Thioamide Analogues of MHC I Antigen Peptides

Abstract

Short, synthetic peptides that are displayed by the major histocompatibility complex I (MHC I) can stimulate CD8 T cells in vivo to destroy virus-infected or cancer cells. The development of such peptides as vaccines that provide protective immunity, however, is limited by rapid proteolytic degradation. Introduction of unnatural amino acid residues can suppress MHC I antigen proteolysis, but the modified peptides typically display lower affinity for the MHC I and/or diminished ability to activate CD8 T cells relative to native antigen. Here, we report a new strategy for modifying MHC I antigens to enhance resistance to proteolysis while preserving MHC I affinity and T cell activation properties. This approach, replacing backbone amide groups with thioamides, was evaluated in two well-characterized antigens presented by HLA-A2, a common human MHC I. For each antigen, singly-modified thioamide analogues retained affinity for HLA-A2 and activated T cells specific for the native antigen, as measured via IFN-γ secretion. In each system, we identified a highly potent triply-substituted thioamide antigen (“thio-antigen”) that displayed substantial resistance to proteolytic cleavage. Collectively, our results suggest that thio-antigens may represent a general and readily accessible source of potent vaccine candidates that resist degradation.

Introduction

CD8 T cell receptors (TCRs) recognize short peptide antigens, typically 8–10 residues, presented by a major histocompatibility complex I (MHC I) on the surface of an antigen-presenting cell.^1,2 Interaction between the TCR and the peptide+MHC I complex (pMHC I) activates the T cell and triggers a cascade of responses that can lead to elimination of abnormal and infected cells.^3–5 Synthetic peptides that correspond to MHC I epitopes are competent to induce antigen-specific CD8 T cell responses;^6,7 such peptides have long been of interest as a basis for vaccine development.^8–10 Extensive prior study has revealed a variety of factors, including the stability of pMHC I, that influence the potency of natural and modified antigen peptides in stimulating CD8 T cell activation.^{3,5,7,11–22} However, this approach to vaccine development has been hampered by the fact that short peptides comprised of proteinogenic L-α-amino acid residues are highly susceptible to cleavage by proteases.

Multiple strategies have been explored to suppress proteolytic susceptibility of MHC I peptides in vivo, such as replacement of one native L-α residue with a D-α- or β-amino acid residue,^23–29 with a peptoid (N-alkyl-glycine) subunit,³⁰ or multiple such replacements.^23,24,30 Although these approaches can be very effective for inhibiting protease action,^31,32 both antigen-MHC I affinity and TCR recognition of the pMHC I are very sensitive to structural alterations within the antigen. No general strategy has been identified for modifying peptide antigens in a manner that retains the necessary MHC I and TCR recognition properties while diminishing cleavage by proteases.

Here, we evaluate a type of modification that has not previously been explored in the context of MHC antigen peptides, to our knowledge: replacement of a backbone amide group with a thioamide. This alteration is relatively subtle in terms of molecular structure, but thioamide linkages between amino acid residues are less susceptible to protease action relative to the natural amide linkages.^33–41 Key physical differences are: (1) a secondary thioamide is a stronger H-bond donor than the analogous amide;⁴² (2) the sulfur atom is a weaker H-bond acceptor than the oxygen atom;⁴³ and (3) the C=S bond is longer than the corresponding C=O bond.⁴⁴ Because of these physical differences, amide-to-thioamide modification can influence the folding propensities of polypeptides.^45–49

It is not possible to predict a priori whether or how swapping backbone amide groups with thioamide groups might affect MHC I antigen properties. To address this uncertainty, we have evaluated the effects of single thioamide replacements in the context of two well-studied antigens, one viral and the other cancer-associated. Both antigens are restricted by the human leukocyte antigen-A2 (HLA-A2), an MHC I allele widely distributed in the human population.^50,51 Most of these “thio-antigens” containing a single thioamide unit displayed strong HLA-A2 binding and evoked potent T cell responses, as measured by stimulation of secretion of interferon-γ (IFN-γ), a proinflammatory cytokine. Guided by the data for these single-replacement thio-antigens, we designed analogues containing two or three dispersed thioamide units. These analogues were comparable to the natural antigen or more potent in terms of IFN-γ secretion. In addition, some thio-antigens with multiple substitutions displayed substantial resistance to proteolytic cleavage. Our findings suggest that thioamide derivatives of MHC I antigens may represent a general basis for peptide vaccine development.

Results and Discussion

Experimental design.

We selected two well-known HLA-A2-restricted antigens, GILGFVFTL (GIL) and ELAGIGILTV (ELA), for evaluation of the thio-antigen approach. HLA-A2 is found in ~50% of the human population.^50,51 Both GIL and ELA have been evaluated in vitro or in clinical trials as potential peptide vaccines.^24,52–56 Antigens containing 9 or 10 residues appear to be optimal for HLA-A2 presentation,^57,58 and we wished to evaluate one antigen of each length. Crystallographic evidence shows that ELA and GIL differ considerably in presentation of the HLA-A2-bound peptide to cognate TCRs (Figure S1). HLA-A2-bound ELA adopts a conformation common among MHC I antigens,⁵⁹ with a central bulge that protrudes away from the binding groove on HLA-A2 (Figure S1A).⁶⁰ This bulged segment makes contacts with the TCR. In contrast, GIL in complex with HLA-A2 lacks a central bulge, and the central residues of GIL are oriented toward HLA-A2 rather than the TCR (Figure S1B).^61–64 Because of these structural differences, we reasoned that results obtained for thioamides derived from both antigens would collectively test the generality of the thio-antigen design hypothesis. Beyond these two examples of HLA-A2-bound peptide conformations, additional examples of uncommon peptide conformations when bound to the MHC I have been reported (summarized by Hopkins et al.⁶⁵).

Thiobenzotriazolide derivatives of L-α-amino acids, with Fmoc protection of the backbone amine, were used for incorporation of thioamide units via solid-phase synthesis.^66–69 Initially, a single thioamide group was introduced at most sites within GIL or ELA. In each case, we did not examine thioamides at primary anchor residues, i.e., Ile2 or Leu9 for GIL, or Leu2 or Val10 for ELA, because we wished to preserve strong affinity between these peptides and the MHC I. Side chains of anchor residues project into pockets at the bottom of the antigen-recognition groove of the MHC I.⁷⁰ Engagement of these pockets is critical for the stability of the pMHC I complex.⁷¹ In addition to the anchor residue side chains, the backbone nitrogen and oxygen atoms of the anchor residues form many intimate interactions with MHC I residues (Figure S2, S3 and Table S2, S3). Future studies may reveal that amide-to-thioamide replacements are tolerated at anchor positions.

Three assays were used to assess the performance of the thio-antigens. The first measured the ability of thio-antigens to stabilize HLA-A2 on the surface of T2 cells. Peptides that bind with higher affinity to HLA-A2 will generate greater levels of pHLA-A2 complexes on the surface of these cells.^11,72 The amount of pHLA-A2 on the cell surface can be quantified by flow cytometry using a fluorescently labeled anti-HLA-A2 antibody. The data generated by varying the thio-antigen concentration enabled determination of EC₅₀ (i.e., the thio-antigen concentration required to produce half-maximal response) and % Max (i.e., the maximum amount of HLA-A2 stabilized by a peptide). The natural antigens, GIL and ELA, were used to define 100% response for each comparative dataset.

To complement the cell-based HLA-A2 stabilization measurements, we evaluated binding of the thio-antigens using purified HLA-A2 complexes. These experiments measured the ability of each thio-antigen to compete with a radiolabeled peptide known to bind strongly to HLA-A2 (the Phe6→Tyr analogue of HBV core 18–27, FLPSDYFPSV, for which K_D = 2.5 nM).⁷³ These measurements provided IC₅₀ values (i.e., the thio-antigen concentration required to produce half-maximal inhibition of radiolabeled tracer binding).

T cell activation assays were carried out using T2 cells as the antigen-presenting cells and antigen-specific CD8 T cells, which can recognize the pHLA-A2 complex displayed by the T2 cells. The CD8 T cells displayed specificity for either the GIL or ELA antigen; therefore, our assays determined how well thio-antigens derived from GIL or ELA mimic the parent antigen in terms of recognition by a cognate TCR in the context of the pHLA-A2 complex. We quantified T cell response by measuring levels of secreted IFN-γ using ELISA. The data generated by varying the thio-antigen concentration enabled determination of EC₅₀ and % Max values; the maximal responses with GIL or ELA were defined as 100%.

Antigen-specific CD8 T cells used in this study were isolated from human HLA-A2+ donors, which means that they carried the MHC I necessary for presenting the GIL or ELA antigen to their own T cells. These T cells (97–98% CD8+; Table S4) are presumed to have been polyclonal, which means that more than one TCR was represented within the sample. We reasoned that a polyclonal antigen-specific T cell population would provide the most useful insight on T cell activation by the thio-antigens in vivo because each person is likely to carry more than one TCR within the T cell population that recognizes a specific pMHC I. Among the GIL-specific cell sample, 46% of the CD8 T cells stained positively with the GIL antigen tetramer, while among the ELA-specific cells, 95% of the CD8 T cells stained positively with the ELA antigen tetramer (Table S4).

Thio-antigen analogues of GIL.

GIL is a highly conserved immunodominant epitope derived from the influenza A virus M1 protein.^74,75 Unlike most antigens presented by HLA-A2, which form a central bulge that contacts the TCR,⁵⁹ GIL is presented with side chains from the central region inaccessible to the TCR. Only the side chain of Thr8, near the C-terminus of this antigen, is exposed to solvent (Figure S1B).^61–64 This manner of antigen presentation is evident in multiple pMHC I+TCR structures involving the JM22,⁶³ F50,⁶² F6⁶⁴ or F8⁶⁴ TCR, all of which have evolved to recognize HLA-A2-bound GIL largely via the peptide backbone.

Analysis of a crystal structure of a ternary complex containing the GIL antigen, the HLA-A2 MHC I and the JM22 TCR revealed a large network of contacts between backbone amide units of the antigen and the MHC I or the TCR (Figure S3 and Table S3).⁶³ (We selected JM22 for this analysis because this TCR has been studied to a greater extent than the F50, F6 or F8 TCRs.) The two exceptions were backbone amide oxygen and nitrogen of Phe5 and Phe7 in GIL; however, the side chains of Phe5 and Phe7 make contacts with MHC I residues.⁷⁰ Backbone amide groups in the Gly1-Leu3 and Thr8-Leu9 segments of GIL make extensive contacts with MHC I residues, while central residues of GIL, Gly4 and Val6, make contacts with five residues of JM22 TCR (Table S3). The backbone oxygen atom of Gly4, alone, contacts four TCR residues. Beyond these direct interactions, there are nine H-bond contacts via bridging water molecules between the TCR and pMHC I.⁶³

Nearly all of the singly-substituted GIL derivatives (GIL-1 to GIL-8; Figure 1A) were comparable in potency to GIL itself in terms of stabilizing HLA-A2 on the T2 cell surface, as indicated by EC₅₀ values (Figure 1B, Figure S4, Table 1). The lone exception to this trend was GIL-7, which contained an O→S replacement at Phe7, a known HLA-A2 secondary anchor position;⁷¹ the HLA-A2 stabilization assay suggested that the pMHC I complex formed with GIL-7 is relatively weak (Figure 1B, Figure S4, Table 1). The other singly-substituted thio-antigens achieved maximum HLA-A2 levels on the cell surface comparable to or higher than that achieved by the native antigen. Those analogues that exceeded the maximum achieved with GIL may form a more stable complex with HLA-A2 relative to the native antigen as a result of amide-to-thioamide substitution.

Figure 1.

Structures and activities of thioamide-containing derivatives of the GIL antigen. (A) Sequences of GIL and analogues; sites of amide-to-thioamide replacement are indicated by blue ovals. (B) HLA-A2 stabilization on the surface of T2 antigen presenting cells by GIL and thioamide variants. Data points represent the average of ≥2 independent experiments. (C) Activation of GIL-specific CD8 T cells by GIL and thioamide variants, as measured by levels of secreted IFN-γ. Data points were generated from ≥2 independent experiments. All error bar uncertainties are expressed as S.E.M.

Table 1.

HLA-A2 stabilization and in vitro HLA-A2 affinity responses for GIL and thio-antigens.

	HLA-A2 Stabilization				In vitro Competition HLA-A2 Affinity
Peptide	pEC50	EC50 (μM)	EC50 relative	% Max	pIC50	IC50 (nM)	IC50 relative
GIL	5.8 ± 0.07	1.5	1	100 ± 4	7.8 ± 0.11	16	1
GIL-1	6.1 ± 0.05	0.84	0.6	112 ± 3	ND	<0.25	<0.016
GIL-3	5.8 ± 0.06	1.7	1	144 ± 5	7.5 ± 0.04	34	2
GIL-4	5.3 ± 0.07	5.6	4	194*	ND	<0.25	<0.016
GIL-5	5.4 ± 0.05	4.1	3	141 ± 4	7.1 ± 0.26	71	4
GIL-6	5.3 ± 0.05	5.2	4	176*	7.0 ± 0.15	112	7
GIL-7	4.3 ± 0.16	49	30	93*	5.6 ± 0.04	2600	160
GIL-8	5.5 ± 0.07	3.5	2	126 ± 5	6.8 ± 0.20	150	9
GIL-1,8	5.9 ± 0.05	1.2	0.8	100 ± 2	7.0 ± 0.04	99	6
GIL-1,4,8	6.2 ± 0.06	0.67	0.4	177 ± 5	7.1 ± 0.08	88	6
GIL-1,5,8	5.9 ± 0.06	1.2	0.8	183 ± 5	7.0 ± 0.08	94	6

Left: HLA-A2 stabilization pEC₅₀, EC₅₀ and % Max values are obtained from ≥2 independent experiments. pEC₅₀ indicates the negative logarithm of the half-maximal effective concentration (EC₅₀). EC₅₀ relative implies the HLA-A2 stabilization potency normalized to GIL: (thioamide analogue/GIL). pEC₅₀ and % Max uncertainties are expressed as S.E.M.

^*Value shown indicates % Max at the highest peptide concentration (50 μM); the fitted curve did not reach a saturation point. Right: In vitro HLA-A2 affinity data are represented by pIC₅₀ and IC₅₀ values obtained from ≥6 independent experiments. pIC₅₀ indicates the negative logarithm of the half-maximal inhibitory concentration (IC₅₀). pIC₅₀ uncertainties are expressed as S.E.M. ND = Not Determined. IC₅₀ relative indicates the HLA-A2 affinity normalized to GIL by the quotient (thioamide analogue/GIL). GIL-1 and GIL-3 bound HLA-A2 with an affinity below the threshold of the assay, as indicated by approximated values: <0.25 and <0.016 IC₅₀ and IC₅₀ relative, respectively.

The in vitro HLA-A2 affinity assay is considerably more sensitive than the HLA-A2 stabilization assay involving T2 cells.²⁵ The range of IC₅₀ values observed among the seven GIL analogues containing a single thioamide was larger than the range of EC₅₀ values from the T2 assay (Table 1). These two assay formats were consistent in identifying GIL-7 as having substantially lower affinity relative to all other members of this thio-antigen series. The in vitro assay suggested that GIL-1 and GIL-4 bind to HLA-A2 with significantly higher affinity than does the native GIL antigen; however, this feature was not evident in the T2 assay (Table 1). These observations raise the possibility that some thioamide locations can increase affinity of a peptide for the MHC I relative to the native antigen.

In the reported crystal structure of the complex formed by the antigen GIL, the MHC I HLA-A2 and the TCR JM22, there are contacts (≤4 Å) that seem to represent H-bonds between the MHC I and backbone oxygen and/or nitrogen atoms of the antigen that could be affected by the amide-to-thioamide replacements in GIL-1 and GIL-8 (Figure S3 and Table S3).⁶³ Nevertheless, both GIL-1 and GIL-8 display substantial affinity for the MHC I, as do the thio-antigens that contain two or three substitutions, GIL-1,8, GIL-1,4,8 and GIL-1,5,8 (Table 1). In the less sensitive HLA-A2 stabilization assay, all five of these thio-antigens display EC₅₀ values that are very similar to the EC₅₀ for GIL itself (Figure 1B, Figure S4 and Table 1). In the more sensitive in vitro assay, there is a larger variation (Table 1). GIL-8, GIL-1,8, GIL-1,4,8 and GIL-1,5,8 seem to bind to HLA-A2 moderately less strongly than does GIL, but the IC₅₀ values for these thio-antigens are within ~10-fold of the GIL IC₅₀. The IC₅₀ value for GIL-1 indicates significantly stronger binding relative to GIL. Overall, considering the MHC I binding data in Table 1 in the context of the ternary complex crystal structure suggests that antigen-MHC I affinity is not substantially diminished and may even be enhanced by amide-to-thioamide replacement at backbone positions in the antigen that form H-bonds with the MHC I. The crystal structure does not suggest an explanation for the loss of affinity observed for GIL-7.

Each of the singly-substituted GIL analogues stimulated IFN-γ secretion from CD8 T cells, but none matched the potency of the native antigen in T cell activation, as manifested by EC₅₀ values (Figure 1C, Figure S5, Table 2). The most potent among these analogues was GIL-1, which was the only GIL analogue containing a single thioamide to achieve a maximum in IFN-γ secretion that was comparable to the maximum observed with the native antigen. Among GIL-3 to GIL-8, EC₅₀ values were ~20- to 300-fold higher than that of the native antigen, and these thio-antigens achieved maxima 50% to 74% of the native antigen’s maximum (Table 2).

Table 2.

T cell activation responses for GIL and thio-antigens derived from dose-responses in Fig.1C.

	T Cell Activation
Peptide	pEC50	EC50 (pM)	EC50 relative	% Max
GIL	10.8 ± 0.2	17	1	95 ± 4
GIL-1	10.0 ± 0.2	97	5.7	100 ± 5
GIL-3	8.9 ± 0.2	1300	76	74 ± 5
GIL-4	9.3 ± 0.4	540	32	63 ± 7
GIL-5	9.4 ± 0.6	370	22	50 ± 9
GIL-6	8.6 ± 0.4	2600	150	68 ± 8
GIL-7	8.3 ± 0.3	5300	310	71 ± 7
GIL-8	9.1 ± 0.3	850	50	61 ± 6
GIL-1,8	11.9 ± 0.3	1.4	0.082	78 ± 6
GIL-1,4,8	12.3 ± 0.2	0.49	0.029	72 ± 4
GIL-1,5,8	12.2 ± 0.2	0.69	0.041	70 ± 4

T cell activation pEC₅₀, EC₅₀ and % Max values are derived from ≥2 independent experiments. pEC₅₀ implies the negative logarithm of the half-maximal effective concentration (EC₅₀). EC₅₀ relative implies the T cell activation potency normalized to GIL: (thioamide analogue/GIL). pEC₅₀ and % Max uncertainties are expressed as S.E.M.

We examined GIL derivatives containing multiple amide-to-thioamide replacements because our long-term goal is to generate functional antigens that are less susceptible to proteolysis relative to antigens constructed entirely of α-amino acid residues. A single thioamide is unlikely to offer protection from proteolysis across the entire length of a 9-mer peptide. The sites of substitution were selected based on the T cell activation results obtained for the single substitution series (Figure 1C, Figure S5, Table 2) and a desire to distribute thioamide groups along the backbone. GIL-1,8 contains one thioamide near each end. GIL-1,4,8 and GIL-1,5,8 share these two terminal substitutions and contain a third substitution near the center. All three were comparable to the native antigen in the HLA-A2 stabilization assay with T2 cells, and all three displayed moderately lower affinity (approximately six-fold higher IC₅₀) relative to the native antigen in the in vitro HLA-A2 binding assay (Figure 1B, Figure S4, Table 1). We were surprised to observe that GIL-1,8, GIL-1,4,8 and GIL-1,5,8 were significantly more potent than the native antigen in the T cell activation assay, although in each case the maximum IFN-γ secretion was modestly lower than for the native antigen (Figure 1C, Figure S5, Table 2). The consistently low EC₅₀ values observed for the thio-antigens containing multiple substitutions raise the possibility of favorable cooperativity among the thioamide units in terms of T cell activation.

In the reported crystal structure of the complex formed by the antigen GIL, the MHC I HLA-A2 and the TCR JM22, there are contacts (≤4 Å) that seem to represent H-bonds between the TCR and backbone oxygen and/or nitrogen atoms in the antigen that would be affected by the amide-to-thioamide replacements in GIL-4, GIL-5 and GIL-6 (Figure S3 and Table S3).⁶³ Each of these thio-antigens was less potent than the native antigen in terms of stimulating IFN-γ secretion (higher EC₅₀; Figure 1C, Figure S5, Table 2), and each caused a lower level of maximum IFN-γ secretion relative to the native antigen, but these behaviors were seen also for GIL-3, GIL-7 and GIL-8. Therefore, it is possible that altering antigen-TCR contacts via amide-to-thioamide replacement can cause a decline in T cell activation potency, but similar or larger declines can result from replacements that do not involve direct TCR contacts. The potent T cell stimulation observed with GIL-1,4,8 and GIL-1,5,8 suggests that the impact of backbone replacements at TCR contact sites, such as Gly4 or Phe5 of the GIL antigen, depends on context.

We examined the susceptibility of GIL-1,4,8 and GIL-1,5,8 to proteolysis with proteinase K, an aggressive protease with low substrate specificity.⁷⁶ MHC I antigens would not naturally be exposed to proteinase K, a fungal serine protease; however, this protease provides a stringent test for the ability of amide-to-thioamide replacements to protect short peptides from enzymatic cleavage. The native antigen was used as a point of comparison. The PeptideCutter tool in Expasy predicted six cleavage sites for GIL (after Ile2, Leu3, Phe5, Val6, Phe7 and Thr8; Figure 2A). Under our assay conditions, GIL displayed a half-life of 1.3 min (Figure 2B, Figure S6). GIL-1,4,8, with thioamides at Gly1, Gly4 and Thr8, was not significantly protected from cleavage by proteinase K relative to GIL (half-life of 2.1 min; Figure 2B, Figure S6). This outcome may reflect the fact that two of the O→S modifications in GIL-1,4,8, at Gly1 and Gly4, occur at positions that are not predicted to be cleavage sites for proteinase K (Figure 2A). Shifting the central thioamide by one position, to generate GIL-1,5,8, led to a significant increase in half-life (60.4 min; Figure 2B, Figure S6), suggesting that locations of amide-to-thioamide replacements are critical for achieving global proteolytic protection, at least with proteinase K. Future studies will reveal whether comparable positional sensitivities are observed in other proteolytic environments (e.g., serum).

Figure 2.

(A) Predicted cleavage sites of GILGFVFTL antigen for proteinase K; prediction was performed by the PeptideCutter tool in the Expasy resource. Numbers indicate positions of predicted cleavage sites. (B) Proteolysis by proteinase K for GIL, GIL-1,4,8 and GIL-1,5,8 peptides evaluated in DPBS, pH 7.4. Peptide sequences are shown in Fig. 1A. Numbers above bars indicate peptide half-life (min).

Thio-antigen analogues of ELA.

ELA is an anchor-modified heteroclitic variant of two natural epitopes derived from the melanoma antigen A protein, AAGIGILTV (AAG) and EAAGIGILTV (EAA).⁵⁶ Replacement of Ala with Leu (underlined) causes ELAGIGILTV to bind HLA-A2 with increased affinity relative to either of the natural antigens.^56,72,77 Enhanced HLA-A2 binding may explain why ELA is more immunogenic than natural antigen, i.e., why ELA induces far greater frequencies of antigen-specific CD8 T cells both in vitro and in vivo compared to AAG or EAA.^56,78,79 CD8 T cells elicited by vaccination with ELA induce lysis of cells that display the natural EAA antigen in pMHC I complexes.⁸⁰ HLA-A2-bound ELA adopts a bulged conformation in complex with the MEL5 TCR (Figure S1A); this TCR was isolated from the CD8 T cell clone generated by stimulating peripheral blood mononuclear cells of a healthy HLA-A2+ donor with the ELA peptide.⁶⁰ The TCR is centered on the bulged region, residues 4–7 of ELA, and these residues make the majority of peptide-TCR contacts.⁶⁰ The architecture of the ELA pMHC I+TCR complex differs from the architecture observed in multiple GIL pMHC I+TCR complexes, as noted above.^61–64

Analysis of a crystal structure of a ternary complex containing the ELA antigen, the HLA-A2 MHC I and the MEL5 TCR revealed that 9 out of 10 ELA residues make contacts between backbone amide atoms (nitrogen or oxygen) and the MHC I and/or the TCR (Figure S2 and Table S2).⁶⁰ (We chose MEL5 TCR for this analysis because it is the only TCR for which there is a reported structure when bound to the ELA-HLA-A2 complex.) Only Leu8 of ELA lacks direct contacts involving backbone atoms. Similar to the GIL antigen, the first three N-terminal and the last two C-terminal residues of ELA interact with the MHC I, while the central segment, Gly4-Ile7, mainly interacts with TCR residues. Gly6 is an exception to these trends, with a very close contact involving Gln155 of HLA-A2. Unlike the GIL antigen, which forms backbone interactions with five different TCR residues (Figure S3 and Table S3), two MEL5 TCR residues (Gln31 and Leu98) are responsible for contacts with the ELA antigen backbone (Figure S2 and Table S2).

Nearly all of the singly-substituted thio-antigens (ELA-1 to ELA-9, Figure 3A) were comparable in potency to ELA itself in stabilizing HLA-A2 on the surface of T2 cells, as revealed by the EC₅₀ values in Table 3. This trend parallels our observations for the GIL series (Table 1). The one exception to this trend, ELA-8, which was generated by placing the thioamide at the third position from the C-terminus (Figure 3B, Table 3), as was previously observed with GIL-7 (Table 1). We interpret this observation to indicate relatively weak affinity of ELA-8 for HLA-A2. Together, the data for GIL-7 and ELA-8 indicate that an O→S replacement at p7 of a 9-mer or p8 of a 10-mer peptide is likely to diminish the peptide’s ability to form stable pMHC I complexes. Maximum HLA-A2 stabilization was very similar among all singly-substituted thio-antigens except ELA-8 (Figure 3B, Figure S7, Table 3).

Figure 3.

Structures and activities of thioamide-containing derivatives of the ELA antigen. (A) Sequences of ELA and derivatives; orange ovals indicates sites of thioamide replacement. (B) Stabilization of HLA-A2 on the surface of T2 cells by ELA and thioamide variants. Data points represent the average of ≥2 independent experiments. (C) Activation of ELA-specific CD8 T cells, as detected via IFN-γ production, by ELA and analogues. Data points were generated from ≥3 independent experiments. All error bar uncertainties are expressed as S.E.M.

Table 3.

HLA-A2 stabilization and in vitro HLA-A2 affinity responses for ELA and thio-antigens.

	HLA-A2 Stabilization				In vitro Competition HLA-A2 Affinity
Peptide	pEC50	EC50 (μM)	EC50 relative	% Max	pIC50	IC50 (nM)	IC50 relative
ELA	6.0 ± 0.05	1.0	1	100 ± 2	7.4 ± 0.04	42	1
ELA-1	5.8 ± 0.08	1.5	2	97 ± 4	6.7 ± 0.08	200	5
ELA-3	6.1 ± 0.09	0.71	0.7	100 ± 4	7.0 ± 0.08	95	2
ELA-4	5.5 ± 0.13	2.9	3	98 ± 7	6.9 ± 0.08	130	3
ELA-5	5.9 ± 0.15	1.2	1	111 ± 8	8.2 ± 0.08	6.9	0.2
ELA-6	5.5 ± 0.11	3.4	3	111 ± 7	7.0 ± 0.04	100	2
ELA-7	5.3 ± 0.09	4.7	5	87 ± 5	6.2 ± 0.08	580	14
ELA-8	4.6 ± 0.04	24	20	77*	5.8 ± 0.08	1500	36
ELA-9	5.7 ± 0.09	1.9	2	104 ± 5	7.5 ± 0.11	33	0.8
ELA-1,9	5.8 ± 0.04	1.7	2	95 ± 2	6.3 ± 0.18	530	13
ELA-1,6,9	5.6 ± 0.06	2.3	2	86 ± 3	6.1 ± 0.15	790	19

Left: HLA-A2 stabilization pEC50, EC₅₀ and % Max values are derived from ≥2 independent experiments. pEC₅₀ indicates the negative logarithm of the half-maximal effective concentration (EC₅₀). EC₅₀ relative implies the HLA-A2 stabilization potency normalized to ELA by the quotient (thioamide analogue/ELA). pEC₅₀ and % Max uncertainties are expressed as S.E.M.

^*Value shown describes % Max at the highest peptide concentration (50 μM); the fitted curve did not reach a saturation point. Right: In vitro HLA-A2 affinity data are represented by mean pIC₅₀ and IC₅₀ values obtained from 6 independent experiments. pIC₅₀ indicates the negative logarithm of the half-maximal inhibitory concentration (IC₅₀). pIC₅₀ uncertainties are expressed as S.E.M. IC₅₀ relative indicates the HLA-A2 affinity normalized to ELA by the quotient (thioamide analogue/ELA).

The in vitro binding assay revealed a slightly larger spread of IC₅₀ values among ELA-1 to ELA-9, ~40-fold, when compared to the ~20-fold spread among the HLA-A2 stabilization EC₅₀ values (Table 3). This trend displays the greater sensitivity of the in vitro assay relative to the T2 cell assay, which was manifested also in the GIL series (Table 1). The two assays that probe the stability of the pHLA-A2 complex were consistent in identifying ELA-8 as the weakest binder among the singly-substituted ELA thio-antigens. ELA-5 appeared to bind more tightly than ELA itself in the in vitro assay (Table 3).

In the reported crystal structure of the complex formed by the antigen ELA, the MHC I HLA-A2 and the TCR MEL5, there are contacts (≤ 4 Å) that seem to represent H-bonds between the MHC I and the backbone oxygen or nitrogen atoms of the antigen that could be affected by the amide-to-thioamide replacements in ELA-1, ELA-3 and ELA-9 (Figure S2 and Table S2).⁶⁰ All three of these thio-antigens display affinities for the MHC I that are very similar to the native antigen’s affinity (Table 3). Thus, as was concluded from the GIL series, it appears that MHC I affinity is not necessarily affected by amide-to-thioamide replacements at antigen sites that form H-bonds to the MHC I. The crystal structure does not suggest an explanation for the loss of affinity observed for ELA-8 (Figure S2 and Table S2).

The trend in T cell activation, as indicated by IFN-γ secretion, among singly-substituted ELA thio-antigens differed from the trend observed among singly-substituted GIL thio-antigens. Five among ELA-1 to ELA-9 were comparable to ELA itself in terms of potency, as indicated by EC₅₀ values (Figure 3C, Figure S8, Table 4). Among these potent thio-antigens, ELA-1, ELA-3 and ELA-6 matched the maximum achieved by ELA itself; ELA-7 and ELA-9 had moderately lower maxima. The least potent thio-antigens were ELA-4 and ELA-5 (Figure 3C, Figure S8, Table 4); backbone atoms at p4 and p5 of ELA are critical for the recognition by the TCR (Figure S2 and Table S2). It is possible that O→S modifications at these sites diminished the peptide’s ability to be recognized by ELA-specific T cells.

Table 4.

T cell activation responses for ELA and thio-antigens derived from dose-responses in Fig.3C.

	T Cell Activation
Peptide	pEC50	EC50 (nM)	EC50 relative	% Max
ELA	9.9 ± 0.1	0.14	1	98 ± 4
ELA-1	10.0 ± 0.1	0.098	0.7	93 ± 4
ELA-3	10.2 ± 0.2	0.064	0.5	100 ± 5
ELA-4	7.5 ± 0.1	29	200	54 ± 2
ELA-5	8.1 ± 0.1	8.9	60	55 ± 3
ELA-6	9.9 ± 0.1	0.13	0.9	101 ± 4
ELA-7	9.9 ± 0.1	0.14	1	85 ± 3
ELA-8	7.4 ± 0.1	37	300	60 ± 4
ELA-9	9.8 ± 0.2	0.16	1	73 ± 4
ELA-1,9	10.4 ± 0.2	0.037	0.3	68 ± 3
ELA-1,6,9	10.1 ± 0.1	0.081	0.6	83 ± 4

T cell activation pEC50, EC₅₀ and % Max values are derived from ≥3 independent experiments. pEC₅₀ implies the negative logarithm of the half-maximal effective concentration (EC₅₀). EC₅₀ relative implies the T cell activation potency normalized to ELA by the quotient (thioamide analogue/ELA). pEC₅₀ and % Max uncertainties are expressed as S.E.M.

We examined two ELA analogues containing multiple thioamides, ELA-1,9, with two substitutions, one near each terminus, and ELA-1,6,9 with a third substitution toward the middle (Figure 3A). Both were comparable to ELA in the HLA-A2 stabilization assay, but the in vitro binding assay indicated that both of these thio-antigens bound less tightly to HLA-A2 relative to ELA (Figure 3B, Figure S7, Table 3). The backbone amide atoms of these sites, Glu1, Gly6 and Thr9, appear important because each one of them makes at least one measurable contact with the MHC I residue (Table S2). Despite what appeared to be significantly diminished MHC I affinity for ELA-1,9 and ELA-1,6,9 relative to ELA, both of these thio-antigens were comparable to the parent antigen in potency for T cell activation, although the maximum level of activation was somewhat lower for both thio-antigens relative to ELA (Figure 3C, Figure S8, Table 4).

In the reported crystal structure of the complex formed by the antigen ELA, the MHC I HLA-A2 and the TCR MEL5, there are contacts (≤ 4 Å) that seem to represent H-bonds between the TCR and the backbone oxygen or nitrogen atoms of the antigen that could be affected by the amide-to-thioamide replacements in ELA-3, ELA-4 and ELA-7 (Figure S2 and Table S2).⁶⁰ ELA-3 and ELA-7 were comparable to the native ELA antigen in potency for stimulating IFN-γ secretion, but ELA-4 was less potent (Table 4). Thus, amide-to-thioamide replacement at a TCR contact site does not necessarily cause a decline in T cell activation potency in this system.

ELA-1,9 and ELA-1,6,9 were evaluated as substrates for proteinase K. The PeptideCutter tool in Expasy predicted seven cleavage sites for ELA (after Glu1, Leu2, Ala3, Ile5, Ile7, Leu8 and Thr9; Figure 4A). Under our assay conditions, the half-life of ELA when treated with proteinase K was 0.4 min (Figure 4B, Figure S9). ELA-1,9 displayed a modest level of resistance, with a half-life of 1.2 min (Figure 4B, Figure S9). Addition of a thioamide near the center of the antigen, to generate ELA-1,6,9, increased the half-life to 12.2 min (~30-fold increase relative to ELA). Thus, as observed in the GIL series, incorporation of three thioamide units dispersed across the antigen can lead to significant protection from proteolysis.

Figure 4.

(A) Prediction of cleavage sites for ELAGIGILTV antigen when treated with proteinase K, as calculated by the PeptideCutter tool in the Expasy resource. Numbers indicate positions of cleavage sites. (B) Proteolytic resistance against proteinase K for ELA, ELA-1,9 and ELA-1,6,9 peptides evaluated in DPBS, pH 7.4; peptide sequences are shown in Fig. 3A. Numbers above bars indicate peptide half-life (min).

Conclusions

We have explored a new strategy for modifying short peptides that serve as MHC I-displayed antigens with the goal of enhancing immunological function. Specifically, we sought to diminish susceptibility to proteolysis while maintaining recognition of the antigen by HLA-A2 and recognition of the peptide+HLA-A2 complex by cognate T cell receptors. Previous studies have shown that these two recognition events are very sensitive to unnatural modifications of peptide antigens.^23–30 We explored amide-to-thioamide replacement, which represents a modest change from the structural perspective but has been shown in other contexts to hinder proteolysis.^33–41 Parallel studies with two well-known HLA-A2-restricted antigens, GILGFVFTL and ELAGIGILTV, revealed that single thioamide substitutions were tolerated at most positions with only moderate loss in HLA-A2 affinity, as indicated by two complementary assays (Table 1, Table 3). In some cases, however, a single thioamide substitution could substantially enhance or substantially diminish HLA-A2 affinity. A larger range of effects was observed for the single thioamide substitutions in terms of T cell activation, which was monitored via IFNγ secretion (Table 2, Table 4).

Based on the T cell activation results of single amide-to-thioamide substitutions, we designed GIL- or ELA-based thio-antigens that contained two or three substitutions. These thio-antigens displayed modestly diminished HLA-A2 affinity relative to the native antigen, but they were quite potent in activating cognate T cells (Figure 1, Figure 3, Tables 1–4). Both of the GIL thio-antigens with three substitutions (GIL-1,4,8 and GIL-1,5,8) were significantly more potent than the native antigen in terms of stimulating IFN-γ secretion from T cells (lower EC₅₀ values), and the ELA thio-antigen with three substitutions (ELA-1,6,9) was slightly more potent in stimulating IFN-γ secretion relative to the native antigen (Table 2, Table 4). In both systems, the maximum level of IFN-γ secretion was moderately lower for the thio-antigens relative to the parent antigen. The T cell activation potency of the triply-substituted analogues was not obviously predictable based on the behavior of the corresponding singly-substituted analogues, which raises the possibility of synergistic effects from multiple amide-to-thioamide replacements.

This study suggests a general experimental approach to discovering analogues derived from other MHC I antigens that can activate cognate T cells and resist proteolytic degradation. A priori prediction of optimal amide-to-thioamide replacement sites does not seem to be possible at this point, but preparation and evaluation of a small set of derivatives of a given antigen, comprising 10–15 analogues, should allow the identification of thio-antigens with promising properties. Synthesis of such thioamide-substituted peptides is straightforward.^{39,41,48,67–69,81–85}

Substantial protection from degradation by proteinase K, a broad-spectrum serine protease, was observed for one of the triply-substituted GIL thio-antigens and for the lone triply-substituted ELA thio-antigen we investigated. The combination of resistance to proteolysis and potent T cell activation observed for two different HLA-A2 antigens suggests that the thio-antigen design strategy may be broadly applicable for generating CD8 T cell-directed vaccines. However, future studies directed toward specific applications of MHC I thio-antigens will require degradation studies conducted in relevant proteolytic environments. Depending on the application, such studies might involve serum (for administration via injection) or simulated gastric fluid and simulated intestinal fluid (for oral administration). Efforts to predict the proteolytic susceptibilities of thioamide-substituted peptides have been reported;⁸⁶ this insightful work suggests that empirical evaluations in complex environments will be important for future thio-antigen studies directed toward specific potential uses.