Light-based tuning of ligand half-life supports kinetic proofreading model of T cell activation

T cells are thought to discriminate stimulatory versus non-stimulatory ligands by converting small changes in ligand binding half-life to large changes in cell activation. Such a kinetic proofreading model has been difficult to test directly, as existing methods of altering ligand binding half-life also change other potentially important biophysical parameters, most notably the stability of the receptor-ligand interaction under load. Here we develop an optogenetic approach to specifically tune the binding half-life of a light-responsive ligand to a chimeric antigen receptor without changing other binding parameters. By simultaneously manipulating binding half-life while controlling for receptor occupancy, we find that signaling is strongly gated by ligand binding half-life. Our results provide direct evidence of kinetic proofreading in ligand discrimination by T cells. One Sentence Summary Direct control of ligand binding half-life with light shows that lifetime, not occupancy, dominates T cell activation.


Main Text:
The T cell response begins when the T cell antigen receptor (TCR) binds its cognate peptide-major histocompatibility complex (pMHC) on an antigen-presenting cell. Ligand discrimination is essential, as falsely recognizing a self-peptide can lead to autoimmunity, while missing a foreign-peptide allows pathogens to evade detection. Compounding this challenge, 5 self-pMHCs generally outnumber foreign-pMHCs by several orders of magnitude (1-4). How the TCR recognizes its foreign-pMHC while ignoring the more numerous self-pMHC is a fundamental open question in immunology.
For many cell surface receptors, occupancy drives the strength of signaling. Lower affinity ligands can signal as well as high affinity ligands when their concentration is increased 10 such that an equal number of receptors are bound (5). This is unlikely to be the case for activation of the TCR, because low affinity pMHCs are significantly less potent than high affinity pMHCs, even when adjusted for occupancy (6). Furthermore, given the large excess of self-pMHC (1-4) and the relatively small differences in binding affinity (7)(8)(9), it is likely that more TCRs are bound to self-pMHC than foreign-pMHC. For a T cell to detect foreign pMHCs, 15 it is likely that some ligand-intrinsic factor makes bound foreign-pMHCs more stimulatory than bound self-pMHCs.
A major model of T cell ligand discrimination is kinetic proofreading (10), which predicts that small differences in ligand binding half-life can be amplified into large differences in signaling. Such a model is attractive because it allows a few receptors bound to long-lived 20 ligands to signal better than many receptors bound to short-lived ligands, potentially explaining why abundant self-pMHCs do not activate T cells, though scarce foreign-pMHCs do. It posits a delay made up of multiple irreversible biochemical steps between ligand binding and 3 downstream signaling. Only ligands that bind persistently signal effectively. These steps only occur while the ligand is bound and quickly reset upon dissociation, preventing two successive short binding events from mimicking one long binding event. This mechanism is based on the kinetic proofreading that gives DNA replication, protein translation (11,12), and mRNA splicing (13) specificity far beyond what would be predicted from equilibrium binding constants alone. 5 Consistent with this hypothesis, kinetic measurements generally show that stimulatory pMHCs have longer binding half-lives, while non-stimulatory pMHCs have shorter binding half-lives (7-

9).
Kinetic proofreading effects need to be strongest when differences in pMHC binding affinities are small, as abundant self-pMHCs will bind more TCRs than scarce foreign-pMHCs. 10 Strong kinetic proofreading ensures that short self-pMHC binding events are unlikely to contribute to downstream signaling. However, recent techniques that measure pMHC binding affinities in more native 2D environments show a larger range of binding affinities than previously measured (14), suggesting that there may not be as large an excess of self-pMHC bound to the TCR as previously thought and raising the question on whether ligand half-life or 15 receptor occupancy dominates ligand discrimination.
To probe kinetic proofreading in T cells, the field most commonly relies on altered peptide series to change binding half-lives. The limitation of this approach is that when the binding interface is altered, both the bond's half-life and stability under tension must change together ( Figure 1A, left). Differentiating the effects of these two variables is essential because 20 recent evidence has shown that the TCR is mechanosenstive (15)(16)(17) and that foreign-pMHC can form catch bonds with the TCR, growing more stable under load (18). Thus, conventional mutational approaches make it difficult to determine whether changes in signaling arise from 4 alterations in binding half-life or reflect changes in the force being applied to the TCR. To specifically probe the role of binding half-life in ligand discrimination, the field needs new techniques to manipulate the receptor-ligand half-life while leaving force transmission and all other aspects of the interaction unchanged. Here we develop an optogenetic approach for T cell activation that addresses this need. 5 For optogenetic control of ligand half-life and T cell activation, we constructed a chimeric antigen receptor (CAR) that binds its ligand in a light-dependent fashion. CARs are single chain mimetics of the native, multi-subunit TCR complex and minimally consist of an extracellular binding domain, a transmembrane domain and an intracellular signaling domain.
Because CARs are not restricted to binding pMHCs, they can be engineered to signal in response 10 a wide variety of ligands by changing the extracellular domain. Once bound, they initiate downstream signaling by becoming phosphorylated and recruiting Zap70 to the receptor (19).
For generating a light-responsive ligand, we choose the LOVTRAP system (20) because it is one of only two optogenetic systems where light drives the dissociation of two proteins on the order of seconds, the approximate physiological range of pMHC half-lives (7-9). By 15 presenting the blue-light sensitive protein LOV2 on a supported lipid bilayer (SLB) and fusing its binding partner (Zdk) to the extracellular domain of a CAR, we generated a CAR with a lightgated ligand. In the dark, the Zdk-CAR binds the LOV2 ligand with nanomolar affinity (20).
When excited by a photon of blue-light (<520 nm), LOV2 changes conformation and unbinds the receptor. Thus, the intensity of blue light controls the length of time individual molecules of 20 LOV2 and the CAR interact. High intensity blue light enforces a short binding half-life, while low intensity blue light enforces a long binding half-life. This optogenetic approach enables one receptor-ligand pair to explore a range of binding half-lives while maintaining an identical 5 binding interface, ensuring the bond's stability under tension remains unchanged. (Figure 1A, We first validated the ability of the LOV2 ligand to photoreversibly bind the Zdk-CAR. Clonal Jurkat cells stably expressing the Zdk-CAR were exposed to SLBs functionalized with purified Alexa-488-labeled LOV2 ( Figure 1B). Because LOV2 diffuses freely in the bilayer and We noticed that within the same field of view, cells could have high or low-toundetectable receptor occupancies (red and white arrows, respectively; Figure 1C, D). However, despite clear differences in receptor occupancy, all cells had similar DAG levels ( Figure 1E). 5 This result suggested that something other than receptor occupancy dominates downstream signaling.
As blue light affects the LOV2-Zdk binding half-life, we sought to measure this relation in more detail. The release of purified Zdk from SLBs functionalized with LOV2 was measured upon acute exposure to different blue light intensities ( Because half-life is a strong predictor of DAG levels across a range of receptor occupancies, these data argue that binding half-life is a major determinant of CAR signaling. 20 Combining the data from all experiments over a range of LOV2 ligand concentrations, we sought to quantify how strongly binding half-life influences CAR signaling. In kinetic proofreading, the delay between ligand binding and downstream signaling is modeled as a series 8 of discrete steps (10). The stronger the proofreading, the greater the number of steps and the more binding half-life dominates downstream signaling. To minimize the number of assumptions, we modeled the DAG response as a saturable system downstream of "strong" proofreading steps, i.e. the forward biochemical reaction is sufficiently slow that the probability of completing a step is proportional to the ligand binding half-life. In such a model, 5 ∝ where R is the receptor occupancy, τ is the enforced ligand binding half-life, n is the degree of proofreading, K is the amount of upstream signal that generates a half-maximal DAG response, and β is basal signaling through the pathway. Fitting our data, we find n = 2.7 ± 0.5 (95% CI; levels. While this result is only true when the signaling pathway is far from saturation, it is reasonable that T cells operate far from saturation when they encounter foreign-pMHCs, given the non-stimulatory nature of self-pMHCs. The parameter n can also be interpreted as the 9 minimum number of proofreading steps in the pathway, indicating there at least three steps (between CAR ligation and DAG signaling) whose forward reaction is in kinetic competition with ligand dissociation (Figure 4B).
How much of ligand discrimination do our results account for? As T cells can respond to between 1-10 foreign-pMHCs (4,28) in the context of ~10 5 self-pMHCs (1-3), foreign-pMHCs 5 must signal at least 10 4 -10 5 fold stronger than self-pMHCs. Given that differences in binding half-lives between self-pMHCs and foreign-pMHCs are on the order of 10-fold (7)(8)(9)14), our observed degree proofreading can account for much, but not all, of ligand discrimination. T cells likely use kinetic proofreading in combination with other mechanisms, such as co-receptor involvement (4), mechanical forces (15,16,18), or signal amplification downstream of DAG 10 (29), to fully discriminate self from non-self.
What biochemical steps underlie the observed kinetic proofreading? We addressed this question by measuring one of the earliest signaling events: the recruitment of Zap70 to the CAR.
Zap70 binds to doubly phosphorylated ITAMs in the CAR's cytoplasmic domain. In contrast to DAG levels, Zap70 recruitment showed no detectable evidence of kinetic proofreading ( Figure   15

4-figure supplement 3).
In other words, Zap70 recruitment was solely dependent on CAR occupancy and was unaffected by the ligand binding half-life. As live cell reporters reflect the cumulative effect of all upstream signaling steps, these results argue that the observed kinetic proofreading occurs somewhere downstream of Zap70 recruitment and upstream of DAG production. 20 This result is surprising, as it shows that ligand discrimination is not complete by the time the CAR signals to downstream components. Rather, proximal signaling steps downstream of ITAM phosphorylation are essential for kinetic proofreading in our system. This explanation 10 runs counter to the conventional view that the proofreading steps correspond to the sequential phosphorylation of the ITAMs and that the TCR alone fully discriminates pMHCs, passing its decision to the proximal signaling steps. In contrast, our data argue that the proximal signaling steps kinetically discriminate the ligands while ITAM phosphorylation is insensitive to the ligand binding half-life. 5 While our results cannot exclude the possibility that TCR ITAM phosphorylation is sensitive to ligand binding half-life, they strongly suggest that the TCR does not fully discriminate ligands on its own. This is consistent with data showing that some Zap70 is associated with phosphorylated ITAMs in mouse thymocytes in vivo, even when they are not stimulated (30,31). Thus, even in primary cells, ligand discrimination is not complete at the 10 level of Zap70 recruitment.
Even if the proofreading steps do not occur at the level of receptor activation, they still need to occur in physical proximity to the receptor so that they terminate quickly upon ligand dissociation. Candidate biochemical steps include the phosphorylation and activation of Zap70 (32,33) and/or the phosphorylation of LAT. Lck was recently shown to act as a bridge between 15 Zap70 and LAT (34), suggesting that at least some fraction of LAT molecules could be tethered to phosphorylated ITAMs. Future experiments should focus on how mutations that change the kinetics of these biochemical steps affect ligand discrimination.
In summary, we addressed a fundamental question in immunology: How do T cells discriminate ligands? We overcame previous technical limitations by developing a new 20 optogenetic approach that directly tunes ligand binding half-life while keeping constant all other parameters that could affect CAR signaling. We found direct evidence that ligand binding halflife strongly controls downstream DAG levels but not upstream signaling events, like Zap70 11 recruitment. Additionally, our experiments with a CAR revealed a similar degree of kinetic proofreading as complementary optogenetic experiments that signaled through the TCR (see the co-submitted Schamel group paper). These results argue against the common view that kinetic proofreading occurs exclusively at the level of TCR ITAM phosphorylation and instead provide evidence that it can emerge from steps in the proximal signaling pathway. 5 To date, optogenetics has been a powerful approach because it tells us how the timing, amount, and localization of an ensemble of signaling molecules modulate cell signaling (35).

Competing interests: Authors declare no competing interests. Data and materials availability:
All data is available in the main text or the supplementary materials.  10 experiments, there is a fixed concentration of LOV2, meaning that both receptor occupancy and binding half-life change in response to blue-light. We measured Spearman's correlation coefficient to ask whether DAG levels were best described as a function of receptor occupancy or binding half-life across different LOV2 concentrations. DAG levels are more correlated with binding half-life (E) than they are with CAR occupancy (D). This is reflected in the fact that the 15 DAG response curves nearly overlap with each other when plotted as a function of binding half- 17 life, indicating changing the concentration of LOV2 has little effect on downstream signaling.
These data are consistent with binding half-life, not receptor occupancy, dominating CAR signaling. Plotted is the mean with a 95% CI (two-sided Student's t-test).