An In Vivo Screen to Identify Short Peptide Mimotopes with Enhanced Antitumor Immunogenicity

Self-antigens (Ag) overexpressed within tumors have long been a focus of cancer immunology (1). Trp2 is a tumor-associated Ag (TAA) expressed in melanomas, and mouse and human sequences share more than 80% amino acid sequence identity (2). Trp2_180–188 (SVYDFFVWL) is a putative HLA-A2–restricted epitope (3). This identical sequence and position within murine Trp2 is also a murine H-2K^b restriction element that can be recognized by murine melanoma CTLs. Epitopes within Trp2 have been explored as vaccine Ag in trials for patients with metastatic melanoma (4, 5) and glioblastoma (6). The Trp2_180–188 epitope has been used as a peptide Ag in melanoma trials, although efficacy was limited (7).

MHC-I–restricted T cells with high affinity for self-Ag are generally deleted in the thymus during negative selection (8), so any self-Ag–reactive T cells that escape central and peripheral tolerance would be expected to be weakly responsive (9). To enhance the immunogenicity of TAAs such as Trp2, immunization with vaccine adjuvants can be used (10–14). Alternatively, T-cell responses against self-Ag can potentially be improved with altered peptide ligands (15). Epitope sequences can be modulated to either enhance MHC-I binding, resulting in improved T-cell responses to the wild-type sequence (16–19), or alter the interaction of the peptide–MHC complex with the T-cell receptor (TCR), resulting in improved T-cell recognition and expansion (20). Such alterations to a wild-type epitope create a mimotope, and mimotopes that induce superior functional results are referred to herein as enhanced mimotopes (e-mimotopes).

Because peptide-binding motifs for MHC-I are known, substituting residues at an MHC-I anchor position of a TAA can enhance peptide–MHC-I binding. This can sometimes improve CTL induction (21–23). This strategy is facilitated by the use of computational algorithms to predict peptide–MHC-I binding affinity, reducing the number of peptides to assess functionally (24). However, amino acid alterations to enhance MHC-I binding still might result in insufficient vaccine-induced T-cell responses, or alternatively might induce T cells that lack cross-reactivity with the wild-type epitope (25). The enhancement of peptide binding to MHC-I is not only limited to anchor sites, and substitutions at other positions may also enhance MHC-I binding and T-cell responses (16). For example, for the Trp2_180–188 epitope, alanine alteration at nonanchor point position 7 is reported to enhance MHC-I binding and lead to delayed tumor growth in mice (26). However, improving epitopes by enhancing MHC-I binding affinity has intrinsic limitations because poor MHC-I–binding epitopes might not be effectively displayed on target cell surfaces. On the other hand, strong MHC-I–binding epitopes might not benefit from further enhancement to MHC-I binding. Thus, as a second strategy, substitutions of amino acids in the wild-type epitope that result in enhanced interactions between peptide–MHC (pMHC) and TCR can be a powerful strategy. Using TCRs with known structure, techniques such as surface plasmon resonance (20) and molecular simulation of pMHC–TCR complexation (27) can be used to identify altered peptides with enhanced immunogenicity. These techniques are limited for vaccination, however, because the knowledge of a diverse TCR repertoire structure is not possible with current approaches. It was predicted that for H-2K^b–restricted peptides, position 2 is important for TCR recognition (28). However, Trp2 peptide with Isoleucine (I) alteration on position 2 induced Trp2-specific T cells but failed to protect mice from tumor challenge (29).

Herein, we present a two-step method to identify short e-mimotope cancer vaccine candidates using the H-2K^b–restricted Trp2_180–188 epitope as a test case. The method is then reproduced with other TAAs; the H-2L^d–restricted gp70_423–431 (AH1) epitope and the H-2D^b–restricted mWT1_126–134 epitope. In these studies, immunization is enabled by the CPQ liposome adjuvant system containing cobalt porphyrin–phospholipid (CoPoP), PHAD-3D6A (a synthetic monophosphoryl lipid A variant), and the QS-21 saponin. Peptides containing an abbreviated polyhistidine-tag (his-tag) spontaneously form stable particles with CPQ, and we recently demonstrated this approach induces strong antitumor immune responses with nanogram peptide doses (30, 31).

Trp2, AH1, and WT1 peptide libraries and peptides were synthesized by GenScript. CoPoP was synthesized as reported previously (32). The following lipids were used to produce liposomes: dioleoylphosphatidylcholine (DOPC; Corden #LP-R4–070), cholesterol (PhytoChol; Wilshire), and synthetic PHAD-3D6A (Avanti; #699855). QS-21 was purchased from Desert King (part number NC0949192). The APC-CD8a (clone: CT-CD8a) antibody was obtained from Accurate Chemical and Scientific Corporation (#ACL168APC). The following antibodies were obtained from BioLegend: APC-CD8a (clone: CT-CD8a; #100712), FITC-CD4 (clone: GK1.5; #100405), AF700-CD45 (clone: 30-F11; #103127), APC-Cy7-CD44 (clone: IM7; #103027), PE/Cy7-CD62 L (clone: MEL-14; #104417), PerCP-Cy5.5-PD-1 (clone: RMP1–30; #109119), pacific blue-IFN-γ (clone: XMG1.2; #505818), BV605-TNFα (clone: MP6-XT22; #506329), PE/Cy7-Granzyme B (clone: QA16A02; #372213). Anti-mouse PD-1 (#BP0146) and anti-mouse CTLA-4 (#BP0131) were acquired from Bio X Cell. Other reagents used were Golgiplug/Brefeldin A (BD Biosciences; #555029), Live/Dead dye (Invitrogen; #L34857), Fc-block (clone: 2.4G2; BD Biosciences, #553142), fixation/permeabilization kit (BD Biosciences, #554714), red blood cell (RBC) lysis buffer (BioVision, #5830), Collagenase Type I (Gibco, #17018–029), and DNase I (Roche, #04536282001).

Liposomes were prepared by an ethanol injection method followed by lipid extrusion, as previously described (33). Ethanol was removed by dialysis in PBS at 4°C, then liposomes were passed through a 0.2-μm sterile filter. Where indicated, QS-21 (1 mg/ml) was added to liposomes overnight at 4°C with a (DOPC: Cholesterol: CoPoP/PoP: PHAD-3D6A: QS-21) mass ratio of 20:5:1:0.4:0.4. Liposome concentration was adjusted to 320 μg/ml CoPoP. To form the liposome-displayed peptide vaccine, CoPoP liposomes were mixed with peptides at a mass ratio of 4:1 (CoPoP:peptide) for 1 hour at room temperature.

To assess peptide binding, peptides were incubated with liposomes for 1 hour at room temperature and the mixture was subjected to microcentrifugal filtration using a 100-kDa membrane (PALL; part number 29300) to separate free peptide from liposomes. Free peptide in the filtrate was quantified with a micro BCA assay (Thermo Fisher Scientific; part number 23235). Dynamic light scattering with a NanoBrook 90 plus PALS instrument was used to measure sizes and polydispersity indexes (PDI) of samples in PBS and zeta potential in water.

An identical procedure was followed for CPQ/Trp2–8W, CPQ/Trp2–8Y, and CPQ/Trp2–8C liposomes. Sample vitrification was performed using a Vitrobot Mark IV (Thermo Fisher Scientific). Holey carbon grids C-flat 2/2–2Cu-T (Electron Microscopy Sciences) were washed with chloroform for 2 hours before sample vitrification. Grids were treated with negative glow discharge (EMS 100 Glow Discharge System) in air at 5 mA for 15 seconds before the sample was applied. For all grids, a sample volume of 3.6 μL of was applied to holey carbon grids and manually blotted using Vitrobot blotting paper (Standard Vitrobot Filter Paper, Ø55/20 mm, Grade 595). Next, a volume of 3.6 μL of the same sample was applied to the holey carbon grids, and each grid was blotted once in the Vitrobot for 3 seconds with a blot force +1 before they were plunged into liquid ethane. The Vitrobot was set at 25°C and 100% relative humidity. Data acquisition was performed using SerialEM (34) software on the Titan Krios electron microscope at FEMR-McGill, operated at 300 kV. Images were collected with a Gatan K3 direct electron detector equipped with a Bioquantum imaging filter. Defocus was set at −2.00 μm. Images were collected using a total exposure of 50 e⁻/Ų at a nominal magnification of 81,000× corresponding to a calibrated pixel size of 1.09 Å.

B16-F10 cells were obtained from ATCC in 2019 and cultured in DMEM (Gibco, #11965118) with 10% FBS (Gibco, # A4766801) and 1% penicillin/streptomycin (Gibco, #15140122). RMA-S cells were kindly provided by Dr. Abrams (Roswell Park, Buffalo, NY) in 2020 and cultured in RPMI1640 (Gibco, #21875034) with 10% FBS, 1% penicillin/streptomycin, and l-glutamine (2 mmol/L; Gibco #25030081). CT26 cells were obtained from ATCC in 2019 and cultured in RPMI1640 with 10% FBS and 1% penicillin/streptomycin. All cells were grown in a humidified incubator with 5% CO₂ at 37°C. Cells were not reauthenticated. Cell lines were used for less than 5 weeks after initial thawing. B16F10 and CT26 cell lines tested negative for Mycoplasma.

For splenocyte analysis, murine spleens were dissociated and passed via a 70-μm cell strainer. The plunger of a 3-mL syringe was used to pass tissue through the strainer, then 5 mL of cold PBS was used to wash cells in a 50 mL tube. Cells were centrifuged for 5 minutes at 500 × g then 5 mL RBC lysis buffer was incubated with the cells for 5 minutes. After 35 mL PBS was added, the cells were centrifuged again, and pellets were collected for further use. For tumor-infiltrating lymphocyte (TIL) studies, tumors were isolated and diced into 1–2 mm sections and digested with collagenase Type I (2 mg/mL) and DNase I (100 μg/mL) for 1 hour at 37°C. Cells were passed through a 70-μm strainer and washed. Peripheral blood mononuclear cells (PBMC) were prepared by lysing 100 μL of mouse whole blood with 2 mL of RBC lysis buffer, then were washed twice for further use. Splenocytes, TILs, and PBMCs were cultured in RPMI1640 supplemented with 10% FBS, 1% penicillin/streptomycin, l-glutamine (2 mmol/L), sodium pyruvate (1 mmol/L; Gibco #11360070), 1× diluted nonessential amino acids solution (Gibco, 11140050), and 50 μmol/L β-mercapethanol (Gibco #21985023).

All animal experiments were approved by the University at Buffalo Institutional Animal Care and Use Committee. Five- to 6-week-old female C57BL/6 mice (Jackson Laboratory) or BALB/c mice (Charles River Laboratory) were intramuscularly immunized with 50 μL vaccine in the right quadriceps. For prophylactic vaccination, mice were vaccinated on days 0 and 7, prior to challenge on day 14 with subcutaneous injection of 1 × 10⁵ B16-F10 or CT26 cancer cells. For the Trp2 therapeutic vaccine tumor model, mice were inoculated with 1 × 10⁵ B16-F10 tumor cells subcutaneously on day 0, and then vaccinated with indicated vaccine on days 4 and 11. For mice treated with anti-programmed cell death protein 1 (anti–PD-1) and anti–CTLA-4, antibodies were administered intraperitoneally on day 6, 8, 13, and 15, with 100 μg PD-1 and 100 μg CTLA-4 per mouse per injection. For the AH1 therapeutic vaccine tumor model, mice were inoculated with 3 × 10⁴ CT26 tumor cells subcutaneously on day 0 and then vaccinated with indicated vaccine on days 5 and 12 or days 8 and 15.

Tumor volumes were calculated as [length×width²/2]. Mice were euthanized when the tumor size reached 1.5 cm in diameter or when animals developed a tumor ulceration. For the experimental lung metastasis tumor model, animals were injected with 1 × 10⁵ B16-F10 cells intravenously via tail vein injection on day 0, and vaccinated intramuscularly on days 2 and 9. Lungs were removed and stained with Bouin solution (Sigma #HT10132) on day 23.

pMHC tetramer complexes were generated at the NIH Tetramer core facility as follows: H-2K^b–restricted Trp2_180–188 (e) peptide was complexed with MHC-I (H-2K^b) and conjugated with PE, gp70_423–431 (AH1; SPSYVYHQF) peptide was complexed with MHC-I (H-2L^d) and conjugated with PE, and the mWT1 peptide_126–134 (RMFPNAPYL) was complexed with MHC-I (H-2D^b) and conjugated with PE. PBMCs from 100 μL of blood or 1 × 10⁶ splenocytes or 5 × 10⁶ tumor-infiltrating cells were prepared for antibody staining. For H-2K^b Trp2_180–188 tetramer staining, cells were incubated with tetramer (500× diluted), CD4 (200× diluted), CD8 (200× diluted), and CD45 (200× diluted) for 30 minutes on ice. For AH1 and WT1 tetramer staining, cells were incubated with tetramer (100× diluted) in 4°C for 1 hour and then with antibody mixture of CD4 (200× diluted), CD8 (× diluted) for 30 minutes on ice. For T-cell phenotyping, antibody mixture of Fc-block (100× diluted), CD8a (200× diluted), CD4 (200× diluted), CD45 (200× diluted), CD44 (200× diluted), and CD62 L (200× diluted) were added to cells. For T-cell exhaustion studies, mixture of Live/Dead fixable dye (500× diluted), tetramer (500× diluted), Fc-block (100× diluted), CD8a (200× diluted), CD45 (200x diluted), CD4 (200× diluted), PD-1 (200× diluted) were added to cells. Cells were incubated for 30 minutes at 4°C, then washed twice and subjected to flow cytometry analysis using a BD LSRFortessa X-20 cytometer. FlowJo software was used for analysis. Information about the antibodies and other reagents used is provided above (see Materials and Methods).

For intracellular cytokine staining, 1 × 10⁶ splenocytes in 100 μL cell culture medium were seeded in 96-well plates and stimulated with 10 μg/mL peptide Ag for approximately 18 hours. Brefeldin A was added at a 1,000× dilution for another 5 hours. Cells were transferred to a 96-well round bottom plate and centrifuged at 1,350 rpm. For Trp2, the cell pellets were washed twice and stained with Trp2 tetramer, live/dead fixable dye (500× diluted), Fc-block (100× diluted), and the following antibodies against CD45 (200× diluted), CD8 (200× diluted), and CD4 (200× diluted) for 30 minutes at 4°C. For AH1 studies, cell pellets were washed twice and stained with tetramer for 1 hour at 4°C then stained with live/dead fixable dye (500× diluted), Fc-block (100× diluted), and the following antibodies against CD45 (200× diluted), CD8 (200× diluted), and CD4 (200× diluted) for 25 minutes at 4°C. Cells were fixed and permeabilized according to the manufacturer's instructions. Cells were also stained with antibodies against IFNγ (200× diluted), TNFα (200× diluted), and Granzyme B (200× diluted) for 30 minutes at 4°C, then washed for flow cytometry. For IFNγ measurement by ELISA, 2.5 × 10⁵ splenocytes were seeded in a 96-well plate and stimulated with 10 μg/mL Ag for 72 hours. Fifty microliters of supernatant was collected from each well and subjected to IFNγ ELISA (Thermo Fisher Scientific; part number BMS606TEN) according to the manufacturer's protocol. Information about the antibodies and other reagents used is provided above (see Materials and Methods).

Class I H-2K^b stabilization assays were performed as previously described (16). RMA-S cells were incubated at 26°C with 5% CO₂ overnight. Cells were washed with RPMI medium and seeded on a 96-well plate with 1×10⁶ cells in 200 μL serum-free RPMI medium. 15 μg/mL or 1.5 μg/mL peptide were incubated with cells for 2 hours at 26°C, then cells were incubated at 37°C for another 2 hours. Cells were washed twice and stained with anti–H-2K^b-PE (BD Pharmingen, #553570) on ice for 30 minutes. Cells were washed twice and analyzed by flow cytometry.

Eight-week-old female C57BL/6 mice were either untreated or injected intramuscularly with CPQ/Trp-8C on days 0 and 7, with a vaccine formulation comprising 0.5 μg peptide, 2 μg CoPoP, 0.8 μg PHAD, and 0.8 μg QS-21 per mouse. On day 14, blood and serum were collected for complete blood cell count and serum panel. Fifteen microliters of blood was assessed by Heska Element HT5 Hematology Analyzer for complete blood cell count. Serum was assessed by the Heska Element DC Chemistry Analyzer. Organs (heart, liver, spleen, lung, kidney) were fixed in formalin, stored in 70% ethanol, and subject to hematoxylin and eosin (H&E) staining and imaging as previously reported (33).

For H-2L^d binding, a 96-well ELISA plate was coated with 25 μg/mL AH1 (without a his-tag) overnight at 4°C, then were washed three times with PBS containing 0.1% Tween 20 (Amresco, part number M147; PBST) and blocked with 100 μL 2% BSA in PBST for 2 hours at 37°C with shaking. Eighty microliters of competing peptides with serial concentrations of 50, 5, 0.5, 0.05, or 0.005 μg/mL and 20 μL of 25 μg/mL H-2L^d (BD Pharmingen, part number 550751) were added. The plate was incubated at 4°C overnight, and after washing three times with PBST, 100 μL 1 μg/mL anti-mouse IgG (Genscript, #A00160) was added for 1 hour at room temperature. The plate was washed three times with PBST, and 100 μL of tetramethylbenzidine (SouthernBiotech, #0412–01) was added to each well prior to analysis by absorption measurement at 450 nm (TECAN Safire).

Mice were vaccinated with CPQ/Trp2–8C or CPQ/Trp2–8Y on days 0 and day 7, then splenocytes were prepared on day 14 and stimulated with 10 μg/mL Trp2–8C and Trp2–8Y, respectively. Then cells were stained with live/dead dye, anti-CD8 and anti-IFNγ and subjected to sorting (BD, FACSArial II). DNA was extracted from IFNγ⁺CD8⁺ T cells using the QIAamp DNA Micro Kit according to the manufacturer's instructions (QIAGEN # 51404). Sequencing of TCRβ CDR3 regions was performed by ImmunoSEQ (Adaptive Biotechnologies). TCR repertoire analysis was performed as described (35, 36). Metrics of the complete TCR repertoire in each sample, including the number of productive rearrangements, productive clonality, and clonal frequencies, were determined with ImmunoSEQ software and confirmed with the LymphoSeq package (37). TCR repertoires were analyzed using the LymphoSeq package and custom scripts with R statistical software. The similarity between the different TCR repertoires was measured using the Morisita-Horn Index (38), using the vegan package. This unitless index ranging from 0 to 1, considers the number of shared sequences between 2 repertoires as well as the contribution of those shared sequences to each repertoire.

NetMHCpan was used for prediction of binding affinity of peptides from positional libraries to H-2K^b (39, 40). Percent rank was used to present binding affinity of the core peptide to H-2K^b.

Analysis was carried out using Prism 9 (GraphPad Software) with the tests described in the figure captions.

The data generated in this study are available within the article and its Supplementary Data files.

The e-mimotope development paradigm is outlined in Fig. 1A. First, nine position-scanning peptide libraries of the Trp2_180–188 MHC-I–restricted peptide were generated by randomizing peptide residues while fixing the remaining ones with the original amino acid sequence. Each library had exactly one position of the native peptide randomized. Next, vaccines were prepared by admixing the positional peptide libraries with CPQ liposomes and C57BL/6 mice were then immunized on days 0 and 7 with 1 μg total peptide and were challenged on day 14 with B16-F10 cells, which express Trp2. Only mice vaccinated with the Trp2_180–188 peptide library randomized at position 8 had smaller tumors than control (Fig. 1B). Next, all 20 of the individual Trp2 peptide variants were synthesized (Supplementary Table S1), with each peptide bearing a different amino acid at position 8. Mice were immunized with these peptides and challenged with tumor cells. Mice immunized with Trp2–8C showed the greatest tumor growth inhibition on day 21, compared with the wild-type sequence (Trp-8W) (Fig. 1C). Immunization with Trp-8Y also completely inhibited tumor growth in one mouse (Fig. 1C). Thus, this screening method identified two mutated Trp2_180–188 peptide candidates for further assessment, without any a priori considerations about how the epitope should be mutated for enhanced function.

To confirm the enhanced antitumor efficacy of Trp2–8Y and Trp2–8C, tumor growth was assessed following immunization with these candidate e-mimotopes and the wild-type epitope (Trp-8W). Mice immunized with CPQ/Trp2–8Y and CPQ/Trp2–8C had significantly delayed tumor growth compared with mice immunized with CPQ/Trp2–8W (Fig. 1D). Control liposomes for CPQ, in which the cobalt is replaced by two hydrogen atoms that are otherwise identical but do not induce peptide particle-formation (termed 2HPQ) did not inhibit tumor growth when admixed with the e-mimotopes (Fig. 1E). Therefore, the use of the CPQ adjuvant was required to observe potent antitumor effects. By day 23, only mice vaccinated with Trp2–8Y and Trp2–8C had tumor sizes smaller than 500 mm³ (Fig. 1E). CPQ/Trp2–8W, CPQ/Trp2–8Y, and CPQ/Trp2–8C vaccines increased the percentage of mice with tumor sizes smaller than 1 cm compared with untreated mice or mice vaccinated with the adjuvant alone. However, CPQ/Trp-8Y and CPQ/Trp2–8C were still significantly more effective than the CPQ/Trp2–8W native epitope (Fig. 1F).

To gain insight into the mechanism of the enhanced response to Trp2–8C and Trp2–8Y, the binding of peptides to CPQ or 2HPQ liposomes was assessed. Analysis of peptides admixed with CPQ or 2HPQ liposomes showed that more than 75% of the Trp2–8W, Trp2–8Y, and Trp2–8C peptides bound to CPQ liposomes, whereas none of these peptides bound to 2HPQ liposomes (Fig. 2A). The liposome size was around 100 nm with or without peptide binding (Fig. 2B). Cryogenic electron microscopy of CPQ/Trp2–8W, CPQ/Trp2–8Y, and CPQ/Trp2–8C confirmed that the sizes of the formed particles were around 100 nm and that the liposomes were spherical (Fig. 2C). The polydispersity of the liposome–peptide particles was less than 0.3 (Supplementary Fig. S1A), and the zeta-potential of the liposome was negative with or without peptide binding (Supplementary Fig. S1B). Mice were vaccinated with CPQ/Trp2–8W, CPQ/Trp2–8Y, or CPQ/Trp2–8C on days 0 and 7; then splenocytes were collected on day 14 for flow cytometry analysis. CD8⁺ T cells were analyzed for changes in effector memory (T_EM; CD44⁺CD62L⁻) or central memory (T_CM; CD44⁺ CD62L⁺) phenotypes. On the basis of the flow cytometry gating shown in Supplementary Fig. S2, more than 20% and 10% of CD8⁺ T cells were of the T_EM phenotype in CPQ/Trp2–8C or CPQ/Trp2–8Y vaccinated mice, respectively. There were much lower percentages of T_EM cells in the CD8⁺ T-cell population in mice immunized with the native CPQ/Trp2–8W, and no significant difference was observed in untreated mice (Fig. 2D). The frequency of T_CM in the CD8⁺ T-cell population in the spleens of CPQ/Trp2–8C vaccinated mice also increased compared with CPQ/Trp2–8W vaccinated mice and untreated mice (Fig. 2E). Intracellular levels of IFNγ and TNFα in splenocytes restimulated with 10 μg/mL of the wild-type peptide Trp2–8W in vitro were analyzed using the flow cytometry gating strategy shown in Supplementary Fig. S3. IFNγ was detected in over 10% of CD8⁺ T cells in the splenocytes of CPQ/Trp2–8C vaccinated mice and approximately 2% of CD8⁺ T cells in the CPQ/Trp2–8Y vaccinated mice. However, the wild-type CPQ/Trp-8W group had less than 1% of CD8⁺ T cells that produced IFNγ (Fig. 2F). Compared with the CPQ/Trp2–8W vaccinated mice, significantly more TNFα-producing cells were present in the CD8⁺ T-cell population with CPQ/Trp2–8Y or CPQ/Trp2–8C vaccination (Fig. 2G). We also observed more Trp2_180–188 tetramer⁺ cells in the CD8⁺ T-cell population in the blood of CPQ/Trp2–8C vaccinated mice compared with the CPQ/Trp2–8W vaccinated mice (Supplementary Fig. S4).

Next, T cells elicited from e-mimotope immunization were assessed for response to wild-type peptide or e-mimotope stimulation. In the splenocytes of CPQ/Trp2–8C vaccinated mice, there were higher percentages of CD8⁺ T cells producing IFNγ upon stimulation with Trp2–8C, Trp2–8Y, or Trp2–8W peptide compared with other vaccination groups (Fig. 2H). Splenocytes from CPQ/Trp2–8C vaccinated mice responded to Trp2–8C stimulation better than stimulation with Trp2–8W and Trp2–8Y. Splenocytes responded to Trp2–8W and Trp2–8Y stimulation with almost the same pattern, with around 5% of the CD8⁺ T cells producing IFNγ at the highest peptide concentration and decreasing to background levels with peptide concentrations lower than 0.1 μg/mL. At low concentrations of Trp2–8C for stimulation, CD8⁺ T cells still produced IFNγ. In the CPQ/Trp2–8Y vaccinated mice, IFNγ⁺CD8⁺ T cells were observed with Trp2–8Y stimulation, but the levels were low when Trp2–8C or Trp2–8W were used for stimulation (Fig. 2I). Splenocytes from the CPQ/Trp2–8W vaccinated mice did not produce detectable IFNγ⁺CD8⁺ T cells following stimulation with Trp2–8C, Trp2–8Y, or Trp2–8W (Fig. 2J). Using a sensitive ELISA detection method, we found that splenocytes from Trp2–8W immunized mice produced IFNγ in response to stimulation with high concentrations of Trp2–8W peptide (Fig. 2K). However, mice immunized with the Trp2–8C e-mimotope produced far more IFNγ when stimulated with the same peptide, even at concentrations 1–2 orders of magnitude lower. Taken together, these data show that immunization with e-mimotopes induced a higher frequency of cytokine-producing T cells, compared with the wild-type epitope.

We next assessed whether the mechanism for the enhanced T-cell response induced by Trp2–8C and Trp2–8Y e-mimotopes compared with Trp2–8W pertained to MHC-I binding or pMHC–TCR stabilization. We first used the NetMHC neural network algorithm to predict the binding affinity of each positional library member to the H-2K^b MHC-I restriction element. The binding percentile of the wild-type Trp2–8W peptide is 0.04%, which is a top MHC-I binder. The library of Trp2-Pos4, Trp2-Pos7, and Trp2-Pos8 was predicted to have, on average, less diminished binding than other positional libraries (Fig. 3A). In addition, all peptides from the Trp2-Pos8 library were predicted to have better binding affinity than Trp2–8W (Fig. 3B). However, in all cases, MHC-I binding was expected to be effective.

RMA-S murine lymphoma cells are TAP-deficient but express H-2K^b on their surface at lower temperatures, such as 26°C, or when stabilized by exogenous MHC-I–binding peptides. Staining for H-2K^b on the cell surface of RMA-S cells thus represents a useful tool for assessing the H-2K^b–binding affinity of putative MHC-I epitopes (41). We found that RMA-S cells that had been incubated with 15 μg/mL Trp2–8W, Trp2–8Y, and Trp2–8C had more H-2K^b expression compared with those incubated with 1.5 μg/mL of peptide or no peptide (Fig. 3C and D). However, the three peptides appeared to have similar MHC-I–binding affinities. A similar result was observed with 1.5 μg/mL of peptide during incubation (Fig. 3E). A higher percentage of RMA-S cells expressed H-2K^b with 15 μg/mL peptide incubation compared with 1.5 μg/mL.

To better understand TCR usage induced by Trp2–8C and Trp2–8Y vaccination, TCR sequencing was performed on IFNγ⁺CD8⁺ splenocytes from mice vaccinated with either CPQ/Trp2–8C or CPQ/Trp2–8Y. The 30 most frequent T-cell clones accounted for more than half the TCR sequences in the CPQ/Trp2–8C vaccinated group and approximately 65% in the CPQ/Trp2–8Y group (multicolored portion of graphs in Fig. 4A). Reciprocally, about half the Ag-specific CD8⁺ T cells induced by CPQ/Trp2–8C and approximately 35% in the CPQ/Trp2–8Y group corresponded to less frequent clones, shown in purple. On the basis of the higher immunogenicity, as indicated by IFNγ staining, we were able to sort a higher number of Ag-specific T cells for TCR sequencing in the Trp2–8C group compared with Trp2–8Y group. Thus, the total numbers of sequenced TCR transcripts and unique CDR3s was more than 2-fold higher in the Trp2–8C group compared with the Trp2–8Y group. However, the overall clonality and gini coefficient and clonal relatedness of the two e-mimotope groups were not significantly different (Supplementary Fig. S5A). We also analyzed the CDR3 length distribution and the frequency of the TRBJ and TRBV segment usage for each animal. A significant shift in CDR3 length was observed between Trp2–8C and Trp2–8Y repertoires. The preferential lengths of CDR3 were 12 amino acids in the Trp2–8Y group and 14 amino acids in the Trp2–8C group (Fig. 4B). A significant increase in the frequency of J02–02 and J02–05 in the Trp2–8C group was observed compared with the Trp2–8Y group. In contrast, the usage of J02–07 was significantly lower in the Trp2–8C group compared with the Trp2–8Y group (Fig. 4C). The Trp2–8C group also showed increased usage of V12–02, V14–01, and V31–01 and decreased usage of V04–01, V13–02, V27–01, and V29–01 compared with the Trp2–8Y group (Fig. 4D). The VJ family combined usage was also significantly different (Supplementary Fig. S5B). Similarity assessment of TCR repertoires between samples showed that TCR heterogeneity existed within vaccine-induced Ag-specific T cells between individual mice, especially in mice from different vaccine groups (Fig. 4E; Supplementary Fig. S5C and S5D). There were some common clonotypes observed between replicates in each vaccinated group, but there were only three clonotypes observed to be in common between the Trp2–8C and Trp2–8Y repertoires (Fig. 4F).

To assess the safety of the e-mimotope and CPQ vaccine, C57BL/6 mice were vaccinated on days 0 and 7 with the functional dose of 500 ng of peptide. The bodyweight of mice showed normal gain before and after vaccination (Supplementary Fig. S6A). No apparent differences in the heart, liver, spleen, lung, or kidney were observed with histology (Supplementary Fig. S6B). A complete blood cell count (Supplementary Fig. S6C) and serum chemistry panel (Supplementary Fig. S6D) showed no significant differences between vaccinated mice and healthy mice in all parameters. These results are consistent with our prior report that CPQ with the A5 mimotope is well-tolerated in mice (31).

To assess e-mimotopes in a therapeutic context, mice were inoculated with B16-F10 tumor cells on day 0 and injected with the peptide vaccines on days 4 and 11. Wild-type CPQ/Trp2–8W vaccination did not delay tumor growth in any mice (0/10). However, CPQ/Trp2–8Y delayed tumor growth in 3 of 10 mice and CPQ/Trp2–8C delayed tumor growth in 5 of 10 mice (Fig. 5A). CPQ/Trp2–8Y and CPQ/Trp2–8C vaccinated mice showed significantly smaller tumor sizes on day 28 (Fig. 5B).

Next, the Trp2–8Y and Trp2–8C e-mimotopes were compared with the Trp2–8W wild-type peptide for immunization in a therapeutic lung metastasis tumor model. Lungs from the CPQ/Trp2–8C vaccinated mice had significantly fewer nodules compared with the lungs from CPQ/Trp2–8W vaccinated mice (Fig. 5C and D). The lung weights reflected the lung tumor nodule counts (Fig. 5E).

Next, we investigated the T-cell response in the tumor microenvironment (TME) on day 28 postinoculation with B16-F10 cells; vaccines were administered on days 4 and 11. Using the flow cytometry gating shown in Supplementary Fig. S7, a much higher frequency of CD45⁺ cells (Fig. 6A) and CD8⁺ T cells (Fig. 6B) infiltrated the TME following Trp2–8Y and Trp2–8C e-mimotope vaccination compared with wild-type CPQ/8W vaccination. Approximately 40% of the CD8⁺ T cells in the TME were Ag-specific with CPQ/Trp2–8C or CPQ/Trp2–8Y vaccination, whereas approximately 20% of CD8⁺ T cells were Ag-specific in the TME with CPQ/Trp2–8W vaccination (Fig. 6C). Even though approximately 20% of the CD8⁺ T cells in the TME were Ag-specific, the overall number of CD8⁺ T cells in the TME was very low. We examined the expression of the inhibitory receptor PD-1 in both the spleen and the TME. A significantly higher frequency of intratumoral tetramer⁺CD8⁺ T cells expressed PD-1 compared with splenic tetramer⁺CD8⁺ T cells (Fig. 6D). Compared with CPQ/Trp2–8W vaccinated mice, CPQ/Trp2–8C and CPQ/Trp2–8Y vaccinated mice had a significantly lower proportion of CD8⁺tetramer⁺ T cells that expressed PD-1 within the TME. However, the overall percentage of PD-1⁺tetramer⁺CD8⁺ T cells in live cells and the actual number of live PD-1⁺tetramer⁺CD8⁺ T cells was higher in the e-mimotope vaccinated group compared with the wild-type peptide vaccinated group within the tumor (Supplementary Fig. S8).

Immune checkpoint blockade (ICB) in the B16-F10 murine melanoma model, with the combination of anti–PD-1 and anti–CTLA-4 antibodies, expands TILs and improves survival (42). We tested e-mimotope vaccination and ICB combination therapy by first inoculating B16-F10 tumor cells subcutaneously on day 0, followed by vaccination with CPQ with the indicated peptides on days 4 and 11, and then providing ICB as a mixture of PD-1 and CTLA-4 antibodies on days 6, 8, 13, and 15. Vaccination alone or the combination of ICB and vaccination delayed tumor growth compared with ICB alone (Fig. 6E). Vaccination combined with ICB inhibited tumor growth significantly more than vaccination alone. With the vaccine and ICB combination therapy, the percentage of mice that developed tumors greater than 1 cm was significantly lower than the groups given vaccine or ICB alone (Fig. 6F). We also compared CPQ vaccination with e-mimotopes Trp2–8Y, Trp2–8C, or the wild-type Trp2–8W. All three vaccines delayed tumor growth compared with the untreated group; however, the Trp2–8C e-mimotope combined with ICB delayed tumor growth significantly better than Trp2–8W given with ICB (Fig. 6G) and led to a greater percentage of mice maintaining tumor sizes smaller than 1 cm (Fig. 6H).

Although the two-step screening method identified effective e-mimotopes for Trp-2, we wanted to demonstrate the versatility of the approach using a different MHC-I haplotype. To do so, we applied an adapted two-step screening method to the H-2L^d–restricted epitope AH1. We tested all peptides from positions 1, 3, 5, 8 (Fig. 7A; Supplementary Table S2), as amino-acid substitution of residues at anchor positions 2 and 9 or the T cell–binding positions 4, 6, or 7 can result in loss of functional activity (20). Mice were vaccinated with CPQ and peptides on days 0 and 7 and then inoculated with CT26 tumor subcutaneously on day 14. With low individual peptide doses of 50 ng per mouse, amino acid substitutions with C, Q, H, I, L, M, F, W, Y, V at position 1; E, H, P at position 3; A, C, G, T at position 5; or N, D, H at position 8 inhibited CT26 tumor growth (Supplementary Fig. S9). Thus, the AH1 epitope appeared to be much more tolerant to amino acid substitutions compared with the Trp2 peptide. Indeed, it has been shown that this epitope can have multiple amino acid substitutions while maintaining activity (43). To screen which peptide elicits the best antitumor activity, we tested these peptide variants in a therapeutic tumor model, administering vaccine on days 5 and 12 after CT26 tumor cell inoculation. Numerous peptides provided full tumor-free survival (Fig. 7B). Those were then further tested in a more advanced tumor model in which vaccine was given on days 8 and 15 after tumor inoculation. Of all peptides assessed, AH1–5C had the best antitumor efficacy (Fig. 7B). Using this approach, the AH1–5A mimotope was also identified to have better antitumor efficacy than the wild-type peptide, which was previously identified in the course of studying pMHC–TCR interactions and has subsequently been used frequently (20). The AH1–5C e-mimotope delayed tumor growth (Fig. 7C) and increased the percentage of mice with tumor sizes smaller than 1 cm (Fig. 7D) relative to the wild-type peptide.

We next tested the mechanism of the enhanced antitumor efficacy of the AH1–5C e-mimotope relative to the wild-type AH1 peptide. Compared with CPQ/AH1 peptide vaccination, CPQ/5C vaccination induced a higher percentage of tetramer⁺ cells (Fig. 7E) and T_EM in the CD8⁺ T-cell population (Fig. 7F). Immunization also induced a higher frequency of CD8⁺ T cells that produced IFNγ (Fig. 7G) and granzyme B (Fig. 7H) in the spleen after peptide stimulation (10 μg/mL).

To understand why AH1–5C had better antitumor efficacy than other e-mimotopes, we used the well-established AH1–5A e-mimotope as a comparator (20). Splenocytes were stimulated with different concentrations of wild-type AH1 peptide. The percentage of IFNγ-producing cells in the AH1 tetramer⁺CD8⁺ T cells was assessed and then normalized by comparing to the percentage of IFNγ-producing cells in the AH1 tetramer⁺CD8⁺ T cells at the maximum stimuli concentration (Supplementary Fig. S10A). The half-maximal effective concentration (EC₅₀) of AH1 stimuli in the AH1 and AH1–5C vaccinated mice was less than in the AH1–5A vaccinated group, indicating that these IFNγ-producing T cells elicited by AH1 and AH1–5C had higher sensitivity for AH1 peptide compared with T cells elicited by AH1–5A (Supplementary Fig. S10B). An ELISA method was used to test the H-2L^d binding of these peptides. Both AH1–5A and AH1–5C bound to H-2L^d (Supplementary Fig. S10C) and both bound moderately better than AH1 (Supplementary Fig. S10D). However, because these three peptides exhibit efficient H-2L^d binding, it is unlikely that the enhancement of MHC-I binding accounted for the dramatic enhancement of immunity induced by immunization with AH1–5A or AH1–5C. We then investigated whether the immunity enhancement of AH1–5C over AH1–5A reflected the nature of the peptide–MHC-I binding versus the pMHC–TCR interaction (20). Because there was no significant difference in MHC-I binding between AH1–5A and AH1–5C, we presumed that, as with the Trp2–8C e-mimotope, it is the pMHC–TCR interaction that accounts for the AH1–5C enhancement of immunity. We also verified that there were no significant differences between AH1–5C, AH1–5A, and wild-type AH1 groups with regard to the binding to liposome, liposome sizes, polydispersity (PDI), or zeta-potential (Supplementary Fig. S11).

To further validate the two-step screening methodology, the murine Wilms' Tumor Gene 1 epitope mWT1_126–134 (RMFPNAPYL) was analyzed. This epitope has been investigated as a vaccine target for murine cancers (44, 45) and human WT1 is a known TAA (46). mWT1_126–134 is H-2D^b restricted. In the first step of the screening method, CPQ/Pos6 and CPQ/Pos7 positional microlibrary vaccination increased mWT1 tetramer⁺ cells (Supplementary Table S3; Supplementary Fig. S12A) and T_EM in the CD8⁺ T-cell population compared with CPQ/wild-type peptide vaccination (Supplementary Fig. S12B). In the second step of screening, it was found that amino-acid substitution with G at position 6, or D, E, or W at position 7 in the mWT1 epitope significantly increased immunogenicity (Supplementary Fig. S12C and S12D). Unlike the other e-mimotopes described in this study, the mWT1 e-mimotope was developed without direct antitumor functional screening as a metric and rather using tetramer staining. Further testing of the mWT1 e-mimotopes has not yet been carried out. A summary of all the e-mimotopes identified is found in Supplementary Table S4.

A two-step screening method was developed that identified Trp2_180–188 and AH1(gp70_423–47) as e-mimotopes with improved functional activity for antitumor vaccination. The approach makes no a priori provisions and is based simply on in vivo immunization. Trp2–8C and Trp2–8Y were demonstrated to be more immunogenic in mice compared with the wild-type Trp2–8W. Multiple AH1 e-mimotopes were identified and AH1–5C in particular was identified to have superior immunity. We also demonstrated that immunogenicity enhancement of Trp-8C, Trp2–8Y, and AH1–5C were independent of MHC-I binding, so therefore likely operate by enhancing pMHC–TCR complexation. The two-step method was also used successfully to produce e-mimotopes directed against murine WT1_126–134, based on tetramer staining alone rather than direct antitumor functional screening.

As shown in Supplementary Table S4, a total of 27 e-mimotopes were identified for three CD8⁺ T cell TAA epitopes across three mouse MHC haplotypes (H-2K^b, H-2D^b, and H-2L^d). Given that each of the three MHC molecules presents with distinct features, further studies are required to determine the extent to which single mutation e-mimotopes can be predicted directly from CD8⁺ epitope sequences for each MHC. Initial analysis shows that of the e-mimotopes discovered in this work, none were at position 2 or 4, and the most common replacement amino acids in the e-mimotopes were C, H, and Y. With more data from additionally generated e-mimotopes, we anticipate that development of simple computations prediction algorithms will be possible in the future.

There have been prior reports that describe mutations to Trp2_180–188 position 2 (26, 29), and that in the context of H-2K^b, position 2 of this peptide can change T-cell recognition (28). The binding motif of H-2K^b is tyrosine at position 5, and Alanine and Glycine at position 2, or alternatively phenylalanine at position 5 and Isoleucine at position 2 (47). On the basis of the sequence of Trp2_180–188 (SVYDFFVWL), F is present at position 5, so the change of position 2 to I could enhance MHC-I binding. However, it was found that substitution of I in position 2 did not actually enhance the MHC-I binding, and despite producing T cells with higher affinity, tumor growth was not inhibited (29). These results show that amino-acid substitutions at an MHC-I anchor point do not reliably enhance MHC-I binding and underscore challenges for the rational design of e-mimotopes. Furthermore, even if Ag-specific T cells can be produced, these induced T cells may not lyse tumor cells. By directly selecting for either higher induction of target T cells (with tetramer staining), or functional T cells (with tumor challenge), our screening approach bypasses that limitation and screens a large pool of candidate e-mimotopes. The positional library strategy is enabled by the potent CPQ vaccine adjuvant and effectively results in functional screening of 180 different mimotope candidates based on directly finding unique mutations with strong antitumor efficacy.

Our work is not the first to assess position 8 mutations on H-2K^b epitopes, and it has been predicted that leucine or methionine on position 8 can enhance binding to H-2K^b (48). The NetMHC prediction showed that substitution of many amino acids at position 8 generally increases MHC-I binding. However, our data show that the immune enhancement of Trp2–8C and Trp2–8Y was independent of MHC-I binding. Indeed, of the numerous position 8 mutants screened, few had functional activity, but most increased MHC-I binding. Thus, the mechanism of enhanced immunity of Trp-8Y and Trp2–8C is more likely to be related with TCR interaction. The same conclusion is drawn for the AH1–5C e-mimotope.

T cells elicited by altered peptides can respond poorly to the wild-type peptide (25). However, for Trp2–8C and AH1–5C, the T cells elicited by e-mimotope immunization responded to not only the e-mimotope epitopes, but also the native epitope better than T cells elicited with immunization using the native epitopes. However, the induced T cells were more responsive toward the mutated Trp-8C peptide. Thus, the T-cell repertoire activated by Trp2–8C was sufficient to cross-react with wild-type Trp2–8W as well as the mutated Trp2–8Y, and this phenomenon has been observed with other mutated peptides (49, 50). It is reasonable to assume that an effective e-mimotope is required to activate a similar TCR repertoire compared with the wild-type epitope (51). T cells from Trp2–8C and Trp2–8Y immunized mice responded to wild-type peptide stimulation more strongly than T cells from mice immunized with wild-type peptide. The reason for this could be that Trp2 is a self-Ag, and only T cells with low affinity remain after thymic deletion and would be difficult to be activated with the native epitope, which can be circumnavigated by using Trp2–8C and Trp2–8Y for activation.

One limitation of this study is that it has not yet been established how the two-step method could be adapted to human TAAs. The use of humans to carry out this screening method is not feasible. However, recent advances in the development of human immune system (HIS) mice suggest that the methodology could be adapted to human-relevant TAAs without significant barriers (52). Although the use of conventional transgenic HLA-A2–expressing C57BL/6 mice lacking human T-cell repertoires is not ideal, HLA-A2–expressing NSG mice that carry human T-cell repertoires, which are made by way of engrafting human hematopoietic stem cells (HSC), are available. For example, the HIS mice that possess functional human CD8⁺ T cells (53), human NKT cells (53), and human dendritic cells (54) can be made by transducing genes that encode HLA-class I molecules, human CD1d molecules and selected human cytokines to NSG mice, followed by the engraftment of HLA-matched HSCs. These HIS mice have been shown to elicit human melanoma epitope–specific human CD8⁺ T-cell responses upon immunization with a nanoparticle-based vaccine (55). Therefore, human TAAs could be developed using humanized mouse models and the same two-step approach, although this needs to be tested. As we demonstrated that different e-mimotopes for the same epitope induce different Ag-specific TCR repertoires, functional testing of any new human e-mimotope is required.

An additional limitation of this study is that all the peptides used contained 3 histidine residues at the N terminus to enable spontaneous particle formation with CPQ liposomes. We previously found that different lengths of his-tag on the N terminal peptide slightly inhibited H-2L^d binding, but it did not impact them as vaccine Ag to induce Ag-specific T cells (11). Nevertheless, further work would be useful to determine whether the naked peptides without the CPQ adjuvant system are effective as e-mimotopes. Finally, although this two-step method led to e-mimotope identification for three different CD8⁺ T-cell epitopes, the broad applicability and success rate of the technique requires further investigation. We are currently in the process of using this approach to identify new e-mimotopes from other mouse self-Ag.

In conclusion, the CPQ vaccine adjuvant system enabled a two-step procedure to systematically identify several e-mimotopes. For example, compared with the wild-type Trp-8W epitope, Trp2–8C and Trp2–8Y induced stronger T_EM and T_CM responses and cytokine-producing T cells with wild-type Ag stimulation. The enhanced T-cell response was independent from MHC-I binding and likely relates to enhanced pMHC–TCR complexation. E-mimotopes were effective as vaccine Ag in both therapeutic and lung metastasis tumor models and resulted in more CD8⁺ T-cell infiltration. E-mimotopes also synergized with ICB treatment. Using a similar approach, e-mimotopes were identified for the AH1 and WT1 epitopes with different MHC-I haplotypes. Thus, e-mimotope discovery appears versatile. This methodology makes no a priori assumptions of target epitope sequence mutations, and directly interrogates and selects for functional mutations within hundreds of altered peptides candidates. Overall, further efforts are warranted to develop this simple and straightforward method to rapidly identify e-mimotopes with improved antitumor function.

X. He reports a patent for 63/137,036 pending. J.F. Lovell reports a patent for 63/137,036 pending and holds equity in POP Biotechnologies. No disclosures were reported by the other authors.

X. He: Conceptualization, methodology, formal analysis, visualization, writing–original draft, writing–review and editing. S. Zhou: Methodology. B. Quinn: Methodology. D. Jahagirdar: Formal analysis. J. Ortega: Formal analysis, writing–review and editing. M.D. Long: Formal analysis, visualization. S.I. Abrams: Conceptualization, writing–review and editing. J.F. Lovell: Conceptualization, writing–original draft, project administration, writing–review and editing.

This study was supported by the NIH (R01CA247771 to both J.F. Lovell and S.I. Abrams). The NIH Tetramer Core Facility provided MHC-I tetramers. Cryo-EM was carried out at the Facility for Electron Microscopy Research at McGill University (Montreal, Quebec, Canada), which is supported by the Canadian Foundation for Innovation, the Quebec Government and McGill University (Montreal, Quebec, Canada).

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.