The age of the bone marrow microenvironment influences B-cell acute lymphoblastic leukemia progression via CXCR5-CXCL13

Twenty-five percent of cancers in children are due to B-cell acute lymphoblastic leukemia (B-ALL), the most common cancer in this age group.¹ Enormous progress has been made in the treatment of this disease, although for some patients the disease, its relapse, or the sequelae of its treatment are still fatal. In addition, the incidence of pediatric B-ALL is increasing.¹ Chronic myeloid leukemia (CML), in contrast, is rare in children, although all cases of CML and 3% of cases of pediatric B-ALL (and 25% of cases of adult B-ALL) are due to the same oncogene, the fusion gene BCR-ABL1. BCR-ABL1 confers a poor prognosis to patients with B-ALL.² The difference in incidence of B-ALL and CML in children vs adults has been explained by an increased expression of genes involved in myeloid lineage specification and an increased potential of myeloid differentiation in hematopoietic stem cells of aged individuals.^3,4 However, other factors, such as the bone marrow microenvironment (BMM), contributing to the predominantly lymphoid leukemic phenotype in children have not been studied in detail.

The BMM is a complex arrangement of various cell types and other factors influencing hematopoietic stem and progenitor cells and leukemia stem cells (LSCs).⁵ LSCs are protected by this BMM to prevent therapy-induced apoptosis.^5-7 The BMM has been shown to influence the phenotype of leukemia,⁸ and leukemias remodel the leukemic niche.^9-11 The leukemic BMM can be targeted, leading to a reduction of LSCs.^12,13 Our knowledge about the influence of an aging microenvironment on hematopoiesis and leukemia is basically restricted to the findings that an old BMM provides a selection pressure on certain hematopoietic clones, contributing to leukemia development,¹⁴ and that old mice transplanted with leukemia-initiating cells (LICs) positive for the oncogene AML1-ETO have greater expansion of preleukemic stem and myeloid progenitor cells than young mice.¹⁵ We hypothesized that a young vs an old BMM may specifically promote the development and progression of B-ALL, possibly via the secretion of soluble factors.

In summary, we have shown that similar to the situation in humans, B-ALL development is more efficient in young mice. We observed differences of B-ALL cell behavior in a young vs an old BMM and implicated the B-cell chemoattractant C-X-C motif chemokine (CXCL) 13 secreted by macrophages and bound by its receptor, C-X-C chemokine receptor type 5 (CXCR5), as contributing to the leukemic phenotype in murine models of B-ALL. In human B-ALL of the hyperdiploid subtype, low expression of CXCR5 may act as an additional factor of favorable prognosis, whereas high expression of CXCR5 may be associated with central nervous system (CNS) relapse. The CXCL13-CXCR5 axis may act as a novel target for treatment in human B-ALL.

For B-ALL induction, non–5-fluorouracil (5-FU)-treated BM cells, transduced once with MSCV-IRES green fluorescent protein (GFP) BCR-ABL1 retrovirus, were transplanted into unirradiated young vs old recipients at a dose of 2 × 10⁶ cells (supplemental Methods available on the Blood Web site.)

Survival differences were assessed by Kaplan-Meier (log-rank) test.

Hypothesizing that induction of B-ALL in a young BMM may be more efficient than in an old BMM, we transplanted non–5-FU–pretreated BM from 6-week-old mice, transduced once with BCR-ABL1–expressing retrovirus in the murine transduction/transplantation model of B-ALL¹⁶ into young (3-4 weeks old) vs old (>1.5 years old) unirradiated recipient mice. This approach revealed a significantly higher number of leukocytes (P = .01; Figure 1A) and BCR-ABL1⁺ BP-1⁺ (pre-B) cells (Figure 1B-C; supplemental Figure 1A; supplemental Table 1), higher spleen weights (P = .03; Figure 1D), and a shortening of survival (P = .02; Figure 1E) in young compared with old BALB/c and C57/BL6 (P = .04; Figure 1F) mice. This finding was independent of weight differences between young and old mice because 50% reduction of the transplanted LIC dose also led to increased tumor burden (P = .03; Figure 1G) and accelerated B-ALL progression (P = .01; Figure 1H) in young mice.

Similarly, transplantation of human B-ALL cells in pairs into irradiated (200 cGy) nonobese diabetic severe combined immunodeficiency interleukin-2 receptor γ mice led to significantly higher engraftment of human CD45⁺ leukocytes in young mice (Figure 1I-J; supplemental Figure 1B; supplemental Table 2), and their survival was significantly shortened (P = .04; Figure 1K). In the retroviral model, there was significant reduction of GFP⁺ (BCR-ABL1⁺) BP-1⁺ cells in several organs of young moribund mice (Figure 1L; supplemental Figure 1C). The percentage, but not absolute numbers of GFP⁺ (BCR-ABL1⁺) BP-1⁺ cells, that homed to a young BMM was higher compared with an old BMM (P = .008; Figure 1M; supplemental Figure 1D). In contrast, transplantation of BCR-ABL1–transduced 5-FU–pretreated BM into young vs aged recipients in the retroviral model of CML (in which recipients have to be irradiated) (supplemental Figure 2A), led to no differences in tumor load (supplemental Figure 2B) but a significant prolongation of survival in young compared with aged mice (P = .016; supplemental Figure 2C). As has frequently been reported for inefficient induction of CML in this model,^13,17 the BM cavity of young mice with CML was mostly filled with amorphous eosinophilic material and infiltrating monocytes/macrophages, with the latter being the cause of death.¹³ Histopathological characteristics of CML were found in aged recipient mice (supplemental Figure 2D). No differences in homing of GFP⁺ (BCR-ABL1⁺) CML LICs¹⁸ to a young vs an aged BMM were observed (supplemental Figure 2D; supplemental Table 1). Further, a higher number of proviral clones, representing LICs,¹⁶ engrafted in spleens of aged compared with young recipients (P = .0004; supplemental Figure 2F-G). More myeloid colonies were found when the plated BM was derived from an aged compared with a young BM (P = .0009; supplemental Figure 2H). In summary, these results suggest that a young BMM is more supportive to the engraftment and maintenance of B-ALL LICs compared with an old BMM, whereas an aged BMM may be more conducive to the progression of CML. These findings suggest the importance of the BMM, in addition to LIC intrinsic factors, for the leukemic phenotype.

Because location of LICs in the BMM may correlate with disease aggressivity and progression,¹⁹ we tested by in vivo microscopy of the murine calvarium²⁰ whether the location and behavior of murine BCR-ABL1⁺ BP-1⁺ B-ALL LICs¹⁷ may differ between a young and old BMM (Figure 2A). Indeed, B-ALL LICs, but not normal lineage-negative cells, injected into young mice homed to locations closer to the endosteum (P = .0007; Figure 2B-C; supplemental Movies 1 and 2). Further, the speed of movement (P = .017; Figure 2D) of injected B-ALL LICs was increased, and their migration may be less random (supplemental Figure 3) in a young compared with an old BMM. Taken together, these data suggest that location and migration of B-ALL–initiating cells may differ between a young and an old BMM. Proximity to the endosteum, higher speed, and less random movement may be indicative of specific responses of LICs to factors in a young rather than old BMM and may influence disease outcome.

Monocytes play an important role in the B-ALL microenvironment,²¹ and monocyte/macrophage depletion in B-ALL,²¹ T-cell ALL,²² and chronic lymphocytic leukemia²³ leads to a reduction of leukemia burden and survival prolongation. To test the role of monocytes/macrophages for the support of B-ALL, we obtained this population via its adherence to plastic (supplemental Figure 4A).^24,25 Detailed immunophenotyping of these cells,^26,27 henceforth termed macrophages, was F4/80⁺ CD169⁺ CX3CR1⁻ with little contamination by dendritic cells, monocytes, and neutrophils (supplemental Figure 4B-C), whose percentages in the BM did not differ (except for neutrophils) (supplemental Figure 4D-F). Testing whether macrophages from young vs old mice play a differential role in B-ALL, we showed that numbers of sorted GFP⁺ (BCR-ABL1⁺) BP-1⁺ cells from mice with leukemia (P < .0001; Figure 3A) or BCR-ABL1− but not empty vector-transduced BA/F3 cells (supplemental Figure 5A), a pro-B lymphoid cell line, were significantly higher when plated on young compared with old macrophages. No differences in growth, however, were observed when the myeloid cell line 32D, transduced with empty vector or BCR-ABL1, was plated on young vs old macrophages (supplemental Figure 5B). Plating of BCR-ABL1⁺ BA/F3 cells on macrophages from mice of increasing ages led to a gradual decline of BCR-ABL1⁺ BA/F3 growth (Figure 3B). Plating of sorted BCR-ABL1⁺ BP-1⁺ cells from young or old mice with established B-ALL on young or old macrophages demonstrated an augmentation of the growth of BCR-ABL1⁺ BP-1⁺ cells from an old BMM plated on young macrophages (P = .001; Figure 3C). The increase in B cells was at least partly due to a significant increase in the proportion of cells in the (S)-G2-M fraction of the cell cycle, when primary B-ALL–initiating (P = .003; Figure 3D) or BCR-ABL1⁺ BA/F3 (P = .0008; supplemental Figure 5C) cells were plated on young compared with old macrophages. BCR-ABL1⁺ BA/F3 (P = .02; supplemental Figure 5D) (but not primary B-ALL–initiating) (Figure 3E) cells adhered significantly better to young macrophages, whereas BCR-ABL1⁺ 32D cells adhered better to old macrophages (P = .007; supplemental Figure 5E). The migration of primary B-ALL–initiating (P = .01; Figure 3F) and BCR-ABL1⁺ BA/F3 (P = .02 and P = .03; supplemental Figure 5F) cells was enhanced toward young compared with old macrophages. Suspecting that cytokines differentially secreted by young vs old macrophages may be responsible, we cultured B-ALL cells in conditioned medium (CM) from young vs old macrophages. This approach revealed greater proliferation of primary BCR-ABL1⁺ BP-1⁺ (P < .0001; Figure 3G) and BCR-ABL1⁺ BA/F3 (P = .05; supplemental Figure 5G) cells when cultured in the CM of young macrophages. Further, culture of primary B-ALL LICs from old mice in CM from young macrophages rescued the decreased proliferation observed when the cells were cultured in CM from old macrophages (P < .007; Figure 3H). Taken together, these results suggest that the age of BM macrophages and/or their soluble, secreted factors influence the proliferation, cell cycle, and migration of leukemia cells, although an influence of cell-cell contact or leukemia cell–intrinsic factors cannot be excluded.

Given these differences, we hypothesized that young and old macrophages may differ with respect to key features such as proliferation or metabolism. Indeed, staining of first passage BM macrophages from young vs old mice with carboxyfluorescein succinimidyl ester (CFSE) revealed a significantly higher proliferative index in young macrophages (P = .0001; Figure 4A). Similarly, the proliferative index of young but not old macrophages increased from passage 1 to 2 (P < .0001; Figure 4B). Reactive oxygen species (ROS) (P = .02; Figure 4C) and γH2AX staining as a marker of DNA damage (and/or a higher number of cell division) (P = .031; Figure 4D) were significantly increased in old compared with young macrophages. Furthermore, staining for translocase of the outer membrane 20, a complex of proteins located in the outer membrane of mitochondria, revealed reduced staining and an elongation of mitochondria in old macrophages, which has been associated with mitochondrial stress²⁸ (Figure 4E). A genome-wide transcriptome analysis²⁹ revealed pronounced global differences and overall 473 up- and 569 downregulated differentially expressed genes between cultured young and old macrophages (supplemental Figure 6A-B; supplemental Table 3). Nonredundant enrichment analysis of differentially expressed genes using Metascape revealed significant enrichment of terms related to cytokine function, such as chemotaxis, virus response, and inflammation (supplemental Figure 6C-D). Investigating whether epigenetic changes regulate macrophage function, we performed ATAC sequencing to determine the open chromatin landscape in young vs old macrophages. It revealed pronounced global differences in genome-wide chromatin accessibility patterns (Figure 4F; supplemental Figure 7A-D; supplemental Table 4). Overall, 9496 differentially accessible regions (DARs) were identified (supplemental Figure 7E), which could be annotated to 5465 different genes (Figure 4G; supplemental Figure 7B). Two-thirds (67%) of the DARs showed increased chromatin accessibility in old BM macrophages (supplemental Figure 7A,E). On a global scale, the DARs were associated with the gene expression changes observed between young and old macrophages (supplemental Figure 7F). In analogy to the RNA-seq data, an enrichment analysis of the DARs revealed functional terms related to inflammation and immune response (Figure 4H). In summary, these data suggest that functional differences exist between macrophages isolated from young and old BM, particularly in regards to proliferation, ROS content, DNA damage, mitochondria, and inflammatory response.

We hypothesized that a chemokine, more highly secreted by young macrophages, may be accountable for augmented leukemia cell growth. CXCL13 is a B-lymphocyte chemoattractant mediating the migration of B cells³⁰ and lymphoma cells³¹ to sites of infection and the CNS, respectively. CXCL13 acts as diagnostic marker in the cerebrospinal fluid in neuroborreliosis.³⁰ Further, CXCL13 is included in the MSigDB hallmark gene sets “inflammatory response” and “IL-6_Jak_Stat3_signaling,” which were enriched in DARs (Figure 4H). Therefore, we surmised that CXCL13 levels may be higher in a young vs an old BMM, that CXCL13 may contribute to altered migration of leukemia cells in a young BMM, and that it may play a role in the shortened survival of young mice with B-ALL. Indeed, we found significantly higher levels of CXCL13 in cultured (Figure 5A; supplemental Figure 8A) and a trend in primary (supplemental Figure 8B) BMM-derived macrophages but not other cell types (supplemental Figure 8C-E) from young mice compared with old mice. Further, flushed BM (P = .007; Figure 5B) of young mice and CM from young BM-derived macrophages (P < .0001; Figure 5C) had a higher CXCL13 concentration than the respective old controls, also when the CXCL13 concentration was normalized to the number of leukocytes or macrophages, respectively (supplemental Figure 8F-G). No significant differences of CXCL13 expression were found on a transcriptional level (supplemental Figures 6 and 8H). The CXCL13 concentration was also higher in plasma (P = .03; Figure 5D) and BM (P = .03; Figure 5E) and in the CM of macrophages (P = .05; Figure 5F) from young compared with old mice with B-ALL. Testing the expression of Cxcr5, the receptor for CXCL13, on GFP⁺ (BCR-ABL1⁺) BP-1⁺ B-ALL cells, we found higher Cxcr5 expression on B-ALL cells from young compared with old mice (P = .03; Figure 5G; supplemental Figure 9A). CXCR5 expression was also detected on BCR-ABL1⁺ BA/F3 cells, various human B-ALL cell lines (Figure 5H; supplemental Figure 9B-C), and normal human B and primary human B-ALL cells (Figure 5I; supplemental Figure 9D). These results suggest that CXCL13, a chemokine with B-cell–attracting function, is more highly expressed by macrophages in a young vs an old BMM on a translational level. Cxcr5 is expressed at higher levels on B-ALL cells from a young vs an old BMM and on human B-ALL cells, suggesting that the CXCR5-CXCL13 axis may be of importance, particularly in a young BMM, where it may influence B-ALL cell migration and proliferation.

To test the CXCR5-CXCL13 axis for the proliferation of BCR-ABL1⁺ BA/F3 cells, we transduced BMM-derived macrophages in vitro with sh scrambled or sh Cxcl13-expressing lentivirus (P = .003; supplemental Figure 10A) before coculture with BCR-ABL1–transduced BA/F3 cells. Knockdown of Cxcl13 in macrophages led to a significant reduction in the proliferation of BCR-ABL1⁺ and empty vector-control⁺ BA/F3 cells (supplemental Figure 10B). Consistently, addition of an inhibitory antibody to CXCL13 significantly reduced the proliferation of primary BCR-ABL1⁺ BP1⁺ (P < .0001; Figure 6A) and BCR-ABL1⁺ BA/F3 (supplemental Figure 10C) cells when plated on young macrophages. CXCL13 inhibition did not alter the proliferation of BCR-ABL1⁺ BA/F3 cells plated on old macrophages, nor did it significantly alter the proliferation of BA/F3 cells transduced with empty vector (supplemental Figure 10C). Coculture of BCR-ABL1⁺ BA/F3 cells with young macrophages also led to higher levels of nuclear pAKT, which lies downstream of CXCR5³² and BCR-ABL1³³ and has been implicated in the cell cycle,^25,34 in leukemia cells compared with coculture on old macrophages (Figure 6B). This effect was blocked after treatment with anti-CXCL13 (P < .0001; Figure 6B-C). Conversely, addition of recombinant CXCL13 to cocultures of macrophages with primary BCR-ABL1⁺ BP1⁺ (P = .005 on young macrophages and P = .0002 on old macrophages; Figure 6D) or BCR-ABL1⁺ BA/F3 (supplemental Figure 10D) cells increased leukemia cell proliferation compared with cultures treated with bovine serum albumin (BSA). Treatment of BCR-ABL1⁺ BA/F3 cells with increasing concentrations of CXCL13 led to nonsignificantly increased levels of pAKT (supplemental Figure 10E-F). Further, treatment of primary sorted BP-1⁺ BCR-ABL1⁺ leukemia cells, which had been exposed to a young or an old BMM in vivo, with recombinant CXCL13 significantly promoted the growth of leukemia cells, regardless of whether they had been derived from a young (P = .005; Figure 6E, left) or an old (P = .012; Figure 6E, right) BMM. In confirmation of the pro-proliferative role of CXCL13, in vitro pretreatment of B-ALL LICs with CXCL13 before transplantation significantly accelerated leukemia development in recipient mice compared with B-ALL LICs treated with BSA (P = .002; Figure 6F) by increasing CXCR5 expression (supplemental Figure 10G), homing to BM (supplemental Figure 10H) and pAKT expression (supplemental Figure 10I) of BP-1⁺ BCR-ABL1⁺ cells. Blockade of CXCL13 significantly inhibited the migration of primary BCR-ABL1⁺ BP1⁺ (P < .0001; Figure 6G) and BCR-ABL1⁺ BA/F3 (P = .04; supplemental Figure 10J) cells toward macrophages from a young BMM, whereas addition of CXCL13 increased migration (P = .01; Figure 6G). Ablation of phagocytes in young or old recipients of B-ALL LICs with clodronate²⁵ (supplemental Figure 11A) led to a significant reduction of the GFP⁺ (BCR-ABL1⁺) BP1⁺ tumor burden (P < .0001; Figure 6H), survival prolongation (P = .05; Figure 6I), and a reduction of CXCL13 in blood plasma (P = .001; Figure 6J) and flushed bone (P < .0001; Figure 6K) in young but not old recipients (supplemental Figure 11B-D). Contributions of other ablated cell types to this phenotype cannot be excluded, although their numbers were only minimally reduced (supplemental Figure 11A). Taken together, these data suggest that CXCL13, secreted at least partly by macrophages, promotes the migration and proliferation of B-ALL cells, influencing survival in B-ALL.

We hypothesized that decreased levels of CXCR5 may prolong survival of mice with B-ALL. Indeed, transplantation of BCR-ABL1⁺ wild-type or CXCR5-deficient donor BM into young but not old (data not shown) wild-type recipients led to a significant reduction of the BCR-ABL1⁺ (GFP⁺) BP-1⁺ tumor load (P < .0001; Figure 7A) and survival prolongation in recipients of CXCR5-deficient compared with wild-type B-ALL LICs (P = .0004; Figure 7B). However, the homing of CXCR5-deficient BCR-ABL1⁺ BP-1⁺ B-ALL LICs was not compromised (supplemental Figure 12A). In the human setting, CD68⁺ macrophages and CXCL13 levels were higher in bone sections from pediatric compared with adult patients with B-ALL (Figure 7C; supplemental Figure 12B; supplemental Table 2). Given the role of CD14^dim CD16⁺ nonclassic monocytes for B-ALL progression,²¹ we also identified increased proportions of this cell population (Figure 7D; supplemental Figure 12C), which generates CXCL13 (Figure 7E), in bone sections of patients with pediatric vs adult B-ALL. Testing whether CXCR5 expression may act as predictive marker in human B-ALL, we stratified patients from the Pan-B-ALL study³⁵ into patients with high or low expression of CXCR5 based on RNA-seq results and time-dependent receiver operating characteristic analysis to establish a cutoff. For the B-ALL subset hyperdiploid, normally associated with good prognosis,³⁶ we demonstrated that low CXCR5 expression may be associated with improved event-free survival in the adolescent/young adult and adult age group (P = .008; supplemental Figure 13A), consistent with our murine data. However, patient numbers were small. No significant differences were observed in event-free survival (supplemental Figure 13B-E) or overall survival (supplemental Figure 13F-G) in the low CXCR5 expression group in other B-ALL subtypes. Given the role of CXCR5 for the homing of B cells to the CNS in neuroborreliosis,³⁰ we hypothesized that CXCR5 expression may be predictive of B-ALL relapse in the CNS. Using the analysis of a publicly available clinical dataset of B-ALL blasts from patients with CNS and BM relapses,³⁷ we observed that CXCR5 expression may be higher in CNS compared with isolated BM relapses (P = .002; Figure 7F), although the number of included patients is low. Taken together, these data show that decreased CXCR5 expression in B-ALL cells leads to survival prolongation in mice. The CXCR5-CXCL13 axis may be relevant in human B-ALL, and higher CXCR5 expression in human B-ALL may be associated with CNS relapse.

Here we show that the BMM, in particular macrophages, age dependently influences B-ALL progression. We reveal pronounced differences in genome-wide gene expression and chromatin accessibility between macrophages from young vs old mice. These genes and DARs suggest an elevated inflammatory response in young BMM-derived macrophages. We implicate the CXCR5-CXCL13 axis as an influencing factor in B-ALL progression, although a role for direct cell contact between macrophages and leukemia cells cannot be excluded. We propose that the CXCR5-CXCL13 axis may act as prognostic marker or potential target for treatment of human B-ALL.

In support of our study, targeting of monocytes/macrophages in B-ALL²¹ or chronic lymphocytic leukemia,²³ which is characterized by increased expression of CXCR5/CXCL13,³⁸ led to a reduced leukemia burden, but we do not exclude the possibility that other cell types may also contribute to leukemia support or CXCL13 secretion. The impairment of macrophage function, including the reduction of phagocytic activity,³⁹ the reduced production of inflammatory cytokines,⁴⁰ and altered expression of certain genes, such as MHC class II molecules,⁴¹ by the aging process is well known and was largely also found in our study. Thereby, decreased CXCL13 secretion may be the consequence of decreased macrophage function or may be due to altered macrophage numbers, which in humans are known to decrease with aging. In mice, however, as also found in our study, macrophages have been shown to increase or remain constant with age.^39,42

Consistent with reports on a decrease of mitochondrial biogenesis,⁴³ structural alterations in mitochondria,⁴⁴ and damage to DNA, proteins, and lipids due to oxidative damage in mitochondria,⁴⁵ we showed morphological differences in old macrophages as well as increased ROS and DNA damage.⁴⁶ It is possible that increased ROS may modulate regulators of CXCL13 expression.⁴⁷ On the other hand, no significant differences were found between macrophages from young vs old mice with regards to Cxcl13 transcription, consistent with the absence of strong differences of open chromatin signatures at the transcriptional start site or gene body of Cxcl13. Therefore, the upregulated protein levels of CXCL13 are probably not directly transcriptionally regulated but may be due to differences in translation, protein stability, or degradation or posttranslational modification.

Our findings on a possible association of CXCR5 expression with CNS relapse are consistent with published data on the migration of precursor B-ALL cell lines with higher CXCR5 expression to the CNS, whereby CXCL13 was found to be expressed by the meninges.⁴⁸ CXCL13 has also been identified as a biomarker in CNS lymphoma.³¹ However, although our early human data may suggest a possible role for the CXCR5-CXCL13 axis in human B-ALL, the significance of these data, which are partially underpowered, is less well developed than the murine data and needs to be carefully assessed in the future. Nevertheless, in clear cell renal carcinoma, high expression of CXCR5/CXCL13 has been associated with tumor progression and worse prognosis.⁴⁹ 

Inhibition of the CXCR5-CXCL13 axis by small molecule inhibitors of CXCR5, antibodies to CXCL13, as performed in a murine model of breast cancer,⁵⁰ and, as recently shown, CXCR5-specific chimeric antigen receptor T cells,⁵¹ may be potential therapeutic approaches in B-ALL and could be assessed in the future.

In summary, we have demonstrated that the age of the BMM contributes to the leukemic phenotype, at least partly via increased secretion of CXCL13 by young compared with old BMM macrophages. Our data suggest that the CXCR5-CXCL13 axis may be of prognostic relevance in B-ALL and may offer a possible novel avenue for treatment.

The authors thank the High Throughput Sequencing Unit of the Genomics & Proteomics Core Facility, German Cancer Research Center (DKFZ), for library preparation (RNA-seq) and sequencing (RNA- and ATAC-seq).

This work was supported by grants from the Deutsche Kinderkrebsstiftung (A2016/03 [D.S.K.]), the National Institutes of Health, National Cancer Institute (UG1 CA233338 and U24 CA196171 [C.D.B.] and R35 CA197695 and CA021765 [C.G.M.]), and The Little Princess Trust (CCLGA 2017 13 [C.H.]).

Contribution: C.Z. designed and carried out experiments, analyzed the data, and critically reviewed the manuscript; J.E., E.S.W., R.K., V.R.M., and C.K. assisted with experiments; P.S.G. assisted with the in vivo microscopy experiments; M.H., J.H., K.B., C.P., and D.B.L. performed and analyzed the RNA-seq and ATAC-seq experiments and critically reviewed the manuscript; M.M. provided patient samples; R.P.H., K.B., and S.H. provided patient samples and assisted with histopathology-related questions; Z.G. and K.G.R. provided and analyzed the data on patients with human B-ALL; W.S. and C.G.M. contributed patients for clinical trials; C.D.B. contributed patients for clinical trials and critically reviewed the manuscript; C.H. and A.F.C. contributed and analyzed data on CNS involvement in pediatric patients; N.F. assisted with statistical questions and critically reviewed the manuscript; and D.S.K. designed experiments, supervised the project, analyzed data, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Daniela S. Krause, Georg-Speyer-Haus, Institute for Tumor Biology and Experimental Therapy, Paul-Ehrlich-Str 42-44, 60596 Frankfurt am Main, Germany; e-mail: krause@gsh.uni-frankfurt.de.