Abstract
A major hurdle in designing successful epitope-based vaccines resides in the delivery, stability, and immunogenicity of the peptide immunogen. The short-lived nature of unmodified peptide-based vaccines in vivo limits their therapeutic application in the immunotherapy of cancers and chronic viral infections as well as their use in generating prophylactic immunity. The incorporation of β-amino acids into peptides decreases proteolysis, yet its potential application in the rational design of T cell mimotopes is poorly understood. To address this, we have replaced each residue of the SIINFEKL epitope individually with the corresponding β-amino acid and examined the resultant efficacy of these mimotopes. Some analogs displayed similar MHC binding and superior protease stability compared with the native epitope. Importantly, these analogs were able to generate cross-reactive CTLs in vivo that were capable of lysing tumor cells that expressed the unmodified epitope as a surrogate tumor Ag. Structural analysis of peptides in which anchor residues were substituted with β-amino acids revealed the basis for enhanced MHC binding and retention of immunogenicity observed for these analogs and paves the way for future vaccine design using β-amino acids. We conclude that the rational incorporation of β-amino acids into T cell determinants is a powerful alternative to the traditional homologous substitution of randomly chosen naturally occurring α-amino acids, and these mimotopes may prove particularly useful for inclusion in epitope-based vaccines.
The cellular immune system recognizes processed forms of Ags presented on the surface of APCs in complex with MHC molecules. Because recognition of infectious or abnormal states is dictated by recognition of processed forms of Ags, synthetic peptide-based vaccines offer an attractive alternative to DNA, whole Ag, and attenuated vaccine approaches for treatment of cancer, chronic viral infections, and the generation of prophylactic immunity. The use of peptides offers several advantages over traditional vaccine approaches (reviewed recently by Purcell et al. (1)). However, the use of peptide-based vaccines can be problematic due to relatively poor stability, bioavailability, and rapid modification of the peptide under conditions of formulation and delivery.
Previous attempts to overcome the bioavailability issues of peptide Ags have generally met with limited success. Attempts to use retro-inverso or d-amino acid-containing peptide analogs have largely failed because of the reorientation of the Cα-Cβ bond that is considered essential for optimal MHC binding and T cell recognition (2). Other approaches have sought to enhance peptide stability by inducing chemical modification of the C or N termini or through backbone modifications. For example, Ostankovitch et al. (3) demonstrated that replacement of two successive amino acid residues with a substituted melonate derivative and gem-diaminoalkyl residue of the influenza virus matrix protein peptide (M58–66) maintained T cell recognition and MHC binding. These partially modified retro-inverso pseudopeptides resulted in a 2-fold increase in the serum half-life, while maintaining comparative reactivity with three wild-type-specific clones. Additionally, influenza-infected cells were lysed by CD8+ CTLs raised by the retro-inverso analog. Other groups have reported modifying the N- and C-terminal ends of a peptide to prevent its degradation by exopeptidases (e.g., N-methylation and C-amidation of SIINFEKL increases serum stability, but with detrimental effects on MHC binding) (4). An alternative approach is the use of β-amino acids, which maintains the native Cα-Cβ orientation but incorporates a methyl insertion extending the backbone of the molecule (Fig. 1). The use of β-amino acids as a peptidomimetic approach has emerged in recent years with applications including protease inhibitors and receptor agonists/antagonists for a large variety of bioactive peptides, such as adrenocorticotropin, angiotensin II, gastrin, oxytocin, and more recently for neuropeptides and platelet aggregation factors (5). β-Amino acids have also been used in the design of DNA-binding peptides of defined secondary structure (5), and previous studies have indicated that the CTL epitopes that contain β-amino acids retain MHC-binding capacity and can be cross-recognized by wild-type-specific T cell lines (6, 7), with concomitant enhancement of the serum half-life of these analogs (7). Unlike conservative substitutions of naturally occurring α-amino acids, insertion of a homologous β-amino acid allows for the constitutive side chains to be maintained. Furthermore, the peptide bonds formed when incorporating β-amino acids into a peptide are known to be highly resistant to proteolysis (8). This property reduces the rapid clearance of peptides from animal sera, which has previously limited the effectiveness of natural peptide therapeutics (1, 6, 7, 9).
In this study, we report the systematic replacement of each amino acid in the H-2Kb-restricted SIINFEKL epitope from chicken OVA with its corresponding β-amino acid and examine the effects on MHC binding, T cell recognition, and protease resistance and for selected analogs, the generation of high-affinity, cross-reactive CTLs in C57BL/6 (B6) mice capable of recognizing endogenously presented Ags on tumor cells. We also report the structures of two β-amino acid-containing peptides bound to the murine MHC class I molecule H-2Kb that reveal the molecular basis for their enhanced MHC binding and T cell cross-reactivity. We demonstrate that insertion of β-amino acids into T cell determinants may be a viable strategy for generating proteolytically stable peptide-based vaccines.
Materials and Methods
Cell lines
RMA-S cells (H-2b) (10, 11) were cultured in DMEM (Invitrogen Life Technologies) containing 10% FBS, l-glutamine, and the antibiotics streptomycin and penicillin (here referred to as DMEM-10). The SIINFEKL/Kb-restricted CD8+ T cell hybridomas B3.1 and GA4.2 (12) and the herpes simplex viral-specific T-hybrid HSV2.3 (13) were derived and maintained as described previously. All T cell hybrids were grown in DMEM-10 and kept under selection with 0.5 mg/ml G418. I-3 fibroblast cells (14), which express MHC class I H-2Kb, were grown in DMEM-10. The IL-2-dependent cell line CTLL-2 was used in an IL-2 bioassay to determine the relative amount of IL-2 production by CTL clones as stimulated by peptide analogs, as described previously (14). Briefly, cells were maintained in DMEM-10 supplemented with 100 U/ml rIL-2. Cell lines cultured from immunized mice were restimulated with irradiated EG7 (the thymoma cell line EL4 transfected with chicken OVA) cells and cultured 12 days further in RPMI 1640 containing 10% FBS, l-glutamine, streptomycin, and penicillin (here referred to as RPMI-10) and supplemented with 10 U/ml rIL-2 (Cetus).
Peptides
Peptides were synthesized using solid-phase peptide synthesis and the standard F-moc (N-(9-fluorenyl) methoxycarbonyl) protection strategy. Protected β-amino acids were synthesized by Arndt-Eistert homoligation of commercially available N-F-moc-protected α-amino acids as described previously (15). After synthesis and selective deprotection, complete deprotection and cleavage from the resin was achieved using trifluoroacetic acid (TFA)3. Purification of peptides was achieved by reversed-phase HPLC (RP-HPLC), and purified peptides were characterized by mass spectrometry (MS) using an Agilent Technologies 1100 MSD SL ion trap mass spectrometer. This combination of RP-HPLC purification and, in particular, ion trap MS-based experiments allowed unambiguous structural confirmation of each β-analog.
H-2Kb binding assay
RMA-S cells grown to a density of ∼106 cells/ml at 37°C were maintained at 26°C for 20 h with 5% CO2. Cells were washed and resuspended to give a final concentration of 105 cells/well (200 μl) in a flat-bottom, 96-well plate. Cells were pulsed with graded concentrations of peptide (10 μl), incubated for 1 h at 26°C, and incubated for an additional 2 h at 37°C. After washes, cell surface expression of H-2Kb was detected using the mAb Y-3 (16), which stains properly conformed Kb complexes on the cell surface. After additional washes, Y-3-bound Kb-peptide complexes were detected with FITC-conjugated sheep anti-mouse Ig (Amrad) during flow cytometry performed on a FACScan (BD Biosciences).
T cell recognition and IL-2 bioassay
I-3 cells (H-2Kb-transfected murine fibroblasts) (14) were grown to ∼70% confluence, plated out into a flat-bottom, 96-well plate at a density of 5 × 104 cells/well (90 μl), and incubated at 37°C for 20 h. Cells were then pulsed with graded concentrations of peptide (10 μl) for 1 h at 37°C. T-hybrid clones grown to ∼106 cells/ml were washed and resuspended at a density of 106 cells/ml, and 100 μl were transferred to wells containing the pulsed and washed I-3 cells. T-hybrid clones were then incubated in coculture for 20–24 h. Fifty microliters of coculture supernatant were harvested from each well and added to the CTLL-2 IL-2 bioassay. CTLL-2 cells were maintained in DMEM-10 plus 100 U/ml IL-2, grown to ∼5 × 105 cells/ml, and washed three times in medium containing no exogenous IL-2. A 96-well, flat-bottom plate was seeded at a density of 5000 cells/ well to a final volume of 150 μl. Harvested coculture supernatant from T-hybrid experiments (50 μl) was then added to plated CTLL-2 cells and incubated at 37°C for 18–22 h (depending on visual inspection of control wells with no added IL-2 for CTLL-2 cell death). Cells were pulsed with [3H]thymidine at 1 μCi/well for 6 h. Cells were then harvested onto a glass fiber filter, MicroScint scintillation fluid (Packard Biosciences) was added, and the cells were counted on a TopCount scintillation counter (Packard Biosciences).
Serum stability assay
This protocol is a modification of the method described by Hoffmann et al. (17). Ten microliters of an aqueous peptide solution containing 1 mg/ml peptide were added to 100 μl of 20% mouse serum (pooled from adult male B6 mice) and incubated at 37°C for up to 2 h. At each time point, samples were taken, and serum proteins were removed by precipitation through the addition of 40 μl of 15% TCA. Samples were stored at 4°C for 30 min and centrifuged. Supernatants were then removed and kept on ice. Fifty microliters of each supernatant were analyzed by RP-HPLC on a SMART system liquid chromatograph (Amersham Biosciences) using 0.1% TFA in water (eluent A) and 0.09% TFA in 60% aqueous acetonitrile (eluent B) and the following gradient: 0% B for 5 min, linear gradient to 60% B for 35 min. The column was a Amersham Biosciences μRPC octadecyl silica column (2.1 inner diameter × 100 mm) of 3 μM nominal particle size and 100 Å pore size. The flow rate was 200 μl/min, and UV detection was used at 214, 254, and 280 nm. Values are given as area of peptide peak (214 nm). Peptide amounts at t = 0 min were taken as 100%. Peptide digests were further fractionated by RP-HPLC, and electrospray ionization ion trap MS was used to characterize the degradation pattern of SIINFEKL and β-amino acid analogs as described below for enzyme-mediated proteolysis.
Enzyme proteolysis assay
Ten microliters of aqueous peptide solutions containing 1 mg/ml peptide were added to 10 μl of fresh enzyme solution at an enzyme:peptide ratio of 1:100 and incubated at 25°C. Pronase (Sigma-Aldrich) and proteinase K (PK; Sigma-Aldrich) reaction buffers contained 20 mM Tris (pH 8.0) and 5 mM CaCl2, whereas pepsin (Sigma-Aldrich) reaction buffer contained 10 mM HCl (pH 2.5). Enzyme reactions were performed for 30 min, 2 h, and 6 h. Reaction mixtures were centrifuged for 20 min in a microfuge at 13,000 rpm through a 10-kDa cut-off filter to halt peptide degradation. Peptides were separated on an Agilent 1100 capillary high-performance liquid chromatograph using a Zorbax SB C18 (0.5 inner diameter × 150 mm) RP-HPLC column and a linear gradient of buffer A (0.1% TFA) to 60% B (acetonitrile and 0.09% TFA; 2% per minute), at a flow rate of 20 μl/min. Eluted peptides were monitored at 214 nm and further characterized by an online Agilent 1100 MSD SL ion trap mass spectrometer. Tandem MS experiments were used to characterize the proteolytic fragments of SIINFEKL and selected β-amino acid analogs. Sequence analysis was conducted manually following the theoretical fragmentation patterns of SIINFEKL-related peptides as predicted by the MS-product algorithm of the Protein Prospector package (18).
Generation of TCD8+ cells
Ten-week-old male B6 mice were given s.c. injections at the base of the tail with 50 μg of peptide Ag plus 150 μg of HBVcore128–140 Thelper epitope emulsified in 200 μl of IFA. Mice were immunized in triplicate. After 21 days, 3 × 107 splenocytes were cultured in vitro for 13 days in RPMI-10 supplemented with 10 U/ml rIL-2 with 3 × 105 irradiated EG7 cells or peptide-pulsed irradiated syngeneic splenocytes (30,000 and 2,200 rad, respectively). Viable cells were harvested from Ficoll-Hypaque gradients and enriched for TCD8+ cells by depleting B220+ cells using mAb-coated MACS beads as reported previously (19). TCD8+ cell activities were measured via intracellular cytokine staining (ICS) and 51Cr release cytotoxicity assays.
Intracellular cytokine staining
T cell activation was assessed by ICS as described previously (20). In brief, 1 × 106 spleen cells were stimulated with EG7, peptide-pulsed EL4 cells or control Ags in RPMI-10 supplemented with 10 U/ml rIL-2. Cells were stimulated for 6 h, and brefeldin A was added to a final concentration of 10 μg/ml for the final 4 h of this incubation. The cells were then washed twice with PBS containing 2% FBS and 0.09% sodium azide. The cell phenotype was determined by cell surface staining using PE-conjugated anti-CD8 (BD Biosciences) mAb at 4°C for 30 min. Cells were fixed with 1% paraformadehyde (ProSciTech) and permeabilized with 0.3% Saponin (Sigma-Aldrich), and intracellular IFN-γ was detected with an anti-IFN-γ mouse mAb (BD Biosciences). The percentage of CD8+ T cells producing IFN-γ was determined by flow cytometry using FlowJo software (Tree Star). A total of 300,000 events from each lymphocyte gate was collected for analysis of native and analog-immunized mice.
Cytolytic activity assay
Cytolytic activity was detected by a standard 4-h 51Cr release assay (14). EG7 target cells were labeled with 100 μCi of sodium chromate (51Cr, 100 mCi/ml; DuPont-NEN Research Products), incubated for 1 h with 1 μg/ml peptide, washed twice with DMEM containing 5% FBS, and aliquoted at 5000 cells/well. Target cells were cultured with effector cells at ratios of 1:1, 1:3, 1:10, and 1:30 for 4 h. Cell lysis was measured as the release of intracellular 51Cr by subjecting culture supernatants to solid scintillation plate counting. Spontaneous release never exceeded 10% of the maximum 51Cr uptake. The percentage of specific lysis was determined as follows: 100 × (experimental release − spontaneous release)/(maximum release − spontaneous release).
Expression, purification, crystallization, and structure determination
All crystallization trials were conducted using the hanging drop vapor diffusion technique (21, 22). The crystals were grown under identical conditions for each SIINFEKL analog, by mixing equal volumes of 10 mg/ml Kb/peptide complexes with the reservoir buffer (0.1 M sodium cacodylate, 0.2 M calcium acetate, and 14% (w/v) PEG-8000, pH 6.5), and microseeded from crystals grown in 16% (w/v) polyethylene glycolate 8000 at room temperature. Crystals were frozen after a stepwise transfer from 5 to 10% of the cryoprotectant glycerol with 5 min per condition. The crystals belong to space group P21 with isomorphous unit cell dimensions. A 2.2 and 1.9 Å data set for Kb-β-Phe5 and Kb-β-Leu8 was collected, respectively, and scaled using the HKL suite of programs. The Kb-β analog structures were refined using our recently determined Kb/HSV gB complex as a starting model (23). The progress of refinement was monitored by the Rfree value (3% of the data) with neither a σ nor a low-resolution cut-off being applied to the data. The structure was refined using rigid-body fitting of the individual domains, followed by the simulated-annealing protocol implemented in the Crystallography and NMR System (version 1.0) (24), interspersed with rounds of model building using the program O (25). Tightly restrained individual B-factor refinement was used, and bulk solvent corrections were applied to the data set. Water molecules were included in the model if they were within hydrogen-bonding distance to chemically reasonable groups, appeared in Fo-Fc maps contoured at 3.5σ, and had a B-factor of <60 Å2. See Table II for summary of refinement statistics and model quality. There are two molecules in the asymmetric unit. The electron density for the bound peptides was very clear in both monomers. The root mean square deviation (rmsd) between the Ag binding clefts of monomers in the asymmetric units of the Kb analog complexes is ≈0.2 Å over all Cα atoms, and the structural analysis was confined to one complex in each case.
. | β-Leu . | β-Phe . |
---|---|---|
Data collection statistics | ||
Temperature | 100 K | 100 K |
X-ray source | RU-3HBR | RU-3HBR |
Detector | R-AXIS IV++ | R-AXIS IV++ |
Space group | P21 | P21 |
Cell dimensions (Å) (a, b, c, β) | 66.39, 90.09, 89.22, 111.44° | 66.42, 90.90, 89.44, 111.39° |
Resolution (Å) | 1.9 | 2.2 |
Total no. of observations | 212021 | 114312 |
No. of unique observations | 76528 | 47666 |
Multiplicity | 2.8 | 2.4 |
Data completeness (%) | 99.2 (99.6) | 94.5 (91.4) |
Percentage data >2σ1 | 71.3 (38.4) | 73.9 (58.9) |
I/σI | 13.7 (2.2) | 13.4 (3.9) |
Rmergeb (%) | 7.1 (53.7) | 8.1 (35.9) |
Refinement statistics | ||
Nonhydrogen atoms | ||
Protein | 6284 | 6298 |
Water | 827 | 689 |
Resolution (Å) | 50–1.9 | 50–2.2 |
Rfactorc (%) | 22.9 | 24.1 |
Rfreed (%) | 25.9 | 28.6 |
Rms deviations from ideality | ||
Bond lengths (Å) | 0.005 | 0.006 |
Bond angles (°) | 1.29 | 1.27 |
Impropers (°) | 0.77 | 0.79 |
Dihedrals (°) | 25.2 | 24.9 |
Ramachandran plot | ||
Most favored | 93.2 | 91.5 |
And allowed region (%) | 6.2 | 7.4 |
B-factors (Å2) | ||
Average main chain | 25.49 | 28.47 |
Average side chain | 27.99 | 30.48 |
Average water molecule | 38.63 | 35.86 |
rmsd bonded Bs | 1.60 | 1.61 |
. | β-Leu . | β-Phe . |
---|---|---|
Data collection statistics | ||
Temperature | 100 K | 100 K |
X-ray source | RU-3HBR | RU-3HBR |
Detector | R-AXIS IV++ | R-AXIS IV++ |
Space group | P21 | P21 |
Cell dimensions (Å) (a, b, c, β) | 66.39, 90.09, 89.22, 111.44° | 66.42, 90.90, 89.44, 111.39° |
Resolution (Å) | 1.9 | 2.2 |
Total no. of observations | 212021 | 114312 |
No. of unique observations | 76528 | 47666 |
Multiplicity | 2.8 | 2.4 |
Data completeness (%) | 99.2 (99.6) | 94.5 (91.4) |
Percentage data >2σ1 | 71.3 (38.4) | 73.9 (58.9) |
I/σI | 13.7 (2.2) | 13.4 (3.9) |
Rmergeb (%) | 7.1 (53.7) | 8.1 (35.9) |
Refinement statistics | ||
Nonhydrogen atoms | ||
Protein | 6284 | 6298 |
Water | 827 | 689 |
Resolution (Å) | 50–1.9 | 50–2.2 |
Rfactorc (%) | 22.9 | 24.1 |
Rfreed (%) | 25.9 | 28.6 |
Rms deviations from ideality | ||
Bond lengths (Å) | 0.005 | 0.006 |
Bond angles (°) | 1.29 | 1.27 |
Impropers (°) | 0.77 | 0.79 |
Dihedrals (°) | 25.2 | 24.9 |
Ramachandran plot | ||
Most favored | 93.2 | 91.5 |
And allowed region (%) | 6.2 | 7.4 |
B-factors (Å2) | ||
Average main chain | 25.49 | 28.47 |
Average side chain | 27.99 | 30.48 |
Average water molecule | 38.63 | 35.86 |
rmsd bonded Bs | 1.60 | 1.61 |
The values in parentheses are for the highest resolution bin (approximate interval, 0.1 Å).
Rmerge = ∑ |Ihkl − 〈Ihkl〉| / ∑Ihkl.
Rfactor = ∑hkl | |Fo | − |Fc| | / ∑hkl|Fo| for all data, except for 3% that was used for the Rfreed calculation.
Results
Replacement of β-amino acids within the SIINFEKL determinant alters class I binding
To characterize the effect of main-chain methyl group insertions on H-2Kb binding, analogs of SIINFEKL containing single β-amino acid replacements were synthesized, effectively creating a β-amino acid scan of the epitope (Table I). In this instance, we chose to examine βC3-amino acids, in which the side chain of each residue is located on the Cβ atom. Each analog was tested for binding to H-2Kb in an RMA-S epitope stabilization assay (26). The analogs displayed a spectrum of MHC binding characteristics relative to the wild-type peptide (Fig. 2, A and B, dotted line). β-Amino acid replacements of corresponding α-amino acid residues at positions 1–3 and 7 resulted in analogs that bound poorly to H-2Kb relative to the wild-type peptide (Fig. 2,A), whereas replacement at positions 4–6 and 8 resulted in binding to H-2Kb molecules equivalent or superior to the wild-type peptide (Fig. 2 B).
Sequence . | . | . | . | . | . | . | . | . | . | . | . | . |
---|---|---|---|---|---|---|---|---|---|---|---|---|
Peptide . | Name . | P1b . | P2 . | P3 . | P4 . | P5 . | P6 . | P7 . | P8 . | M.W. . | Purity (%) . | |
1 | W.T. | Ser | Ile | Ile | Asn | Phe | Glu | Lys | Leu | 962.8 | 98 | |
2 | β-Ser1 | β-Ser | Ile | Ile | Asn | Phe | Glu | Lys | Leu | 976.8 | 98 | |
3 | β-Ile2 | Ser | β-Ile | Ile | Asn | Phe | Glu | Lys | Leu | 976.8 | >80 | |
4 | β-Ile3 | Ser | Ile | β-Ile | Asn | Phe | Glu | Lys | Leu | 976.8 | 98 | |
5 | β-Asn4 | Ser | Ile | Ile | β-Asn | Phe | Glu | Lys | Leu | 976.8 | 86 | |
6 | β-Phe5 | Ser | Ile | Ile | Asn | β-Phe | Glu | Lys | Leu | 976.8 | 98 | |
7 | β-Glu6 | Ser | Ile | Ile | Asn | Phe | β-Glu | Lys | Leu | 976.8 | 96 | |
8 | β-Lys7 | Ser | Ile | Ile | Asn | Phe | Glu | β-Lys | Leu | 976.8 | 85 | |
9 | β-Leu8 | Ser | Ile | Ile | Asn | Phe | Glu | Lys | β-Leu | 976.8 | 96 |
Sequence . | . | . | . | . | . | . | . | . | . | . | . | . |
---|---|---|---|---|---|---|---|---|---|---|---|---|
Peptide . | Name . | P1b . | P2 . | P3 . | P4 . | P5 . | P6 . | P7 . | P8 . | M.W. . | Purity (%) . | |
1 | W.T. | Ser | Ile | Ile | Asn | Phe | Glu | Lys | Leu | 962.8 | 98 | |
2 | β-Ser1 | β-Ser | Ile | Ile | Asn | Phe | Glu | Lys | Leu | 976.8 | 98 | |
3 | β-Ile2 | Ser | β-Ile | Ile | Asn | Phe | Glu | Lys | Leu | 976.8 | >80 | |
4 | β-Ile3 | Ser | Ile | β-Ile | Asn | Phe | Glu | Lys | Leu | 976.8 | 98 | |
5 | β-Asn4 | Ser | Ile | Ile | β-Asn | Phe | Glu | Lys | Leu | 976.8 | 86 | |
6 | β-Phe5 | Ser | Ile | Ile | Asn | β-Phe | Glu | Lys | Leu | 976.8 | 98 | |
7 | β-Glu6 | Ser | Ile | Ile | Asn | Phe | β-Glu | Lys | Leu | 976.8 | 96 | |
8 | β-Lys7 | Ser | Ile | Ile | Asn | Phe | Glu | β-Lys | Leu | 976.8 | 85 | |
9 | β-Leu8 | Ser | Ile | Ile | Asn | Phe | Glu | Lys | β-Leu | 976.8 | 96 |
Bold text highlights β-amino acid substitution.
P, Position; W.T., wild type.
Recognition of SIINFEKL analogs by T-hybrids specific for the SIINFEKL-H-2Kb complex
The two unrelated SIINFEKL-specific, Kb-restricted T cell hybridomas GA4.2 and B3.1 have distinct fine specificities (27) and were used to compare the recognition of β-analogs as determined by relative IL-2 secretion (Fig. 2, C and D). The difference in TCR footprint between the two clones can be seen with the β-amino acid scan. All β-analogs were recognized by both hybridomas but generally less efficiently than the wild-type epitope with the following exceptions: β-amino acid replacement at positions 2 and 5 caused no loss of recognition by GA4.2 (Fig. 2,C), whereas replacement at position 6 with B3.1 also had no loss in recognition (Fig. 2 D), reflecting the known insensitivity of the B3.1 clone toward P6 (26). Interestingly, replacement at position 1 resulted in a substantial loss of recognition by both clones. Moreover, the effects of β-amino acid incorporation on T cell recognition had little correlation with their respective H-2Kb binding affinities. For instance, analog 2 was the weakest binder of the series but retained recognition by GA4.2 equivalent to wild-type SIINFEKL. In contrast, analog 4, which showed markedly reduced recognition by B3.1, bound to H-2Kb more efficiently than the wild-type peptide. These results suggest that the recognition of β-amino acid analogs by these T cell clones is mainly dependent on the position of the methyl insertion after β-amino acid replacement and not binding affinity.
Enhanced serum stability of SIINFEKL analogs
The influence of β-amino acid incorporation on the serum stability of the natural peptide was tested by incubating the analogs for 2 h in mouse serum at room temperature (Fig. 3,A). Degradation patterns were monitored using liquid chromatography-tandem MS. The wild-type epitope was rapidly degraded, with only 5% of the intact peptide recoverable after 2 h. The rate of degradation of the β-amino acid analogs was highly dependent on the position of the methyl insertion. For example, little stabilization was observed for analogs with insertion at positions 1, 3, and 4, whereas significant resistance to proteolytic cleavage was generated in analogs replaced at positions 2, 5, and 8 (Fig. 3,A). Furthermore, although the wild-type peptide was degraded into single amino acids and small di- and tri-peptides, incorporation of a β-amino acid in some analogs resulted in protection of cleavage for a large proportion of the peptide. For example, β-Phe5 demonstrated prolonged (>24 h) protection of the entire C-terminal end of the peptide (Fig. 3 B and data not shown).
To further explore the cooperative effects of a single β-amino acid incorporation, the stability of the wild-type peptide was compared with that of the β-Phe5 analog when challenged with the proteases pronase, PK, and pepsin. Fig. 3 C demonstrates the cleavage of both peptides by pronase, an enzyme mixture known to contain a mix of endo- and exo-proteinases and reported to cleave almost any peptide bond (28). It resulted in rapid cleavage of the wild-type peptide after only 2 h, with visible peaks from cleavages at peptide bonds between Asn4-Phe5, Phe5-Glu6, Glu6-Lys7, and Lys7-Leu8. The β-amino acid replacement at position 5 resulted in the protection of bonds Asn4-β-Phe5, β-Phe5-Glu6, and Glu6-Lys7 and partial protection of the Lys7-Leu8 bond. Interestingly, even after 6 h, >50% of the entire β-Phe5 analog remained compared with <15% for SIINFEKL (data not shown).
PK and pepsin readily cleaved a single peptide bond between Phe5 and Glu6 of wild-type SIINFEKL (Fig. 3, D and E). PK is an endolytic protease that cleaves peptide bonds at the carboxylic sides of aliphatic, aromatic, or hydrophobic amino acids, whereas pepsin has specificity toward aromatic residues. The incorporation of a β-amino acid at position Phe5 resulted in the complete abrogation of enzymatic attack from PK and pepsin (Fig. 3, D and E). Efficient generation of CTL responses by oral administration of peptides may be enhanced by the incorporation of pepsin resistance into short CTL epitopes. These results clearly confirm the protective effects of β-amino acid incorporation from proteolysis. Thus, consistent with results in sera, these data suggest that a single replacement of an α-amino acid by a β-amino acid can confer protection not only directly to the site of modification but also to distal peptide bonds, in both complex matrices such as serum or when exposed to specific proteases.
Structure of H-2Kb complexed to SIINFEKL analogs containing β-amino acids at anchor positions
To evaluate the structural consequences of incorporating β-amino acids into the SIINFEKL peptide, two analog-H-2Kb complex structures were compared with the previously reported structure of the wild-type complex (29). We selected β-amino acid replacements at anchor residue positions 5 and 8 because these performed well in our functional analysis and proved most stable against proteases. The structures have been solved to 2.2 Å for β-Phe5-containing and 1.9 Å for β-Leu8-containing analogs (see Table II for a summary of statistics and model quality and Fig. 4). The structures of the two β-amino acid-containing SIINFEKL analogs complexed to H-2Kb were similar to the previously reported structure of the wild-type complex (29). In both structures, the N- and C-terminal H-bonding networks were conserved, and the only perturbations in the structures were restricted to the locality of the inserted methylene group in the β-amino acid at position 5 or 8, respectively. Thus, detailed structural analyses will be confined to these sites.
Effect of incorporation of the β-Phe5 analog
The central binding specificity of Kb-restricted ligands is dictated by interactions between the position 5 amino acid residue of the peptide with the C pocket of the Kb Ag binding cleft. Pool sequencing of naturally presented peptides reveals position 5 is almost always occupied by either a phenylalanine or a tyrosine residue (30). In the H-2Kb-SIINFEKL complex (29), the Phe5 side chain sits within the hydrophobic C pocket making van der Waals’ (VDW) contacts with Asn70, Ser73, Phe74, Val97, Gln114, and Tyr116. Phe5 also makes contact with the Ile2 residue of the peptide. The amide group forms a hydrogen bond to Asn70Oδ1, whereas the carbonyl does not form contact with the main chain. It is clear that the replacement with the β-amino acid has led to the disruption of the peptide main-chain conformation while largely maintaining the VDW contacts within this pocket. Consequently, the main-chain carbonyl of the analog now forms a new H-bond to Ser73, and the Phe5 group sits deeper in the hydrophobic Ag binding pocket, but in approximately the same orientation. This in part explains the enhanced binding of the β-Phe5 analog to H-2Kb (Fig. 2 B).
After superposition of the β-Phe5 complex with the wild-type complex, the Ag-binding cleft demonstrates a rmsd of 0.61 Å for residues 1–180. The most significant changes with respect to the analog are largely restricted to the region of the peptide that accommodates the methyl group insertion (Figs. 4,C and 5 A). Namely, there is a shift in the backbone of Gln72 and Ser73 (0.79 and 0.86 Å, respectively) of the α1 helix of the Kb H chain. In the analog structure, this region is closer to the peptide, presumably reflecting the additional H-bond. After superposition of peptide alone, the rmsd is 0.48 Å, with the largest difference at the site of incorporation (Phe5, 0.8 Å; Glu6, 0.6 Å).
Effect of incorporation of the β-Leu8 analog
After superposition of the wild-type and β-Leu8/H-2Kb complexes, the rmsd between the wild-type and β-Leu8 analog structure is 0.55 Å for residues 1–180 (Figs. 4,D and 5 B), whereas the rmsd for the peptide alone is 0.41 Å. The largest observable difference between the overlaid structures of the peptides is at the site of the main-chain methyl insertion at Leu8 (0.92 Å). In the wild-type structure, the C-terminal side chain of Leu8 sits within the hydrophobic F-pocket of H-2Kb, where the terminal carboxylate is tethered directly by H-bonds to Lys146Nζ, Thr143Oγ1, and Tyr84Oη and by water-mediated H-bonds to Asp77Oδ1 and Thr80Oγ1. Compared with the analog structure, the H-bonds of the peptide backbone to H-2Kb are unchanged. However, the incorporation of the extra methylene group in the terminal leucine residue has altered the position of the side chain. In the β-Leu8 structure, the side chain sits deeper within the pocket such that it forms additional VDW interactions with Thr80, Ile95, and Tyr116. Additionally, a small change was observed in the orientation of the side chains of the cleft residues Phe74 and Tyr116 due to the β-Leu8 sitting more deeply into the F-pocket (the side chains shift to accommodate a more favorable interaction with β-Leu8). However, unlike the β-Phe5 analog structure, there was no significant shift in the Cα backbone of the neighboring α-helices between the structures.
T cell responses after immunization with selected peptide analogs
The reduced ability of the β-amino acid-containing analogs to be recognized by specific T cell clones derived against the native peptide does not necessarily preclude the possibility that they will not be at least as immunogenic as the wild-type peptide in vivo. To investigate whether the β-analogs were able to elicit high-avidity, cross-reactive TCD8+ cells, B6 mice were immunized with the modified β-Phe5 analog, and responses were analyzed for cross-reactivity with the wild-type peptide. Mice were immunized with either the wild-type or β-Phe5 peptide along with the HBVcore Thelper epitope emulsified in IFA and maintained for 21 days. Harvested spleen cells were stimulated ex vivo with either irradiated EG7 cells (OVA-expressing thymoma cell line) or lymphocytes pulsed with 1 μM of either peptide and cultured for 13 days (Fig. 6). Expanded TCD8+ cells from mice immunized with the β-Phe5 analog showed a high level of cross-reactivity and elicited a response of similar magnitude to that of mice immunized with the wild-type peptide (Fig. 6,A). TCD8+ cells elicited and expanded by the β-Phe5 analog also repeatedly displayed superior killing of 51Cr-labeled EG7 cells (a murine thymoma that expresses physiological levels of the parent Ag chicken OVA; Fig. 6,B), suggesting they had enhanced lytic capacity. TCD8+ cells generated by immunization and expansion with the wild-type or analog peptide demonstrated similar low levels of nonspecific lysis of the target cells (Fig. 6,B). Moreover, TCD8+ cells demonstrated equivalent dose responsiveness to SIINFEKL, even when titrated to very low peptide concentrations (Fig. 6 C). TCD8+ cells elicited by the β-Phe5 analog also showed responses of a similar magnitude when targets pulsed with the β-Phe5 analog itself were used in ICS or Cr release assays (data not shown). Thus, these data suggest that the β-Phe5 mimotope is able to expand wild-type-specific TCD8+ cells that exhibit superior lytic activity, yet display similar responsiveness toward the Ag. This difference in cytotoxic potential of the T cells may have applications for in vitro culture of autologous CTL for tumor/viral therapy, in which heterologous stimulation of wild-type-specific T cells by β-analogs may optimally expand Ag-specific CTLs.
Discussion
The rational replacement of β-amino acids for their naturally occurring counterparts may represent a powerful method for generating proteolytically stable peptidomimetics for inclusion in epitope-based vaccines. We demonstrate that incorporation of β-amino acids at single positions in the model peptide Ag SIINFEKL enhances the stability of the entire peptide, retains immunogenicity, and generates functional cross-reactive effector T cells. Our studies suggest that replacement of MHC anchor residues with corresponding β-amino acids may represent a generic strategy for the augmentation of proteolytically susceptible peptide Ags.
In circumstances in which excising epitopes from native Ags results in unstable and poorly bioavailable molecules or in which the immune system fails to mount an effective immune response to the natural Ag, there is an opportunity to further optimize these epitopes to enhance their ability to stimulate an appropriate immune response. Epitope modifications have typically consisted of MHC anchor substitutions, in which suboptimal anchor residues are substituted to provide higher binding affinities. Although these approaches can produce better vaccine components, they have until recently been constrained to naturally occurring amino acid substitutions (recently reviewed by Sette and Fikes (31)). Our work has focused on the introduction of non-natural amino acid analogs into peptide epitopes (32), an approach that not only has improved MHC class I binding and TCR avidity but also introduces favorable biophysical properties to the epitope such as protease resistance and oxidative stability. Stability against proteolysis may play a key role in the delivery of peptide Ags to the immune system as therapeutics, because it has been observed in other systems that processing of an epitope can result in skewing of immune responses toward cryptic determinants with little relevance to protective immunity (33). Such processing events could potentially be circumvented by the incorporation of protease-resistant β-amino acids.
Several studies have explored alternative modifications that not only provide subtle conformational changes to the peptide/MHC structure but also confer resistance against proteases. Results from our work and others (6, 7) have shown that incorporating β-amino acids into epitopes can increase the binding affinity of the mimetic for the MHC molecule relative to the wild-type peptide, although this was highly dependent on the position of the modification. Insertion of a methyl group into some positions of the SIINFEKL peptide resulted in significant loss of MHC binding, indicating a significant shift in peptide conformation and the orientation of anchor residues. However, by actually targeting MHC anchor residues, we have been able to maintain strong MHC binding capacity and retain immunogenicity by minimally perturbing the bound conformation of the peptide ligand. The structure of the strongest binding analog, β-Phe5, clearly shows that the replacement has led to the disruption of the main-chain conformation while largely maintaining the VDW contacts. The consequence of this is that the main-chain carbonyl of the analog now forms a new H-bond to Ser73 and the β-Phe5 side chain sits more deeply in the Kb cleft. Incorporation of the β-Phe residue also results in a shift in the backbone of Gln72 and Ser73 (0.79 and 0.86 Å, respectively) of the α1 helix of the Kb H chain. In contrast, there was no significant shift in the Cα backbone of the neighboring α-helices in the β-Leu8 analog structure, whereas the H-bonding network was fully maintained. The incorporation of the extra methylene group in the terminal leucine residue of the analog altered the position of the side chain, such that it sits deeper within the F-pocket, allowing it to form additional VDW interactions with Thr80, Ile95, and Tyr116. A small change was also observed in the orientation of the side chains of the cleft residues Phe74 and Tyr116 due to the β-Leu8 sitting more deeply into the F-pocket. A summary of the differences in peptide contacts between both analogs and the wild-type complex is given in Table III.
Amino Acid . | Kb Residues . | Type of Interaction . |
---|---|---|
β-Leu8 | Asp77, Thr80, Leu81, Ile95, Tyr116, Tyr123, Thr143, Trp147 | VDW |
β-Leu8N | Asp77-0δ1 | H-bond |
β-Leu80 | Lys146-Nζ | H-bond |
Asp77-0δ1, Thr80-0γ1 | Water-mediated H-bond | |
β-Leu8OXT | Lys146-Nζ | H-bond |
Thr143-0γ1 | H-bond | |
Tyr84-0η | H-bond | |
Leu8 | Leu81, Tyr123, Thr143, Trp147 | VDW |
Leu8N | Asp77-0δ1 | H-bond |
Leu80 | Lys146-Nζ | H-bond |
Asp77-0δ1, Thr80-0γ1 | Water-mediated H-bond | |
Leu8OXT | Lys146-Nζ | H-bond |
Thr143-0γ1 | H-bond | |
Tyr84-0η | H-bond | |
β-Phe5 | Ser73, Phe74, Asn70, Tyr116 | VDW |
β-Phe5N | Asn70-0δ1 | H-bond |
β-Phe50 | Ser73-0γ | H-bond |
Phe5 | Ser73, Phe74, Asn70, Tyr116, Val97, Gln114 | VDW |
Phe5N | Asn70-0δ1 | H-bond |
Phe50 | – |
Amino Acid . | Kb Residues . | Type of Interaction . |
---|---|---|
β-Leu8 | Asp77, Thr80, Leu81, Ile95, Tyr116, Tyr123, Thr143, Trp147 | VDW |
β-Leu8N | Asp77-0δ1 | H-bond |
β-Leu80 | Lys146-Nζ | H-bond |
Asp77-0δ1, Thr80-0γ1 | Water-mediated H-bond | |
β-Leu8OXT | Lys146-Nζ | H-bond |
Thr143-0γ1 | H-bond | |
Tyr84-0η | H-bond | |
Leu8 | Leu81, Tyr123, Thr143, Trp147 | VDW |
Leu8N | Asp77-0δ1 | H-bond |
Leu80 | Lys146-Nζ | H-bond |
Asp77-0δ1, Thr80-0γ1 | Water-mediated H-bond | |
Leu8OXT | Lys146-Nζ | H-bond |
Thr143-0γ1 | H-bond | |
Tyr84-0η | H-bond | |
β-Phe5 | Ser73, Phe74, Asn70, Tyr116 | VDW |
β-Phe5N | Asn70-0δ1 | H-bond |
β-Phe50 | Ser73-0γ | H-bond |
Phe5 | Ser73, Phe74, Asn70, Tyr116, Val97, Gln114 | VDW |
Phe5N | Asn70-0δ1 | H-bond |
Phe50 | – |
Differences in VDW or H-bonding interactions are indicated by bolded residues.
Compared with alanine scanning of an epitope, the β-amino acid scan has revealed a spectrum of reactivity by two unrelated T cell hybridomas specific for the wild-type peptide. The T cell hybridomas GA4.2 and B3.1 express different Vβ elements and have distinct fine specificities (27). All analogs were recognized, but with different fine specificities evident between the two clones. For example, the β-Phe5 analog was recognized well by both clones, whereas the β-Leu8 analog was poorly recognized by the GA4 clone. This loss of recognition may be due to the increased specificity of the GA4 clone toward the C-terminal end of the peptide. In addition to direct effects on T cell recognition, several methyl insertion positions demonstrated an indirect effect of the β-amino acid on T cell reactivity via the introduction of small changes in peptide conformation distal to the point of insertion. For example, alanine scanning mutagenesis of SIINFEKL has indicated that positions 4, 6, and 7 are important TCR contacts for both clones, except that B3.1 lacks specificity for the residue at position 6 (26). Likewise, alanine scanning mutagenesis revealed that positions 3, 5, and 8 are important for binding (26). β-Amino acid replacements at the major anchor residues (positions 5 and 8) resulted in excellent binding and maintenance of T cell recognition (Fig. 1), whereas replacement at position 3 resulted in poor binding presumably because of conformational changes induced in the peptide and poor orientation of the other anchor residues. The ability to replace anchor residues using β-amino acids represents a potential strategy for generating T cell mimotopes due to the retention of the side chains of the substituted amino acids. Moreover, substitution of TCR-accessible residues appears to be better tolerated in some instances compared with alanine substitutions, because instead of removing the side-chain interaction, the peptide backbone is extended. Thus, lack of recognition of a β-substituted peptide represents an altered peptide conformation.
The immunization of B6 mice with the analog β-Phe5 revealed a high level of cross-reactivity of the resultant TCD8+ cells with the wild-type epitope. Additionally, analog-expanded TCD8+ showed a higher level of killing capacity in a chromium release assay compared with the wild-type peptide. Peptide restimulation is known to induce apoptosis of highly stimulated or overstimulated T cells; thus restimulation with the wild-type peptide may result in lower-affinity T cells (34). This method of using mimotopes to expand T cells may have applications in the restimulation of autologous T cells for adoptive transfer to treat a variety of disease states in cases in which autologous stimulators are not available. When these cells were restimulated ex vivo using APCs presenting endogenous SIINFEKL determinants (EG7), the resulting pool of T cells from both analogs was comparable with the wild-type response, indicating that the β-analogs are stimulating the majority of the high-affinity T cell repertoire as efficiently as the wild-type peptide. This result highlights the preservation of high-avidity T cells when stimulating with heterologous epitopes (wild type and those incorporating β-amino acids), which is essential for the successful in vitro expansion of CTLs before adoptive transfer therapy.
Peptides are often overlooked when considering vaccine candidates due to their susceptibility to a wide variety of proteases. Increasing protease resistance thus becomes essential for not only prolonging the in vivo half-life of the Ag but also for maintaining the correct full-length Ag to prevent skewing or unwanted responses. The insertion of methyl groups into the center of the peptide backbone provided a significant improvement to the serum stability of the entire molecule. This novel cooperative effect of protecting distal bonds from the site of modification enables significant improvements in protease resistance while not sacrificing peptide conformation at TCR-sensitive sites. This is essential for maintaining cross-reactivity with the wild-type target, because modifications that produce structures too different from the wild-type target will not be of therapeutic value.
Our data has shown that the incorporation of β-amino acids at some positions within an epitope can maintain the full immunogenic properties of the natural ligand. Furthermore, these modifications can significantly improve protease resistance in an extracellular setting. Thus, the use of βC3-amino acids may represent a unique approach for the successive development of protease-resistant epitopes for therapeutic applications in infectious disease and cancer.
Acknowledgments
We thank the staff at BioCARS and the Australian Synchrotron Research Program for assistance.
Disclosures
The authors have no financial conflict of interest.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by the Australian Research Council, the National Health and Medical Research Council (Australia), Monash University Small Grants, and the Juvenile Diabetes Research Foundation. J.R. and W.C. were supported by Wellcome Trust Senior Research Fellowships in Biomedical Science in Australia.
Abbreviations used in this paper: TFA, trifluoroacetic acid; ICS, intracellular cytokine staining; MS, mass spectrometry; PK, proteinase K; rmsd, root mean square deviation; RP-HPLC, reversed phase HPLC; VDW, van der Waals’.