Tobacco Documents Online

7 Oncogen=c" Transcription Factors in the Human Acute Leukemias A. Thomas Look Chromosomal translocations in the human acute leukemias rearrange the regulatory and coding regions of a vadety of transcription factor genes. The resultant protein products can interfere with regulatory cascades that control the growth, differentiation, and sur- vival of normal blood cell precursors. Support for this interpretation comes from the results of gene manipulation studies in mice, as well as the sequence homology of oncogenic transcription factors with proteins known to regulate embryonic development in primitive organisms, including the nematode Caenorhabditis elegans and the fruit fly Drosophila me/anogaster. Many of these genetic alterations have important prognostic implications that can guide the selection of therapy. The insights gained from studies of translocation-generated oncogenes and their protein products should hasten the de- velopment of highly specific, and hence less toxic, forms of leukemia therapy. The human acute leukemias arise from blood cell progenitors developing in thi~ lymphoid or myeloid pathway or from primitive stem cells with multilineage po- tential. Careful analysis of clonal chromo- somal abnormalities in blast cells from leukemia patients has had a profound im- pact on our understanding of the molecu- lar changes involved in leukemogenesis (1, 2). Most striking has been the finding of nonrandom, somatically acquired chro- mosomal translocations or inversions (hereafter grouped as translocations) in up to 65% of the acute leukemias (Fig. 1) (3). These structural rearrangements affect gene expression in ways that subvert nor- mal programs of cell proliferation, differ- entiation, and survival, and they likely act in concert with other classes of genetic lesions (for example, those affecting tumor suppressors) in multistep pathways that culminate in leukemic transformation. The most frequent targets of chromo- somal translocations in the acute leukemias are genes that encode transcription factors, emphasizing the critical role of these "mas- ter" regulatory proteins in the control of blood cell development (4). The modular structure of transcription factors--includ- ing discrete DNA-binding, dimerization, and trans-effector domains---allows normal- ly unrelated sequences from different chro- mosomes to be recombined into hybrid genes that encode fusion products with al- tered function (2). Activation of transcrip- tion factor genes by chromosomal translo- cations take two main forms (Fig. 2). In T- or B-lymphoid progenitors, such genes are The author is at the Del:)artment of Expenmental Oncol- ogy, St. Jude Children's Research Hospital, Memphis, IN 38105, and the Deoartment of Ped=atncs, University of Tennessee College of Medicine, Memph=s, IN 38163, USA. E-mail: thomas.look@stlude.org frequently mobilized into the vicinity of genes encoding discrete chains of the T-cell receptor (TCR) or immunoglobulin. (Ig) molecules, resulting in inappropriate ex- pression of the translocated proto-onco- genes. More commonly, the coding exons of genes disrupted by a reciprocal transloca- tion are incorporated into a single "fusion" gene, which generates a chimeric protein with unique properties. Translocations that inappropriately acti- vate transcription factor genes in acute lym- phoblastic leukemia (ALL) and acute my- eloid leukemia (AML) show remarkable specificity for hematopoietic cells blocked in defined stages of differentiation (Fig. 1). This property suggests that the different oncoproteins produced by these chromo- somal changes specifically interfere with transcriptional networks that normally function in concert with growth factors and their receptors to regulate hematopoiesis. This hypothesis is reinforced by the results of gene manipulations in mice, which have demonstrated the profound and varied ef- fects on normal hematopoiesis of transloca- tion-targeted transcription factors, includ- ing TALl (SCL), LM02, AMLI, and C8FI3 (4). The genes targeted by chromosomal trans- locations appear to stand near the top of evolutionarily conserved regulatory cascades, as indicated by their extensive sequence ho- mology with genes controlling the earliest stages of embryonic development in primitive organisms (5). Only a few of the downstream responder genes regulated by these oncopro- reins have been identified, however, leaving substantial gaps in contemporary mtxtels of leukemogenesis. Here, 1 will review recent progress in relating the activities of dysregu- luted transcription factors active in leukemic transformation to signal transduction path. ways responsible for the control of normal hematopoiesis. I will also briefly describe emerging applications of molecular genetic findings in the clinical management of newly diagnosed patients with acute leukemia. Homeotic Genes as Targets of Oncogenic Transcription Factors tn most lymphoid leukemias of B- or T-cell origin, translocation-generated oncopro- reins appear to transform committed pro- genitors whose stages of differentiation par- allel those of the majority of cells in the leukemic clone. In many AML cases, by contrast, the tmnslocations seem to aber-, randy activate genes in primitive stem cells that have retained both multilineage and self-renewal capacity (6). These alterations are likely to affect pathways that are critical for normal proliferation and differentiation of hematopoietic progenitors. Emerging ev- idence from analysis of hematopoietic pro- genitor cells in mice and humans suggests that pathways centered around the regula- tion of the major clusters of homeobox- containing HOX genes can provide a con- ceptual framework for the actions of many of the oncogenic transcription factors shown in Fig. 1. Human cells contain 39 major HOX genes, grouped in clusters (HOX-A to HO%-D) on four separate chromosomes. These genes share extensive homology with" the HOM-C genes of Drosophikl and play important roles in axial morphogenesis and patterning (7). The HOX genes are further divided into 13 paralog groups on the basis of their structural and functional identity with individual HOM-C genes (Fig. 3). Of immediate relevance to the origin of acute leukemia, the expression of individual HOX genes in blood cell progenitors follows tightly regulated programs that are specific for the stage and lineage of progenitor cell development, with universal down-regula- tion of HOX gene expression as the progen- itors differentiate into mature blood cells (8). At least 22 of the 39 HOX genes are expressed by different subpopulations of CD34+ human bone marrow progenitor cells. High levels of expression of the 3' genes within the A and B complexes (HOXB3, for example) are found in the most primitive subsets of hematopoietic stem cells. These genes are subsequently down-regulated, and the HOX loci closer to the 5' end are expressed as progenitor cells become committed to the myeloid or ery- throid lineage (8). In mice reconstituted with retrovimlly infected murine bone marrow cells pro- grammed to overexpress individual HOX genes, there are dramatic and highly lin- eage-specific effects.on the proliferative and www.sciencemag.org • SCIENCE • VOL. 278 • 7 NOVEMBER 1097 1059

self-renewal capacities of hematopoietic stem cells and committed progenitors. Each HOX'gene appears to have a distinct effect. For example, HOXB4 overexpression in- duces selective expansion of primitive he- matopoietic stem cells that retain their ca- pacity to differentiate, whereas HOXB3 overexpression produces mice with small thymuses and increased numbers of imma- ture thymocytes that cannot differentiate (9). Mice overexpressing HOXB3 also have defects in early B-cell development and a myeloproliferative disorder characterized by splenomegaly and expansion of myeloid clonogenic progenitors. By contrast, over- expression of HOXAIO leads to selective expansion of the megakaryocytic cell com- partment with diminished numbers of monocytic and B-l~vnphoid progenitors. Af- ter a 5- to 8-month latency period, a sub- stantial proportion of these mice develop AML (I0). The leukemogenic potential of aberrantly expressed HOX proteins is also known from studies of transgenic mice (I I ) and of murine AML cells expressing endog- enous Hox genes activated by retmviml sertion (12). Figure 3B shows seven oncogenic tmn- scription factors oriented according to their predicted roles in the regulation of HOX gene expression in myeloid and lymphoid progenitors. The gene MLL, the mammali- an counterpart of the rr/thorax (trx) gene of Drosophila, has a major impact on human leukemogenesis, because it has been impli- cated in fusions with more than 25 other genes in both ALL and AML (2). In fruit flies, trx-G proteins have a positive role in the maintenance of cell type-specific pat- terns of HOM-C gene expression, apparent- ly through epigenetic mechanisms that es- tablish and sustain a receptive chromatin configuration (13). An analogous role for MLL is supported by biochemical studies (14) and phenotypic analysis of mice lack- ing or haploinsufficient for functional Mll genes (15). Thus, leukemogenic MLL fu- sion proteins are predicted to disrupt criti- cal patterns of HOX gene expression in hematopoietic progenitor cells, in ways that selectively contribute to myeloid (for exam- ple, MLL-AF9) or lymphoid (for example, MLL-AF4) acute leukemias. Conversely, Bmi-1 is the murine ortholog of the Dro- soph//a Psc protein, a member of the Poly- comb (Pc-G) complex of proteins, which collectively oppose the actions of trx-O proteins and act as highly specific silencers of HOM-C gene expression (13). Gene dis- ruption and biochemical studies support the prediction, based on sequence homology, that Bmi-I functions in an analogous man- ner in murine embryonic development (I6). Thus, when activated by retroviral insertion in murine B- and T-cell leukemias and lymphomas, Bmi-1 may down-regulate Hox target genes that are normally ex- pressed by lymphoid progenitor cells. Acting at the far end of the postulated cascade of HOX gene activation are the PBX proteins, the mammalian counterparts of the Drosoph//a extmdenticle (exd) pro- tein (17). Like exd, PBX1 forms complexes with specific subsets o{ HOX proteins (18). In a significant fraction of pre-B leukemias in children, PBX1 is specifically rearranged with the E2A gene, and the small segment of PBXI that mediates interactions with HOX proteins is critical for the transform- ing activity of the E2A-PBX1 chimera (19). This interpretation is supported by the find- ing that murine Pbx-like genes are frequent- ly co-activated with HOX genes by retrovi- ral insertion events that promote the devel- ALL AML Random 25% None Random LYL1 HOX11TALl LM01 TAL2 LM02 ,~'C, 3% Qr 14q11/TCR¢~ 4% t(8;14),t(2;8),t(8;22) 20% T~L.AMLI E~-PBX/ 6% t(1221) t(17;19) t(4;11),t(l ;11 ),t(11 ;19) [] Tcell [] Bcetl [] Pre-Bce11 [] Pro-B cell 4% t(9;22) Rg. 1. Distribution of tra~slocation-ge~erated oncoganes ~ong the ~e leuk~i~ ~ childr~ ~d you~ aduRs (1, ~. ~e pmdu~s of ~ ~ ~nes ~e m~ o~ n~l~ p~eins a~e in t~scfip~,. ~ ~e no~ ~on of BCR-~L, ~ich ~ a ~opl~m~c proton cont~ni~ ~ a~at~ ABL ~si~ ~n~e domain. In ~, gene ~si~s t~d to ~ s~- i~ ~iat~ ~ ~e of ~e ~m~ ~niz~ immunol~ic sub~ of the dise~e. ~ i~t~ ~ the ~r c~{ng on ~e c~. ~ T~I I~kemias (o~ge) ~e ch~a~e~ ~ tr~si~ons ~t ~sr~utate ~ e~re~ion ~ proto~en~ ~m~h ~e m~h~ism ~i~ in F~, ~. ~e ~ g~e is ~ ~s~ulat~ ~rough t~sl~tion into t~ ~cin~ of ~e ~ the Ig I~i ~n B~I leukemia ~nk) ~d Bu~'s lymp~ma. In the p~B ~ell~) ~d p~-B (mag~ta) immunologic sub~s of ~ ~sion g~ ~coding chimedc o~op~teins ~e pr~u~ by the m~h~ism ou~in~ in Rg. 2B. ~e BCR-~L ~d M~ g~e ~sions have ~so ~ identi~ ~ the AML1-ETO t(8;21) inv(16) NUP98.HOXA9 t(7;11) t(9:11) PML-R~R a t(3,'v) t(6;9) PLZF.RAR ~ 1% NPM.RAR~ t(15;17),t(11;17), t(3;5) [] Myeloblastic • Myek~onocytt¢ • Monob~astic chronic and acute myeloid leukemias, respectively, suggesting that they arise in primitive stem cells with multilineage potential. "Random" refers to sporadic translocatiofis tha~ have been observed only in leukemic cells from single cases. "None" refers to leukemias that tack identifiable gene abnormaJi'des. In AML, the genes show specific associations with disease subtypes corre- sponding to the morphologic stages of normal myeloid cell dev~qopment. Although both theAML1-ETO and CBFI~-MYH11 fusion genes lead to alter- ations in the CBF transcription complex, they appear to disrupt hematopoiesls through distinct mechanisms, because they are associated with different mor- phologic subtypes of AML. Other rean~gements affect the RARo~ gene on chromosome 17 in promyelocytic progenitors, resuiting in fusions with PML or, rarely, PLZF or NPM. Less frequent translocations give rise to fusion genes associated with myelodysplastic syndrome, progressing to AML (NPM-MLF1, DEK-CAN, and NUP98-HOXAg}. [Adapted with permission from (2, 34)] 1060 SCIENCE • VOL. 278 • 7 NOVEMBER 1997 • www.sciencemag.org

opment of murine AML (12). Regulators of HOX gene expression are aberrantly acti- vared in human leukemias much more fie- quendy than the HOX genes themselves, possibly because single HOX gene alter- ations are less potent than' concerted dys- regulation of subsets of these genes. It is clear, however, that at least one of the HOX genes, HOXA9, can serve as a target for chromosomal translocations, as demon- strated by the recent detection of NUP98- HOXA9 fusion genes in cases of AML with the t(7;11) translocation (20). A more speculative aspect of the model shown in Fig: 3B is the prediction that oncogenic fusion proteins involving RARer and the core binding factor (CBF) complex will ultimately be shown to act as upstream regulators of the HOX genes in leukemo- genesis. Retinoic acid (RA), acting through its nuclear receptors (RARs), has profound effects on the patterns of Hox gene expres- sion during embryogenesis (7). Individual Hox genes exhibit differential sensitivity to the concentration of retinoic acid: The 3' paralog groups that regulate anterior struc- nares of the embryo are sensitive to high levels of retinoic acid early in development, whereas the 5' groups that regulate posteri- or structures act later and respond to lower levels. A reasonable prediction is that the RARa fusion products in human acute pro- myelocytic leukemia (APL) (for example, PML-RARa, PLZF-RARe~, and NPM- RARa) disrupt .myeloid cell development at the promyelocyte stage by altering the normal sequential pattern of HOX gene expression. Similar arguments lead to the prediction that the AML1 protein (a close counterpart of the Drosophila pair-rule protein runt) and its physiologic binding partner, CBFI3 (21), act as upstream regulators that participate in the initiation of specific patterns of HOX gene expression. Fusions involving AML1 or CBFf3 generate several hybrid proteins, pri- marily AML1-ETO, TEL-AML1, AML1- lEVI I, and CBFI3-MYHll. The pair-rule pro- teins have major roles in initiating patterns of HOM-C gene expression early in Drosoph- ila embryogenesis (22); by analogy, the runt- related leukemogenic fusion proteins may interfere with the ability of the endogenous CBF complex to establish the normal pat- tern of HOX gene expression during hema- topoiesis. Amino acid motifs contributed by the different AMLI and CBFI3 fusion part- ners appear, to result in different alterations of HOX gene regulation in progenitors of different lineages. This would account for the association of AML1-ETO, AML1- EVIl, and CBFI3-MYH11 with distinct mor- phologic subtypes .of AML across ~11 age groups, whereas TEL-AML1 occurs only in pro-B cell ALL, predominately in children. Little is known about the downstream targets of the vertebrate HOX proteins, making it difficult to expand the conceptu- al model in Fig. 3 beyond immediate regu- lation of the HOX genes themsel~,es. Some clues have emerged from studies with Dro- sop/a//,*, which have demonstrated the im- portance of autoregulatory loops among the HOM-C genes and have identified addi- tional HOM targets, such as the decapen- tap/egic, Distal-less, teashirt, and wingless genes (7). The ability of HOX and HOM-C genes to influence cellular identity and re- gional body structure during morphogenesis implies that targets of HOX regulatory pathwa~ must ultimately include proteins that regulate cell proliferation, survival, ad- hesion, and migration---aspects of cell phys- iology that likely contribute to many of the recognized abnormalities of leukemic blast cells. , The pleiotropic potential of homeotic proteins is best illustrated by the orphan homeobox-containing HOX11 protein, which does not lie in the major HOX gene clusters mentioned thus far, and which con- tributes to T-cell ALL when its expression is dysregulared by chromosomal rearrange- ments. Murine Hoxll is required for spleen development, through a mechanism that appears to affect survival of splanchnic pre- cursors (23). In addition, HOXll interacts with phosphatases that normally function within a G2-phase checkpoint; thus, its dys- regulation in T cells may interfere with this checkpoint and cause aberrant entry into Fig. 2, Two distinct mech- anisms by which chromo- somal translocations ab- errantly activate gene~ en- coding transcription fac- tors (TF). {A) Transcription factor proto-oncogenes that are silent or ex- pressed at low levels in the progenitor cells of a partic- ular lineage may be acti- vated when pl~ under the control of potent en- hancer elements within the regulatory region (R) of a gene that is norrnally highly expr&ssed. Typical. ly, the regulaton/region in these cases is contributed by one of the Ig or TCFI genes present in lymphoid progenitors of either the B or T lineage. (B) More commonly, chmmosornal mitosis (24). Overall, because the HOX genes appear to regulatd multiple pathways that control cell fate, they provide a com- pelling focus for research on the proximal sites of action of many of the oncogenic transcription factors. Subversion of Apoptosis in Leukemic Transformation ! ! Aberrant regulation of HOX gene expres- sion is by no means the sole mechanism through which oncogenic transcription fac- tors transform cells. A crucial step in hema- topoietic development is the elimination of B- and T-lymphoid progenitors that fail to rearrange their antigen receptor genes in a productive manner. Defective cells, esti- mated to include up to 75% of B-cell and 95% of T-cell precursors, are destroyed by apoptosis. This death program is thought to be triggered by damaged DNA or by the inability of cells to receive survival signals from antigen receptors expressed at the cell surface (25). The idea that subversion of cell death pathways might lead to malignant transfor- mation gained momentum with the discov- ery of the BCL2 proto-oncogene at the site of atranslocation between chromosomes 14 and 18 in human follicular lymphomas (2). Overexpression of BCL2 has litde or no effect on cell differentiation or prolifera- tion, but rather prevents lymphocytes from initiating apoptosis in response to a number of stimuli (26). Ultimately, BCL2 was 2063633077 ! Ig Or TCR gene TF proto-oncogene TFoncogene ! ! T~I~O ~ ,~1~ TF gene Chimeric TF oncogene breakpoints occur within introns, between the coding sequences of each of two transcription factor genes on different chromosomes, producing a fusion gene that encodes a chimenc transcription factor with altered function. The regulatory sequences that ddve expression of the hybrid gene are generally derived from the gene that contributes the amino-terminal amine acids to the chimeric protein; the carboxyl-terrr~al amino acids often derive from a gene that is not normally expressed in the progenitor cells in which the chimedc oncoprotein arises. Arrows indicate gene expression. www.sciencemag.org • SCIENCE • VOU 278 • 7 NOVEMBER 1997

found to be a structural and functional ho- molog of CED-9, a dominant repressor of programmed cell death in the nematode C. elegans (27). Thus, by dysregulating a key anti-apoptotic protein in lymphocytes, the t(14;18) translocation promotes the accu- mulation of cells that otherwise would die. These cells then acquire additional muta- tions needed for full malignant conversion. Although BCL2 does not appear to be dysregulated by chromosomal rearrangement in the acute leukemias, investigations of the oncogenic E2A-HLF fusion protein, formed by the t(17;19) in pro-B lymphocytes, sug- gest that it disrupts an early step in a con- served cell death pathway that censors im- mature B lymphocytes. Dominant-negative inhibition of E2A-HLF in wansformed lym- phocyte progenitors induces apoptosis (28), suggesting that the chimeric protein increas- es the numbers of immature lymphoid cells by pre,Centing their death. Consistent with this hypothesis, E2A-HLF blocks apoptosis in growth factor-deprived mouse pro-B cells and inhibits p53-mediated apoptosis trig- gered by ionizing radiation (28). Sequence homology between the basic leucine zipper (bZIP) transcription factors HLF (hepatic leukemia factor) and CES-2, a cell death specification protein in C. elegam, suggests the model shown in Fig. 4 (29). In . the developing nematode, loss-of-function mutations of ces-2 are associated with in- creased activity of ces-l. This correlation leads to the aberrant survival of two super- fluous serotonin-containing neurons (NSM sisters) through down-regulation of an apo- ptotic program that includes death effectors ced-3 and ced-4 or through up-regulation of the survival gene ced-9, each of which has Fig. 8. A model based on con- A served pathways of homeotic gene action that integrates the actM'des of seemingly diverse transcription fac- tors in leukemogenesls. (A) Studies of Drosophila embryogenssis have established that the body plan of the fly is determined by a regulato~ cas- cade centered around the control of homeobox gene expression within the HOM-C complexes (7). Genes at the 3' end compdse theAntenna- pedia complex (/ab, fob, Dfd, Scr, and Antp) and encode proteins that control the formation of anti, or structures during embryonic devel- oprnent; more posterior segments are controlled by proteins encoded: by genes of the B/thorax complex (Ubx, abd-A, and Abd-B). Upstream regulators of genes within these complexes include the gap and pa~- rule genes, which act in cono~t to . initiate the expression of spedfic HOM-C genes acsord~g to the spatiai domains occupied by eech embryonic progenitor ceil. Cont~J- ous maintenance of HOM-C gene expression dudng Drosophtla devel- opment requires two groups of tran- scription factors: trithorax (trx-g), whose members positively regulate and sustain the active state of HOM-C gene expression in cells, and Polycomb (Pc-G) proteins, which oppose the effects of trx-g proteins and continuously extin- Target gen e,.=s B Unknown target genes _1_ ' Human IHOX-B ~ m m ~n~ i~x'~ le~ .. ~ • L"°x'° ~----~'_ . ~ Paraloggroups 3' 1 2. 3 4 5 6 7 8 9 10 11 12 13 guish HOM-C gene expression in cells in which these genes are meant to remain inactive. The products of different HOM-C genes form complexes with other transcription factors, such as extradenticte (exd), which modulate binding-site and target-gene specificity. (B) The vertebrate HOX genes also dictate the body plan dudng embtyogenesis. The human counterparts of the Drosoph/la HOM-C genes, ~ on a single chromosome in the ~, are designated HOX-A through HOX-D and are arranged on four separate chromosomes (four rows of colored squares). Genes within the HOM-C and HOX ctuste~ show striking structural and functional conservation, as indicated by the color coding. Most of the HOXgenes appear to have regulatory roles in nom~ hematopoiesls (8). In the model, seven translocation-assoclated proteins are predicted to regulate HOX gene expression, on the basis of the effects of their Drosophila homologs. The chimeric oncoproteins involved in leukemogenesis (Rg. 1) are postulated to act by disrup'dng the activity of their normal counterparts in HOX gene regulation. 1062 counterparts in mammalian cell death pro- grams (27, 30). In human pro-B h.anpho- cytes, E2A-HLF is postulated to compete with a mammalian CES-2-1ike protein for a common promoter binding site and to tram- activate (rather than repress) a ces-l-like gene, resulting in reduced apoptotic activity and prolonged survival of pro-B cells that otherwise would be targeted for destruction. The acquisition of additional mutations by these aberrantly surviving cells is probably required for the generation of a fully trans- formed leukemic clone. At least one other oncogenic transcrip- tion factor, PML-RARa, has anti-apoptotic properties (3 l). Paradoxically, the MYC and E2A-PBX1 transcription factors, though clearly associated with the induction of hu- man leukemias in vivo, promote apoptosis when overexpressed in experimental mod- els (32). This result suggests that in lym- phoid cells expressing these oncogenic pro- teins, additional mutations capable of dis- abling apoptotic pathways are needed to permit leukemic transformation. Therapeutic Significance Treatment of the acute leukemias has pro- gressed from uniform strategies devised for large groups of patients to more refined pro- tocols tailored to the risk of relapse in discrete subgroups (33). Although routinely recorded features, such as the blast cell immunopheno- type and the presenting white blood cell count, provide useful criteria for risk assess- ment, molecular genetic changes appear to offer the most sensitive markers of potential leukemia cell aggressiveness and hence are the best guides to treatment (33, 34). In ALL, the BCR-ABL and MLL-AF4 fusion genes are among the few prognostic markers indicating a requirement for hema- topoietic stem cell transplantation in first remission (Table 1). Although the leuke- mias identified by these changes may re- spond well to standard treatment, they al- most always acquire drug resistance within the first year or two after diagnosis, perhaps because the initial molecular lesion arises in pluripotent stem ceils with an unlimited capacity for self-renewal (35). The E2A- PBXI chimera distinguishes an important subgroup of children with pre-B leukemia who have suboptimal responses to antime- tabolite chemotherapy that is effective for most B-lineage leukemias but fare well when placed on more intensive regimens similar to those used for T-cell ALL and other high-risk cases (36)• At the opposite end of the spectrum, the TEL-AML1 fusion gene, found in ~20% of all children with pro-B-cell ALL, is associated with extended survival and probable cure in more than 85% of patients (37). Moreover, its prog- SCIENCE • VOL. 278 • 7 NOVEMBER 1997 • www.sciencemag.org 2063633078

nostic importance exceeds that of the hge and white blood cell count of the patient at diagnosis, the gold standards of risk classi- fication in this disease. Thus, by careful identification of TEL-AMLl-positive leu- kemias at the time of diagnosis, one can administer well-tolerated antimetabolite- based treatments with a high likelihood of a favorable outcome. Indeed, this molecularly defined subgroup affords the opportunity to investigate new therapies that could reduce short- and long-term toxicities without ieopardizing high cure rotes. Although therapy for AML has become increasingly complex, more than half of the patients treated with chemotherapy alone can still be expected to relapse and succumb to their disease (33). Until recently, there were few prognostic markers that could identify AML patients who were likely to respond well to therapy. It now appears that the AMLI-ETO and CBFf~-MYHII fusion genes will identify patients with a favorable prognosis when treatment consists of inten- sive chemotherapy including high-dose cyt- ambine (38), and PML-RARa distinguishes cases of APL that are particularly sensitive to all-trans-retinoic acid, a compound that induces differentiation of leukemic promy- elocytes (39). When this agent is incorpo- rated with anthmcyclines into chemother- apy for APL, excellent long-term remission rotes are achieved without hematopoietic stem cell transplantation (40). Chimeric messenger RNAs (mRNAs) transcribed from fusion oncogenes provide unique sig- natures for molecular detection of minimal residual leukemia after treatment (33), par- ticularly in cases of APL harboring PML- RARo~ (41). However, the persistence of AMLI-ETO or CBF~3-MYHI I transcripts in the bone marrow and peripheral blood of AML patients in long-term remission after chemotherapy or bone marrow transplanta- tion (42) raises questions about the general applicability of this strategy. Thus far, the use of retinoic acid for therapy of APL affords the only example of effective treatment directed to a chimeric transcription factor. However, transloca- tion-generated fusion proteir~ are truly "tu- mor-specific" and, as such,~rovide novel targets for therapy. One approach would be to use antisense oligonucleotides (43) or ribozymes (44), designed to inactivate the mRNAs encoding chimeric oncoproteins or their downstream effectors. An alternative form of targeted therapy may soon be pos- sible, based on a new technolog3" that per- mits the design of small molecules that repress the transcription of specific genes (45). A decided advantage of such a meth- od would be the reduced likeliho~xt of the outgrowth of resistant cells, a maior liability of contemporary" forms of cancer chemo- Table 1. Therapeutic implications of commonly disrupted transcription factor genes in the acute leukemias of childhood. Leukemia Risk of treatment Recommended treatment~L Altered genes subtype* affected failuret TEL-AML I E2A-PBX1 MYC MLL-AF4 • BCR-ABL AML 1-ETO CBFB-MYH 11 PML-RARe= Acute lymphoblastic leukemia Pro-B cell Low Pre-B cell Intermediate B cell High CD10- pro-B cell Very high Pro-B cell (predominantly) Very high Acute myeloid leukemia Acute myeloblastic Intermediate leukemia with maturation (M2 morphology) Acute myelomonocytic Intermediate teukemia with eosinophils (M4Eo morphology) Acute promyelocytic Intermediate leukemia (M3 morphology) Well-tolerated chemotherapy (antimetabolites primarily) Intensive chemotherapy (genotoxic drugs and antimetabolites) Intensive chemotherapy (rotation of genotoxic drugs) AIIogeneic stem cell transplantation Allegeneic stem cell transplantation Intensive chemotherapy (including high-dose cytarabine) Intensive chemotherapy (including high-dose cytarabine) Intensive chemotherapy (including all-trans-retinoic acid and anthracyclines) "Subclassifications of AML are those of the French-American-British (FAB) group. "~As determined in standard programs of chemotherapy (without hematopoietic stem cell rescue). Treatment failure refers either to remission induction or to remission maintenance, or both. The average rates of long-term, leukemia-free survival in children and adolescents with ALL or AML range from 60 to 70% end from 30 to 40%, respectively (33). :~:The choice of therapy is based on detection of the indicated fusion gane at diagnosis by cytogenetic analysis, Southern blotting, o~ RNA-potymerese chain reaction assays for chimeric mRNAs. Normal C. elegan$ ~ Reduced activity NSM sisters die Human die Abnormal binding NSM sisters survive activity Pro-B cells (leukemogenesis) Fig. 4. Model of E2A-HLF action in pro-B lymphocytes, based on the role of the CES-2 cell death specification protein in C. elegans. The CES-2 basic leucine zipper (bZIP) transcription factor negatively regulates the (still uncloned) ces-I gene, leading to timely elimination of two'serotonergic neurons (NSM sisters) during development. Loss-of-function mutations that inactivate ces-2 lead to increased activity of ces- 1, resulting in aberrant surviva/of these neurons in worms that survive to adulthood. The CED (cell death abnormal) proteins, including the CED-3 and CED-4 death effectors and the CED-9 survival protein, control the execution of CES-2-targeted neurons in C. elegans (30). By analogy, a similar cell death specification program may eliminate defective pro-B cells during development of the human immune system, through a process that appears to be disrupted by the E2A-HLF chimeric oncoprotein. The E2A (E12 and E47) proteins contain two domains with potent trans-activating activity (46), both of which are included in the amino-terminal portion of the E2A-HLF chimera. Because CES-2 and E2A-HLF b=nd to the same dyed-symmetric DNA sequence element (29), the hybrid protein is predicted to oppose the trans-repressor activity of a putative mammalian CES-2 ortholog, activating an evolutionarily conserved surwval pathway in pro-B cells. [Adapted with permission from (29)] www.sciencemag.org • SCIENCE • VOL. 278 • 7 NOVEMBER 1907

19. 20. 21. 22. 23. 24, 25. 26. 27. 28. 29. 30. therapy. In the not-too-distant future, it may be possible to link the different path- ways controlling apoptosis, differentiation, and proliferative capacity in hematopoietic cells, providing a more comprehensive pic- ture of the genetic basis of acute leukemia and perhaps new molecular targets for the design of effective therapy. REFERENCES AND NOTES 1. T. H. Rabbitts, Nature 372, 143 (1994). 2. A. T. Lcok, in The Metabolic Basis of Inherited Dis- ease CD-ROM, C. R. Scriver and B. Vogelstein, Eds. (McGraw-Hill, New York, 1997). 3. J. D. Rowley, Semin. HernatoL 27, 122 (1990); E. Solomon, J. Borrow, A. D. Goddard, Sck~ce 254, 1153 (1991); S. C. Raimondi, Blood81,2237 (1993). 4. R. A, Shivdasani and S. H. Orkin, Blood 87, 4025 (1995). 5. A. T. Lo~k, Adv. CancerRe& 67, 25 (1995). 6. D. Bonnet and J. E. Dick, NatureMed. 3, 730 (1997). 7. R. Krumlauf, Ce//78,191 (1994); M, Maconochie, S. Nonchev, A. Morrison, R. Krumlauf, Annu. Rev. Gen- et. 30, 529 (1996). 8. G. Sauvageau eta/., Proc. Nat/. Acad. ScL U.S~. 91, 12223 (1994); H. J. Lawrence and C. Largrnan, B/cod 80, 2445 (1992); H. J. Lawrence eta/., Exp. Hematol. 23, 1160 (1995). 9. G. Sauvageeu eta/., Immunity6, 13 (1997); G. Sau- vageau eta/., Genes Dev. 9, 1753 (1995). 10. U. Thorsteinsdottir et aL, Mol, Cell. Biol. 17, 495 (1997), 11. A. Perkins, K. Kongsuwan, J. Visvader, J. M. Adams, S. Cory, Proc. Natl. Acad. ScL U.S.A. 87, 8398 (1990). 12. C. Blatt, D, Aberdam, R. Schwartz, L. Sachs, EMBO J. 7, 4283 (1988); K. Kongsuwan, J. A~lan, J. M. Adams, Nucleic Acids Res. 17, 1881 (1989); T, Na- kamura, D. A. Largaespada, J. D. Shaughnessy Jr., N. A. Jenkins, N. G. Copeland, Nature Genet. 12, 149 (1995). 13. A. $chumacher and T. Magnuson, Trends Genet. 13, 167 (1997). 14. P. L. Broeker eta/., B/cod87, 1912 (1996). 15. B.D. Yu, J. 1_ Hess, S. E. Homing, G. A. Brown, S. J. K~,~meyer, Nature 378, 505 (1995). 16. M.J. Alkema, N. M. van der Lugt, R. Co Bobeldijk, A. Betas, M, van Lohu[zen, ibid. 374, 724 (1995); M. J. A~kema eta/., Genes Dev. 11,226 {1997). 17. C. Rauskotb, M. Pellet, E. Weischaus, Ca//74, 1101 (1993~; M. A. Van Dijk and C. Murre, ibid. 78, 617 (1~94). 18. C. P. Chang et al., Genes Dev. 9, 663 (1~95~; Q. Lu and M. P. Kampa, MoL Cell. BIO/. 16, 1632 (1996); M. A. Van Dijk, L. T. Peltenburg, C. Murre, Mech. Dev. 52, 99 (1995). C. P. Chang, I. de V'rvo, M. L. Cleary, Mo/. Ca/L Biol. 17, 81 (1997). J. Borrow eta/., Nature Genet. 12, 159 (1996); T. Nakamura et aL, ibid., p. 154. N. A. Speck and T. Stacy, C#t. Rev. Eukaryot. Gene E~or. 5, 337 (1995). M. Bianz and J, Muller, B~3eesays 17, 775 (1995). C. W. M. Roberts, J. R. Shutter, S. J. Kersmeyer, Nature 368, 747 (1994); T. N. Dear eta/., Develop- merit 121, 29O9 (1995). T. Kawabe, A. J. Muslin, $. J. Korsmeyer, Nature 385, 454 (1997). A. Streeser, Curt. Opin./mmuno/. 7, 228 (1995). D. !_ Vaux, S. Cory, J. M. Adams, Nature 3,35, 440 (1988); G. Nunez eta/., J. /mmunoL 144, 3602 (1990). M. O. Hengartner and H. R. Horvitz, Cell 76, 665 (1994). T. Inaba et aL, Nature 382, 541 (1998). M, M. Metzstein, M. O. Hangartner, N. Tsung, R. E. Ellis, H. R. Horvitz, ibid., r~. 545. H. R. Horvitz, S. Shaham, M. O. Hengartner, Cold Spring Hadoor Symp. Quant. Biol. 59, 377 (1994); H. Zou, W. J. Henzel, X. Uu, A. Lutsohg, X. Wang, Cell 90, 405 (1997); E. S. AJnemn et al.. ibm. 87, 171 (19963. 31. F. Gdgnani etaL, Cell 74, 424 (1993), 32. G. Peckham and J, L. Cleveland. Biochim. Biophys. Acta 1242, 11 (1995); K. S. Smith, Y. Jacob& C, P. Chang, M. L, Cleary, Cncogene 14, 2917 (1997). 33. C. H. Pui, N. EngL J. Med. 332, 1618 (1995). 34. A.T. Look. =n Hematology, R. Hoffman, Ed. (Churchill LMngstone. New York, ed. 2, 1995), pp. 1047-1066. 35. M. I_ Gishizky and O. N. Witte, Science 256. 836 (1992), 36. W. M. Crist eta/., Blood 74, 1970 (1989); S. C. R~mondi eta/,, J. C/in. Once. 8, 1380 (1~)0). 37. S.A. Shurtteffeta/., Leuke~n~ 9, 1985 (1995); T. W. McLean etaL, Blood 88, 4252 (1996); J. E. Rubnitz eta/., ibid. 89, 1143 (1997); J. E. Rubn~ eta/.. J. Clio. OncoL 15, 1150 (1997); A. Borkhandt eta/., Blood 90, 571 (1997). 38. J.A. Maninez-Climent et aL, Leukemia 9. 98 (1998); K. Mrozek, K. Heinonen, A. de la Chapelle, C. D. Bk)om~ld, Semin. Oncol. 24, 17 (1997). 39. M. E. Huang et al., Blood 72, 567 (1988); S. Castalgne et al., ibid. 78, 1704 ( 1990); R. P. Warrell Jr. eta/., N. Engl. J. Meal. 324, 1385 (1991). 40. P. Fe~aux et aL, Leuk. Lyrnphoma 16, 431 (1995). 41. Do Divedo et aL, Blood 82. 3556 (1993); W. H. Miller Jr. eta/,, ibid,, p. 1689. 42. G. Nucifora, R. A. Larson, J. D. Rowley, ibid., p. 712; K. Tobal, P, R. Johnson, M, J. Saunders, J. A. Yin, 43. 44. 45. 46. 47. Br, J. Haernatol. 91, 104 (1995); J. Judander et aL, Blood 88. 2183 (1996). T. Skorsk~ eta/,, Blood 88, 1005 (1996); F. K. Askad and W. M. McDonnell, N. Engl, J. Med. 334, 316 (1996). U. Pace et aL, CancerRes. 54, 6365 (1994). J. M. Gottesle!d, L. Needy, J. W, Trauger, E. E. Baird, P. B. Dervan, Nature 387, 202 (1997). P. Henthorn, M. Kiledjian, T. Kadssch, Science 247, 467 (1990); A. Aronheim, R. Shiran, A. Rosen, M. D. Walker, Prec. Natl. Acad. Sci. U.S.A. 90. 8063 (1993); M. W. Quong, M. E. Massan, R. Zwart, C. Murre, MoL Cell. BioL 13, 792 (1993). I thank S. J. Baker, J. L. Ctevaland, J. R. Downing, S. P. Hunger, J. D, Ucht. S. W. Morns, S. D. Nimer, C.-H. Pui. C. J. Sherr, and J. van Deursen for hetpful discussions and comments regarding the manu- script, L S. Rawlinson for assistance with the ~g- uras, and J. R. Gilbert for cdtic81 comments and editorial assistance. Supported by NIH grants CA- 59571, CA-20 t 80, and CA-21765 and by the Amer- ican Lebanese Sydan Associated Charities, St. Jude Children's Research Hospital. A.T.L holds the had Nassar A]-Reshid Chair of Leukemia Research. Because of space limitations, it was not possible to include a comprehensive list of references for all of the work discussed. Readers interested in additional pdmary articles on translocation-ganerated onco- genes in the acute leukemias are referred to (2). Integrating Genetic Approaches into the Discovery of Anticancer Drugs Leland H. Hartwell, Philippe Szankasi, Christopher J. Roberts, Andrew W, Murray, Stephen H. Friend* The discovery of anticancer drugs is now driven by the numerous molecular alterations identified in tumor cells over the past decade. To exploit these alterations, it is necessary to understand how they define a molecular context that allows increased sensitivity to particular compounds. Traditional genetic approaches together with the new wealth of genomic information for both human and model organisms open up strategies by which drugs can be profiled for their ability to selectively kill cells in a molecular context that matches those found in tumors. Similarly, it may be possible to identify and validate new targets for drugs that would selectively kili tumor cells with a particular molecular context. This article outlines some of the ways that yeast genetics can be used to streamline anticancer drug discovery. 2063633080 "['he recent remarkable progress in identi- fying molecular alterations in human rumor cells has unfortunately not been paralleled in the field of anticancer drug discovery. The shortage of effective anticancer drugs is due in part to the fundamental difficulties associated with the development of any safe effective drug. For example, it remains a formidable task to design small molecules L H. Hertwe{I, P, Szankasi, and S. H. Fdend are at the Seattle Project, Molecular Pharmacology Department, Fred Hutchinson Cancer Research Center, Seattle, WA 98109, USA. C. J. Roberts L~ at Rosetta Inphannatics, Incorporated, 12040 115th Street NE, Kiddand, WA 98034, USA. A. W. Murray is in the Department of Phys- iolo~, University of California at San Francisco, 513 Par- nassus Avenue, San Francisco, CA 94142-0444, USA. *To whom correspondence should be addressed. that alter ~e function of macromolecules with both sensitivity and specificity (for example, an enzyme with a small active site). It is even more difficult to inhibit protein-protein interactions m&tiated over a large surface, or to restore function to a defective protein (such as an inactive tumor suppressor protein). Even when successful, massive efforts are required---often mea- sured in years to decades--from dozens of chemists, biochemists, and toxicologists. There are also many difficulties specific to anticancer drug discovery programs. An effective chemotherapeutic must selectively kill rumor cells. Most anticancer drugs have been discovered by serendipity, and the mo- lecular alterations that provide selective ru- 1064 SCIENCE • VOL. 278 • 7 NOVEMBER 1997 • www.sciencemag.org