|Home | About | Journals | Submit | Contact Us | Français|
DNA methylation is essential for development and in diverse biological processes. The DNA methyltransferase Dnmt1 maintains parental cell methylation patterns on daughter DNA strands in mitotic cells, however, the precise role of Dnmt1 in regulation of quiescent adult stem cells is not known. To examine the role of Dnmt1 in adult hematopoietic stem cells (HSCs), we conditionally disrupted Dnmt1 in the hematopoietic system. Defects were observed in Dnmt1 deficient HSC self-renewal, niche retention, and in the ability of Dnmt1 deficient HSCs to give rise to multilineage hematopoiesis. Loss of Dnmt1 also had specific impact on myeloid progenitor cells, causing enhanced cell cycling and inappropriate expression of mature lineage genes. Dnmt1 regulates distinct patterns of methylation and expression of discrete gene families in long-term HSCs, multipotent and lineage-restricted progenitors, suggesting that Dnmt1 differentially controls these populations. These findings establish a unique and critical role for Dnmt1 in the primitive hematopoietic compartment.
DNA methylation is an epigenetic mechanism essential for normal development, influencing cellular events such as transcription, genomic imprinting and genome stability (Jaenisch, 1997). DNA methyltransferases (Dnmts) are the group of enzymes responsible for establishment and maintenance of genomic DNA methylation. They include the de novo methyltransferases Dnmt3a and Dnmt3b, and the maintenance methyltransferase Dnmt1. All three are essential genes for embryonic development (Lei et al., 1996; Li et al., 1992; Okano et al., 1999), and have been shown to be critical for survival of many somatic cell types, including mouse embryonic fibroblasts (Jackson-Grusby et al., 2001) and CNS neurons (Fan et al., 2001). This is in clear contrast to embryonic stem cells, which can be derived and maintain their stem cell properties without Dnmt1, Dnmt3a or Dnmt3b; essentially in the absence of DNA methylation (Tsumura et al., 2006). The question remains as to whether adult somatic stem cells are critically regulated by DNA methylation, like their differentiated counterparts, or are less dependent upon DNA methylation and thus more similar to embryonic stem cells.
Interestingly, a conditional knockout study of the de novo methyltransferases Dnmt3a and Dnmt3b in adult hematopoietic stem cells (HSCs) demonstrated that DNA methylation by these enzymes is critical for self-renewal of HSCs but not for their differentiation to progenitors and mature cells (Tadokoro et al., 2007). Here, we utilized an inducible, conditional knockout approach to examine consequences of loss of the maintenance methyltransferase Dnmt1 in HSCs and hematopoiesis in vivo. We demonstrate that Dnmt1 is essential for HSC self-renewal, niche retention and proper differentiation of the myeloid lineage. We suggest that adult somatic stem cell fate transitions, critical for stepwise differentiation, changes in cycling and the process of mobilization, are especially dependent upon maintenance methylation induced by Dnmt1.
To evaluate the role of Dnmt1 in maintenance of hematopoiesis, we utilized Dnmt1fl/fl mice (Jackson-Grusby et al., 2001). Crossing Dnmt1fl/fl and Gata1-Cre transgenic mice to elicit germline deletion (Jasinski et al., 2001) resulted in no viable Dnmt1 deficient (Dnmt1Δ/Δ) animals (Table S1), consistent with the embryonic lethal phenotype of conventional Dnmt1 knockout mice (Li et al., 1992). We then inactivated Dnmt1 in the hematopoietic system by breeding Dnmt1fl/fl with interferon-inducible Mx-Cre transgenic animals (Figure 1A). Genomic PCR and Western blot from whole BM confirmed Dnmt1 deletion (Figure 1B and 1C). Analysis of the peripheral blood (PB) of control (Mx-Cre−Dnmt1fl/fl, pIpC injected) and Dnmt1Δ/Δ mice revealed comparable numbers of PB leukocytes, platelets, red blood cells (RBCs) and hemoglobin levels (Figure 1D). At 12 weeks post pIpC, no changes were found in Dnmt1Δ/Δ PB lineage distribution, total BM cellularity, or BM lineage composition (Figure S1). These data demonstrate that hematopoiesis is essentially normal in the absence of Dnmt1.
As Mx-Cre has been observed to induce deletion in both hematopoietic cells and the BM microenvironment (Walkley et al., 2007), we adopted a transplantation strategy to determine the intrinsic role of Dnmt1 in HSCs and progenitors. Whole BM from Mx-Cre−Dnmt1fl/fl or Mx-Cre+Dnmt1fl/fl mice (both CD45.2+) was transplanted into lethally irradiated recipient mice (CD45.1+), and all mice were treated with pIpC at 4 weeks post-transplant (Figure 1E). Following pIpC treatment, the PB chimerism of Dnmt1Δ/Δ cells was significantly decreased and continued to decline over the following 12 weeks (Figure 1E). By 16 weeks post-transplant, Dnmt1Δ/Δ cells were able to generate only 50-80% of mature cells in the BM (Figure 1F). Donor CD45.2+ BM cells were then re-isolated and injected into secondary recipients. This revealed a further reduction in the ability of Dnmt1Δ/Δ HSCs to generate mature cells (Figure 1G). Interestingly, the contribution of Dnmt1Δ/Δ HSCs to the myeloid lineage was most significantly affected as opposed to the lymphoid lineages (Figure 1G). Thus, Dnmt1 is required for presumptive self-renewal divisions of HSCs as well as differentiation, with particular defects observed in the ability of Dnmt1Δ/Δ HSCs to generate mature myeloid cells.
We further evaluated Dnmt1Δ/Δ HSC functional capacity in a competitive repopulation assay. Whole BM from Mx-Cre−Dnmt1fl/fl or Mx-Cre+Dnmt1fl/fl mice was transplanted at a 1:1 ratio with wild-type (WT) competitor BM (CD45.1+CD45.2+) into lethally irradiated recipient mice and all mice were treated with pIpC at 4 weeks post-transplant (Figure 2A). Following pIpC, the PB chimerism of Dnmt1Δ/Δ cells was significantly decreased and continued to decline over 16 weeks (Figure 2B). At 16 weeks post-transplant, Dnmt1Δ/Δ cells were able to generate only 5-20% of mature cells in the BM (Figure 2C). To closely examine this defect, we evaluated the kinetics of loss of PB multilineage cells. The initial contribution of Mx-Cre−Dnmt1fl/fl or Mx-Cre+Dnmt1fl/fl cells to mature myeloid and lymphoid lineages was equivalent (Figure 2D). By 4 weeks post pIpC, the contribution of Dnmt1Δ/Δ cells to the myeloid lineage was barely detectable and remained low for 16 weeks (Figure 2D). In contrast, the contribution of Dnmt1Δ/Δ cells to the B and T lymphoid lineages remained at levels equal to or greater than control (Figure 2D). These results demonstrate that the myeloid lineage is particularly sensitive to Dnmt1 loss in the context of competition with WT BM cells. To assess the self-renewal capacity of Dnmt1Δ/Δ HSCs, donor BM cells were re-isolated at 16 weeks post-transplant, mixed with fresh competitor BM cells and injected into secondary recipient mice. Dnmt1Δ/Δ HSCs were unable to engraft secondary recipients in the presence of WT BM cells (Figure 2E). This phenotype may be accounted for by several mechanisms, including decreased self-renewal, defects in homing to the BM niche, or niche retention.
To evaluate the mechanism of action of Dnmt1 loss on HSC engraftment, we first assessed the ability of HSCs and progenitor cells to home to the BM by intravenous (IV) injection of CFSE-labeled control and Dnmt1Δ/Δ Lin−Sca-1+ cells into lethally irradiated hosts. No difference was observed in the ability of Dnmt1Δ/Δ CFSE+ cells to home to the BM compared to control cells (Figure 3A). The ability of Dnmt1 deficient HSCs to be retained in the BM niche was examined by non-ablative transplant where whole BM or purified long-term HSCs (LT-HSCs, LKS+CD34−/lo) from WT mice (CD45.1+CD45.2+) were transplanted into control or Dnmt1Δ/Δ mice (Figure 3B). The level of engraftment of WT cells into unconditioned control animals was similar to background staining (Figure 3C), indicating that the HSC niche space in the BM remains stably occupied by host cells. Comparatively, the level of engraftment of WT cells into unconditioned Dnmt1Δ/Δ mice is nearly 50-fold higher (Figure 3C), demonstrating that Dnmt1Δ/Δ HSCs are outcompeted for presence in the BM niche by WT cells. Taken together, these data indicate that retention within the BM niche, rather than homing to the niche, contribute to the observed defect in Dnmt1 deficient HSC engraftment. Interestingly, the total number of LT-HSCs and short-term HSCs/multipotent progenitors (ST-HSC/MPPs, LKS+CD34+) within the BM of Dnmt1Δ/Δ mice were not significantly different from controls (Figure 3D), indicating that these cells are retained in the BM in the absence of competing WT cells.
To determine the molecular mechanism underlying the observed defects in Dnmt1 deficient HSCs, we performed global gene expression analysis and bisulfite sequencing. We examined the short-term consequences of Dnmt1 deletion (4 days post pIpC) to favor identification of the most direct effects of Dnmt1 loss. We first examined DNA methylation of the well-characterized intracisternal A-type particle (IAP), an endogenous retrotransposable element that is methylated in somatic cells by Dnmt1 (Gaudet et al., 2004). Bisulfite sequencing of the IAP promoter region demonstrated an extensive reduction in methylation in Dnmt1Δ/Δ LT-HSCs, ST-HSC/MPPs and myeloid progenitors (Figure S2a). Consistent with this loss in methylation, IAP transcript was upregulated in these populations (Figure S2b).
Having confirmed in our system that the loss of Dnmt1 derepresses a known target, we then compared gene expression profiles of Dnmt1Δ/Δ versus control LT-HSCs, ST-HSC/MPPs and myeloid progenitors (denoted “Cre+/Cre−”). 484 genes were found to be differentially expressed with an absolute fold change >1 (Figure S3 and Table S2). Selected candidates were validated by real-time PCR (Figure S4). Overall, the expression changes in Dnmt1Δ/Δ LT-HSCs, ST-HSC/MPPs and myeloid progenitors (Figure S3) suggest that Dnmt1 regulates distinct patterns of gene expression in subsets of hematopoietic stem and progenitor cells. For example, carbonic anhydrase I (Car1), an early marker of erythroid differentiation (Villeval et al., 1985), was upregulated specifically in Dnmt1Δ/Δ ST-HSC/MPPs but not in LT-HSCs or myeloid progenitors (Figure S5A and S5B). Methylation of the Car1 proximal promoter was high (85-98%) in LT-HSCs and ST-HSC/MPPs, and Dnmt1 loss resulted in extensive demethylation (Figure S5C). In contrast, myeloid progenitors were observed to have low-level methylation (42%) and loss of Dnmt1 correlated with an increase in methylation (Figure S5C). These results suggest that: (1) demethylation of the Car1 locus in LT-HSCs is not sufficient to activate gene transcription, (2) another enzyme is able to methylate the Car1 promoter in the absence of Dnmt1 in myeloid progenitors, and (3) Dnmt1 maintains distinct patterns of methylation and gene expression in LT-HSCs, ST-HSC/MPPs and myeloid progenitors.
To further pursue the molecular mechanism underlying Dnmt1Δ/Δ phenotypic defects, we examined genes involved in the processes of HSC self-renewal and BM niche retention. Genes implicated in HSC mobilization and homing (N-cadherin, Casr, Gnas, Rac1, Rac2, Cxcr4, CD49D, Egr1) did not show significant changes upon Dnmt1 loss (data not shown). Genes associated with HSC self-renewal (including Tgm2, Gja1, Thy1, Cttn, Gata1, and Klf4) were found to be increased in expression in Dnmt1Δ/Δ LT-HSCs (Figure 3E), which does not account for reduced self-renewal of these cells. Two genes associated with HSC function, CD62L and Ski, were downregulated specifically in Dnmt1Δ/Δ LT-HSCs (Figure 3E). While this may partially account for our observed phenotype, this data set overall supports the existence of novel, uncharacterized regulators of HSC function to be explored from candidates in our microarray data set following loss of Dnmt1.
Additional, potential contributors to the Dnmt1Δ/Δ phenotype include defects in the ability of HSCs to give rise to multipotent and lineage restricted progenitor cells. Dnmt1Δ/Δ LT-HSCs and ST-HSC/MPPs were found to have decreased ability to generate myeloid colonies in the in vitro colony-forming unit assay (Figure 4A). Examining the myeloid progenitor compartment within the BM of Dnmt1Δ/Δ mice revealed decreased frequencies of common myeloid progenitors (CMP, LKS−IL-7Rα− FcγRloCD34+), granulocyte-macrophage progenitors (GMP, LKS−IL-7Rα−FcγRhiCD34+), and megakaryocyte-erythrocyte progenitors (MEP, LKS−IL-7Rα−FcγRloCD34−) at 4 days post pIpC (Figure 4B). These data demonstrate that the myeloid restricted progenitor compartment is acutely affected in Dnmt1Δ/Δ mice.
Loss of myeloid restricted progenitors may be due to altered in apoptosis, cell cycling and/or differentiation. Acute loss of Dnmt1 was not found to impact the frequencies of apoptotic (AnnexinV+7AAD−) or dead (AnnexinV+7AAD+) cells in total BM, CMPs, GMPs or MEPs (Figure S6A and S6B). While it remains a possibility that these cells are decreased in viability but cleared too rapidly from the system to observe changes, we found by gene ontology (GO) analysis that the molecular signature of apoptosis is under-represented in ST-HSC/MPPs and not altered in LT-HSCs or myeloid progenitors (Figure S6C), supporting that apoptosis does not account for the loss of Dnmt1Δ/Δ myeloid progenitors. To examine cell proliferation, we analyzed cell cycle status by BrdU incorporation (Figure 4C). Dnmt1Δ/Δ mice demonstrated significantly increased percentages of CMPs, GMPs, and MEPs in the S phase compared to controls (Figure 4D). This increase in cycling was further supported by higher transcript expression of Ccnd1 (Cyclin D1) in Dnmt1Δ/Δ myeloid restricted progenitors (LKS−) (Figure 4E). Taken together, these results suggest a concomitant increase in cycling but decrease in overall frequency of Dnmt1Δ/Δ myeloid progenitors that cannot be accounted for by increased apoptosis.
To gain insight into a potential functional mechanism that can explain why increased cycling does not lead to myeloid accumulation/expansion and neutrophilia in Dnmt1Δ/Δ mice, we revisited our GO analysis and observed that genes upregulated in Dnmt1Δ/Δ myeloid progenitors are significantly enriched in immunoglobulin and immune system-related GO terms (Figure 4F). Selected genes from within these GO terms, including the histocompatibility regions H2-D1 and H2-T22 as well as the tetraspanin CD81, were confirmed to be upregulated in Dnmt1Δ/Δ myeloid restricted progenitors (Figure 4G). This suggests that Dnmt1 is involved in maintaining myeloid progenitors in their primitive state through suppressing transcription of more differentiated lineage programs. Together, increased cycling and differentiation of Dnmt1Δ/Δ myeloid restricted progenitors, coupled with defects in differentiation of Dnmt1Δ/Δ LT-HSCs and ST-HSC/MPPs to replace these cells, offers a mechanism by which the myeloid progenitor pool is lost over time. This process may be exacerbated when in competition with WT cells, where we see a more acute loss of the mature myeloid lineages.
Here, we have examined effects of conditional deletion of Dnmt1 on the hematopoietic system. Mature hematopoietic lineages were unaffected by loss of Dnmt1, however, this enzyme was required for HSC self-renewal, niche retention and differentiation. In addition, Dnmt1 deletion resulted in increased cycling and differentiation of the myeloid restricted progenitor pool. DNA methylation and global gene expression analyses revealed that Dnmt1 regulates distinct patterns of methylation and gene expression in HSCs, multipotent progenitors and lineage-restricted progenitors.
Mature cells of the hematopoietic system were found to be stable following deletion of Dnmt1, however, we observed impaired hematopoiesis when Dnmt1 was deleted following transplantation and when WT cells were transplanted into Dnmt1 deficient recipients. There are at least two explanations that may account for this; (1) The timepoint at which we assessed steady state effects (12 weeks post pIpC) may not be long enough to observe modest self-renewal and differentiation defects of HSCs and progenitor cells in the absence of competing WT cells, and (2) the DNA methyltransferases Dnmt3a and 3b may be able to partially compensate for loss of Dnmt1 certain contexts. This is consistent with our observation of increased methylation of the proximal promoter of Car1 in Dnmt1Δ/Δ myeloid restricted progenitors.
The hematopoietic phenotype observed following Dnmt1 deletion is complex and appears to involve most of the primitive hematopoietic compartment including HSCs, multipotent and myeloid lineage restricted progenitor cells. Defects in HSC engraftment are accounted for by decreased niche retention, decreased self-renewal capacity and differentiation defects. The enhanced proliferative activity of Dnmt1 deficient myeloid progenitors coupled with differentiation indicates stress on the system to produce more mature myeloid cells. While all of these functional consequences of Dnmt1 deletion will certainly compound and contribute to one another in an in vivo setting, it was interesting to see the wide variety of gene expression changes and different methylation patterns following acute loss of Dnmt1 in LT-HSCs, ST-HSC/MPPs and myeloid progenitors. Within the ST-HSC/MPP population, in the case of both IAP and Car1, loss of methylation was associated with activation of gene expression. It is possible then that the ST-HSC/MPP population is most directly regulated by Dnmt1 or by loss of methylation maintained by Dnmt1, as opposed to the LT-HSC or myeloid progenitor populations. Comprehensive methylation profiling in the presence and absence of Dnmt1, Dnmt3a and Dnmt3b, coupled with global gene expression analysis, will provide a means to systematically classify gene loci that are regulated by the various Dnmts in hematopoietic stem and progenitor populations.
Certainly, our evidence suggests that epigenetic regulation, at least with respect to DNA methylation, of adult stem cells is distinct from embryonic stem cells and other somatic cell types. For example, there are clear distinctions between functional roles of de novo methylation and maintenance methylation. De novo methylation induced by Dnmt3a and Dnmt3b is necessary for LT-HSC self-renewal but not differentiation (Tadokoro et al., 2007). In contrast, we have observed that maintenance methylation induced by Dnmt1 is especially important for HSC and progenitor cell state transitions, such as the stepwise differentiation of HSCs to ST-HSC/MPPs, ST-HSC/MPPs to myeloid progenitors, mobilization, and regulating cell cycle entry. We suggest that these unique roles of Dnmt1 may be a common mechanism by which maintenance methylation regulates cell state transitions of adult somatic stem cells.
Dnmt1fl/fl mice generously provided by Dr. L. Jackson-Grusby (Children's Hospital Boston, Boston, MA, USA) (Jackson-Grusby et al., 2001) and Mx1-Cre transgenic mice (Kuhn et al., 1995) were maintained on a C57BL/6 background. pIpC (Sigma) was administered via IP injection as described at a dose of 25μg/kg (Hock et al., 2004). To confirm deletion of Dnmt1, genomic DNA was extracted from whole BM and PCR genotyped using described primers (Jackson-Grusby et al., 2001). The Children's Hospital Boston Animal Ethics Committee approved all experiments.
Whole cell lysates from control or Dnmt1Δ/Δ BM cells were separated by SDS-PAGE. Proteins were detected by a Dnmt1 mouse monoclonal antibody (Active Motif) and HRP-conjugated goat anti-mouse IgG1 (Novus Biologicals) or HRP-conjugated β-actin (Abcam).
Single cell suspensions of BM were prepared from pooled femurs, tibiae and iliac crests and PB was isolated via the retro-orbital plexus. Whole PB differential counts were determined using an AcT10 analyzer (Beckman Coulter). RBCs were lysed using ammonium chloride buffer prior to staining. Fluorochrome-conjugated antibody clones were obtained from eBioscience: Gr-1 (RB6-8C5), CD11b (M1/70), F4/80 (BM8), IgM (II/4I), CD43 (S7), B220 (RA3-6B2), CD71 (R17217), Ter-119, CD4 (GK1.5), CD8a (53-6.7). Flow cytometric analysis was performed on a FACSCalibur (Becton Dickinson, BD) and all data were analyzed using FlowJo (Tree Star, Inc.).
Congenic female B6.SJL (CD45.1+) mice (Taconic Farms) were irradiated with a split dose of 11 Gy and injected with 2×106 whole BM cells from Dnmtfl/fl Cre− or Cre+ mice (CD45.2+). The PB chimerism of recipient mice was assessed using fluorochrome-conjugated antibodies against CD45.1 (A20), CD45.2 (104) (eBioscience) and multilineage antibodies as described. For secondary transplants, 9×105 CD45.2+ donor cells from primary recipients were injected into lethally irradiated, secondary CD45.1+ recipients. For competitive transplantation, competitor BM cells were obtained from the first-generation cross of C57BL/6 and B6.SJL mice. Dnmtfl/fl Cre− or Cre+ BM cells were co-injected with competitor BM cells at a ratio of 1:1 (2×106 cells each). For secondary competitive transplantation, 9×105 FACS-isolated donor CD45.2+ cells were co-injected with 3×105 freshly isolated competitor CD45.1+CD45.2+ BM cells. For non-ablative transplant, 40×106 BM cells or 5000 purified Linc−Kit+Sca-1+CD34−/lo LT-HSCs from CD45.1+CD45.2+ mice were retro-orbitally injected into control or Dnmt1Δ/Δ mice. Donor cell engraftment was determined at 12 weeks post-transplant.
Lin−Sca-1+ cells were isolated from control or Dnmt1Δ/Δ BM, labeled with −5(and −6)-carboxyfluorescein diacetate succinimidyl ester (CFSE, Molecular Probes), washed and 2×105 cells were transplanted retro-orbitally into lethally irradiated mice. After 16 hours, BM cells were isolated from the femurs, tibiae and iliac crests of recipient mice, and CFSE+ cells detected by collection of at least 5×106 live events per sample on a FACSAria (BD).
Lineage depletion of BM, spleen or PB cells from control or Dnmt1Δ/Δ mice was performed as described by the manufacturer (Dynabeads Sheep anti-Rat IgG, Invitrogen) using affinity purified rat anti-mouse antibodies (eBioscience) against CD3ε, CD4, CD5, CD8a, B220, Gr-1, CD11b, Ter119. Lineage-depleted cells were stained with fluorochrome-conjugated antibodies (eBioscience) recognizing CD117 (c-Kit; 2B8), Sca-1 (E13-161.7), CD34 (RAM34), CD16/CD32 (FcγIII/II Receptor; 2.4G2), and CD127 (IL-7Rα; A7R34) and analyzed on a LSRII (BD). Colony-forming unit (CFU) assays were performed by plating cells into MethoCult™ GF M3434 (StemCell Technologies) and scoring colonies after 10 days. To assess cell cycling, mice were injected with 200μl of 10mg/mL 5-bromo-2-deoxyuridine (BrdU) by IP injection and sacrificed 12 hours later. Cell cycle status was determined by incorporation of BrdU and 7-AAD as described by the manufacturer (BrdU Flow Kit, BD Pharmingen).
Purified Lin−c-Kit+Sca-1+CD34−/lo LT-HSCs (2 replicates), Lin−c-Kit+Sca-1+CD34+ ST-HSC/MPPs (2 replicates), and Lin−c-Kit+Sca-1− myeloid progenitors (3 replicates) were isolated by FACSAria (BD) to >97% purity from control or Dnmt1Δ/Δ mice. Total RNA was extracted using the RNeasy Micro Kit (Qiagen), treated with DNaseI, reverse transcribed and amplified using the WT-Ovation Pico RNA Amplification System (NuGEN Technologies). Single-stranded cDNA amplification products were purified with QIAquick PCR Purification Kit (Qiagen) and labeled with the FL-Ovation cDNA Biotin Module V2 (NuGEN). Hybridization to GeneChip Mouse Genome 430A 2.0 Arrays (Affymetrix), washing and scanning were performed by the DFCI Microarray Core Facility (Boston, MA). CEL files were imported into dChIP software (Schadt et al., 2001) for data normalization and extraction of signal intensities. Independent biological repeats were combined by averaging the signal intensities of each probe represented on the microarray. Probes with an averaged signal intensity of >100 from at least one of the repeats were further analyzed (a total of 8723 probes).
GO analysis was performed using DAVID (Dennis et al., 2003). The top 10% upregulated and downregulated genes in the LT-HSC, ST-HSC/MPP, and myeloid progenitor gene sets were annotated for GO terms. GO terms with enrichment scores larger than two and p-values less than 1e−4 were considered significant.
RNA was isolated from cells with the RNeasy Micro Kit (Qiagen) and treated with DNaseI prior to reverse transcription. cDNA was prepared using SuperScript III (Invitrogen). Semi-quantitative real-time PCR was performed using iQ SYBR Green mix (Bio-Rad) on an iCycler (Bio-Rad) and Ct values were normalized to (β-actin levels.
Statistical analyses were performed with unpaired Student t tests.
We thank Drs. C.R. Walkley, L.E. Purton, J. Xu, D.A. Williams, G.C. Yuan, and M.D. Kiers for helpful discussion and critical comments; J. Chong for technical assistance; Children's Hospital Animal Facility Staff for care of experimental animals; J. Daley and S. Lazo-Kallanian of the DFCI HemNeo Flow Facility for cell sorting. J.J.T. and J.W.S. are Fellows of the Leukemia & Lymphoma Society, J.K. is a HHMI Research Associate, and S.H.O. is an Investigator of the HHMI.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Raw data from the Affymetrix microarray expression profiling were deposited in the Gene Expression Omnibus under accession number GSE15840.