Portal de Periódicos da CAPES

Loops, Lineage, and Leukemia

1998; Cell Press; Volume: 94; Issue: 1 Linguagem: Inglês

10.1016/s0092-8674(00)81215-5

1097-4172

Tariq Enver, Mel Greaves,

Zebrafish Biomedical Research Applications

Adopting a lineage from amongst two or more options is a fundamental developmental decision in multicellular organisms—which song to sing from a diverse genetic repertoire. One way or another, an exclusive option has to be chosen or imposed. This brief review concerns itself with recent insights into this problem in the context of the hemopoietic system, where a small pool of multipotent hemopoietic stem cells maintain the multiple cell lineages that constitute blood. While an extra- or intraembryonic origin for blood stem cells is still being debated, it is clear that blood is a mesodermal derivative and hemopoietic specification is subject to the same concentration gradients of signals that pattern the mesoderm and give rise to vasculature, muscle, etc. Several candidate extrinsic signals that pattern the blood component of mesoderm have been identified, as have some key transcription factors on the "blood" pathway (16Orkin S.H Zon L.I Annu. Rev. Genet. 1997; 31: 33-60Crossref PubMed Scopus (79) Google Scholar, and references therein). In principle therefore, embryological specification of blood is a developmental pattern formation problem, i.e., the positionally dependent execution of a temporally discrete program of gene expression. It is unclear however how commitment to blood versus muscle is registered and maintained. How does the blood stem cell "know" that it no longer has muscle as an available option but does have eight or more lympho-myeloid lineages at its beck and call? If we knew more about this ground state, we might be better positioned to tackle the perennial problem of how the output progeny of the stem cell adopts one out of its several lineage fates—an issue usually framed as a stochastic–deterministic debate. Whatever the mechanisms involved, they are likely to be coupled within a set of regulatory rules for the stem cell pool that govern dormancy versus cycling. Alternative fates include self-renewal, differentiation, and cell death (Figure 1A). The decision to proliferate in concert with commitment and differentiation rather than self-renew might, for example, be governed by cycling kinetics and extrinsically regulated. Moreover, the decision to differentiate might be independent of the particular lineage fate adopted (15Morrison S.J Shah N.M Anderson D.J Cell. 1997; 88: 287-298Abstract Full Text Full Text PDF PubMed Scopus (866) Google Scholar). Leukemia is one major consequence of a subversion of these rules. We can presume that the process of lineage or cell type restriction involves the formation of stable transcription complexes that initiate and consolidate exclusive programs of gene expression. Transcription factors have therefore been intensively studied as candidate instigators of lineage choice decisions (19Shivdasani R.A Orkin S.H Blood. 1996; 87: 4025-4039Crossref PubMed Google Scholar, for review). Candidature is merited either by virtue of their binding to cis-acting regulatory elements of lineage-specific genes or through association with leukemia-associated chromosomal translocations. While factors identified in this way probably only represent a subset of potential regulators, knockouts have yielded considerable insight into some of the roles they play in blood cell development and, in particular, the developmental level at which their activity first becomes critical. Thus, in many cases there have been gross or lethal perturbations in the quality or quantity of mature effector cells of a given lineage. However, measuring commitment by a cellular readout of this kind may be misleading. For example, while the knockout of GATA-1 results in a dramatic loss in red cells, closer inspection reveals that the block occurs via an arrest, postcommitment, at the proerythroblast stage (reviewed in 16Orkin S.H Zon L.I Annu. Rev. Genet. 1997; 31: 33-60Crossref PubMed Scopus (79) Google Scholar). While the jury may still be out on the commitment roles of a few factors, such as SCL/TAL and PU-1 for panhemopoiesis and Ikaros for lymphoid cells (19Shivdasani R.A Orkin S.H Blood. 1996; 87: 4025-4039Crossref PubMed Google Scholar, 16Orkin S.H Zon L.I Annu. Rev. Genet. 1997; 31: 33-60Crossref PubMed Scopus (79) Google Scholar), in most instances lineage specification appears to be largely intact at the commitment level when the activity of individual factors is ablated by gene knockout. To some extent, these data on transcription factor knockouts parallel those obtained in knockout studies of growth factors and their receptors. For example, knockouts of EPO or its receptor, while seriously affecting mature red cell outputà la GATA-1, have little or no consequence at the level of erythroid commitment (reviewed in 16Orkin S.H Zon L.I Annu. Rev. Genet. 1997; 31: 33-60Crossref PubMed Scopus (79) Google Scholar). These investigations have identified key properties of the system in terms of the combinatorial effects and redundancy. However, with respect to the question of how a cell comes to adopt lineage 1 versus lineage 2, we lack insight into the rules that specify how a winning run is set up. These adventures in genetic manipulation therefore leave a major component of mechanism unresolved. Developmental fates in biology, whatever their mechanisms, are constrained by historical antecedents. We can perhaps redefine the lineage restriction conundrum in hemopoiesis by addressing the prior issue of how multipotential hemopoietic cells register their exclusive but diverse hemopoietic potential. At some level this must be a question of program accessibility. Recent studies facilitating the isolation of multipotential cells and the establishment of equivalent growth factor–dependent cell lines with differentiative potential in vitro have provided an opportunity to interrogate the cells of interest with respect to the accessibility and expression of those genes intimately involved as modulators of lineage choice or indicators of lineage-affiliated function. The results provide novel insight into the ground state of these cells and what it may mean, at the DNA level, to be a hemopoietic stem cell but still uncommitted to wearing red or white. In essence it appears that, in cycling multipotential cells, components of competing or alternative lineage pathways are simultaneously active at the level of chromatin accessibility and gene expression in varying patterns or combinations reflecting considerable intrinsic heterogeneity and developmental plasticity (see 6Cross M.A Enver T Curr. Opin. Genet. Dev. 1997; 7: 609-613Crossref PubMed Scopus (125) Google Scholar, and references therein for review). Much earlier work by Weintraub and colleagues showed that developmental genes become primed or poised for transcriptional activation via changes in chromatin structure and accessibility prior to measurable expression (22Weintraub H Cell. 1985; 42: 705-711Abstract Full Text PDF PubMed Scopus (319) Google Scholar). This has led to the broadly accepted notion that selective gene activation is a multistep process involving chromatin remodeling (1Boyes J Felsenfeld G EMBO J. 1996; 15: 2496-2507Crossref PubMed Scopus (143) Google Scholar, and references therein). Similarly, the idea that during embryogenesis undifferentiated cells might indulge in low-level or promiscuous gene expression requiring subsequent repression has a long historical pedigree. In multipotential hemopoietic cell populations, the enhancer regions for several lineage-specific genes, including CD3δ, IGH, MPO, and β-globin, appear to be simultaneously accessible. Furthermore, these cell populations coexpress lineage-affiliated transcription factors that knockout experiments register as necessary for lineage commitment and/or expression (see 6Cross M.A Enver T Curr. Opin. Genet. Dev. 1997; 7: 609-613Crossref PubMed Scopus (125) Google Scholar, for data review). However, population level analysis doesn't tell us what is going on at the single-cell level. New single-cell RT-PCR-based data provides us with more cogent insight (2Brady G Billia F Knox J Hoang T Kirsch I.R Voura E Hawley R.G Cumming R Buchwald M Siminovitch K et al.Curr. Biol. 1995; 5: 909-922Abstract Full Text Full Text PDF PubMed Scopus (147) Google Scholar, 5Cheng T Shen H Giokas D Gere J Tenen D.G Scadden D.T Proc. Natl. Acad. Sci. USA. 1996; 93: 13158-13163Crossref PubMed Scopus (151) Google Scholar, 10Hu M Krause D Greaves M Sharkis S Dexter M Heyworth C Enver T Genes Dev. 1997; 11: 774-784Crossref PubMed Scopus (588) Google Scholar). These recent studies reveal that blood stem cells coexpress different hemopoietic lineage genes at low levels but not nonhemopoietic genes and in patterns or combinations that appear complex and not a priori predictable (10Hu M Krause D Greaves M Sharkis S Dexter M Heyworth C Enver T Genes Dev. 1997; 11: 774-784Crossref PubMed Scopus (588) Google Scholar). To plagiarize Weintraub, these cells' genetic program appears primed, but the priming is neither untethered promiscuity nor unilineage; alternative potential fates are being signaled. In this primed state, individual cells can express most of the known key regulators of lineage options (i.e., transcription factors, growth factor receptors) plus genes encoding lineage-exclusive function (e.g., β-globin, myeloperoxidase) yet remain in an apparent state of indecision (5Cheng T Shen H Giokas D Gere J Tenen D.G Scadden D.T Proc. Natl. Acad. Sci. USA. 1996; 93: 13158-13163Crossref PubMed Scopus (151) Google Scholar, 10Hu M Krause D Greaves M Sharkis S Dexter M Heyworth C Enver T Genes Dev. 1997; 11: 774-784Crossref PubMed Scopus (588) Google Scholar). Induction of differentiation focuses multilineage accessibility and expression into a unilineage pattern by consolidation and up-regulation of the chosen one and a shutdown of alternatives. One interpretation of these data is that individual lineage-affiliated loops of gene expression coexist within the progenitor population in a dynamic state of flux generating the diverse patterns observed in single cells. At some level these loops must be interactive and competitive (Figure 1B). Within this setting one clear possibility is that lineage choice becomes a threshold event that is probabilistic in nature: a single song emerging from noise. These heterogeneous profiles of gene expression within a relatively homogeneous cell population are reminiscent of those seen with prokaryotic genes in clonal populations. In this latter context, 14McAdams H.H Arkin A Proc. Natl. Acad. Sci. USA. 1997; 94: 814-819Crossref PubMed Scopus (1388) Google Scholar have argued recently that stochastic or random patterns of gene expression can produce probabilistic outcomes with respect to the adoption of alternative pathways or fate and that, in principle, this is likely to apply also to eukaryotic cells. Deciphering the molecular mechanisms underlying priming and its relation to lineage commitment in part awaits a complete picture or profile of the molecular circuitry that prevails in each of the blood lineages, but some questions immediately present themselves. Are all the components of all lineage circuits primed to some degree in stem cells or is a key or skeleton crew recruited? How and when in intramesodermal development is the initial priming achieved and how does it relate to the end game status?—it seems unlikely that all hemopoietic genes would be simultaneously primed, and evidence is accumulating for loops of gene expression and competition between them. Studies of genes that are primed at the chromatin level such as GATA-1, EpoR or MPO reveal that only a subset of the full complement of cis-acting regulatory elements are made accessible (6Cross M.A Enver T Curr. Opin. Genet. Dev. 1997; 7: 609-613Crossref PubMed Scopus (125) Google Scholar, for review; 18Ronchi A Cirò M Cairns L Basilico L Corbella P Ricciardi-Castagnoli P Cross M Ghysdael J Ottolenghi S Genes Func. 1997; 1: 245-258Crossref PubMed Scopus (20) Google Scholar). The absence of full-scale chromatin remodeling at these loci (of the type seen ultimately in the appropriate mature lineage cells) probably ensures only low-level or trickle transcription from these loci, and it is not hard to see how this could manifest as sporadic transcription and thus as "stochastic" expression of key lineage-affiliated regulators. From a mechanistic point of view, opening only a subset of cis-acting regions simplifies the problem of early priming and such a tactic allows for stepwise activation, i.e., a locus passing through various threshold states on its way to full-scale activation. The probabilistic or "all-or-none" activation of hypersensitive sites in response to transcription factor occupancy (1Boyes J Felsenfeld G EMBO J. 1996; 15: 2496-2507Crossref PubMed Scopus (143) Google Scholar) as well as variagated expression (see 11Kioussis D Festenstein R Curr. Opin. Genet. Dev. 1997; 7: 614-619Crossref PubMed Scopus (121) Google Scholar) may be important facets of this process. Similarly, priming regulatory regions with "stand in" factors such as we suggested for the MPO enhancer may be one of the mechanisms that allows a locus to remain effectively in an uncommitted holding position prior to commitment (7Ford A.M Bennett C.A Healy L.E Towatari M Greaves M.F Enver T Proc. Natl. Acad. Sci. USA. 1996; 93: 10838-10843Crossref PubMed Scopus (118) Google Scholar). Another example might be that of GATA-2 standing in for GATA-1 at the β-globin LCR. Partial activation also simplifies the problem of silencing unwanted loci once a lineage decision has been taken. Recent evidence that transcription factors, such as Ikaros, may recruit genes to transcriptionally inactive, heterochromatic regions of nuclei may also be important in this regard (4Brown K.E Guest S.S Smale S.T Hahm K Merkenschlager M Fisher A.G Cell. 1997; 91: 845-854Abstract Full Text Full Text PDF PubMed Scopus (649) Google Scholar). Evidence for the interactivity or competitiveness of loops is also accumulating. For example GATA-1 contributes to regulation of the EPO-receptor and has been implicated in the transcriptional regulation of the SCL locus; there is also considerable evidence for cross-talk between GATA-1 and GATA-2 (reviewed in 16Orkin S.H Zon L.I Annu. Rev. Genet. 1997; 31: 33-60Crossref PubMed Scopus (79) Google Scholar). These sorts of data raise the possibility that loops may not have absolutely fixed starting points, are flexible, and may thus be divertable. And, consistent with this, forced expression of transcription factors can in certain cellular environments apparently "select" a lineage outcome, presumably by "tipping" the balance. For example, forced expression of GATA-1 in the mouse oligopotent 416B cells drives megakaryocytic differentiation (21Visvader J.E Elefanty A.G Strasser A Adams J.M EMBO J. 1992; 11: 4557-4564Crossref PubMed Scopus (156) Google Scholar) and overexpression in multipotential chicken progenitors (MEP cells) directs erythroid, megakaryocyte or, eosinophil differentiation depending on GATA-1 levels achieved (12Kulessa H Frampton J Graf T Genes Dev. 1995; 9: 1250-1262Crossref PubMed Scopus (341) Google Scholar). Part of the solution clearly has to do with the now widely accepted view of the combinatorial action of transcription factors. This notion has grown now to accommodate the idea that factors themselves interact physically to form large complexes or possibly even lineage-specific "machines" (see 6Cross M.A Enver T Curr. Opin. Genet. Dev. 1997; 7: 609-613Crossref PubMed Scopus (125) Google Scholar, 16Orkin S.H Zon L.I Annu. Rev. Genet. 1997; 31: 33-60Crossref PubMed Scopus (79) Google Scholar). For example, it has been suggested that Lmo2 (Rbtn2/Ttg2) may form a transactivating complex in erythroid cells that includes SCL (TAL-1), GATA-1, E2A, and a recently identified protein, Ldb1/NLI. GATA-1 has been shown to interact with a specific cofactor, FOG, and more recently to interact with CBP, a transcriptional integrator complex (for which many different factors may compete) with intrinsic as well as associated histone acetyl transferase activity. It will be important in the future to delineate the general sensitivity and acetylation status of primed lineage-affiliated loci in stem cells. Evidence for negative regulation and competitiveness is also available. This, it would seem, is inevitable as selection of a given outcome from a multiply primed starting point necessarily includes rejection of unwanted or nonselected pathways/options. Graf and colleagues have shown that the myelomonocytic transcription factor Maf B binds Ets 1 and inhibits the expression of the transferrin receptor gene, which in turn is required for erythroid differentiation (20Sieweke M.H Tekotte H Frampton J Graf T Cell. 1996; 85: 49-60Abstract Full Text Full Text PDF PubMed Scopus (248) Google Scholar). This provides a basis for the reciprocal relationship between the activation of appropriate genes and the inactivation of inappropriate ones during lineage differentiation. The adoption or restriction of a single lineage must involve the stable re-setting of an equilibrium in favor of one loop, at the expense of alternatives. The loop that has passed the threshold barrier for dominance is likely to be consolidated by positive autoregulation whilst at the same time antagonizing an alternative loop either by competition for a common component or by direct suppression of transcription mechanisms. In this context it is not difficult to see how in principle environmental signals could modulate probability of a particular lineage outcome (i.e., endorse a loop) and/or favor, by direct clonal selection, the survival and proliferation of cells that signal their adoption of this choice. Such an arrangement in multipotential hemopoietic progenitors would still be compatible with two rather different patterns of lineage outputs: either (a) random assortment of lineage commitments amongst progeny cells in variable proportions as argued by Ogawa, Suda, and colleagues (see 15Morrison S.J Shah N.M Anderson D.J Cell. 1997; 88: 287-298Abstract Full Text Full Text PDF PubMed Scopus (866) Google Scholar, for review); or (b) a structured or preferential output of lineages, for example a hierarchical sequence of cell fate decisions that are predominantly binary as 3Brown G Bunce C.M Lord J.M McConnell F.M Differentiation. 1988; 39: 83-89Crossref PubMed Scopus (24) Google Scholar suggested. The latter pattern prevails in a number of invertebrate and vertebrate developmental systems and in at least some mammalian progenitors including those in the neuronal–glial and germ cell lineages. Many of these pathways appear to be orchestrated as a succession of binary choices involving default-dominant pairings; it would be surprising if this antique but successful mechanism did not penetrate into hemopoiesis. That there is some degree of hierarchical structure with nearest neighbor affiliations of lineages and stable oligo- or bipotential intermediates within hemopoiesis is not really in doubt. However, not all such intermediate states have been rigorously scrutinized for their full developmental potential, and the structured lineage or fate map may not be as secure or rigid as usually portrayed. Nevertheless, these patterns of output may reflect selective but shared transcriptional requirements, for example, GATA-1/NF-E2/FOG in the erythroid and megakaryocytic lineages (16Orkin S.H Zon L.I Annu. Rev. Genet. 1997; 31: 33-60Crossref PubMed Scopus (79) Google Scholar). They might also in part be an evolutionary legacy reflecting the sequential elaboration of lineages in phylogeny. A preferential output or pattern need not imply a deterministic process (14McAdams H.H Arkin A Proc. Natl. Acad. Sci. USA. 1997; 94: 814-819Crossref PubMed Scopus (1388) Google Scholar). Indeed, since the biochemical and energy demands of the different lineage loops or circuits are certain to vary, an apparently structured outcome could be probabilistic in nature. A key issue is whether or how extrinsic signals can modulate a segregated lineage pattern of commitment. This conundrum is usually posited in the form of a bipolar instructive–permissive debate (15Morrison S.J Shah N.M Anderson D.J Cell. 1997; 88: 287-298Abstract Full Text Full Text PDF PubMed Scopus (866) Google Scholar). No decisive experiment exists with multipotential stem cells, and what data there are generally argue against an exclusively instructive role (see 6Cross M.A Enver T Curr. Opin. Genet. Dev. 1997; 7: 609-613Crossref PubMed Scopus (125) Google Scholar). There are many experiments that suggest that an external signal can influence bilineage choice or bring about a switch between two lineages. Macrophages feature prominently in such experiments hinting that they might constitute the phylogenetically imprinted default for blood. But whether, in this context, inductive equals instructive, or "permissive", in a cell that has only one alternative over the intrinsic default option is unclear. The issue may be in part semantic and anyway we won't understand the process until the complexities of transcription factor assembly and function are unravelled. The new data and models clearly require much more analysis, but they will almost certainly hold valid currency for improving our understanding of pathological processes in multipotential blood stem cells and especially for leukemia. In fact earlier studies on promiscuous gene expression in human acute leukemia clones in differentiation arrest provided some of the first indirect evidence for the multilineage priming concept (9Greaves M.F Chan L.C Furley A.J.W Watt S.M Molgaard H.V Blood. 1986; 67: 1-11Crossref PubMed Google Scholar). The molecular genetics of acute leukemia has been very effectively explored in recent years, and a plethora of alterations has been found. Some of these are associated selectively with subtypes of leukemia originating from stem cells with different lineage affiliations; others are shared between different types of leukemia and indeed with other cancers. Whilst the latter often involve genes encoding ubiquitous cell cycle regulators, a remarkable aspect of the former is the extent to which they involve changes in genes encoding transcription factors and, in particular, in-frame fusion of two genes resulting in hybrid transcription factor proteins (reviewed in 17Rabbitts T.H Nature. 1994; 372: 143-149Crossref PubMed Scopus (1376) Google Scholar, 13Look A.T Science. 1997; 278: 1059-1064Crossref PubMed Scopus (961) Google Scholar). Identification of their participation in these illegitimate liaisons has fingered some of these genes as key players in hemopoietic differentiation. Similar genetic recombinants are described in sarcomas of mesenchymal origin (17Rabbitts T.H Nature. 1994; 372: 143-149Crossref PubMed Scopus (1376) Google Scholar), but it is striking that none, or very few to our knowledge, has been detected in the major human cancer subtype—epithelial carcinoma. This is unlikely to be explained by a lack of clonal selection for epithelial stem cells in differentiation arrest in carcinomas. At present, it cannot be excluded that such chromosomal translocations and the resultant gene fusions have been missed in carcinomas for technical reasons. On the other hand, we are intrigued by the possibility that the difference may be real and reflect some interesting biological distinction between epithelial versus hemopoietic and mesenchymal stem cells. One possibility in line with the speculations above is that redirection of transcription factors via gene fusion effectively bars exclusive unilineage adoption or expression. The result is then an identity crisis: perpetually "looping" leukemias, running but staying in the same place, developmentally speaking (Figure 1B5). The occurrence of such gene fusions corrupting regulatory factors that lie at the apex of a cascade of gene expression underpinning differentiation might then be expected to have a strong selective value as a component in transformation (17Rabbitts T.H Nature. 1994; 372: 143-149Crossref PubMed Scopus (1376) Google Scholar), particularly in concert with constitutive cell cycling or self-renewal. Imposition of differentiation arrest has long been seen as a consistent feature of malignant hemopoiesis, and there is some recent functional evidence that transcription factors encoded by chimeric fusion genes do have this anticipated property. The same phenomenon might occur less frequently in epithelial cancers if, for example, the genes involved in alternative lineage fates were not coexpressed in the "target" stem cells and physically accessible for recombination. Alternatively, even if they did occur, chimeric transcription genes might have less selective currency for clonal expansion. In line with these possibilities, it has been suggested that epithelial lineages may represent a default pathway with respect to the minimal involvement of lineage-specific transcriptional regulators (8Frisch S.M Bioessays. 1997; 19: 705-709Crossref PubMed Scopus (57) Google Scholar). The reciprocal benefits of studying normal and malignant cell regulation are likely to continue.