computed by assigning random clusters features (2 or 4 variables, red and blue lines respectively) to the same data and calculating the DB index 10000 times. (C–E) The mean inter-repertoire cosine similarity values of the Vβ gene distribution versus the mean inter-repertoire cosine similarity values for the 350 most abundant CDR3βAA triplets. (C, D) Each repertoire (spleen only) is compared to the young naive repertoires for CD4+, CD8+ (C) and CD4+ Treg (D). In E, each repertoire is compared to each other repertoire from the same compartment. The error bars represent SEM. Arrows show shift from young (red dots) to adult (blue dots).

Figure 6

The impact of LCMV infection on repertoire organization. (A) Summary of the LCMV induced T cell compartments and epitope –specific cells isolation for TCR repertoire sequencing and analysis. (B) The mean inter-repertoire cosine similarity values of the Vβ gene distribution versus the mean inter-repertoire cosine similarity values for the 350 most abundant CDR3βAA triplets. In the upper panel each repertoire (spleen only) from LCMV infected mice is compared to the young naive repertoires for CD4+ and CD8+ compartments. In the lower panel each repertoire (spleen only) is compared to each other repertoire from the same compartment. The error bars represent SEM. (C) Triplets over-represented in CD8+ effector repertoires after 8 days of LCMV infection compared to young uninfected healthy mice. Each dot represents a single triplet. The y-axis shows FDR-adjusted t-test p-values. The x-axis shows the log 2-fold-change, calculated between mean triplets from young and LCMV infected mice (6-8 samples in each). Significance thresholds are marked in blue lines: at y=1.3 (equivalent to p-value of 0.05) and x= ± 1 (denoting a total fold-change of 2). Representative triplets above both thresholds are labeled with red text and dots. Significantly enriched triplets that are labeled in red text are also found in the TCRs of epitope specific (NP396, NP205, and GP92 tetramer sorted) T cells.