Opposite effects of endogenous peptide/class I MHC on T cell activity in the presence and absence of CD8

Abstract

Non-stimulatory or endogenous pepMHC presented on the surface of APCs, either alone or alongside agonist pepMHC, play various roles in T cell selection and activation. To examine these properties in more detail, here we explored several model systems of TCR and pepMHC ligands with sufficient affinity to be activated in the absence of CD8. The TCRs had a range of affinities for agonist and non-stimulatory ligands, and were restricted by class I MHC alleles with different properties. We observed CD8-independent antagonism from TCR:pepMHC interactions with very low affinities (e.g. K_D = 300μM). In addition, endogenous pep/L^d complexes on APCs antagonized activation of co-receptor (CD8)-negative 2C T cells even by the strong agonist QL9/L^d. In contrast, TCRs m33 and 3D-PYY, restricted by K^b and D^b, respectively, did not show signs of antagonism by endogenous pepMHC in the absence of CD8. This did not appear to be an inherent difference in the ability of the TCRs to be antagonized, as altered peptide ligands could antagonize each TCR. In the presence of CD8, endogenous pepMHC ligands acted in some cases as co-agonists. These results show that endogenous pepMHC molecules exhibit complex behavior in T cells, leading to either reduced activity (e.g. in cases of low co-receptor levels) or enhanced activity (e.g. in presence of co-receptor). The behavior may be influenced by the ability of different TCRs to recognize endogenous pepMHC, but also perhaps by the inherent properties of the presenting MHC allele.

Introduction

T cells become stimulated when their clonotypic T cell receptors (TCRs) detect stimulatory peptide bound to products of the major histocompatability complex (MHC) on the surface of antigen-presenting cells (APCs). Typically, though, when a T cell interacts with an APC, it is exposed to a diverse set of peptide/MHC (pepMHC) complexes, most of which are non-stimulatory. A subset of the non-stimulatory endogenous pepMHC may still interact with the TCR with low affinity; in fact, a requirement for recognizing endogenous pepMHC drives T cell positive selection (reviewed in (1)). A sharp affinity threshold has been defined where weaker TCR:pepMHC interactions result in positive selection of clones, while higher affinity interactions result in negative selection (2), showing the importance of weak binding to endogenous complexes. Self peptides mediating positive selection signals have been isolated for CD8 (2–5) and CD4-resticted (6, 7) TCRs. Endogenous pepMHCs are also involved in providing T cell survival signals in the periphery, in the absence of fully activating pepMHC (6–8). The affinity ranges for TCR binding to endogenous pepMHC are quite low and difficult to measure by surface plasmon resonance; in fact, very few positively selecting pepMHC:TCR interactions have been explicitly measured. Those that have been measured ranged from 20 to 210 μM (measured at 10–25°C, (9–11)).

In addition to signals resulting from interactions with non-stimulatory pepMHC alone, there have also been reports describing a co-agonist role for these complexes alongside stimulatory pepMHCs. The presence of non-stimulatory pepMHCs has been shown to enhance activation by agonist pepMHC for CD4⁺ T cells (12, 13) and CD8⁺ T cells (14, 15). CD4 T cells have shown a restricted peptide specificity for the non-stimulatory co-agonists (12, 13), while CD8 cells have been reported to respond to any class I pepMHC tested (14, 15), suggesting that in the latter case, the interaction with CD8, and not necessarily the TCR, is critical.

In this study, we took advantage of TCR:pepMHC interactions that were of sufficient affinity to result in CD8-independent stimulation (16, 17). This allowed us to compare the responses of T cells to non-stimulatory pepMHC presented at the same time as an agonist pepMHC, with or without the expression of the co-receptor CD8. We found that in the absence of CD8, T cells experienced peptide-specific antagonism by non-stimulatory complexes that cannot be explained by peptide competition for MHC. Even high affinity, in vitro engineered TCRs were subjected to efficient antagonism. Quite weak interactions of at least K_D=300μM were able to mediate this antagonist effect. Antagonism was also mediated by endogenous pepMHC complexes when the 2C TCR was stimulated via its allospecific ligand QL9/L^d in the absence of CD8. When CD8 was expressed, however, the non-stimulatory complexes were converted into co-agonists.

Materials and Methods

Peptides and Cell Lines

SIY (SIYRYYGL), SIY R4A (SIYAYYGL), OVA (SIINFEKL), QL9 (QLSPFPFDL), QL9 F5A (QLSPAPFDL), QL9 F5R (QLSPRPFDL), MCMV (YPHFMPTNL), and WT1 (RMFPNAPYL) peptides were synthesized by the Macromolecular Core Facility of the Section of Research Sources, Penn State College of Medicine. Peptides were purified by reverse phase chromatography using a C-18 column; mass was confirmed by MALDI. Peptide quantification by amino acid analysis was performed at the Molecular Structure Facility, UC Davis (Davis, CA).

T2 (TAP-deficient lymphoblastoid cell line transfected with either H-2K^b, H-2D^b, or H-2L^d, P815 (L^d+ mastocytoma), EL4 (K^bD^b+/+ thymoma), MC57-L^d (K^bD^b+/+ fibrosarcoma transduced with H2-L^d, a gift from Hans Schreiber (University of Chicago)) and the 58^−/− T cell hybridoma (TCR αβ negative, CD8 co-receptor negative, H-2^k) were maintained in RPMI-1640 complete media (supplemented with 10% fetal calf serum, L-glutamine, penicillin and streptomycin). 58^−/− T cell hybridomas were also transduced to express 2C, m33, Y48_βA, Y26_αA, or 3D-PYY TCRs with the MSCV retrovirus as previously described (16), and the cells were maintained similarly in complete RPMI with or without co-transfection with the CD8αβ co-receptor under G418 selection. The CD8-independent TCR 3D-PYY was engineered from the 3D TCR, which was derived from a T cell clone isolated from a B6 mouse that was primed with activated RMFPNAPYL peptide-pulsed adherent splenocytes, and boosted with recombinant Listeria monocytogenes encoding the RMFPNAPYL epitope. The 3D-PYY mutant was identified using a T cell display system as has been previously described (18) and selection with WT-1/D^b Ig DimerX (BD Biosciences).

Protein expression and purification

High-affinity scTCR constructs were expressed as inclusion bodies in E. coli. The inclusion bodies were solubilized in 6M guanidine at 37°C, then added to 400mL of denaturing solution (3M Urea, 2mM reduced glutathione, 0.2mM oxidized glutathione, pH 8) at 4°C. Dilution buffer (200mM NaCl, 50mM Tris, pH 8) was added dropwise over a 36 hours, and the protein was purified using Ni-NTA agarose beads (Qiagen) followed by Superdex 200 gel filtration column (GE Healthcare). Purified scTCR was chemically biotinylated using the EZ-Link Sulfo-NHS-LC-Biotinylation Kit (Pierce) and re-purified by gel filtration shortly before use.

Detection of specific pep/MHC on the surface of APCs

Antigen-presenting cells were incubated with the indicated concentrations of peptide in complete RPMI media for 2 hours at 37°C. The cells were washed in phosphate-buffered saline (PBS) with 1% bovine serum albumin (BSA) and incubated with saturating amounts (more than 20 times higher than the K_D) of soluble, high-affinity scTCR (20μg/mL biotinylated m3 scTCR, and 40μg/mL biotinylated m67 scTCR for QL9/L^d and SIY/K^b, respectively) for 2 hours on ice. The cells were washed with ice-cold PBS with 1% BSA, and incubated with streptavidin labeled with phycoerythrin (BD Pharmingen) at a 1:100 dilution in PBS with 1% BSA for 30–45 minutes on ice. The cells were washed again and analyzed by flow cytometry using a BD FACS Canto (BD Biosciences) or a C6 flow cytometer (Accuri Cytometers, Inc.).

T cell stimulation and detection of secreted IL-2

TCR-transduced 58^−/− cells with or without CD8αβ co-expression (5×10⁴ per sample) were incubated with equal numbers of the indicated antigen-presenting cells and the indicated peptide mixture in complete RPMI media for 24 hours at 37°C. After the incubation, IL-2 in the supernatant was detected by sandwich ELISA as described previously (16); briefly, 2.5μg/mL anti-mIL-2 (BD Pharmingen) was coated in PBS onto 96-well Immulon 2HB plates, wells were blocked with 1% BSA in PBS, and then cell supernatant was incubated in the wells. Bound IL-2 was detected using 6.7μg/mL biotinylated anti-mIL-2 (BD Pharmingen) followed by streptavidin labeled with horseradish peroxidase (BD Pharmingen) at a 1:10,000 dilution. Plates were developed with a TMB substrate (KPL), and the reaction was stopped with 1N H₂SO₄. Absorbance at 450nm was measured using an EL_x800 universal microplate reader (Bio-Tek Instruments). Data were fit by least squares using Graph Pad or Microsoft Excel to the following equation:

$$y=\frac{a-b}{1+{\left(\frac{x}{c}\right)}^d}+b$$

where x is the concentration of peptide added, y is the IL-2 release, a and b are the maximum and minimum IL-2 release measured, respectively; c is the IC₅₀ or SD₅₀, and d is the slope of the increase or decrease.

Results

Antagonism of CD8-negative T cells by lower-affinity pepMHC ligands

The 2C TCR is a well-defined receptor system consisting of both syngeneic (SIY/K^b, (19)) and allogeneic (QL9/L^d, (20)) pepMHC ligands. In addition to the wild-type receptor, a series of engineered TCR mutants that bind with various affinities to both ligands have been made and characterized (16, 21, 22). Antagonism has been shown previously for the 2C TCR in CD8-positive T cells by various synthetic peptides presented by the syngeneic MHC H2-K^b (23); however, antagonism had not been shown for this TCR without expression of the CD8 co-receptor. To examine the relationships of TCR binding affinity and antagonism in the absence of co-receptor, we determined whether T cell hybridomas that express CD8-independent TCRs (2C or a higher affinity 2C mutant m33 and its derivatives) without CD8, could be antagonized by altered peptide ligands (APLs). A summary of binding properties and effects on stimulation or antagonism is shown in Table I for the TCR:pepMHC interactions explored below.

Table I.

Binding and Activation or Antagonism Properties for TCR:pep/MHC interactions.

TCR	pep	MHC	KD1 (μM)	CD8− effect	CD8− IC50 (μM pep)	CD8− SD50 (μM pep)	CD8+ effect	CD8+ SD50 (μM pep)	Refs2
2C	SIY	Kb	30	null	----	----	active	0.0001	(16, 21)
2C	QL9	Ld	1.6	active	----	0.0001	active	0.0001	(17, 20, 24, 25)
2C	QL9 F5R	Ld	300	antag. 100pM QL9	0.00044± 0.00027	----	active	0.01	(17, 25)
2C	MCMV	Ld	----	null/inhib.	1.7	----	null	----
Y48βA	SIY	Kb	2.9	antag. 200pM QL9	0.0011± 0.0002	----	active	0.00002	(16)
Y48βA	QL9	Ld	1.43	active	----	----	active	----
Y48βA	QL9 F5R	Ld	----	antag. 100pM QL9	0.0047± 0.0004	----	active	----
Y48βA	MCMV	Ld	----	null/inhib.	4.8	----	null	----
Y26αA	SIY	Kb	0.017	active	----	0.0002	active	0.00003	(16)
Y26αA	QL9	Ld	----	active	----	----	active	----
Y26αA	QL9 F5A	Ld	----	antag. 10pM SIY	0.0013± 0.0008	----	----	----
Y26αA	MCMV	Ld	----	null/inhib.	>10	----	null	----
m33	SIY	Kb	0.03	active	----	0.017	active	0.030	(16)
m33	SIY R4A	Kb	>34	antag. 5pM SIY	0.00015	----	active	0.34	(25)
m33	OVA	Kb	----	null	0.00029	----	null	----
3D-PYY	WT1	Db	0.15	active	----	8.9	active	0.076

¹K_D measured by surface plasmon resonance at 25°C unless otherwise noted.

²References for affinity and SD₅₀ values not included in this paper.

³K_D estimated by competitive binding of soluble high-affinity scTCR, see Supplementary Figure 1.

⁴Affinity and SD₅₀ values measured for the closely-related TCR m67.

⁵K_D estimated by titration of WT1 peptide-loaded D^b-Ig dimer (DimerX, BD Pharmingen) on CD8-negative T cell hybridomas transduced with the 3D-PYY TCR, analyzed by flow cytometry (data not shown).

Cells that express the 2C TCR can be stimulated by QL9/L^d in the absence of CD8, owing to an interaction affinity above the CD8-dependent threshold (K_D ~1 μM, (20, 24, 25)). 2C T cells are stimulated to release IL-2 by 100 pM QL9 peptide presented by T2-L^d cells (Figure 1A), which are TAP deficient and lack the ability to present endogenous peptides on class I MHC. To determine if this CD8-independent activity could be antagonized by low affinity interactions, we used the altered peptide QL9 F5R. F5R binds well to L^d (K_D=0.9–2nM estimated from MHC stabilization assays, (17, 25)), but the affinity of the 2C TCR for F5R/L^d complex is very low (K_D = 300μM, (25)). The F5R peptide was able to antagonize 2C activity with an IC₅₀ of 0.44±0.27 nM, whereas null peptide, MCMV, which also binds well to L^d (K_D=0.5nM estimated from MHC stabilization, (26)), exhibited over 1,000-fold lower inhibitory activity (average IC₅₀ of 1.7μM, Figure 1A). This result indicates that antagonism can be mediated by the TCR:pepMHC interaction alone, without a requirement for CD8 participation. When CD8 is expressed, 2C can recognize QL9 F5R/L^d as an agonist peptide, whereas MCMV/L^d remains a “null” interaction (25).

Fig. 1. Activation of CD8-negative T cells is antagonized by specific petide/MHC ligands that have lower affinity for the TCR.

(A) Activation, detected as IL-2 secretion in duplicate, of 2C T cells by QL9 presented by T2-L^d cells, with effects of added peptide QL9 F5R (antagonist, triangles) or MCMV (null, circles). Data are representative of 2 independent experiments. (B) Activation of m33 Y48_βA T cells in duplicate by QL9 presented by T2-L^d cells, with effects of added peptide QL9-F5R (antagonist, triangles) or MCMV (null, circles). Data are representative of 2 independent experiments. (C) Activation of m33 T cells by SIY presented by T2-K^b cells, with effects of added peptide SIY-R4A (antagonist, triangles) or OVA (null, circles). Data are representative of 4 independent experiments.

We compared the 2C antagonism by QL9 F5R/L^d to the effects seen for the related TCR, m33 Y48_βA. The m33 Y48_βA TCR has an affinity for QL9/L^d comparable to 2C (1.4μM and 1.2μM, respectively; Supplementary Fig. 1), while its affinity for the syngeneic pepMHC ligand SIY/K^b is 2.9μM, 12-times higher than 2C (16). Stimulation of T cells with the Y48_βA TCR by 100pM QL9 on T2-L^d cells was also efficiently antagonized by QL9 F5R (IC₅₀ = 4.7±0.4nM), while MCMV inhibited activity only at higher concentrations (average IC₅₀ of 4.8μM, see Figure 1B). Thus, a different receptor showed similar antagonist properties in the absence of CD8.

Antagonism of high-affinity TCR:pepMHC interactions on CD8-negative T cells

Next, we analyzed whether it was possible to antagonize a very high-affinity TCR:pepMHC interaction. The m33 TCR binds to SIY/K^b with 1000-fold higher affinity (K_D =30 nM, (16, 24)) than the wild-type 2C TCR (K_D=30 μM, (16, 21)), and as such T cells that express the m33 TCR are stimulated by SIY/K^b on T2-K^b cells even in the absence of CD8 (16, 17). The SIY peptide variant R4A binds to K^b with the same affinity as SIY and the complexes have the same half-lives (~10 hours) on the surface of T2-K^b cells (27, 28). These K^b complexes are less stable than OVA/K^b, which has a half-life of 23 hours (28). The binding of m33 TCR to R4A/K^b is reduced by over 100-fold, compared to SIY/K^b, as this residue is energetically important for TCR binding (27). R4A/K^b is unable to stimulate m33 T cells in the absence of CD8. Nevertheless, R4A was able to inhibit the activity of SIY at approximately a 20-fold lower concentration than OVA (IC₅₀ = 0.15 nM and 2.9 nM, respectively, Figure 1C). The R4A effect is likely due to antagonism, rather than competition for K^b, since OVA/K^b complexes are more stable than R4A/K^b complexes, and yet are less efficient at inhibiting activation by SIY. Thus, even very high-affinity TCR interactions can undergo antagonism with lower affinity ligands in the absence of CD8.

Antagonism signals can be delivered through TCR interactions with a different MHC allele

As shown above with null peptide ligands, inhibition can also occur through competition with the agonist peptide for MHC binding. To avoid this complicating effect, it is possible to use an APC that expresses two different MHC alleles, and a single TCR that can interact with peptides bound to each. MC57-L^d is an H-2^b-derived tumor cell line that has been transfected with the gene for L^d. This cell line expressed high levels of H2-K^b and H2-L^d (Figure 2A,B). CD8-negative T cells that express the m33 Y48_βA TCR were stimulated with the agonist QL9/L^d, and SIY was added to assess antagonist activity when presented by K^b. SIY/K^b is a full agonist for the Y48_βA TCR when CD8 is co-expressed, but it does not stimulate the T cells when Y48_βA is expressed without CD8 (16). Addition of SIY peptide caused potent antagonism of Y48_βA T cell stimulation by QL9/L^d (average IC₅₀ = 1.1±0.2 nM, Figure 2C). As expected, a higher level (five-fold) of agonist QL9/L^d required greater levels of SIY peptide, roughly tenfold, for the same antagonist effect. This result supports the view that inhibitory signals are delivered through TCR binding to the antagonist pepMHC, even in the absence of CD8.

Fig. 2. Peptide antagonism using separate, non-competing MHC alleles on MC57-L^d.

(A–B) Cells were stained on ice with (A) B.8.24.3, specific for H2-K^b, or (B) 30-5-7, specific for H2-L^d. Binding of the primary antibody was detected with a polyclonal goat anti-mouse Alexa 488 F(ab′)₂ (Invitrogen Molecular Probes), and analyzed by flow cytometry. In each panel, the gray histogram depicts the cells stained with the MHC-specific antibody, and the black histogram represents cells stained with the secondary antibody only. (C) Activation of m33 Y48_βA T cells in triplicate by two concentrations of QL9 presented by H2-L^d on the surface of MC57-L^d cells, with effects of added SIY peptide, which is presented by H2-K^b on the surface of the antigen-presenting cell. Data are representative of 4 independent experiments. (D) Activation of m33 Y26_αA T cells by SIY peptide presented by H2-K^b on MC57-L^d cells, with effects of added MCMV (null, circles) or QL9 F5A (antagonist, triangles), which are presented by H2-L^d on the surface of the antigen-presenting cell. Representative data are shown from 2 independent experiments.

Using the same MC57-L^d cells, we confirmed that a high affinity interaction could be fully antagonized with the reciprocal combination of MHCs as agonist (K^b in this case) and antagonist (L^d in this case). The m33 Y26_αA TCR has high-affinity for SIY/K^b such that the T cell line is stimulated by SIY in the absence of CD8 (K_D =17 nM, (16)). T cells that express the m33 Y26_αA TCR were stimulated in the presence of 10 pM SIY and MC57-L^d cells, and various concentrations of QL9 variant F5A (antagonist) or MCMV (null) peptides, both presented by H2-L^d. Once again, the high-affinity agonist interaction with SIY/K^b could be fully antagonized by QL9 F5A/L^d (IC₅₀ = 1.3±0.8 nM) at more than 10,000-fold lower concentration than MCMV (IC₅₀ > 10 μM, see Figure 2D). The minimal inhibition at very high MCMV concentrations might result from increased levels of stabilized MCMV/L^d complexes on these cells competing for a common pool of β₂M, resulting in lower total levels of K^b on the surface.

Antagonism of CD8-negative 2C T cells by endogenous pep/L^d complexes

Typically, antigen-presenting cells are TAP-competent, and present a variety of endogenous pepMHC along with potential agonist pepMHC. Various roles for endogenous pepMHC have been proposed, most often involving participation of co-receptor (12, 14, 29). In principle, the inhibitory signals associated with antagonist ligands as described in the lower affinity altered peptide systems here could also be delivered through weak interactions with endogenous pepMHC. To analyze the effect of endogenous pepMHC for co-receptor negative 2C T cells, we compared their stimulation by QL9/L^d presented by TAP-competent (MC57-L^d and P815) and TAP-negative (T2-L^d) antigen-presenting cells. Despite CD8-independent sensitivity to QL9/L^d, wild-type 2C T cells lacking CD8 were impaired when responding to QL9/L^d on the surface of TAP-competent cells (see Figure 3A). For both TAP-competent lines, the maximum level of IL-2 secretion for 2C cells was significantly reduced. In contrast, the SD₅₀ values for QL9 peptide were similar for T2-L^d and MC57-L^d (115±100 pM and 28±19 pM, respectively), with a shift to a slightly higher concentration for P815 (350±70 pM).

Fig. 3. Activation of CD8-negative 2C T cells is antagonized by self peptide/Ld ligands.

(A) IL-2 secretion detected in triplicate from activation of 2C T cells by QL9 presented by T2-L^d cells (open circles), P815 (closed diamonds), or MC57-L^d (closed squares). Data are representative of at least 3 independent experiments. (B) Staining in duplicate of antigen-presenting cells for specific stimulatory complex QL9-L^d, using the soluble, high-affinity scTCR m3. Error bars represent standard deviation of measurements, and data are representative of 3 independent experiments. (C) High affinity m3 scTCR staining and IL-2 secretion from 2C T cells stimulated by T2-L^d (128pM QL9, highlighted with arrows labeled “a” in panels (A) and (B)), P815 (16nM QL9, highlighted with arrows labeled “b” in panels (A) and (B)), and MC57-L^d (16nM QL9, also indicated with arrows labeled “b”).

We considered that there could simply be less agonist peptide loaded onto antigen-presenting cells when other, endogenous complexes are present, since existing peptide would have to be displaced from the MHC before the exogenously added peptide could be loaded. To investigate this possibility, we used a soluble, biotinylated, high-affinity TCR (called m3, K_D = 27 nM for QL9/L^d) to specifically detect the levels of agonist peptide-MHC on the surface of the different antigen-presenting cells by flow cytometry (25). The peptide loading levels at equivalent exogenous peptide concentrations were, indeed, different depending on whether endogenous pep/L^d complexes were present (Figure 3B). However, we examined in more detail the QL9 concentrations that yielded the highest QL9/L^d surface levels on the P815 and MC57-L^d, compared to lower surface levels on T2-L^d (Figure 3C). For example, at 16 nM QL9 much higher levels of QL9/L^d were found on MC57-L^d and P815 than at 128 pM QL9 with T2-L^d (Figure 3C). However, at these concentrations, there was considerably more activity with T2-L^d as the antigen-presenting cell, despite lower levels of QL9/L^d (Figure 3C). This is consistent with the possibility that endogenous pepMHC ligands act as an antagonist on P815 and MC57-L^d.

To further analyze the role of endogenous pepMHC in T cell activation (in the absence of CD8), we examined the K^b system, comparing T2-K^b and the TAP-sufficient line EL4. Here we used the m33 TCR that has high-affinity for SIY/K^b (K_D = 30 nM) and thus is stimulated by SIY in the absence of CD8. As with the QL9/L^d system, reduced sensitivity was observed for high-affinity m33 T cells that lacked CD8 when stimulated with SIY presented by cells with endogenous pep/K^b (EL4) versus TAP-deficient cells (T2-K^b) (Figure 4A). In this case, the maximum stimulation level was equivalent for both antigen-presenting cell types, but the SD₅₀ of SIY peptide was increased 100-fold when presented by EL4 (560±210 nM) versus T2-K^b (17±15 nM). We used the soluble, high affinity TCR m67 (K_D=16 nM, (21)) to specifically detect SIY/K^b on the surface of antigen-presenting cells, similar to the approach with soluble m3 TCR for quantification of QL9/L^d levels. The level of SIY/K^b detected on EL4 cells began to rise at about 10-fold higher SIY peptide concentrations than on T2-K^b cells. Unlike the observations with QL9/L^d, there were no SIY concentrations where higher levels of SIY/K^b were presented by EL4 than T2-K^b, yet there was correspondingly lower levels of stimulation by EL4 (Figure 4C). Hence, in this case, the lower sensitivity in the presence of APCs with endogenous pepMHC could be explained, most simply, by reduced SIY loading and presentation by K^b on EL4.

Fig. 4. Antagonism is not evident for other CD8-independent TCRs beyond what can be explained by reduced agonist presentation.

(A) Activation of m33 T cells by SIY/K^b in duplicate and (B) Staining of antigen-presenting cells in duplicate for specific stimulatory complex SIY/K^b using the soluble, high-affinity scTCR m67 for T2-K^b cells (open circles), or EL4 (closed diamonds). Error bars represent standard deviations of measurements, and data are representative of 2 independent experiments. (C) High affinity m67 scTCR staining and IL-2 secretion from m33 T cells stimulated by EL4 (10 mM SIY, highlighted with arrows labeled “a” in panels (A) and (B)) and T2-K^b (2 mM SIY, highlighted with arrows labeled “b” in panels (A) and (B)). (D) Activation of 3D-PYY T cells in duplicate by WT1 presented by T2-D^b cells (open circles), or EL4 (closed diamonds). Data are representative of 3 independent experiments.

Bonafide CD8-independent TCRs (i.e. those that are active in the absence of CD8, rather than those that are simply inhibited by anti-CD8 antibodies, (30)) are uncommon. However, we recently engineered a mouse TCR against WT1 peptide/D^b antigen that has sufficient affinity to be CD8-independent (unpublished data). T cells that express this TCR, 3D-PYY are stimulated specifically by WT1/D^b complexes in the absence of CD8. Unlike the TCRs examined above, the 3D-PYY T cell hybridoma was stimulated equally well with WT1 peptide presented by EL4 (TAP-competent) or T2-D^b (TAP-negative) (SD₅₀s roughly 10 μM, Figure 4D). Thus, in this case, there is no evidence that 3D-PYY binds sufficiently to endogenous peptide/D^b complexes to elicit antagonist behavior. These results also show that one can not assume that an exogenous peptide loaded on a TAP-deficient line will stimulate T cells more effectively than the same peptide loaded on a cell line that contains endogenous peptides.

Further evidence that CD8 can allow endogenous pepMHC to act as co-agonists for some TCRs

Previous reports for CD4 (12, 13, 31) and CD8 (14, 15) T cells have shown that non-stimulatory pepMHC complexes can increase the stimulatory capacity of agonist pepMHC, acting as co-agonists when co-receptor is expressed. However, this behavior can be difficult to observe, apparently depending on the differentiation state of the T cell, the effector function examined, and the strength of the agonist. Since in our cases all of the TCR:agonist reactions are strong, such that they were genuinely CD8-independent, we analyzed whether the antagonist behaviors observed in the absence of CD8 would demonstrate co-agonist behavior in the presence of CD8.

When 2C CD8⁺ T cells were stimulated with QL9/L^d presented by the different APCs, similar SD₅₀ values were seen in the presence or absence of endogenous pepMHC (SD₅₀ = 42±39 pM for MC57-L^d, 30±10pM for P815, and 120±110pM for T2-L^d, Figure 5A). Since 2C was originally generated as an alloreactive T cell against P815 (32), we observed, as expected, that P815 cells acted as a stimulus for 2C CD8⁺ T cells in the absence of added QL9. It is not clear why MC57-L^d did not also yield higher levels of stimulation in the absence of QL9. Nevertheless, to compare the additional agonist effects of exogenous QL9, the stimulation levels in the absence of peptide were subtracted to yield the QL9-mediated agonist activity (Figure 5B). The findings showed that the 2C CD8⁺ T cell response to MC57-L^d and P815 presented QL9 were superior to T2-L^d. Thus, in the presence of the CD8 co-receptor, the endogenous pep/L^d complexes switched from providing an overall negative signal to an overall positive signal to 2C T cells.

Fig. 5. Stimulation of CD8-positive T cells can be enhanced by self peptide/MHC ligands.

(A) Raw absorbance and (B) absorbance corrected by subtracting the absorbance of no peptide added measured for IL-2 secretion by 2C T cells stimulated in triplicate by QL9 presented by T2-L^d cells (open circles), P815 (closed diamonds), or MC57-L^d (closed squares). Data are representative of at least 3 independent experiments. (C) Raw absorbance or (D) absorbance corrected by subtracting the absorbance of no peptide added measured for IL-2 secreted by m33 T cells stimulated in duplicate by SIY presented by T2-K^b cells (open circles), or EL4 (closed diamonds). Data are representative of at least 3 independent experiments. (E) IL-2 secreted by 3D-PYY T cells stimulated by WT1 presented by T2-D^b cells (open circles), or EL4 (closed diamonds). Data are representative of at least 2 independent experiments.

One consequence of using our system of altered peptide ligand antagonism assays (Figures 1,2) is that all of the peptides that can antagonize without CD8 co-expression can act as full agonists when CD8 is present ((17, 25, 27), and data not shown). Nevertheless, we attempted to analyze the effects of two of these peptides (SIY R4A and QL9 F5A) on activation by strong and weak agonists for the 2C TCR. In this altered peptide ligand assay, we did not see a co-agonist effect for any of the peptide combinations tested for stimulation of 2C⁺ CD8 T cells (See Supplementary Figure 2). The co-agonist effect seen with endogenous complexes is in contrast to the lack of co-agonist effect seen by altered peptide ligands. It is possible that the individual altered peptide ligands tested do not bind with properties in the exact range of some present among the many endogenous pepMHC complexes that can mediate either antagonist or co-agonist behavior seen for 2C T cells with or without co-receptor, respectively. Alternatively, it is possible that the many different endogenous peptides among the peptide milieu may mediate a range of effects, and it is an averaged effect that differs when CD8 is or is not co-expressed.

We also analyzed the effects of endogenous pepMHC on CD8⁺ T cell responses where there was less evidence for antagonism in the absence of CD8. Simulation of m33 CD8⁺ T cells by TAP-competent EL4 cells was significant without added peptide (Figure 5C), as observed previously (21). However, subtraction of this basal level, as done with P815 and 2C CD8⁺ cells, showed that the sensitivity for SIY was not increased in the presence of endogenous pepMHC (SD₅₀ = 1.4±1.0 μM for EL4, and 30±26 nM for T2-K^b, Figure 5D). Similarly, stimulation of 3D-PYY CD8⁺ T cells showed minimal effects in the presence of endogenous pepMHC (SD₅₀ =210±110 nM for EL4, and 550±80 nM for T2-D^b, Figure 5E). Thus, like antagonism mediated in the absence of CD8, co-expression of CD8 resulted in one strong agonist:TCR interaction with enhanced effects of endogenous pepMHC, but two high-affinity, engineered agonist:TCR interactions with no evidence of an effect of endogenous pepMHC. Enhanced agonist:TCR affinity reduced the apparent positive and negative effects of non-agonist pepMHC, even though cells with high-affinity TCRs without CD8 can be antagonized by defined altered peptide ligands (Figures 1 and 2). This is in agreement with results showing that there is a reduced co-agonist effect mediated by endogenous pepMHC for strong agonists as compared to weak agonists (15).

Discussion

Specific endogenous peptides bound to class II MHC can serve as both positive selecting ligands and as co-agonists of mature CD4⁺ T cells (6, 7). Endogenous peptides bound to class I MHC can also act as co-agonists for CD8⁺ T cells (14, 15) and provide cooperative binding in a CD8-dependent mechanism (29). However, in contrast to observations with CD4⁺ T cells, there has been no evidence that TCRs on CD8⁺ T cells contribute to the co-agonist effects by binding to specific peptide-class I complexes. Using class I-restricted TCRs that have affinities for agonist allowing them to function in the absence of CD8, we show evidence that endogenous peptide-class I ligands can have effects on signaling directly through the TCR.

Our strongest evidence for the role of endogenous pepMHC came from comparing the responses of 2C TCR T cells in the presence or absence of CD8, to the exact same APCs. Accordingly, the activity with QL9 peptide for CD8⁻ cells was much stronger when endogenous pepMHC were absent, whereas the activity with QL9 peptide for CD8⁺ cells was much stronger when endogenous pepMHC were present. Quantification of the total number of QL9/L^d complexes verified that even though there were greater levels of complex at some concentrations, the presence of endogenous pepMHC inhibited stimulation. We ruled out that this effect was due to inhibitory ligands such as PD-L1 on the APCs (data not shown), and by the observation that these same cell lines stimulated CD8⁺ cells very effectively.

In cases of the TCR 2C, where there appeared to be antagonist activity by endogenous pepMHC, there was strong evidence for co-agonist activity in the presence of CD8. In the cases (m33 and 3D-PYY) where there was no evidence of antagonist activity by endogenous pepMHC, there was also no evidence for co-agonist activity in the presence of CD8. The result with 2C further supports the view that CD8 is critical in the mechanism that involves the co-agonist behavior of endogenous pepMHC. However, the results also suggest that the CD8-dependent, co-agonist behavior will vary among TCRs, depending on their interaction with the endogenous pepMHC. This finding may be somewhat contrary to the observations of others, but we propose the discrepancy may be due to differences in the TCRs, as we used fully CD8-independent TCRs, and also perhaps due to differences in the MHC alleles that have been examined (see below).

We were unable to distinguish whether some of the differences in behavior between TCR systems were a consequence of differences in interactions with endogenous pepMHC (e.g. 2C vs m33 vs 3D-PYY), or whether there was a difference among MHC alleles (e.g. L^d vs K^b vs D^b), or both. While a previous study suggested that TCRs do not distinguish between endogenous pepMHC in CD8-dependent co-agonist effects, our results suggest that different TCRs may indeed have differences in the antagonist behavior of endogenous pepMHC in the absence of CD8. If this hypothesis is correct, than one would also predict that TCRs could also differ in endogenous pepMHC co-agonist activity in the presence of CD8. The relationships between TCR affinity for agonist, TCR cross-reactivity with endogenous ligands, and apparent antagonist or co-agonist effects have not been resolved. Li et. al. presented results suggesting that for CD4 T cells, increased agonist affinity would allow for a larger fraction of endogenous pepMHC complexes to contribute to T cell recognition (33). However, our results are more similar to experimental evidence from Yachi et. al. showing that the most significant effects of endogenous pepMHC were seen for weak, rather than strong, agonists for CD8 T cells (15). It is possible that results will vary not only based on the affinity of these interactions, but on the inherent cross-reactivity of each TCR with the array of endogenous pepMHC (34).

Alternatively, perhaps some MHC alleles are more likely to be involved in antagonist or co-agonist activity, depending on intrinsic properties such as cell surface life-times of the peptide complexes. In this respect, it is interesting that L^d complexes showed the most clear-cut evidence of antagonist and co-agonist behavior. L^d has been shown to be unique among other well-studied murine class I MHC alleles, including K^b and D^b. L^d is processed and trafficked more slowly (35, 36), has a lower affinity for β₂M (35, 37), has a shorter average half-life on the cell surface (36, 38, 39), has a lower surface expression level (35, 40), and is present more frequently in a putative peptide-empty form (36, 41), presumably due to rapid peptide dissociation. These features may result in less stable TCR interactions, as a given pep/L^d complex that binds to a TCR and dissociates may be more likely than other alleles to fall apart before potential re-binding by the TCR (42). In this scenario, in the absence of CD8, an accumulation of abortively-triggered TCR complexes could result.

Mathematical modeling suggested that the less exhaustively a given TCR population has been exposed to pep/MHC complexes for negative selection by a given allele, the more likely that cross-reactive TCRs restricted by that allele will be present in the T cell population (43). The 2C TCR was raised from an H-2^b mouse in an allospecific reaction against an H-2^d stimulus, and thus the 2C TCR was not negatively selected against the array of pep/L^d complexes. Alloreactive TCRs have been shown to have varying degrees of peptide specificity (or degeneracy, reviewed in (44)). While the 2C TCR is clearly not reactive with L^d in a peptide non-specific manner, it is possible that it binds weakly to an unusually high fraction of pep/L^d complexes in the affinity range that leads to antagonism when CD8 is not present. Hence, a higher level of cross-reactivity due to incomplete or inefficient negative selection may enhance effects of endogenous pepMHC. Less efficient peptide presentation by L^d in the thymus could lead to less efficient negative selection even for endogenous TCRs, leading to an L^d-restricted TCR population that is more tolerant of epitope mutations, as has been seen for elite HIV controllers carrying the HLA-B57 allele (43).

Physiological occurrences of antagonism by self pepMHC have been proposed to be common, and possibly play a role in disease resistance or susceptibility (reviewed in (45)). The antagonism seen here for CD8-negative T cells by endogenous complexes may have relevance to the selection of TCRs for adoptive T cell tumor therapy, which is a technique that has shown some promise as a cancer immunotherapy (46, 47). Efforts to re-direct the activity of CD4 T cells to class I-restricted tumor antigens have involved introduction of a TCR that binds to a class I ligand with sufficient affinity. Previously, we have shown that CD8 independent activity would only be attained with a K_D around 1μM or less (16). Furthermore, the presence of non-cognate co-receptor (CD4) would further reduce the sensitivity of T cells to their class I pepMHC targets (30). The current results show that once a CD8-independent TCR is obtained, it may also be subject to antagonism by endogenous pepMHC. Since even high-affinity TCRs (such as Y26_αA and m33, Figs. 1 and 2) are antagonized by lower affinity pepMHC, it would be useful to test different candidate therapeutic TCRs for possible antagonist activity in the presence of relevant endogenous pepMHC.

The exact mechanism of inducing negative, antagonistic signals to a T cell is not clear. Co-receptor recruitment has been proposed as an important distinguishing factor between agonist and antagonist ligands (48, 49). As with the effects of co-agonist pepMHC, the mechanisms of antagonism may, in fact, be different for CD4 and CD8 T cells. Studies of CD8⁺ T cells expressing two, independent TCRs showed that providing an antagonist signal for one TCR did not decrease activation through the other TCR (50), suggesting that antagonism of CD8⁺ cells occurs at the level of individual TCRs. However, parallel studies of CD4⁺, dual-TCR T cells showed that cross-antagonism from one receptor to the other occured (51), suggesting that antagonism of CD4⁺ T cells resulted in a dominant negative signal. These results imply co-receptor involvement in the antagonism, although differences in activation read-outs for CD8⁺ T cells (specific target cell lysis, (50)) and CD4⁺ T cells (proliferation, (51)) between the two studies may be a confounding factor. Our findings here agree with previous findings for CD4 T cells and class II pepMHC (52, 53) indicating that co-receptor is not required for antagonism, and the effect can be mediated by TCR alone. One possibility is that pepMHC binding by the TCR outside of an optimal range initiates abortive signaling, accumulation of partially phosphorylated TCR complexes (54), recruitment of negative regulators such as SHP-1 (55, 56), and/or inefficient formation of the immunological synapse and cell polarization (57).

Our investigation shows that TCR:pepMHC affinities up to 300μM are sufficient to result in antagonist activity in the absence of CD8, while the same interaction can mediate activation when CD8 is present (25). It is possible for the binding properties that result in antagonism to differ with and without co-receptor participation, and the K_D threshold for antagonism for CD8⁺ T cells might be significantly higher than 300 μM. Many naturally occurring TCR interactions with endogenous or weak agonist pepMHC are likely to fall in this intermediate range of affinity (antagonists without CD8 and co-agonists or agonists with CD8). Since previous studies have shown that only a very small fraction of TCRs are constitutively associated with CD8 (58, 59), an interesting combination of signals must be integrated when a T cell encounters an antigen-presenting cell: negative or abortive signals from endogenous or even weak agonist pepMHC binding to TCR not pre-associated with CD8 must be balanced against agonist interactions with the participation of CD8. Further investigation into this phenomenon is required.