A set of 880 K^d-restricted peptides had been detected across rejecting B6.K^d skin grafts, B6.K^d spleen cells and C57BL/6 hepatocytes transduced with AAV-K^d (2). From this set, 100 peptides were chosen for further evaluation, based on abundance (as estimated by spectral intensity) and predicted IC50 ≤ 500 nM [Net MHC pan 4.1 (10), Table 1]. MHC haplotypes of donor and recipient strains are shown in Figure 1A. Liver leukocyte populations enriched for activated alloreactive T cells were generated following priming of C57BL/6 mice with a B6.K^d skin graft, and boosting by inoculation with AAV-K^d (Figure 1B). On d7 after inoculation, liver leukocytes were isolated and stained using each of 100 individual K^d-peptide tetramers to determine the proportion of CD8⁺ T cells recognising each epitope. (Figure 1B). The gating strategy is shown in Figure 1C. Alloreactive CD8⁺ T cells are CD44⁺ and PD-1^hi, whereas bystander cells are PD-1⁻. Tetramer binding results are presented in Table 1 and Figure 1D,E. Of the 100 peptides screened, 22 were recognised by ≥5.0% of activated alloreactive T cells, with a ratio of ≥5 between binding of activated and bystander cells, and these were designated as highly immunogenic.

A further 34 peptides were recognised by between 2.0 and 4.99% of activated CD8⁺ cells (moderately immunogenic), while 43 peptides were recognised by ≤1.99% of activated CD8⁺ cells (non-immunogenic). These proportions imply significant cross-reactivity in recognition of different pMHC species by alloreactive T cells. Representative examples of tetramer staining for each of these groups are shown in Figure 1D, while the summary of tetramer binding results appears in Figure 1E. One peptide, KFIATLQYI, was consistently recognised by both activated (9.63%) and bystander cells (4.81%). Eight peptides were selected for inclusion into a panel for detection of alloreactive CD8⁺ T cells (Figure 1F,G). This panel detected >30% of activated CD8⁺ T cells expanded using the same prime-boost strategy outlined above.

Similar to our earlier findings for K^b-bound peptides (2), the strongest predictor of immunogenicity for the peptide cargo of K^d was peptide abundance, here estimated as the product of the peptide spectral intensity across rejecting skin grafts, transduced hepatocytes and spleen (Figure 2A). Both abundance and T cell binding values for K^d-SYFPEITHI were considerably higher than for the remaining peptides. Robustness of the correlation was tested by a) converting numerical values to ranks and using a non-parametric test (Spearman's correlation; Supplementary Figure S1A) and b) removing Kd-SYFPEITHI. In both cases, significant correlations were maintained. The binding affinity (BA) score from NetMHC pan 4.1 (10) also correlated with immunogenicity (Figure 2B), as did the overall peptide hydrophobicity as measured by the Kyte-Doolittle scale (Figure 2C) (11). As for abundance, these correlations remained significant when Spearman's test was employed. Hydrophobicity at P3 or P6 was associated with increased immunogenicity, while the opposite was true of hydrophobicity at P7 (Figure 2D, Supplementary Figure S1B). Hydrophobicity at other non-anchor positions was not significantly associated with immunogenicity (Supplementary Figure S1B). Multivariable analysis showed that peptide abundance, BA score and overall hydrophobicity were independent of each other and all three were significantly associated with peptide recognition by activated alloreactive CD8⁺ T cells (Figure 2E).

In contrast to these findings, there was no significant correlation between the IEDB immunogenicity score (12) and the proportion of CD8⁺ T cells recognising a given peptide presented by H-2K^d (Figure 2F). H-2K^b-presented peptides are shorter than those favoured by H-2K^d, and biochemically distinct, with an unusual binding motif featuring an aromatic residue at peptide position (P)5. In this instance, neither the overall hydrophobicity (Supplementary Figure S1C) nor the hydrophobicity score for any individual residue, was correlated with immunogenicity. The IEDB immunogenicity score, designed for 9-mers, was not applied to the H-2K^b 8-mer peptide dataset.

To determine whether our core 8-tetramer panel would be able to detect alloreactive T cells responding to a different target tissue in the setting of a full allogeneic mismatch, BALB/c hearts were transplanted into C57BL/6 recipients (Figure 3A). The haplotypes of this donor-recipient strain combination are shown in Figure 3B, while Figure 3C depicts the gating strategy. Heart-infiltrating leukocytes, splenocytes and lymphocytes from the parathymic (draining) nodes (DLN) were isolated on day 7 post-transplantation and stained with the 8-tetramer panel described above (Figure 3D). 9.51 ± 1.03, 6.38 ± 2.14 and 7.73 ± 0.89% of CD8⁺ T cells from rejecting heart, spleen and DLN respectively bound the 8-tetramer panel, compared with 3.59 ± 0.68, 2.68 ± 0.56 and 2.58 ± 0.25% binding a single tetramer of the dominant epitope K^d-SYFPEITHI. Binding to the control syngeneic tetramer of K^b loaded with the abundant self-peptide SNYLFTKL was less than 0.75% for CD8⁺ T cells isolated from each tissue (Figure 3E).

Tetramer-positive proportions were consistently higher in activated T cells than in non-activated cells. In the graft infiltrate, (Supplementary Figure S2) 93.1% of CD8⁺ T cells were activated (CD44 ⁺ PD-1^hi); of these, 10.07 ± 1.1% bound the tetramer panel, compared with 1.25 ± 0.14% of cells that were PD-1⁻. Conversely, in the DLN, the majority of cells were naïve (64.1% CD44⁻), with 6.45% CD44⁺ PD-1^hi (Supplementary Figure S3). The proportion of tetramer-positive activated cells was very high at 31.7 ± 4.08% with 4.14 ± 0.82% of naïve cells binding the tetramer panel (Supplementary Figure S3). Roughly equal numbers of spleen CD8⁺ T cells were found in the activated and naïve compartments (Supplementary Figure S4). In the spleen, tetramer-positive fractions were 12.99% ± 3.06 and 3.16 ± 1.74% of activated and naïve cells respectively.

Peptides used in this study were synthesised with an average of 98% purity (GL Biochem Shanghai Ltd.). 10% DMSO was used in reconstituting lyophilised peptides and aliquots were stored at −80°C. 5 mM stock of Dasatinib (Sigma-Aldrich, catalogue# CDS023389) reconstituted in DMSO was stored at −80°C. Primary and secondary antibodies used in this study are summarised in Supplementary Table S1.

C57BL/6J^Arc (H-2^b) and BALB/c^Arc (H-2^d) mice (herein termed C57BL/6 and BALB/c) were purchased from the Animal Resources Centre, Perth, Australia. B6.Kd mice (20, 21) express an H-2K^d transgene ubiquitously on a C57BL/6 (H-2^b) background, and were bred at Australian Bioresources, Moss Vale, Australia. B6.Kd mice were backcrossed for 4 generations to C57BL/6J^Arc, prior to use. 8–12 week old male mice were used in this study.

Skin transplantation was performed as outlined previously (21). C57BL/6 recipient mice received full-thickness grafts of 1 × 1 cm tail skin from B6.Kd donor mice. Grafts were deemed rejected when less than 20% of the viable skin graft remained.

AAV vectors encoding H-2K^d were prepared as previously described (21). Briefly, H-2K^d cDNA was cloned into the pAM2AA backbone incorporating the liver-specific human α-1 antitrypsin promoter and human ApoE enhancer flanked by AAV2 inverted terminal repeats. This construct was packaged into an AAV2/8 vector, purified, and quantitated by the Vector and Genome Engineering Facility, Children's Medical Research Institute, Westmead, Australia. Vector aliquots were stored at −80°C. AAV-K^d was used at a dose of 5 × 10¹¹ vgc/mouse.

The heart transplant procedure was carried out according to a published protocol (22). Donor hearts were perfused with 1.0 ml of cold heparinised saline through the inferior vena cava (IVC). The donor heart was retrieved and then stored in cold saline on ice prior to transplantation. The donor aorta and pulmonary artery were anastomosed end to side to the recipient aorta and IVC, respectively. Following the release of cross-clamps, haemostasis was achieved using gentle pressure with a cotton bud. C57BL/6 mice (H-2^b) received fully mismatched allogeneic heart grafts from BALB/c donor mice (H-2^d).

For liver leukocyte isolation, the IVC was cannulated and the hepatic portal vein was transected. The liver was flushed with 20 ml of PBS at RT and after gallbladder removal, the liver was mashed through a 100 μm cell strainer and washed through with cold RPMI-1640 medium supplemented with L-glutamine (Lonza, catalogue# 12-702F) and 2% FCS (Sigma-Aldrich, catalogue# 13K179) (RMPI/FCS2 medium). The liver slurry was centrifuged at 400 g for 10 min and washed twice, then purified using isotonic Percoll PLUS (GE Healthcare Life Sciences) gradient separation - centrifuged at 500 g for 15 min at RT. The liver leukocyte pellet was collected, washed twice, and resuspended in red cell lysis buffer for 2 min at RT. Cells were used following two further washes with PBS containing 2% FCS.

For isolation of heart infiltrating leukocytes, the heart grafts were flushed by injecting 3 ml of cold PBS slowly into the abdominal aorta. The grafts were resected and stored in cold DMEM (Lonza, catalogue# BW12-709F). Grafts were washed with cold PBS and then transected lengthways. The bisected heart grafts were washed again with cold PBS. 500 μl of DMEM containing 400 U/ml collagenase Type II (Gibco, catalogue# 17101015) and 48 U/ml DNase I (Thermo Scientific, catalogue# EN0525) was injected directly into the parenchyma of the bisected heart grafts at multiple sites. Hearts were incubated for 10 min at 37°C with gentle shaking.

Following this, the grafts were diced and incubated further with 4.5 ml of DMEM containing collagenase Type II and 48 U/ml DNase I for 1 h at 37°C with gentle shaking. Digested heart tissue was gently pushed through a 70 μm nylon mesh strainer. Dissociated cells were washed with DMEM medium containing 15% FCS (DMEM/FCS15). The cells were centrifuged at 300 g for 3 min and then resuspended in 42% isotonic Percoll PLUS (GE Healthcare Life Sciences)/PBS. The resuspended cells were centrifuged at 800 g for 20 min at RT. The supernatant was discarded, and the heart infiltrating leukocyte pellet was then resuspended in red cell lysis buffer for 2 min at RT. Cells were used following two further washes with PBS containing 2% FCS. This method was adapted from Prosser et al. (23).

To obtain splenocytes, the spleen was pressed through a 70 μm nylon mesh strainer, washed and resuspended in red cell lysis buffer for 2 min at RT. The splenocytes were washed twice before use. To isolate lymphocytes from draining lymph nodes, the nodes were ruptured through a 40 μm nylon mesh strainer and then prepared as for splenocytes, with the omission of the red cell lysis step.

BioLegend Flex-T™ H2-K^d Monomer UVX (BioLegend, custom order) kits were utilised to generate biotinylated K^d monomers incorporating the selected peptides. These monomers were then assembled with PE-streptavidin (BioLegend, catalogue# 405203). 20 μl of selected peptide at 400 μM concentration was added to 20 μl of Flex-T monomer. The mixture was illuminated with long-wave 366 nm UV light on ice for 30 min, then incubated at 37°C for 30 min in the dark. The final exchanged monomer solution is at 2 μM concentration.

To assemble PE-conjugated tetramers, 2.0 μl of PE-streptavidin at 0.2 mg/ml concentration (BioLegend, catalogue# 405203) was added to 18 μl of the exchanged monomer. The reaction mixture was incubated at 4°C for 30 min protected from light. Blocking solution was prepared by mixing 1.6 μl of 50 mM D-Biotin (Sigma-Aldrich, catalogue# 711610) and 6 μl of 10% (w/v) sodium azide (Sigma-Aldrich, catalogue# S2002-5G) in 192.4 μl of PBS. 1.4 μl of the blocking solution was added to the reaction mixture and incubated at 4°C for at least 16 h protected from light. 5.8 μl of the PE-conjugated tetramer (approximately 0.5 μg) was used for each single test.

The pMHC multimer staining method was adapted from Dolton et al. (24). Cells were first incubated with a protein kinase inhibitor, 50 nM dasatinib, in staining buffer (2% FCS in PBS) for 30 min at 37°C. PE-conjugated tetramers were centrifuged at 16,000 g for 1 min to remove aggregates. Cells were stained with 0.5 μg of pMHC tetramer in 50 μl for 30 min at 4°C. Following pMHC multimer staining, the cells were washed with cold staining buffer twice. Samples were incubated with mouse Fc Block (BD Biosciences, catalogue# 553141) for 10 min at 4°C and mouse anti-PE antibody was added at 0.5 μg/100 μl. The cells were washed and incubated with a cocktail of antibodies against surface markers for 30 min on ice (Supplementary Table S1). Cells were washed twice with PBS before staining with viability dyes Zombie NIR (BioLegend, catalogue# 423105) for 15 min at RT. Cells were then washed with staining buffer. Sample data was acquired using either of the Fortessa X-20 or LSR-II (both BD Biosciences) instruments and analysed using FlowJo v10.

Data are represented as mean ± SEM unless otherwise stated. Unpaired Student's t-tests were performed to calculate statistical differences in a single variable between the means of two groups. The relationships between overall peptide abundance, binding affinity score or hydrophobicity score and alloreactive T cell binding were analysed using linear regression and Pearson correlation tests. Spearman's correlation was performed as an additional test for robustness. P < 0.05 was considered significant. Multivariable analysis using least squares regression was then carried out. Statistical tests were performed using GraphPad Prism version 8.01 (GraphPad Software, La Jolla CA).

All animal procedures were approved by the University of Sydney Animal Ethics Committee (protocol 2017/1253 and protocol 2022/2092) and carried out in accordance with the Australian code for the care and use of animals for scientific purposes. The use of genetically-modified mice was covered by University of Sydney Institutional Biosafety Committee approvals 18NO13 and 23NO23, while AAV vector production and use was approved under NLRDs 17NO28 and 22NO12.