A human fetal liver-derived infant MLL-AF4 acute lymphoblastic leukemia model reveals a distinct fetal gene expression program

Although 90% of children with acute lymphoblastic leukemia (ALL) are now cured, the prognosis for infant-ALL remains dismal. Infant-ALL is usually caused by a single genetic hit that arises in utero: an MLL/KMT2A gene rearrangement (MLL-r). This is sufficient to induce a uniquely aggressive and treatment-refractory leukemia compared to older children. The reasons for disparate outcomes in patients of different ages with identical driver mutations are unknown. Using the most common MLL-r in infant-ALL, MLL-AF4, as a disease model, we show that fetal-specific gene expression programs are maintained in MLL-AF4 infant-ALL but not in MLL-AF4 childhood-ALL. We use CRISPR-Cas9 gene editing of primary human fetal liver hematopoietic cells to produce a t(4;11)/MLL-AF4 translocation, which replicates the clinical features of infant-ALL and drives infant-ALL-specific and fetal-specific gene expression programs. These data support the hypothesis that fetal-specific gene expression programs cooperate with MLL-AF4 to initiate and maintain the distinct biology of infant-ALL.

A characteristic and baffling feature of MLL-r infant-ALL is the fact that this single oncogenic hit before birth seems to be sufficient to induce a rapidly proliferating therapy-resistant leukemia without the need for additional mutations, unlike many cases of childhood-ALL, which also originate in utero but only develop into full-blown leukemia after a second postnatal hit 23 . One reason for this could be that the specific fetal progenitors in which the translocation arises provide the permissive cellular context necessary to cooperate with MLL-r to induce infant-ALL [24][25][26] .
Here, using the most common MLL-r infant-ALL, MLL-AF4, as a disease model, we identify fetal-specific gene expression programs in primary human hematopoietic cells and show that MLL-AF4 infant-ALL, but not MLL-AF4 childhood-ALL, maintains expression of fetal-specific genes. CRISPR-Cas9 gene editing of primary human fetal liver (FL) CD34+ cells to produce a t(4;11)/MLL-AF4 translocation replicates the clinical features of infant-ALL in a xenograft model and drives infant-ALL-and fetal-specific molecular programs. These data support the hypothesis that developmentally regulated features of human fetal cells cooperate with MLL-AF4 to initiate and maintain the distinct biology of infant-ALL.

Results
MLL-AF4 infant-ALL is molecularly distinct from MLL-AF4 childhood-ALL. Using gene expression profiles from patients with an identical driver mutation (t(4;11)/MLL-AF4), we first set out to determine whether infant-ALL had a distinct molecular profile compared to childhood-ALL. It has already been shown by others that MLL-AF4 leukemias can be divided into different subtypes based on a HOXA lo or HOXA hi gene expression profile [19][20][21][22] . In our analysis, we wanted to determine if there was a unique infant-ALL molecular signature that was independent of these subtypes.
Using a previously published patient bulk RNA-sequencing (RNA-seq) dataset 23 , we carried out differential gene expression analysis between all MLL-AF4 infant-ALL (n = 19) and MLL-AF4 childhood-ALL (n = 5) samples. We identified 617 significantly differentially expressed genes (false discovery rate (FDR) < 0.05), 193 of which were upregulated in MLL-AF4 infant-ALL and therefore represented an infant-ALL-specific gene expression profile (Fig. 1a, Supplementary Fig. 1a, b, Supplementary Data 1). When patients were classified as HOXA lo or HOXA hi based on HOXA9 expression ( Supplementary Fig. 1c), we found that HOXA lo MLL-AF4 infant-ALL and HOXA hi MLL-AF4 infant-ALL did not separate based on these 617 genes, suggesting that this was a true infant-ALL-specific gene expression profile, irrespective of other molecular characteristics (Fig. 1a). In fact, of the top ten most significantly upregulated genes that could separate MLL-AF4 infant-ALL from MLL-AF4 childhood-ALL (Fig. 1b), no significant differences were observed between HOXA lo and HOXA hi MLL-AF4 infant-ALL subsets (Fig. 1c).
Fetal gene expression programs drive the distinct molecular profile of MLL-AF4 infant-ALL. We next sought to determine the extent to which normal fetal gene expression programs contribute to the distinct molecular profile of MLL-AF4 infant-ALL. We compared bulk RNA-seq data for sorted human FL hematopoietic stem and progenitor cell (HSPC) subpopulations previously generated in our lab 27 to a human adult bone marrow (ABM) HSPC RNA-seq dataset 28 . We carried out differential gene expression analysis between comparable subpopulations of FL and ABM HSPCs along the B-lineage differentiation pathway (Fig. 1d). The hematopoietic stem cell (HSC) subpopulation showed the greatest number of differentially expressed genes between FL and ABM (3787 genes), reducing at each subsequent stage of B-lineage differentiation (1509 genes differentially expressed between FL committed B progenitors (CBPs) and ABM common lymphoid progenitors (CLPs)) ( Fig. 1d). A total of 5709 genes were differentially expressed between FL and ABM in at least one HSPC subpopulation when we combined all differentially expressed gene lists ( Fig. 1d and Supplementary Data 2).
We carried out clustering analysis of the patient dataset based on these 5709 genes and found that they were capable of separating MLL-AF4 infant-ALL from MLL-AF4 childhood-ALL ( Fig. 1e and Supplementary Fig. 2a). Comparing differentially expressed genes in both the normal and leukemic setting, we found 72 genes that were significantly upregulated in at least one normal FL HSPC subpopulation and also in MLL-AF4 infant-ALL (~40% of all genes upregulated in MLL-AF4 infant-ALL compared to MLL-AF4 childhood-ALL) (Supplementary Fig. 2b and Supplementary Data 2). These included some of the most significantly upregulated genes in MLL-AF4 infant-ALL such as HOXB4 and IGF2BP1, a member of the LIN28B gene expression pathway that has previously been reported to positively regulate HOXB4 expression ( Supplementary Fig. 2c) 29,30 . Together, these data suggest that the molecular profile of human fetal HSPCs plays a key role in determining the distinct gene expression profile of MLL-AF4 infant-ALL.
A CRISPR-Cas9-induced MLL-AF4 translocation can transform human FL HSPCs in vitro. To test the hypothesis that to accurately model MLL-AF4 infant-ALL, the MLL-AF4 translocation should be expressed in human fetal HSPCs, we directly induced the most common t(4;11)/MLL-AF4 translocation in infant-ALL in primary human FL HSPCs. The translocation was induced in 13-15 postconception week (pcw) human FL CD34+ cells by CRISPR-Cas9 gene editing 31,32 . Single guide RNAs (sgRNAs) were designed to target intron 11 of MLL and intron 3 of AF4 ( Supplementary Fig. 3a), chosen because of their prevalence as the most common breakpoint region in infant-ALL 18    and MLL-AF4 childhood-ALL (chALL (orange), n = 5) based on 617 significantly differentially expressed genes (FDR < 0.05, edgeR exact test; Supplementary Data 1). HOXA lo (light green, n = 11) and HOXA hi (purple, n = 13) MLL-AF4 subsets of infant-ALL and childhood-ALL are annotated. Color scale = log 2 counts per million (logCPM). b Bar plot showing significance (−log10(FDR)) for the ten most significantly upregulated genes in MLL-AF4 infant-ALL. c Expression (log 2 transcripts per million (log 2 TPM)) of the top ten most significantly upregulated genes in MLL-AF4 infant-ALL in HOXA lo MLL-AF4 infant-ALL (iALL (light green), n = 11), HOXA hi MLL-AF4 infant-ALL (iALL (purple), n = 8), and HOXA hi MLL-AF4 childhood-ALL (chALL (orange), n = 5). Data are shown as mean ± SEM (n.s.: p > 0.05; one-way ANOVA with Tukey's correction for multiple comparisons). Source data are provided as a Source Data file. d (left) Schematic representation of differential gene expression analysis between FL and ABM. Equivalent HSPC subpopulations were compared and significantly differentially expressed genes (FDR < 0.05, edgeR exact test) for all four comparisons were combined into a master list of genes that were differentially expressed in at least one HSPC subpopulation (HSC hematopoietic stem cell, MPP multipotent progenitor cell, LMPP lymphoid-primed multipotent progenitor cell, CBP committed B progenitor, CLP common lymphoid progenitor). CD34+ cells was comparable between CRISPR MLL-AF4+ and control cultures from week 3 ( Supplementary Fig. 3d). The majority of CRISPR MLL-AF4+ CD34+ cells were CD19+ B progenitors, suggesting that MLL-AF4-driven B-lineage specification might occur at a progenitor stage (Fig. 2d, right). More detailed immunophenotyping showed that the majority of CRISPR MLL-AF4+ cells were CD34−CD19+CD10+IgM/IgD− preB cells, of which~10% aberrantly expressed the leukemiaassociated marker CD133 (Fig. 2c, d), a direct gene target of the MLL-AF4 fusion protein 33 . By week 7 of coculture, when control cultures no longer produced any detectable human cells, the number of human cells in CRISPR MLL-AF4+ cultures began to decline ( Fig. 2a and Supplementary Fig. 3c), suggesting that MS-5 stroma may not be optimal for long-term maintenance of FLderived CRISPR MLL-AF4+ cells.  Table 1 and Supplementary Fig. 4d) confirmed the presence of a heterozygous MLL-AF4/t(4;11) translocation in CRISPR MLL-AF4+ cells. Greater than 87% of cells were positive for MLL-AF4 by FISH (range 87-99%, n = 3) ( Supplementary  Fig. 4c). In addition, Sanger sequencing and FISH analysis confirmed that the wt allele of MLL had neither gained indels that would affect its expression (Supplementary Fig. 4e and Supplementary Table 1) nor translocated to any of the four most common MLL fusion partners other than AF4 (AF6, AF9, AF10, and ENL) ( Supplementary Fig. 4f), and karyotyping confirmed that no other major structural abnormalities had been caused post-CRISPR-Cas9 editing (Supplementary Fig. 4g and Supplementary Table 2). Sanger sequencing of potential off-target editing sites 31,32 predicted in silico showed that, in the CRISPR MLL-AF4+ clones that grew out in primary recipient mice (n = 3; donors 1 and 2), no indels were present at these loci (Supplementary Table 1). The B-ALL in CRISPR MLL-AF4+ mice recapitulated key phenotypic features of infant-ALL, including circulating blasts in the PB ( Supplementary Fig. 4h) and blast infiltration into the spleen and liver ( Fig. 3c and Supplementary Fig. 4h). CRISPR MLL-AF4+ mice also had central nervous system (CNS) disease, with extensive parameningeal blast cell infiltration (Fig. 3d); a key clinical feature of infant-ALL. As chemo-resistance is also an important feature of MLL-r infant-ALL, we compared the responses of CRISPR MLL-AF4+ ALL blasts to prednisolone and L-asparaginase with ALL patient-derived xenograft (PDX) samples and the SEM and KOPN8 MLL-r cell lines and found that CRISPR MLL-AF4+ blasts showed similar levels of in vitro drug resistance to previous reports of treatment-resistant patient samples [34][35][36] (Supplementary Fig. 4i and Supplementary Table 2). Secondary (n = 4; donors 1 and 2) and tertiary (n = 3; donors 1 and 2) recipient mice all developed B-ALL with significantly reduced latency compared to primary recipients (median survival 11.5 weeks in secondary (p < 0.02) and 8 weeks in tertiary (p < 0.03)) ( Fig. 3a).
Although the clinicopathological features were the same in all leukemic mice (Supplementary Table 2), 2/3 CRISPR MLL-AF4+ mice had a CD19+CD10−CD20−IgM/IgD−CD34+/− proB ALL immunophenotype ( Fig. 3e and Supplementary Fig. 5a), while the remaining mouse had a preB ALL immunophenotype, with the majority of cells being CD19+CD10+CD20−IgM/IgD −CD34− ( Supplementary Fig. 5a, b). Further characterization of CRISPR MLL-AF4+ proB ALL (the most common type seen in patients) revealed that it recapitulated the immunophenotype of MLL-AF4 infant-ALL, including heterogeneous expression of CD133 33 , NG2 37 , and CD24 ( Fig. 3e and Supplementary Fig. 5c). Sequencing of the IgH locus showed that CRISPR MLL-AF4+ ALL was clonal (Supplementary Table 2). The proportion of blasts that were CD34+ did not correlate with CD10 expression (Supplementary Fig. 5a and Supplementary Table 2), nor did it increase significantly in secondary recipients (Supplementary Fig. 5d and Supplementary Table 2). This is in keeping with data from primary MLL-r patient samples, where CD34 expression is known to be heterogeneous ( Supplementary Fig. 5d). In fact, all clinicopathological and immunophenotypic features of primary CRISPR MLL-AF4+ ALL were maintained in secondary recipients, including CNS disease (Supplementary Table 2). Together, these data show that a CRISPR-Cas9-induced t(4;11)/MLL-AF4 translocation in human FL HSPCs is sufficient to promote a rapidly progressive, fatal, transplantable B-ALL that recapitulates key features of infant-ALL.
CRISPR MLL-AF4+ ALL recapitulates the molecular profile of MLL-AF4 ALL in patients. To characterize the transcriptomic changes underlying leukemic transformation, human CD19+ cells were sorted from the BM of primary CRISPR MLL-AF4+ (n = 3; donors 1 and 2) and control (n = 3; donors 1 and 2) mice for bulk RNA-seq ( Supplementary Fig. 6a). Among the 1068 differentially expressed genes between CRISPR MLL-AF4+ ALL and control CD19+ cells were many genes known to be upregulated in MLL-AF4 ALL, including FLT3, MEIS1, and RUNX1 ( Supplementary Fig. 6b). We compared bulk RNA-seq from control and CRISPR MLL-AF4+ BM to two independent patient datasets 23,27 and found that, on a transcriptome-wide level, CRISPR MLL-AF4+ ALL more closely resembled MLL-AF4 ALL patients compared to MLLwt ALL patients (Fig. 4a, b and Supplementary Fig. 6c). In addition, we carried out differential gene expression analysis between HOXA lo MLL-AF4 infant-ALL, HOXA hi MLL-AF4 infant-ALL, and HOXA hi MLL-AF4 childhood-ALL patients to derive a list of 765 genes that separate these patient subsets (Supplementary Data 3). Based on these genes, we found that all CRISPR MLL-AF4+ ALLs clustered with HOXA lo MLL-AF4 infant-ALL, suggesting that they all represent HOXA lo MLL-AF4 infant-ALL, regardless of their immunophenotype (Fig. 4c). Interestingly, while all CRISPR MLL-AF4+ mice showed negligible HOXA9 expression characteristic of this molecular profile, the expression pattern of IRX1, a gene that is usually overexpressed in HOXA lo ALL 20,21 was variable. This finding was similar to HOXA lo MLL-AF4 infant-ALL patients (Fig. 4d).
By chromatin immunoprecipitation-seq (ChIP-seq), we observed a clear genome-wide correlation between the MLL-AF4-binding profile in CRISPR MLL-AF4+ ALL, the MLL-AF4 B-ALL SEM cell line 38 , and a primary MLL-AF4 ALL patient sample 39 . In addition, we observed a substantial overlap in MLL-AF4 target genes (2323 genes bound by MLL-AF4 in all three datasets) (Fig. 4e, Supplementary Fig. 6d, and Supplementary hCD45-AF700    Data 4) with strikingly similar binding profiles within target genes, such as RUNX1 (Fig. 4f).
FL-derived CRISPR MLL-AF4+ ALL specifically recapitulates MLL-AF4 infant-ALL. Finally, we wanted to ask whether inducing an MLL-AF4 translocation in human FL HSPCs gave rise to a model that specifically recapitulated the molecular profile of MLL-AF4 infant-ALL. The only humanized mouse model of MLL-AF4 ALL that has previously been published introduced a chimeric MLL-Af4 fusion gene into human cord blood (CB) HSPCs (hereafter referred to as CB MLL-Af4+ ALL) 40 . While we acknowledge the fact that it was generated using a different technique and did not harbor the reciprocal AF4-MLL translocation, we hypothesized that this model may represent a neonatally derived (non-fetal) ALL to which our model could be compared.
To examine the fetal and postnatal gene expression programs that are key to determining the age-related differences between MLL-AF4 ALLs, we used the 139 genes up-or downregulated in both FL (compared to ABM) and MLL-AF4 infant-ALL (compared to MLL-AF4 childhood-ALL) (Supplementary Data 2). Clustering analysis based on this core fetal-specific infant-ALL gene list showed that CRISPR MLL-AF4+ ALL clustered with MLL-AF4 infant-ALL, whereas both MLL-AF4 childhood-ALL and CB MLL-Af4+ ALL formed their own, separate clusters (Fig. 5a). To explore this in more detail, we carried out differential gene expression analysis between CRISPR MLL-AF4+ ALL and CB MLL-Af4+ ALL followed by gene set enrichment analysis (GSEA). We found that CRISPR MLL-AF4+ ALL was significantly enriched for genes upregulated in both MLL-AF4 infant-ALL (p < 0.03) and FL HSPCs (p < 0.001) compared to CB MLL-Af4+ ALL (Fig. 5b).
Comparing MLL-AF4 binding at promoters genome-wide in both models, we found that MLL-AF4 in CRISPR MLL-AF4+ ALL showed greater enrichment (normalized ChIP-seq reads/bp) at the promoters of infant-ALL-and FL-specific genes compared to MLL-Af4 in CB MLL-Af4+ ALL. However, at all other genes, MLL-AF4/MLL-Af4 enrichment was comparable (Fig. 5c). At infant-ALL-and FL-specific genes IGF2BP1 (Fig. 5d) and HOXB4 (Fig. 5e), we observed an MLL-AF4 peak at the promoter in CRISPR MLL-AF4+ ALL but not CB MLL-Af4+ ALL. These data suggested that MLL-AF4 may play an active role in maintaining fetal gene expression programs in infant-ALL.
Increased levels of H3K79me2 are a commonly used marker of MLL-AF4 activity 5,6,33,41 . Therefore, using one of the unique features of our model, we carried out H3K79me2 ChIP-seq in identical primary human FL HSPCs before and after leukemic transformation. We observed increased levels of H3K79me2 in infant-ALL-and FL-specific genes such as IGF2BP1 (Fig. 5d) and HOXB4 (Fig. 5e) in CRISPR MLL-AF4+ ALL, further suggesting that MLL-AF4 actively maintains the expression of these fetalspecific genes in MLL-AF4 infant-ALL.

Discussion
The mechanisms by which the same MLL-r driver mutation could cause more aggressive disease and worse outcomes in infant-ALL compared to childhood-ALL have always been unclear. We hypothesized that there must be intrinsic biological differences between infant-ALL and childhood-ALL blasts, unrelated to their shared driver mutation, that underlie these age-related differences. To explore this, we used the most common MLL-r infant-ALL, MLL-AF4, as a disease model and set out to identify agerelated differences on the transcriptomic level. Using primary patient data 23 , we have identified the unique molecular profile of MLL-AF4 infant-ALL. Importantly, we find that this profile is present regardless of other well-studied molecular characteristics of the ALL, such as HOXA status [19][20][21][22] . Reasoning that this profile drives the distinct phenotype of infant-ALL, we then set out to identify factors that could explain it. We find that maintenance of fetal-specific gene expression programs accounts for a large proportion (~40%) of the unique molecular profile of MLL-AF4 infant-ALL, suggesting that it is the fetal target cells in which the translocation arises that provide the permissive cellular context for aggressive infant-ALL.
Human fetal HSPCs are more proliferative than ABM HSPCs 42,43 and they differentiate down distinct developmental pathways 44,45 , some of which are virtually absent in postnatal life. Therefore, maintenance of fetal HSPC characteristics provides a possible explanation for the highly proliferative, therapy-resistant nature of infant-ALL. However, one of the biggest challenges to understanding the biology of infant-ALL has been the lack of an appropriate model that captures the unique characteristics and aggressive nature of the disease. By targeting a t(4;11)/MLL-AF4 translocation to primary human FL HSPCs, we have created a faithful humanized MLL-AF4 infant-ALL model. Our results confirm that a human fetal cell context is permissive to give rise to an ALL that recapitulates key phenotypic and molecular features of poor prognosis MLL-AF4 infant-ALL.
We targeted the t(4;11)/MLL-AF4 translocation to CD34+ FL cells, which represent a mixture of different HSPC types. The immunophenotypic heterogeneity we observed among primary CRISPR MLL-AF4+ mice, with 2/3 showing a proB and 1/3 showing a preB immunophenotype, may be a consequence of the translocation occurring in different progenitor cell types. Interestingly, however, no other significant differences were observed between proB and preB CRISPR MLL-AF4+ ALL. First, no clinicopathological differences were observed, which may suggest that it is the shared fetal characteristics, more so than a cell-typespecific context, that drive the aggressive phenotypic features of infant-ALL, such as treatment resistance and CNS disease. Second, all CRISPR MLL-AF4+ ALLs represented the HOXA lo subset of MLL-AF4 infant-ALL. While this may draw an interesting parallel with the higher frequency of the HOXA lo subset observed in MLL-r infant-ALL patients [19][20][21][22] , we cannot draw conclusions from these data about the specific cell of origin of infant-ALL and/or the drivers of the HOXA lo/ HOXA hi molecular profiles. It will be interesting in the future to target the t(4;11)/MLL-AF4 translocation to specific fetal HSPC subsets to ask whether leukemic transformation and HOXA status is determined by gestational age, hematopoietic site, or progenitor cell type. Finally, all CRISPR MLL-AF4+ ALLs showed expression of both reciprocal fusion genes, MLL-AF4 and AF4-MLL. The contribution of AF4-MLL to the initiation of MLL-AF4 leukemia has been a topic of debate in the field 46,47 . However, our editing approach has not allowed us to address the question of the relative importance of AF4-MLL to transformation. In the future, our editing approach could potentially be adapted to express only one or both of the reciprocal fusion genes.
As well as providing insights into MLL-AF4 function in a human fetal cell context, CRISPR MLL-AF4+ ALL provides a preclinical model for translational studies that specifically recapitulates poor prognosis infant-ALL. For example, the CNS disease observed in CRISPR MLL-AF4+ ALL is a common clinical feature of infant-ALL that can lead to CNS relapse in these patients 48 . Therefore, the ability of novel treatments to eradicate blasts from the CNS is an important consideration, and this can be tested in CRISPR MLL-AF4+ ALL. Going forward, the unique, age-related molecular profile of MLL-AF4 infant-ALL defined here can be mined for potential novel targets to specifically treat poor prognosis infant-ALL, and these novel treatments can then be tested in CRISPR MLL-AF4+ ALL.  processing and were cryopreserved for future use as described previously 49 . ALL patient samples were obtained from Blood Cancer UK Childhood Leukemia Cell Bank, UK after appropriate review of our research project to ensure that it was covered under their ethics approval granted by NHS HRA South West -Central Bristol Research Ethics Committee (REC: 16/SW/0219). Informed consent was obtained from all participants or those with parental responsibility, and participants did not receive any monetary compensation. Infant and pediatric MLL-r ALL patients being treated at Great Ormond Street Hospital for Children, London had immunophenotypic analysis performed as part of their diagnostic workup after informed consent was obtained from all participants or those with parental responsibility. All patient samples/data were anonymized at source, assigned a unique study number, and linked. MLL-r and ETV6-RUNX1+ ALL PDX cells were provided by the Halsey lab.
Animals. All experiments were performed under a project license approved by the UK Home Office under the Animal (Scientific Procedures) Act 1986 after approval by the Oxford Clinical Medicine Animal Welfare and Ethical Review Body; and in accordance with the principles of 3Rs (replacement, reduction, and refinement) in animal research. All experimental animals were 8-12-week-old female NSG mice (n = 18). Mice were housed in individually ventilated cages, and kept at a 12-h light/dark cycle, 21-22°C temperature, and 45-65% relative humidity. They had red tunnels or houses and balconies in the cages as enrichment.
CRISPR-Cas9 MLL-AF4 translocation. CRISPR-Cas9 genome editing was carried out using a previously described protocol 50 . MLL and AF4 sgRNAs (Synthego; Supplementary Table 3) were first tested for editing efficiency individually in FL CD34+ cells. Cryopreserved CD34+ cells from a single primary human FL sample were thawed and placed into suspension culture at a density of 2.5 × 10 5 cells/ml in StemLine II (Sigma) supplemented with stem cell factor (SCF) (100 ng/ml), FLT-3ligand (FLT3L) (100 ng/ml), and thrombopoietin (100 ng/ml) (Peprotech) for 12 h. Cells were harvested and electroporated with either (i) Cas9 protein (IDT) only or (ii) a Cas9/sgRNA ribonucleoprotein (RNP) using a Neon TM Transfection System (Thermo Fisher). Electroporated cells were placed into fresh suspension culture media to recover overnight. Cells were harvested and bulk genomic DNA was extracted using a DNeasy Blood and Tissue Kit (Qiagen). An~1 kb region of DNA around the target cut site was amplified by PCR and Sanger sequenced (Eurofins). Sanger sequencing traces from samples edited with RNPs were compared to traces from Cas9-only controls using the ICE Analysis online tool (Synthego, https:// ice.synthego.com). Editing efficiency is reported as the percentage of indels detected ( Supplementary Fig. 3a). For each CRISPR-Cas9 MLL-AF4 translocation experiment, cryopreserved CD34+ cells from a single 13-15 pcw primary human FL underwent suspension culture as described. Cells were harvested and electroporated with either (i) Cas9 protein (IDT) only, (ii) Cas9 protein plus MLL-sgRNA only, as biologically matched controls, or (iii) a 1:1 mix of Cas9/MLL-sgRNA and Cas9/AF4-sgRNA RNPs using a Neon TM Transfection System (Thermo Fisher). Electroporated cells were placed into fresh suspension culture media to recover overnight before subsequent in vitro culture and in vivo transplantation experiments.
Xenograft transplantation. The 8-12-week-old female NSG mice were sublethally irradiated with two doses of 1.25 Gy 6 h apart (2.5 Gy total) and injected via the tail vein with 25,000-35,000 edited FL cells ( CRISPR MLL-AF4+, n = 3; Cas9 control, n = 5; or Cas9 plus MLL-sgRNA control, n = 1) plus 30,000 wt, unedited, sexmismatched FL CD34+ carrier cells. Engraftment was monitored by PB sampling every 3 weeks. Human CD45+ cells were sorted from PB samples to carry out MLL-AF4 and AF4-MLL RT-qPCR for the detection of successfully edited cells. Animals were monitored regularly using a standardized physical scoring system, and any mouse found to be in distress was humanely killed. Mice were considered leukemic if they met at least three of the following criteria: (i) overt signs of disease (hunching, lack of movement, weight loss, paralysis), (ii) splenomegaly, (iii) PB blast count over 50%, (iv) peripheral organ infiltration, and (v) detection of the MLL-AF4 translocation in both BM and spleen.
Flow cytometry. Cells were stained with fluorophore-conjugated monoclonal antibodies in phosphate-buffered saline with 2% FBS and 1 mM EDTA for 30 min and analyzed using BD LSR II or Fortessa X50 instruments using BD FACSDiva software (v8.0.2). Antibodies used are detailed in Supplementary Table 4. Flow cytometry antibodies were validated by titration in-house using primary human fetal mononuclear cells or NSG mouse BM. Analysis was performed using FlowJo software (v10.7.1), where gates were set using unstained and fluorescence minus one control.
Murine heads were decalcified and processed as described previously 51 . Following paraffin wax embedding, 2.5 μm sections were cut onto poly-L-silanecoated slides and stained with Gill's hematoxylin and Putt's eosin (both made inhouse). Slides were imaged on a NanoZoomer Digital Pathology (NDP) slide scanner (Hamamatsu) and analyzed with NDP.view 2 software.
Karyotype. G-band analysis was performed on metaphase spreads obtained after 24 h unstimulated culture (RPMI-1640, Colcemid, 5% CO 2 ). Cells were harvested, slides made according to the laboratory standard-operating procedure. G-band staining has been done with the use of an automated staining machine (Leica Autostainer XL). Karyotype analysis was performed with the use of CytoVision v 7.7 software (Leica Ltd). For each case, ten metaphase spreads were analyzed unless an insufficient number of metaphase spreads were found. A constitutional Robertsonian der(14;21)(q10;q10) was present in donor 2-derived CRISPR MLL-AF4 ALL and was confirmed to be present in the original unedited FL cells from this donor (Supplementary Table 2).  Table 5). FISH setup and wash were performed following the manufacturer's (Cytocell Ltd) standard protocol. Olympus BX41 fluorescent microscope equipped with the filters for FITC, Cy3, TexasRed, Aqua, DAPI, and double filter set for FITC/TexasRed was used for analysis. For each case, 200 interphase nuclei were examined and patterns scored.
In vitro drug sensitivity assays. Cryopreserved blast cells harvested from the spleen of CRISPR MLL-AF4+ mice, MLL-r, or ETV6-RUNX1+ PDX ALL models, as well as SEM and KOPN8 cell lines, were assayed in vitro for their response to prednisolone and L-asparaginase using the MTT assay (Roche, Cell Proliferation Kit I) as described previously 34,35 . Briefly, cells were resuspended at 2 × 10 6 cells/ml in RPMI supplemented with 15% heat-inactivated FBS, 100 U/ml penicillin, 100 µg/ml streptomycin, and 1% ITS Liquid Media Supplement (Sigma). In flatbottom 96-well plates, 100 µl cell suspension (0.2 × 10 6 cells) was treated with a range of concentrations of prednisolone (Sigma, final concentrations: 0.05-900 µg/ ml) or L-asparaginase (Cambridge Biosciences, final concentrations: 0.003-10 IU/ ml) based on previously published studies. For prednisolone, dimethyl sulfoxide was added to untreated control wells.
After 48 h incubation (or 96 h incubation for SEM and KOPN8 cell lines) at 37°C and 5% CO 2 , 10 µl MTT reagent was added to each well, after which the plates were incubated for another 4 h. During this time, the tetrazolium salt MTT is reduced into a purple-colored formazan product by living but not dead cells. One hundred microliters of Solubilization Solution was added to each well to dissolve the formazan crystals. The optical density (OD) of each well was measured on a SPECTROstar Nano (BMG Labtech) microplate reader at 570 nm. For each drug concentration, leukemia cell survival (LCS) was calculated by the following equation: LCS = (OD treated well/OD untreated well) × 100%. Drug resistance was expressed by the LC50, the drug concentration lethal to 50% of the cells.
RT-qPCR. Total RNA was extracted from cells using an RNeasy Micro Kit (Qiagen). Complementary DNA (cDNA) was generated from polyA messenger RNA (mRNA) using a SuperScript III Kit (Invitrogen). qPCR was carried out on cDNA using SYBRGreen Master Mix (Thermo Fisher) and a QuantStudio3 Real-Time PCR System (Thermo Fisher). For a list of qPCR primers used, see Supplementary  Table 3. ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-27270-z CRISPR-Cas9 off-target editing analysis. Potential off-target editing sites for both MLL and AF4 sgRNAs were predicted using the Synthego CRISPR guide verification tool (https://design.synthego.com/#/validate). None of the potential off-target sites had >3 mismatches in the guide sequence. To test off-target sites with three mismatches (total of nine loci), genomic DNA was extracted from the spleen of primary recipient CRISPR MLL-AF4+ mice (n = 3 mice from two individual FL samples) and from biologically matched, unedited FL cells (n = 2 individual FL samples) using a DNeasy Blood and Tissue Kit (Qiagen). An~1 kb region of DNA around the target cut site was amplified by PCR and Sanger sequenced (Eurofins). Sanger-sequencing traces from primary CRISPR MLL-AF4+ samples were compared to matched wt FL traces using the ICE Analysis online tool (Synthego, https://ice.synthego.com). The absence of off-target editing is reported as the percentage of reads in the CRISPR MLL-AF4+ edited cells that match the unedited cells at each of these loci.
RNA-sequencing. Approximately 3 × 10 5 CD45+CD19+ cells were sorted from the BM of three primary CRISPR MLL-AF4+ recipient mice and three control primary recipient mice (Cas9 control, n = 2; Cas9 plus MLL-sgRNA, n = 1). Total RNA was extracted using an RNeasy Mini Kit (Qiagen). Poly(A) purification was conducted using the NEB Poly(A) mRNA magnetic isolation module as per the manufacturer's protocol. Library preparation was carried out using the Ultra II Directional RNA Library Prep Kit (NEB, E7765). RNA libraries were sequenced by paired-end sequencing using a 150-cycle high output kit on a Nextseq 500 (Illumina). RNA-seq protocols for sorted subpopulations of FL HSPC have been described previously in ref. 27 .
IgH rearrangement analysis. Samples were screened for IgH complete (VH-DH-JH) and IgH incomplete (DH-JH) rearrangements using BIOMED-2 protocols to detect clonality. DNA was extracted from cells from the BM of three primary CRISPR MLL-AF4 + recipient mice. IgH rearrangements were analyzed as described in ref. 45 .
ChIP-sequencing. The full protocol is described in ref. 38 . In short, up to 5 × 10 7 cells were sonicated (Covaris) following the manufacturer's protocol and incubated with an antibody overnight. Magnetic protein A and G beads (Thermo Fisher Scientific) were used to isolate antibody-chromatin complexes. Antibodies used are detailed in Supplementary Table 4. Beads were washed three times using a solution of 50 mM HEPES-KOH (pH7.6), 500 mM LiCl, 1 mM EDTA, 1% NP40, and 0.7% sodium deoxycholate and once with Tris-EDTA. Samples were eluted and proteinase K/RNase A-treated. Samples were purified using a ChIP Clean and Concentrator Kit (Zymo). DNA libraries were generated using the NEBnext Ultra DNA Library Preparation Kit for Illumina (NEB, E7103). Libraries were sequenced by paired-end sequencing using a 75-cycle high output kit on a Nextseq 500 (Illumina).
NGS analysis. For RNA-seq, following sequencing, QC analysis was conducted using the fastQC package. Reads were mapped to the human genome assembly using STAR (https://github.com/alexdobin/STAR/). The featureCounts function from the Subread package (http://subread.sourceforge.net/) was used to quantify gene expression levels using standard parameters. This was used to identify differential gene expression globally using the edgeR package (https:// bioconductor.org/packages/release/bioc/html/edgeR.html). Differential gene expression was defined by an adjusted p value (FDR) of <0.05. Infant-ALL RNAseq datasets were analyzed as described previously 45 . GSEA analysis was performed using the fgsea function in the fgsea R package to determine the positive enrichment score and enrichment p value of gene sets within differentially expressed genes, nperm = 1000 52 .
To derive an FL vs ABM gene signature, bulk RNA-seq for sorted subpopulations of FL HSPC 27 were compared to matched sorted subpopulations of ABM HSPC 28 (FL HSC vs adult BM HSC, FL MPP vs ABM MPP, FL LMPP vs ABM LMPP, and FL CBPs vs ABM CLP). Genes that were differentially expressed between FL and ABM in at least one matched HSPC subpopulation were included in the gene signature. Genes that showed a significant change in opposite directions in different HSPC subtypes (e.g., upregulated in FL HSC vs ABM HSC, but downregulated in FL LMPP vs ABM LMPP) or in the normal vs leukemic setting (e.g., upregulated in FL HSPC vs ABM HSPC, but downregulated in MLL-AF4 infant-ALL vs MLL-AF4 childhood-ALL) were filtered out of the gene signature to leave a total of 5709 genes (Supplementary Data 2).
To analyze the effects of HOXA status and age on MLL-AF4 ALL blasts, we used the generalized linear model (glm) functionality of the edgeR package to carry out a three-way comparison between HOXA lo MLL-AF4 infant-ALL, HOXA hi MLL-AF4 infant-ALL, and HOXA hi MLL-AF4 childhood-ALL. This identified 765 marker genes, which could differentiate the three patient subsets from one another. UMAP analysis including the patient samples and CRISPR MLL-AF4+ ALL was carried out based on this 765 gene signature.
For ChIP-seq, quality control of FASTQ reads, alignment, PCR duplicate filtering, blacklisted region filtering, and UCSC data hub generation was performed using an inhouse pipeline as described 53 . The HOMER (http://homer.ucsd.edu/homer/) tool makeBigWig.pl command was used to generate bigwig files for visualization in UCSC, normalizing tag counts to tags per 1 × 10 7 . ChIP-seq peaks were called using the HOMER tool findPeaks.pl with a ChIP input sample used to estimate background signal. Gene profiles were generated using the HOMER tool annotatePeaks.pl.
Statistics. Two-tailed Mann-Whitney, log-rank (Mantel-Cox) tests, and analysis of variance followed by multiple comparisons testing were used to compare experimental groups as indicated in the figure legends. Statistical analyses were performed using GraphPad Prism v7.04 or R v4.0.1. Data are expressed as mean ± SEM, unless otherwise indicated.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The RNA-seq and ChIP-seq data generated in this study have been deposited in the NCBI GEO database under accession code NCBI GEO: GSE162041. These data are included in Figs. 1, 4, and 5 and Supplementary Figs. 1, 2, and 6. Source data are provided with this paper.

Code availability
ChIP-seq data were analyzed using an in-house pipeline 53 . Further information and requests for resources and reagents may be directed to and will be fulfilled by the corresponding authors.