Distinct roles of mesenchymal stem and progenitor cells during the development of acute myeloid leukemia in mice

Acute myeloid leukemia (AML) is characterized by accumulation of immature myeloid blasts in blood, bone marrow (BM), and other organs.^1,2  The efficacy of current treatments for AML targeting leukemic cells has been disappointing, and relapse is common.^3,4  Therefore, there is urgent need to identify new therapeutic targets to develop additional treatment strategies for AML. It has been believed that the disease persistence is attributed to residual leukemia-initiating cells or leukemic stem cells (LSCs) that are protected by a specialized BM microenvironment, the so-called hematopoietic stem cell (HSC) niche.^5-9  These leukemic cells outcompete normal HSCs for the niche occupancy,¹⁰  which ultimately causes disruption of normal hematopoiesis and mortality. Therefore, efforts have been put into untangling the complex interactions between leukemic cells and neighboring stromal cells.^11,12  Some of the niche components have been proposed to be critical as the novel candidate target for therapies in AML.^13,14 

The BM HSC niche is composed of various types of stromal cells, including osteoblasts, adipocytes, perivascular cells, endothelial cells (ECs), mesenchymal stem cells (MSCs), and mesenchymal progenitor cells (MPCs).^11,15  MSCs are the precursors of mesenchymal lineages like osteoblasts, adipocytes, and chondrocytes.¹⁶  BM MSCs are enriched in CD45⁻TER119⁻CD31⁻CD44⁻ stromal cells¹⁷  and can be isolated based on their expression of SCA1 and CD51 or PDGFRA/CD140A.^18-20  The MSCs (SCA1⁺CD51⁺ or SCA1⁺PDGRA⁺) are functionally estimated by their ability to form fibroblast colony-forming units (CFU-Fs) in vitro, and they can generate more differentiated MPCs (SCA1⁻CD51⁺) with single or bilineage potential but with little to no CFU-F activity.^15,21  The SCA1⁻CD51⁺ MPCs, largely overlapping (75%) with Nestin-GFP⁺ cells, have been shown to be enriched for HSC maintenance genes.²² 

The HSC niche tightly regulates normal hematopoiesis by controlling HSC fate and contributes to leukemogenesis.¹¹  The niche requirement for AML development was considered to be disease-stage specific in a study showing that the niche was required for leukemogenesis in a preleukemic stage, although it became permissive once leukemia was established.²³  Then, malignant hematopoietic cells could actively reprogram the BM niche into a self-reinforcing leukemic niche, thereby propelling leukemia.^24,25  A similar finding was reported in acute lymphoid leukemia.²⁶ 

The niche function and structure in leukemia has been reported to be disease-type specific,²⁷  emphasizing that a thorough BM niche characterization in different types of leukemia is required in order to identify possible disease-specific therapeutic targets. Although BM niche alterations have been reported in the MLL-AF9 AML mouse model,¹⁴  the dynamic alterations and the roles of different BM cellular niche components in AML remain largely unexplored.

We here demonstrate dynamic functional and molecular alterations of BM MSPCs in MLL-AF9 AML mouse BM that are associated with AML burden. We also show that early B-cell factor 2 (Ebf2)⁺ cells, a highly purified primitive MSC population identified in mouse BM,²⁸  are dramatically expanded in AML mouse BM and primed toward differentiation. Ebf2⁺ cells constitute the major fraction of AML niche. Most importantly, depletion of Ebf2⁺ cells leads to an accelerated development of the AML. The dysregulations of HSC niche factors in AML MSPCs may aid in identification of novel therapeutic targets for treatment of AML.

Wild-type CD45.2 C57BL/6J or CD45.1 B6.SJL-Ptprca Pepcb/BoyJ (The Jackson Laboratory) or Ebf2-Egfp reporter FVB/N mice²⁸  at 8 to 12 weeks were used for transplantation of MLL-AF9 AML cells. Triple-transgenic Ebf2-Egfp × Ebf2-Cre^ER × Rosa26-^loxpStop^loxp-Tomato mice were generated by crossing Ebf2-Egfp with Ebf2-Cre^ER × Rosa26-^loxpStop^loxp-Tomato mice and used to trace Ebf2⁺ cells. Ebf2-Cre^ER × Rosa26-^loxpStop^loxp-Dta mouse models were used for specific deletion of Ebf2⁺ cells in vivo. Mice carrying one of the transgenes were used as controls. Mice were injected with tamoxifen (TAM) (Sigma) intraperitoneally at 3 mg/20 g body weight every second day 3 times to induce recombination. All mice were maintained in specific-pathogen–free conditions in the animal facility of Karolinska Institute. Animal procedures were performed with approval from the local ethics committee (ethical number S40-14) at Karolinska Institute (Stockholm, Sweden).

Human and mouse MSCs were isolated as described previously.¹⁷  See supplemental Materials and methods for the detailed procedure.

The AML mouse model was generated by transplanting mouse BM KIT⁺ cells transduced with MLL-AF9 retrovirus as described previously.²⁹  BM CD45.1⁺ KIT⁺ cells for virus transduction were first enriched by magnetic-activated cell sorting using KIT-microbeads (Miltenyi Biotec) and then sorted by FACS from 8- to 10-week-old C57BL/6J or FVB/N mice. Cells were then transduced with MLL-AF9 retrovirus. Transduced cells were clonally propagated and selected by colony assay in methylcellulose M3434 and subsequently by transplantation. MLL-AF9–expressing cells were sorted by FACS from primary recipient mice that developed AML and then expanded in culture in presence of interleukin-3 (10 ng/mL) in RPMI + 10% fetal bovine serum for secondary transplantation to establish the AML mouse model. Finally, 50 000, 250 000, or 1 million MLL-AF9–expressing cells were intravenously transplanted into nonirradiated mice.

BM samples were collected from adult or pediatric patients with AML at diagnosis and healthy donors (30-45 years old). The experiments were approved by the local ethical committee at Stockholm (2012/4:10, 2013/3:1 and 2013/1248-31/4), and informed consent was obtained from the patients or guardians and healthy donors. Mononuclear cell isolation from the BM samples was done as previously described.¹⁷ 

Primary BM or peripheral blood (PB) mononuclear cells from adult (n = 3) and pediatric patients (n = 3) with AML were transplanted via intrafemoral injection into human cytokine engineered immunodeficient NSG-SGM3 mice (Jackson Laboratory) at doses of 100 000-500 000 cells/mouse. One of 6 patient samples was with MLL-AF9 mutation. The recipient mice were subjected to sublethal irradiation (220 cGy). The BM was collected from the recipients at 6 to 8 weeks after transplantation for analyzing leukemia engraftment and BM stromal cell alterations.

CFU-F assay was performed as described previously.^17,28 

Unpaired Student t and Mann-Whitney U tests were used to determine the differences based on data distribution. Pearson or spearman correlation was applied to analyze the correlation analysis. The Kaplan-Meier survival curve of the mice was generated using a log-rank (Mantel-Cox) test in Prism 6.0. All reported P values were obtained using Prism 6.0 software, and P < .05 was considered statistically significant.

See supplemental Materials and methods for additional methods.

In order to explore leukemia-induced alterations in the BM niche, we used a syngeneic AML mouse model²⁹  by transplanting MLL-AF9–transduced CD45.1⁺ BM KIT⁺ cells into nonirradiated CD45.2⁺ wild-type mice after clonal selection in culture and transplantation. When 250 000 MLL-AF9–expressing cells were transplanted, most C57BL/6 mice developed AML with thrombocytopenia and splenomegaly within 27 to 30 days after transplantation (Figure 1A-C). At the late stage of AML, the number and frequencies of total stromal cells (CD45⁻TER119⁻) in the BM remained unchanged in AML mice compared with control mice transplanted with normal KIT⁺ hematopoietic stem and progenitor cells (HSPCs) (Figure 1D-E; supplemental Figure 1A). However, consistent with the previous findings,¹⁴  total CD45⁻TER119⁻CD31⁺ ECs and SCA1⁺CD31⁺ arteriolar ECs, but not sinusoidal ECs, were dramatically increased in AML mice (Figure 1E-G; supplemental Figure 1B). The proportions of CD45⁻TER119⁻ CD31⁻CD44⁻CD51⁺SCA1⁺ MSCs and CD45⁻TER119⁻CD31⁻CD44⁻CD51⁺SCA1⁻ MPCs were significantly increased in AML mice compared with control mice (Figure 1D,H-I). This finding was further supported by the increased frequencies and absolute counts of CFU-Fs in AML mouse BM. The expansion of AML MSPCs might be due to reduced apoptosis, since we detected increased expression of the antiapoptotic gene Bcl2 and reduced apoptotic cells in AML MSPCs (supplemental Figure 2A-C). There was no significant change in either cell-cycle status or cell-cycle regulator expression in AML MSCs (supplemental Figure 2D-F). Nevertheless, we observed an increase of AML MPCs residing in S/G2/M stage, even though Cdk6 expression declined in AML MPCs (supplemental Figure 2E-F).

To assess the functionality of the AML MSCs, we performed a CFU-F and multilineage differentiation assay on the sorted MSCs. AML MSCs contained a frequency of CFU-Fs similar to that of control mice (Figure 1J-K), and the colony size derived from AML MSCs was comparable to control MSCs (supplemental Figure 3A). However, AML MPCs appeared to form more CFU-Fs but fewer osteoblast colonies (supplemental Figure 3B-C). We did not observe any difference in population doubling time between AML MSCs and control MSCs (supplemental Figure 3D), suggesting normal proliferation capacity of the AML MSCs.

However, AML MSCs exhibited enhanced osteogenic and adipogenic differentiation potential (Figure 1L-M; supplemental Figure 3E). Similar to a previous report,¹⁴  micro–computed tomography analysis indicated a reduced femoral bone density in AML mice (supplemental Figure 3F). However, we did not detect significant difference in osteocalcin expression between control and AML bones (n = 3; supplemental Figure 3G-H).

To determine the dynamic changes of the BM niche in relation to AML engraftment, we analyzed the kinetics of BM cellular niche composition and AML engraftment. As expected, normal HSPC engraftment in PB, spleen, and BM remained <0.1% in the nonirradiated recipients after transplantation, whereas AML engraftment was <0.01% of total CD45⁺ cells within 7 days after transplantation but progressively increased with time thereafter (Figure 2A-B). BM cellular niche alterations closely correlated with AML burden in recipient BM (Figure 2C) and appeared at 21 days after transplantation (Figure 2D; supplemental Figure S4). However, BM cellular niche components remained relatively constant after transplantation of normal HSPCs (supplemental Figure 4A), indicating the niche alterations are AML specific.

Taken together, our data suggest that AML cells alter BM cellular niche composition in a leukemia burden–dependent manner. Moreover, the AML skewed MSC differentiation potential toward adipocytes and osteoblasts.

To test whether primary BM AML cells from patients and mouse AML cells exert a similar effect on the BM niche, we transplanted primary AML mononuclear cells from adult or pediatric patients directly into femurs of sublethally irradiated NSG-SGM3 mice.³⁰  Control mice were injected with phosphate-buffered saline (n = 12) or healthy donor CD34⁺ cells (n = 2). Human AML cell engraftment in the injected femur varied from 3.8% to 94.6% at 6 to 20 weeks after transplantation (Figure 3A-B). The engraftment of patient AML cells did not lead to any changes in total BM cellularity of the mice (Figure 3D). Similar to the finding in the MLL-AF9 mouse model, both the frequencies and numbers of the ECs, MSCs, and MPCs were significantly increased in recipient BM engrafted with patient AML cells compared with control mice (Figure 3C,E). These data suggest that primary AML cells from patients could remodel BM stroma.

To investigate the molecular mechanisms underlying the massive proliferation of AML cells and impaired normal hematopoiesis, we next examined the expression of HSC niche genes and mesenchymal lineage–associated genes by quantitative polymerase chain reaction (qPCR) (Figure 4A-C). Angiopoietin-like 1(Angptl1) and collagen type I (Col1a1) were significantly reduced, while Kit ligand (Kitl) was higher in AML MSCs than control MSCs (Figure 4A). These genes as well as C-X-C motif chemokine ligand 12 (Cxcl12), interleukin-7 (Il7), and Nov were downregulated in AML MPCs, while osteopontin (secreted phosphoprotein 1 [Spp1]), Il6, Lama4, and Jag1 were upregulated in these cells (Figure 4B). Tgfb1 and Lama5 were significantly increased in AML ECs (Figure 4C). Consistent with the increased in vitro adipogenic and osteogenic differentiation potential of AML MSCs, Fabp4 was increased in AML MSPCs. However, the early osteoblast differentiation gene Sp7 (osterix) and Runx2 remained unaltered, while the late-stage osteoblast marker Col1a1 and Bglap (osteocalcin) decreased in AML MSPCs, indicating a possible osteoblast maturation block in AML. Interestingly, Sipa1, whose expression in the BM niche is critical for maintaining normal hematopoiesis,³¹  was significantly decreased in AML MSCs and MPCs (Figure 4A-B).

In line with the dynamic changes in BM cellular niche composition in AML mice, the major BM niche population MPCs showed molecular alterations 14 days after transplantation (supplemental Figure S5; Figure 5) and prior to the massive expansion and egress of AML cells (Figure 2). Alterations in most of the HSC niche factors in AML MSPCs seemed to be closely associated with the level of AML engraftment (supplemental Figure 5C; Figure 5C). These data provide molecular evidence for dynamic changes of HSC niche factors in AML MSPCs.

We recently identified a stromal cell population highly enriched with MSCs marked by Ebf2 in mouse BM.²⁸  Ebf2⁺ cells account for ∼6% of total CD51⁺SCA1⁺ MSCs and partly overlap with nestin⁺ cells in mouse BM.²⁸  To evaluate the involvement of Ebf2⁺ cells in AML niche reprogramming, we next analyzed Ebf2⁺ MSPCs after AML establishment using Ebf2-reporter mice (Ebf2-Egfp) as recipients for the transplantation (Figure 6A). We found an expansion of Ebf2⁺ cells in AML mouse BM compared with that in control mice (Figure 6B-C). However, further analysis revealed a reduced MSC fraction (SCA1⁺CD51⁺) but an increased fraction of the more differentiated MPC population (SCA1⁻CD51⁺) within Ebf2⁺ cells (Figure 6D). CFU-Fs in AML Ebf2⁺ cells seemed to be smaller than those in control Ebf2⁺ cells (supplemental Figure 6A), even though CFU-F frequencies in Ebf2⁺ cells were similar between AML and normal mice (Figure 6E). Notably, more CFU-Fs from AML Ebf2⁺ cells contain adipocytes (8 out of 9 in AML and 1 out of 3 in control) (supplemental Figure 6B), supporting the notion of functional alterations of AML Ebf2⁺ cells with a biased differentiated capacity toward adipocytes. Furthermore, Cxcl12 mRNA was significantly reduced in AML Ebf2⁺ cells (Figure 6F). These data suggest that AML cells induce a shift of Ebf2⁺ MSCs toward differentiation into a less primitive MPC population. Similar to a previous report,¹⁴  more MSPCs, including Ebf2⁺ cells, appeared in the central marrow of AML mice than in controls (supplemental Figure 6C-D).

To further determine the contribution of Ebf2⁺ cells during AML progression, we took advantage of triple-transgenic Ebf2-Egfp × Ebf2-Cre^ER × Rosa26-^loxpStop^loxp-Tomato mice, where Ebf2⁺ cells and their progeny are reported by enhanced GFP and tomato protein, respectively. AML cells were injected 1 month after activation of tomato expression by TAM treatment. The progeny (tomato⁺) derived from Ebf2⁺ cells were detected by FACS (Figure 6G). Interestingly, Ebf2⁺ MSPCs could give rise to all fractions of BM stromal cells, including the CD44⁻ stromal cells enriched with MSPCs and CD44⁺ mature stromal cells (Figure 6H). In line with the enhanced proliferation of Ebf2⁺ MSPCs in AML BM, the frequency of the Ebf2⁻tomato⁺ stromal cell fraction representing cells generated from Ebf2⁺ MSPCs was significantly increased in AML mouse BM (Figure 6I). Both BM MSCs and MPCs within tomato⁺ cells were significantly increased in AML mice (Figure 6I). Tomato⁻ MSPCs, possibly derived from non-Ebf2⁺ cells or Ebf2⁺ cells that escaped TAM treatment, were also increased in the AML BM (Figure 6J). However, tomato⁺ MSCs and MPCs accounted for approximately two-thirds of the total MSPC population (Figure 6K), indicating significant contribution of Ebf2⁺ cells to AML niche formation. This result suggests that Ebf2⁺ cells might be a key niche element contributing to the formation of the AML BM niche.

The alteration of Ebf2⁺ cells and their participation in AML BM niche formation suggest that these cells have a potentially important role in symptomatic AML onset and progression. To functionally determine the role of Ebf2⁺ cells in AML development, we used transgenic Ebf2-Cre^ER × Rosa26-^loxpStop^loxp-Dta mice as recipients for establishing an AML mouse model. At steady state, in vivo depletion of Ebf2⁺ cells led to enhanced myelopoiesis 2 months after TAM injection (supplemental Figure 7), suggesting an important role for these cells in maintaining normal hematopoiesis. We intravenously transplanted CD45.1⁺MLL-AF9 AML cells into nonirradiated CD45.2⁺ double-transgenic Ebf2-Cre^ER × Rosa26-^loxpStop^loxp-Dta or single-transgenic control mice either 1 month after Ebf2⁺ cell depletion or 1 week prior to depletion (Figure 7A). Surprisingly, we observed a shorter latency of symptomatic AML and an overall reduced survival time in Ebf2⁺ cell–depleted recipient mice compared with control mice, regardless of whether Ebf2⁺ cell depletion was induced before or after AML cell transplantation (Figure 7B-C). The onset of symptomatic AML was observed at ∼29 to 55 days after transplantation in control mice and 22 to 32 days in Ebf2⁺ cell–depleted mice. FACS analysis indicated much higher total engraftment of MLL-AF9 AML cells (CD45.1⁺) in the PB of Ebf2⁺ cell–depleted recipients than in nondepleted recipient mice 22 days after transplantation (Figure 7D-E). Disease acceleration following Ebf2⁺ cell depletion was manifested by faster development of splenomegaly, increased circulating leukemic blasts in the PB, and more profound AML cell infiltration in the spleen and liver of Ebf2⁺ cell–depleted recipient mice compared with control mice (supplemental Figure 8A-B; Figure 7E). Normal HSPCs were lost in Ebf2⁺ cell–depleted recipient BM (supplemental Figure 8D-E). While HSCs, granulocyte/macrophage progenitors, and pre–erythrocyte colony-forming units remained in the Ebf2⁺ cell–depleted recipient spleen, pre–megakaryocyte-erythroids were diminished in recipient spleen (supplemental Figure 8C-D). In line with this, total AML cell engraftment and the frequency of AML KIT⁺ cells enriched with LSCs²⁹  were dramatically higher in the Ebf2⁺ cell–depleted recipient BM and spleen than in nondepleted mice (supplemental Figure 8E). Notably, the increased AML cell engraftment and AML acceleration were not related to the initial homing (3 hours after transplantation) of AML cells, since there was no obvious homing advantage of AML cells into Ebf2⁺ cell–depleted recipient BM and spleen (Figure 7F-G). Taken together, our data suggest that deletion of Ebf2⁺ cells accelerated AML development. Thus, Ebf2⁺ cells are important for maintaining normal hematopoiesis and may act as a suppressor for AML development.

BM HSC niche contribution to leukemia initiation and progression has been increasingly recognized.^11,12  However, the contributions of BM MSPCs to leukemia niche formation and progression remain poorly defined. Thus, there is great need to understand the role of BM MSPCs in leukemia in order to identify disease-specific therapeutic targets in the niche. We here demonstrate that MLL-AF9⁺ AML cells reconstructed the BM HSC niche by inducing abnormal expansion and differentiation of BM MSPCs. In vivo fate-mapping experiments suggested an involvement of Ebf2⁺ MSPCs in AML niche formation. Strikingly, the depletion of Ebf2⁺ MSPCs resulted in AML acceleration, emphasizing a critical role for Ebf2⁺ cells in AML onset and progression and providing new evidence for the contribution of MSPCs to AML development.

Our dynamic characterization reveals a correlation between BM cellular niche alterations and AML cell engraftment. The molecular alterations partially appeared before the massive expansion of AML cells. The progressive deregulation of HSC maintenance genes and inflammatory cytokines in AML BM MSPCs was closely associated with AML burden. This may represent molecular mechanisms by which normal hematopoiesis is impaired while AML cells massively proliferate. Altogether, our findings provide cellular and molecular evidence for dynamic alterations of the BM HSC niche in AML. AML cells may actively reconstruct the HSC cellular niche into a potentially self-reinforcing leukemic niche. The “educated” BM niche could, in turn, contribute to the massive proliferation of leukemic cells and failure of normal hematopoiesis. Strategies to restore BM niche homeostasis may provide a means to eliminate dormant LSCs after treatment and thereby impede AML progression.

Previous studies showed that MLL-AF9 AML¹⁴  and BCR-ABL CML cells²⁵  led to increased bone remodeling with an accumulation of osteoblast-primed MSPCs accompanied by reduced numbers of mature osteoblasts. In line with these findings, our data suggest an enhanced capacity of AML MSCs to generate more differentiated MPCs than MSCs, which is reflected in the expansion of MPCs but the reduction of the MSC fraction within the defined Ebf2⁺ MSPCs. The differentiation bias of MSCs is consistent with their stronger potential to generate osteoblasts and adipocytes in vitro. However, the reduction of Col1a1 and Bglap/osteocalcin transcripts in native AML MPCs suggests that the differentiation of MPCs to mature osteoblasts is blocked at a late development stage in vivo, as supported by micro–computed tomography and a previous study,¹⁴  and this block could also contribute to the accumulation of MPCs in AML.

Importantly, this study shows the critical role of Ebf2⁺ cells in AML BM niche remodeling, as illustrated by their contribution to BM stromal cell turnover. Consistent with our previous finding,²⁸  lineage-tracing experiments demonstrated that Ebf2⁺ cells could generate all BM stromal cell subsets, including MSPCs. These findings suggest a high hierarchical position of Ebf2⁺ MSCs in BM mesenchymal cell development and a potential role of these cells in maintaining BM cellular niche homeostasis and in AML niche formation. During AML establishment, Ebf2⁺ cells preferentially generated more differentiated MPCs than the MSCs. The depletion of Ebf2⁺ MSPCs may further disrupt the balance between primitive MSCs and more differentiated MPCs, thereby accelerating AML progression.

During AML progression, leukemic cells outcompete normal hematopoietic cells, which leads to pancytopenia and infiltration of leukocytes in the BM, blood, and other organs.³²  However, the underlying molecular mechanisms remain to be defined. Cxcl12, Kitl, Nov, Igf1, and Angptl1 are critical for maintaining normal hematopoiesis through keeping HSC in quiescence^21,33-36  and lodgment in their niche.³⁷  Decreased expression of these genes in AML MSCs and MPCs could affect HSC retention and quiescence and eventually exhaust the normal HSC pool while promoting leukemic cell proliferation, leading to the progression of leukemia, as reported previously.^13,14,38 

Accumulated evidence indicates that inflammatory cytokines such as Tgfb1 and Il6 are involved in the development of myeloid malignancies^11,12  and are often increased in patients with myeloproliferative neoplasms.^39,40  We here report that Il6 and Tgfb1 transcripts are upregulated in AML BM MSPCs and ECs, respectively. In addition, it has been shown that SPP1 (OPN) is a critical negative regulator of normal HSC self-renewal,⁴¹  and the interactions between SPP1 and lymphoblastic leukemic cells can induce dormancy of leukemic cells.⁴²  In this study, the upregulation of Spp1 and Il6 in the BM MSPCs has already occurred before the expansion of AML. Thus, the increased Spp1 in the AML BM niche might contribute to the exhaustion of normal HSCs and thereby to AML progression, although more work is required to test this hypothesis. Spp1 upregulation is also consistent with the biased osteoblastic differentiation of AML MSCs, which was reflected in the preferential expansion of Ebf2⁺ MPCs in AML mice. These data provide evidence of the molecular mechanisms by which AML cells remodel the BM HSC niche to create a self-reinforcing leukemic niche.

Laminin α4 chain (Lama4) and α5 chain (Lama5) are active chains forming the cell-binding domains in the extracellular matrix proteins laminin-411, laminin-421, laminin-423 and laminin-511 and laminin-521, respectively.⁴³  Previous studies have shown that the interactions between HSC and laminins are critical for adult BM HSC homing and engraftment.^44,45 Lama4 upregulation was detected in CML mice treated with placental growth factor, a factor promoting CML proliferation.⁴⁶  The upregulation of Lama4 and Lama5 in the AML BM niche warrants further investigation of the role of the laminin in AML progression. Furthermore, Sipa1 reduction in AML MSCs and MPCs might contribute to AML progression, since Sipa1 loss could induce BM niche alterations, causing myeloproliferative neoplasms.³¹ 

In summary, our findings provide new evidence for MLL-AF9 AML–induced dynamic niche alterations and the temporal roles of BM MSPCs during the development of AML. BM native Ebf2⁺ cells suppress AML onset and progression. However, these cells can be functionally altered by AML cells to constitute the AML niche. In vivo fate-mapping using Ebf2 reporter mouse models demonstrated the significant contribution of Ebf2⁺ cells to AML niche formation. The suppressive role of native Ebf2⁺ cells in AML onset and progression warrants future exploration of the underlying molecular mechanisms involved. Importantly, molecular tools that restore BM MSC function and the molecular niche may be a potentially powerful strategy to suppress AML development.

Confocal images were obtained at the Live Cell Imaging unit, Department of Biosciences and Nutrition, Karolinska Institute, Huddinge, Sweden. The authors are grateful to Sten Eirik W. Jacobsen at Karolinska Institute for his scientific input on the study. The authors acknowledge the MedH Core Flow Cytometry facility (Karolinska Institute) for providing cell-sorting/analysis services. The authors thank the Nordic Society of Pediatric Hematology and Oncology (http://www.nopho.org) for scientific discussion and providing precious samples from their biobank.

This study was supported by the Swedish Research Council (K2013-99X-22241-01-5), the Swedish Childhood Society (TJ2013-0048, PR2015-0142, and PROJ12/081), the Swedish Cancer Society (CAN 2015/652 and CAN 2012/891), the Åke Olsson Foundation, Radiumhemmets Forskningsfonder, and Karolinska Institute Wallenberg Institute for Regenerative Medicine (H.Q.).

Contribution: H.Q. and P.X. designed and performed experiments, analyzed data, and wrote the manuscript; L.S. designed and performed experiments, analyzed data, and participated in manuscript writing; Y.H. generated MLL-AF9 AML cells and helped with transplantation experiments; M.K. performed, analyzed data and participated in the manuscript writing; T.B., M.D., and A.-S.J. assisted in part of the study and with manuscript review; J.W., M.S., and M.E. provided scientific input on the study and reviewed the manuscript; and all authors approved the final version of the manuscript.

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

Correspondence: Hong Qian, Center for Hematology and Regenerative Medicine, Department of Medicine, Karolinska Institute, Karolinska University Hospital, SE-141 86 Stockholm, Sweden; e-mail: hong.qian@ki.se.