Abstract
MHC proteins that present peptide ligands for recognition by TCR form nanoscale clusters on the cell membrane of APCs. How the extent of MHC clustering controls productive TCR engagement and TCR-mediated signaling has not been systematically studied. To evaluate the role of MHC clustering, we exploited nanoscale discoidal membrane mimetics (nanolipoprotein particles) to capture and present peptide–MHC (pMHC) ligands at various densities. We examined the binding of these model membrane clusters to the surface of live human CD8+ T cells and the subsequent triggering of intracellular signaling. The data demonstrate that the proximity of pMHC ligands, high association rate of CD8–MHC interactions, and relatively long lifetime of cognate TCR–pMHC complexes emerge as essential parameters, explaining the significance of MHC clustering. Rapid rebinding of CD8 to MHC suggests a dual role of CD8 in facilitating the T cells’ hunt for a rare foreign pMHC ligand and the induction of rapid T cell response. Thus, our findings provide a new understanding of how MHC clustering influences multivalent interactions of pMHC ligands with CD8 and TCR on live T cells that regulate Ag recognition, kinetics of intracellular signaling, and the selectivity and efficiency of T cell responses.
Introduction
Major histocompatibility complex proteins bearing peptide Ags form nanoscale clusters on the surface of APCs (1–3). The clusters contain 2–20 of tightly packed MHC molecules (3–5). Although it is thought that MHC clustering could control Ag presentation and T cell responses (6), rigorous analysis of the effect of MHC clustering has been challenging. Available approaches (1, 7) do not allow to systematically evaluate how the clusters’ size, density, and composition influence TCR engagement and triggering of TCR-mediated signaling.
To model membrane peptide–MHC (pMHC) clusters, we used nanolipoprotein particles (NLPs) containing functional nickel-chelating lipids (NiNLPs), allowing conjugation of His-tagged proteins (8, 9). NLPs are nanoscale discoidal membrane mimetics that are readily self-assembled upon mixing purified lipids and apolipoproteins and provide a versatile platform for in vitro and in vivo applications (10–12). The density of the attached proteins can be reliably controlled by changing the percentage of nickel-chelating lipids in the NiNLP bilayer. These combined attributes make NiNLPs an ideal platform to evaluate how pMHC clustering affect Ag recognition and T cell signaling.
In this study, we analyze how the density of stimulatory (cognate) and nonstimulatory (noncognate or self) pMHC ligands within model membrane clusters influences their equilibrium binding and association kinetics with live CTL as well as the induction of TCR-induced Ca2+ signaling. The latter controls expression of three quarter of the genes involved in T cell activation (13).
Our results demonstrate that the density of pMHCs, not just their numbers, regulates multivalent interactions of the model pMHC clusters with TCR and CD8 on the T cells. Elucidating the role of MHC clustering provided basis for understanding of how variations in MHC clustering regulates Ag recognition and efficiency of T cell responses against cancerous and virus-infected cells.
Materials and Methods
Cells
Human CTL clone 68A62 recognizing HIV-derived peptide ILKEPVHGV (IV9) bound to HLA-A2 MHC class I (14) was a kind gift from B.D. Walker. Human CTL clone CER43 specific for influenza matrix protein peptide GILGFVFTL (GL9) in context with HLA-A2 (15, 16) was kindly provided by A. Lanzavecchia. The peptides IV9 and GL9 were generous gifts from H.N. Eisen. The LLFGYPVYV peptide from human T lymphotropic virus type 1 (Tax) (17) was synthesized by Research Genetics.
Production of soluble MHC
Preparation of pMHC/NiNLP conjugates
Preparation of NiNLPs has been previously described in detail (9–11). pHLA-A2/NiNLP conjugate formation was driven by the interaction between the His6 tag of pHLA-A2 and NLPs functionalized with lipids bearing nickel-charged nitrilotriacetic acid (NiNTA) headgroups. NiNLPs were mixed with pHLA-A2 ligands at various molar ratios in DPBS buffer at final concentration of 500 nM. The mixtures were incubated at room temperature for 10 min and diluted with DPBS containing 1% BSA to achieve the required NiNLP concentrations.
Equilibrium binding of pMHC/NiNLP to T cells
68A62 or CER43 CD8+ T cells were incubated with increasing amounts of freshly prepared pHLA-A2/NiNLP conjugates at a pHLA-A2 to NiNLP ratio of 10:1 or 30:1. The mixtures were incubated with live T cells for 1 h at the indicated temperature, and the cell fluorescence was immediately measured by a flow cytometer. The NiNLP loaded with noncognate Tax–HLA-A2mut protein and unconjugated NiNLP were used to measure background binding. The avidity of interactions was calculated with the GraphPad Prism 7 program, using the law of mass action:
where Y is the mean fluorescence intensity measured at concentration of pHLA-A2/NiNLP equal X, Bmax is the maximum specific binding (plateau value), and Kd is an apparent equilibrium binding constant.
Binding kinetics of pMHC/NiNLP to T cells
CTL at a density of 107 cells per ml were combined with the conjugates at 25 or 5 nM, and the mixture was incubated at different temperatures in DPBS/1% BSA. At designated time points, 10 μl of the mixture was diluted in 200 μl of chilled buffer, and the cell fluorescent intensities were immediately recorded using a flow cytometer. NiNLPs loaded with noncognate Tax–HLA-A2mut protein or unconjugated NiNLPs were used to measure the background binding. The background-subtracted data were fitted to an exponential equation with the GraphPad Prism 7 program using one- or two-phase exponential models:
where ki = 1/τi and Ai, ki, and τi are the amplitude, rate, and time constants, respectively.
To distinguish between one- and two-phase exponential models, an extra sum-of-squares F test was applied.
Measuring the kinetics of CTL Ca2+ signaling
CTL were loaded with Fluo-4 Ca2+–sensitive fluorophore as previously described (20, 22). Time-dependent changes in intracellular fluorescent intensity were measured using a BD LSR II Flow Cytometer. After measuring the background fluorescence, freshly prepared pHLA-A2/NiNLP were promptly added to 1 ml of the cell suspension (106 cells per ml). The data were analyzed with FlowJo software. Noncognate pHLA-A2/NiNLP conjugate were used as a negative control.
Results
Preparation and characterization of pMHC/NiNLP
NLPs were assembled to include a functional nickel-chelating lipid (NiNLPs), allowing for conjugation of His6-tagged proteins (Fig. 1A). The surface density of attached proteins can be reliably controlled by varying the percentage of nickel-chelating lipids in the NLP bilayer. The conjugation was assessed by analytical size exclusion chromatography at different pHLA-A2:NiNLP ratios (Fig. 1B). Analysis of the integrated area under the pMHC/NiNLP peak allowed to determine pHLA-A2:NiNLP molar ratio for each NiNLP containing distinct amounts of nickel-chelating lipids (Fig. 1C). For the experiments described in this paper, we choose NLPs containing 5 and 25 mol% of functional nickel-chelating lipid that bear 10 and 30 pHLA-A2 molecules, respectively.
NiNLP assembly and noncovalent conjugation of pMHC to NiNLP platform. (A) NiNLPs are assembled with a functional nickel-chelating lipid, which allows for surface conjugation of His6-tagged proteins. (B and C) Analytical size exclusion analysis (aSEC) of pMHC binding to NiNLPs. NiNLPs were incubated with pMHC at increasing pMHC:NiNLP molar ratios. An increase in molar ratio resulted in a gradual increase in the area under the main pMHC/NiNLP peak, which then plateaued at higher ratios concomitant with the appearance of a peak corresponding to free, unbound protein (B). The integrated area under the pMHC/NiNLP peak was analyzed as a function of the pHLA-A2:NiNLP ratio for NiNLP containing various concentrations of nickel-chelating lipids (C). Molar ratios for 5 and 25% of NiNLPs were found to be 10:1 and 30:1, respectively.
NiNLP assembly and noncovalent conjugation of pMHC to NiNLP platform. (A) NiNLPs are assembled with a functional nickel-chelating lipid, which allows for surface conjugation of His6-tagged proteins. (B and C) Analytical size exclusion analysis (aSEC) of pMHC binding to NiNLPs. NiNLPs were incubated with pMHC at increasing pMHC:NiNLP molar ratios. An increase in molar ratio resulted in a gradual increase in the area under the main pMHC/NiNLP peak, which then plateaued at higher ratios concomitant with the appearance of a peak corresponding to free, unbound protein (B). The integrated area under the pMHC/NiNLP peak was analyzed as a function of the pHLA-A2:NiNLP ratio for NiNLP containing various concentrations of nickel-chelating lipids (C). Molar ratios for 5 and 25% of NiNLPs were found to be 10:1 and 30:1, respectively.
The diameters of NiNLPs and MHC molecules are approximately 20 nm (9, 10) and 5 nm (23, 24), respectively. As such, 15 pMHCs can be captured on each bilayer surface of the nanoparticle having 25 mol% of NiNTA. These NiNLPs produce high-density pMHC/NiNLP, with close proximity between pMHC proteins and reflect the clustering of pMHC found at the surface of live cells (3). NiNLPs containing 5 mol% of NiNTA immobilized 10 molecules per NiNLP, which are referred to in this paper as low-density pMHC/NiNLP.
To evaluate multivalent receptor–ligand interactions between model pMHC membrane clusters and live CD8+ T cells, we used NiNLPs containing fluorescent-labeled Apo protein, which allowed us to measure the amount of membrane-bound pHLA-A2/NiNLPs by flow cytometry.
Equilibrium binding of pHLA-A2/NiNLPs with different stoichiometry to live T cells
To calculate apparent affinity constants (Kapp) for pMHC/NiNLP binding to live T cells, we used the following pMHC ligands: 1) cognate (strong agonist), 2) noncognate (“self”), and 3) the agonist ligands containing HLA molecules with mutated CD8 binding site. The last two ligands were analyzed to dissect contribution of CD8–pMHC and TCR–pMHC interactions to multivalent binding.
The high-density pHLA-A2/NiNLPs (30:1) bound to live 68A62 CTL with higher apparent affinity compared with low-density pHLA-A2/NiNLPs (10:1) regardless of the kind of pHLA-A2 and temperature (Fig. 2, Table I). Binding experiments with CER43 CTL produced similar results (data not shown). To quantify the difference, we introduced an enhancement factor, which was calculated as a ratio between apparent binding constants for low- and high-density pHLA-A2/NiNLPs. The analysis revealed that the enhancement factor for the noncognate conjugates was much higher than for the mutated and intact cognate conjugates (Table I); the difference was even more profound with temperature increase. Indeed, the amount of cell-bound low-density noncognate pHLA-A2/NiNLP was barely detectable at 37°C as opposed to high-density pHLA-A2/NiNLP (Fig. 2C, 2D). In contrast to noncognate conjugates, elevated temperature enhanced the binding of high- and low-density cognate conjugates with mutated CD8 binding site by 1.6- and 4.14-fold, respectively (Table I), providing evidence for the opposite effect of the temperature changes on multivalent CD8–MHC and TCR–pMHC interactions. In accord with these findings, the apparent binding constant of low-density cognate pHLA-A2/NiNLPs increased by 2-fold after the temperature was raised from 4°C to 22–24°C, whereas the binding of high-density cognate pHLA-A2/NiNLPs showed only a slight temperature dependence (Table I). These data provide compelling evidence that the pMHC density within model membrane clusters strongly affects binding to the T cell surface and that CD8–MHC interactions play a major role in determining contribution of self-pMHC to Ag recognition by T cells. Importantly, at 37°C, low-density noncognate pMHC model clusters essentially do not bind to the T cell surface (Fig. 2D), indicative of higher selectivity of T cell responses at physiological conditions.
Equilibrium binding of pMHC/NiNLPs to the surface of live CTL. 68A62 CD8+ T cells were mixed with increasing amounts of preformed pMHC/NiNLP conjugates at a pMHC to NLP ratio of 30:1 (A and C) or 10:1 (B and D) at indicated temperature. The data are representative graphs for binding of strong agonist IV9–HLA-A2/NiNLP at 4°C (filled circle), noncognate Tax–HLA-A2/NiNLP at 4°C (filled square), cognate IV9–HLA-A2mut/NiNLP at 4°C (filled triangle), noncognate Tax–HLA-A2/NiNLP at room temperature (open square), and noncognate Tax–HLA-A2/NiNLP at 37°C (cross-hatched square).
Equilibrium binding of pMHC/NiNLPs to the surface of live CTL. 68A62 CD8+ T cells were mixed with increasing amounts of preformed pMHC/NiNLP conjugates at a pMHC to NLP ratio of 30:1 (A and C) or 10:1 (B and D) at indicated temperature. The data are representative graphs for binding of strong agonist IV9–HLA-A2/NiNLP at 4°C (filled circle), noncognate Tax–HLA-A2/NiNLP at 4°C (filled square), cognate IV9–HLA-A2mut/NiNLP at 4°C (filled triangle), noncognate Tax–HLA-A2/NiNLP at room temperature (open square), and noncognate Tax–HLA-A2/NiNLP at 37°C (cross-hatched square).
NiNLPa . | pMHCb . | Kapp, nMc 4°C . | Enhancement Factord . | Normalized Binding Plateaue . |
---|---|---|---|---|
5% NiNTA | Cognate | 4.61 ± 1.96 | NA | 1 |
5% NiNTA | Noncognate | 9.43 ± 1.81 | NA | 0.340 ± 0.069 |
5% NiNTA | Cognate, mut | 31.2 ± 7.32 | NA | 0.463 ± 0.005 |
25% NiNTA | Cognate | 0.829 ± 0.368 | 5.6 | 1 |
25% NiNTA | Noncognate | 0.286 ± 0.116 | 33 | 0.390 ± 0.085 |
25% NiNTA | Cognate, mut | 5.60 ± 0.509 | 5.6 | 0.415 ± 0.010 |
NiNLPa . | pMHCb . | Kapp, nMc 4°C . | Enhancement Factord . | Normalized Binding Plateaue . |
---|---|---|---|---|
5% NiNTA | Cognate | 4.61 ± 1.96 | NA | 1 |
5% NiNTA | Noncognate | 9.43 ± 1.81 | NA | 0.340 ± 0.069 |
5% NiNTA | Cognate, mut | 31.2 ± 7.32 | NA | 0.463 ± 0.005 |
25% NiNTA | Cognate | 0.829 ± 0.368 | 5.6 | 1 |
25% NiNTA | Noncognate | 0.286 ± 0.116 | 33 | 0.390 ± 0.085 |
25% NiNTA | Cognate, mut | 5.60 ± 0.509 | 5.6 | 0.415 ± 0.010 |
5 and 25% NiNTA containing NLPs displayed 10 and 30 pMHC molecules per nanoparticle, respectively.
pMHCs were: IV9–HLA-A2 (cognate), Tax–HLA-A2 (noncognate), and IV9–HLA-A2mut (cognate, mut, cognate with mutation in HLA-A2 that diminishes CD8–HLA-A2 interactions).
Kapp, apparent equilibrium binding constants, mean ± SD are based on three independent experiments (n = 3).
The ratio between Kapp measured for low- and high-density pHLA-A2/NiNLPs.
Binding plateau values were measured from the fit of experimental data and normalized relative to the plateau value for the cognate conjugates.
The observed effects may be attributed exclusively to a larger number of pHLA-A2 proteins displayed on high-density pHLA-A2/NiNLPs (i.e., higher multivalency of the nanoparticles). To decouple the role of density and valency in the binding of model membrane clusters, we used pMHC/quantum dots (pMHC/QD[520]). The proximity between the pMHCs on QD (21) was similar to that for pMHCs presented on high-density NiNLPs (see above), and the number of the ligands per QD nanoparticle capable to interact with TCR/CD8 was much the same as for low-density pMHC/NiNLPs. Thus, pHLA-A2/QD conjugates have significantly lower valency but similar density when compared with high-density pMHC/NiNLPs. We found that both cognate and noncognate pHLA-A2/QD conjugates were bound to the surface of activated T cells (Supplemental Fig. 1). The difference between Kapp as well as normalized binding plateau values were similar to those observed for cognate and noncognate high-density but not low-density pHLA-A2/NiNLPs (Table I). Thus, the density, but not multivalency, was found to affect the efficiency of MHC–CD8 interactions and controls the ability of noncognate conjugates to bind to T cell surface.
Despite the higher expression level of CD8 relative to TCR/CD3 on the surface of T cells (25) (Supplemental Table I), the plateau values of high-density noncognate pHLA-A2/NiNLP binding were lower than the maximum of the binding for cognate pHLA-A2/NiNLPs (Fig. 2A, Table I). This suggests that merely a fraction of CD8 molecules on the T cell surface were capable of interacting with noncognate pHLA-A2/NiNLPs. The results are consistent with previous findings demonstrating that only a fraction of CD8 molecules on the surface of activated, but not naive, T cells, are coclustered with the TCR (26). To support this notion, we compared binding of cognate and noncognate conjugates to the surface of activated and naive 2C T cells. Although cognate SIYRYYGL-Kb/NiNLPs interacted equally well with the cell surface of naive and activated 2C cells, the binding of noncognate SIINFEKL-Kb/NiNLP to naive T cells was barely detectable despite similar level of expression of CD8 coreceptor on both naive and activated T cells (Supplemental Fig. 2).
Effect of pMHC density on the kinetics of pMHC/NiNLP binding
Multivalent interactions typically require a long time (many hours) to achieve equilibrium, whereas interaction between CTL and target cells is a very dynamic process that occurs very rapidly (minutes). To determine how pMHC cluster density influences the kinetics of their interactions with CD8 T cells, we compared the binding kinetics of various pMHC/NiNLPs conjugates to CTL surface.
For both high- and low-density cognate conjugates, biphasic time courses with similar contribution of the fast phase were observed (Fig. 3, Table II). However, comparison of the time constants for high- and low-density pHLA-A2/NiNLPs indicated that the latter bound to live T cells with slower kinetics. To understand these differences, the binding of noncognate and mutated cognate pMHC/NiNLPs conjugates was analyzed. The time courses of both high- and low-density cognate pHLA-A2mut/NiNLPs were monophasic, with very similar large time constants and low amplitudes. In contrast, the binding kinetics of low- and high-density noncognate pMHC/NiNLPs was very different. The association of high-density noncognate pHLA-A2/NiNLPs was biphasic, with time constant values close to that of the cognate conjugate. In contrast, real-time association curve of low-density noncognate pMHC/NLPs was monophasic, with relatively large time constant. This indicates that CD8–MHC interaction is responsible for the rapid kinetics observed for high-density cognate pHLA-A2/NiNLP interaction with CD8 T cell surface.
Association kinetics for the binding of pMHC/NiNLPs to the surface of live CTL. CER-43 CD8+ T cells were mixed with 5 nM pMHC/NiNLP conjugates at 18°C, and binding was measured as a function of time by flow cytometry. The experimental data and best fit are shown for strong agonist GL9–HLA-A2/NiNLP (circle), noncognate Tax–HLA-A2/NiNLP (square), and cognate GL9–HLA-A2mut/NiNLP (triangle) conjugates at pMHC to NiNLP ratios of 30:1 (A and C) and 10:1 (B and D). For comparison, normalized data and their fits (C and D) are presented. Results of representative experiment (n = 3) are shown.
Association kinetics for the binding of pMHC/NiNLPs to the surface of live CTL. CER-43 CD8+ T cells were mixed with 5 nM pMHC/NiNLP conjugates at 18°C, and binding was measured as a function of time by flow cytometry. The experimental data and best fit are shown for strong agonist GL9–HLA-A2/NiNLP (circle), noncognate Tax–HLA-A2/NiNLP (square), and cognate GL9–HLA-A2mut/NiNLP (triangle) conjugates at pMHC to NiNLP ratios of 30:1 (A and C) and 10:1 (B and D). For comparison, normalized data and their fits (C and D) are presented. Results of representative experiment (n = 3) are shown.
NiNLPa . | pMHCb . | Association Model . | τfast, sc . | τslow, sc . | % Fast Component . |
---|---|---|---|---|---|
5% NiNTA | Cognate | Two-phase | 164 ± 21.6 | 1301 ± 91 | 30.5 ± 4.9 |
Noncognate | One-phase | 831 ± 138 | NA | NA | |
Cognate, mut | One-phase | 2310 ± 50.2 | NA | NA | |
25% NiNTA | Cognate | Two-phase | 78.7 ± 35.4 | 703 ± 178 | 37.6 ± 5.1 |
Noncognate | Two-phase | 58.1 ± 14.6 | 831 ± 138 | 20.6 ± 5.3 | |
Cognate, mut | One-phase | 1751 ± 92.6 | NA | NA |
NiNLPa . | pMHCb . | Association Model . | τfast, sc . | τslow, sc . | % Fast Component . |
---|---|---|---|---|---|
5% NiNTA | Cognate | Two-phase | 164 ± 21.6 | 1301 ± 91 | 30.5 ± 4.9 |
Noncognate | One-phase | 831 ± 138 | NA | NA | |
Cognate, mut | One-phase | 2310 ± 50.2 | NA | NA | |
25% NiNTA | Cognate | Two-phase | 78.7 ± 35.4 | 703 ± 178 | 37.6 ± 5.1 |
Noncognate | Two-phase | 58.1 ± 14.6 | 831 ± 138 | 20.6 ± 5.3 | |
Cognate, mut | One-phase | 1751 ± 92.6 | NA | NA |
5 and 25% NiNTA containing NLPs displayed 10 and 30 pMHC molecules per nanoparticle, respectively.
pMHCs were: GL9–HLA-A2 (cognate), Tax–HLA-A2 (noncognate), and GL9–HLA-A2m (cognate, mut, cognate with mutation in HLA-A2 that diminishes CD8–HLA-A2 interactions).
τ is time constant in Eq. 2.
To evaluate the effect of valency versus proximity of the pHLA-A2 ligands, the binding kinetics of QD/pMHC(520) was analyzed. As evident from Supplemental Fig. 1, the association kinetics of both cognate and noncognate pHLA-A2/QD was very similar, resembling the binding kinetics of the high-density pHLA-A2/NiNLP conjugates (Fig. 3) despite lower valency of pHLA-A2/QD conjugates. These data provide evidence that the proximity of the pMHC ligands rather than the valency has a stronger impact on binding kinetics of pMHC/NiNLPs with various pMHC densities.
The role of pMHC density in regulating the kinetics of TCR-mediated Ca2+ signaling
To evaluate the effects of pMHC ligand density and MHC mutation that weaken CD8–MHC interactions on TCR-mediated signaling, we measured intracellular calcium flux induced by various pMHC/NiNLPs in real time. Induction of Ca2+ signaling by pMHC/NiNLPs occurs very rapidly, and the assay conditions preclude formation of cell–cell conjugates and allow the evaluation-integrated responses by individual T cells. For comparison reasons, the changes in intracellular Ca2+ level were measured at concentration of pMHC/NiNLP conjugates close to the apparent KD of their interactions with the T cells.
Fig. 4A shows that high-density cognate pHLA-A2/NiNLPs were superior in the induction of robust Ca2+ signaling as opposed to low-density pHLA-A2/NiNLP. To further explore the role of the proximity between pHLA-A2, we used high-density pMHC/NiNLP loaded with a mixture of cognate and noncognate pMHC ligands. The total number of pMHC per particle was kept the same, but the number of cognate pMHC was adjusted to that displayed on the low-density pMHC/NiNLP. Fig. 4B shows that the total MHC density, but not the density of cognate pMHC, controls the induction of the robust Ca2+ signaling. When cognate pHLA-A2 were replaced with cognate pHLA-A2mut, the signaling kinetics was significantly diminished (Fig. 4C). This effect became even more pronounced when the experiment repeated with the low-density conjugates (Fig. 4C). These data demonstrate that high density of pMHC molecules capable of interacting with CD8 within the membrane MHC clusters facilitates the kinetics of clustered pMHC interactions with the T cell surface, which regulates the kinetics and magnitude of Ca2+ signaling.
Variations in the kinetics of intracellular Ca2+ accumulation in CTL induced by pMHC/NiNLP conjugates with different pMHC density. Fluo-3–labeled 68A62 CD8+ T cells were mixed with cognate pMHC/NiNLP conjugates at indicated concentrations, and changes in intracellular fluorescent intensity were recorded by flow cytometry. NiNLPs were loaded with: (A) 30 (red) and 10 (black) cognate IV9–HLA-A2 molecules; (B) 30 (red) or 10 (black) cognate IV9–HLA-A2 molecules or a mixture of 10 cognate IV9–HLA-A2 molecules and 20 noncognate Tax–HLA-A2 molecules (blue); (C) 30 cognate IV9–HLA-A2 molecules (red), 30 cognate IV9–HLA-A2mut molecules (green), and 10 IV9–HLA-A2mut (violet). Data shown are representative from two to four independent experiments.
Variations in the kinetics of intracellular Ca2+ accumulation in CTL induced by pMHC/NiNLP conjugates with different pMHC density. Fluo-3–labeled 68A62 CD8+ T cells were mixed with cognate pMHC/NiNLP conjugates at indicated concentrations, and changes in intracellular fluorescent intensity were recorded by flow cytometry. NiNLPs were loaded with: (A) 30 (red) and 10 (black) cognate IV9–HLA-A2 molecules; (B) 30 (red) or 10 (black) cognate IV9–HLA-A2 molecules or a mixture of 10 cognate IV9–HLA-A2 molecules and 20 noncognate Tax–HLA-A2 molecules (blue); (C) 30 cognate IV9–HLA-A2 molecules (red), 30 cognate IV9–HLA-A2mut molecules (green), and 10 IV9–HLA-A2mut (violet). Data shown are representative from two to four independent experiments.
Discussion
The major finding of this study is that the density of pMHC molecules displayed on model membrane clusters strongly influences their binding to live CTL and the kinetics of intracellular Ca2+ signaling. Importantly, the binding of pMHC/NiNLPs to the T cell surface is regulated by both TCR–pMHC and CD8–MHC multivalent interactions, each contributing to the binding in a unique way. Although CD8–MHC interactions are characterized by high association rate constant and very short lifetime, TCR typically interacts with cognate pMHC with a much slower rate but form more stable complexes. Although CD8–MHC and TCR–pMHC interactions in vitro have been previously studied (27, 28), the contribution of these interactions on multivalent binding at the cell membrane have not yet been investigated. In this paper, we have shown that close proximity between the pMHC ligands involved in the multivalent interactions favors rebinding of the ligands to CD8 coreceptor which, along with the higher stability of the TCR–pMHC bonds, control the strength of pMHC/NiNLP association with the cell surface. Therefore, each interaction regulates apparent affinity of the multivalent binding in a different way. Consistent with this, computer simulation of bivalent ligand interactions with membrane-bound molecules have shown that the highest gain in apparent affinity of the interactions is achieved when the interaction of monovalent ligand with the receptor have a rapid kinetic constant of association with moderate lifetime of the complex (29). Thus, the proximity of pMHC ligands, high association rate of CD8–MHC interactions, and relatively long lifetime of cognate TCR–pMHC complexes emerges as essential parameters, explaining significance of MHC clustering in controlling multivalent interactions of pMHC membrane clusters with TCR/CD8 proteins on live T cells.
Analysis of equilibrium parameters of multivalent binding suggests an essential role for CD8–MHC interactions in the rapid scanning of target cell surface to identify a rare pMHC ligand. In contrast, rapid rebinding of CD8 to MHC facilitates binding kinetics of dense clusters to the T cell surface, which in turn facilitates the kinetics of intracellular signaling and T cell response. Therefore, CD8 plays a dual role facilitating the “hunt” by T cells for a rare or recently presented foreign pMHC ligand and rapid T cell response.
The important role of MHC density became even more apparent upon comparing the kinetics of Ca2+ signaling in T cells induced by high- and low-density cognate pMHC/NLPs (Fig. 4A). The superior potency of the high-density pMHC/NLPs was still evident when the number of cognate pMHC was decreased to match that presented on low-density conjugates with all others being replaced by noncognate pMHC (Fig. 4B). These data provide further evidence that self-pMHC within high-density pMHC clusters facilitate recognition of a small amount of cognate pMHC ligands by CD8 T cells (20, 21, 30). Close proximity of CD8 coreceptors, which are bound to foreign and self-MHC molecules within high-density clusters, may initiate p56lck trans-phosphorylation that also induces activation of TCRs bound to self-pMHC ligands within the TCR/CD8 cluster, a mechanism that has been referred to as signal spread (20). Cooperation of cognate and noncognate pMHC within high-density membrane clusters facilitates rapid accumulation of intracellular Ca2+ in CTL that is linked to effective delivery of cytolytic granules and potent cytolytic response (31, 32).
The cooperation of cognate and self-pMHCs has been previously delineated (20, 33), and clustering of self-pMHCs has been shown to be particularly important, as they are always present in a large excess relative to foreign pMHC on APCs (21, 34, 35). Thus, in the dense nanocluster, self-pMHC proteins could effectively facilitate recognition of even a single foreign pMHC ligand within the same cluster, resulting in a robust T cell response (20, 36). In contrast, poor clustering of MHC proteins on transformed cells could contribute to inefficient potency of T cells to attack on cancer cells.
The significance of close proximity of pMHC ligands within MHC clusters is also evident from our failure to observe cooperation of cognate and self-pMHC ligands incorporated into bilayers (N. Anikeeva, K. Somersalo, M.L. Dustin, and Y. Sykulev, unpublished observations). Even with the bilayers containing artificial barriers that limit diffusion of pMHCs (37–39), it is impossible to achieve the pMHC density similar to that observed for pMHC clusters assembled on nanoparticles or those presented on the cell membrane (5, 21). The density of pMHC on nanoparticles is significantly higher when compared with planar bilayers where pMHC ligands could not be positioned close enough to each other to cooperate during TCR and coreceptor engagement.
Upon delivery to the surface, pMHC ligands appear as large patches containing ∼100–150 pMHC molecules (40), which are relatively short-lived with t1/2 of around 30 s (41). The pMHC molecules then diffuse away from the patches, generating smaller clusters and individual molecules (40). The appearance and subsequent disaggregation of large pMHC clusters at the cell surface is a continuous process, ensuring sustained presentation of newly generated peptides (42). Comparison of the potency of high- and low-density pMHC/NiNLPs suggests that the most efficient recognition of newly processed peptides occurs within a large MHC membrane cluster. This is essential for early detection of infected cells by CTL during acute viral infection and initial containment of the infection. In contrast, low-density pMHC clusters possess much lower efficiency to stimulate T cell responses because of diminished ability of self-pMHC ligands to facilitate recognition of cognate pMHC (Fig. 2D). The clustering of cell surface receptors and their ligands involved in multivalent receptor–ligand interactions may also allow the formation of different patterns of engaged receptors, each possibly linked to a particular cellular response and efficacy. Thus, clustering of immune receptors appears to be an essential mechanism regulating T cell responses.
Because TCR and CD8 on naive T cells are not coclustered (26, 43), Ag recognition by naive T cells appears to be less efficient (44). This is expected to favor T cells’ responses to strong peptide epitopes enhancing specificity and diminishing cross-reactivity. During chronic infection, activated T cells whose TCR and CD8 are coclustered respond more efficiently to clustered pMHC on APC, allowing the detection of minute amounts of viral peptides as well as weak agonist peptides and mutated viral peptides displayed on the infected cells. Thus, responses of activated T cells are more sensitive but less selective (i.e., could target a wider range of peptide Ags). These considerations are consistent with the difference in the order of TCR and CD8 engagement on the surface of activated T cells described in this paper and naive T cells previously studied by Zhu and colleagues (45).
We have previously shown that MHC clusters on APC may include heterotypic membrane proteins such as ICAM-1, a ligand for LFA-1 integrin, which augment the sensitivity of the T cell response (1). This is consistent with our recent findings showing that integrin-mediated signaling regulates the efficiency of proximal signaling in cytotoxic lymphocytes (46, 47). Productive engagement of other receptors leads to recruitment of adaptor and signaling proteins to the TCR-containing clusters and formation of signalosomes (48) that integrate receptor-mediated signaling. Thus, clustering of immune receptors and proximal signaling proteins represent a highly sensitive regulatory mechanism that likely controls the balance of positive and negative signaling to ensure efficient responses against legitimate targets and to avoid overstimulation and the induction of T cell death.
Acknowledgements
We are grateful to Bruce Walker and Antonio Lanzavecchia for providing human CD8+ T cell clones. We also thank the Flow Cytometry Facility of the Sidney Kimmel Cancer Center for excellent technical support.
Footnotes
This work was supported by the National Institutes of Health (Grant R21 AI113819 to Y.S., which includes subcontract DE-AC52-07NA27344 to Lawrence Livermore National Laboratory) and in part by National Institutes of Health Grant R01 CA217714 to Y.S.
The online version of this article contains supplemental material.
Abbreviations used in this article:
- GL9
human CTL clone CER43 specific for influenza matrix protein peptide GILGFVFTL
- IV9
human CTL clone 68A62 recognizing HIV-derived peptide ILKEPVHGV
- mol%
mole percent
- NiNLP
NLP containing functional nickel-chelating lipid
- NiNTA
nickel-charged nitrilotriacetic acid
- NLP
nanolipoprotein particle
- pMHC
peptide–MHC
- QD
quantum dot
- Tax
LLFGYPVYV peptide from human T lymphotropic virus type 1.
References
Disclosures
The authors have no financial conflicts of interest.