Figure 1. Statistics of human T cell repertoire organization.

(A) T cells with highly diverse receptors are created from progenitor cells through genetic recombination (left), which then undergo clonal selection (middle) together shaping the immune repertoire. The T cell receptor (TCR) locus acts as a natural barcode for clonal lineages, which can be read out by sequencing (right). (B, C) Clone size distributions in two large cohort studies of human blood samples using disparate sequencing protocols display a power-law relationship between the rank and size of the largest clones. Each line shows the size distribution of all T cell clones in an individual in an unsorted blood sample, that is independently of the phenotypes of the cells making up the different clones. Ages are color coded as indicated in the legend. The black line shows a power law with a slope of -1 for visual comparison. Normalized clone sizes were defined as the number of reads of a given receptor’s sequence divided by the total number of reads within a sample and a factor equal to the average fraction of T cells with memory phenotype at different ages to account for variations in sampling depth and in the subset composition of peripheral blood, respectively (Figure 1—figure supplement 3). Only a single individual is displayed per 2-year age bracket to improve visibility. (D) Power-law exponents as a function of the age (legend: linear regression slope and coefficient of determination). Data sources: Britanova et al., 2016, Emerson et al., 2017.

Figure 1—figure supplement 1. Cohort age distributions.

Figure 1—figure supplement 2. Clone size distributions in phenotypically sorted T cell subsets.

(A) The naive cell fraction as determined by flow cytometry and the fraction of singletons are closely correlated in the Britanova cohort. To diminish the influence of sampling depth variations, we computationally subsampled all repertoires to an equal sample size of 5 · 10⁵ counts. (B,C) Analysis of unsorted (TCR sequencing from all peripheral blood mononuclear cells), memory (CD3+, CD45+), and naive (CD3+, CD45RA+) blood samples from the same individual (Data source: Chu et al., 2019). (B) Clone size distributions in the different T cell compartments. Filtering naive clones that are also found in the memory compartment removes most large naive clones. (C) Frequency of large clones in the memory sample is shifted upwards relative to their frequency within the unsorted sample. Color represents logarithm of local kernel density estimate in regions with overplotting. The solid lines are guides to the eye (black line represents equal frequency, green line 2.6-fold higher frequency in the memory compartment). (D) The fraction of naive cells decreases with age (Data source: Britanova et al., 2016) starting in early infancy (Data source: Pediatric AIDS Clinical Trials Group et al., 2003). The legend shows the fitted time constant of exponential decay (± SE).

Figure 1—figure supplement 3. Influence of normalization procedure on clone size distributions.

(A,D) Raw clone size distributions show large variability due to different sample sizes. (B,E) A normalization by sampling depth removes much of this variation. (C,F) A normalization by the fraction of memory cells at different ages further collapses the tails of the clone size distributions.

Figure 1—figure supplement 4. Dependence of power-law exponent on age by cytomegalovirus (CMV) infection status and sex.

(A) Chronic infection with CMV drives large clonal expansions (Lindau et al., 2019; Sylwester et al., 2005). We thus repeated the analysis of Figure 1C separating individuals based on their CMV infection status (fitted lines shown in legend, regression results displayed as offset + slope · (age in years - 40)/10). Overall, CMV-positive individuals have a smaller α than uninfected individuals, which is independent of age. The average exponent in CMV-negative individuals decreases slowly with age, and in old age coincides those of CMV-positive individuals. Combining CMV infection status and age explained a significantly larger proportion of the variance in scaling exponents (17%) than age alone. (B) Many immune determinants differ markedly between the sexes (Klein and Flanagan, 2016). We thus analyzed whether α depends on sex. We find that the dependence on age is similar among the sexes, but men have on average a slightly smaller exponent than women indicating a more skewed repertoire organization. Data source: Emerson et al., 2017.

Figure 1—figure supplement 5. Clone size distributions in cordblood.

Each line shows the distribution in one individual. The black line shows a power law with a slope of -1 for visual comparison. The fitted power-law exponents α = 2.1 ± 0.1 (mean ± SE) are larger than in adult repertoires, but clone sizes are already remarkably broad. Data source: Britanova et al., 2016.