The DNA plasmids designed here were based on previously published vectors encoding five or thirteen 27-mer CT26 in silico predicted neoepitopes from the mouse tumor cell line CT26, linked by (Glycine-Serine (GS))₅ as poly-epitopes on a string, called Neo5 and Neo13. The sequence and selection strategies of the neoepitopes used for these studies have been previously described by Viborg et al. (33).

The APC-targeting pDNA constructs were based on the above. The APC-targeting DNA inserts contained i) an APC-binding molecule, ii) a dimerization module consisting of hinge 1, hinge 4, and the CH3 domain (h1h4CH3) from human IgG3 (21), and iii) the poly-neoepitope unit consisting of five or thirteen neoepitopes. Each of the modules was fused by Glycine-Leucine (GL) linkers. The sequences of the APC-binding molecules chemokine (C-C motif) ligands 3, 4, 5, 19, 20, 21 (CCL3, CCL4, CCL5, CCL19, CCL20, CCL21), XCL1, GM-CSF, Fv αDEC205, Fv αClec9a, and Celc9a ligand are found in Supplementary Table 1 . In the absence of an endogenous secretion signal in the APC-binding molecule (Fv αDEC205, Fv αClec9a, and Clec9a ligand), the secretion signal of murine serum albumin (MKWVTFLLLLFVSGSAFS) was inserted in frame upstream. To design a non-targeted pDNA construct (NT_Neo5) encoding a secretion fusion protein unable to target any receptor, we introduced the secretion signal peptide from CCL3 upstream of the dimerization module.

To study the implications of the dimerization module in the immunogenicity of the delivered neoepitopes, we designed two additional constructs where the sequence of h1h4CH3 was either eliminated to design a monomeric APC-targeting DNA construct (CCL19_Neo5 monomer) or substituted with the dimerization sequence MHD2 from human IgM (CCL19_hMHD2_Neo5) (34).

All DNA insert sequences were optimized for codon adaptation in mice, GC content, repeated sequences, and mRNA-free energy using previously described tools (33). The DNA inserts were synthesized and subcloned into the expression vectors pUMVC4a (Aldevron, #4038) or pTVG4 (35) using EcoRI and NotI sites and upscaled by Aldevron (US) as in vivo-grade DNA.

To strengthen the translatability of the pDNA designs, empty pTVG4, and CCL19_Neo13 DNA plasmids were upscaled at Cobra Biologics (UK) as clinical-grade DNA.

BALB/c syngeneic colon cancer cell line CT26 (ATCC, #CRL2638) was cultured in RPMI (Gibco, #72400-021) supplemented with 10% heat-inactivated fetal calf serum (FCS) (Gibco, #10500-064). HEK293T cells (ATCC, #CRL-1573) were cultured in DMEM (Merk, #D6546) supplemented with 10% FCS, 1% GlutaMAX (ThermoFisher, #35050061), and 1% Penicillin/Streptavidin. CHO-K1 cells (ATCC, #CCL-61) were cultured in RPMI supplemented with 10% FCS and 1% Penicillin/Streptavidin. All cell lines were kept at 37°C and 5% CO_2, following the supplier’s instructions. All cell lines were regularly tested for mycoplasma.

Adherent HEK293T or CHO-K1 cells were seeded in 6-well plates (0.25 x 10⁶ cells/well) and transfected after 48h using Lipofectamine 2000 (ThermoFisher, #L3000015). According to the manufacturer’s instructions, 1 µg of DNA plasmid and 6 µL of Lipofectamine were diluted in 125 µL of OptiMEM (ThermoFisher, #31985062) and added to each well. The supernatants were collected 48 hours after for analysis.

To evaluate the correct molecular weight of the different fusion proteins, the supernatant of transfected HEK293T cells was run with β-mercaptoethanol (Sigma-Aldrich, #63689) on a 4-20% Mini-Protean® TGX Stain-Free™ Protein Gel (BioRad, #4568093). After transfer to PVDF membranes (ThermoFisher, #88518), the membranes were blocked with 1x Pierce™ Clear Milk Blocking Buffer (ThermoFisher, #37587) and incubated overnight with Mouse anti-Human IgG (CH3 domain) Secondary Antibody (Invitrogen, #MA5-16557). After washing with TBS-Tween (ThermoFisher, #28360), the membranes were incubated for 45 min with Anti-mouse IgG HRP-conjugated (R&D systems, #HAF007) and developed using SuperSignal™ West Atto Ultimate Sensitivity Substrate (ThermoFisher, #A38555). Images were obtained using the iBright FL1500 Imaging System (Invitrogen, #A44241) and analyzed with iBright Analysis Software Version 5.1.0 (Invitrogen).

The expression levels of CCL19_Neo5 monomer, CCL19_Neo5, and CCL19_hMHD2_Neo5 were assessed in the supernatant of transfected CHO-K1 cells by murine CCL19 DuoSet sandwich ELISA (R&D systems, #DY440). CCL19/MIP-3 beta antibody pairs were used to capture and detect the fusion proteins. Plates were developed using DuoSet ELISA Ancillary Reagent Kit 2 (R&D Systems, #DY008) according to the manufacturer’s instructions.

Mice: 6 to 8 weeks old BALB/c JrJ females were acquired from Janvier Labs (France). All animal experiments were conducted under the license 2017-15- 0201-01209 from the Danish Animal Experiments Inspectorate in accordance with the Danish Animal Experimentation Act (BEK nr. 12 of 7/01/2016), which is compliant with the European directive (2010/63/EU).

Tumor challenges were conducted as previously described (33). Briefly, on the day of tumor cell inoculation (defined as study day 0), 5 x 10⁵ or 2.5 x 10⁵ CT26 cells were injected subcutaneously (s.c.) in the right flank of mice. The tumor volume was measured three times per week and calculated using the following formula: tumor volume = π 6 * ( d 1 *d 2 ) 3 / 2 . Experiments were terminated when the majority of tumors in the control groups reached a 12 mm diameter in any direction. Individual mice were euthanized upon reaching humane endpoints (i.e.15% loss of body weight or tumor ulcerations). When depicting longitudinal tumor volumes, missing data was mitigated by applying Last-Observation-Carried-Forward (LOCF).

The induction of neoepitope-specific CD8+ T cells was monitored during the animal studies as described previously (33). In brief, tail-vein blood from a representative number of mice was stained with fluorochrome-labeled antibodies to allow for the identification of CD8+ T cells, and fluorochrome-labeled MHC class I neoepitope-specific tetramers (murine allele H-2K^d loaded with C1 minimal peptide KFKASRASI, from hereon: C1 multimer) purchased from Tetramer Shop (Denmark). The full gating strategy is illustrated in Supplementary Figure 7A .

Upon termination of the studies, spleens from a representative number of mice were harvested and collected in RPMI supplemented with 10% FCS (Gibco, #10500-064), processed into single cells suspensions by GentleMACS processing (Miltenyi Biotec, C-tubes, #130-096-334 and Dissociator, #130-093-235) and cryopreserved.

To detect neoepitope-reactive T cells induced by the vaccination, splenocytes were re-stimulated with synthetic neopeptides and analyzed by intracellular cytokine staining (ICS) or enzyme-linked immunosorbent spot (ELISpot).

Synthetic neopeptides: 27-mer peptides, corresponding to the neoepitopes encoded in the pDNA constructs, were synthesized by GenScript (New Jersey, USA). The 27-mer peptides feature the mutated AA in the central position. The lyophilized peptides were dissolved in dimethyl sulfoxide (DMSO) (Merck, #D8418) at 10 mg/mL for peptide re-stimulation.

ICS: 2 x 10⁶ splenocytes per well were stimulated with synthetic neopeptide pools corresponding to the vaccine content. Following re-stimulation, frequencies of interferon γ (IFNγ) and tumor necrosis factor α(TNFα)-producing CD4+ and CD8+ T cells were analyzed by flow cytometry and quantified as previously described (33). The full gating strategy is illustrated in Supplementary Figure 7B . Samples where less than 500 CD8+ T cells were acquired were excluded from the analysis.

ELISpot: 0.5 x 10⁶ splenocytes per well were plated in anti-IFNγ antibody-coated ELISpot plates and stimulated either with synthetic neopeptide-pools corresponding to the vaccine content or with the individual peptides overnight. IFNγ-positive cell spots were developed as previously described (33). The plates were imaged using a CTL ELISPOT Analyzer.

To evaluate the performance of the different APC-binding molecules, the pDNA vaccines were ranked according to i) anti-tumor efficacy, ii) induction of reactive CD8+ T cells, and iii) induction of reactive CD4+ T cells in the spleen compartment measured by ICS. The readouts for three features were normalized, scaling the maximum values to 1, and averaged to generate a score that is used to rank the efficacy of the APC-binding modules in the immunogenicity and anti-tumor efficacy of neoepitope-encoding pDNA designs.

GraphPad Prism 9 for Mac OS X was used for graphs and statistical analyses. Data were subjected to the Shapiro-Wilk test for normality (alpha = 0.05). Parametric data were analyzed by One-way ANOVA with Šidák´s test for multiple comparisons. Non-parametric data were analyzed by the Kruskal-Wallis test with Dunn’s multiple comparison correction. For the tests described above, the following levels of statistical significance are applied: ns p ≥ 0.05, *p< 0.05, **p< 0.01, ***p< 0.001, ****p< 0.0001. Kaplan-Meier survival curves were assessed by the Mantel-Cox test, and significance levels were corrected for multiple comparisons by the Bonferroni method. For these tests, *p< 0.0167.

To evaluate the efficacy of the APC-targeting strategy for the delivery of in silico predicted neoepitopes, we selected 11 APC-binding molecules reported to bind to receptors on the surface of cDCs. Specifically, we chose molecules binding receptors implicated in i) the endocytosis of antigens, ii) the regulation of the cDC migration, and iii) the maturation process of cDCs ( Supplementary Table 2 ).

Using these molecules, we designed APC-targeting pDNA constructs encoding fusion proteins consisting of i) an APC-binding molecule, ii) the dimerization module h1h4CH3 from human IgG3, and iii) 5 in silico predicted CT26 neoepitopes previously published (33) ( Figure 1A ).

We confirmed that the different APC-targeting pDNA constructs resulted in the expression secretion of fusion proteins with the expected molecular weight in the supernatant of DNA-transfected HEK293T cells ( Figure 1B ). While acknowledging the limitations of immunoblotting for precise quantification, our data suggest potential differences in the expression levels of the vaccine candidates. Specifically, NT_Neo5, CCL20_Neo5, and Celc9a_Neo5 show the highest expression levels, while CCL19_Neo5 exhibits the lowest.

We assessed the potential of the different neoepitope-encoding APC-targeting pDNA vaccines (APCt_Neo5) to enhance T-cell responses and tumor rejection in the BALB/c syngeneic model of colon carcinoma, CT26, in a prophylactic treatment setting. In two separate experiments, mice were immunized i.m. with 5 µg of research-grade APCt_Neo5 pDNA. The elicited T-cell responses and tumor-rejection efficacy were compared to the non-targeted pDNA construct NT_Neo5, which contains the same selection of neoepitopes but not an APC-binding molecule.

To evaluate the potential benefits of incorporating APC-binding molecules in the tumor-control capabilities of the delivered neoepitopes, we intentionally opted for a suboptimal dose of 5 µg of pDNA. The dose selection was informed by prior research demonstrating that 50 µg of pDNA vaccine encoding Neo5 leads to nearly complete tumor abrogation in the CT26 tumor model (33), limiting our ability to evaluate improvements in the vaccine design.

In the first experiment, we screened APCt_Neo5 pDNA constructs with the following APC-binding molecules: CCL3, CCL4, CCL5, XCL1, and CCL19. As expected, NT_Neo5 failed to induce significant tumor rejection ( Figures 2A, B ). However, it elicited immunogenicity, as evidenced by the presence of neoepitope-reactive T cells in the spleen compartment ( Figures 2C, D ). On the contrary, most mice receiving an APC-targeting vaccine developed smaller tumors than the mock control, which received an empty DNA plasmid ( Figures 2A, B ). The constructs harboring CCL4 and CCL19 as APC-binding molecules were the most efficient, with CCL19_Neo5 leading to statistically significant tumor control compared to the non-targeted construct NT_Neo5. Efficient tumor control correlated with the induction of high frequencies of neoepitope-reactive T cells measured by the presence of T cells that simultaneously produce TNFα and IFNγ upon neopeptide re-stimulation. APCt_Neo5 pDNA encoding CCL4 and CCL19 induced in average twice as many neoepitope-reactive CD8+ (2.43% and 2.19% respectively) and CD4+ (0.27% and 0.29% respectively) T cells compared to NT_Neo5 (1.06% CD8+ and 0.15% CD4+ reactive T cells) ( Figures 2C, D ). However, given the relatively high standard deviation in these parameters, the results need to be interpreted with caution.

Tumor control, and induction of reactive CD8+ and CD4+ T cells, are among the main readouts for determining the efficacy of TCVs. We ranked the different DNA constructs according to these three parameters to evaluate the performance of the different APC-binding molecules. The results designated CCL19 as the most promising APC-binding molecule for delivering in silico-predicted cancer neoepitopes, closely followed by CCL4 ( Supplementary Table 3 top).

In a subsequent experiment, we compared the efficacy of CCL19, CCL20, CCL21, GM-CSF, Fv αDEC205, Fv αClec9a, and Clec9a ligand as APC-binding molecules. All APC-targeting pDNA vaccines conferred superior tumor control compared to the mock control ( Figures 2E, F ) with CCL19, CCL20, and CCL21_Neo5 displaying statistically significant improvements. CCL19_Neo5 induced the highest levels of neoepitope-reactive reactive CD8+ and CD4+ T cells (3.23% and 0.26%, respectively) ( Figures 2G, H ) in the spleen compartment at the termination of the study. The ranking of the APC-binding molecules tested in the second experiment based on improved tumor control, and induction of reactive CD8+ and CD4+ T cells, pointed to CCL19_Neo5 as the most efficient APC-targeting construct ( Supplementary Table 3 bottom).

Despite the potentially lower expression levels, CCL19 performed as the best APC-binding molecule when combined with 5 CT26-predicted neoepitopes and, therefore, was selected to further develop an APC-targeted TCV for cancer neoepitopes.

It is believed that, upon secretion of the APC-targeting construct, the APC-binding molecules interact with receptors in the surfaces of APCs, facilitating antigen internalization and presentation in MHC molecules and enhancing the immunogenicity of the antigens. This mechanism is possible because a covalent link connects both elements. In addition, APC-binding molecules such as chemokines can act as genetic adjuvants, modulating APC activation and recruitment to the vaccination site -processes independent of the covalent connection between the two elements-.

To address the implications of the covalent link between the APC-binding molecule CCL19 and the cancer neoepitopes in CCL19_Neo5, we designed a tumor experiment comparing the efficacy of immunizing with the fusion construct CCL19_Neo5 or with the combination of NT_Neo5 and a pDNA encoding CCL19 (pCCL19) ( Figure 3A ). In addition, we evaluated the contribution of the secretion signal by comparing the efficacy of co-delivering pCCL19 and NT_Neo5 and co-delivering pCCL19 and Neo5, which contains the same neoepitopes but not a secretion signal (33).

Co-administration of pCCL19 and NT_Neo5 and treatment with CCL19_Neo5 led to complete tumor prevention in most animals, hampering the analysis of the potential impact of the link between CCL19 and the neoepitopes in preventing tumor development. Similarly, co-delivery of pCCL19 and Neo5 achieved tumor control in most of the animals ( Figures 3C, D ). These data suggest that neither the secretion signal nor the covalent link between CCL19 and the neoepitopes are essential for the tumor control capabilities of the CCL19_Neo5 pDNA vaccine. Nevertheless, given the limitations of the current setup, the results do not rule out the influence of these two factors in tumor prevention.

MHC-I multimer staining of the blood three weeks after the first immunization showed that immunization with CCL19_Neo5, which harbors a covalent link connecting CCL19 and the cancer neoepitopes, led to an average of 3,61% C1-specific CD8+ T cells, a significant increase when compared to the levels obtained by the co-administration of pCCL19 and NT_Neo5 ( Figure 3E ). On the other hand, introducing a secretion signal does not affect the levels of neoepitope-specific CD8+ T cells, with both co-administration regimes rendering ~1.5% C1-specific CD8+ T cells.

Analysis of the functional T-cell responses upon the study’s termination showed comparable levels of cytokine-secreting CD8+ (~3.5%) and CD4+ (~0.2%) T cells between treatment with CCL19_Neo5, co-administration of pCCL19 and NT_Neo5 and co-administration of pCCL19 and Neo5 ( Supplementary Figures 1A, B ).

Together the data suggest that including a secretion signal to pDNA-encoded neoepitopes does not impact the vaccine’s efficacy, while the introduction of a covalent connection between CCL19 and the cancer neoepitopes might favor early induction of C1-specific CD8+ T cells. Nevertheless, the benefit of the covalent link is restricted to the onset of the immune response as analysis of the spleen compartment as the end of the study show similar immunogenicity for the APC-targeting vaccine and the co-delivery regime. Administration of pCCL19 alone did not induce tumor protection or specific T-cell responses, underlying the neoepitope-dependency of the results.

Next, we evaluated the role of the dimerization module h1h4CH3 in the CCL19_Neo5 pDNA construct for which we generated the monomeric pDNA construct CCL19_Neo5 monomer and the homodimeric construct CCL19_hMDH2_Neo5, where the dimerization sequence hMDH2 substitutes h1h4CH3 ( Figure 3B ). Replacing h1h4CH3 with hMDH2 did not reduce the efficacy of the APC-targeted neoepitope vaccine, as both constructs achieved complete tumor rejection in most animals ( Figures 3F, G ). Removing the dimerization unit (i.e., CCL19_Neo5 monomer) also resulted in good tumor rejection, with no significant differences observed when comparing this group to those immunized with dimeric constructs. With the current setup, we are unable to scrutinize in depth the potential impact of substituting or eliminating the h1h4CH3 dimerization domain.

When analyzing the presence of neoepitope-specific T cells in circulation three weeks after the first immunization, we observed that CL19_h1h4CH3_Neo5, and CCL19_hMHD2_Neo5 induced on average twice as high frequencies of C1-specific CD8+ T cells (1.79% and 2.22%) than the monomeric construct CCL19_Neo5 monomer (1.25%) ( Figure 3H ). However, no statistically significant differences could be measured due to the limited group size and the high variation of the data set. Analysis of the spleen compartment indicated similar levels of neoepitope-reactive T cells between CL19_h1h4CH3_Neo5, and CCL19_ Neo5 monomer with the group receiving CCL19_hMHD2_Neo5 displaying the highest levels of reactive T cells ( Supplementary Figure 1C ). The changes in the dimerization domain did not affect the expression levels of the plasmids ( Supplementary Figure 1D ).

Together, the data indicate that in the current setup, including a dimerization module, regardless of its nature, has no significant impact on the anti-tumor efficacy of APC-targeting vaccines. Still, it could contribute to an earlier onset of the specific immune response.

To bridge our preclinical results to a clinical setting for pDNA TCVs, we upgraded the backbone plasmid in CCL19_Neo5 from pUMVC4a to the more immunogenic pTVG4 (containing additional CpG motifs). The change in backbone plasmid did not alter the tumor control capability, which remained comparable for both pDNA designs ( Supplementary Figure 2A ). But we observed an improvement in the frequencies of C1 neoepitope-specific and cytokine-producing T-cell in the pTVG4-based construct ( Supplementary Figures 2B–D ).

Previous studies have demonstrated that delivering 13 neoepitopes, compared to 5, can prime the T cells to recognize and attack more targets, increasing the likelihood of tumor recognition and rejection (33). Following this strategy, we designed a pTVG4-based APC-targeted DNA construct containing 13 27-mer neoepitopes predicted from the CT26 cell line, called CCL19_Neo13, and manufactured it at clinical grade.

Change of the backbone plasmid to pTVG4 results in higher levels of CCL19_Neo5 in the supernatant of pDNA HEK293T-transfected cells. On the contrary, the introduction of additional neoepitopes has a detrimental effect on the expression level of CCL19_Neo13. Nevertheless, the CCL19_Neo13 fusion protein detected by immunoblot against the CH3 part of the protein, presents the expected molecular size, indicating that the neoepitopes are fully transcribed ( Supplementary Figure 2E ).

We evaluated the ability of the APC-targeted construct CCL19_Neo13 and the previously published, non-targeted pDNA design, Neo13 (33), to induce neoepitope-specific T cells and tumor control in a prophylactic setting at different pDNA doses. The results revealed a clear positive correlation between the dose of pDNA and the effectiveness of the treatment, determined by the frequency of circulating C1-specific CD8+ T cells and the ability to reject the tumors ( Figure 4 ). Neo13 prevents tumor development at 5 μg of pDNA. The effect is lost when the pDNA dose is reduced to 1 μg. CCL19_Neo13 at 1 μg pDNA achieved significantly lower tumor volumes than the non-targeted construct. Notably, at the lowest tested dose, 0.5 µg of pDNA, CCL19_Neo5 retained partial tumor control when compared to the mock control, with five out of 13 animals remaining tumor-free upon the termination of the study ( Figures 4A–C ).

Overall, CCL19_Neo13 proved five times more potent than non-targeted version Neo13. 1 µg of CCL19_Neo13 resulted in 10 out of 13 tumor-free animals, comparable with the effect obtained with 5 µg of Neo13, where 9 out of 13 animals remained tumor-free. The frequency of C1-neoepitope-specific CD8+ T cells in circulation was comparable between the CCL19_Neo13 (1 µg) and Neo13 (5 µg) groups. CCL19_Neo13 at 0.5 µg retained immunogenicity evident by the presence of C1-neoepitope specific CD8+ T cells ( Figure 4D ).

In a separate experiment we interrogated the T-cell responses induced in the spleen compartment by 5 µg of CCL19_Neo13 and Neo13. We found significant differences in the frequencies of cytokine secreting CD8+ and CD4+ T cells between these two groups. CCL19_Neo13 induced an average of 2.5% and 0.36% reactive CD8+ and CD4+ T cells respectively, four times more than Neo13 ( Supplementary Figure 3 ).

In a subsequent immunogenicity study, BALB/c mice were administered increasing doses of CCL19_Neo13, ranging from 5 to 500 µg of pDNA. At lower doses of CCL19_Neo13 pDNA, vaccine-induced T-cell responses correlate with the pDNA dose ( Supplementary Figure 4 ). Thirteen days after the first immunization, 50 µg of pDNA vaccine induced over three times higher frequencies of C1-specific T-cell responses than 5 µg ( Supplementary Figure 3A ). The measured immune response plateaued at 50-100 µg, beyond which increasing the pDNA dose resulted only in marginal improvements in the frequencies of neoepitope-specific and reactive T cells.

To evaluate the duration of the vaccine-induced T-cell responses, we subjected groups of mice to one or four immunizations with 100 µg of CCL19_Neo13. C1-neoepitope-specific CD8+ T cells increased up to 20 days after the last immunization (study day 41) in both groups, representing 2.1% of the total CD8+ T cells in mice receiving four immunizations, and 1.2% in the group receiving one immunization. After study day 41, C1-neoepitope-specific T cells decreased over time, persisting at 0.8% and 0.28%, respectively, 105 days after the initiation of the study ( Supplementary Figure 5 ).

Next, we investigated if immunizing with CCL19_Neo13 post-tumor inoculation could achieve tumor control in the CT26 tumor model.

The fast kinetics of tumor growth in the syngeneic CT26 model provide a short time interval for evaluating the effectiveness of therapeutic vaccination regimes. To induce more rapid and robust cellular immune responses, we utilized EP-assisted pDNA vaccination. In a previous immunogenicity study, EP-assisted immunization with the CCL19_Neo13 DNA construct yielded stronger neoepitope-specific immune responses than poloxamer-formulated pDNA ( Supplementary Figure 6 ).

To investigate the efficacy of APC-targeting DNA vaccines post-tumor inoculation, one day after s.c. inoculation of CT26 cells, mice received five EP-assisted immunizations of 50 µg of CCL19_Neo13 spaced two to three days apart. Mice immunized with the APC-targeting DNA vaccine exhibited partial tumor control with generally lower end-tumor volume ( Figures 5A–C ). Kaplan-Meier curves depicting tumor-free mice confirmed superior tumor control driven by CCL19_Neo13 with complete tumor abrogation in 5 out of 13 mice at the end of the study ( Figure 5D ). Multimer analysis of the blood nine days after the first immunization confirmed the presence of 0.5% C1-neoepitope-specific CD8+ T cells in mice receiving CCL19_Neo13 ( Figure 5E ). On day 21, upon the termination of the study, 8% of the CD8+ and 0.3% of the CD4+ T cells responded to neopeptide re-stimulation, producing both IFNγ and TNFα cytokines ( Figures 5F, G ). Further analysis of the neoepitope recognition profile of CCL19_Neo5 showed that T-cell reactivity is driven by several neoepitopes being C1, C2, C10 and C12 the predominant contributors to the vaccine-specific T-cell repertoire ( Figure 5H ).