In this issue, Roman-Gomez et al. (1) report on the correlation of a positive CpG island methylator phenotype (CIMP) status, defined by aberrant hypermethylation of four genes (DKK3, sFRP2, PTEN, and P73) with outcome in children with acute lymphoblastic leukemia (ALL) who harbor the TEL-AML1 gene rearrangement.
In childhood, pre-B-cell ALL TEL-AML1 is the most common genetic abnormality, with ∼25% of cases expressing this cryptic t(12;21) rearrangement (2–4). Patients with TEL-AML1 have an event-free and overall survivals superior to other subtypes of childhood ALL. Recently, Loh et al. (5) reported that the event-free and overall survivals of patients treated on the Dana-Farber Cancer Institute ALL Consortium Protocol 95-01 was 89% and 97%, respectively. In the cooperative group setting,3 the event-free and overall survivals were similarly high. In spite of this success, a small percentage of children still experience relapse. It is particularly discouraging that some of these relapses occur while on therapy (6) and up to 12 years after completing therapy (7) in a group of patients considered to have an excellent prognosis.
The prenatal origin of the TEL-AML1 rearrangement has been reported by several groups (8–10). These data suggest that the TEL-AML1 rearrangement occurs in utero as an early event possibly in more than one clone. Additional mutational steps are required for leukemogenesis to be initiated and maintained. Zuna et al. (11) have shown that relapse in TEL-AML1-positive patients occurs from a reservoir of preleukemic clones, with a second TEL deletion as a secondary event in leukemogenesis. Similarly, Rothman et al. (12) have also described the coexistence of multiple TEL-AML1 clones at diagnosis with submicroscopic deletion of the uninvolved AML1 allele.
The prevalence of multiple TEL-AML1 clones before and after birth, at risk for a secondary mutagenic or epigenetic event, generates molecular heterogeneity in this subtype of leukemia. This heterogeneity may cause a drug resistance phenotype that confers a higher risk of relapse. Among the many different mechanisms that underlie this heterogeneity, aberrant hypermethylation and silencing of a key gene [e.g., a tumor suppressor gene(s) (TSG)] can affect changes in downstream targets with detrimental effects on drug sensitivity and outcome.
Papadhimitriou et al. (6) described an on-therapy relapse in a child with pre-B-cell ALL who had a deletion in the short arm of chromosome 9 in addition to a TEL-AML1 rearrangement. The 9p deletion resulted in a loss of p16ink4a protein expression that was identified retrospectively at diagnosis and relapse. These findings suggest that silencing of p16ink4a in the setting of a TEL-AML1 rearrangement may confer a poor outcome. Although in this case loss of p16ink4a expression was due to a deletion, an alternative model of epigenetic silencing of one or several genes via CpG island aberrant methylation may identify a subgroup of patients with poor outcome who have an evaluable biomarker that can be targeted with chromatin remodeling agents (13, 14).
Aberrant methylation is emerging as an important factor in the biology of leukemogenesis. CIMPs (CIMP+) were originally described in epithelial-derived malignancies (15) but have subsequently been described in myeloid and lymphoid leukemias (16–18). The criteria for declaring particular tumors CIMP+ or CIMP− have varied significantly from publication to publication. The lack of consensus over how many genes to screen, what methodology to use, and how to uniformly report methylation status reflect our lack of understanding of this phenomenon. Epigenetic abnormalities affecting known or candidate TSGs are found with varying frequency in many types of cancer (19).
The findings of Roman-Gomez et al. (1) provide an insight into a model where leukemogenesis, epigenetic genome regulation, and opportunity for drug discovery and development intersect (Fig. 1 ). One can envision two general models whereby these anomalies might arise. In one scenario, global dysregulation of the epigenetic machinery may lead to generalized hypomethylation and aberrant methylation of CpG islands in a nonspecific manner. This could result, for example, from inappropriate expression of de novo DNA methyltransferases (3a or 3b) or by loss of specificity by the maintenance enzyme DNA methyltransferase 1. When this global phenomenon results in the silencing of a key growth or survival control gene, that cell would gain a proliferative advantage and may eventually acquire further mutations or epigenetic changes on its way to transformation and maintenance of the leukemogenic state. In the second model, upstream signaling events (e.g., resulting from constitutive pathway activation) lead to the transcriptional silencing of a TSG. The unoccupied promoter now becomes subject to methylation by a seeding and spreading mechanism, ultimately acquiring sufficient methyl-CpG content to stabilize the transcriptional off-state (Fig. 2 ). This model would also require some amount of selection for the permanent loss of expression of a particular TSG. Needless to say, these are not mutually exclusive models and may explain why some tumors seem to have more methylated genes than others.
In the global dysregulation model, TSGs are caught up in a process analogous to DNA mismatch repair defects that result in microsatellite instability. By its genome-wide nature, a global defect in epigenetic programming is more likely to produce a methylator phenotype. Methylation of a TSG silenced as a result of upstream phenomena would play more of a secondary role in the transformation process and would be unlikely to result in methylation of multiple genes. The report by Roman-Gomez et al. suggests that, within a subset of childhood leukemia that, as a whole, responds well to therapy, the methylator phenotype distinguishes those patients who will fare poorly. One possible interpretation of these results would be that the pathway by which epigenetic changes are acquired can have significant effect on the clinical behavior of the disease. Again, by analogy to DNA mismatch repair defects, the mechanism that leads to global dysregulation of DNA methylation specificity remains active in the transformed cell. This epigenetic instability would provide a leukemia cell with the flexibility to rapidly acquire resistance to chemotherapeutic agents. Although the success in treating TEL-AML1 rearranged patients is unparalleled compared with other subtypes of leukemia, understanding the biology underlying this increased risk of relapse in the CIMP+ phenotype could have an effect on other subtypes of childhood leukemia.
Sweetser et al. (20) studied de novo childhood acute myelogenous leukemia samples for evidence of loss of heterozygosity. In contrast to childhood solid tumors, childhood ALL, and adult acute myelogenous leukemia, they found very little evidence of involvement of loss of heterozygosity in the development or maintenance of leukemia, suggesting that other mechanisms, such as methylation and point mutations, are involved in the development and maintenance of de novo childhood acute myelogenous leukemia. Many reports of aberrant methylation in adult and childhood acute myelogenous leukemia (21–27) continue to accumulate, begging for a standardization of detection methods and uniformity in reporting the prevalence of specific patterns of gene methylation in leukemia.
Additional retrospective and prospective investigations are needed in cohorts of patients using CIMP profiling to determine if this approach is valid and whether it can be used to develop risk stratification schemas for therapeutic interventions. Nevertheless, the work of Roman-Gomez et al. argues that targeting epigenetic instability using inhibitors of DNA methyltransferases may be a useful therapeutic approach adjunct to standard chemotherapy approaches in patients exhibiting the methylator phenotype.
Drug development in the field of chromatin remodeling is ongoing. It will be important to find disease models, such as the one reported by Roman-Gomez et al. in this issue. Available banked specimens from cooperative groups should be used to evaluate retrospectively and prospectively subtypes of leukemia with epigenetic changes that are associated with outcome. Combination therapy with chromatin remodeling agents should be pursued once a biomarker(s) is identified, preferably a biomarker that is amenable to evaluation before and after combination drug therapy to determine the minimal effective pharmacologic doses of these agents with cytotoxic chemotherapy.