To the Editor:

While childhood B cell acute lymphoblastic leukemia (ALL) has an excellent prognosis at diagnosis, outcomes for relapsing patients have been disappointingly low. This grim picture has recently improved with the advent of bispecific T cell engagers (BiTEs) and chimeric antigen receptor (CAR) T cells targeting CD19 with, for instance, CART cells achieving complete remission in over 80% of patients with B-ALL. Nevertheless, 30–50% of the patients still experience relapse within one year [1] with three quarters of relapses showing loss of CD19 surface expression [2]. In particular KMT2A-rearrangements, which independently predict poor outcome [3], are prone to treatment failures resulting from lineage-switched CD19-negative relapse [4]. Recently, we demonstrated that lineage switch can originate either in the ALL blast population or from an immature progenitor population and that KMT2A-rearranged infant ALL is characterized by an early lymphocyte precursor (ELP) signature that was not detectable in the lineage-switched myeloid relapse [5, 6]. These results complement similar findings from mouse models of t(4;11) ALL [7, 8] and TCF3::ZNF384 BCP-ALL also pointing towards an early progenitor with lymphoid potential pre-VDJ recombination [9, 10]. These combined findings raise the question about the nature of the cell of origin of this type of B-ALL. Here we demonstrate the presence of KMT2A-rearrangements in progenitor cells harboring both lymphoid and myeloid potential and their capacity to initiate leukemia, potentially acting as the cellular origin of CD19-negative relapse.

We examined seven infants with ALL and KMT2A-rearrangements at diagnosis: five with KMT2A::AFF1 and two with KMT2A::MLLT3 fusions (Fig. 1A, B). The following hematopoietic stem and progenitor cell populations were isolated: HSCs (CD34+CD38−CD45RA−CD90+), multipotent progenitor cells (MPPs, CD34+CD38−CD45RA−CD90−), lymphoid-primed multipotent progenitor cells (LMPPs, CD34+CD38−CD45RA+), common myeloid progenitor cells (CMPs), granulocyte monocyte progenitor cells (GMPs), mature monocytes, and T cells. We evaluated the presence of fusion genes in these purified populations by PCR. In three cases, the KMT2A::AFF1 fusion gene was found in LMPPs. In two cases, KMT2A::AFF1-positive cells were also present either in the MPP or HSC fraction. In two KMT2A::MLLT3 samples we detected only fusion gene-positive CMP-like and blast populations (Fig. 1B). Notably, the two KMT2A::AFF1-positive patients MA4_2 and MA4_3, had undetectable levels of the fusion gene in the HSC-MPP-LMPP subsets and did not undergo relapse (Fig. 1A, B). However, a larger cohort is required to follow up if there was any correlation between fusion gene positivity in early progenitor populations and incidence of relapse. Moreover, it is important to note that a negative PCR result does not formally exclude the presence the corresponding fusion gene at levels below the detection limits of the assay.

Fig. 1: Identification of KMT2A-rearrangement in different hematopoietic subsets.
figure 1

A Clinical information patient cohort. B Summary of KMT2A::AFF1 and KMT2A::MLLT3 presence in different subsets of hematopoietic population. Representative patients MA4_2, MA4_4, and MA9_2 breakpoint sequences were shown by Sanger sequencing. C Single-cell sorting of patient MA4_1 for monocyte (HLA-DR+CD14+CD11c+) populations. The two KMT2A::AFF1-positive cells, cells 11 and 18, are highlighted in red throughout the sorting strategy (lower panel). D Amplification of the KMT2A::AFF1 fusion in MA4_1 in cells 11, 18, and bulk sample shows the expected 299 bp band. E Summary KMT2A::AFF1, MAGED1, and NCOA2 mutation assessment on sorted populations MA4_1 diagnosis and relapse. MAGED1 is mutated gene only found at diagnosis. NCOA2 is mutated gene only found at relapse.

Full size image

In concordance with the translocation occurring in early progenitor populations, we identified the KMT2A::AFF1 fusion gene also in CD34–CD19–CD3–HLA-DR+ monocytes/dendritic cells in 4 of 5 cases, providing further evidence of an early KMT2A-r progenitor with bilineage, i.e., lymphoid and myeloid, potential (Fig. 1B). To exclude the possibility of sorting impurities, these results were confirmed by single-cell PCR of sample MA4_1. Monocytes were sorted into 96-well plates, followed by whole-genome amplification and PCR amplification of KMT2A::AFF1. The causative translocation was identified in 2 of 22 monocytes analyzed (Fig. 1C, D). These results imply an early KMT2A::AFF1 progenitor-like cell with both lymphoid and myeloid potential that might serve as a source for lineage switch.

Indeed, patient MA4_1’s disease relapsed four years after diagnosis with an AML harboring the same KMT2A::AFF1 breakpoint [5, 6]. We performed bulk whole-exome sequencing and identified, in addition to the shared KMT2A::AFF1 fusion gene, mutations that were exclusively present at diagnosis (MAGED1) or at relapse (NCOA2). Assessment of each of these mutations in sorted cell populations detected KMT2A::AFF1 in the MPP populations of both presentation and relapse (Fig. 1E). Mutated MAGED1 was present in all diagnostic progenitor populations except HSCs and CMP-like cells. In contrast, mutated NCOA2 was found in LMPP and GMP-like populations, but not in more immature cell populations (Fig. 1E). These findings identify a KMT2A::AFF1-positive cell as the cell of origin for both diagnostic ALL and relapse AML and show that secondary mutations were acquired at later stages. These data are also supported by our previous observation that immunoglobulin rearrangements were detected only at diagnosis but not at relapse of patient MA4_1 [6]. Therefore, these data suggest that the cell of origin had not initiated VDJ rearrangement, suggesting an MPP-like or even more immature phenotype.

To further characterize and functionally investigate the KMT2A-rearranged precursor cells, we generated patient-derived xenograft (PDX) models by transplanting the unsorted diagnostic samples into NOD-scid/IL2Rγ–/– (NSG) mice. Initially we selected two KMT2A::AFF1 (MA4_1 and MA4_6) samples and one KMT2A::MLLT3 (MA9_3) sample. Following engraftment, haematopoietic cells were collected and analyzed by flow cytometry. We observed a CD19+ blast population and, in addition, a more immature CD34+CD19– population. These populations were sorted, and the fusion gene expression was assessed by qPCR. Fusion gene transcripts were observed in CD34+CD19– cells in all three samples (Fig. 2A–C), albeit ~2-fold lower compared to the CD19+ compartment (Fig. 2C). Notably, serial transplantation across 4 generations of mice maintained this human HSC compartment confirming its self-renewal potential (Fig. 2D, E). Finally, we isolated CD19+ and CD34+CD19− populations and transplanted them into NSG mice. We observed that the CD34+CD19− population could reconstitute the disease by having both CD19+ and CD19– subsets (Fig. 2F), consistent with previous studies [8, 9].

Fig. 2: CD19− CD34+ population can reconstitute CD19+ and CD19− blasts.
figure 2

PCR identification of A KMT2A::AFF1 and B KMT2A::MLLT3 fusions within sorted human CD19–CD34+ and CD19+ on PDX MA4_1 and MA9_3, respectively. 293T cells provided a fusion gene negative control. C qRT-PCR analysis of KMT2A::AFF1 expression in CD34+CD19- and CD19+ cells from PDX MA4_6. D Flow cytometric analysis of bone marrow-derived hematopoietic precursors from normal human bone marrow control, and across four generations of NSG mouse xenografts. The CD34+CD38-CD45RA-CD90 + HSC population is expanded and maintained across four generations. E Proportion CD34+CD38−CD45RA−CD90+ HSC gate of CD19− CD34+ cells patient MA4_1 across sequential transplantation. F Sorting of CD34+CD19− and CD19+ populations from patient MA4_1, left panel. Both populations were transplanted into NSG and evaluated following the engraftment, right panel.

Full size image

This study confirms recent findings that the cell of origin in B-ALL is located at an early progenitor stage preceding the ELP stage and may be of pre-leukemic nature [5, 7, 11]. CD19-negative populations contain KMT2A::AFF1 or KMT2A::MLLT3 fusion genes and can progress towards malignancy, raising the question of the nature of factors co-operating with KMT2A-rearrangements to produce full transformation. Our data support the need to involve other targets, such as dual CD19/CD22, to prevent relapse caused by CD19-negative cells. The involvement of CD22 is of potential interest because its expression starts at the LMPP stage [11]. Alternatively, we propose targeting the KMT2A fusion gene, which is present in both CD19+ and CD19– populations at the transcript level or via fusion peptides presented by major histocompatibility complex and recognized by T cells [12,13,14,15]. These studies support the feasibility of discovering other fusion gene–reactive T cells, including reactivity for KMT2A-rearrangements.