ReportImmunology

Heterogeneous Differentiation Patterns of Individual CD8+ T Cells

See allHide authors and affiliations

Science  14 Mar 2013:
1235487
DOI: 10.1126/science.1235487

Abstract

Upon infection, antigen-specific CD8+ T lymphocyte responses display a highly reproducible pattern of expansion and contraction that is thought to reflect a uniform behavior of individual cells. We tracked the progeny of individual mouse CD8+ T cells by in vivo lineage tracing and demonstrated that, even for T cells bearing identical T cell receptors, both clonal expansion and differentiation patterns are heterogeneous. As a consequence, individual naïve T lymphocytes contributed differentially to short- and long-term protection, as revealed by participation of their progeny during primary versus recall infections. The discordance in fate of individual naïve T cells argues against asymmetric division as a singular driver of CD8+ T cell heterogeneity and demonstrates that reproducibility of CD8+ T cell responses is achieved through population averaging.

The murine naïve T cell repertoire contains approximately 100-1000 CD8+ T cells specific for a given antigen (1, 2), which upon antigen-recognition proliferate to produce up to 107 progeny (24). After pathogen clearance, the numbers of antigen-specific T cells decline by 90-95%, leaving behind a population of memory T cells (5). This characteristic T cell response kinetic is accompanied by differentiation into functionally distinct subsets (6, 7).

Although the patterns of T cell differentiation and response kinetics are highly reproducible for a given infection, it is unclear how this reproducibility is controlled. Single-cell tracing studies have unambiguously shown that individual naïve T cells are able to yield both effector and memory cells (8, 9). However, the relative contribution of individual naïve T cells to the effector and memory pool has not been determined. Reproducibility of T cell responses could be controlled at the single cell level, with each naïve T cell producing the same types and amount of progeny, and asymmetric division of T cells (10) would provide a mechanism to ensure such equal representation. Alternatively, individual T cell families (i.e., one naïve T cell and its progeny) may show distinct sizes or phenotypes, with reproducibility manifesting itself only at the population level. Both scenarios predict strikingly different mechanisms behind protection to renewed infection. In the former, a defined fraction of each family provides long-term memory, whereas in the latter, some T cell families could primarily convey short-term protection, whereas others could mostly yield memory cells.

Here, we aimed to distinguish between these two potential mechanisms for T cell reactivity during primary and renewed infections, by in vivo lineage tracing of individual naïve CD8+ T cells. Combination of a previously established DNA-barcode-based lineage tracing technology (9, 1113) with second-generation sequencing allowed quantification of individual T cell families with substantial accuracy (quantification/correction procedures: fig. S1 and (14)). To track individual T cells in a system in which differences in T cell receptor (TCR) affinity do not influence cell fate, naïve TCR-transgenic OT-I T cells (which recognize the SIINFEKL peptide presented on H-2Kb) were generated that each carry a unique DNA barcode (9). Mice transferred with physiological numbers of these cells were then infected with Listeria monocytogenes expressing SIINFEKL (LM-OVA), and barcode sequencing was used to quantify the progeny of individual naïve T cells.

Analysis of the contribution of individual T cells at the peak of the immune response revealed the number of progeny produced by individual T cells to be highly variable (Fig. 1A, controls in fig. S2) and, on average, the ‘dominant’ naïve T cell produced 400-fold more offspring than the median naïve T cell within the same animal (Fig. 1B). As a consequence, approximately 60% of the total OT-I T cell pool was derived from only 5% of naïve OT-I cells (Fig. 1C). While ~50% of OT-I families consisted of fewer than 200 daughter cells, these ‘mini-families’ had formed as a consequence of TCR triggering, rather than bystander or homeostatic proliferation (fig. S3A). The numerical dominance of a small number of T cell families was unrelated to rearrangement of the endogenous TCR loci, as LM-OVA-induced responses of OT-I Rag2−/− T cells were also biased toward few families (fig. S3B-D). Furthermore, family size disparity was independent of the number of responding OT-I families (range evaluated: 17 to 874 families/animal, fig. S3E), arguing against competition between the transferred cells as a confounding factor.

Fig. 1 Single-cell output disparity within CD8+ T cell responses.

Recipients of 800 naïve, barcode-labeled OT-I T cells were infected with LM-OVA. At day 7 post-infection, OT-I T cell family sizes in spleen and LN were quantified. (A) Average size of the largest to smallest OT-I family, represented as fraction of the overall OT-I T cell response derived from on average 228 naïve T cells. (B) Number of OT-I cells constituting the largest and median OT-I family. Symbols: individual mice. Bar: average. (C) Average (+SD) fraction of overall OT-I response formed by the indicated fractions of OT-I families, ordered by dominance. Data from one experiment (4 mice), representative of 13. (D-E) Disparity between T cell families in recipients receiving (D) 1000 or (E) 640 barcode-labeled OT-I cells and infected with (D) indicated doses of LM-OVA and (E) LM-OVA (N4) or an LM-OVA variant expressing SIITFEKL (T4). Left: OT-I T cell response size (# GFP+ cells). Middle: Fraction (mean + SD) of OT-I T cell response formed by the indicated fractions of OT-I families, ordered by dominance. Right: Index of disparity for OT-I family sizes. Upper limit of the index of disparity is 1, occurring when one family dominates. lower limit is 0, occurring when all families occupy an equal fraction. Data from one experiment per experimental setting (≥4 mice/ group), representative of 2 each. (A-D) Lineage tracing quality controls in fig. S2.

To evaluate whether these results were influenced by the retroviral integration sites of barcodes we developed a transgenic mouse strain (BCM) that allows DNA barcode generation in vivo at a defined genomic locus (fig. S4A). After LM-OVA infection, OT-I T cells harboring such endogenously generated barcodes also displayed a marked variability in clonal output (fig. S4B-C). Thus, T cell fate is also disparate when cells are tagged at a specific genomic site.

Strong disparity in the number of progeny of different naïve T cells constitutes a robust feature of T cell responses, observed upon infection with different doses of LM-OVA (Fig. 1D), in high- and low-affinity TCR-antigen interactions (Fig. 1E) and upon pulmonary influenza infection (fig. S5). Kinetic analyses revealed that clonal disparity was already apparent at day 5 and remained similarly pronounced during the following days (Fig. 2A-D, fig. S6). An appealing explanation would be that naïve T cells enter their first division at different time-points post infection and hence differ in the time window of clonal expansion. However, mathematical modeling based on CFSE-data suggests that the time of first division cannot be the main decisive factor (fig. S7). Furthermore, strong variation in family size was also observed when naïve T cells were transferred 48h after infection (fig. S8), indicating that individual ‘late-comer T cells’ (15) also show disparate behavior.

Fig. 2 Kinetics of T cell family disparity and reproducibility of T cell responses.

Recipients of 1600 naïve, barcode-labeled OT-I T cells were infected with LM-OVA. At the indicated days post-infection, spleen and LNs were harvested and individual OT-I T cell family sizes quantified. (A) OT-I T cell response size (# GFP+ cells). (B) Number of T cell families detected. (C) Number of OT-I T cells constituting the largest (circles) and median (triangles) OT-I family for each day. (D) Index of disparity for OT-I family sizes. Data are from one experiment (4 mice per group), representative of 2 independent experiments. Lineage tracing quality controls are shown in fig. S6. (E) Reproducibility of OT-I response size (90% confidence interval, out of 100 simulations) as a function of the number of contributing OT-I families. Note that T cell response magnitude only becomes reproducible within a narrow range when the output of more than approximately 50 individual T cells is averaged.

Collectively, the above experiments reveal that the numerical output of individual T cells is highly variable. As a consequence, the magnitude of T cell responses only becomes reproducible when the output of some 50 T cell families is combined (Fig. 2E). This heterogeneity appeared consistent with two distinct scenarios for the formation of T cell memory. In a first scenario, some naïve T cells are imprinted to expand more strongly during the primary expansion, and the resulting families would remain similarly dominant during recall infection, or become increasingly prominent should this imprinting still exist (fig. S9A). In an opposite model, family size disparity reflects a ‘division of labor’, in which some naïve T cells mainly contribute offspring for primary responses whereas others primarily seed T cell memory (fig. S9B).

The second model predicts that different T cell families show significant variation with respect to phenotype and - more directly – with respect to contribution to recall responses. To evaluate this, we first separated barcode-labeled OT-I T cells after LM-OVA infection on the basis of expression of three different markers of T differentiation state, KLRG-1, CD27 and CD62L. Analysis of the distribution of individual T cell families over these subsets revealed that individual T cell families showed pronounced variation in phenotype, as revealed by the fraction of cells expressing KLRG-1, CD27 or CD62L (Fig. 3A, fig. S10). Thus, not only the magnitude but also the phenotypic properties of antigen-specific T cell responses are controlled ‘by averaging’ disparate behaviors of individual T cell families.

Fig. 3 Individual T cells yield phenotypically distinct progenies.

Recipients of 500-700 naïve, barcode-labeled OT-I T cells were infected with LM-OVA. At day 11 post-infection, spleen and LNs were harvested. OT-I T cells were isolated on the basis of expression of KLRG1, CD27 or CD62L by cell sorting, and T cell family sizes in the indicated phenotypic subsets were quantified. Plots depict pooled data from 4 mice. (A) For each T cell family, the fraction of members expressing the indicated marker is depicted. (B) For each T cell family, the fraction of members expressing the indicated marker is depicted as a function of the relative size of this family. Lineage tracing quality controls are provided in fig. S10. The Spearman correlation r was calculated for all families contributing to the marker-positive and -negative populations. Data are from one experiment (4 mice) per sort, representative of 2 independent experiments per sort.

We next analyzed whether T cell family size was correlated to the differentiation state of its members. This revealed a strong inverse correlation between T cell family size and the percentage of CD62L-expressing cells (Fig. 3B), indicating that clonal expansion and loss of CD62L expression may share a common regulator, as previously proposed (16). In contrast, the fraction of KLRG-1-positive T cells showed a weaker (positive) correlation with family size. Furthermore, the fraction of CD27-positive cells and family size were only marginally correlated, both at days 7 and 11 (Fig. 3B), indicating differential regulation of family size, KLRG-1 and CD27 expression.

To directly evaluate whether the offspring of individual T cells contributes differently to early protection and long-term immunity, we quantified T cell families longitudinally within the same animal. First, comparison of clonal dominance patterns during primary response and resting memory phase showed that the extent of contraction (i.e., the fraction of cells that survived into the memory phase) was essentially independent of family size (fig. S11). Subsequently, we quantified family sizes during three consecutive infections (Fig. 4A). Unsupervised clustering showed that barcode distribution was much more alike between secondary and tertiary infections than between any of these and the primary T cell response (Fig. 4B). Comparison of the size of individual T cell families during primary and secondary infection revealed that many of the dominant T cell families during the primary immune response were present at substantially lower frequency during secondary infection and vice versa (Fig. 4C). This highly variable participation in secondary responses reflected an intrinsic difference between T cell families in their ability to re-expand, rather than stochastic variation, as the pattern of clonal dominance during the secondary response was very well preserved upon tertiary infection (Fig. 4D-E). In line with this, when memory T cells from one mouse were transferred into two recipient mice, the patterns of clonal dominance were similar in these paired recipients upon subsequent infection, and - as expected (Fig. 4C) - distinct from the pattern observed during the primary T cell response within the matched donor mouse (Fig. 4 F-H, fig. S12).

Fig. 4 Division of labor between antigen-specific T cell families.

(A) Experimental setup and lineage tracing controls for (B-E). (B) Euclidian distance heatmap representing contribution of individual naïve T cells (rows) to the primary-tertiary response (columns). Data from one representative mouse of 2 experiments with at least 5 mice/experiment. (C) Comparison of family size (barcode frequencies) between primary and secondary response. (D) As (C), but for secondary and tertiary response. Data for C-D pooled from 5 mice of one experiment, representative of 2, with each mouse receiving 500-600 cells. (E) Spearman correlation r between family sizes in indicated samples, calculated for individual mice. Pooled data from 2 experiments. Significance of differences analyzed by paired t test. (F-H) Recipients of barcode-labeled OT-I T cells were infected with LM-OVA. At d59, CD8+ T cells were transferred into secondary recipients, followed by LM-OVA infection (fig. S12B). OT-I family sizes were quantified at d8 after primary and d6 after secondary infection. (F) Comparison of family size between primary and secondary responses (G) Comparison between matched secondary recipients. Data for F-G pooled from 4 mice of one experiment, representative of 2, with each mouse receiving 500-600 cells. (H). Spearman correlation, as in (E). Pooled data from 2 experiments. Depicted barcode frequencies are normalized to 105 per half-sample. Read counts of 0 set to 0.05 for log-scale visualization. Grey lines: cut-off for barcode-detection. Spearman correlations were calculated for all families contributing to both depicted responses. Lineage tracing controls in fig. S12.

Our experiments demonstrate a division of labor between T cell families in immune protection to renewed infection (S9B). As T cell families show disparate behavior along multiple axes that are only partially correlated (e.g., clone size versus fraction CD27+, Fig. 3), we consider it unlikely that variation in one single factor is responsible for this heterogeneity. Furthermore, our data imply that at least some T cell fates must be heritably imprinted before clonal expansion becomes substantial. Specifically, the observation that T cell families of similar size can either consist of cells that are largely devoid of CD27, or that are essentially all CD27-positive, implies an early fixation of this fate.

Collectively, our data and the data from Buchholz et al. (17) demonstrate that the behavior of individual T cells varies markedly with respect to clonal expansion, differentiation and recall capacity. This variability may be due to both intrinsic stochastic variation, as has been observed for B cells in vitro (18), and to variation in extrinsic signals, such as local differences in antigen-presenting cells or cytokines. In either case, the current observations exclude models in which each naïve T cell yields progeny with the same distribution of cells with either short- or long-term potential. Thus, although our data do not evaluate the role of asymmetric division as a mechanism to generate daughter cells with different fates, they do show that asymmetric division by itself cannot explain the disparity between individual T cell families that is experimentally observed. Rather, a strong variation between families in the expansion of proximal and distal daughter cells needs to be invoked to arrive at the heterogeneity observed here. Finally, the observation of strong heterogeneity at the single cell level indicates that T cell responses are made up of ‘averages’, a behavior reminiscent of recent analyses of stem cell renewal (19). Thus, although the differentiation/expansion of the combined T cell population follows a uniform course, the fate of individual naïve T cells is unpredictable.

Supplementary Materials

www.sciencemag.org/cgi/content/full/science.1235487/DC1

Materials and Methods

Supplementary Text

Fig. S1

Table S1

References (2023)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: The authors thank A. Pfauth and F. van Diepen for cell sorting; E. Borg for technical assistance; D. Zehn for provision of LM-OVA (T4); S. Harkema for helpful discussions; and G. Bendle, M. C. Wolkers, and E. A. Moseman for critically reading the manuscript. The data presented in this manuscript are tabulated in the main paper and in the supplementary materials. The authors declare no conflict of interest. This research was funded by ERC AdG Life-his-T to T.N.M.S., a Walter-Hitzig-fellowship from the Center of Chronic Immunodeficiency (CCI) (BMBF 01EO0803) and a Deutsche Forschungsgemeinschaft (DFG RO 4120/1-1) research fellowship to J.C.R., a Marie Curie Intra European Fellowship to L.P., a NWO Veni grant (916.86.080) to J.B.B., and a LLS Career Development Fellowship to S.H.N.
View Abstract

Navigate This Article