1. Introduction
Acute myeloid leukemia (AML) is the most common type of leukemia in adults. Despite the high rate of initial complete remission in AML patients treated with intensive chemotherapy, relapses frequently occur [
1]. Allogeneic hematopoietic stem cell transplantation has been established as a treatment that effectively reduces the risk of relapse and increases the overall survival of AML patients [
2]. The anti-tumor effect of allogeneic stem cell transplantation is induced by strong alloreactive T-cell responses, indicating that cellular immunotherapy has the potential to eradicate AML [
3]. Although allogeneic stem cell transplantation increases overall survival, up to half of AML patients still relapse [
4], emphasizing the need to develop new treatment strategies for AML.
We and others have shown that neoantigens encoded by recurrent genetic aberrations in AML can be targeted using immunotherapy [
5,
6,
7,
8,
9,
10]. Neoantigens are attractive targets for therapy since they are highly specific for cancer cells [
11]. We previously isolated a T-cell receptor (TCR) for an HLA-A*02:01-binding neoantigen derived from mutant NPM1 [
5]. T-cells engineered with this TCR have been shown to effectively target AML in vitro as well as in mice engrafted with AML.
In AML, mutations have been found in genes involved in epigenetic regulation, including genes affecting DNA (hydroxy)methylation, such as DNA methyltransferase 3A (
DNMT3A), Tet oncogene family member 2 (
TET2), and isocitrate dehydrogenase 1 and 2 (
IDH1/2). Mutations in
DNMT3A,
TET2,
IDH1, and
IDH2 each occur in 7–23% of AML cases [
12], and are also found in clonal hematopoiesis of indeterminate potential (CHIP), a condition prevalent in elderly people and patients in remission after treatment for AML [
13]. Individuals with CHIP are at increased risk of developing AML.
DNMT3A,
TET2,
IDH1, and
IDH2 mutations in hematopoietic stem cells from individuals with CHIP are usually sub-clonal. Targeting these mutations in patients with AML may not only effectively eradicate leukemia, but also pre-malignant hematopoietic stem cells, thereby reducing the risk of relapse.
Neoantigen identification benefits greatly from human cell line models where somatic mutations are introduced via Clustered Regularly Interspaced Short Palindromic Repeats (CRISPRs) and CRISPR-associated protein 9 (Cas9), thereby allowing analysis of neoantigen surface expression under more physiological conditions [
14]. Conventional CRISPR techniques use complexes consisting of Cas9 and a custom-designed single-guide RNA (sgRNA), which guides Cas9 to the target locus, where it induces a double-stranded DNA break. These breaks are repaired through mechanisms like nonhomologous end joining or homology-directed repair, which facilitates the introduction of a repair template. Base editing is a new CRISPR method that enhances precision genome editing. It involves a Cas9 nickase, which creates single-stranded breaks, and an enzyme that chemically modifies a single base, which is corrected by endogenous cellular repair mechanisms. Unlike homology-directed repair, which is limited to dividing cells, base editing can modify both dividing and nondividing cells [
15].
Here, we explored whether mutations in genes involved in DNA methylation or hydroxymethylation encode neoantigens that can be targeted with immunotherapy. We focused on hotspot mutations in
DNMT3A,
IDH1, and
IDH2 [
12], as they may encode recurrent neoantigens that are shared between multiple patients. Two HLA class I-binding neopeptides derived from DNMT3A
R882H and IDH2
R140Q were identified using tandem mass spectrometry, and used to search for specific T-cell clones. Isolated T-cell clones were screened against human cell lines expressing a natural
DNMT3AR882H mutation [
16] or
IDH2R140Q mutation introduced by CRISPR-Cas9 using homology-directed repair [
17]. Among the various isolated T-cell clones, one clone showed specific reactivity against patient-derived AML cells with
DNMT3AR882H, thereby validating the surface presentation of a neoantigen that can potentially be targeted with immunotherapy.
2. Materials and Methods
2.1. Human Samples and Cell Culture
Peripheral blood and bone marrow samples from AML patients were acquired from the Leiden University Medical Center Biobank for Hematological Diseases. Peripheral blood mononuclear cells (PBMCs) used for T-cell isolations were freshly used and obtained from healthy individuals (Sanquin, Amsterdam, The Netherlands). PBMCs and bone marrow mononuclear cells were isolated using Ficoll-Isopaque density gradient centrifugation. Patient AML samples were thawed and incubated in Iscove’s Modified Dulbecco’s Medium (IMDM) supplemented with 10% human serum (Sanquin) and 100 U/mL penicillin/streptomycin (pen/strep; Lonza, Basel, Switzerland). Epstein–Barr virus-transformed human B-lymphoblastoid cell lines (EBV-B cell lines), K562, and HL-60 cell lines (ATCC, Manassas, MA, USA) were cultured in IMDM supplemented with fetal bovine serum (FBS; Thermo Fisher Scientific, Bremen, Germany), 200mM L-glutamine (Lonza), and pen/strep. The human K562-R140Q cells had a homozygous
IDH2R140Q mutation introduced via CRISPR-Cas9-guided genome editing [
17]. SET-2 (DSMZ), which is an AML cell line with a heterozygous
DNMT3AR882H mutation [
16], was cultured in Dulbecco’s Modified Eagle Medium (DMEM; Lonza) supplemented with FBS, L-glutamine, pen/strep, and 1 × 10
5 M β-mercaptoethanol (Sigma Aldrich, Burlington, MA, USA). T-cells were cultured in T-cell medium (TCM) consisting of IMDM with FBS, human serum, L-glutamine, pen/strep, and 100 IU/mL IL-2 (Novartis, Macquarie Park, NSW, Australia). T-cells were (re)stimulated every 14 days with 0.8 µg/mL phytohemagglutinin (PHA) and irradiated (40 Gy) allogeneic feeder cells.
2.2. Gene Constructs
DNMT3A, IDH1, and IDH2 hotspot mutations were combined as minigenes in an MP71 retroviral construct containing the nerve growth factor receptor (NGFR) as a marker gene. Each minigene consisted of 87 nucleotides coding for 29-mer peptides with the mutated amino acid at position 15. The construct contained the following minigenes from the 5′ to 3′ ends: DNMT3A c.2645G>A, IDH2 c.419G>A, IDH1 c.395G>A, IDH1 c.394C>T, DNMT3A c.2644C>T, IDH2 c.515G>A, IDH1 c.394C>G, and IDH1 c.394C>A. Constructs with full-length DNMT3AR882H or IDH2R140Q genes were inserted into the LZRS retroviral vector containing NGFR as a marker gene. Genes encoding HLA-A*01:01 or HLA-B*07:02 were inserted into the LZRS vector containing mouse CD19 as a marker gene. After retroviral transduction, EBV-B cell lines, K562-R140Q, and SET-2 cells were purified on NGFR or mouse CD19 expression using magnetic beads (MACS beads; Miltenyi Biotec, Westphalia, Germany).
2.3. Samples for Peptide Elution
To identify the neopeptides encoded by
DNMT3A,
IDH1, or
IDH2 hotspot mutations, cell pellets (1 × 10
9 cells) were collected from five EBV-B cell lines transduced with the minigene construct (
Table S1), as well as from SET-2 transduced with HLA-A*01:01 (and the full-length
DNMT3AR882H gene) and K562-R140Q transduced with HLA-B*07:02 (and the full-length
IDH2R140Q gene). In addition, cell pellets (100 × 10
6 cells) from two patient AML samples were collected. AML samples were selected based on the HLA type and
DNMT3A or
IDH2 mutations as identified via RNA-Seq and validated with Sanger sequencing [
18].
2.4. HLA Class I Peptide Elution
Peptide elution was performed as described previously [
5]. Briefly, cell pellets were lysed and HLA class I molecules were immunoprecipitated using an immunoaffinity column to extract peptide–HLA (pHLA) complexes using an HLA class I antibody (W6/32, ATCC). pHLA complexes were dissociated from the column using 10% acetic acid. Peptides and HLA class I molecules were separated via filtration through a 10 kDa membrane. The collected peptide pools were fractionated with strong cation exchange chromatography (SCX) and analyzed using mass spectrometry.
2.5. Tandem Mass Spectrometry
For data-dependent acquisition (DDA) tandem mass spectrometry (MS/MS), the peptides were lyophilized, dissolved in a solution of 95/3/0.1
v/
v/
v water/acetonitrile/formic acid, and analyzed online with C18 nano-HPLC MS/MS using an Easy nLC 1200 gradient HPLC system (Thermo, Bremen, Germany) and a LUMOS mass spectrometer (Thermo). Fractions were injected onto a homemade precolumn (100 μm × 15 mm; Reprosil-Pur C18-AQ 3 μm, Dr. Maisch, Ammerbuch, Germany) and eluted via a homemade analytical nano-HPLC column (30 cm × 50 μm; Reprosil-Pur C18-AQ 3 μm). The gradient ran from 2% to 36% solvent B (20/80/0.1 water/acetonitrile/formic acid (FA)
v/
v/v) over 120 min. The nano-HPLC column served as the electrospray needle of the MS source, with a tip diameter of ~5 μm. The LUMOS mass spectrometer operated in the data-dependent MS/MS mode with a cycle time of 3 s, HCD collision energy at 32 V, and recording of the MS2 spectrum in the orbitrap. In the master scan (MS1), the resolution was set at 60,000× and the scan range was from 300 to 1400, with a ‘standard’ AGC target and a maximum fill time of 50 ms. Dynamic exclusion after
n = 1 with an exclusion duration of 20 s was applied. Charge states included 1 (precursor selection range: 800–1400), 2 (precursor selection range: 400–800), and 3 (precursor selection range: 300–600). For MS2, precursors were isolated with the quadrupole using an isolation width of 1.2 Da. The MS2 scan resolution was 30,000 with the ‘standard’ AGC target at a ‘dynamic’ maximum fill time. For the post-analysis processing, raw data were first converted to peak lists using Proteome Discoverer version 2.1 (Thermo Electron), and then submitted to the Uniprot Homo sapiens minimal database (20,596 entries), using Mascot v. 2.2.07 (
www.matrixscience.com) for protein identification. Mascot searches were carried out with 10 ppm and 0.02 Da deviation for the precursor and fragment mass, respectively, and no enzyme was specified. Methionine oxidation and cysteinylation of cysteine were set as variable modifications. The false discovery rate was set to <1% and, in addition, peptides with mascot ion scores <35 were discarded. We searched for 9, 10, and 11mer peptides matching amino acid sequences encoded by
DNMT3A or
IDH2 hotspot mutations, but not wildtype
DNMT3A or
IDH2 sequences. NetMHCpan version 4.1 was used to predict peptide binding to common HLA alleles [
19].
For parallel reaction-monitoring mass spectrometry (PRM-MS), the samples were lyophilized and resuspended in buffer A containing 4 fmol/uL of each heavy peptide. HLA–eluates were injected together with a mix of heavy labeled peptides (20 fmol each). The Orbitrap Fusion LUMOS mass spectrometer was operated in the PRM mode. DNMT3A (YTD
VSNMSH, YTDVSNMSH
LA) and IDH2 (S
PNGTIQNIL) peptides were monitored (heavy amino acids in bold and underlined).
Table S2 shows the selected peptides, transitions, and collision energies. The Q1 isolation width was 1.2 Da and the MS2 resolution was 30,000 at an AGC target value of 1 million and with a maximum fill time of 100 ms. The gradient was the same as that used for the discovery mode experiments. PRM data analysis and data integration were performed in Skyline 3.6.0.10493.
2.6. Antibodies, Peptide–HLA Tetramers, and Flow Cytometry
For T-cell isolation, PBMCs were stained with CD8 Alexa Fluor 700 (Invitrogen (Waltham, MA, USA)/MHCD0829), CD4 FITC (BD (Franklin Lakes, NJ, USA)/555346), CD14 FITC (BD/555397), CD19 FITC (BD/555412), and in-house generated PE-conjugated pHLA tetramers produced as outlined previously with minor modifications [
20]. T-cell clones were stained with pHLA tetramers for 15 min at room temperature. AML samples were stained with HLA class I FITC (Bio-Rad (Hercules, CA, USA)/MCA81), CD54/ICAM1 APC Fire 750 (BioLegend (San Diego, CA, USA)/353121), CD58/LFA-3 BUV395 (BD/752794), and CD102/ICAM2 BB700 (BD/746179). For the cytotoxicity assays, the AML cells were stained with CD33 PE (Biolegend/397225), CD34 APC (BD/343510), CD14 V450 (BD/561390), and CD206 BV711 (Biolegend/321136); to exclude the T-cells, CD3 BV421 (BD/562427) and CD8 BV421 (Biolegend, 344748) were used. Conventional flow cytometry was performed on a BD FACS LSR-II 4L Full (BD Biosciences, San Jose, CA, USA) using BD FACSDiva version 6 software. Spectral flow cytometry was performed on a 5L Cytek Aurora flow cytometer (Cytek Biosciences, Fremont, CA, USA). Raw spectral flow cytometry data were unmixed using SpectroFlo (Cytek Biosciences). Data were analyzed using OMIQ flow cytometry software (Dotmatics,
www.omiq.ai).
2.7. Isolation of Neopeptide-Specific CD8 T-Cells
The PBMCs were stained with a mix of PE-conjugated pHLA tetramers for 1 h at 4 °C, washed, and then isolated using magnetic anti-PE beads according to the manufacturers’ instructions (Miltenyi Biotec). Isolated T-cells were stained with antibodies, and single pHLA-tetramer-positive CD8 T-cell clones were sorted with a BD FACSAria III cell sorter (BD Biosciences) using BD FACSDiva version 6 software in a 96-well round-bottomed plate containing 5 × 104 irradiated PBMCs (40 Gy) in 100 µL TCM with 0.4 µg/mL PHA.
2.8. T-Cell Reactivity Assays
T-cell clones (4000 T-cells/well) were tested for their reactivity against various stimulator cells (10,000 cells/well) after overnight co-incubation assays in 384-well plates via an IFNγ ELISA (R&D Systems, Minneapolis, MN, USA). For measuring reactivity against AML cell lines and patient samples with DNMT3AR882H or IDH2R140Q mutations, higher T-cell numbers (50,000 T-cells/well) were co-incubated with stimulator cells (100,000 cells/well) in 384-well plates, and the IFNγ release was measured in another ELISA (Diaclone, Besançon, France). The AML samples were thawed one day before the flow cytometry and co-incubation with T-cells.
In flow-cytometry-based killing assays, the AML samples were co-incubated with T-cells directly after thawing in transparent IMDM with FBS, human serum, L-glutamine, pen/strep, and 100 IU/mL IL-2. Target cells (50,000 cells/well) were co-incubated with T-cells (50,000 or 150,000 cells/well) in duplicate for 48 h in 96-well plates. Cells were washed and stained with Zombie-Red viability dye (Biolegend, San Diego, CA, USA) for 15 min at room temperature followed by washing and incubation in 10 µL PBS with 2.5% human serum (block). After 15 min, the cells were stained with antibodies for 30 min at 4 °C, washed, and resuspended in 85 μL. A 30 µL amount of each sample was recorded on a 5L Cytek Aurora flow cytometer (Cytek Biosciences, Fremont, CA, USA). Specific lyses were determined by measuring the number of viable target cells per sample. T-cells were excluded from this analysis.
2.9. Statistical Analyses
Statistical analyses were performed using Prism software V9.3.1. (GraphPad Software, San Diego, CA). p-values < 0.05 were considered significant. The mean numbers of alive cells after the co-incubation killing assays were compared with unpaired t-tests adjusted for multiple comparison according to the Benjamini method. Significance levels are indicated as * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001.
4. Discussion
AML cells often harbor hotspot mutations in the DNMT3A, IDH1, or IDH2 genes involved in DNA (hydroxy)methylation. In this study, we explored whether these AML hotspot mutations encode HLA class I-binding neoantigens. Immunopeptidomics via tandem mass spectrometry revealed three peptides that can be processed and presented on the cell surface. For two of these peptides, i.e., DNMT3AR882H-derived YTDVSNMSHLA and IDH2R140Q-derived SPNGTIQNIL, pHLA tetramers were generated to search for specific T-cells. Of all isolated clones, one T-cell clone for DNMT3AR882H showed specific recognition and lysis of patient-derived AML samples, thereby confirming the peptide as a neoantigen in AML that has the potential to be targeted with immunotherapy.
Of the various single-amino-acid substitutions in DNMT3A-R882, IDH1-R132, IDH2-R140, and IDH2-R172, 46 peptides showed predicted binding to common HLA class I alleles when analyzed using NetMHCpan4.1. Three of these peptides were identified with DDA-MS on the EBV-B cell lines in which hotspot mutations were introduced as minigenes. Using PRM-MS, which is a highly sensitive technique that allows the detection of a few pHLA surface complexes [
24], two of these peptides were also shown to be present on AML cell lines transduced with the mutant gene. However, the surface presentation of these neopeptides could not be directly demonstrated on patient-derived AML cells with the mutation. Other studies have also used mass spectrometry and demonstrated that IDH2
R140Q-derived SPNGTIQNIL is present on COS-7 cells transfected with HLA-B*07:02 and the
IDH2R140Q gene [
25]. In contrast to SPNGTIQNIL, we were able to detect the wildtype IDH2 peptide SPNGTIRNIL on patient-derived AML cells, which is in line with the stronger HLA-B*07:02 binding predicted for this peptide (
Table S3). Despite it not being detected with mass spectrometry, we successfully isolated a T-cell clone that specifically reacts to HLA-A*01:01- and
DNMT3AR882H-positive patient-derived AML samples, thereby confirming that YTDVSNMSHLA is an HLA-A*01:01-binding neoantigen. This demonstrates that although HLA class I peptidomics via mass spectrometry is a reliable technique for measuring neopeptides, a lack of detection does not necessarily indicate absence on the cell surface.
To investigate whether neoantigens derived from DNMT3A
R882H or IDH2
R140Q can be recognized on target cells by T-cells, we isolated and tested T-cell clones sequentially against cell types increasingly resembling patient-derived AML cells. The clones were tested stepwise against EBV-B cells pulsed with mutant or wildtype peptides, cell lines transduced with full-length mutant or wildtype genes, AML cell lines carrying the mutation naturally or with the mutation introduced by CRISPR-Cas9, and finally against patient-derived AML samples with the mutation. Of the 43 and 5 T-cell clones reacting against EBV-B cells pulsed with exogenous YTDVSNMSHLA or SPNGTIQNIL peptides, 5 and 3 T-cell clones recognized cell lines transduced with the
DNMT3AR882H or
IDH2R140Q gene, respectively. In a recent study by Leung et al. [
10], PBMCs from 17 healthy donors were screened for T-cells recognizing neopeptides derived from 14 hotspot mutations in AML. The T-cells were stimulated with mature monocyte-derived dendritic cells pulsed with mixes of 15-mer overlapping peptides. T-cells were detected against IDH1
R132H peptides in 7 donors, against FLT3
D835Y peptides in 6 donors, and against IDH2
R140Q peptides in 11 donors, including SPNGTIQNIL. T-cell recognition of patient-derived AML cells has been demonstrated for two IDH2
R140Q peptides, i.e., IQNILGGTVF in HLA-B*35:43 and TIQNILGGTV in HLA-B*15:01, but not for SPNGTIQNIL in HLA-B*07:02. T-cell recognition of cell lines overexpressing the
IDH2R140Q gene has also been shown by Hwang et al. [
26], who developed CAR-T-cells containing a TCR mimic antibody domain specifically targeting SPNGTIQNIL in HLA-B*07:02. These CAR-T-cells exhibited reactivity against the exogenous peptide with low avidity (EC50 ~10 μg/mL). In comparison, our T-cell clones, particularly clone 1.3H12, displayed a higher avidity (EC50 44 ng/mL). Despite this higher avidity, our three T-cell clones failed to react against
IDH2R140Q patient-derived AML cells, as well as K562 cells engineered with a homozygous
IDH2R140Q mutation through CRISPR-Cas9 genome editing. Our data indicate that T-cell clones capable of recognizing AML cell lines overexpressing the mutant gene often fail to react against patient-derived AML cells. This underscores the importance of validating neoantigens by demonstrating T-cell recognition of tumor cells carrying an endogenous mutation.
Though efficient, our strategy to identify HLA class I-restricted neoantigens encoded by recurrent mutations in AML also has some limitations. First, we searched for specific T-cell clones using pHLA tetramers that were produced for neopeptides identified with mass spectrometry on five selected EBV-B cell lines transduced with a minigene construct with DNMT3A, IDH1, and IDH2 hotspot mutations. Three HLA class I-binding neopeptides were successfully identified, but other neopeptides may have been missed due to inaccurate expressions of the hotspot mutation, inefficient processing and presentation of the neopeptide, or an absence of the relevant HLA allele on the selected EBV-B cell lines. Second, by screening thirty healthy donors, three T-cell clones were isolated that recognized K562 transduced with HLA-B*07:02 and the IDH2R140Q gene. These T-cell clones failed to react against patient-derived AML samples with the relevant HLA and mutation. However, based on these observations, SPNGTIQNIL cannot be excluded as an IDH2R140Q neoantigen in AML, since the affinity of our isolated T-cell clones may have been too low to detect the surface expression of this neoantigen. Similarly, based on the finding that T-cell clones for DNMT3AR882H did not efficiently kill all patient-derived AML cells in vitro, it cannot be concluded that the surface expression of YTDVSNMSHLA on AML cells is too low to be effectively targeted with immunotherapy.
Of the five T-cell clones with specific reactivity against cell lines transduced with
DNMT3AR882H, one T-cell clone was able to recognize and lyse HLA-A*01:01-positive
DNMT3AR882H patient-derived AML cells, demonstrating that YTDVSNMSHLA is a neoantigen in AML.
DNMT3AR882H is considered a founder mutation that is often present in leukemic stem cells, as well as in pre-leukemic hematopoietic stem cells in clonal hematopoiesis [
13]. Therefore, immunotherapy targeting YTDVSNMSHLA in AML may be an effective treatment to prevent relapse. T-cell clone 8.6F10 showed high peptide avidity (EC50 1.33 ng/mL) and specific recognition and lysis of patient-derived AML cells, especially of mature subpopulations. Despite this reactivity profile, we consider the affinity of the TCR to be too low for clinical development, given that clone 8.6F10 did not efficiently kill AML subpopulations with more immature phenotypes. Nevertheless, our data confirm the surface presentation of YTDVSNMSHLA on
DNMT3AR882H AML cells, underscoring the potential significance of this neoantigen as a target for immunotherapy, for instance by treating HLA-A*01:01-positive patients with
DNMT3AR882H AML with high-affinity TCRs or TCR mimic antibodies incorporated into CAR-T-cells or employed in bispecific T-cell engagers [
27,
28].