SF3B1^K700E neoantigen is a CD8⁺ T-cell target shared across human myeloid neoplasms

Abstract

Acquired mutations in spliceosome genes in early hematopoietic stem/progenitor cells are common events in myelodysplastic neoplasms (MDS) and related myeloid malignancies. Mutations in the spliceosome factor subunit B1 (SF3B1) gene occur in ≥20% of MDS cases at conserved hotspots and in early neoplastic clones as driver events. Neoantigens from aberrant SF3B1 proteins could serve as shared T-cell therapy targets for SF3B1-mutated myeloid neoplasms. We identified a candidate neoantigen from the prevalent SF3B1^K700E variant using in silico predictions of epitope processing and presentation, then validated presentation and immunogenicity in vitro. CD8⁺ T cells recognizing SF3B1^K700E demonstrated high functional avidity and killed neoplastic myeloid cell lines and primary cells in an antigen-specific manner. We then sequenced, cloned, and transduced a SF3B1^K700E-specific T-cell receptor (TCR) into 3^rd-party T cells and confirmed that TCR transfer conferred antigen specificity and killing of neoplastic myeloid cells in vitro and in vivo. The data indicate that the SF3B1^K700E neoantigen represents a promising T-cell target for patients with SF3B1-mutated MDS and acute myeloid leukemia.

Introduction

The spliceosome factor subunit B1 (SF3B1) gene is recurrently mutated in hematologic neoplasms (1, 2) and, in fact, SF3B1 mutations define a unique subtype of myelodysplastic neoplasms (MDS) (3, 4). MDS are clonal bone marrow disorders that result in ineffective hematopoiesis, causing cytopenias and risk of transformation to secondary acute myeloid leukemia (sAML), especially in patients with intermediate or higher risk features (5). In MDS, mutations in SF3B1 appear to arise in primitive hematopoietic progenitors and play a disease-initiating role (reviewed in (3)), making mutated SF3B1 an attractive therapeutic target that is unlikely to be lost through deletion or transcriptional repression. The most prevalent mutation in SF3B1 leads to the substitution of a glutamic acid for the lysine residue at position 700 in the SF3B1 protein (SF3B1^K700E) and occurs in ~50% of patients with SF3B1-mutated MDS (1), making this particular mutation a potential shared therapeutic target. Because MDS is susceptible to T cell-mediated killing (6, 7), T cells targeting SF3B1^K700E neoantigens could provide an effective, potentially curative treatment for patients whose disease harbors this mutation.

We previously identified MDS neoantigens created from a recurrent mutation in another spliceosome gene, U2AF1, and demonstrated that U2AF1 neoantigen–specific T cells killed neoplastic cells but did not recognize normal hematopoietic cells (8). Using a similar reverse-immunology approach, we have discovered a nonameric peptide that is processed from endogenous SF3B1^K700E protein and presented on human leukocyte antigen (HLA)-B*40:01, and isolated a T-cell clone with high avidity for the SF3B1^K700E epitope. We sequenced, modified, and transferred the T-cell receptor (TCR) from this clone into 3^rd-party T cells. The resulting SF3B1^K700E-specific TCR–expressing T cells (TCR-T) efficiently eliminated cell lines presenting the neoantigen and controlled disease in a patient-derived xenograft (PDX) model. These results suggest that SF3B1^K700E could be used as a target in neoantigen-directed precision medicine approaches for MDS and sAML, including TCR-T cell therapy.

Materials and Methods

Ethics approval and consent to participate

Blood and bone marrow samples were obtained from healthy volunteer donors (n = 3) and patients (n = 15) with MDS and AML after the individuals provided written informed consent in accordance with the Declaration of Helsinki to participate in Institutional Review Board (IRB)-approved research protocols. Samples from healthy donors for immunogenicity screening assays were collected on Fred Hutchinson Cancer Center (FHCC) protocols 985 and 2684, and samples from healthy donor samples for third-party T cells for TCR transduction were collected on protocol 985. CD34-enriched cells from GCSF-mobilized peripheral blood stem cell products were obtained on FHCC protocol 2684. Samples from patients were obtained through the FHCC/University of Washington (UW) Hematopoietic Diseases Repository (protocol #1690), FHCC MDS Repository (protocol #1713), FHCC protocol 956, and FHCC protocol 2684. Additional patient samples were obtained according to a protocol approved by the Washington University Human Studies Committee (WU #01–1014). All patients previously provided consent to have samples banked and were not newly recruited for this study.

Human samples

Mononuclear cells were isolated from apheresis obtained from healthy volunteer donors or from whole blood (PBMC) or bone marrow (BMMC) obtained from patients with MDS or AML by Ficoll-Hypaque (Perkin-Elmer) density gradient centrifugation. PBMCs and BMMCs were cryopreserved in aliquots in RPMI 1640 (Gibco) supplemented with 20% fetal bovine serum (FBS, Gibco) and 10% dimethylsulfoxide (DMSO, Sigma-Aldrich) in vapor-phase liquid nitrogen until use.

Cell lines

Epstein-Barr virus (EBV) transformed lymphoblastoid cell lines (B-LCL), including the TM-LCL feeder cell line, were generated as described previously (9) at Fred Hutch and maintained in RPMI 1640, 10% FBS, 2 mM L-glutamine (Gibco), and 1% penicillin/streptomycin (Gibco) (LCL medium) (10). HEK293T cells (FHCC Cooperative Center of Excellence in Hematology Vector Core, RRID: CVCL_0063) used in lentivirus (LV) production were maintained in DMEM (Gibco) supplemented with 10% FBS, 25mM HEPES (Gibco), 2mM L-glutamine, and 1% penicillin/streptomycin, and detached for passage using 0.05% trypsin-EDTA (Gibco). T cells were maintained in RPMI 1640, 10% human serum (Bloodworks Northwest), 1% penicillin/streptomycin, 3mM L-glutamine, and 50μM β-mercaptoethanol (Sigma-Aldrich) (CTL medium). NB-4 (DSMZ# ACC 207, RRID: CVCL_0005) and HNT-34 (DSMZ# ACC 600, RRID: CVCL_2071) myeloid cell lines were purchased from Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ) and maintained in LCL medium with 10–20% FBS. Jurkat 76 cell line (subclone of Jurkat E6.1, RRID: CVCL_0367, {12869497}) were maintained in LCL medium with 10% FBS. Mycoplasma testing (Lonza MycoAlert testing kit) was performed on a monthly basis for all cell lines involved in the study. Morphology and proliferation rates were checked weekly to ensure B-LCL and cell lines appeared normal. Cell lines were checked for known myeloid marker CD33 at the time of each experiment. Cell line authentication by STR was not performed. Time between cell line thaw and use in experiments was kept below a one-month maximum and thawed cells came from banked low passage aliquots.

Class I HLA typing of normal donors and cell lines

Genomic DNA was isolated from cell lines, donor PBMCs or primary neoplastic cells (QIAamp DNA Blood Kit, Qiagen), then used for HLA typing by PCR (Allset Gold Low-Resolution ABC Kit, One Lambda) or by next-generation sequencing (NGS) using the ScisGo HLA v6 typing kit (Scisco Genetics Inc.). Briefly, the NGS employs an amplicon-based 2-stage PCR, followed by sample pooling and sequencing using a MiSeq v2 PE500 (Illumina). Bioinformatic analysis of NGS HLA-typing was performed by Scisco.

HLA binding predictions, processing and presentation predictions, and similar peptides

HLA binding predictions of peptides 8–12 amino acids in length were made with netMHCpan 4.1 (11–13) using both binding affinity (BA) and eluted ligand (EL) predictions for the 23 amino acids spanning SF3B1^K700E (IIEHGLVDEQQEVRTISALAIAA) or the wild-type equivalent (IIEHGLVDEQQKVRTISALAIAA) as input sequences. Predictions were made using the online interface (https://services.healthtech.dtu.dk/services/NetMHCpan-4.1/). Predicted values for SF3B1^K700E and equivalent wild-type peptides for each class I HLA were -log10 transformed and visualized as heatmaps in GraphPad Prism (RRID: SCR_002798). Raw data used for generating heatmaps is included in an Excel spreadsheet as Supplementary Data File S1. Predictions were also made using the Immune Epitope Database (IEDB) analysis resource (http://tools.iedb.org/main/) Consensus tool (14), which combines predictions from NetMHC (4.0) (15–17), SMM (18), and Comblib (19). netMHCpan 4.1 predictions were made on 3/19/2024 and IEDB predictions on 3/20/2022 using the most recent versions of the algorithms at the time. The 22 HLA class I molecules evaluated for predicted binding were HLA-A*01:01, -A*02:01, -A*03:01, -A*33:01, -A*33:03, -A*11:01, -A*24:02, -B*07:02, -B*08:01, -B*15:01, -B*35:01, -B*40:01, -B*44:02, -B*44:03, -C*03:03, -C*03:04, -C*04:01, -C*05:01, -C*06:02, -C*07:01, -C*07:02, and -C*12:03. Cytotoxic T lymphocyte (CTL) epitope predictions assessing processing and presentation were performed using netCTLpan 1.1 (20) for the same 22 HLA class I molecules. We identified similar peptides using the motif search option in the online ScanProsite tool (https://prosite.expasy.org/scanprosite/), setting taxonomy to Homo sapiens and otherwise using default parameters. Control and epitope peptides were synthesized (GenScript) as described previously (8).

Immunogenicity screening and identification of spliceosome neoantigen-specific CD8⁺ T cells

Immunogenicity screening and identification of neoantigen-specific T cells were performed as described previously (21). Briefly, CD8⁺ T cells were purified from HLA-typed healthy donor PBMC and autologous dendritic cells (DC) were generated from monocytes. DCs were incubated with epitope or control peptides, then irradiated prior to co-culture in 96 well plates with CD8⁺ T cells at a ratio of 30:1 T cells:DCs, in the presence of IL-12 (Biotechne) at 10 ng/mL. IL-15 (Biotechne) was added at 10 ng/mL on day 7. 51Cr-release cytotoxicity assays (CRAs) were performed on day 12–13 to identify positive wells containing T cells that lysed peptide-pulsed B-LCL targets. Peptide-specific T cells were then cloned by limiting dilution, screened again by CRA, and expanded using the previously described rapid expansion protocol (REP) using OKT3 (Milteny), IL-2 (Miltenyi), and feeder cells (pooled healthy donor PBMC [AllCells] and TM-LCL) (22) prior to further functional testing as detailed below.

TCR sequencing, transfer into lentiviral vectors, and transduction of T cells

SF3B1^K700E neoantigen-specific TCR beta and alpha chains were sequenced by next-generation sequencing (Adaptive Biotechnologies) as described previously (21).

The dominant unique nucleotide sequence for each chain was as below.

D1.C24 TCR beta variable region:

ATGCTGAGTCTTCTGCTCCTTCTCCTGGGACTAGGCTCTGTGTTCAGTGCTGTCATCTCTCAAAAGCCAAGCAGGGATATCTGTCAACGTGGAACCTCCCTGACGATCCAGTGTCAAGTCGATAGCCAAGTCACCATGATGTTCTGGTACCGTCAGCAACCTGGACAGAGCCTGACACTGATCGCAACTGCAAATCAGGGCTCTGAGGCCACATATGAGAGTGGATTTGTCATTGACAAGTTTCCCATCAGCCGCCCAAACCTAACATTCTCAACTCTGACTGTGAGCAACATGAGCCCTGAAGACAGCAGCATATATCTCTGCAGCGACCGGGGACAGGTCGTAGGAGAGACCCAGTACTTCGGGCCAGGCACGCGGCTCCTGGTGCTCG

D1.C24 TCR alpha variable region:

ATGCTCCTGGAGCTTATCCCACTGCTGGGGATACATTTTGTCCTGAGAACTGCCAGAGCCCAGTCAGTGACCCAGCCTGACATCCACATCACTGTCTCTGAAGGAGCCTCACTGGAGTTGAGATGTAACTATTCCTATGGGGCAACACCTTATCTCTTCTGGTATGTCCAGTCCCCCGGCCAAGGCCTCCAGCTGCTCCTGAAGTACTTTTCAGGAGACACTCTGGTTCAAGGCATTAAAGGCTTTGAGGCTGAATTTAAGAGGAGTCAATCTTCCTTCAATCTGAGGAAACCCTCTGTGCATTGGAGTGATGCTGCTGAGTACTTCTGTGCTGTGGCCGCGAAGGATAGCAGCTATAAATTGATCTTCGGGAGTGGGACCAGACTGCTGGTCAGGCCTG

The transgenic TCR was constructed by pairing the sequences encoding the dominant TCR beta and alpha chains in the SF3B1^K700E-specific clone D1.C24 (above), then codon-optimized (ThermoFisher GeneArt), cysteine modified, and the constant regions murinized, as described previously (23). The constructs were then cloned into the pRRL.PPT.MPSV.WPRE LV (modified from (24), kindly donated by Dr. Stanley Riddell at FHCC), also including the RQR8 selection marker (25), and transduced into normal donor TCR knockout (TCR^KO) CD8⁺ T cells as described previously (26) and as detailed below.

Flow cytometry monoclonal antibodies and instruments

A complete list of monoclonal antibodies used in all assays below is provided in Supplemental Table S1. Flow cytometry was performed on a 4-laser (405 nm, 488 nm, 552 or 532 nm and 628 or 640 nm) Celesta, or 5-laser (355 nm, 405 nm, 488 nm, 552 or 532 nm and 628 or 640 nm) Symphony instrument (BD) Cell sorting was performed on a 4-laser (405 nm, 488 nm, 561 nm, 638 nm) MA900 (Sony) or 4-laser (405 nm, 488 nm, 561 nm, 637nm) FACSAria III (BD) device. All data was analyzed with FlowJo software version 10.1 or higher (BD, RRID: SCR_008520). Fluorochrome-conjugated peptide-HLA (pHLA) tetramers were produced in-house by the FHCC Immune Monitoring Lab core facility.

Knockout of endogenous TCR and TCR transduction

CD8⁺ T cells were enriched from non-mobilized apheresis product collected from HLA-typed healthy donors by immunomagnetic bead enrichment (Miltenyi Biotec), then cryopreserved. Knockout of endogenous TCR by targeting the TCR alpha constant region (TRAC) and TCR beta constant region (TRBC) loci after CD3/CD28 bead stimulation and prior to subsequent TCR transduction was performed as described previously (26). On the day of nucleofection, duplexes of synthetic CRISPR RNA (crRNA) and trans-activating crRNA (tracrRNA) were complexed by incubating 1 μL of Atl-R tracrRNA (IDT) and 0.5 μL each of TRAC (AGAGTCTCTCAGCTGGTACA) and TRBC (GGAGAATGACGAGTGGACCC) guide crRNAs (IDT) at 37°C for 30 minutes. Cas9-NLS protein (IDT) at 6 μM was prewarmed to 37°C for 5 minutes and added to tracrRNA–crRNA complexes, then incubated at 37°C for 15 minutes to generate Cas9-ribonucleoprotein (RNP) complexes. Prior to nucleofection, CD3/CD28 beads were removed, and the cells centrifuged and held on ice. Nucleofection was performed using the P3 Primary Cell 4D-Nucleofector X kit and 4D Nucleofector unit (Amaxa) by adding cells resuspended in nucleofection buffer to RNP complexes and nucleofecting using pulse code EH-115. Immediately following nucleofection, cells were diluted in serum-containing medium with 180 U/mL IL-2 at 37°C for 48 hours prior to TCR transduction. Immediately before TCR transduction, residual CD3⁺ T cells were depleted using anti-CD3 beads (Miltenyi Biotec) to achieve >95% CD3⁻CD8⁺ cells.

TCR^KO T cells were transduced directly after CD3 depletion by adding concentrated LV particles (details below) at a multiplicity of infection (MOI) of 2 in the presence of polybrene (8 μg/mL, Sigma-Aldrich). Transduction efficiency was assessed by flow cytometry 72–96 hours later by staining for the CD34 epitope (Q [QBEnd10] component of RQR8), CD3, and HLA-B*40:01/QEVRTISAL peptide-HLA tetramer. TCR-transduced T cells were enriched to >80% purity by fluorescence activated cell sorting (FACS), expanded by REP, and cryopreserved until use.

T-cell functional assays

Cytotoxicity was measured in short-term (4-hour) CRA using ⁵¹Cr-labeled target cells. Briefly, target cells (including autologous and allogeneic B-LCL with and without peptides in varying concentrations) were labeled with ⁵¹Cr overnight at 37°C and 5% CO₂. Effector cells were added to labeled target cells at E:T ratio of 20:1 and incubated for 4 hours. After co-incubation, supernatant was harvested for γ-counting. Spontaneous release (SR) in counts per minute (CPM) was determined from target cells incubated with media alone. All targets were washed to remove excess ⁵¹Cr and residual peptide before co-culture initiation. Maximal release (MR) was determined from target cells incubated with detergent. The calculation for percent specific lysis by effector cells (T-cell clones or TCR-transduced T cells) in experimental wells was performed using the standard formula (27):

$$\%lysis=((averageCP{M}_{EXPERIMENTAL})-(averageCP{M}_{SR}))/((averageCP{M}_{MR})-(averageCP{M}_{SR}))$$

Longer-term cytotoxicity of T cells against the HNT-34 and NB-4 cell lines was measured in co-culture up to 96 hours by assessing survival of targets using flow cytometry. Briefly, effector and targets cells were plated at a 1:1 ratio (2–4 × 10⁴ each) in four replicate wells in CTL 20% with IL-2 medium in a 96-well plate, and co-cultured at 37°C in 5% CO₂. Wells with only effectors and only targets were included as controls. At various timepoints, the cultures were centrifuged, and cell pellets stained with antibodies against CD34 (QBEnd10), CD8, and DAPI (0.002 μg/mL; Sigma-Aldrich). In some assays, CD20 mAb (rituximab, R component of RQR8 tag) was added to distinguish target cells expressing transgene constructs. Fluorescent CountBright counting beads (ThermoFisher) were used to calculate absolute numbers of live (DAPI-negative) target cells according to the manufacturer’s instructions. Percent survival was calculated as (absolute number of live targets with effector)/(absolute number of live targets without effector)×100.

In cytotoxic degranulation (CD107a) assays, effector T cells and stimulator cells were washed and plated in a 1:1 ratio in LCL medium with GolgiStop transport inhibitor (BD) and anti-CD107a mAb (BD). Stimulator cells were pre-incubated in interferon-γ at 500 units/mL with or without azacitidine at 2 nM for at least 24 hours prior to co-incubation. Effectors and stimulators were co-incubated for 5 hours at 37°C. Cells were then washed and stained with mAbs against CD8, CD33 (clone P67.6), CD34 (clone QBEnd10), and DAPI.

Generation and titration of LV

HEK293T cells were transfected with the LV backbone plasmids along with psPAX2 (Addgene, RRID:Addgene_12260) and pMD2.G (Addgene, RRID: Addgene_12259) packaging plasmids using Lipofectamine 2000 (Invitrogen) per the manufacturer’s protocol. Virus particles were harvested after 48 hours and concentrated using PEG-It (System Biosciences) prior to use according to the manufacturer’s protocol.

Virus was titrated using the naturally TCR-null Jurkat 76 cell line. Concentrated virus was added to Jurkat 76 cells (1×10⁵ per well) in serial 4-fold dilutions starting at 1:20, alongside no virus control wells. Polybrene 4 μg/μL was included in all conditions. 72 hours later, transduction efficiency was evaluated by flow cytometry based on staining for the RQR8 tag (CD34 QBEnd10), CD3, and peptide-HLA tetramer. Transduction units per mL (TU/mL) were calculated using the following formula:

$$TU/mL=(numberofcellstransduced\times \%fluorescence\times dilutionfactor)/(transductionvolumeinmL)$$

Virus volume based on the TU/mL and MOI was calculated using the following formula:

$$Virusvolume=(numberofcellstransduced\times MOI)/(TU/mL)$$

These standard formulas were obtained from Addgene (https://www.addgene.org/protocols/lentivirus-ddpcr-titration/).

Generation of SF3B1^K700E+ HLA-B*40:01⁺ AML cell lines

NB-4 cells were transduced with a codon-optimized nucleotide sequence encoding either a) the 25 amino acids spanning SF3B1^K700E (EIIEHGLVDEQQEVRTISALAIAAL) or b) the wild-type equivalent sequence (EIIEHGLVDEQQKVRTISALAIAAL). An epitope-based marker gene, RQR8, was included upstream of the fusion in the construct for tracking and selection of transduced cells (25). To ensure coordinated gene expression, the transgene components were separated by 2A elements from the porcine teschovirus (P2A). Transgenes were codon-optimized and synthesized as GeneArt (Life Technologies), cloned into the pRRL.PPT.MPSV.WPRE LV backbone plasmid by restriction digestion and ligation, and confirmed by Sanger sequencing. Virus production, cell transduction, and purification were performed as below.

HNT-34 cells were transduced with an HLA-B*40:01 sequence, with the same upstream RQR8 marker described above. To ensure coordinated gene expression, the transgene components were separated by a P2A sequence. Transgenes were codon-optimized and synthesized as GeneArt (Life Technologies), cloned into the pRRL.PPT.MPSV.WPRE LV backbone plasmid by restriction digestion and ligation, and confirmed by Sanger sequencing. Virus production, cell transduction, and purification were performed as below.

Virus production was performed as described in Generation and titration of LV. For cell line transduction, virus was not titrated. Virus particles were harvested after 48 hours and filtered through a 0.45 μm filter prior to use, and this LV supernatant plus 1μg/mL of polybrene was added to cells, followed by 90 minutes of centrifugation at 800×g, after which cells were returned to the incubator at 37°C, 5% CO2. Transduction efficiency of AML cell lines and CD4⁺ T cells was assessed by flow cytometry 72 hours after spin inoculation based on staining for the CD34 epitope (CD34 clone QBEnd10) or CD20 epitope (rituximab). Transduced cell lines were then enriched to >95% purity by flow sorting, expanded, and cryopreserved until use.

Induced pluripotent stem cell (iPSC)-derived multipotent hematopoietic progenitor cell lines (MPP-5F) generation and maintenance

BMMCs were obtained from a HLA-B*40:01⁺ patient with SF3B1^K700E MDS according to the institutional guidelines approved by the FHCC IRB. CD34⁺ BMMC and PBMC were reprogrammed to iPSC colonies as previously described (28). Reprogrammed iPSC colonies were maintained on 90% confluent CF-1 mouse embryonic fibroblasts in human embryonic stem cell (hESC) medium: DMEM/F12 (Gibco) supplemented with 20% KnockOut-serum replacement (Invitrogen), 1 mM L-glutamine, 1 mM non-essential amino acids (Gibco), 0.1 mM β-mercaptoethanol, and 10 ng/ml bFGF (PeproTech). iPSCs were differentiated into embryoid bodies (EBs) as described previously (28) and dissociated on day 14. Dissociated EBs were frozen in 10% DMSO, 40% FBS freezing solution and stored under liquid nitrogen until progenitor sorting. Dissociated EBs were thawed and sorted for CD34⁺CD45⁺ EB progenitors using a BD FACS Aria II cell sorter (CD34 clone 8G12). Sorted CD34⁺CD45⁺ EB progenitors were immediately seeded on RetroNectin-coated 96 well plates and infected with ERG, MYB, RORA, HOXA9, and SOX4 lentiviral constructs as described previously (28) to generate MPP-5F cells. After gene transfer, MPP-5F cells were cultured in StemSpan SFEM (StemCell Technologies) with 50 ng/mL SCF, 50 ng/mL FLT3, 50 ng/mL TPO, 50 ng/mL IL-6, and 10 ng/mL IL-3 (all from PeproTech). Doxycycline was added at 2 μg/mL (MilliporeSigma). Cells were maintained in culture at a density of <1 × 10⁶ cells/mL and medium was changed every 3 to 4 days. MPP-5F cells were routinely cultured between 20 and 80 days without loss of normal karyotype and differentiation potential.

Patient PBMC stimulation

PBMC from patients with SF3B1^K700E-bearing myeloid malignancies who had not recently received antithymocyte globulin or other immunosuppressive agents were thawed and stimulated using a pool of tiled 15-mer peptides spanning SF3B1^K700E (GenScript). The 15-mers allowed for processing and presentation by the target cells to evaluate PBMC from patients with diverse HLA types including HLA-B*40:01. 10 units/mL IL-2 was added to the stimulation at 24 hours and 10 ng/mL IL-15 was added at 120 hours. Stimulation with a pool of 15-mer peptides from 5 influenza proteins (HA, MP1, MP2, NA, NP; Miltenyi Biotec PepTivators) and influenza HLA-A*02:01 epitope GILGFVFTL was performed on an aliquot of each PBMC sample in parallel as a control. Based on data from the Immune Epitope DataBase (iedb.org) and HLA typing on the patient samples, the influenza peptides used should have captured influenza-specific T-cell responses in ≥90% of the samples. After 11–14 days in culture following stimulation, the expanded T cells were re-stimulated with autologous peptide-pulsed target cells (T cells expanded non-specifically by REP) and evaluated for degranulation via CD107a expression and, in some cases, for SF3B1^K700E/B*40:01 pHLA tetramer staining, by flow cytometry. HLA typing was performed as described above on gDNA isolated from expanded cells.

Patient-derived xenograft

A patient-derived xenograft model was generated using NSG-SGM mice engrafted with the iPSC-derived MPP-5F cells above, as detailed in (29) and described below. Animal experiments were performed in accordance with institutional guidelines approved by UW and Fred Hutch animal care committees. The IACUC review boards of the University of Washington and Fred Hutchinson Cancer Center approved these studies (UW protocol #4398–01; Fred Hutch protocol #51082).

NOD.Cg-PrkdcscidIl2rgtm1WjlTg(CMV-IL3,CSF2,KITLG)1Eav/MloySzJ (NSGS) mice were purchased from Jackson Labs (RRID: IMSR_JAX:013062). Intra-femoral transplants were performed as described previously (29): 8 to 14-week-old male and female mice were irradiated at 275 rads 24 hours before transplantation. Following irradiation, enrofloxacin (MWI Animal Health) was administered in drinking water at 0.27 mg/mL to prevent infections. Mice were temporarily sedated with isoflurane to undergo transplantation. The right (injected) femur was drilled with a 27g needle and 5 × 10⁶ MPP-5F were injected in 25 μL volume of cell suspension using a 29.5g insulin needle. Thereafter, throughout the experiment, animals received doxycycline-containing drinking water (1.0 mg/mL) to maintain transgene expression in MPP-5F cells. Approximately eight weeks after MPP-5F injection, GMP-grade recombinant human IFNγ 25,000 units (R&D Systems/Biotechne) was administered parenterally daily for a total of five days to increase surface expression of HLA class I on MPP-5F. The effect of in vivo IFNγ exposure on HLA class I expression was assessed by flow cytometry on human CD45-expressing BMMC (i.e., MPP-5F cells) harvested from an animal euthanized after ≥48 hours of IFNγ and before TCR-T treatment; BMMC from an animal who did not receive IFNγ was used as a control for basal HLA class I expression.

After ≥48 hours of supplemental IFNγ, 10⁷ TCR-T cells (specific for either SF3B1^K700E or control U2AF1^Q157R) were administered IV. IFNγ was continued for two days (total of five days) after TCR-T administration to maintain class I expression. Two weeks later, animals were euthanized and BMMC from the injected femur were collected. MPP-5F cells (human CD45⁺CD33⁺) and TCR-T (human CD45⁺CD8⁺) cells were assessed in aliquots of terminal BMMC by flow cytometry. To confirm that the transferred TCR-T cells were still functional, human CD45⁺CD8⁺ cells were isolated from aliquots of terminal BMMC by FACS, pooled (either from experimental or control animals), then expanded by REP. Expanded TCR-T were then stimulated in vitro with HLA-matched B-LCL pulsed with either SF3B1^K700E QEVRTISAL or control U2AF1^Q157R peptide at 1 μg/mL and evaluated in a CD107a assay.

Data Availability

The raw data generated in this study that was used to produce Figure 1 is available in Supplementary Data File S1. All other data generated in this study is available upon request from the corresponding author.

Figure 1.

Predicted binding affinity and eluted ligand ranking of SF3B1^K700E peptides to 20 prevalent HLA class I alleles. Heatmaps of A, the predicted binding affinity or B, the eluted ligand rank as an output from a netMHCpan 4.1 analysis of the SF3B1^K700E-spanning 18-mer amino acid sequence for the 20 most prevalent HLA class I alleles in the U.S. population. For both heatmaps, data is −log₁₀ transformed and shown for mutant (gray) and wild-type (black) equivalent peptides (indicated at the top of heatmaps). Created in BioRender. Lok, J. (2025) https://BioRender.com/4i4i1dy

Results

Identification of an immunogenic SF3B1^K700E epitope that elicits high-avidity T-cell responses

Presentation of aberrant peptide on HLA molecules is necessary for T-cell recognition of a neoantigen. A foundational approach to neoantigen discovery is to select mutant peptides with strong predicted binding to HLA, focusing on those where predicted binding of the corresponding wild-type peptide is weaker (30). Building on this foundation, we evaluated the 18-mer amino sequence spanning SF3B1^K700E for predicted binding to twenty prevalent HLA class I alleles using netMHCpan 4.1 (Figure 1 and Supplementary Data File S1) and identified 3 distinct SF3B1^K700E peptides with strong predicted binding to 5 HLA. Of those, one nonameric peptide (QEVRTISAL) had strong predicted binding for the relevant HLA (HLA-B*40:01, -B*44:02, and -B*44:03) while the wild-type equivalent (QKVRTISAL) did not.

These findings from a broad in silico screen of numerous HLA suggested QEVRTISAL as a promising SF3B1^K700E epitope. We previously developed a strategy combining multiple in silico analysis methods to identify candidate neoantigen peptides with a high likelihood of being processed from endogenous protein as well as presented on HLA (8) based on three criteria: 1) strong predicted HLA binding of the mutant peptide by at least two of four publicly available algorithms; 2) high ranking (% rank <1) based on a composite score from the processing and presentation algorithm netCTLpan (20); and 3) weak predicted binding of the equivalent wild-type peptide to the relevant HLA. QEVRTISAL (Table 1, Figure 2A) met all criteria to be a candidate neoantigen peptide created from the SF3B1^K700E mutation. QEVRTISAL was predicted to bind three HLA molecules: HLA-B*40:01, -B*44:02, and -B*44:03 and to be processed from the parent protein and loaded onto HLA. Binding of QKVRTISAL, the wild-type equivalent peptide, to HLA-B*40:01, -B*44:02, and -B*44:03 was predicted to be weak. Because the predicted affinity for HLA-B*40:01 was >20-fold higher than for HLA-B*44:02 or -B*44:03 (14 nM versus 278 nM or 250 nM, respectively in Table 1; Figure 2), we focused on HLA-B*40:01 as the most promising predicted HLA restriction for QEVRTISAL as a candidate SF3B1^K700E neoantigen.

Table 1.

In silico HLA binding and processing/presentation (“CTL epitope”) predictions for a candidate SF3B1^K700E neoantigen peptide. HLA binding predictions for the equivalent wild-type (wt) peptide are also shown. Variant residue (position 2) is underlined. Abbreviations: ANN, Immune Epitopes DataBase (IEDB) artificial neural network method for HLA-binding prediction; SMM, IEDB stabilized matrix method for HLA-binding prediction; BA, binding affinity-based prediction of HLA binding; EL, mass spectrometry eluted ligand-based prediction of HLA binding; netCTLpan score components: HLA, HLA-binding prediction score; TAP, transporter associated with antigen processing transport efficiency prediction score; Cle, peptide cleavage prediction score; Comb, combined processing and presentation score; % Rank, % rank based on combined processing and presentation scores.

HLA	K700E peptide sequence	Predicted binding affinity (nM)			% Rank by netMHC pan EL	Total HLA-binding	netCTLpan scores					Predicted processed/ presented?	Wild-type (wt) peptide sequence	wt % Rank by netMHC pan EL	wt predicted binding affinity (netHMC pan, nM)	wt binds (BA)?	wt binds (EL)?	wt binds (either)?	Meets all criteria for optimal epitope?
		ANN	SMM	netMHC pan BA			HLA	TAP	Cle	Comb	%Rank
B*40:01	QEVRTISAL	14	55	14	0.034	4	0.686	0.861	0.959	0.923	0.1	YES	QKVRTISAL	2.676	10334	no	no	no	YES
B*44:02	QEVRTISAL	577	2107	278	0.211	2	0.314	0.861	0.959	0.551	0.8	YES	QKVRTISAL	8.304	24567	no	no	no	YES
B*44:03	QEVRTISAL	187	524	250	0.174	3	0.358	0.861	0.959	0.595	0.8	YES	QKVRTISAL	7.222	25391	no	no	no	YES

Figure 2.

The SF3B1^K700E neoantigen epitope is immunogenic and primes a high-avidity epitope-specific CD8⁺ T cell clone. A, Schematic of altered SF3B1^K700E protein and neoantigen epitope peptide. B, A clone was identified after primary in vitro stimulation of CD8⁺ T cells from an HLA-B*40:01-positive donor with QEVRTISAL peptide-pulsed DC and tested in peptide titration CRA against autologous LCL pulsed with varying peptide concentrations (3 technical replicate experiments). C, HLA restriction of D1.C24 was confirmed by testing in CRA against a panel of HLA-typed LCL with single HLA overlap with the original T cell donor. LCL were pulsed with QEVRTISAL peptide at 1000 ng/mL prior to coculture (≥5 biological replicates per HLA). D, D1.C24 was tested for recognition of known immunogenic HLA-B*40:01-presented peptides in CRA using autologous LCL pulsed with 1000 ng/mL of each peptide. QEVRTISAL peptide and the wild-type equivalent QKVRTISAL were included as controls (3 technical replicate experiments). E, Alanine scanning for D1.C24 was performed using autologous LCL pulsed with a panel of peptides (1000 ng/mL) with alanine residues substituted at each position, along with two peptides with either a glycine or valine substitution at position 8, a natural alanine residue in the QEVRTISAL peptide (3 technical replicate experiments). These data were used to identify critical residues for HLA and TCR binding and define the core motif xExRTIxxL. F, Four peptides derived from wild-type human proteins and sharing the xExRTIxxL motif were identified using the Scan ProSite tool. To evaluate for cross-recognition of these peptides by D1.C24, autologous LCL were pulsed with each peptide (1000 ng/mL) and used as targets for D1.C24 in CRA (3 technical replicate experiments). For all experiments, mean and SEM are shown. Created in BioRender. Lok, J. (2025) https://BioRender.com/pijsqsm

We next evaluated whether the QEVRTISAL peptide could stimulate T-cell responses (i.e., was immunogenic). Primary in vitro stimulation of CD8⁺ T cells from an HLA-B*40:01⁺ donor with peptide-pulsed autologous monocyte-derived DCs yielded a QEVRTISAL-specific clone (Figure 2B). The D1.C24 clone demonstrated high functional avidity in a CRA with half-maximal lysis around 0.1 ng/mL (Figure 2B). D1.C24 recognized only QEVRTISAL peptide-pulsed B-LCLs bearing HLA-B*40:01 and not B-LCL with other HLA shared by the original T-cell donor (Figure 2C), confirming HLA-B*40:01 as the restricting allele. Although QEVRTISAL was predicted to bind strongly to HLA-B*40:02 (binding affinity 25 nM; netMHCpan 4.1), which is structurally similar to HLA-B*40:01 and may present the same peptides (21), D1.C24 did not recognize the peptide presented on HLA-B*40:02 (Supplemental Figure S1). To exclude cross-reactivity against other known HLA-B*40:01-presented epitopes, CRA were performed using B-LCL pulsed with one of several peptides as targets and it was confirmed that D1.C24 lysed only autologous B-LCL pulsed with QEVRTISAL peptide, not those pulsed with wild-type (QKVRTISAL) peptide or any of three known immunogenic, HLA-B*40:01-presented peptides (Figure 2D).

To further examine D1.C24’s potential for off-target recognition of non-SF3B1^K700E peptides, we performed alanine scanning, which demonstrated that residues at positions 2, 4, 5, 6, and 9 were required for D1.C24 to recognize QEVRTISAL (Figure 2E). Positions 2 and 9 are known anchor residues for HLA-B*40:01-binding (31). To assess potential cross-recognition of other similar human peptides, we performed an in silico search for proteins containing the xExRTIxxL motif using the ScanProsite tool (32, 33) and identified four wild-type human proteins with the motif: alkylglycerone-phosphate synthase (AGPS), mitochondrial trifunctional enzyme subunit alpha (HADHA), Gem-associated protein 5 (GEMIN5), and spermatogenesis-associated protein 1 (SPATA1). However, none of the peptides with the xExRTIxxL motif were recognized by D1.C24 (Figure 2F), indicating that the clone was highly specific for QEVRTISAL.

SF3B1 ^K700E/HLA-B*40:01 is a bona fide MDS and sAML neoantigen

We then investigated whether the QEVRTISAL epitope was naturally presented on neoplastic myeloid cells, first utilizing cell lines. To create target cells that were genotypically positive for both SF3B1^K700E and HLA-B*40:01, we transduced the naturally SF3B1^K700E-positive sAML cell line HNT-34 to express HLA-B*40:01. In a flow cytometry–based cytotoxicity assay, HLA-B*40:01-transduced HNT-34 cells were efficiently eliminated by D1.C24 within 24 hours, but not by a control clone specific for an irrelevant neoantigen (U2AF1^Q157R/A*33:03) and D1.C24 did not eliminate HNT-34 that had not been transduced to express HLA-B*40:01 (Figure 3A; gating strategy Supplemental Figure S2). We evaluated recognition of the SF3B1^K700E/HLA-B*40:01 putative neoantigen in a second cell line model in which we transduced a minigene encoding SF3B1^K700E (Supplemental Figure S3) into the naturally HLA-B*40:01⁺ myeloid cell line NB-4. D1.C24 eliminated SF3B1^K700E minigene-transduced NB-4 in the cytotoxicity assay, but not non-transduced wild-type NB-4, showing specific lysis of the SF3B1^K700E/HLA-B*40:01 neoantigen (Figure 3B), consistent with the HNT-34 experiment results.

Figure 3.

The SF3B1^K700E neoantigen is naturally processed and presented on neoplastic myeloid cells. A, Percent survival of HNT-34/B*40:01 cells cocultured with either D1.C24 (blue) or irrelevant neoantigen-specific clone (gray) in a flow cytometry cytotoxicity assay. Mean and SEM from ≥3 technical replicates shown. B, Percent survival of SF3B1^K700E minigene transduced NB-4 cells cocultured with either clone D1.C24 (blue) or irrelevant neoantigen-specific clone (gray) in flow cytometry cytotoxicity assay. Mean and SEM from ≥3 technical replicates shown. C, Representative flow plots from CD107a assay demonstrating D1.C24 degranulation in response to genotypically SF3B1^K700E-positive HLA-B*40:01-positive primary neoplastic myeloid cells (MDS or sAML) but not controls lacking either the mutation or HLA. D, Summary data from CD107a assays of D1.C24 response to primary neoplastic myeloid cells (≥1 biological replicates, ≥2 technical replicates). E, Representative flow plots from CD107a assay demonstrating D1.C24 degranulation with genotypically SF3B1^K700E-positive HLA-B*40:01-positive MPP-5F line but not SF3B1^K700E-negative HLA-B*40:01-positive isogenic control. MPP-5F lines were pre-cultured with IFNγ for at least 24 hours. F, Summary data from CD107a assays of D1.C24 response to MPP-5F lines under various conditions (≥3 biological replicates). For summary figures, mean and SEM are shown. Created in BioRender. Lok, J. (2025) https://BioRender.com/6rz6fds

To determine whether the QEVRTISAL epitope was naturally processed and presented by primary malignant myeloid cells, the high-avidity SF3B1^K700E-specific clone (D1.C24) was co-cultured with hematopoietic cells from patients with active MDS or AML, then assessed for antigen recognition in a CD107a degranulation assay. D1.C24 showed specific degranulation in response to primary SF3B1^K700E/HLA-B*40:01-positive samples (n=1) compared to samples lacking either the mutation (n=3) or restricting HLA (n=3) (Figure 3C and D). These results indicate that the SF3B1^K700E epitope is naturally processed and presented on primary neoplastic hematopoietic cells and is thus a bona fide MDS/sAML neoantigen.

Hematopoietic progenitor lines generated from SF3B1^K700E primary MDS cells present antigen

One challenge in developing TCR-T therapies for MDS is obtaining adequate numbers of primary cells from patient samples. iPSC can be differentiated into multipotent hematopoietic progenitor cell lines (MPP-5F) that recapitulate the genotype and phenotype of primary patient MDS cells (28). We thus evaluated whether SF3B1^K700E-specific clones could recognize SF3B1^K700E iPSC-derived hematopoietic cells. Two separate clones from hematopoietic stem/progenitor cells isolated from bone marrow from an HLA-B*40:01⁺ individual with SF3B1^K700E MDS were reprogrammed, and two iPSC lines established: SF3B1^K700E and an isogenic control with wild-type SF3B1. Primary bone marrow cells from this patient were known to be recognized by D1.C24 (Figure 3C and D). MPP-5F were then generated from iPSC lines by doxycycline-dependent expression of five HSPC transcription factors (28). Baseline expression of HLA class I on MPP-5F was low but was augmented by IFNγ exposure (Supplemental Figure S4). Although neither the SF3B1^K700E nor control MPP-5F were recognized by D1.C24 without pre-culture with IFNγ, IFNγ-exposed SF3B1^K700E MPP-5F elicited strong recognition by D1.C24 in a CD107a assay, while the isogenic control did not (Figure 3E and F). These data indicate that iPSC-derived hematopoietic progenitor cell lines can recapitulate presentation of neoantigens and enable comparison of neoantigen-specific T-cell recognition of isogenic SF3B1^K700E and wild-type clones in the context of endogenous HLA class I expression.

Functional SF3B1^K700E-specific T cells are not detected in patient PBMC

Having determined that SF3B1^K700E/B*40:01is naturally processed and presented on MDS and sAML, we asked whether patients with SF3B1^K700E-bearing myeloid malignancies might develop natural T-cell responses to SF3B1^K700E epitopes. We stimulated PBMC from patients with SF3B1^K700E-positive MDS, including four who were genotypically positive for HLA-B*40:01, using pools of tiled 15-mer peptides spanning SF3B1^K700E to expand antigen-specific T cells. Influenza peptides were used to stimulate a separate aliquot as a control. Post-stimulation PBMC were evaluated for CD8⁺ T-cell recognition of autologous peptide-pulsed target cells by flow cytometry (Supplemental Figure S5). Influenza-stimulated controls responded to influenza peptide-pulsed targets (10 of 15 samples), indicating that T cells from patients could function and the assay was suitable for evaluation of antigen-specific T-cell responses. In contrast, SF3B1^K700E-stimulated CD8⁺ T cells showed only background CD107a expression even following re-stimulation with SF3B1^K700E peptides. pHLA tetramer staining of expanded PBMC post-stimulation with SF3B1^K700E peptides did not identify convincing SF3B1^K700E-specific T cells (Supplemental Figure S6).

SF3B1^K700E/B*40:01-specific TCR can be transferred and confers specificity

The lack of detectable functional patient T-cell responses to SF3B1^K700E (Supplemental Figures S5, S6) suggests that this neoantigen may not be optimal for immunotherapy strategies that rely on stimulating or augmenting existing responses, like vaccination. However, T cells can be engineered to express transgenic TCRs (34) and thus target cancer-specific antigens, including neoantigens. TCR-T can augment natural responses against cancer antigens or even replace them when natural responses are defective or absent (35). We assessed the feasibility of transferring a TCR specific for the SF3B1^K700E/B*40:01 neoantigen as a step towards clinical translation. We first sequenced the TCRα and β chains from D1.C24, then cloned the paired TCRα and β chains into a LV containing the RQR8 selection marker (25) and transduced 3^rd-party CD8⁺ T cells from which the endogenous TCR chains had been knocked-out using CRISPR/Cas9. Primary human CD8⁺ T cells transduced with the SF3B1^K700E/B*40:01-specific TCR, but not non-transduced controls, stained with SF3B1^K700E/B*40:01 pHLA tetramer along with a monoclonal antibody specific for CD34, encoded by the RQR8 sequence (Figure 4A). T cells transduced with the SF3B1^K700E/B*40:01-specific TCR killed QEVRTISAL-pulsed B-LCL less efficiently at low peptide concentrations than the parental D1.C24 in a 4-hour CRA (Figure 4B) but killed HLA-B*40:01-transduced HNT-34 cell lines bearing natural SF3B1^K700E by 48 hours (Figure 4C). Similarly, the SF3B1^K700E/B*40:01-specific TCR-T killed SF3B1^K700E-transduced NB-4 cell lines, which are naturally HLA-B*40:01-positive (Figure 4D). Taken together, these results indicate that transfer of an SF3B1^K700E-specific TCR is feasible and confers antileukemic activity in vitro.

Figure 4.

Transfer of the SF3B1^K700E/B*40:01-specific TCR confers specificity and function. A, Representative flow plots demonstrating expression of the D1.C24 SF3B1^K700E/B*40:01-specific TCR (TCR24) transduced (TD) into primary human CD8⁺ T cells after CRISPR/Cas9-mediated knock-out of endogenous TCR alpha and beta chains (middle) showing staining for the RQR8 transduction marker and the transgenic TCR with SF3B1^K700E/B*40:01-pHLA tetramer. B, Cytolytic activity of TCR24 TD T cells and corresponding T cell clone in CRA (technical triplicates). C, Percent survival of HNT-34 cells without and with transduction of HLA-B*40:01 cocultured with TCR24 TD T cells, parental clone, or controls in flow cytometry cytotoxicity assay (≥3 technical replicates). D, Percent survival of NB-4 cells without and with transduction of SF3B1^K700E minigene cocultured with TCR24 TD T cells, parental clone, or controls in flow cytometry cytotoxicity assay (≥3 technical replicates). Created in BioRender. Lok, J. (2025) https://BioRender.com/xtc8n5z

We then evaluated the functionality of SF3B1^K700E-specific TCR-T in vivo using a PDX model in which NSGS mice were engrafted with iPSC-derived MPP-5F cells. Informed by in vitro experience with MPP-5F (Supplemental Figure S4), recombinant human IFNγ was administered for two days before and after TCR-T administration to increase HLA class I on MPP-5F (Supplemental Figure S7A). In animals treated with SF3B1^K700E-specific TCR-T, the percent of CD33-expressing human CD45⁺ cells (engrafted MPP-5F cells) were decreased in terminal BMMC compared to those treated with control TCR-T (Supplemental Figure S7B), although this difference did not achieve statistical significance. The percent of CD8-expressing human CD45⁺ cells in terminal marrow significantly decreased in the experimental group compared to control (Supplemental Figure S7C), consistent with some activation-induced cell death as we have previously observed with antigen recognition by adoptively transferred neoantigen-specific T cells in a different PDX model (21). Human CD45⁺CD8⁺ cells isolated from terminal BMMC and expanded in vitro degranulated in response to B-LCL bearing the appropriate restricting HLA and pulsed with either SF3B1^K700E QEVRTISAL or control peptide (Supplemental Figure S7D), indicating that transferred, persisting TCR-T were functional. Taken together, these results suggest that SF3B1^K700E-specific TCR-T have in vitro and in vivo activity against SF3B1^K700E HLA-B*40:01⁺ myeloid cells.

Discussion

There is a paucity of potentially curative antigen-specific T-cell therapies for MDS and sAML, in part due to lack of suitable target antigens. Although neoantigen discovery is now decades old (reviewed in (30)), recent studies highlight the continued relevance of neoantigen discovery for the treatment of these diseases (36, 37), and novel approaches may accelerate the ability to identify translationally relevant neoantigens (38). Spliceosome mutations generally, and SF3B1 mutations in particular, are initiating mutations that form founding neoplastic clones (39). Therapies redirecting T cells to recognize spliceosome mutation–derived neoantigens could therefore eradicate neoplastic myeloid cells without adversely affecting normal hematopoiesis. Since MDS is a genetically heterogeneous disease, a toolbox of multiple spliceosome neoantigens is needed to develop broadly applicable T-cell therapies. In this study, we combined in silico and in vitro screening to identify a neoantigen created from SF3B1^K700E mutation, which is found in approximately 10% of MDS (1). CD8⁺ T cells that were naturally specific for the neoantigen or engineered to express an SF3B1^K700E-specific TCR recognized and killed neoplastic myeloid cells bearing both the mutation and restricting HLA-B*40:01. We were unable to detect functional T-cell responses to SF3B1^K700E peptides in patients with SF3B1^K700E-bearing myeloid malignancies, consistent with our and others’ experience with undetectable natural responses to shared neoantigens in de novo AML (21, 40). We did consider that T cells in patients might be dysfunctional. Given that terminal differentiation and increasing dysfunction of CD8⁺ T cells is associated with preserved cytotoxicity but decreased cytokine production and in vitro proliferative capacity (41), our approach of measuring degranulation, a proxy for cytotoxicity, should have captured all but the most profoundly dysfunctional SF3B1^K700E-specific T cells. Furthermore, we were unable to detect pHLA tetramer-staining T cells in stimulated PBMC from HLA-B*40:01⁺ individuals. The lack of robust endogenous responses to SF3B1^K700E provides a compelling rationale for investing in TCR-T to augment immunity in SF3B1^K700E-mutated MDS and sAML.

The goal of our research is to help build the “toolbox” of potential TCR-T therapies available to patients with myeloid malignancies. Our discovery of an SF3B1^K700E neoantigen expands the toolbox of known MDS and AML neoantigens with published preclinical data supporting their targeting with TCR-T therapies (reviewed in (42) and Supplemental Table S2 (8, 21, 36, 40, 43–54)). We recently identified two neoantigens created from the recurrent U2AF1^Q157R spliceosome mutation with therapeutic potential (8). Other known shared neoantigens relevant to MDS and sAML include those created from recurrent mutations in TP53 and in Ras family genes (KRAS, NRAS, and HRAS) that produce identical peptide epitopes (42). Neoantigens resulting from KRAS and TP53 mutations were identified in a solid tumor context and occur in cancers broadly, including unfavorable-risk myeloid neoplasms, where mutations in TP53 and in Ras family members are associated with sAML transformation and chemotherapy resistance. Three neoantigens pertaining to de novo AML, created from a leukemia-initiating CBFB-MYH11 gene fusion, a recurrent NPM1 frameshift mutation, and a FLT3 gain-of-function mutation have also been identified (21, 36, 40). Because of the genetic heterogeneity of myeloid neoplasms and HLA restriction, neoantigen-directed TCR-T cell therapies for MDS and AML are inherently semi-personalized rather than “one-size-fits-all” or even “one-size-fits-most.” Like the SF3B1^K700E neoantigen, most published MDS/AML neoantigens currently in the toolbox individually apply to less than 10% of all patients (Supplemental Table S2), so each new neoantigen increases the overall applicability of this therapeutic toolbox and works towards the goal of providing at least one TCR-T option to every patient in need.

In MDS, SF3B1 mutations are often associated with a favorable risk profile. However, a subset of patients with SF3B1-mutated AML present with higher risk features such as blasts >5%, intermediate- to very poor-risk cytogenetics, and/or certain cooperating mutations and have worse overall survival (3). This subgroup of patients with SF3B1-mutated MDS, along with those who have SF3B1-mutated sAML would derive the greatest benefit from potentially curative SF3B1^K700E-directed T-cell therapy. In addition, T-cell therapies targeting SF3B1^K700E have applicability beyond myeloid neoplasms. SF3B1 mutations are common events in chronic lymphocytic lymphoma (CLL), occurring in >15% of patients with high-risk disease (2). As in MDS, mutations occur at hotspots and about 50% of patients with SF3B1-mutated CLL have the K700E variant (2). SF3B1 mutations have also been detected in solid tumors including pancreatic cancer (4%) (37), renal cancer (3%) (55) and breast cancer, where SF3B1^K700E mutations are found in 16% of papillary breast cancers and 6% of mucinous breast cancers (37). SF3B1^K700E-directed T-cell therapies thus could fill an unmet need for effective treatments for certain high-risk patients with a variety of cancers.

One challenge for TCR-T therapy generally is the ability of malignant cells to escape immune surveillance by losing antigen presentation through any of several mechanisms (56). SF3B1^K700E is advantageous as a target since it arises in a founding clone and has a disease-initiating role. These features have two implications: First, the mutation should be present in all aberrant cells such that targeting would eradicate all neoplastic clones. Second, loss of SF3B1^K700E expression should to lead to diminution or loss of the malignant phenotype, making immune escape by downregulation or silencing of SF3B1^K700E expression unlikely (56). Alterations in processing or presentation by malignant myeloid cells could lead to resistance, but targeting SF3B1^K700E along with other neoantigens, using a combination of different TCR-T deployed either simultaneously or sequentially, could further reduce the risk of immune escape.

We additionally present the use of iPSC-derived hematopoietic (MPP-5F) cell lines as target cells for pre-clinical evaluation and validation of MDS antigen-specific T-cell immunotherapy. We found that MPP-5F lines processed and presented the SF3B1^K700E epitope as expected, if exposed to IFNγ to increase expression of the restricting HLA. While disease-related cytopenias and ethical considerations on sample sizes limit the availability of primary cells from MDS patients, iPSC-derived lines can be generated in large numbers. In addition, since multiple iPSC-derived hematopoietic lines can be generated from different clones within the same individual, using them as targets allows for direct comparison of T-cell recognition of cells with identical endogenous HLA expression differing only in the presence or absence of the mutation. Although further study is needed to broadly assess how presentation of diverse antigens by iPSC-derived hematopoietic lines compares to primary cells, our findings with the SF3B1^K700E neoantigen suggests that these cells could prove valuable tools to validate novel MDS antigens as targets for immunotherapy.

We have identified a shared neoantigen created from the recurrent mutation SF3B1^K700E which, because it arises in a founding clone and has a disease-initiating role, has curative potential as a target for TCR-T therapy. The knowledge that the SF3B1^K700E epitope is immunogenic could suggest its incorporation as part of vaccine studies for myeloid neoplasms (57). However, our observations indicated that SF3B1^K700E-specific T cells were absent or very dysfunctional in patients, making vaccination unlikely to succeed in this context and providing further rationale for the development of TCR-T for MDS. Developing precision medicine T-cell therapies that are broadly applicable to patients with MDS and sAML requires a large repertoire of semi-personal neoantigens representing a wide range of mutations and HLA-restrictions. Discovery of the SF3B1^K700E epitope represents a key contribution to this effort.

Synopsis:

There is a paucity of potentially curative antigen-specific T-cell therapies for myeloid neoplasms. The authors identify a neoantigen created from the recurrent disease-initiating SF3B1^K700E mutation as a promising T-cell therapy target for these diseases.