The hematopoietic stem cell niche—home for friend and foe?


  • Daniela S. Krause,

    Corresponding author
    1. Department of Pathology, Massachusetts General Hospital, Boston, Massachusetts 02114
    2. Center for Regenerative Medicine and Cancer Center, Massachusetts General Hospital, Boston, Massachusetts 02114
    3. Harvard Stem Cell Institute, Department of Stem Cell and Regenerative Biology, Harvard University, Boston, Massachusetts 02114
    • 185 Cambridge Street; Boston, MA 02114, USA
    Search for more papers by this author
  • David T. Scadden,

    1. Center for Regenerative Medicine and Cancer Center, Massachusetts General Hospital, Boston, Massachusetts 02114
    2. Harvard Stem Cell Institute, Department of Stem Cell and Regenerative Biology, Harvard University, Boston, Massachusetts 02114
    Search for more papers by this author
  • Frederic I. Preffer

    1. Department of Pathology, Massachusetts General Hospital, Boston, Massachusetts 02114
    Search for more papers by this author

  • How to cite this article: Krause DS, Scadden DT, Preffer FI. The hematopoietic stem cell niche—home for friend and foe? Cytometry Part B 2013; 84B: 7–20.


The hematopoietic stem cell (HSC) niche is involved in the maintainance and regulation of quiescence, self-renewal and differentiation of hematopoietic stem cells and the fate of their progeny in mammals dealing with the daily stresses to the hematopoietic system. From the discovery that perturbations of the HSC niche can lead to hematopoietic disorders, we have now arrived at the prospect that the HSC niche may play a role in hematological malignancies and that this HSC niche may be a target for therapy. This review attempts to capture the discoveries of the last few years regarding the normal and malignant hematopoietic stem cell niche and possible ways to target this niche. © 2012 International Clinical Cytometry Society

A hematopoietic stem cell (HSC) niche was first conceptualized by Schofield in 1978 and was defined as “an entity in which the stem cell's maturation is prevented and the properties of ‘stemness’ are preserved” (1) [a list of abbreviations can be found in Table 3]. The HSC niche, which consists of osteolineage cells, sinusoidal endothelial cells, mesenchymal stromal and stem cells, sympathetic neurons, and the extracellular matrix, controls the number, quiescence, self-renewal, proliferation, differentiation, and localization of HSCs. It is the site in which 500 billion hematopoietic cells constituting our blood and hematopoietic organs are generated daily and where appropriate responses to stressors, injury, and repair processes are carefully orchestrated. During the embryonic phase of mammals, HSCs are first identified in the aorto-gonad-mesonephros region, from where they migrate to the fetal liver. After birth they migrate to the bone marrow, a cavity surrounded by rigid bone and filled with arterial vascular and sinusoidal systems, stroma and various HSC-supportive cell types (Table 1). Our understanding of the HSC niche has grown considerably over the last few years, at least partly due to the advancement in the technology of analytic flow cytometry, fluorescence-activated cell sorting (2), and in vivo imaging. While the nature of the HSC niche or niches and the question of the niche being endosteal and/or vascular has been somewhat controversial, this article aims to reconcile these differing theories. We aim at corroborating the hypothesis that HSCs, depending on their more active or more resting state, migrate between different niches.

Table 1. Characteristics of Normal Hematopoietic Stem and Progenitor Cells
Normal HSPCMolecular signatureCell-surface phenotype
  1. Adapted from Krause DS, Van Etten RA, 2007, Trends in Molecular Medicine;13(11): 470.

  2. LTR = long-term repopulating; ND = not done; CD = cluster of differentiation; HSC = hematopoietic stem cell; LKS = lineage- c-Kit+ Sca-1+; SLAM = signaling lymphocyte activation molecule; Hu = human; Mo = mouse; d = day; YS = yolk sac; AGM = aorto-gonado-mesonephros; WBM = whole bone marrow; ESC = embryonic stem cell; STR = short-term repopulating; MPP = multipotent progenitor cells; CMP = common myeloid progenitor; GMP = granulocyte-macrophage progenitor; MEP = megakaryocyte-erythroid progenitor

LTR-HSCNDLin- CD34+ Thy-1+; CD33+/−(82) (Hu)
Activated Notch (83); Activated Hedgehog (84); Activated Wnt (85); Bmi-1 (86); HoxB4 (87)Lin- c-Kit+ Sca-1+ Thy-1low (88) (Mo)
LKS SLAM Lin- c-Kit+ Sca-1+ CD150+ CD48− (Mo)
9.5d YS HSC c-Kit+ CD41+ CD150− CD48− CD45− CD34+ (89) (Mo)
10.5d AGM HSC c-Kit+ CD41+ CD150− CD48− CD45+/− CD34+ (89) (Mo)
12.5d placenta HSC c-Kit+ CD41+/− CD150− CD48− CD45+/− CD34+ (89) (Mo)
14.5d fetal liver c-Kit+/− CD41+/− CD150+/− CD48− CD45+/− CD34+/− (89) (Mo)
WBM HSC c-Kit+ CD41− CD150+ CD48− CD45+ CD34− (89) (Mo)
ESC-HSC c-Kit+ CD41+ CD150+ CD48+/− CD45+/−CD34− (89) (Mo)
 Lin- c-Kit+ Sca-1+ Thy-1lo Flk-2+ (88) (Mo)
 Lin- c-Kit+ Sca-1+ Thy-1lo Flk-2+ (88) (Mo)
CMP Lin- CD34+ CD38+ CD123lo CD45RA- FcγR- (90) (Hu)
 Lin- c-Kit+ Sca-1- CD34+ FcγRlo (91) (Mo)
GMP Lin- CD34+ CD38+ CD123lo CD45RA+ FcγR- (90) (Hu)
 Lin- c-Kit+ Sca-1- CD34+ FcγRhi (91) (Mo)
MEP Lin- CD34+ CD38+ CD123- CD45RA- FcγR- (90) (Hu)
 Lin- c-Kit+ Sca-1- CD34- FcγRlo (91) (Mo)
Table 2. Characteristics of Leukemic Stem Cells
LSCMolecular signatureCell-surface phenotype
  1. Adapted from Krause DS, Van Etten RA, 2007, Trends in Molecular Medicine;13(11): 470.

  2. CP = chronic phase; Mbc = myeloid blast crisis; B-ALL = B-cell acute lymphoblastic leukemia; T-ALL = T-cell acute lymphoblastic leukemia

CML CP LSCBCR-ABL1+CD34+ CD38− (92, 93) (Hu)
BCR-ABL1+Lin- Sca-1+ c-Kit+ (94, 95); CD44hi (53) (Mo)
CML LT-HSCBCR-ABL1+Lin- Sca-1+ c-Kit+ Flk-2− CD150+ CD48− (96) (Mo)
CML ST-HSCBCR-ABL1+Lin- Sca-1+ c-Kit+ Flk-2− CD150− CD48− (96) (Mo)
CML mBC LSCBCR-ABL1+; NUP98-HOXA9+ (97)Sca-1+ CD34+ c-Kitlo Flt3+ CD150− (97) (Mo)
AML LSCActivated NF-κB (98); Activated PI3K (99); Activated mTOR (100)CD34+ CD38− (101, 102); CD123+ (75); CD33+/− (82); CLL−1+/− (103) (Hu)
MLL-AF9+Lin- c-Kit+ CD34+ FcγR+ (104, 105); CD11b+ (105) (Mo)
CALM-AF10+B220+ CD11b- Gr-1− (106) (Mo)
Over-expressed HoxA9/Meis1; Bmi-1 (107)ND (Mo)
Pten−/−; activated mTOR (108)Lin- c-Kit+ Sca-1+ Flk-2− (108) (Mo)
Mutated NPM1 CD34+ CD38− CD123+ CD33+ CD90− or CD34− (109, 110) (Hu)
 Lin+ or CD38+ or CD45RA+ and Lin- CD38− (111) (Hu)
B-ALLETV6-RUNX1+ CD34+ CD38− CD19+ (112) (Hu)
T-ALL CD34+ CD4− or CD7− (113) (Hu)
Table 3. Abbreviations in Alphabetical Order
ALLAcute lymphoblastic leukemia
AMLAcute myeloid leukemia
BMBone marrow
CARCXCL12-abundant reticular cells
CD29Integrin beta-1, involved in cell adhesion and processes such as embryogenesis, tissue repair, immune responses, metastasis
CD31Platelet endothelial cell adhesion molecule (PECAM-1) expressed on platelets, monocytes, neutrophils, types of T cells, endothelial cells
CD34Cell-cell adhesion factor; expressed on early hematopoietic and vascular-associated tissue
CD44Cell-surface glycoprotein involved in cell–cell interactions, cell adhesion and migration
CD45Protein tyrosine phosphatase located on hematopoietic cells except erythrocytes and platelets; common leukocyte antigen

Belongs to subfamily of immunoglobulin-like receptors which includes SLAM (signaling lymphocyte activation molecules); see Table 1

CD51Integrin alpha-V, vitronectin receptor, expressed on activated T cells, polymorphonuclear granulocytes, platelets, blastocysts, and some osteolineage cells.
CD735′-nucleotidase (5′-NT) marker of lymphocyte differentiation
CD90= Thy1, glycophosphatidylinositol (GPI) anchored cell surface protein, present on stem cells and axons
CD105Endoglin, type I membrane glycoprotein, expressed on endothelial cells, activated macrophages, fibroblasts, smooth muscle cells
CD117c-Kit; Mast/stem cell growth factor receptor, tyrosine-protein kinase, identifies certain hematopoietic progenitors
CD146Melanoma cell adhesion molecule (MCAM) cell adhesion molecule used as a marker for endothelial cell lineage

Signaling lymphocytic activation molecule; see Table 1

Ter119Molecule associated with glycophorin A, marks the late stages of the murine erythroid lineage
CMLChronic myelogenous leukemia
HSCHematopoietic stem cell
HSPCHematopoietic stem and progenitor cell
LSCLeukemic stem cell
MDSMyelodysplastic syndrome
MPLMyeloproliferative leukemia virus oncogene (thrombopoietin receptor)
MPNMyeloproliferative neoplasm
MPSMyeloproliferative syndrome
MSCMesenchymal stem cell
NOD SCIDNon-obese diabetic severe combined immunodeficiency mouse
PTHParathyroid hormone
PTHrpParathyroid hormone-related protein
RARRetinoic acid receptor

Furthermore, with our increasing knowledge about the normal hematopoietic stem cell niche(s) it is compelling to contemplate the nature of the leukemic stem cell (LSC) niche, which may have parallels or may differ substantially from the normal HSC niche. Nonetheless, the LSC niche is considered to act as a sanctuary for malignant cells during therapy with cytostatic agents or tyrosine kinase inhibitors and possibly during the graft-versus-leukemia-effect after allogeneic hematopoietic stem cell transplantation and, thus, possibly represents the origin of disease progression and relapse. Flow cytometry has added substantially to the diagnosis of leukemias and lymphomas and our knowledge of certain types of cancer stem cells in clinical medicine and basic research over the last 20 years. However, the importance of flow cytometry for understanding chemotherapy-resistant, quiescent cancer stem cells, which likely reside in a tumor niche, has been minimal due to the limitations of detecting these rare cells outside of their niche. We also aim at increasing the impact flow cytometry may have in the detection and analysis of cancer stem cells.

Different Cell Types Constitute Specialized Hematopoietic Niches

The ability to visualize the HSC niche by confocal 2-photon microscopy in real time at a single cell level has confirmed that transplanted hematopoietic stem and progenitor cells (HSPCs) home to bone marrow vascular domains, in which the expression of the chemokine (C-X-C motif) ligand 12 (CXCL12) is highest (3) [a list of technical and functional terms can be found in Table 4]. Further studies revealed that osteoblasts are in close proximity to vascular networks and that positioning of transplanted HSPCs is nonrandom in irradiated mice (Fig. 1). Localization of HSPCs is dependent on the maturation stage of the transplanted cells with more mature cells homing to locations further away from the endosteum and bone than more immature HSPCs. The bone marrow endosteum is thought to maintain HSCs under normal circumstances and to foster their expansion in response to bone marrow damaging agents (4, 5).

Table 4. Technical and Functional Terms
Alpha-V integrinAn integrin mediating cell adhesion and migration
Ang-1Angiopoietin-1; vascular development and angiogenesis, mediation of interactions between the endothelium and surrounding matrix and mesenchyme
BMPBone morphogenetic protein; cytokine which induces the formation of bone and cartilage, also provides morphogenetic signals influencing the architecture of tissues
CXCL12Chemokine (C-X-C motif) ligand 12 (= SDF-1α); chemotactic for lymphocytes
CXCR4C-X-C chemokine receptor type 4, receptor specific for stromal-derived-factor-1 (SDF-1 or CXCL12)
Dicer 1Endoribonuclease that cleaves double-stranded RNA (dsRNA) and pre-microRNA (miRNA) into short double-stranded RNA fragments called small interfering RNA (siRNA)
Dkk1Dickkopf 1; inhibitor of the WNT signaling pathway, involved in embryonic development
E-selectinCell adhesion molecule expressed on endothelial cells activated by cytokines
Gal-3Galectin-3; member of the lectin family, involved in cell adhesion, cell activation, chemoattraction, cell growth, differentiation, cell cycle, apoptosis
GCSFGranulocyte colony stimulating factor; cytokine which stimulates the bone marrow to produce granulocytes and stem cells and to release them into the blood
GsalphaSubunit of the heterotrimeric G protein which activates the cAMP-dependent pathway by activating adenylate cyclase
Hif1-αHypoxia-inducible factor 1α; the protein encoded by HIF1 is found in mammalian cells growing at low oxygen concentrations, plays a role in cellular and systemic responses to hypoxia
ImatinibTyrosine kinase inhibitor used for treatment of CML and other diseases
JaggedLigand for the receptor notch 1
Jak2Janus kinase 2; tyrosine kinase implicated in signaling from type II cytokine receptors, the GM-CSF receptor family, the gp130 receptor family and the single chain receptors
Jak2 V617FMutation in the Jak2 tyrosine kinase associated with some myeloproliferative neoplasms
Mind bomb 1E3 ubiquitin-protein ligase; involved in regulating apoptosis
MLL-AF9Oncogene associated with different types of acute leukemia
Mx1Interferon-induced GTP-binding protein
NestinType VI intermediate filament (IF) protein, found mostly in nerve cells
NotchSingle-pass transmembrane receptor protein
OsterixTranscription factor essential for osteoblast differentiation and bone formation
PI-3KPhosphatidylinositol-3-kinase; family of enzymes involved in cell growth, proliferation, differentiation, motility, survival and intracellular trafficking
PlerixaforCXCR4-antagonist, used to mobilize hematopoietic stem cells
PSGL-1P-selectin glycoprotein ligand; glycoprotein found on white blood cells that binds to P-selectin
RANKLReceptor activator of nuclear factor kappa B ligand; ligand for osteoprotegerin and key factor for osteoclast differentiation and activation.
RbRetinoblastoma protein; tumor suppressor protein involved in several cancers
Sca-1Stem cell antigen 1; expressed on hematopoietic stem cells
SCFStem cell factor (kit ligand); binds to the c-Kit receptor (CD117)
SDF-1αStromal derived factor 1α (= CXCL12); see CXCL12
Stat3Signal transducer and activator of transcription 3; transcriptional activator which is phosphorylated by receptor-associated kinases in response to cytokines and growth factors

(= CD90); see Table 3

Tie-2Cell-surface receptor that binds and is activated by the angiopoietins
TNFαCytokine involved in systemic inflammation, produced by activated macrophages and occasionally lymphocytes
VCAM-1Vascular cell adhesion molecule 1; expressed on large and small vessels after stimulation of the endothelial cells by cytokines
VEGF(R)Vascular endothelial growth factor; involved in angiogenesis
VLA-4Very late antigen 4; integrin expressed on leukocytes
Wnt pathwaySignaling pathway which controls proliferation and differentiation in the embryo and adult
Wt1Wilms tumor protein; role in the normal development of the urogenital system, but also expressed on a variety of hematological malignancies
Figure 1.

Schematic depicting the microanatomy of the normal hematopoietic stem cell niche. The normal bone marrow microenvironment consists of a variety of cell types including osteoblasts, osteoclasts, mesenchymal stem cells, adipocytes, endothelial cells, perivascular reticular cells, sympathetic neurons, macrophages, and other hematopoietic cells. The location, proliferation, differentiation, and quiescence of hematopoietic stem cells are regulated by cytokines secreted by constituents of the bone marrow microenvironment, the extracellular matrix, the oxygen tension, and the nervous system. Not all interactions could be included. ECM = extracellular matrix; CXCL12 (SDF1alpha) = stromal derived factor alpha; Scf = stem cell factor; Ang-1 = angiopoietin-1; IL-7 = interleukin-7; OPN = osteopontin; HSC = hematopoietic stem cell; TPO = thrombopoietin; MSC = mesenchymal stem cell; OB = osteoblast; OC = osteoclast; Retic = reticular cell; Adip = adipocyte; EC = endothelial cell; Adventitital retic. Cell = adventitial reticular cell.

The BM stroma consists of adipocytes, reticular cells, macrophages, vascular endothelial cells, smooth muscle cells, and mesenchymal stem cells and is responsible for the production and deposition of the extracellular matrix, the production and concentration of cytokines, and growth factors. The stroma interacts with HSPCs via cell surface receptors, secreted growth factors, and adhesive ligands.

Specialized niches have been defined in the bone marrow, such as those for the different stages of B-cell development or for megakaryocytic platelet shedding. In the former case, pre-pro-B cells are found in close proximity to HSPCs and CXCL12-expressing stromal cells, whereas more mature pro-B cells are located closer to interleukin- (IL-) 7-expressing cells. Most mature plasma cells reside close to CXCL12-expressing cells (6). In the latter case, proplatelets were shown to be shed from megakaryocytic protrusions into microvessels at defined locations in response to blood flow-induced shear stress (7).

Osteolineage Cells Modulate HSC Number and Function

The knowledge that osteoblastic cells are constituents of the normal HSC niche dates back to 1996, when Taichman et al. described that human osteoblasts can expand hematopoietic long-term culture-initiating cells and progenitors three- to fourfold in ex vivo culture systems (8). Two papers published in 2003 showed that osteoblastic cells regulate the number of hematopoietic stem cells in vivo. In one paper, mice with osteoblastic-cell specific constitutively activated receptor for parathyroid hormone (PTH) and PTH-related peptide (PTHrp) (col1-caPPR mice) were found to have increased numbers of HSPCs, which exhibited superior engraftment compared to HSPCs from wild-type (wt) controls (9). In subsequent studies, application of PTH, an important regulator of bone, to wt mice also resulted in increased numbers of HSPCs suggesting that osteoblasts regulate HSPC number (10). The second paper demonstrated that mice with conditional inactivation of BMP receptor IA (Mx1-Cre; Bmpr1a) had increased numbers of osteoblasts and HSPCs, arguing that osteoblasts also control the size of the niche. Acute deletion of osteoblastic cells in a ganciclovir-based inducible murine system led to progressive bone loss; decreased bone marrow cellularity and loss of erythroid, lymphoid, and myeloid progenitors and HSPCs. Withdrawal of ganciclovir, however, led to reappearance of osteoblasts and medullary hematopoiesis confirming the interplay between mesenchymal and hematopoietic elements in the bone marrow (11). Sorted CD45- Tie-2- CD105+ alphaV+ Thy1.1- Thy1- cells from fetal bones 15.5 days post-conception transplanted under the renal capsule in mice could recruit host-derived blood vessels, produce ectopic bones via a cartilaginous stage and form a marrow cavity populated by HSPCs. In contrast, CD45- Tie-2- CD105+ alphaV+ Thy1.1+ cells form bone without a cavity. Suppression of factors inducing endochondral ossification such as osterix and vascular endothelial growth factor inhibited niche generation. This argued for the importance of endochondral ossification via a cartilaginous stage for formation of the HSC niche (12). In a mixed multicellular spheroid in vitro model osteoblastic cells, which are CD45- CD105+ CD31- Ter119- Sca-1- CD51+, and CD105+ osteoprogenitor cells acted as constituents of a 3-dimensional HSC niche producing factors like CXCL12 (stromal-derived factor 1 alpha; SDF-1α) and osteopontin (OPN) to retain HSCs in their niche (13). OPN, an extracellular structural protein secreted by osteoblasts and other cell types, is also known to negatively regulate HSPC pool size (14). In addition, osteoblastic cells secrete angiopoietin-1 (Ang-1), membrane-bound stem cell factor (SCF) and thrombopoietin maintaining HSCs in a quiescent state via MPL (myeloproliferative leukemia virus oncogene), the thrombopoietin receptor (15) (Fig. 1).

Many other factors expressed on or derived from the osteoid lineage have been described to be important for HSPC maintenance, but the role of N-cadherin-expression in osteolineage cells for retaining HSCs, for example, has been controversial (16, 17). Mice deficient for biglycan, a leucine-rich repeat proteoglycan found in bone, cartilage, and tendon, had significantly reduced trabecular bone and osteoblasts, but numbers or function of HSPCs remained unchanged (16). Similarly, the bone anabolic agent strontium increased osteoblasts, bone volume, and trabecular thickness, but did not increase HSPCs (18), suggesting that only certain subtypes of osteoblastic cells may be involved in HSPC maintenance and regulation.

There has been evidence that osteoclasts also play a role in hematopoiesis. Specifically, stimulation of osteoclasts by receptor activator of nuclear factor kappa-B ligand (RANKL) an important regulator of osteoclast function, proliferation, and survival, led to reduction of SDF-1α and SCF, both of which are also cleaved by the bone-resorbing proteinase cathepsin K, and osteopontin. This leads to subsequent mobilization of HSPCs (19). Furthermore, mice treated with bisphosphonates, known to promote osteoclast apoptosis, have decreased numbers of HSPCs, which are inferior to control HSPCs in competitive transplantation assays (20).

Recently, a role in the regulation of hematopoiesis was also ascribed to osteocytes, which are the most abundant cell type in bone and are embedded in the bone matrix. Mice with osteocyte-specific deletion of the Gsα-subunit (OCY-GsαKO mice) were found to have leukocytosis, neutrophilia, and splenomegaly. Transplantation studies revealed that this effect was the consequence of an altered bone marrow microenvironment and not intrinsic to the hematopoietic cells. In addition, in vitro co-culture of wt bone marrow cells with conditioned media from osteocyte-enriched bone explants from OCY-GsαKO mice were found to have increased levels of GCSF and significantly increased myeloid colony formation (21). Another study demonstrated that GCSF-induced mobilization of HSPCs was significantly impaired in transgenic mice with osteocyte-specific expression of the receptor for diphtheria toxin (22). The importance of osteolineage cells for HSPCs also manifests itself indirectly, as the concentration of calcium ions in the endosteal area is sensed by the calcium receptor expressed on HSPCs (Fig. 1). HSCs from the fetal liver of mice deficient for the calcium receptor, for example, are normal in number, their ability to proliferate and differentiate, yet HSCs are defective at homing to the endosteal area presumably based on sensing of a calcium gradient. This demonstrates a prominent role for the calcium receptor to retain HSCs in their niche (23).

Mesenchymal Stem Cells Secrete Factors Involved in HSC Maintenance

Mesenchymal stem cells (MSC) marking positive for nestin, CD73, CD105, CD44, CD29, and CD90 are recently identified components of the HSPC niche. They are strictly perivascular and are localized mostly in central areas of the BM, but are also found in the vicinity of the endosteum. They are closely associated with HSCs and adrenergic nerves and highly express genes like CXCL12, SCF, Ang-1, IL-7, and OPN, which are involved in HSC maintenance (Fig. 1). Depletion of nestin+ MSC leads to reduced retention of HSPCs and to reduced homing of transplanted hematopoietic progenitor cells (24).

The HSC Niche is also Regulated by the Nervous System, the Extracellular Matrix, Cytokines, and Adipocytes

Trafficking of HSCs between the niche and the bloodstream is regulated by the sympathetic nervous system with circadian rhythms controlling egress of HSCs and their progenitors from the bone marrow niche. Release of CXCL12 by cells in the BM microenvironment is coordinated by circadian secretion of noradrenaline acting on β-adrenergic receptors on stromal cells and thereby leading to HSC release from the BM during periods of rest in mammals (Supporting Information, Animation 1) (25). In addition, it was recently shown that non-myelinating Schwann cells, which ensheathe autonomic nerves, are in close contact with HSCs in the BM. They also regulate the HSC pool by expressing molecules that activate latent transforming growth factor β, a negative regulator of HSCs (Fig. 1) (26).

The extracellular matrix of the bone marrow consists of fibronectin, hyaluronan, collagen types I and IV, laminin, cytokine-binding glycosaminoglycans, heparin sulfate, and chondroitin sulfate and represents a reservoir for growth factors, cytokines, and metalloproteinases (Fig. 1). It provides gradients of soluble factors that influence cell growth, differentiation, motility, and viability of hematopoietic cells by interactions via integrins, CD44, and other adhesion molecules.

For example, osteopontin is secreted by preosteoblasts, osteoblasts, osteocytes, and other cell types and is known to be a negative regulator of HSPCs, as an increase in the stromal Notch-ligand Jagged-1, angiopoietin-1, and in HSPCs were demonstrated in an OPN-null microenvironment (14). Furthermore, conditional deletion of GRP94, an endoplasmatic reticulum chaperone essential for the expression of certain integrins, in the hematopoietic system led to significant reduction of integrin α4 on HSPCs and an increase of HSCs. It also enhanced HSPC mobilization, impaired homing and impaired binding to fibronectin. Myeloid and lymphoid differentiation was also altered in these mice (27). HSC-regulation by cytokines became evident by the analysis of mice deficient for thrombopoietin, which have a 150-fold decreased number of HSCs. In addition, in the post-transplant state HSC expansion was highly dependent on thrombopoietin and its receptor MPL (28).

By the use of flow cytometry, competitive repopulation assays and colony forming activity assays it was recently demonstrated in genetic and pharmacological mouse models of adipocyte deletion that HSC engraftment after irradiation is accelerated in mice lacking adipocytes. Bone marrow adipocytes were found to act as predominantly negative regulators of hematopoiesis in this study (29).

Numerous other regulators of the HSC niche, including the Notch- and Wnt/β-catenin (30) pathways, have been identified (reviewed in (31, 32), respectively) but are not the focus of this review.

Macrophages Play a Role in Maintaining and Retaining HSC

Mononuclear phagocytes positive for F4/80, the mouse homologue to the human epidermal growth factor-like module-containing mucin-like hormone receptor-like 1 (Emr1) protein, which is expressed on macrophages, form a roof-like structure around endosteal osteoblasts (Fig. 1). They have been identified as important components of the niche promoting the maintenance and retention of HSCs. In recent papers, it was demonstrated that administration of GCSF to mice led to reduction of osteoblasts and of endosteal macrophages (osteomacs), which support osteoblast function. In vivo modeling of macrophage loss by administration of clodronate-loaded liposomes or the use of macrophage Fas-induced apoptosis (Mafia) transgenic mice recapitulated the loss of endosteal osteoblasts. It also led to reduction of HSC-maintaining cytokines at the endosteum followed by HSC mobilization into the blood (33). Further work demonstrated that reduction of macrophages in the bone marrow by conditional deletion led to reduced levels of HSC-retention factors like CXCL12, SCF, Angiopoietin-1 (Ang-1), and VCAM-1 in nestin+ cells. The increased egress of HSCs and progenitors mentioned earlier was also confirmed, identifying macrophages as positive regulators of the nestin+ MSC niche, although mediating factors are still unknown (34).

Perivascular Cells in the Niche Secrete Stem Cell Factor, which Supports HSC

Bone marrow sinusoids are thin-walled blood vessels with a diameter of > 50 μm, into which arterioles open directly. They are lined by endothelial cells and are covered by adventitial reticular cells on ∼60% of their abluminal surface. CXCL12-expressing CD146+ MSCs with self-renewal capacity are located on sinusoidal surfaces, contribute to the sinusoidal wall structure, and produce Ang-1. These nestin+ perisinusoidal cells, called CXCL12-abundant reticular (CAR) cells (Fig. 1), coincide with osteoprogenitors/ mesenchymal skeletal stem cells and are involved in regulation of HSCs (24).

There has been much debate about the relationship between the vascular and the osteoblastic niche within the bone marrow microenvironment, which arose due to experimental evidence that transplanted HSCs home to vascular locations in non-irradiated hosts and that HSCs also reside at a distance from the endosteum. 60% of lineage- Sca-1+ c-Kit+ CD150+ CD48- (LKS SLAM) HSCs were actually localized close to sinusoidal endothelial cells (35). In addition, mice deficient for the protein Bis that is related to apoptosis and cellular stress responses exhibited loss of HSCs and defective B-cell development. However, transplantation experiments revealed that these alterations were due to microenvironmental perturbations such as a defective sinusoidal endothelium and loss of CXCL12- and IL-7-expressing stromal cells, while having an intact osteoblastic niche (36). Further evidence for a vascular niche was provided, when increased CXCL12- and vascular endothelial growth factor (VEGFR)-expression after myelotoxic injury was found in perivascular structures, and HSC stressors like tumor necrosis factor α (TNF α) and lipopolysaccharide (LPS) were shown to upregulate Jagged2 on BM endothelial cells (37). It was also demonstrated that vascular endothelial cells promote hematologic recovery and improved survival in mice following total body irradiation (38). These regenerative cues prompting HSCs to enter the cell cycle were provided by type I or II interferon (39), GCSF, or other injury signals.

A recent, elegant study showed that stem cell factor (SCF; kit ligand) which is an essential niche component for HSC maintenance, is primarily expressed by perivascular cells. The authors systematically deleted Scf from hematopoietic cells, osteoblasts, and nestin-cre or nestin-cre-estrogen receptor (ER) expressing cells, but could not find a defect in HSC frequency or function. However, when Scf was deleted from endothelial cells or leptin-expressing perivascular stromal cells, HSCs were depleted from the bone marrow, arguing for a prominent role in the vascular niche. Nevertheless, conditional deletion of Scf was only performed in some cell types and non-conditional deletion in others. Therefore, it is not clear whether the effect of Scf loss was due to a developmental effect while the HSC was being formed or due to an effect on the maintenance of the HSC in the adult (40).

An example of how gene disruption can affect hematopoietic stem cells via alteration of the osteoblastic and vascular niches was provided by the analysis of mice deficient for the tumor suppressor and mediator of cell contact inhibition Nf2/merlin. These mice have an increased number of HSCs with quantitatively significant location in the circulation. These changes were entirely dependent on the bone marrow microenvironment, which was characterized by an increase of trabecular bone, bone marrow vascularity and expression of VEGF (41).

The Vascular and Osteoblastic Niches Cooperate

Several pieces of evidence argue that the vascular and osteoblastic niches are not as distinct as originally perceived. There are niche cells around vessels that are not found on the surface of bone. Whether these collectively are a niche or represent specific sites with a specific function is unclear. For example, it has been proposed that there may be a niche for dormant HSC, and a niche for self-renewing or cycling HSC. This hypothesis is appealing as it represents a way of organizing functions with niche localization, but this is presently still hypothetical at this time. Differing oxygen levels may also contribute variation to the contour of HSC cycling. For example, hypoxia-inducible factor α (Hif1-α), a transcription factor responding to oxygen levels in the environment, controls expression of HSC-regulating genes like CXCL12 and its receptor CXCR4. But the precise distribution of oxygen or other nutrients in the bone marrow is still unclear.

Hematopoietic Cells Modulate their Niche

Although less explored, there is evidence that the interactions between the niche and hematopoietic cells are reciprocal. 48 h after stressing murine bone marrow by an acute bleed, it was revealed that HSCs secrete bone morphogenetic proteins (BMP) 2 and 6 directing the fate of MSCs toward the osteoblastic lineage in vitro and in vivo. However, this response was mitigated in aging and osteoporotic animals (42). Megakaryocytes, localized close to the endosteum, in addition, have been shown to stimulate osteoblasts via increased levels of BMP-2, −4, and −6 (Fig. 1) (43, 44). This argues that hematopoietic elements are involved in bone formation and activities within the niche.

Hematopoietic Abnormalities can Arise due to a Defective Bone Marrow Microenvironment

Several publications over the last few years have provided evidence that a defective bone marrow microenvironment can lead to hematopoietic abnormalities. Deficiency of the nuclear receptor retinoic acid receptor (RAR) γ led to a myeloproliferative syndrome (MPS) with an increase of granulocytic/macrophagic progenitors and increased granulocytes in peripheral blood, bone marrow, and spleen, which was entirely dependent on the bone marrow microenvironment. This effect was mediated by increased levels of tumor necrosis factor alpha (TNFα) in the altered bone marrow microenvironment (45). Deletion of retinoblastoma protein (Rb) in the hematopoietic system led to an MPS and loss of HSCs from the bone marrow niche due to mobilization. However, these abnormalities were not intrinsic to the hematopoietic cells, but rather dependent on the role of Rb in the interplay between myeloid cells and their bone marrow microenvironment (46). Inactivation of a component of Notch ligand-mediated endocytosis, mind bomb 1 (Mib1) also led to a fatal MPS in mice with death being due to infiltration of organs by myeloid cells. The MPS was microenvironment-dependent, as transplantation of wild-type hematopoietic cells into a Mib1-deficient microenvironment resulted in myeloproliferation. In addition, reintroduction of the constitutively active intracellular domain of Notch1 into Mib1-null mice led to reduction of disease (47). As mentioned earlier, osteocyte-specific deletion of the Gsα subunit of the G-protein signaling cascade, which lies downstream of the receptors for parathyroid hormone, prostaglandin, and some β-adrenergic receptors, resulted in myeloproliferation and increased egress of myeloid progenitor cells from the bone marrow. This phenotype was entirely due to the altered bone marrow microenvironment and increased secretion of GCSF from mutated osteocytes (21). Furthermore, a myelodysplasia-like syndrome with rare cases of acute myeloid leukemia were observed in mice with osteoprogenitor cell-specific disruption of Dicer1, a protein required in the RNA interference and microRNA pathways (48).

It is a daunting thought that the genetic aberrations found in the bone marrow stroma cells of patients with myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML) may be causative of the hematological disease (49). It is also fathomable that certain genetic syndromes with increased risk of development of MDS/AML may possibly be partly due to the skeletal abnormalities in disease syndromes such as Diamond-Blackfan-, Shwachman-Diamond Syndrome, or Fanconi Anemia. In addition, donor-derived hematopoietic disease has been known to occur in recipients of allogeneic hematopoietic stem cell transplants (50). This suggests that hematological disease may have occurred due to perturbations of the bone marrow microenvironment in patients heavily pretreated with chemo- and radiation therapy.

Leukemia-Initiating Cells Interact with Their Niche

About 70-80% of patients with AML achieve a complete remission after chemotherapy, but most patients relapse. Despite progress in treatment the 5-year survival rate in AML is only 50% and the 4-year disease-free survival rate is 44%. Although the treatment for chronic myelogenous leukemia has been revolutionized and augmented by the tyrosine kinase inhibitor imatinib and several second line tyrosine kinase inhibitors, achieving hematologic remission in > 95% of patients, cures involving molecular remissions are rare. In addition, the disease frequently relapses, when imatinib is discontinued. Continuation and intensification of the search for improved therapies, including those targeting the bone marrow niche, considered a sanctuary for dormant leukemic stem cells (LSC) from where disease progression and resistance to therapy are thought to evolve, is warranted. An LSC is defined as a rare cell type that has acquired the ability to self-renew and to be resistant to most chemotherapeutic agents, while the bulk of the highly proliferative leukemic cells are eradicated (Table 2).

By in vivo confocal imaging of the microvasculature in the calvarium of live mice specialized, discontinuous stretches of endothelium were found with high expression of E-selectin or SDF-1α, which uniquely influenced the homing of injected tumor cell lines (Fig. 2). Disruption of the interactions with SDF-1, for example, inhibited the homing and eventual engraftment of the acute lymphoblastic leukemia cell line Nalm-6 to these vessels. This suggested for the first time that molecularly distinct vasculature may be responsible for differential engraftment of tumor cells in the bone marrow microenvironment (3). In a follow-up study it was demonstrated that leukemia cell growth in the bone marrow microenvironment disrupts normal HSPCs in their niche and reduces HSPC number and mobilization via the secretion of stem cell factor by leukemic cells leading to a “hijacking” of the niche (Supporting Information, Animation 2) (51).

Figure 2.

Schematic depicting the microanatomy of the leukemic stem cell niche and strategies to target the LSC in its niche. Strategies include the blocking of cytokines or adhesion molecules like CD44 or integrins, the mobilization of LSCs out of their niche by GCSF or AMD3100, the blocking of proteins secreted by the bone marrow stroma like galectin, which promotes drug resistance, or the increase of bone remodeling by PTH in CML. For simplicity's sake the authors have integrated interactions of lymphoid and myeloid leukemic cells with their respective niches into one figure. Details can be found in the text. CXCL12 (SDF1alpha) = stromal derived factor alpha; IL-7 = interleukin-7; HSC = hematopoietic stem cell; OB = osteoblast; VLA-4 = very late antigen 4; GCSF = granulocyte colony stimulating factor; PTH = parathyroid hormone; CXCR4 = C-X-C chemokine receptor type 4.

Mechanistic description of the interaction of leukemic stem cells with their niche was provided by two papers which depicted the type I transmembrane glycoprotein and adhesion molecule CD44 as a mediator between leukemia-initiating cells and the bone marrow microenvironment (Fig. 2) (52, 53). CD44 is widely expressed on hematopoietic and non-hematopoietic tissues, binds to hyaluronan (54), osteopontin, and E-selectin (55, 56) and is involved in lymphocyte homing, embryonic development, and tumor cell metastasis. In the first paper, an activating antibody of CD44 significantly reduced the repopulation of NOD SCID mice with human AML cells and reduced the frequency of LSCs in these animals as tested in secondary transplantation assays. These effects were found to be due to decreased migration to LSC-supportive niches within the bone marrow microenvironment and to direct alteration of LSC fate (52). In the second paper, the authors showed that CD44 is required for homing and engraftment of chronic myelogenous leukemia (CML) -initiating cells. Wild-type recipient mice receiving CD44 deficient syngeneic bone marrow transduced with the oncogene which induces CML, BCR-ABL1, had significantly prolonged survival and the efficiency of CML-induction was impaired. However, intrafemoral injection of the BCR-ABL1+ CD44-deficient donor graft or coexpression of CD44 bypassed the requirement for CD44 and restored CML. Compared to normal HSCs (53), a blocking antibody to CD44 impaired CML induction in recipient animals suggesting a prominent role of CD44 as a provider of selectin ligands for the homing and engraftment of BCR-ABL1+ LSCs. In a similar study, the same authors showed that E-selectin deficiency on bone marrow endothelium impaired the engraftment of BCR-ABL1+ CML-initiating cells, but this could be overcome by intrafemoral injections. BCR-ABL1+ CML-initiating cells deficient for P-selectin glycoprotein ligand-1 (PSGL-1) or the selectin ligand-synthesizing enzymes core-2 β1,6-N-acetylglucosaminyltransferase or fucosyltransferases IV/VII were impaired for engraftment, while destruction of selectin ligands by neuraminidase completely blocked leukemic engraftment. These results establish that BCR-ABL1+ leukemic stem cells depend to a large extent on selectins and their ligands, including CD44, for homing and engraftment (57). Another study showed that a neutralizing antibody to interleukin-3 (IL-3) receptor alpha chain (CD123) (clone 7G3) impaired the homing and engraftment of AML-LSCs in non-obese diabetic/severe-combined immunodeficient (NOD/SCID) mice and prolonged survival (58).

The Hematopoietic Niche is Protective for Leukemia-Initiating Cells

Perturbations of the bone marrow microenvironment due to hematological disease such as plasma cell myeloma are well-described entities and are not the focus of this article. Alterations of the bone marrow microenvironment have also been reported in patients with primary myelofibrosis and recently in AML. In an immunocompetent murine model of AML serum levels of osteocalcin, a marker of bone formation, were decreased signifying inhibition of osteoblastic cells via the secretion of the chemokine CCL3 by leukemic cells. Furthermore, osteoprogenitor cells, as well as endosteum-lining osteopontin+ cells were reduced leading to decreased mineralization of bone. This suggested that pancytopenia associated with some cases of AML may be due to inhibition of osteoblastic cells (59).

However, many publications of the last few years have focused on the protective effects of bone marrow microenvironmental components on the survival of leukemic cells. The bone marrow microenvironment is thought to provide a sanctuary for leukemic cells during chemotherapy, therapy with tyrosine kinase inhibitors or during the graft-versus-leukemia-effect after allogeneic hematopoietic stem cell transplantation. The bone marrow microenvironment has been considered to contain minimal residual disease and to provide survival signals for leukemic cells. This is supported by the finding that levels of Wilm's tumor (WT1) gene transcripts, a well-recognized marker for minimal residual disease in leukemia, are higher in the BM than in peripheral blood of patients with AML and acute lymphoblastic leukemia (ALL) (60). By using xenotransplantation assays of human AML cells into NOD/SCID interleukin (IL)-2 receptor gamma null mice the authors showed that AML cells home to the osteoblast-rich areas of the bone marrow, where they become quiescent and are protected from chemotherapy-induced apoptosis (61). In a follow-up study, the authors describe that forcing AML LSCs to enter cell cycle by the administration of GCSF renders them more sensitive to apoptosis-induction by cytostatic agents and results in elimination of LSCs and prolongation of survival of secondary recipients (Fig. 2) (62). Similarly, complete remission rates were higher (11%) in AML patients receiving chemotherapy and GCSF in a combined treatment regimen called “priming” compared to patients receiving chemotherapy alone. This could be due to differentiation of some AML cells or enhanced neutrophil function, but may also be associated with abrogation of protective stromal effects on chemotherapy-induced apoptosis, if AML cells are mobilized into the peripheral blood (63).

Several leukemic cell-specific factors contribute to their attraction to the niche. Overexpression of β-integrins, mostly very late antigen (VLA)-4, on leukemic blasts enhances the adhesion to fibronectin in the bone marrow microenvironment. Thereby it contributes to resistance to therapy and persistence of minimal residual disease and correlates with sensitivity to chemotherapy via the phosphatidylinositol-3-kinase (PI-3K)/AKT/Bcl-2 signaling pathway. A 100% survival rate in mice with leukemic minimal residual disease was achieved by combining cytosine arabinoside with a blocking antibody to VLA-4 arguing for the importance of adhesion via the integrin axis for survival of leukemia cells (64). Similarly, blockade of VLA-4 by natalizumab reduced adherence of lymphomatous B cells to stromal fibronectin and overcame resistance to the anti-B cell antibody rituximab (65).

Furthermore, CXCR4, the receptor for CXCL12 (= SDF-1α) may be overexpressed on leukemic blasts or CML cells and may portend a poor clinical prognosis by increasing the adhesion of leukemic cells to stroma, which may provide antiapoptotic signals. In the case of CML this mechanism has been shown to provide protection to cell death induced by the tyrosine kinase inhibitor imatinib via reduction of caspase 3-activation and modulation of the anti-apoptotic protein Bcl-XL (66). Blocking the interaction of CXCR4 on CML cells with stroma by the CXCR4-antagonist plerixafor (Fig. 2) leads to reduced adhesion in vitro and to a more significant reduction of leukemia burden in vivo in combination with the second line tyrosine kinase inhibitor nilotinib than nilotinib alone (67).

The bone marrow microenvironment, and in particular the stroma, also contributes to survival of normal and malignant HSPCs by the secretion of soluble factors in the extracellular matrix, although the stroma itself may be abnormal in particular hematological malignancies like CML (68). The type and concentration of secreted factors, however, may differ between certain subdomains of the bone marrow microenvironment. In in vitro experiments with BCR-ABL1+ K562 cells, stroma-derived conditioned media was sufficient to induce drug resistance to imatinib mesylate via a Stat3-dependent mechanism, while knockdown of Stat3 reversed drug resistance to imatinib (69). One factor identified as a mediator for survival of leukemia cells or for rescue from drug-induced apoptosis produced by bone marrow stroma is interleukin-7 in the case of T-cell acute lymphoblastic leukemia (70). In addition, galectin-3 (Gal-3) a member of the β-gal-binding galectin family of proteins, was induced in CML cell lines upon co-culture with the stromal cell line HS-5. Enforced expression of Gal-3 led to increased proliferation of CML cells and increased resistance to tyrosine kinase inhibitors. In vivo Gal-3 overexpression increased the lodgement of CML cells in BM (Fig. 2) suggesting that Gal-3 may be a reasonable target to decrease the proximity of CML to stroma cells in the BM microenvironment (71). Using human and mouse cells positive for the Jak2 V617F mutation, associated with the myeloproliferative neoplasms (MPN) polycythemia vera, essential thrombocythemia, and primary myelofibrosis, it was demonstrated in in vitro assays that response of these cells to the Jak2 inhibitor atiprimod was diminished upon coculture with the human stromal cell lines in a noncontact fashion. Treatment of the cocultures with neutralizing antibodies to IL-6, fibroblast growth factor, or CXCL10 abrogated the protective stromal effects on MPN cells, indicating that humoral factors secreted by stromal cells are responsible for protection of Jak2 V617F+ cells from Jak2 inhibitors.

Finally, the importance of environmental cues for the localization and differentiation of leukemic progenitors has also been established. It was demonstrated that the localization of acute myeloid leukemia-initiating cells associated with the oncogene MLL-AF9 is unimpaired by osteoblast-specific expression of the Wnt-inhibitor Dickkopf1 (Dkk1) and is independent of microenvironmental Wnt signaling. In contrast, the localization of normal HSCs is significantly reduced in such an environment (72). Furthermore, expression of MLL-AF9 in human CD34+ cord blood progenitor cells could give rise to lymphoid, myeloid or biphenotypic leukemias. This was dependent on growth factors or the recipient strain of mice arguing for microenvironmental cues for leukemic differentiation (73).

Can the LSC Niche Be Targeted in Conjunction with Other Therapies?

Targeting of the LSC niche in conjunction with conventional treatments directed at the LSC itself may represent a new paradigm for the therapy of certain cancers (74). First, approaches which target self-renewal pathways like the Wnt or Notch pathways or cytokine-mediated signaling such as blocking the interleukin-3-receptor (CD123) (75), a known marker of LSCs, are gaining significance (76). Second, strategies to block survival signals of LSCs mediated by stromal cells are another feasible strategy. It is currently being tested as to whether increased parathyroid hormone (PTH)-mediated bone turnover in a murine model of CML-like MPN can suppress CML-initiating cells and prolong survival in mice with osteoblastic cell-specific constitutive expression of the receptor for PTH and PTH-related protein (Fig. 2). Early encouraging data argues that increased bone turnover may be directly suppressive of CML and puts forward the hypothesis that the bone may be targeted, to positively influence the course of a leukemia (77). Third, it has been demonstrated that adhesion molecules like CD44 or integrins can be targeted, to impair the interaction of LSCs with their niche (52, 53) or to decrease adhesion to fibronectin in the bone marrow microenvironment (64), respectively. In addition, strategies to mobilize LSCs out of their chemoprotective niche, for instance by G-CSF (78, 63) or AMD3100 (67), have at least partly already been successful (Fig. 2). And finally, it is conceivable that the hypoxic milieu in bone marrow niches in which LSCs may likely reside and which provide resistance to chemotherapeutic agents could be targeted. In addition, angiogenesis inhibitors may decrease the survival of non-LSC leukemic cells in the bone marrow microenvironment.

What Needs to Be Done?

In summary, many open questions remain on the role of the bone marrow microenvironment, both in health and in malignant states. It may be speculated how useful murine studies may be for understanding the HSC niches in humans. However, many discoveries, for instance the use of GCSF for the mobilization of HSCS (79) or the expansion of HSCs by prostaglandins (80), are originally derived from murine studies. It is an underlying assumption that murine niche biology resembles that of humans and that lessons learned from murine studies will be valuable and translatable to the clinic, but this will ultimately need validation. For the normal HSC niche, efforts may focus on secreted or membrane-bound cytokines and adhesion receptors and the vast entity of extracellular matrix binding factors which play a role in HSPC homeostasis. Other efforts may go toward establishing the relationship between nestin+ MSCs and CAR cells or subtypes of phagocytes, which may themselves regulate nestin+ MSCs, present in the niche. The dynamic relationship between the osteoblastic and the vascular niche and factors regulating the possible shuttling of HSCs and their progenitors between these niches need to be elucidated, as well as the questions about which cell types actually directly contact HSCs. Furthermore, future efforts will be directed at what factors osteoblastic cells may secrete to support HSCs and whether other, possibly unknown cell types, may play a role in the bone marrow microenvironmental niche. It is conceivable that three-dimensional hematopoietic microenvironments consisting of de-cellularized tissue matrices and synthetic polymer niches will be experimentally constructed. These surrogate environments will mimic interactions of cells, cytokines, and mechanic forces with each other within a given biochemical milieu to model these intricate relationships in vitro (81). Applying these lessons to the LSC niche may inform about distinctions that would point to vulnerabilities of the LSC. These studies may then guide the development of novel therapeutics that can complement our current armamentarium.


We thank Mitchell Preffer for artistically drawing the figures and on-line animations. This work was supported by grants 5KO8CA138916-02 and T32 CA009216 to D.S.K., grant 2-R01HL044851-21 and The Ellison Foundation to D.T.S. and grant 1S10RR020936-01 to F.I.P. D.T.S. is a consultant for Genzyme and Bone Therapeutics, a stockholder at Fate Therapeutics and a member of the scientific advisory board at Fate Therapeutics and Hospira. The other authors have no relevant conflicts of interest.