The ERGonomics of hematopoietic stem cell self-renewal

Abstract

Self-renewal refers to cell division events during which at least one of the two daughter cells remains identical to its mother. Although examples exist for self-renewal of differentiated cell types, it is largely a process used by stem cells to make more stem cells. Self-renewal is controlled by the interactions of cell-intrinsic molecular pathways with dynamic extrinsic signals presented by the cell's microenvironment. The natural environment for stem cells is termed the niche: a tissue-specific anatomical residence that often changes throughout development and provides the proper milieu for regulated transitions between stem cell quiescence, self-renewal, and differentiation. When exposed to the proper signals, stem cells exit a quiescent state and go through symmetrical cell divisions to expand their number, or undergo asymmetrical cell divisions. Tissue stem cells arise during development, acquire self-renewal abilities, and maintain this ability during the steady state of an adult organ, when self-renewal can be activated during regeneration in response to tissue injury.

Hematopoietic stem cells (HSCs) are the best characterized of all tissue stem cells. In the mammalian embryo, definitive HSCs are first born at embryonic day 10.5 (E10.5) as clusters budding off the ventral wall of the dorsal aorta—a region termed the aorta–gonad mesonephros (AGM) (Bertrand et al. 2010; Boisset et al. 2010; Kissa and Herbomel 2010). From the AGM HSCs migrate to niches in the fetal liver and thymus prior to taking up permanent residence in the bone marrow, where they reside from E15.5 to E18.5 through the life of the organism. HSCs are multipotent, giving rise to the myeloid and lymphoid cell lineages and a number of differentiated cell types in a hierarchical fashion. Numerous tools exist for the characterization of HSC self-renewal and differentiation. They can be prospectively isolated via fluorescent-activated cell sorting (FACS) with appropriate cell surface markers and transplanted, following chemical- or irradiation-induced ablation, into recipients where they can reconstitute the entire blood system. Versions of this transplantation assay have been used in the clinic for years in treating various blood diseases, and it remains as the gold standard for measuring HSC self-renewal and differentiation. Tissue imaging of embryonic and adult blood populations and niches, as well as the use of transgenics for tissue-specific genetic manipulation, provide researchers with more direct methods for characterizing HSC biology. Establishment of these tools has proven paradigmatic, as scientists studying other tissue stem cells look to implement methods analogous to those used for HSCs (Orkin and Zon 2008).

It is not surprising that the molecular control of stem cell self-renewal involves the cell cycle machinery. While embryonic stem cells (ESCs) show unlimited self-renewal due, in part, to their mitogen-independent Rb status, control of G1–S transitions in tissue stem cells is highly dependent on signaling events that act positively and negatively through the p16^{–CDK4/6–Rb and p19–p53–p21pathways. When knocked down or overexpressed, components of these pathways show various HSC defects. Developmental pathways such as Bmp, Wnt, and Notch have been shown to influence HSC activity, as have chemical modulators such as prostaglandin E2 and retinoic acid. These signals are coordinated spatially and temporally during development to balance HSC quiescence with self-renewal and differentiation, and to specify, together with the cdx-hox code, the cell identity of any progeny (}Lengerke et al. 2008). At the DNA level, chromatin factors (such as BMI1, MLL, and DNMT3A/B) and transcription factors (such as SOX17) establish the HSC epigenetic playing field with regions of transcriptionally silent heterochromatin and active euchromatin. As stem cells are signaled out of quiescence and into asymmetric self-renewal divisions, master transcriptional regulators are induced to implement tissue specification. These transcription factors are thought to work in a combinatorial and dose-dependent fashion, both early and late during differentiation, to skew lineage choices. The best-characterized example of this occurs at the branching point between red blood cell and white blood cell specification. In a common precursor cell, the erythroid regulator GATA1 and the myeloid regulator PU.1 interact physically at the protein level and also repress each other's transcription, leading to a model of “molecule counting” for lineage choice (Zon 2008; Orkin and Zon 2008; He et al. 2009).

All of these molecular events control different aspects of HSC biology, yet, as our knowledge base grows, the assignment of a particular factor to one process—i.e., self-renewal or specification—is increasingly dependent on technical advances, subtle experimental design, and careful data interpretation. Several factors have been shown to be required for the specification of HSCs, but show dispensable or subtle roles for HSC maintenance or function. For example, loss of RUNX1/AML1 or SCL during development results in a complete lack of all definitive blood, while conditional ablations later in life result in few hematopoietic defects (Mikkola et al. 2003; Chen et al. 2009). This may be due to redundancy of family members, as was found recently for the basic helix–loop–helix factors SCL and LYL, in which a double knockout does lead to a self-renewal defect in adult HSCs (Souroullas et al. 2009). The regulation of HSC self-renewal or maintenance has also been molecularly separated from specification: Conditional loss of either ZFX or TEL/ETV6 results in adult HSC depletion with appropriate lineage differentiation before exhaustion, and SOX17 is required for maintenance of fetal liver HSCs (Hock et al. 2004b; Galan-Caridad et al. 2007; Kim et al. 2007). Some genes are required for multiple aspects of hematopoiesis: PU.1 has been implicated in myeloid differentiation and HSC maintenance (Burda et al. 2010), GFI1 is required for neutrophil maturation and adult HSC maintenance (Hock et al. 2003; Zeng et al. 2004), and GATA2 is required for specification of embryonic HSCs, mast cell differentiation, and adult HSC homeostasis (Tsai et al. 1994; Tsai and Orkin 1997; Rodrigues et al. 2005). Most of these observations were only made possible with sophisticated techniques, such as cre-mediated inducible knockout alleles, which allow the separation of early, lethal phenotypes from crucial roles later in life.

One gene that has become increasingly interesting is the transcription factor ERG. ERG belongs to the ETS family of DNA-binding proteins, whose members are critical for the development of multiple tissues. As a whole, the ETS family is thought to regulate mesenchyme–epithelial and extracellular matrix interactions (Trojanowska 2000). A number of ETS proteins—including PU.1, FLI1, and TEL/ETV6—have been shown to play critical roles at every level of hematopoiesis (Maroulakou and Bowe 2000). ERG itself is truncated, translocated, or overexpressed in multiple types of cancers—most notably prostate cancer and leukemia—and acts as an oncogene for acute megakaryoblastic leukemia (Clark and Cooper 2009; Salek-Ardakani et al. 2009). Initial studies on mice with a germline mutation disrupting ERG transactivation showed several hematopoietic phenotypes. Homozygous mutants display normal primitive hematopoiesis but die in utero between E10.5 and E13.5 with defective definitive hematopoiesis, as measured by yolk sac culture colony assays. Heterozygous animals display low leukocytes, platelets, and spleen colony-forming units (CFUs), forming half as many bone marrow progenitor colonies spread across all lineages, and heterozygous marrow fails to compete with wild-type marrow in transplantation assays (Loughran et al. 2008). Together, the data suggest a role for ERG in HSC self-renewal and/or function. Since these studies, ERG has been shown to regulate megakaryopoiesis, angiogenesis, and endothelial apoptosis, and to be required for ESC differentiation toward endothelial fate (Birdsey et al. 2008; Kruse et al. 2009; Nikolova-Krstevski et al. 2009; Stankiewicz and Crispino 2009).

In the February 1, 2011, issue of Genes & Development, Taoudi et al. (2011) expand embryonic hematopoietic analysis of a previously published mutation in the transactivation domain of ERG. Since the homozygous mutant embryos show a steep death curve between E10.5 and E11.5, Taoudi et al. (2011) set out to characterize the details of the hematopoietic defects during these time points using more thorough techniques than were used previously. While primitive hematopoietic specification was normal in ERG homozygotes between E8.5 and E9.5, cells of all blood lineages were exhausted by E10.5, as measured by CFU-S and CD45 positivity in the yolk sac. Interestingly, despite having normal definitive HSC formation at E10.5, AGM explants followed by in vitro HSC induction show that ERG mutant HSC aortic clusters cannot support hematopoiesis following transplant into irradiated mice. While mutant CD45 ^{cells were normal in the AGM at earlier time points and remained normal following the explant procedure, CFU-C counts from the same assay points show a loss of explanted cell activity that reads out across all blood cell lineages. Despite this loss of sustained expansion, mutant HSCs are capable of homing to the liver and placenta.} Taoudi et al. (2011) used chimera analysis to ask how well mutant cells can contribute to hematopoietic compartments within an otherwise healthy animal. GFP ^{wild-type, heterozygous, or homozygous mutant morulas were aggregated with GFP wild-type embryos, grown to blastulas, and injected into pseudo-pregnant females. Mutant cells were able to contribute to peripheral blood, fetal liver, and thymic CD45 cells comparable with wild-type cells at E14.5, but much less so at E18.5. On the flip side, homozygous and heterozygous cells displayed lower chimerism in the HSC compartment at both time points. Importantly, a similar low level of CD45, Lin, and HSC chimerism from mutant cells was observed in the E18.5 bone marrow as the E18.5 fetal liver, suggesting that a seeding defect is not responsible for the adult phenotypes. Finally,} Taoudi et al. (2011) show decreased Gata2 and Runx1 mRNA expression in mutant fetal liver cells, as well as liver-specific enrichment for ERG at both the Gata2 and Runx1 enhancers via chromatin immunoprecipitation (ChIP) assays. A model is presented in which ERG is required to activate Gata2 and Runx1, which are synergistically required for HSC maintenance after hematopoietic specification (Fig. 1).

Figure 1.

Specification of HSCs in the AGM of the early mouse embryo is dependent on the activity of both RUNX1 and GATA2, but is independent of ERG. Taoudi et al. (2011) observed normal levels of early progenitors and HSC intra-aortic clusters in the Erg mutant embryos. Mutants for Gata2, Runx1, and Erg all show homing ability to fetal liver and bone marrow niches. Within these niches, GATA2 and RUNX1 show a redundant requirement for HSC self-renewal. Taoudi et al. (2011) suggest that the essential requirement for ERG in self-renewal is via direct activation of both Gata2 and Runx1, such that loss of ERG activity results in low levels of each factor specifically in the later stages and not during specification. We thank Steven Moskowitz for assistance in producing the figure.

One aspect of this study gets right to the heart of key questions in stem cell self-renewal: How do defects in self-renewal manifest at the cellular level, and is self-renewal a separable phenotype from other aspects of HSC biology? Exhaustion of HSCs in the ERG mutant embryos is intriguing: The normal number of intra-aortic clusters rules out specification defects, and the detection of mutant cell contribution to fetal liver, placenta, and bone marrow hematopoiesis, albeit greatly reduced, also rules out defects in homing and occupation of secondary niches. If normal symmetric or asymmetric HSC self-renewal divisions become differentiation events in the mutants, then one would expect a burst of hematopoiesis in the mutants, yet Taoudi et al. (2011) detect no such burst. If self-renewal divisions were occurring normally, but the resulting HSCs were dying off, then one would expect an increase in apoptosis, but no such increase was detected. Ruling out technical limitations, one possibility is that normal symmetric self-renewal divisions (expansion) become asymmetric self-renewal events. This would result in a low level of normal HSC function that is consistent with the observed phenotypes.

A second aspect of this study directly addresses molecular mechanisms underlying the transition between HSC specification and HSC maintenance. Perhaps the most striking experiments performed by Taoudi et al. (2011) are the ChIPs for ERG at the Runx1 and Gata2 loci. It is worth noting that Taoudi et al. (2011) do not rule out that the difference between the AGM and liver ERG signal could simply reflect the percentage of relevant cell types in the two tissues. Also, since at least one other gene important for HSC maintenance—GFI1—has been implicated as a direct ERG target (Wilson et al. 2010b), it is possible that the ERG phenotype is due to loss of this or other interactions. Despite these caveats, the reduction of Gata2 and Runx1 mRNA levels in the liver, combined with the ChIP data, provides a compelling argument for the proposed model, especially in light of recent studies showing strong definitive hematopoietic defects in compound heterozygous mutants for Runx1 and Gata2 (Wilson et al. 2010a). This model places ERG at the molecular transition between HSC specification and HSC maintenance—the first being ERG-independent and individually dependent on both GATA2 and RUNX1, and the second being dependent on the synergistic activity of RUNX1 and GATA2 that is stimulated by ERG transactivation.

This study raises several interesting questions. (1) Why is there a need for synergistic RUNX1 and GATA2 activity? Perhaps the chromatin domains in the liver HSC are different than those in the AGM, leading to the need for stronger activation or suppression of a similar set of genes. Or perhaps an entirely new set of RUNX1 and GATA2 targets are required in the liver, and the two factors are acting within protein complexes composed of different cofactors. There is evidence that GATA2 and RUNX1 can interact physically with each other at the protein level (Wilson et al. 2010a). It is possible that the two are in the same complex in the liver, but in separate complexes during AGM specification. Comparing direct RUNX1 and GATA2 targets with ChIP-seq from the two tissues would be enlightening in this regard. (2) How do developmental regulators interface with ERG and the change in RUNX1 and GATA2 requirements? As discussed above, Bmp, Wnt, and Notch pathways are known to influence hematopoiesis during development. RUNX1 can rescue HSCs in notch mutant zebrafish embryos, and GATA2 has been placed downstream from Bmps during embryonic hematopoiesis (Maeno et al. 1996; Burns et al. 2005; Dalgin et al. 2007). ERG was shown recently to activate the Wnt pathway in prostate cancer studies (Gupta et al. 2010). It is likely that developmental signals are integrated at the level of tissue-specific transcription factors, but the molecular mechanisms governing this integration are largely unknown. Transcriptional profiling of different HCS populations during development would shed light on these topics, as would ChIP for nuclear components of the signaling pathways in different hematopoietic cell types.

ERG may be one of the best examples of a hematopoietic cell-autonomous factor functioning mainly in stem cell self-renewal through all stages of life. BMI1, the best-characterized HSC self-renewal factor, differs from ERG in the timing of its phenotype (fetal for Erg and postnatal for BMI1) and where it functions downstream from the HSC (lymphocytes for BMI1 and megakaryocytes for ERG) (van der Lugt et al. 1994; Park et al. 2003; Kruse et al. 2009). Erg phenotypes differ from those of ZFX in that Erg is required for both embryonic and adult HSC self-renewal while ZFX is only required in adult HSCs, and the self-renewal defect of ZFX can be explained with an increase in apoptosis while the Erg defect cannot (Galan-Caridad et al. 2007). Sox17 also has a cell death explanation for its fetus-specific HSC self-renewal defect (Kim et al. 2007), as does Etv6 for its adult-specific HSC phenotype. Aside from its role in lymphoid cell and neutrophil production, Gfi1 also shows an adult-specific HSC self-renewal phenotype, but maintains relatively normal levels of HSCs (Hock et al. 2004a; Zeng et al. 2004). Although it will be interesting to see a full conditional Erg knockout in the hematopoietic compartment to more directly compare its adult phenotypes with these other factors, Erg's fetal defects stand out among known HSC self-renewal genes.

One additional theme in the ERG story deserves highlighting. The ERG allele used in these studies was found through an impressive mutagenesis screen on a background that was mutant for the thrombopoietin receptor (mpl). Although the ERG mutant allele has a phenotype in the absence of mpl mutation, the embryonic studies might not have been pursued without the strong Mpl ^Erg+/− adult phenotypes. As we tease out more molecular details concerning HSC specification and self-renewal, it will become important to go beyond single-gene analysis and test more complex genetic interactions between regulatory pathways. This is further exemplified by the synergistic requirements for RUNX1 and GATA2 in maintaining fetal HCSs, and by the redundant activities of LYL1 and SCL (Souroullas et al. 2009; Wilson et al. 2010a). We will only tease out these elusive relationships by taking advantage of the latest models and methods.

How can we find new factors involved in HSC self-renewal? Forward genetic screens in the mouse are increasingly feasible methodologies for HSC biology, with the increasing throughput of assays such as FACS and hematological analysis (Loughran et al. 2008; Papathanasiou et al. 2010). Reverse genetic methods involving RNAi have been used successfully to identify factors that regulate HSC self-renewal (Hope et al. 2010). In addition, other leukemic translocation partners will be studied, and advanced live imaging of the transgenic reporter will continue to help characterize precise in vivo phenotypes (Bertrand et al. 2010; Boisset et al. 2010; Kissa and Herbomel 2010). Biochemistry together with next-generation sequencing and quantitative mass spectrometry are powerful, rapidly developing technologies that have already proven useful and hold great promise (Baek et al. 2008; Spooncer et al. 2008; Wilson et al. 2010a). Chemical genetics, particularly in the zebrafish embryo, has been fruitful in identifying hematopoietic regulatory pathways (North et al. 2007; Yeh et al. 2009; Paik et al. 2010). The methods used by Taoudi et al. (2011) represent key advances in cell culture and chimera studies, and highlight the continued power of more established approaches. Ultimately, the combination of old and new technology will undoubtedly yield deep insight into HSC self-renewal.

Footnotes