The Rules of Human T Cell Fate in vivo

Abstract

The processes governing lymphocyte fate (division, differentiation, and death), are typically assumed to be independent of cell age. This assumption has been challenged by a series of elegant studies which clearly show that, for murine cells in vitro, lymphocyte fate is age-dependent and that younger cells (i.e., cells which have recently divided) are less likely to divide or die. Here we investigate whether the same rules determine human T cell fate in vivo. We combined data from in vivo stable isotope labeling in healthy humans with stochastic, agent-based mathematical modeling. We show firstly that the choice of model paradigm has a large impact on parameter estimates obtained using stable isotope labeling i.e., different models fitted to the same data can yield very different estimates of T cell lifespan. Secondly, we found no evidence in humans in vivo to support the model in which younger T cells are less likely to divide or die. This age-dependent model never provided the best description of isotope labeling; this was true for naïve and memory, CD4 + and CD8 + T cells. Furthermore, this age-dependent model also failed to predict an independent data set in which the link between division and death was explored using Annexin V and deuterated glucose. In contrast, the age-independent model provided the best description of both naïve and memory T cell dynamics and was also able to predict the independent dataset.

The Impact of Model Assumptions in Interpreting Cell Kinetic Studies

“…with the explicit models showing good accuracy albeit with very wide credible intervals. However, upon widening the prior from [0, 1] to [0, 10] we found that the accuracy of the explicit kinetic heterogeneity models broke down resulting in very inaccurate estimates and extremely wide credible intervals (Fig. 7B).…”

“…The one exception to these priors is in Figure 7B where we explored the impact of prior assumptions. Here the prior distribution for p in the homogeneous model is [0, 10]. For the implicit heterogeneity model, priors are [0, 10] for p and d ∗ .…”

“…Here the prior distribution for p in the homogeneous model is [0, 10]. For the implicit heterogeneity model, priors are [0, 10] for p and d ∗ . For the explicit heterogeneity model with N = 2, priors are [0, 10] for p 1 and p 2 (the prior for α 1 is unchanged at [0, 0.5]).…”

Cyton2: A Model of Immune Cell Population Dynamics That Includes Familial Instructional Inheritance

“…The model is informed by, and the worked examples are for, data from in vitro experiments where stimulation is provided to a group of T or B cells, and the resulting proliferation occurs in a burst that can be followed by division tracking dyes or direct filming. Those population dynamics follows the pattern of an exponential rise, a period of division cessation, and then of cell loss that characterises immune responses in vivo ( Veiga-Fernandes et al, 2000 ; Costa Del Amo et al, 2020 ). As such, as with the original Cyton model ( Hawkins et al, 2007 ; Subramanian et al, 2008 ; Marchingo et al, 2014 ), Cyton2 can be successfully fit to in vivo data (data not shown).…”

The dynamics and longevity of circulating CD4+memory T cells depend on cell age and not the chronological age of the host

“…The temporal heterogeneity (TH) model assumed that each memory subset (T CM /T EM , host/donor) divides at its own uniform rate, but that Ki67 high cells, which are actively dividing or divided recently, are lost at a different rate to that of more quiescent Ki67 low cells. This is an instance of a more general model in which a cell's risk of death changes with the time since its last division [32]. Our TH model is itself a simple generalisation of a purely homogeneous model in which cells divide and die at constant, uniform rates, independent of their expression of Ki67.…”