Comparative analysis of murine T-cell receptor repertoires

Abstract

For understanding the rules and laws of adaptive immunity, high-throughput profiling of T-cell receptor (TCR) repertoires becomes a powerful tool. The structure of TCR repertoires is instructive even before the antigen specificity of each particular receptor becomes available. It embodies information about the thymic and peripheral selection of T cells; the readiness of an adaptive immunity to withstand new challenges; the character, magnitude and memory of immune responses; and the aetiological and functional proximity of T-cell subsets. Here, we describe our current analytical approaches for the comparative analysis of murine TCR repertoires, and show several examples of how these approaches can be applied for particular experimental settings. We analyse the efficiency of different metrics used for estimation of repertoire diversity, repertoire overlap, V-gene and J-gene segments usage similarity, and amino acid composition of CDR3. We discuss basic differences of these metrics and their advantages and limitations in different experimental models, and we provide guidelines for choosing an efficient way to lead a comparative analysis of TCR repertoires. Applied to the various known and newly developed mouse models, such analysis should allow us to disentangle multiple sophisticated puzzles in adaptive immunity.

Keywords: T cell; T-cell receptor repertoires; aging; diversity; functional T-cell subsets.

Figure 1

Extracting and comparing T‐cell receptor (TCR) repertoires. TCR repertoires can be extracted from targeted TCR sequencing (TCR‐seq) performed using genomic DNA or cDNA methods, with or without unique molecular identifiers (UMI), e.g. using MiGEC and MiXCR software tools. Alternatively, TCR repertoires can be extracted from bulk RNA‐seq data using MiXCR RNA‐seq mode. The latter approach works most efficiently for samples enriched with T cells or representing pure sorted T cells.

Figure 2

Comparison of the CDR3 length distributions for mice and humans. (a) T‐cell receptor‐β variable (TRBV) CDR3 length histogram for C57BL/6J mouse (3 months old) and human (male, 35 years old) peripheral blood mononuclear cells. CDR3 is defined starting from the last codon of TRBV (cysteine, position 92) and ending at the phenylalanine in the conserved TRBJ segment motif FGXG. (b) Added N‐nucleotides histogram. The data sets were processed using MiXCR, with correction for the probability of zero insertions.71

Figure 3

Analysis of T‐cell receptor‐β (TCR‐β) CDR3 repertoire diversity in young and old mouse peripheral blood mononuclear cells (PBMC). (a, b) Spearman correlations for the diversity metrics and age or naive T cells percentage in the blood of young (n = 8) and old (n = 8) mice. Black circles correspond to the MiXCR data sets. Red squares correspond to the unique molecular identifiers (UMI)‐based analysis (MIGEC+MiXCR). Correlations are shown for the non‐normalized data sets: all obtained UMI or sequencing reads (a), and normalized data sets (down‐sampled to 15 000 sequencing reads or 15 000 UMI) (b). (c) Schematic representation of the TCR repertoire diversity assessment using different diversity metrics. Examples of a young (left) and an old (right) mouse peripheral blood TCR‐β CDR3 repertoire are shown. Three circles represent observed diversity (dark green), Shannon–Wiener index (light green), and Chao1 (yellow) metrics. Segment of the TCR repertoire that contributes mainly to a metric is shown by the bold part of the circle. Observed diversity takes into account all clonotypes, Chao1 depends mostly on the representation of singletons and doubletons (clonotypes represented by one and two reads, correspondingly).70 Shannon–Wiener index characterizes structural complexity of the TCR repertoire based on assessment of the evenness of distribution of relatively abundant clonotypes. (d–f) Diversity metrics for TCR‐β CDR3 repertoires in young and old mice. Data sets were normalized by down‐sampling to 15 000 randomly chosen UMI. Two‐tailed unpaired Mann–Whitney test showed that all diversity estimates were significantly different between old and young mice. (g) Percentage of naive T cells of all CD3⁺ T cells in peripheral blood of young and old mice. Each symbol represents an individual mouse; small horizontal line indicates the mean. Two‐tailed unpaired Mann–Whitney test was applied. (h) Share of the whole repertoire occupied by the top 100 most frequent clonotypes.

Figure 4

Analysis of T‐cell receptor‐β (TCR‐β) CDR3 repertoire diversity in mouse tissues. (a–c) Diversity metrics (n = 6 mice). Data sets obtained for the samples from lymph nodes (LN), spleen (SP), thymus (THY) or peripheral blood mononuclear cells (PBMC) were normalized by down‐sampling to 30 000 randomly chosen unique molecular identifier (UMI)‐labelled cDNA molecules. Observed diversity (number of clonotypes per 30 000 T cells (a), Shannon–Wiener index (b), and Chao1 (c) metrics are shown. Kruskal–Wallis test showed that diversity metric values were significantly different across various tissues on each plot. (d) The average percentages of T‐cell subsets according to flow cytometry analysis. CD3‐positive lymphocyte gates were defined using CD62 and CD44 staining among thymocytes, splenocytes, PBMC and lymphocytes isolated from lymph nodes. (e–g) The same diversity metrics calculated without data normalization. (h) Correlation of observed TCR‐β diversity with the number of analysed UMI‐labelled cDNA molecules.

Figure 5

Visualizing repertoire overlaps. Multi‐dimensional scaling (MDS) analysis of the T‐cell receptor‐β (TCR‐β) CDR3 repertoires in young (red circles, n = 8) and old (blue circles, n = 8) mouse peripheral blood (upper panels). Metrics F2, R and D show that young (red dots) data sets show a high similarity to each other, whereas the old ones demonstrate distinctive features of the TCR‐β repertoires, due to the decreasing proportion of naive T cells and expansion of different antigen‐experienced clones with age. For the D metric, cluster analysis was restricted to the top 5000 clonotypes per sample. Bottom panels show schematic (for the F2 metric) and exemplary (for R and D metrics) pairwise analyses of samples overlap.

Figure 6

Analysis of the T‐cell receptor‐α (TCR‐α) repertoires on a fixed TCR‐β chain background. Amino acid TCR‐α CDR3 repertoires for the sorted T‐cell subsets of the Foxp3 gfp Tcra ^−/+ mice (n = 5) bearing the DO11.10 TCR‐β transgene were analysed. (a) Dendrogram showing ‘R’ overlap for the regulatory (green bars), naive (blue bars) and effector (red bars) CD4 T cells from spleen (dark grey bars), lymph nodes (LN, light grey) and thymus (brown). (b,c) Multi‐dimensional scaling (MDS) analysis of the TCR‐α repertoire overlaps using ‘F2’ (b) and ‘R’ (c) metrics. (d) Dendrogram showing ‘R’ overlap for the CXCR3‐positive and CXCR3‐negative T regulatory (green bars), and effector (red bars) CD4 T cells from spleen (dark grey bars) and lymph nodes (LN, light grey). (e) Amino acid composition characteristics of the CDR3 middle part. Analysis of the TCR repertoires derived from thymic (Thy), lymph node (LN) and splenic (Spl) CD4⁺ effector, CD4⁺ naive, and CD4⁺ regulatory T cells of Foxp3 gfp Tcra ^−/+ mice (n = 5) bearing the DO11.10 TCR‐β transgene. Weighted analysis is shown (i.e. the size of clonotypes was considered). (a–c) and (e) data are from ref. 46, (d) data are from ref. 47.

Figure 7

Analysis of the T‐cell receptor‐β variable (TRBV) gene segment usage. (a) Unweighted and weighted analysis of TRBV gene segment usage in old (n = 8) and young (n = 8) C57BL/6J mouse peripheral blood samples. (b) CD4/CD8 ratio in the young (n = 8) and old (n = 8) C57BL/6J mice. Two‐tailed Mann–Whitney test. (c) Unweighted and weighted analysis of TRBV gene segment usage in sorted thymic and splenic naive CD4 and CD8 T cells of 1‐year‐old C57BL/6J mice (n = 5). (d) Jensen–Shannon divergence of TRBV gene segment usage distributions in sorted thymic and splenic naive CD4 and CD8 T cells of 1‐year‐old C57BL/6J mice (n = 5).