The cell lines used, including 624.Mel 38, SK.Mel 37, LB373.Mel, MZ2.Mel 43, K562 and BV173, and their maintenance are listed in the Supplementary Table S1 . K562-A2 cell line was generated by transducing K562 cells with HLA-A2 cDNA. The ABL-minigene-E255V-expressing variants and the SPTG4-expressing variants were generated by transducing cancer cells with ABL-minigene-E255V or STPG4 cDNA linked to GFP or mCherry by an internal ribosomal entry site (IRES), where ABL-minigene-E255V cassette was constructed of a DNA fragment of a size of 1045 bp (position 19-1063 of ABL open reading frame, ORF) comprising the E255V mutation. The mutated CML cell clones harboring the E255V⁺ BCR-ABL gene were generated by CRISPR-Cas9 system. All cell culture reagents were purchased from Life Technologies unless otherwise indicated.

For priming, ABab-A2 mice were injected subcutaneously with 100 nmol of peptide, ABL-E255V (KLGGGQYG V , Genscript) or ABL-T315I (YII I EFMTYG, Genscript), in a 200 μl 1:1 solution of incomplete Freund’s adjuvant and PBS supplemented with 50 μg CpG. Sequential boosts were performed in 4 weeks interval. The presence of responsive CD8⁺ T cells in the peripheral blood was assessed by in vitro peptide stimulation and subsequent intracellular cytokine staining by BD cytofix/Cytoperm solution kit (BD biosciences) 7 days after each boost. All animal experiments were performed according to institutional and national guidelines and regulations after approval by the responsible authority (Landesamt für Gesundheit und Soziales, Berlin).

Splenocytes from responding mice were prepared at day 10 after the last boost and cultured with 10^-9 M peptide for 7 days. Responsive T cells were labeled by IFNγ secretion assay (Milteny) and sorted directly into RLT Plus lysis buffer (Qiagen) by flow cytometry after peptide (10^-6 M) stimulation for 3 hr. Total RNA from sorted cells was extracted (RNeasy Plus Micro kit, Qiagen), and first-strand cDNA synthesis and 5’-RACE PCR was performed using SMARTer™ RACE cDNA amplification kit (Clontech), where TCR-specific amplification was carried out with following primers: hTRAC (5’-CGGCCACTTTCAGGAGGAGGATTCGGA AC-3’) or hTRBC (5’-CCGTAGAACTGGACTTGACAGCGGAAGTGG-3’). The RACE PCR products were cloned using a Zero Blunt TOPO PCR Cloning Kit (Life Technologies). Plasmids from individual clones were isolated and sequenced using the T3 primer (5’-AATTAACCC TCACTAAAGGG-3’) at Eurofins Genomics.

All TCR transgene cassettes were codon-optimized and synthesized by GeneArt (Life Technologies). Pairing TCR α and β chains were linked with the porcine teschovirus-1 derived self-cleavage peptide P2A, in which the human TCR constant regions were either replaced by murine constant regions or minimally murinized by introducing individual amino acids from the mouse counterparts and an additional cysteine bridge (T9141-mmc) (21–24). The transgenes were cloned into pMP71 vector using NotI and EcoRI restriction sites (25).

Transduction was carried out as described (26). Packaging HEK-GALV cells (HEK-293 cells expressing stably GALV-env and MLV-gag/pol) were transfected with transgene in the presence of Lipofectamine2000 (Life Technologies). Retrovirus-containing supernatant was harvested 48 and 72 hours after transfection. Human PBMCs were activated by anti-CD3 (OKT3, BD Pharmingen) and anti-CD28 (CD28.2, BD Pharmingen) antibodies in the presence of 300 U/ml hIL-2 (PeproTech) and transduced 48 and 72 hours after T cell activation by spinoculation with virus supernatant for 90 min or with virus-preloaded retronectin (Takara)-coated plates for 30 min at 800 g and 32°C. Transduced T cells were expanded for 1 week in the presence of 300 U/ml h IL-2 and rested for 3 days with 30 U/ml hIL-2 before being used for experiments.

The two CRISPR RNA (crRNA), crRNA3 (5’-AGGTCTTCACGGCCACCGTCAGG-3’) and crRNA6 (5’-CAGTACGGGGAGGTGTACGAGGG-3’), targeting the ABL gene were designed using CRISPOR program (27). The ribonucleoprotein (RNP) for genome editing was generated following the manufacturer’s instruction of Integrated DNA Technologies (IDT). To prepare the delivery of RNP, a total of 3×10⁵ cells were resuspended in 20 µl supplemented Nucleofector solution R (Lonza Bioscience) and mixed with 5 µl RNP, 1 µl of 100 µM single strand oligo DNA (ssODN) and 1 µl of 100 µM electroporation enhancer (IDT). The mixture was electroporated by using DN-100 program of 4D-Nucleofector device (Lonza Bioscience). The electroporated cells were immediately transferred to pre-warmed conditioned medium and cultured for expansion and selection. All CRISPR-Cas9 components were purchased from IDT. The ssODNs and the primers used for screening and sequencing are listed in Supplementary Tables S2 and S3 (Eurofins Genomics).

All co-culture experiments were performed by incubating T cells and target cells for 16-18 hours at an effector-to-target (E:T) ratio of 1:1 (1×10⁴ cells of each), unless otherwise indicated. T cells cultured with 50 ng/ml Phorbol-12-myristate 13-acetate (PMA) and 5 µg/ml Ionomycin served as positive control. Human IFNγ secretion in the supernatant was measured by ELISA (BD OptEIA).

Total RNA was extracted using Quick-RNA™ miniprep kit (Zymo Research) and reverse transcribed using ProtoScript^® II First Strand cDNA synthesis Kit (BioLabs). The genes selected for Taqman^® gene expression assay (ThermoFisher) were BCR-ABL (b2a2, Hs03024541_ft) and ABL (exon 3-4, Hs00245445_m1). ABL (exon 8-9, Hs01104728_m1) served as an endogenous control. The qPCR was performed using QuantStudio3 Real-Time PCR system (ThermoFisher). The results were analyzed according to the comparative 2^-ΔΔCt Method.

The following conjugated antibodies were used at 1:100 dilution and purchased from BioLegend: anti-hCD3 (HIT3α), anti-hCD8 (SK1), anti-hHLA-A2 (BB7.2), anti-mCD3 (145-2C11), anti-mCD8 (53.6-7), anti-mTCRβ (H57-597), anti-mIFNγ (XMG1.2). HLA-A2/ABL-E255V dextramer (Immudex) staining was performed according to the manufacturer’s instruction. Cells were analyzed by BD FACSCanto™ II (BD Biosciences) or MacsQuant (Miltenyi Biotech). Cell sorting was performed with BD FACSAria™ II (BD Biosciences). FACS data analysis was performed using FlowJo (TreeStar).

All statistical analyses and data display were performed using GraphPad Prism version 9.4.0. The number of independent experiments, biological and technical replicates, and statistical tests used are indicated in the relevant figure legends.

To discover potential neoepitopes generated from mutated BCR-ABL, we evaluated frequently occurring mutations in silico with NetMHC 4.0 and IEDB (28–30). Two candidate peptides, ABL-E255V (KLGGGQYG V ) and ABL-T315I (YII I EFMTYG), were selected for short peptide immunization ( Table 1 ). The parental sequences of the two candidates were confirmed to be identical between human and mouse, ensuring that the imatinib-induced somatic mutation is the only foreign amino acid in the mice ( Figure 1A ). ABab-A2 mice expressing a diverse human TCR repertoire and HLA-A2 molecules were boosted in four weeks interval. The response of each immunized mouse was tested 7 days after each boost by intracellular cytokine staining following ex vivo peptide re-stimulation of blood. As a result, specific CD8⁺ T cell responses were observed in several ABL-E255V-immunized mice, whereas no response was found in ABL-T315I-immunized mice ( Figures 1B, C ).

Upon detection of CD8⁺ T cell responses against ABL-E255V, splenocytes from responding mice were prepared and cultured to expand specific CD8⁺ T cells. ABL-E255V-responsive CD8⁺ T cells were labelled by the mouse IFNγ secretion kit and isolated by flow cytometry ( Figure 1D ). Identified TCRα and β chains were matched and linked by P2A element for equimolar expression ( Figure 2A ). Of note, human TCR constant regions were replaced with murine constant regions, allowing preferential pairing and avoiding mispairing with endogenous TCRs (31). Finally, two TCRs, T9141 and T8922, were identified from two individual ABab-A2 mice ( Table 2 ).

Human T cells were retroviral transduced to re-express the paired TCRs ( Figure 2B ). To verify the specificity of isolated TCRs, TCR-transduced T cells were co-cultured with T2 cells loaded with ABL-E255V or parental peptide ABL-E255WT (KLGGGQYG E ), and the release of IFNγ was used as an indicator of T cell activation. As a result, both isolated TCRs specifically recognized ABL-E255V ( Figure 2C ). Next, we investigated their functionality by co-culturing TCR-transduced T cells with T2 cells loaded with titrated ABL-E255V peptide. T9141-TCR responded to ABL-E255V up to 10^-11 M with an EC₅₀ of 10^-10 M, while T8922-TCR responded up to 10^-10 M with an EC₅₀ of 10^-9 M ( Figures 2D, E ). For clinical purposes, T9141-TCR with superior avidity was cloned with minimally modified human constant regions (T9141-mmc-TCR) for testing (22). T9141-mmc-TCR showed comparable TCR expression by dextramer (HLA-A2/ABL-E255V) staining ( Figures 2F, G ) and achieved similar IFNγ production in response to titrated ABL-E255V peptide ( Figures 2H, I ), suggesting that the expression and functionality of the TCR were not affected by the mmc-constant region.

To verify the expression of ABL-E255V in human cells and the recognition by specific TCRs, we co-cultured TCR-transduced T cells with cancer cells transduced with ABL-minigene-E255V. This minigene is an approximately 1 kb DNA fragment derived from ABL gene exon 1 to exon 6 with the point mutation E255V in exon 4 ( Figure 3A ). As determined by GFP expression, the three transduced melanoma cell lines, 624.Mel 38, SK.Mel 37 and LB373.Mel, expressed comparable levels of ABL-minigene-E255V, whereas the transduced CML cell line, BV173, had a lower level of expression ( Figure 3B ). The HLA-A2^- melanoma cell line, MZ2.Mel 43, was included as control.

Both isolated TCRs recognized HLA-A2⁺/ABL-minigene-E255V⁺ cells ( Figure 3C ). T9141-TCR exhibited significant responses, despite a weaker response to transduced BV173 cells. This result was consistent with the expression level of minigene detected in the transduced cell lines. In contrast, T8922-TCR showed a robust response to transduced 624.Mel 38 cells, but exhibited significantly reduced responses to transduced SK.Mel 37 and LB373.Mel cells, and unexpectedly failed to recognize transduced BV173 cells, suggesting that T8922-TCR has a too low avidity. On the other hand, these results showed that the transduced cell lines expressed different levels of ABL-E255V despite comparable levels of minigene and HLA-A2 expression, indicating that the efficiency of peptide processing varies among cell lines ( Figures 3B, D ). In summary, ABL-E255V is processed and presented in human cells and is recognized by ABL-E255V-specific TCRs.

To prevent unforeseen immune responses, T9141-TCR was tested for off-target toxicity. As ABab-A2 mice express HLA-A2 but no other human HLA alleles, we analyzed allo-reactivity of T9141-TCR by co-culturing TCR-transduced T cells with a panel of B lymphoblastoid cell lines (B-LCLs) expressing different HLA molecules ( Supplementary Table S4 ). No allo-reactivity was detected ( Figure 4A ).

As T9141-TCR was selected in the murine thymus, it may respond non-specifically to certain human peptides. First, we analyzed the cross-reactivity of T9141-TCR with 114 known HLA-A2-restricted human self-peptides loaded on T2 cells at 10^-5 M (32). No cross-reactivity was detected ( Figure 4B ). Next, we analyzed the recognition motif of T9141-TCR by measuring its response to ABL-E255V mutated sequentially to alanine at each position ( Supplementary Table S5 ). Position 2, 5, 6, and 7 (x-L-x-x-G-Q-Y-x-x) were identified as the TCR binding motif ( Figure 4C ). Of 36 human peptides containing this motif but which were absent in mice ( Supplementary Table S6 ), eight peptides were evaluated as candidates for cross-reactivity ( Figure 4D ). Here, peptides eliciting responses below 250 pg/ml of IFNγ release were neglected, because peptides that are poorly recognized at 10^-5 M are unlikely to elicit responses at physiological levels. The 8 peptides were investigated by titration assay, of which one peptide derived from the STPG4 gene (sperm-tail PG-rich repeat containing 4) was recognized by T9141-TCR at a concentration as low as 10^-8 M ( Figure 4E ). We further co-cultured T9141-transduced T cells with full-length STPG4-transduced cancer cells ( Supplementary Figures S1A, B ). No recognition was detected ( Figure 4F ). Importantly, T9141-transduced T cells responded poorly to all cell lines externally loaded with STPG4, except T2 cells, even at 10^-4 M, suggesting that large amounts of STPG4 are needed to compete with endogenous peptides for HLA-A2. In other words, STPG4 peptide is unlikely to cause cross-reactivity, as it cannot be expressed at such high levels under normal physiological conditions. Taken together, no relevant cross-reactivity of T9141-TCR was detected.

The processing and presentation of ABL-E255V has been demonstrated by ABL-minigene-E255V-transduced cancer cells ( Figure 3 ). However, it was important to investigate mutated CML cells that acquire the E255V mutation in the endogenous BCR-ABL gene, as the overexpressed minigene may not accurately reflect the natural scenario. Due to the lack of CML cell lines that are HLA-A2⁺/E255V⁺, we performed CRISPR-Cas9 to introduce the E255V mutation into BV173 cells. After screening and sequencing ( Supplementary Figure S2 ), three mutated BV173 clones were obtained ( Figure 5A ). We then co-cultured T9141-transduced T cells with the three mutated BV173 clones. Unexpectedly, T cell responses were elicited only when the mutated BV173 clones were externally loaded with ABL-E255V peptide ( Figure 5B ). To assess the functionality of the clones’ peptide processing machinery, the mutated BV173 clones were transduced with ABL-minigene-E255V for co-culture ( Figures 5B, C ). Although all minigene-transduced cell samples elicited specific responses, the mutated BV173 clones on average induced lower IFNγ production than the parental cells, suggesting that the mutated BV173 clones had a less efficient processing machinery after a series of manipulations and long-term culture. As BCR-ABL is required for neoplastic transformation and survival of leukemic cells in CML, its overexpression is frequently observed (33–36). To compare the expression of the BCR-ABL gene and ABL-minigene-E255V in each BV173 sample, we performed q-PCR. In contrast to the BCR-ABL gene, the expression of ABL-minigene-E255V was 20-fold higher ( Figure 5D ). These results may explain why the mutated BV173 clones were not recognized, since they expressed the BCR-ABL gene at much lower levels than the minigene and had a less efficient processing machinery, thereby presenting insufficient levels of ABL-E255V for T cell recognition.

As CML cell lines can express varying levels of BCR-ABL and have different processing efficiencies, we included an additional CML cell line, K562, transduced to express HLA-A2 and edited using CRISPR-Cas9. Through screening and sequencing, the K562-A2 clone 1 was determined to carry the E255V mutation in the BCR-ABL gene ( Figure 6A ). After co-cultured with T9141-transduced T cells, this mutated clone induced a specific, albeit weak T cell response ( Figure 6B ). Nonetheless, T cell response to the mutated K562-A2 clone was significantly improved by using a higher number of transduced T cells and by adjusting the ratio of effector cells to target cells to 5:1 ( Figure 6C ). These results demonstrate that ABL-E225V derived from the mutated BCR-ABL gene can be presented in HLA-A2⁺ CML cells. Additionally, the mutated K562-A2 clone that underwent CRISPR-Cas9-editing and screening exhibited a well-functional peptide processing machinery, as the mutated clone transduced with ABL-minigene-E255V elicited a T cell response comparable to transduced parental cells ( Figures 6B, D ). And K562-A2 cells were also verified to express similar levels of HLA-A2 as cell lines that naturally express HLA-A2 ( Figures 3D , 6E ). In summary, T9141-transduced T cells can recognize HLA-A2⁺/E255V⁺ CML cells. Although the response is highly dependent on the BCR-ABL expression in CML cells, it can be improved by optimizing the number of TCR-transduced T cells.