In this study, we have presented a comprehensive methodology to leverage the variation of protein abundance within cell populations in order to quantify variability in cytokine sensitivity and investigate its underlying mechanisms. This method relies on the demonstrated ability of flow cytometry to precisely measure the abundance of multiple proteins simultaneously in single cells (). We have used this methodology to study receptor signaling in the γc chain family, uncovering unexpected relationships between related receptors.
During activation of isogenic T cells, the abundance of IL-2Rα varies by up to five orders of magnitude (17
). Our co-staining experiments verified that these measurements of IL-2Rα precisely define IL-2Rα abundance in these cells (, fig. S1
). This allowed us to study the effect of this broad range of IL-2Rα abundance on cytokine sensitivity, as defined by the EC50
of STAT5 phosphorylation stimulated by the cytokine. In agreement with previous findings, we found that IL-2Rα abundance strongly correlates with IL-2 sensitivity, with an EC50
decreasing from 100pM to 1pM as IL-2Rα abundance increases from 102
proteins (, fig. S2
). This is consistent with Cantrell & Smith's observation that cells endowed with higher IL-2Rα abundance proliferate more robustly in vitro (22
). Strikingly, we observed that this increase of IL-2 sensitivity is concomitant with a 20-fold decrease of IL-7 sensitivity (). The dependency of IL-7 sensitivity on IL-2Rα abundance could not be explained by anti-correlation of IL-7Rα and IL-2Rα abundance or by a potential inhibition of IL-7Rα signaling capacity as a function of IL-2Rα abundance within activated effector cells (fig. S3–S4
). Such anti-correlation of receptor abundance has, however, been observed both during T cells' transitions from naïve to effector and effector to memory stages in vivo: naïve cells have abundant IL-7Rα, but no IL-2Rα; memory cell precursors are marked by relatively high abundance of IL-7Rα (25
) and low abundance of IL-2Rα (18
). Additionally, we found that highly activated effector cells, with large amounts of IL-2Rα, are inhibited from sensing IL-7: this implies that they would rely chiefly on IL-2 for survival. This could reinforce the suppressive effect of IL-2 depletion by T regulatory cells, preventing IL-7 from rescuing effector cells when IL-2 becomes scarce (17
). In this context, the inhibitory effect of IL-2Rα abundance on IL-7 sensitivity represents an additional mechanism to enhance separation of cytokine specificities across different T cell subtypes, with naïve and memory cells relying chiefly on IL-7, while effectors rely on IL-2.
Addressing the impact of IL-2Rα abundance on IL-7 sensitivity at the computational level was made possible because this system is well characterized at the biophysical level (20
). Binding interactions between receptor subunits, and between receptor subunits and cytokines have been quantified (29
). However, because many of these measurements have been made in different systems and with varying methodologies—on the surface of cells or with soluble fractions— multiple models have been proposed for the function of these receptors (28
). Our CCVA methodology () allows us to add detailed quantitative data, across a large range of protein abundance, thereby placing tight constraints on any proposed model of IL-2 and IL-7 signaling. In the present work, we show that equilibrium binding models incorporating the measured affinity constants as priors quantitatively reproduce aspects of the dependency of IL-7 and IL-2 sensitivities on IL-2Rα (). The switch of sensitivities between IL-2 and IL-7 as a function of IL-2Rα abundance can be readily attributed to the preformation of the full heterotrimeric IL-2 receptor, which sequesters the γc
chain away from the IL-7 receptor. By comparing all possible equilibrium binding models of formation of signaling cytokine-receptor complexes, we have identified a minimal equilibrium binding model with affinity constants constrained around the measured values that accounts quantitatively for the CCVA data. This model was selected using a statistical criterion comprising a cost for deviation from both biochemically measured parameters and CCVA data, and penalizing for inclusion of a larger set of parameters. As such, this selection procedure does not exclude other models with more parameters, representing other paths of receptor formation (35
), but more data would be needed to justify their inclusion. The selected model presents the practical advantage of providing a straightforward interpretation of the scaling of IL-7 sensitivity as a function of IL-2Rα abundance. The increased EC50
at high IL-2Rα abundance is a direct consequence of the sequestration of the γc
chain in the preformed IL-2 receptor. As a consequence, a cytokine receptor subunit with a high affinity for a shared component (i.e. γc
or β) should make it difficult for the IL-2 receptor to sequester this component, and the effect of IL-2Rα abundance on the sensitivity to the corresponding cytokine should be weak.
Extending this minimal model to other cytokines of the γc
family, we recapitulated different measured effects of IL-2Rα on cytokine sensitivities (). Specifically, we found that IL-2Rα abundance negatively impacts sensitivity to IL-7 and to a lesser extent to IL-15, but not at all to IL-4 or IL-21. In our model, non-dependence of sensitivity to IL-2Rα for these two cytokines implies high affinity between the corresponding α chains and the shared component(s). The ability of each α chain to tune not only the sensitivity to its respective cytokine, but also cross-antagonism of other signals provides a functional separation between γc
-utilizing cytokines that matches their respective functions in the context of T cell differentiation (40
). IL-7 is dedicated to the homeostasis of naïve and memory cells, but not to the expansion of effectors, whereas IL-15 affects both effectors and memory cells (41
). Therefore, segregating the homeostatic functions of IL-7 from the expansion function of IL-2 may be advantageous (24
). In contrast, IL-4 and IL-21 are both involved in the differentiation of effector cells into helper subtypes, and are experienced during expansion (21
). For these two cytokines, cross-antagonism from large amounts of IL-2Rα would thus be functionally counter-productive.
Our CCVA method can be applied to virtually any cellular signaling investigation wherein cellular heterogeneity and signaling can be measured, such as other cytokines, chemokines, or growth factors (42
). The number of signaling regulators that can be simultaneously measured is limited only by technical constraints of flow cytometry. As the limitations of flow cytometry recede with the introduction of machines capable of recording dozens of parameters simultaneously (10
), our method is poised to identify and quantify regulatory effects in signaling networks. This technical framework, accompanied by quantitative modeling efforts, will improve mechanistic understanding of signal transduction cascades in living cells (2