|Home | About | Journals | Submit | Contact Us | Français|
Chromosomal translocations that disrupt transcriptional regulators are frequently involved in the etiology of leukemia. To gain an understanding of the normal and pathologic roles of these transcriptional regulators, both gain- and loss-of-function mutations have been examined in the context of steady-state hematopoiesis. These studies have identified a remarkable number of genes whose loss-of-function phenotype includes a perturbation of hematopoietic stem cell (HSC) proliferation. As more of these models are generated and analyzed using commonly available tools, the regulatory pathways that control HSC quiescence and proliferation are becoming clearer. An emerging theme is that leukemia-associated transcriptional regulators coordinate the balance of proliferation and quiescence within the HSC pool by modulating the number and frequency of cells transiting the cell cycle. Uncoupling proliferation from differentiation by the aberrant generation of chimeric oncogenes that retain some, but not all of the attributes of the original transcription factor is likely to be an important step during leukemogenesis.
One approach to understanding the genetic pathways deregulated in leukemia is to study the normal functions of the proto-oncogenes involved in chromosomal translocations that typify particular leukemia subsets. In contrast to the proto-oncogenes and tumor suppressors made familiar by classic solid tumor studies,1, 2 the players in leukemia are numerous and less familiar. This may be due to the fact that in contrast to most solid tumors, leukemia arises in cells in which proliferation and differentiation are ongoing processes. Leukemia has been described as a disease in which a block in differentiation plays an important role in the transformation process.3 This observation may simply result from the fact that it is common for the coordination between cell identity and proliferation to be performed by individual transcriptional regulators. Studies characterizing the roles of individual cell-cycle regulators in adult hematopoiesis have demonstrated that loss of core cell-cycle components (including cyclin-dependent kinases [CDKs] and their inhibitors) have distinct effects on stem and progenitor homeostasis and function as reviewed by Walkey et al.4 These studies provide an essential point of reference for the work reviewed here, but are not discussed in detail in this review.
Here we review a selection of recent studies that characterize the function of specific genes in hematopoiesis, most of which have been implicated in leukemia by their identification as frequent targets of chromosomal translocation. However, we also discuss several genes whose involvement in leukemia was implicated by other experimental evidence. The genes discussed below are divided broadly into three basic groups (Fig. 1): those whose mutation results in 1) an increase in phenotypically defined HSCs with an equivalent increase in HSC function (“Homeostatic Negative Regulators”), 2) an increase in phenotypically defined HSCs accompanied by a decrease in HSC function (“Homeostatic Balancers”), and 3) a concomitant reduction in phenotypically defined HSC number and function (“Homeostatic Positive Regulators”).
This first group of genes include those that when mutated, result in an increased HSC pool size accompanied by an increase in HSC function (Figure 1, Table 1). The very existence of this category illustrates that active processes are in place to suppress HSC numbers under steady-state conditions.
Loss-of-function mutations of the proto-oncogene c-Myb demonstrate its essential role during hematopoietic development.5–7 In addition, a co-activator recruited by c-Myb, p300, is required for differentiation of multiple hematopoietic cell types.8, 9 However, HSCs in which c-Myb has selectively lost the ability to interact with p300 exhibit augmented steady-state HSCs and enhanced HSC function. This selective disruption is characteristic of a particular mutant allele of c-Myb (Met303->Val, isolated by chemically induced mutagenesis).10 Competitive transplantation assays using bone marrow from these homozygous mutant animals revealed a 5–10 fold elevation in HSC frequency, which was consistent with the increase in HSC pool size as measured by cell surface phenotype. In this and other studies described in the following sections, the lineage-negative, Sca-1 positive, and c-Kit positive (LSK) population serves as a surrogate measure of the HSC pool, as this population is highly enriched in HSCs.11 Analysis of LSK cells from the Met303->Val c-Myb mutant revealed roughly twice as many of these cells in the S/G2/M phase of the cell cycle relative to wild-type. Despite the increase in proliferation within the LSK pool, the c-Myb mutant had functionally overabundant HSCs as revealed through 3 rounds of serial transplantation, suggesting that the increased number of cycling cells has no negative consequences on HSC function.10 This example, like that of Hoxb4 over-expression12, also illustrates that an increase in HSC pool size can be maintained without necessarily resulting in leukemia.
An interesting contrast is exemplified by the Ets-domain protein MEF/Elf-4 which has been implicated in M2/M3 acute myelogenous leukemia (AML).13, 14 Constitutive loss of Elf4 results in an increase in the phenotypic long-term (LT)-HSC (here defined as LSK/CD34−/Flt-3−) with some key differences with respect to proliferation. Elf4 knockout cells out-compete wild-type cells in transplantation assays and knockout animals recover faster than wild-type mice from myelo-ablation.15 However, in contrast to the c-Myb mutant described above, steady-state or cytokine-stimulated Elf4-null HSCs exhibit a reduction in S/G2/M cells, reduced BrdU uptake, and a resistance to enter the cell cycle upon cytokine stimulation. Despite this cell-cycle perturbation within the HSC pool, Elf4 knockout animals apparently maintain normal bone marrow cellularity. These examples illustrate that the loss of transcriptional regulators or their interaction with co-activators can lead to coordinately enhanced HSC number and function. Importantly, enhanced HSC pool size and function can be achieved in conjunction with enhanced or reduced proliferation within the HSC pool, suggesting there are multiple mechanisms by which HSC numbers are normally suppressed.
Considering that the ability to maintain a quiescent state is an important characteristic of HSCs, a more intuitive response to deregulated proliferation might be the loss of HSC function, which is a common phenotype in many of the knockout models analyzed to date. The Gfi1 gene knockout was one of the first targeted mutants to exhibit the combination of an increase in cycling HSCs at the expense of HSC function.16, 17 Gfi1 was identified as a proto-oncogene activated by proviral insertion in a mouse model of T cell lymphoma.18 Although bone marrow HSC numbers were slightly increased in one study16 and decreased in another,17 both Gfi1 knockout models exhibited an increase in cycling cells within the HSC pool, a large decrease in expression of the cyclin-dependent kinase inhibitor p21, and profoundly reduced HSC function in transplantation experiments. Thus, under normal homeostasis, Gfi1 is thought to suppress the proliferation of HSCs thereby preventing their depletion.
The excision of the tumor suppressor gene, JunB, results in some phenotypes in common with the Gfi1 knockout described above. Due to the embryonic lethality, JunB excision was achieved in HSCs using the polyinositide-polycytidylic acid (polyI:C) inducible Mx1-Cre transgene19. Upon JunB loss, an expanded HSC pool was observed, specifically in the LT-HSC population (here defined as LSK/Flt-3−/Thy-1low). This expansion is accompanied by a greater percentage of LT-HSCs in S/G2/M and reduced engraftment.20, 21 These animals eventually succumb to a myeloproliferative disorder that is initiated in the HSC pool, demonstrating that the accumulated HSCs are susceptible to additional events that result in leukemia.21
MYC was discovered as a proto-oncogene in lymphoma, and has over the intervening two decades been implicated in a wide variety of solid tumors and leukemias.22, 23 The acute loss of Myc in adult bone marrow results in the transient accumulation of HSCs as measured by phenotype (LSK/Flt3−), then a progressive loss of bone marrow cells which is maximal at 8 weeks after Myc deletion. Perinatal deletion of Myc also using the Mx1-Cre system results in the accumulation of lineage negative, Sca-1 positive, c-Kit negative cells, which are proposed to reflect an aberrant, senescent primitive progenitor.24 Myc-deficient HSCs produced by either perinatal or adult excision fail to engraft in a competitive or non-competitive setting, demonstrating that the accumulated cells are functionally defective.24, 25 The accumulation of phenotypic HSCs appears to result from a combination of a slight proliferative increase within the HSC pool (possibly due to N-Myc substitution at Myc targets24, 26) and mis-expression of cell surface molecules that may retain HSCs within the bone marrow microenvironment.24, 25
RUNX1/AML1 is disrupted by chromosomal translocation in a majority of childhood leukemias, often producing a fusion protein that is thought to act by a dominant-interfering mechanism.27 Excision of Runx1 in adult bone marrow, again using the Mx1-Cre transgene, caused the stem and early progenitor fraction of cells to expand by as much as 3 fold.28–31 Interestingly, LT-HSCs isolated from these mice exhibit a proliferation-associated gene expression signature, similar to that of the short-term HSC/multipotent progenitor population (ST-HSC/MPP) despite retaining the LT-HSC cell surface phenotype.28, 32 As with the examples above, Runx1-deficient HSCs exhibited reduced repopulating activity in both competitive and non-competitive transplant settings.29 Furthermore, transplant recipients of Runx1-null bone marrow cells experienced a progressive loss of Runx1-deficient cells in peripheral blood.29 Whereas the decline of peripheral white blood cells in Myc-deficient mice was accompanied by a progressive decrease in bone marrow cellularity, cell number in Runx1-deficient bone marrow actually increased by a factor of 2, likely due to the subtle, but persistent myeloproliferation that follows conditional ablation of Runx1 with Mx1-Cre.29 In summary, the examples discussed above exhibit an early reduction of quiescent cells within highly purified HSC populations, demonstrating that these transcriptional regulators act within HSCs to affect the appropriate balance of cycling versus quiescent cells.
Disruption of several important developmental signaling pathways can also result in similar phenotypes as described for the transcriptional regulator knockout models described above. For example, acute activation of Wnt signaling, accomplished by excising exons encoding negative regulatory sequences of β-catenin19, 33 using the Mx1-Cre model, resulted in a transient increase in the number of LSK cells and an increase in the proportion that are in cycle.34, 35 These cells also failed to sustain hematopoiesis in a transplant setting or in the original Mx1-Cre animal, which succumbs to bone marrow failure within several weeks.34 The rapid lethality may result from both the depletion of HSC function and a block in the development of multiple progenitors.34, 35
Constitutive activation of the Hedgehog signaling pathway in Patched (Ptc) heterozygotes similarly results in the steady-state accumulation of phenotypically-defined HSCs and an increase in the proportion of cycling cells within this population. In contrast to activation of the Wnt pathway described above, enhanced short-term engraftment in primary recipients is observed using Ptc +/− bone marrow cells. However, HSC activity upon secondary transplantation is reduced ~3 fold, indicating the functional exhaustion of the HSC pool in this mutant.36
The disruption of phospho-inositide-3 kinase (PI3K) signaling pathways through PTEN or FoxO1/3/4 deletion using the Mx1-Cre transgene exhibits features in common with those models described above. In both the PTEN and FoxO1/3/4 conditional knockout models, acute loss of gene function is associated with a transient increase in phenotypically-defined HSCs, accompanied by an increase in the proportion of HSCs that are cycling.37–40 Both the ability of these cells to persist in the bone marrow of chimeras and to engraft secondary recipients is highly compromised. A slight increase in apoptosis appears to play a role in the homeostasis of HSCs in the FoxO1/3/4 knockout but not in the PTEN knockout.37 Ultimately, both animals acquire a myleoproliferative syndrome or leukemia, but the defects in HSC proliferation and function are temporally and clonally separable from these diseases.
Lastly, an Mx1-Cre mediated knockout of cdc42, a Rho family GTPase, similarly results in a transient increase in ST-HSCs (CD34+ LSK immunophenotype), coupled with a rapid shift from quiescence to a proliferative state within the LSK pool overall.41 In this particular model, HSC mobilization (migration out of the bone marrow) to the blood and peripheral organs is substantial, although it is also observed to a lesser extent in the FoxO1/3/4 knockout.37, 41 Similar to the other examples, Cdc42-deficient cells exhibit a severe reduction in their ability to engraft recipient animals.
This group of mouse models collectively illustrate that it is common for gene disruptions (JunB, Runx-1, c-Myc, FoxO1/3/4, cdc42) or mutations that activate pathways (Wnt, Hh and PI3K) to allow an accumulation of HSCs at the expense of the long-term preservation of HSC functions. This relationship suggests that multiple independent pathways are actively restricting proliferation within the stem cell pool, and that there exist signals in the environment to promote HSC proliferation when the above pathways are disrupted. The frequent co-occurrence of enhanced proliferation and reduced function within the HSC pool is consistent with studies illustrating that these processes are mechanistically linked.42, 43
This category of genes is required for HSC homeostasis as demonstrated by a continual decline of HSC number and function in loss-of-function models (Fig. 1). Within this category are the Polycomb (PcG) and trithorax (trxG) family genes (Table 1). These gene families were initially identified genetically in fruit flies and were demonstrated to influence the expression of their target genes in a manner that is transmissible through daughter cell divisions.44, 45 Mammalian genes related to both groups have been implicated in leukemia and other cancers.46
Bmi-1 represents one of the best-characterized Polycomb family members with respect to function within multiple stem cell types. Bmi-1-deficient mice exhibit a reduction in phenotypically-defined HSCs; these cells engraft poorly and exhaust prematurely in serial transplantation experiments. Both gain- and loss-of-function experiments have demonstrated that Bmi-1 plays a role in HSC self-renewal, with an initial focus on the role of Bmi-1 as a negative regulator of the CDK inhibitor, p16INK4A.47–49
Several other Polycomb family members, M33, Mel-18, and Rae28, have also been shown to be important for maintaining HSC function.50–52 Cell-cycle defects are not apparent in most of these loss-of-function models, but some studies report observations consistent with the premature or ectopic activation of a senescence pathway,53 although additional relevant processes are likely to be affected.
One of our groups (Jude et al.) and McMahon et al. 54, 55 have recently placed the TrG protein, MLL, in this category. Both studies demonstrated a reduction in HSC number and function using distinct Cre transgenes and gene disruption strategies and identified a role for Mll in maintaining quiescent versus cycling HSCs. Jude et al. demonstrated that upon Mll deletion, the LSK/CD48− subpopulation (enriched in quiescent cells) exhibited ectopic entry into G1. Shortly thereafter, an increased number of cells in S-phase were observed in the total LSK population based on BrdU labeling studies.54 Although these observations predict the type of transient accumulation in LSK cells as described in the previous section for β–catenin activation or Myc loss, no such increase was observed. This suggests that the increase in cycling HSCs was tightly coupled to the differentiation of these cells, thereby eliminating self-renewing cell divisions. A function for MLL in the negative regulation of proliferation specifically in HSCs was surprising, as MLL fusion oncoproteins are generally though to act by a gain-of-function mechanism.56 Therefore the simplest prediction for the normal role of MLL during hematopoiesis might have been that it maintains the expression of pro-proliferative target genes, which are then hyper-activated or maintained inappropriately in the context of oncogenic MLL fusion proteins. However, the finding that Mll loss leads to enhanced entry into the cell cycle may indicate that a distinct HSC-specific set of Mll target genes coordinate the proper balance of proliferation versus quiescence.
The constitutive knockout of the histone acetyltransferase MOZ results in some features in common with the Mll knockouts described above, likely due to its participation in some common genetic pathways. MOZ (encoded by the MYST3 gene) was identified at the site of the 8;16 translocation in AML and has been implicated in the transactivation function of Runx1.57 Myst3, like Mll, positively regulates Hoxa9.55 In addition, the closely related MOF protein has been isolated as part of the MLL multi-protein complex.58 Fetal liver cells from Myst3−/− animals lack both short-term and long-term engraftment potential and exhibit a reduction in phenotypically defined HSCs.59, 60 Although the reduced expression of Hoxa genes (amongst others) indicate that Mll and Myst3 may co-regulate shared targets, the distinct phenotypes suggest that this overlap is not extensive.60 Alterations in proliferation within the fetal HSC pool were not reported, but it will be interesting to determine whether proliferation in this model or in a conditional knockout of MOZ are affected as described above.
The genes reviewed here are divided into three categories based on the dynamic perturbation of the HSC pool, as defined both functionally and phenotypically. Many of the gene manipulations discussed here result in embryonic lethality, hence the use of inducible, tissue-selective knockout models has been essential to elucidate functions in adult HSCs. Due to the frequent use of the Mx1-Cre transgenic model, a unique opportunity for direct comparisons exists. Although the categorization presented here may represent an oversimplification in that many of the genes within a category may impinge on very different pathways to maintain a proper balance of HSCs, it is informative to appreciate shared and distinct features within these groups (Table 1). These comparisons should inform further studies aimed at identifying distinct pathways required to maintain the HSC pool in vivo. Beyond these initial reports, several groups have begun to assess compound mutants in which cell-cycle regulatory mutations are combined with mutants described in this review.61, 62 Such analyses will establish epistatic relationships among the leukemia-associated transcriptional regulators, signaling pathways, and core cell cycle machinery and hence help define the redundant and unique mechanisms that govern the regulation of HSC proliferation and the decision to self-renew or differentiate.
Due to the dynamic and flexible nature of hematopoiesis, the effects of gene knockouts (either constitutive or conditional) can be complicated by several factors. First, the loss of a given gene product simultaneously at several stages of differentiation can produce an accumulation of particular populations by lengthening the time to differentiate, increasing the production of the input to a particular progenitor pool, or blocking a subsequent differentiation step. This is discussed in particular in the analyses of Runx1 and β-catenin-activation mutants, where an independent block in the development of lymphoid and megakaryocytic, or erythroid, myeloid and lymphoid lineages, respectively, is clearly demonstrated.28–31, 33 In addition to performing careful kinetic analyses of populations after Cre-mediated excision in the inducible knockout models, an informative approach has been to perform analyses of the flux of cycling cells through stem and progenitor pools.25 Second, an important aspect of these knockout studies is the relationship between proliferation and migration out of the bone marrow niche through the circulation.63 Although the extent to which normal HSCs leave the bone marrow during normal steady state hematopoiesis is controversial,64, 65 it is clear that a number of perturbations, including agents that produce cytopenia, can increase the rate at which HSCs flux through the circulation and can shift the G0/G1 ratio within the HSC pool. Thus it is important to determine whether these proliferation phenotypes are cell-intrinsic and whether the primary defect is only in the regulation of quiescence, or in aspects of niche retention, which then influence cell proliferation. Future studies will clarify whether the disruption of particular pathways reveals an obligate linkage between bone marrow retention and HSC quiescence, or whether some transcription factor knockouts will reveal an intrinsic control of HSC quiescence.
In conclusion, the detailed analyses of the roles of leukemia-associated transcriptional regulators in adult hematopoiesis has revealed that a common function for these proteins is to maintain the appropriate balance of proliferation and quiescence within the HSC pool. The manipulation of particular signaling pathways and core cell cycle machinery components demonstrates that this field is well on its way toward defining a network of genes whose function is essential to maintain HSC function and that this information will be critical for understanding and interfering with the process of leukemogenesis.
We extend apologies to our colleagues whose work was not discussed due to space constraints or unintentional omission on the part of the authors. This work was supported by NIH grants DK067119, HL090036 and a NCRR COBRE award (P.E.), AI051427 (N.S.) and T32 AR07576 (J.G.). We are grateful to members of our laboratories for discussion and to R. Mako Saito and David Traver for critical review.