Close this search box.

Selective haematological cancer eradication with preserved haematopoiesis – Nature

Structural dataset and computational analysis

The experimentally determined 3D structure of the CD45 extracellular domain was retrieved from the PDB (5FMV)22. The per-residue relative solvent accessibility area was computed using a previously published algorithm45 implemented in FreeSASA46 using default parameters. Prediction of B-cell epitopes was based on BepiPred-2.0 (ref. 47) using the default threshold (0.5) for epitope residues nomination. The EV mutation sequence analysis framework48 was used to search for the CD45 sequence with the non-redundant UniProtKB database49. A multiple-sequence alignment was built using five iterations of the jackhammer HMM search algorithm50 with default significance score for the inclusion of homologous sequences.

Plasmid cloning

For sgRNA cloning into the px458 host vector (a gift from F. Zhang) (Supplementary Table 1), forward and reverse primers containing complementary CRISPR RNA (crRNA) sequences flanked by BbsI restriction sites were used (Supplementary Table 4). The px458 plasmid was double digested with AgeI-HF (NEB, R3552S) and EcoRI-HF (NEB, R3101S) to eliminate the regions coding for GFP and Cas9. The px458 vector was then digested using BbsI (ThermoFisher Scientific, ER1012), gel purified and ligated with the phosphorylated and annealed crRNA oligonucleotides (called sgRNA plasmid once cloned).

To transiently overexpress CD45 mutants, we introduced each variant of interest in a plasmid expressing WT CD45RO. Briefly, we digested the pCD45RABC plasmid (Sino Biological) (Supplementary Table 1) with HindIII-HF (NEB, R3104S) and XcmI (NEB, R0533L) to remove the alternative spliced exons A, B and C. The point mutations of the human CD45 variants were then introduced into the plasmid expressing CD45RO using PCR (Supplementary Table 4).


All ligations were transformed in JM109-competent bacteria (Promega, P9751). BE plasmids were from Addgene (SPACE-NG, ABE8e-NG, ABEmax-SpRY, ABEmax-SpG, CBE4max-NG, CBE4max-SpG and xCas9(3.7)-BE4) (Supplementary Table 1).

BE mRNA and sgRNAs

ABE8e-NG mRNA (capped (cap 1) using CleanCap AG; fully substituted with 5-methoxy-U; 120A polyA tail) and ABE8e(TadA-8e V106W)-SpRY mRNA (capped (cap 1) using CleanCap AG 3′-O-methylation; fully substituted with N1-methyl-pseudo-U; 80A polyA tail) were from Trilink Biotechnologies and Tebu-bio. We used 100-base lyophilized chemically modified sgRNAs from Synthego using their CRISPRevolution sgRNA EZ Kit service and resuspended at 100 µM (3.2 µg µl−1) in 1× TE buffer from Synthego (10 nM Tris, 1 mM EDTA, pH 8.0; chemical modifications include 2′-O-methylation of the three first and last bases and 3′ phosphorothioate bonds between the first three and last two bases of each sgRNA).

Genomic DNA extraction, PCR and Sanger sequencing

Cells from the BE plasmid screening were lysed in tail lysis buffer (100 mM Tris pH 8.5, 5 mM Na-EDTA, 0.2% SDS, 200 mM NaCl) containing proteinase K (Sigma-Aldrich) at 56 °C (1,000 rpm) for 1 h. The DNA was precipitated with isopropanol (1:1 volume ratio) and washed in 70% ethanol. The DNA was then resuspended in H2O and the genomic DNA concentration was measured with a NanoDrop device (Thermo Fisher).

For samples containing few cells, genomic DNA was extracted using QuickExtract (Lucigen, QE09050). Cell pellets were resuspended in 30 µl QuickExtract, incubated at 60 °C for 6 min, vortexed for 1 min and subsequently re-incubated at 98 °C for 10 min.

PCR was performed using GoTaq G2 Green Master Mix (Promega, M782B). The gDNA of samples analysed by NGS was extracted using QuickExtract (Lucigen, QE09050) or the Quick-DNA 96 Plus kit (Zymo, D4070) and the genomic DNA concentration was measured with a Qubit device (Thermo Fisher).

For Sanger sequencing, different PCR primers were used depending on the CD45 exon targeted by the sgRNA and the sequencing technology (Supplementary Table 4). Sequencing of PCR amplicons was done at Microsynth and sequencing chromatograms were analysed using the EditR R package51 to quantify BE efficiencies.

Next-generation amplicon sequencing

For NGS, targeted amplicon libraries were generated using a three-step PCR protocol. In brief, nested PCRs were done on genomic DNA samples using KAPA HiFi HotStart polymerase (Roche) (Supplementary Table 4). After Illumina barcoding (Nextera indices, Illumina) using KAPA HiFi HotStart polymerase (Roche), PCR samples were pooled, purified using AMPure XP beads (Beckman Coulter) and quantified using Qubit dsDNA HS assay kit (Thermo Fisher). Libraries were paired-end sequenced on an Illumina Miniseq instrument using the Illumina Miniseq Mid output kit (300 cycles) with 50% PhiX spike-in (Illumina). After demultiplexing, each sample was assessed for quality using FastQC52 and processed using the CRISPResso2 tool53. For each of the samples, we provided the reference amplicon sequence (hg38) and the guide RNA sequence (reverse complement) and defined the quantification window centre to −10, the quantification window size to 15 and the plot window size to 30. We applied minimum paired end reads overlap between 10 and 200 and provided the following Trimmomatic sentence: ILLUMINACLIP:NexteraPE-PE:2:30:10 LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:36. Finally, we used a custom R script ( to count and translate into amino acid each allele from the CRISPResso2 output file Alleles_frequency_table.txt in the quantification window. Alleles with less than 0.8% frequency were considered as ‘other’.

Genetic determination of human chimerism

Genetic discrimination between PDX- and HSPC-derived cells was performed using the Devyser Chimerism NGS kit (Devyser) according to the manufacturer’s recommendations. In brief, sequencing libraries were prepared from genomic DNA targeting 24 polymorphic insertion–deletion markers distributed on 16 different chromosomes. The libraries were sequenced on a MiniSeq (Illumina) instrument using the high-throughput flow cell, generating 74-bp pair-end reads. An informative marker set was defined for the PDX donor/HSPC donor pair. It consisted of 10 markers that reliably discriminated between DNA from PDX and HSPC donor cells. The average proportion of leukaemia donor-specific reads to total reads was calculated to determine the proportion of PDX cells in each sample. It was confirmed that the genomic DNA of the mouse host did not interfere with the analysis.


Genomic DNA was extracted from human peripheral blood mononuclear cells (PBMCs) using the Puregene tissue kit (Qiagen, 158063) according to the manufacturer’s instructions including the proteinase-K and RNase steps (Qiagen, 158143 and 158153). CHANGE-seq-BE was adapted and modified from the original CHANGE-seq method54 to validate genome-wide activity for ABEs28. Similar to CHANGE-seq, purified genomic DNA tagmented with a custom Tn5-transposome to generate an average length of 650 bp and followed by gap repair with Kapa HiFi HotStart Uracil + DNA Polymerase (KAPA Biosystems, KK2802) and Taq DNA ligase (NEB, M0208L). Gap-repaired DNA was treated with USER enzyme (NEB, M5505L) and T4 polynucleotide kinase (NEB, M0201L). Intramolecular circularization of the DNA was performed with T4 DNA ligase (NEB, M0202L) and residual linear DNA was degraded by a cocktail of exonucleases containing plasmid-safe ATP-dependent DNase (Lucigen, E3110K), lambda exonuclease (NEB, M0262L), exonuclease I (NEB, M0293L) and exonuclease III (NEB, M0206L). The circularized DNA was then treated with Quick CIP (NEB, M025L) to dephosphorylate 5′ and 3′ ends of any residual linear DNA. Circularized genomic DNA (125 ng) was treated with ABE8e–SpRY:sgRNA-49.3 complexes in vitro in a 50 μl reaction for 24 h at 37 °C. ABE RNP complexes nicked the targeted DNA strand and deaminated adenine bases to inosine in the non-targeted stranded DNA of both on- and off-target sites. Further enzymatic steps were included with ABE treatment in the CHANGE-seq-BE method to generate double-strand breaks. Nicked DNA circles were treated with endonuclease V in 10× NEB buffer 4 (NEB, M0305S). Endo V cleaved DNA adjacent to inosines to generate linear DNA with 5′ overhangs. Gaps were filled with klenow fragments (3′ > 5′ exo) and deoxyribonucleotide triphosphates (dNTPs) (NEB, M0212L) in NEB buffer 2. End-repaired DNA products were A-tailed and further ligated with a hairpin adapter using an HTP library preparation kit (Kapa, KK8235), USER treated and amplified by PCR-barcoded universal primers with NEBNext multiplex oligonucleotides for illumina (NEB, E7600S), using Kapa HiFi HotStart uracil master mix. PCR libraries were quantified by quantitative PCR (KAPA Biosystems, KK4824) and sequenced with 151-8-8151 cycles on an Illumina NextSeq 2000 instrument. CHANGE-seq-BE data analyses were performed using open-source software:


Validation of off-target sites was performed using the rhAmpSeq system from IDT. rhAmpSeq primer panels for targeted amplification were generated using the rhAmpSeq design tool defining the insert size between 150 and 250 bp. Applied primer sequences are listed in Supplementary Table 4. A rhAmpSeq CRISPR library was prepared according to the manufacturer’s instructions and sequenced on an Illumina MiniSeq instrument (MiniSeq high output kit, 300 cycles). Custom python code and open-source bioinformatic tools were used to analyse rhAmpSeq data. First, we generated FASTQ format files by demultiplexing high-throughput-sequencing BCL data files. Next, the reads were processed using CRISPRessoPooled (v.2.0.41) with quantification_window_size 10, quantification_window_center −10, base_editor_output, conversion_nuc_from A, conversion_nuc_to G. The allele frequency table from the output files was used to calculate the A•T-to-G•C editing frequency. Specifically, the editing frequency for each on- or off-target site was defined as the ratio between the number of reads containing the edited base (that is, G) in a window from positions 4 to 10 of each protospacer (where the GAA PAM is positions 21–23) and the total number of reads. To calculate the statistical significance of off-target editing, we applied a method previously described55. In brief, a 2-by-2 contingency table was constructed using the number of edited reads and the number of unedited reads in the treated sample and its corresponding control sample. Next, a χ2 test was done. The FDR was calculated using the Benjamini–Hochberg method. Significant off-targets were defined on the basis of: first, FDR ≤ 0.05 and second, the difference in editing frequency between treated and control (≥1%).

Cell line culture conditions

All cancer cell lines (listed in Supplementary Table 5) were cultured in RPMI-1640 (Sigma-Aldrich, R8758) supplemented with 10% heat-inactivated FCS (Gibco Life Technologies) and 2 mM GlutaMAX (ThermoFisher Scientific, 35050061) at 37 °C.

All cell lines were retrovirally transduced with MI-Luciferase-IRES-mCherry (gift from X. Sun; Supplementary Table 1). Cells were then FACS-sorted on the basis of mCherry expression. After expansion, MOLM-14 and OCI–AML2 were profiled for short tandem repeats and tested negative for mycoplasma before being frozen until further use. Jurkat and NALM-6 were purchased from ATCC and were therefore not profiled for short tandem repeats.

DF-1 cells were cultured in DMEM high-glucose medium (Sigma-Aldrich, D5796) supplemented with 10% non-heat-inactivated FCS and 2 mM GlutaMAX at 39 °C (Supplementary Table 5).

Human primary T cell culture and activation conditions

Leukocyte buffy coats from anonymous healthy human donors were purchased from the blood-donation centre at Basel (Blutspendezentrum SRK beider Basel, BSZ). PBMCs were isolated by density centrifugation using SepMate tubes (StemCell Technologies, 85450) and the density gradient medium Ficoll-Paque (GE Healthcare) according to the manufacturer’s protocol. Frozen PBMCs were thawed and human primary T cells were then isolated using an EasySep human T cell isolation kit (Stemcell Technologies, 17951) following the manufacturer’s protocol. T cells were cultured overnight without stimulation at a density of 1.5 × 106 cells per ml in RPMI-1640 medium supplemented with 10% heat-inactivated human serum (AB+, male; purchased from BSZ), 10 mM HEPES (Sigma-Aldrich), 2 mM GlutaMAX, 1 mM sodium pyruvate, 0.05 mM 2-mercaptoethanol and 1% MEM non-essential amino acids (all from Gibco Life Technologies). The next day, the human primary T cells were activated with interleukin-2 (IL-2) (150 U ml−1, proleukin, University Hospital Basel), IL-7 (5 ng ml−1, R&D Systems), IL-15 (5 ng m−1, R&D Systems) and Dynabead Human T-Activator CD3/CD28 (1:1 beads:cells ratio) (Gibco, 11132D). The activated cells were de-beaded before electroporation.

hCD34+ HSPC isolation and culture conditions until electroporation

Leukopaks were purchased from CytoCare and hCD34+ HSPCs were isolated by the LP-34 process using CliniMACS Prodigy (Miltenyi). Isolated hCD34+ HSPCs were thawed and grown in HSPC medium for two days until electroporation (StemSpan SFEM II (StemCell, 09655) supplemented with 100 ng ml−1 human stem cell factor (hSCF) (Miltenyi, 130-096-695), 100 ng ml−1 human FMS-like tyrosine kinase ligand (hFlt3)-ligand (Miltenyi, 130-096-479), 100 ng ml−1 human thrombopoietin (hTPO) (Miltenyi, 130-095-752) and 60 ng ml−1 hIL-3 (Miltenyi, 130-095-069).

Electroporation conditions

K562 cells (2 × 106) were resuspended in buffer T and mixed with 5 µg BE plasmid (Supplementary Table 1) and 1.5 µg sgRNA plasmid for co-electroporation using a Neon transfection system (ThermoFisher, MPK10096; 1,450 V, 10 ms, 3 pulses). To monitor the electroporation efficiency, GFP expression was evaluated 24 h after electroporation using an optical microscope. In Extended Data Fig. 1f, BE results are displayed using a custom BE score: log((sum of editing frequencies per condition/number of edited positions per condition)+1).

De-beaded human activated T cells (1 × 106) were resuspended in 100 µl Lonza supplemented P3 electroporation buffer with 7.5 µg BE mRNA (1 µg µl−1) and 7.5 µg sgRNA (3.2 µg µl−1) (Supplementary Table 2) and electroporated using the 4D-Nucleofector system (Lonza) with program EH-115. Immediately after electroporation, 900 µl pre-warmed human T-cells medium was added directly in the cuvettes and incubated for 20 min at 37 °C for the T cells to recover. Cells were then transferred in 48-well flat-bottomed plates (Corning, 3548) and the medium was supplemented with 500 U ml−1 IL-2. The medium was renewed every two days.

hCD34+ HSPCs (1 × 106) were electroporated 48 h after thawing with 7.5 µg BE mRNA (1 µg µl−1) and 13.6 µg sgRNA (3.2 µg µl−1) (Supplementary Table 2) with a 1:100 BE:sgRNA molar ratio, following the same protocol as for human T cells but with program CA-137. Electroporated hCD34+ HSPCs were kept in culture at 0.5 × 106 cells per ml in a six-well flat-bottomed plate (Corning, 3516) in HSPC medium supplemented with 100 ng ml−1 hSCF, 100 ng ml−1 hFlt3–ligand and 100 ng ml−1 hTPO for in vivo applications and with the addition of 60 ng ml–1 hIL-3 for in vitro assays. The medium was renewed every five days. Edited hCD34+ HSPCs prepared for in vivo injection were frozen two days after electroporation in cryo-preservation CryoStor CS10 medium (Stem Cell Technologies, 07930) at a density of 10 × 106 cells ml−1.

CFU assays

The CFU assay was started 72 h after gene editing. For each condition, 1.1 ml semi-solid methylcellulose medium (StemCell Technologies) containing 200 cells was plated in a well of a SmartDish (StemCell Technologies, 27370) in duplicates. The cells were incubated at 37 °C, for 14 days. The resulting progenitor colonies were counted and scored using STEMVision Analysis (StemCell Technologies) according to the manufacturer’s instructions. The mean of the total number of colonies in the NTC samples for each experiment was set as 1.

DF-1 cell-transfection conditions

Plasmid (6.5 µg) encoding WT hCD45RO or its variants was mixed with 200 µl serum-free DMEM medium and 19.5 µl polyethylenimine (1 mg ml−1; Chemie Brunschwig, POL23966-100). The transfection mix was added dropwise to 1 × 106 DF-1 cells plated the day before in a six-well plate. Cells were analysed 48 h later by flow cytometry.

In vitro ADC-mediated killing assays

For in vitro ADC killing assays, 5,000 base-edited human activated T cells or 25,000 base-edited hCD34+ HSPCs were plated five days after electroporation in 96-well plates (flat-bottomed for T cells and round-bottomed for HSPCs; Corning 3596 and 3799, respectively) in 100 µl of corresponding medium (supplemented with only 50 U ml−1 of IL-2 for human T cells). For ADC killing assays involving saporin, a 100 nM stock was prepared by incubating the biotinylated antibody (BC8 or MIRG451 mAbs) and saporin–streptavidin (ATS-Bio, IT-27-1000) at a 1:1 molar ratio for 30 min at room temperature.

For in vitro ADC killing of co-cultures, 12,500 Jurkat cells were stained for 20 min with CTV (Invitrogen, C34557A) at 37 °C and then seeded at a 1:1 cell ratio with 12,500 base-edited hCD34+ HSPCs five days post-electroporation in 96-well round-bottom plates in 100 µl HSPC medium with corresponding concentrations of CIM053–SG3376 (ADC Therapeutics). Cells were incubated for 72 h at 37 °C, stained for flow cytometry or cell sorting and analysed using a BD LSRFortessa. Genomic DNA was extracted for sequencing.

For in vitro ADC killing of mCherry–luciferase-marked tumour cell lines (Jurkat, NALM-6, OCI-AML-2 and MOLM-14), 2,000 cells were plated in 384-well plates in medium with or without 30 min pre-incubation at 37 °C with 50 µg ml−1 (333.33 nM) naked CIM053 antibody (40 µl final total volume per well). Following a 72 h incubation period, 5 μl firefly d-luciferin (0.75 mg ml−1 resuspended in medium (Biosynth, L-8220)) was added to each well and incubated for 5 min at room temperature. Luminescence readouts were recorded using a BioTek Synergy H1 plate reader.

Expression and purification of soluble CD45wt and CD45 variants

For precise antibody–protein affinity measurements and biophysical characterization, CD45wt and variants containing only D1 and D2 of the ECD were produced. The protein sequence (residues 225–394) was histidine tagged at the carboxy terminus and contains few N- and C-terminal added amino acids (full WT sequence: ETGIEGRKPTCDEKYANITVDYLYNKETKLFTAKLNVNENVECGNNTCTNNEVHNLTECKNASVSISHNSCTAPDKTLILDVPPGVEKFQLHDCTQVEKADTTICLKWKNIETFTCDTQNITYRFQCGNMIFDNKEIKLENLEPEHEYKCDSEILYNNHKFTNASKIIKTDFGSPGEGTKHHHHHH, SEQ ID 57, Uniprot ID P08575). Expi293F GnTI cells (Thermo Fisher, A39240) that lack N-acetylglucosaminyltransferase I (GnTI) activity and therefore lack complex N-glycans were used for protein expression. After collection, the protein was purified using Ni-NTA chromatography followed by digestion of high-mannose glycans with endoglycosidase H (EndoHf; New England BioLabs, P0703S) at 37 °C overnight. EndoHf was removed from the protein solution with amylose resin and the CD45 protein was further purified by size-exclusion chromatography in buffer comprising 20 mM HEPES, pH 7.4, 150 mM NaCl. Peak monomer and dimer fractions (where needed) were concentrated using a 10 kDa cut-off Amicon centrifugal filter (UFC8010) and protein aliquots were flash-frozen in liquid nitrogen before storage at −150 °C. Variant CD45 proteins were produced using the same experimental procedure. The monomer content percentage for each protein was taken from the size-exclusion chromatogram.

Binding analysis of soluble CD45 proteins by BLI

Analysis of MIRG451 and BC8 binding to the selected variants was performed on an Octet system RED96e (Sartorius) or R8 (Sartorius) at 25 °C with shaking at 1,000 rpm using 1× kinetic buffer (Sartorius, 18-1105). The selected variants were screened for their ability to bind to MIRG451 and BC8 using different concentrations of CD45 (WT or variant). MIRG451 was captured by an anti-human Fc-capture biosensor (AHC) (Sartorius, 18-5060) for 300 s at 0.5–1 µg ml−1. As an analyte, human CD45wt and variants, containing only domains 1 and 2, were titrated at seven different concentrations, from 1,000 nM to 15.6 nM or from 50 nM to 0.78 nM with a 1:2 dilution series. Association of the analyte to MIRG451 was monitored for 600 s and dissociation of the analyte from MIRG451 was monitored for 1,800 s. Reference subtraction was performed against buffer-only wells. AHC tips were regenerated using 10 mM Gly-HCl, pH 1.7. Data were analysed using the Octet data analysis software HT 12.0. Data were fitted to a 1:1 binding model. Kinetic rates Ka and Kd were globally fitted.

To analyse binding to BC8, streptavidin biosensors (Sartorius, 18-5020) were first coated with CaptureSelect biotin anti-LC-κ (murine) conjugate (Thermo Scientific, 7103152100) for 600 s at 1 µg ml−1. BC8 was then captured by the coated streptavidin biosensors for 300 s at 0.5–1.0 µg ml−1. Analyte titration was performed as for MIRG451. Association of the analyte to BC8 was monitored for 300 s and dissociation of the analyte from BC8 was monitored for 900 s. Reference subtraction, regeneration and data analysis were performed as for MIRG451.

Characterization of CD45 variants by nanoDSF

The thermostability of CD45 D1–D2 variants was analysed by differential scanning fluorimetry and monitoring tryptophane fluorescence using Nanotemper Prometheus NT.48 NanoDSF or a Nanotemper Prometheus Panta (NanoTemper Technologies)56,57,58. Thermal denaturation was monitored by tryptophane/tyrosine fluorescence at 350 and 330 nm and an excitation wavelength of 280 nm was used. CD45wt and variants were prepared at 0.25–1.0 mg ml−1 in 20 mM HEPES, 150 mM NaCl, pH 7.4. Then 10 μl was put into the capillaries and placed into the sample holder. Each protein was measured in triplicates per experiment and the CD45wt was measured in four different experiments. The temperature was increased from 20 °C to 90 °C or 95 °C. The analysis was performed using the ratio of the fluorescent intensities at 350 and 330 nm. The software of the instrument was used to calculate Tonset and TM as well as the mean and s.d. of the triplicates. The melting temperature was determined as the inflexion point of the sigmoidal curve and compared with that of CD45wt.

Flow cytometry analysis and sorting

Flow cytometry was done on BD LSRFortessa instruments with BD FACSDiva software. Data were analysed with FlowJo software. Antibodies used for flow cytometry are listed in Supplementary Table 6. Cells were sorted with BD FACSAria or BD FACSMelody cell sorter instruments. Sorted cells were resuspended in 30 µl QuickExtract. PCRs were performed and sent for Sanger sequencing.

CD45 expression of AML samples

The CD45 expression of 27 people diagnosed with AML at University Hospital Basel was assessed using routinely acquired flow cytometry data as part of the diagnostic work-up. Gating for AML blasts, lymphocytes and erythrocytes was performed manually using FlowJo 10.10.0. Owing to the experimental set-up (threshold for SSC-A and FSC-A to exclude debris), a distinct erythrocyte population could not be distinguished in all samples (23 of 27). Data were analysed with GraphPad Prism 10 and statistical significance was calculated using mixed-effects analysis. All patients gave written informed consent to the analysis of clinical data for research purposes and the study was approved by the local ethics committee (BASEC-Nr 2023-01372).

Animal experiments

All animal work was done in accordance with the federal and cantonal laws of Switzerland. Protocols were approved by the Animal Research Commission of the Canton of Basel-Stadt, Switzerland. All mice were housed in a specific pathogen-free condition in accordance with institutional guidelines and ethical regulations. NBSGW (stock 026622) female mice were purchased from Jackson Laboratories. HSPCs were edited as described above. Two days after electroporation, cells were collected and frozen in CryoStor CS10 medium. Cells were thawed on the day of injection, washed and resuspended in PBS. Recipient NBSGW female mice (4 weeks old) were injected intravenously into the tail vein with HSPCs (the number of cells injected varied between 0.6 and 1.1 million and is indicated in each figure). Chimerism was analysed by flow cytometry in blood after ten weeks. Mice were treated with saline or CIM053–SG3376 at the dose(s) and intervals indicated in each figure. For tumour experiments in humanized mice, 1 × 106 MOLM-14–mCherry–luc cells were injected into the tail vein. Then, 10 or 12 days after tumour inoculation, the mice were treated with saline or 1 mg per kg CIM053–SG3376. The mice received a second antibody dose of 0.5 mg per kg CIM053–SG3376 10 or 25 days after the first dose. Mice were euthanized 43 or 45 days after tumour inoculation or when reaching the maximum allowed clinical score. To monitor tumour growth, mice were injected intraperitoneally with 100 μl d-luciferin (BioSynth, L-8220) and were subjected to Newton7.0 imaging (Vilber).

For secondary transplant, NSG–SGM3 female mice (stock 013062) were purchased from Jackson Laboratories. Recipient mice were irradiated the day before the BM transplant with 200 cGy. Primary transplant mice were euthanized, the BM was isolated and 40% of it was re-injected into the new host. Mice from secondary transplants were euthanized 8 weeks after humanization.

MOLM-14–mCherry–luc (1 × 106), OCI-AML-2–mCherry–luc (2 × 106), Jurkat–mCherry–luc (5 × 106) or NALM-6–mCherry–luc (0.5 × 106) cells were injected into the tail vein of NBSGW mice. After tumour inoculation, mice were monitored regularly (for behaviour, weight and imaging). Mice were treated with saline, control-SG3376 or CIM053–SG3376 at the dose and intervals indicated in the relevant figure and euthanized 21 days after tumour inoculation or when reaching the maximum allowed clinical score.

Deidentified patient-derived AML samples were obtained from the PDX repository59,60 (Cancer Research Center of Toulouse, France). Signed written informed consent for research use in accordance with the Declaration of Helsinki was obtained from patients and approved by the Geneva Health Department Ethic Committee. PDX cells (0.6 × 106) were injected into the tail vein of humanized NBSGW mice (8 weeks after HSPC injection). The weight of the mice was monitored regularly. Mice were treated with saline or CIM053–SG3376 at the doses and intervals indicated in the figure and the mice were euthanized 54 days after tumour inoculation. Some control mice were euthanized 3 days before antibody treatment.

Tissue collection and processing

After the mice were euthanized, 0.2 ml blood, both hind legs (femur and tibia) and the spleen were collected from each mouse. Cell suspensions were generated, red blood cells were lysed using ACK lysis buffer and then the cell suspensions were filtered. For tumour experiments, organs were collected on the day of euthanasia, single cell suspensions were generated and frozen in cryo medium. Samples from all mice were thawed and stained on the same day to minimize experimental variability. Cells were stained for different antigens and 30 μl Accucheck counting beads (1,066 microspheres per μl; Invitrogen, PCB100) were added to each sample and the results were analysed by FACS using a BD LSRFortessa instrument.

Statistics and reproducibility

Statistical analyses were done using GraphPad Prism 9 and 10 software. In all figure legends, n refers to the number of experimental replicates. For multiple comparisons, two-way ANOVA tests were used with significance levels indicated. Data are presented as mean ± standard deviation. Survival curves were analysed using the log-rank Mantel–Cox test. rhAmpSeq was analysed using a χ2 test. The FDR was calculated using the Benjamini–Hochberg method.

Some data points of the in vivo experiments were excluded after visual inspection of samples if the FACS time gate showed irregularities. One mouse that did not engraft HSPCs was excluded from Fig. 5 and Extended Data Fig. 10. Cell numbers in the sgNTC group treated with CIM053–SG3376 were so low that analysis of some assays became unreliable (NGS, genetic chimerism analysis). We therefore excluded this group from NGS.

The number of biological replicates is specified for each experiment in the relevant figure legend. Several key experiments were performed by different people at times in different laboratories, and reagents were shared. For instance, identification of variants, characterization of recombinant variants and FACS validation were performed by different people. Some experiments were performed in the academic lab and validated in Cimeio labs and vice versa. To avoid unconscious bias when assigning mice to saline or the CIM053–SG3376 groups, we always assigned the mice with the largest tumour mass to the ADC group.

The investigator who determined genetic chimerism (NGS and analysis) was blinded and provided the results to the investigator in charge of supervising in vivo experiments. The people who performed CHANGE-Seq_BE, rhAMPSeq and analysed the data were blinded and provided the results to the investigator in charge of supervising the in vivo experiments.

Availability of materials

Non-proprietary materials are freely available on reasonable request. Restrictions apply to proprietary, commercial material.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.