Fusion-protein truncation provides new insights into leukemogenesis

AML1-ETO (also known as RUNX1-ETO) is the fusion-protein transcription factor product of the 8;21 translocation. The translocation is present in up to 40% of leukemias of the French-American-British M2 subtype and is one of the most common events associated with myeloid leukemia. Clarifying the role of AML1-ETO in leukemogenesis has been difficult because the expression of the oncoprotein is not sufficient to cause disease. The challenge is reflected by the number of models that have been devised to study AML1-ETO in mice. In this issue of PNAS, Zhang and colleagues (1) advance the understanding of AML1-ETO another step with the discovery that a C-terminal truncation mutation produces penetrant leukemia.

Mouse Models of AML1-ETO

The earliest attempts to model AML1-ETO-associated leukemia in mice showed that the oncoprotein has detrimental developmental effects when expressed during embryogenesis (2, 3). The phenotype of the AML1-ETO knock-in embryos strongly resembled AML1 knockout embryos. The similarity provided compelling in vivo evidence supporting the hypothesis that AML1-ETO interferes with AML1. Although the embryonic lethal phenotype of the AML1-ETO knock-in was a temporary setback in the production of a leukemia model, the observations provided important clues to the action of AML1-ETO.

Subsequent mouse models were engineered to bypass the embryonic lethality of AML1-ETO. Transgenic models, inducible systems, and bone marrow transplant strategies for expressing AML1-ETO were unable to reliably produce leukemia in mice even after 24 months (4–7). However, the failure to cause leukemia was extremely enlightening. When stem cells were transduced with AML1-ETO and transplanted into lethally irradiated recipient animals, the stem cell compartment expanded dramatically (6). Similarly, direct targeting of AML1-ETO expression to stem cells by using the SCA-1 promoter enhanced myeloid progenitor expansion (7). These experiments suggested that AML1-ETO could promote the expansion of a compartment that ultimately would acquire additional “hits” leading to disease.

In later-generation models, AML1-ETO was analyzed in the context of cooperating mutations. Animals expressing AML1-ETO were mutagenized with the alkylating agent N-ethyl-N-nitrosourea (ENU). In two independent systems, mutagenized AML1-ETO-expressing mice developed myeloid leukemia or granulocytic sarcoma at frequencies greater than ENU-treated WT animals (5, 8). These important results confirmed that AML1-ETO predisposes a myeloid precursor population to transformation.

Nature of Second Hits

The requirement for cooperating events is not surprising in hindsight. AML1-ETO has a paradoxical inhibitory effect on cell proliferation and is notoriously difficult to stably express in cell lines (2, 9). Expression of AML1-ETO reduces expression of target genes, including cyclin D3 and CDK4 (9–11). The cellular consequence is impairment in the transition from G₁ to S phase and slower proliferation. Enhanced rates of apoptosis secondary to reduced BCL2 levels also contribute to reduced proliferation (9). Given the dichotomous actions of AML1-ETO, it is plausible that leukemia might develop only in animals that overcome the inhibitory influences of AML1-ETO on cell growth.

Analysis of AML1-ETO on mutant backgrounds provided insight into the nature of second hits. Tumors from mutagenized AML1-ETO mice had increases in CDK4 and cyclin D2 levels and in phosphorylation of retinoblastoma (5). However, p53, p19ARF, and p16INK4A pathways were ostensibly intact (5). Another important clue was that leukemic cells survived in culture in the absence of cytokines. The survival contrasts with the cytokine-dependent preleukemic cells from the same animals. The implication is that activation of a cytokine signaling pathway is important for transformation. Indeed, common mutations in human leukemia associated with AML1-ETO include activating mutations of FLT-3 and c-KIT (12). Formal evidence that activated cytokine signaling pathways can collaborate with AML1-ETO came from bone marrow transplant experiments where stem cells coexpressed TEL-PDGFβR with AML1-ETO. In the coexpression system 100% of animals developed a transplantable leukemia that depended on an intact platelet-derived growth factor pathway (13). The analysis of AML1-ETO on mutant backgrounds showed that cooperating events can support the leukemic expansion of a predisposed population.

Zhang and colleagues (1) provide another advance in the understanding of secondary events that promote leukemia. In the study that used retrovirally transduced bone marrow expressing AML-ETO, one animal developed leukemia within a period of just 14 weeks. Consistent with previous experience, >30 other animals remained leukemia-free. The Zhang laboratory cloned the retroviral integrant from the leukemic line and identified a point mutation that produced a premature stop codon and a truncated protein. Remarkably, transplantation of stem cells expressing the variant form of AML1-ETO produced a penetrant leukemia in 27 mice with a mean survival of 20 weeks. The AML1-ETO truncation eliminates the cellular proliferation defect associated with the full-length oncoprotein. Interestingly, a similar truncated variant is present in some human leukemias as a result of alternative splicing, although, until now, it has been largely ignored (14). The correlation between in vivo disease and restored proliferation rates suggests that impaired proliferation might account for why AML1-ETO is not sufficient to cause leukemia.

Mechanism of Action of Truncated AML1-ETO

Future investigations are likely to address the mechanism by which the AML1-ETO truncation promotes leukemia. Four domains of ETO have homology to the Drosophila protein Nervy and have been the focus of many structure–function studies. Nervy homology regions (NHRs) 3 and 4 are deleted by the C-terminal truncation. Whereas NHR3 has not been shown to be essential for AML1-ETO function, NHR4 is important in selected systems (15). Nevertheless, the role of these domains remains a subject of interest and debate. The C terminus of AML1-ETO binds the corepressor complexes associated with N-CoR (nuclear receptor corepressor) and SMRT (silencing mediator for retinoid and thyroid hormone receptor) (15). It also makes independent contacts with histone deacetylases (16). However, numerous additional domains, including NHR2, and flanking sequences contact corepressors. Based on these in vitro studies, the C-terminal truncation of AML1-ETO would be predicted to retain several corepressor recruitment domains and significant activity (16). Although the AML1-ETO truncation lacks inhibitory effects on cellular proliferation, the protein's influence on differentiation is clearly preserved (1). The divergence could represent different sensitivities of different cellular processes to AML1-ETO or different corepressor requirements at different target genes. The in vivo consequences of the AML1-ETO truncation raise many questions regarding corepressor recruitment and transcriptional repression.

Clues to the effect of the AML1-ETO truncation might come from other leukemia models. Recent studies of promyelocytic leukemia (PML) retinoic acid receptor (RAR) α and PU.1 models, among others, have demonstrated that an oncoprotein's transforming capabilities often are balanced by toxic effects. Attempts to model acute PML by using transgenic approaches and knock-ins produced only animals that expressed PML-RARα at low levels. These observations inspired the hypothesis that high levels of the oncoprotein were toxic and that penetrant disease was produced when expression levels were optimized (17). Similarly, PU.1 was capable of promoting leukemia when activity was titrated to an ideal level (18). Mice that were PU.1-haploinsufficient or PU.1-null did not develop disease. However, hypomorphs with 20% of WT PU.1 activity developed leukemia. One interpretation of the data presented by Zhang and colleagues (1) is that AML1-ETO activity is reduced to its optimum by the loss of the N-CoR/SMRT recruitment domains. Reduced activity is supported by a lack of transcriptional effects of truncated AML1-ETO on cell cycle genes (1). Future studies likely will evaluate the influence on genes important for differentiation and self-renewal. If the AML1-ETO truncation is, indeed, partially defective, it would be another example of leukemogenesis in a setting of titrated oncoprotein activity.

Potential Mechanisms of C-Terminal Inactivation

At this time, it is unclear how AML1-ETO's C terminus might be inactivated in human disease (Fig. 1). One means by which truncated AML1-ETO could be generated is alternative splicing. Humans have an alternative splice form that incorporates an alternative exon with a premature stop codon (14). Interestingly, Zhang and colleagues (1) reported that another mouse that developed leukemia expressed a truncated protein without an obvious mutation at the level of DNA sequence. Proteolytic cleavage of AML1-ETO might explain such an observation. A precedent exists in acute PML where neutrophil elastase expression and PML-RARα cleavage are critical events for the development of disease (19). Alternatively, cooperating events within cytokine signaling pathways might functionally achieve the same effect as truncation of AML1-ETO's C-terminal N-CoR/SMRT interaction domain. Consistent with this notion, common inactivating mutations of FLT-3 promote the nuclear export of SMRT and inactivation of SMRT-dependent repressors (20). The importance of the observations in mice will undoubtedly be an important topic of investigation in human disease.

Fig. 1.

Potential mechanisms of inactivating AML1-ETO's C terminus. Full-length AML1-ETO is not sufficient to cause leukemia. Three mechanisms might promote leukemia by interfering with the activity of AML1-ETO's C terminus. First, a premature stop codon could be introduced by mutation or transcriptional splicing (1). Second, proteolytic processing analogous to PML-RARα could occur (1, 19). Third, activating mutations of FLT3 could export corepressors such as SMRT and N-CoR to the cytoplasm (20).

This combination of serendipity and perseverance has yielded an important breakthrough in modeling AML1-ETO-induced leukemia. Truncation of AML1-ETO reduces the protein's ability to block G₁ to S phase progression and is correlated with penetrant leukemia. It will be important to learn whether functional inactivation of C-terminal sequences commonly takes place in human disease or if second hits take other forms. This exciting study (1) represents the first structure–function analysis of AML1-ETO in vivo and inspires reevaluation of the mechanistic paradigm accepted to date.