Different affinity windows for virus and cancer-specific T-cell receptors – implications for therapeutic strategies

Summary

T-cell destiny during thymic selection depends on the affinity of the T-cell receptor (TCR) for autologous peptide ligands presented in the context of MHC molecules. This is a delicately balanced process; robust binding leads to negative selection, yet some affinity for the antigen complex is required for positive selection. All TCRs of the resulting repertoire thus have some intrinsic affinity for an MHC type presenting an assortment of peptides. Generally, TCR affinities of peripheral T cells will be low towards self-derived peptides, as these would have been presented during thymic selection, whereas, by serendipity, binding to pathogen-derived peptides which are encountered de novo could be stronger. A crucial question in assessing immunotherapeutic strategies for cancer is whether natural TCR repertoires have the capacity for efficiently recognizing tumor associated peptide antigens (TAPAs). Here, we report a comprehensive comparison of TCR affinities to a range of HLA-A2 presented antigens. TCRs which bind viral antigens (VAs) fall within a strikingly higher affinity range than those which bind cancer-related antigens. This difference may be one of the key explanations for tumor immune escape and for the deficiencies of T-cell vaccines against cancer.

Introduction

The repertoire of circulating CD8⁺ cytotoxic T lymphocytes (CTLs) is a result of both positive and negative selection processes that occur in the thymus [1]. Immature, double-positive (DP: CD4⁺CD8⁺) T-cells are positively selected in the thymic cortex through their interaction with peptide / MHC molecules on the surface of cortical epithelial cells (cTECs). cTEC-presented peptides arise through processing by the unique thymoproteasome and a full range of thymoproteasome-processed self-peptides are required to produce a T-cell repertoire that can efficiently recognize pathogens [2].

In the medulla, dendritic cells (DCs) and thymic epithelial cells (mTECs) present a vast repertoire of self-peptides that are involved in negative selection at the single-positive T-cell stage (SP: CD4⁺CD8⁻ or CD4⁻CD8⁺). Both mTECs and medullary DCs process peptides through the constitutive proteasome and the more efficient immunoproteasome, and therefore potentially display the same repertoire of self-peptides as presented in the periphery [3]. mTECs exhibit the unique phenomenon of promiscuous gene expression (pGE) induced by the autoimmune regulator (Aire) (reviewed in [4]), in addition to possible epigenetic mechanisms [5]. Such processes result in the potential expression of genes from all tissues of the body. As a result, T cells with a high affinity for self-peptides are deleted from the repertoire, conferring a level of immune tolerance to the host and preventing autoimmune disease.

One key to T-cell selection in these processes is the T-cell receptor (TCR). The αβ TCR expressed by CD8⁺ T cells, recognizes peptides (usually 8–10 amino acids in length) derived mainly from endogenous proteins and presented in the context of MHC class I (pMHC1 (or pHLA in human cells)) [6]. The degenerate nature of TCRs within the T-cell repertoire conferred by positive selection [7, 8] ensures that robustly binding TCRs are available for a broad range of pathogen-derived antigens, leading to a vigorous CTL response.

Tumor cells, on the other hand, appear to evade the CTL response (for a review of immune evasion strategies see [9]). The explanation for such evasion may include down-regulated antigen presentation and the secretion of immune-regulating factors that prevent tumor infiltration and cause CTL fatigue. However; the primary reason for the deficiencies of the CTL response against malignant T cells might simply be recognition failure, caused by a lack of high affinity TCRs. Indeed, the identification of CTLs possessing TCRs with sufficient antigen-sensitivity to recognize tumor associated peptide antigens (TAPAs) is far more challenging than isolation of viral antigen (VA)-specific CTLs. Moreover, despite the observation that a small number of cancer patients, and even some healthy donors, do generate vigorous CTL responses to TAPAs, the success of vaccination strategies, in all but a very few cases, has been dismal [10, 11]. Here, we report a comprehensive single-site study to investigate antigen recognition by VA- and TAPA-specific TCRs. We observe a clear difference between the affinities of both TCR groups. These findings are discussed in terms of thymic selection and their implications for development of cancer therapeutics.

Results and Discussion

Comparison of published HLA-A2 restricted TCR affinities

A number of studies comparing TCR affinities have been reported ([12–14], and references therein), and a tentative assertion made of differing affinities between VA and TAPA specific TCRs, with the virus-specific TCRs binding tighter than cancer-specific ones [12, 14–16]. However, definitive conclusions are difficult to draw for two reasons; first, only a rather limited data set is currently available, and second, small variations in affinity measurements are difficult to resolve given the inevitable methodological differences between individual laboratories. To address these issues a comprehensive panel of TCRs was investigated here (Table 1) and their affinities determined using identical methodology and equipment. The peptide antigens investigated were limited to those presented by HLA-A2, to prevent any influence from variations in CD8 co-receptor affinity between different HLA types.

Table 1.

Binding parameters of virus and cancer specific TCRs to their corresponding target antigen (pHLA-A2*0201)

	TCRa)	Target antigen	Peptide sequence	KD (µM)b)	ΔG (kcal mol−1)	KA (mM−1)	Koff (s−1)	Kon (M−1s−1)c)	t1/2 (s)
Virus specific TCRs	HIVgag (868)	HIVgag77–85 (SL9)	SLYNTVATL	0.18 (+/− 0.01)	−9.20	5550	0.03 (+/− 0.0003)	14 × 104	27
	A6_Tax	HTLV-1 Tax11–19	LLFGYPVYV	2.0 (+/− 0.7)	−7.77	500	0.10 (+/− 0.006)	5.2 × 104	6.7
	JM22_flu	FLU-MP58–68	GILGFVFTL	5.0 (+/− 0.2)	−7.23	200	0.14 (+/− 0.002)	2.8 × 104	4.9
	HIVpol	RT-Pol476–484	ILKEPVHGV	5.6 (+/− 0.4)	−7.16	178	0.67 (+/− 0.08)	12 × 104	1.0
	EBV	EBV LMP2A426–434	CLGGLLTMV	23 (+/− 3)	−6.33	43.5	0.30 (+/− 0.02)	1.3 × 104	2.3
	HCV-1	NS5b2594–2602	ALYDVVTKL	25 (+/− 2)	−6.28	40.0	0.44 (+/− 0.10)	1.8 × 104	1.6
	HCV-2	NS5b2594–2602	ALYDVVTKL	3.9 (+/− 0.1)	−7.37	254	0.20 (+/− 0.010)	5.1 × 104	3.5
	HCV-3	NS31406–1415	KLVALGINAV	3.1 (+/− 0.3)	−7.51	321	0.07 (+/− 0.011)	2.2 × 104	10
	HCV-4	NS31406–1415	KLVALGINAV	13 (+/− 0.8)	−6.66	75.9	0.21 (+/− 0.01)	1.6 × 104	3.3
	HCV-5	NS31073–1081	CINGVCWTV	1.5 (+/− 0.1)	−7.95	676	0.10 (+/− 0.014)	6.5 × 104	7.2
Cancer specific TCRs	1G4 (NY-ESO)	NY-ESO-1157–165	SLLMWITQC	11 (+/− 0.8)	−6.78	93.5	0.38 (+/− 0.02)	3.6 × 104	1.8
	gp100	gp100280–288	YLEPGPVTA	26 (+/− 0.5)	−6.25	38.0	0.50 (+/− 0.06)	1.9 × 104	1.4
	Telomerase	Telomerase540–548 (hTERT)	ILAKFLHWL	34 (+/− 2)	−6.10	29.4	0.14 (+/− 0.01)	0.4 × 104	5.1
	WT1-1	WT1126–134	RMFPNAPYL	45 (+/− 2)	−5.93	22.2	>0.69	NDd)	<0.5
	WT1-2	WT1126–134	RMFPNAPYL	31 (+/− 3)	−6.15	32.3	0.53 (+/− 0.09)	1.7 × 104	1.3
	WT1-3	WT137–45	VLDFAPPGA	156 (+/− 53)	−5.19	6.41	0.12 (+/− 0.01)	0.08 × 104	5.8
	AFP	AFP158–166	FMNKFIYEI	51 (+/− 5)	−5.86	19.6	0.26 (+/− 0.01)	0.5 × 104	2.7
	PSCA	PSCA14–22	ALQPGTALL	48 (+/− 12)	−5.89	20.8	0.69 (+/− 0.01)	1.4 × 104	1.0
	Her2/Neu	Her-2/Neu369–377 (E75)	KIFGSLAFL	53 (+/− 3)	−5.83	18.7	1.1 (+/− 0.14)	2.1 × 104	0.6
	P450	P450CYP1B1190–198	FLDPRPLTV	86 (+/− 12)	−5.55	11.6	0.22 (+/− 0.02)	0.3 × 104	3.1
	Prostein	Prostein31–39	CLAAGITYV	147 (+/− 7)	−5.23	6.80	>0.69	NDd)	<0.5
	Imp-3e)	Imp-3199–207	RLLVPTQFV	387 (+/− 3)	−4.65	2.58	>0.69	NDd)	<0.5
	Trp-p8e)	Trp-p8187–195	GLMKYIGEV	182 (+/− 4)	−5.10	5.49	>0.69	NDd)	<0.5
	5T4	5T4107–115	GAFEHLPSL	95 (+/− 6)	−5.49	10.5	>0.69	NDd)	<0.5

^a)HIV, Human Immunodeficiency Virus; HTLV-1, Human T-Lymphotrophic Virus-I; EBV, Epstein-Barr Virus; HCV, Hepatitis-C Virus; NY-ESO, New York Esophageal; gp100, Pre-Melanosome Protein 100; WT1, Wilms Tumor 1; PSCA, Prostate Stem Cell Antigen; Her-2/Neu, v-erb-b2 Erythroblastic Leukemia viral homolog 2; AFP, alpha-fetoprotein; IMP-3, Insulin-like growth factor-II mRNA-binding protein 3; Trp-p8 (also known as TRPM8), melastatin related transient receptor potential protein.

^b)Error values for K_D were determined from the fit data using BIAevaluation software.

^c)Error values for k_off represent one standard deviation determined from at least 4 independent fits.

^d)ND = not determined.

^e)Maximal TCR binding for K_D fitting was constrained according to the level of active pHLA.

TCR affinities and half-lives depend on the origin of the target antigen (pHLA)

TCR genes were isolated from blood samples, and expressed and prepared as soluble TCRs from E. coli as described in the Materials and Methods. Binding of the 24 TCRs to their specific pHLA-A2 complexes (10 VAs and 14 TAPAs) was analyzed by surface plasmon resonance (SPR) at 25°C. The affinities, in terms of dissociation constants, (K_D) and the dissociation rate constants (k_off) were determined (Table 1). The half-lives (t_1/2) and association rate constants (k_on) were subsequently calculated from the measured values (Table 1). Due to the limitations of SPR resolution (t_1/2 = 0.5 s), dissociation rate constants could not be determined for a number of TAPA specific TCRs that have particularly fast off-rates. Representative binding data for high, intermediate and low affinity TCRs are shown in Supporting Information Figure 1.

A clear pattern was observed in TCR binding parameters correlating with the origin of the target peptide. TCRs recognizing VAs (such as those derived from HIV and influenza) exhibited relatively high affinity with K_D values, between 0.18 and 25 µM (mean = 8.25 µM) while the affinity of TCRs for TAPAs ranged from 10.7 to 387 µM (mean = 96.6 µM). The half-lives were, in general, longer for the VA specific TCRs (mean = 6.8 s) than for the TAPA specific TCRs (mean = <1.8 s). These data are presented graphically in Figure 1.

Figure 1. Binding parameters of virus and cancer specific TCRs to the corresponding pHLA.

TCRs recognizing virus or tumor-associated antigens were isolated and prepared as described in the Materials and Methods. Binding parameters were assessed by Surface Plasmon Resonance (SPR) using a BIAcore3000. Data shown are from one experiment. The calculated values for (A) affinity (K_D) and (B) half-life (t_1/2) shown are from one independent measurement. The horizontal bar denotes the mean. For graphing purposes, TCRs with t_1/2 values, of < 0.5s were assigned a value of 0.5s. The difference between the two population means is statistical significance at the 95% confidence interval determined using an unpaired t test (p = < 0.0001 for K_D and p = 0.0028 for t_1/2). Equal variance, determined using an F test, was first achieved by taking the log of each data point. (C) The relationship between, K_D and t_1/2 is also shown. The different windows for TCR recognition of viral and tumor antigen-specific TCRs are defined by the shaded areas.

This represents comprehensive, single-study evidence for a variation in binding parameters between human TCRs recognizing VA and TAPA pHLAs. Where available, we find no substantial differences between the biophysical data presented here and that reported in the literature. Since each isolated TCR represents one random selection event (with the possibility of higher or lower-affinity TCRs for the same antigen being present in other donors), it was fundamental to investigate a large number of responses. Previous exposure of individuals to antigen can influence their T-cell repertoire by enriching it with antigen-specific memory T cells, resulting from a successful response to the antigen challenge, and presumably connected with expression of high-affinity antigen-specific TCRs [17]. In most cases the medical condition of T-cell donors for our study was unknown, but in all probability some had been previously infected with common viruses such as influenza and EBV, which may have introduced a bias towards higher affinity TCRs for these antigens. However, in cases where previous antigen exposure of the donor is highly likely, it has not always led to selection of robust TCR affinity. For example, the Her-2/Neu TCR, isolated from a breast cancer patient, has a relatively low affinity for the antigen (K_D = 53 µM – Table 1). In contrast, the PSCA TCR was cloned from a healthy donor but has a slightly higher antigen affinity (K_D = 48 µM – Table 1). We therefore suggest it is unlikely that the higher affinities observed for VA-specific TCRs manifest themselves solely as a consequence of previous antigen exposure in the donors.

Binding Parameters and T-cell Activation

The observed differences in binding parameters between TCRs recognizing VAs or TAPAs will confer significantly different levels of antigen sensitivity to T cells and are likely to affect their signaling pathways. T-cell activation is first and foremost driven by TCR binding to antigen, although it remains unclear whether the affinity or kinetics of binding is the determining factor; discrepancies in the correlation of a single binding parameter with T-cell activation have been reported ([18–20] and reviewed in [13]). Despite this debate it is established that, in the naturally selected affinity range, T-cells with TCRs which bind pMHCs with higher affinities and longer half-lives elicit a stronger and more effective immune response. It therefore follows from the data presented here that in general VAs will draw a stronger CTL response than TAPAs. Indeed we have shown that cancer-specific CTLs give a poor functional response to physiological levels of antigen (data not shown).

Thymic selection

The lower affinity of TAPA-specific TCRs, in comparison with their VA-specific counterparts, could be a consequence of negative selection during T-cell maturation within the thymic medulla. Negative selection, in response to antigenic presentation of self peptides, leads to the deletion of T -cells bearing high affinity TCRs to self antigens. Since many TAPAs are also self antigens, high affinity TAPA- specific T cells will be simultaneously deleted from the repertoire. Even for antigens such as NY-ESO-1 [21], whose expression is usually restricted to immune privileged sites, low-levels of mRNA have been detected in thymus [22]. Nevertheless, some TAPA-specific TCRs possessing low to moderate antigen affinity (between 10 and 400 µM – Table 1) do escape thymic deletion; this may occur as a result of promiscuity within the T-cell repertoire. Given that the circulating T cells are required to respond to an extremely large number of foreign antigens, there is a significant degree of degeneracy within the repertoire [7, 8, 23, 24]; indeed, it has been estimated that one TCR can respond to over 1 × 10⁶ peptides when presented in the context of a single MHC type [25]. Therefore, the escape of T cells bearing TCRs with some degree of affinity toward TAPAs is probable. Furthermore, differences in the presentation of certain antigens, resulting from variable gene expression [26] and instability within the peptide MHC complex [27], may also contribute to thymic escape.

Therapeutic Implications

The clear difference in binding parameters between VA- and TAPA- specific TCRs has implications for therapeutic approaches. Vaccines rely on the activation of pre-existing T cells to target tumors; however, since TAPA-specific T cells possess TCRs with relatively low affinities for antigen, vaccines may be largely ineffective in eliciting an effective anti-tumor CTL response. This may provide one explanation for the limited success of such approaches [10, 11]. A more promising strategy, for modulating the immune system to target tumors is through adoptive therapy [28], especially if this is combined with genetically engineered TCRs designed to have a “VA-TCR-like” affinity. Indeed, T cells carrying these enhanced affinity TCRs have been shown to recognize tumor antigens with high avidity [29]. The construction of enhanced affinity TCRs is also central to emerging cancer therapies comprising soluble, bi-specific proteins, such as the recently described ImmTACs. These molecules combine a genetically engineered, picomolar affinity, soluble TCR, with a humanized anti-CD3 antibody, capable of redirecting non-tumor specific T cells [30, 31]. Similar fusions which rely on monoclonal antibody binding to redirect the CTL response have been applied with success [32]. However, the antigens targeted by antibodies are limited to those produced as integral membrane proteins; TCRs meanwhile can recognize the larger pool of intracellular-derived peptides presented in the context of the MHC. Therefore therapeutic agents exploiting enhanced affinity TCRs hold substantial promise.

Concluding Remarks

Immune tolerance to tumors is a critical issue to overcome in the development of effective immunotherapies against cancer. By comparing the binding parameters of individual TCRs to their respective pHLAs, the data presented here provide an enhanced understanding of the role of TCR affinity in tumor immune evasion, informing on the most appropriate strategies for successful therapeutics.

Materials and Methods

Generating peptide specific T-cell lines

CD8⁺ T cells from donors were enriched from freshly prepared peripheral blood by negative selection using microbeads according to the manufacturer’s instructions (Dynal). Dendritic cells (DCs) and activated B-cells were generated as described in [20, 33]. Purified CD8⁺ cells were cultured in CTL medium: IMDM (Invitrogen), 10% human AB serum (Sera Laboratories Int.), 100 U/ml penicillin, 100 µg/ml streptomycin, 1% glutamine (Invitrogen), supplemented with IL-7 at 10 ng/ml and autologous peptide pulsed irradiated DCs were added in a 5:1 ratio (T cells : DCs). For secondary and subsequent stimulations peptide pulsed irradiated activated B-cells were used, mixed at 4:1 ratio (T cells : B cells) and from day 18 IL-2 (100 U/ml) was added to the culture. Following three stimulations T-cells were stained with specific pMHC tetramers, and positive cells were sorted using FACSaria cell sorter (BD Biosciences). Sorted cells were then grown to 500 cells per well to produce cell lines. Alternatively, peptide-specific CD8 T cells were generated from whole peripheral blood mononuclear cells stimulated with cognate peptides and rIL2 at 100U/ml for 10 days, stained with specific pMHC tetramers and FACS-sorted for tetramer+ CD8 T cells before RNA extraction for TCR analysis.

Production of soluble mTCRs

Soluble mTCRs were produced as previously described [34]. Briefly, DNA coding α and β chains of the TCRs were isolated from peptide specific T-cell lines by PCR using cDNA as a template and cloned into a bacterial expression vector. TCR chains were then expressed in E. coli as inclusion bodies and soluble disulphide-linked heterodimeric mTCRs were refolded from denatured inclusion bodies and purified by anion exchange and size exclusion chromatography.

Production of soluble pMHCs

Specific peptides (>95% purity) were obtained from Peptide Protein Research and dissolved in DMSO at 4 mg/ml prior use. BirA tagged human HLA-A2*0201 and β-2 microglobulin were expressed in E. coli, purified as inclusion bodies and refolded with appropriate peptide [35]. Refolded pMHCs were purified by anion exchange and size exclusion chromatography and biotinylated in vitro using BirA ligase (Avidity) [36].

Biophysical measurements using BIAcore Surface Plasmon Resonance (SPR)

Purified mTCRs were subjected to SPR analysis on a BIAcore3000. Briefly, biotinylated specific and control pMHC monomers were immobilized on to a streptavidin-coupled CM-5 sensor chips. All measurements were performed at 25°C in PBS buffer (Sigma) supplemented with 0.005% Tween (Sigma) at a flow rate of 10 µl/min. To measure affinity, serial dilutions of the mTCR were flowed over the immobilized pMHCs and the response values at equilibrium were determined for each concentration. Typically an initial TCR concentration of at least twice the measured KD value was used. For Imp-3 and Trp-p8 TCRs the starting TCR concentration used was lower than optimal, due to TCR aggregation at high concentrations. To increase accuracy of the fitting we first measured the level of active pHLA on the chip by injecting saturating amounts of high affinity ILT2. In this way curve fitting was improved by constraining theoretical maximum TCR binding according to the level of active pHLA. Equilibrium Dissociation constants (K_D) were determined by plotting the specific equilibrium binding against protein concentration followed by a least squares fit to the Langmuir binding equation, assuming a 1:1 interaction. Dissociation rate constant (k_off) was determined by dissociation curve fitting to 1:1 binding model using BIAevaluation software and half-lives calculated from: t_1/2=ln2/k_off. Association rate constant (k_on) and association constant (K_A) were calculated from: k_on=k_off/K_D; K_A=1/K_D. Gibbs free energy change (ΔG) was calculated from: ΔG= −RT ln(K_D).

Statistical analysis

The differences between the calculated means for virus- and tumor-specific TCRs, in terms of affinity (K_D) and half-life (t_1/2), were evaluated for statistical significance using an unpaired t test. Equal variance, determined using an F test, was first achieved by taking the log of each data point. The reported P values were determined at the 95% confidence interval.