Methylation of the ABL1 Promoter in Chronic Myelogenous Leukemia: Lack of Prognostic Significance

One hundred eighteen cases of CML seen and followed-up at the MD Anderson Cancer Center (MDACC; Houston, TX) between 1988 and 1993 were selected for this study based on the availability of stored leukemic cells. The characteristics of the 109 cases included in the final analysis are detailed in Table 1. Risk group was assigned according to the overall prognostic model.1 Leukemia samples were obtained from the bone marrow of patients at the time of diagnosis or referral to MDACC. Patients were treated at MDACC on a variety of chemotherapy protocols that were approved by the Investigational Review Board of the MDACC in accordance with the policies of the Department of Health and Human Services. All patients gave informed consent for the use of their tissue samples. Therapy varied according to the treatment period. Patients in ECP CML (diagnosis to study entry <12 months) were offered interferon-α (IFN-α)–based regimens. Therapy in late chronic phase, accelerated phase, and blast crisis depended on the time of referral. The different first-time treatments offered to patients are detailed in Table 2.

DNA was extracted from frozen mononuclear cell fractions using standard methods. Southern blot analysis was used to determine the methylation state of the ABL1 CpG island. This island contains a cluster of methylation-sensitive restriction enzymes. We used one of these enzymes, Not I, to study the methylation state of the island.Not I will digest DNA to completion if the two CG sites in its recognition sequence are unmethylated, but will not cut DNA if either of the two CG sites are methylated. Briefly, 5 μg of genomic DNA was digested with 50 U of EcoRI and 100 U of Not I for 16 hours as specified by the manufacturer (New England Biolabs, Boston, MA). The DNA was then precipitated, electrophoresed on a 1% agarose gel, transferred to a nylon membrane, and hybridized with a³²P-labeled probe specific to the 5′ end of theABL1 gene (Fig 1). This probe was generated by PCR amplification from normal genomic DNA using primers (U) CAAACTTCCCTGATGGTGCCCTCTTG (L) TGACGTGTATTGTGCTCTTCCTATGT. The blots were washed using standard solutions and exposed to a phosphor screen for 2 to 4 days before imaging in a phosphorimager (Molecular Dynamics, Sunnyvale, CA). In this analysis, normal (unmethylated) ABL1 alleles are represented by a band at 5.4 kb, whereas methylated alleles, which fail to cut down with NotI, are represented by a band at 6.2 kb, which corresponds to the size of the EcoRI flank (Fig 1). For quantification of ABL1methylation, the relative density of the 5.4 kb was measured using the ImageQuant software (Molecular Dynamics) and expressed as a percentage of the density of all bands in each lane. All methylation studies were performed without prior knowledge of the patient’s characteristics, stage, or outcome.

DNA was restricted with EcoRI, Bgl II, or BamHI and run on Southern blots as described above. The blots were then probed with a BCR cDNA probe (a kind gift from Dr C. Griffith, Johns Hopkins University, Baltimore, MD). The proportion of the rearranged band (an indirect measure of the purity of the samples and the contamination with nonneoplastic cells) was determined by densitometry, as described above.

Demographic, clinical, and outcome data for all patients were collected and evaluated according to standard procedures at the MDACC. Patient and disease characteristics were entered into the CML data base upon referral and updated periodically for treatment effects, including response both hematologically and cytogenetically and survival. Response to therapy is as previously detailed.1 Briefly, a complete hematologic response (CHR) referred to complete normalization of peripheral counts and differential and disappearance of signs and symptoms of disease including palpable splenomegaly. A CHR was further categorized by the degree of Ph suppression: complete cytogenetic response, Ph⁺ cells 0%; partial cytogenetic response, Ph⁺ cells 1% to 34%; minor cytogenetic response, Ph⁺ cells 35% to 90%; and no cytogenetic response, Ph⁺ cells greater than 90%. A major cytogenetic response included both complete and partial response (Ph⁺ <35%). Differences among characteristics of different subsets were evaluated by the χ² or Wilcox tests. Survival was plotted by the Kaplan-Meier method,12 and differences among curves were analyzed using the log-rank test.

The ABL1 Pa promoter is embedded in a typical CpG island that is usually translocated intact to chromosome 22.4 We used quantitative Southern blot analysis of DNA restricted with methylation-sensitive restriction enzymes (as outlined in Fig 1) to measure ABL1 Pa methylation in various samples. No ABL1methylation was observed in any control blood (N = 10) or bone marrow (N = 3; data not shown) or multiple cases of acute myelogenous leukemia (AML; Fig 2). For this study, 118 samples from patients with CML seen and treated at MDACC were selected based on the availability of frozen leukemic cells. DNA was extracted from the leukemic blasts, and ABL1 methylation was quantitated by Southern blot analysis followed by densitometry of the phosphorimager scans. The DNA from 5 samples was degraded and unsuitable for this study, leaving 113 samples for analysis. Two samples were from the same patient (1 from Acc phase and 1 from Blast crisis). In 2 cases, the translocation breakpoint was within theEcoRI flank, and methylation of the translocated allele could not be ascertained. However, in both cases, the nontranslocated allele was completely unmethylated (Fig 3). In the remaining 111 samples, ABL1 methylation ranged from 0% to 62%, with a median of 27% (examples in Fig 2). One patient had Ph⁻ CML and had no evidence of ABL1methylation. This case was excluded from further analyses. To study interactions between ABL1 methylation and clinical variables, we analyzed the data both by treating ABL1 methylation as a continuous variable and by classifying patients into four categories based on ABL1 methylation: (1) negative (<3%, N = 23), (2) low (3% to <15%, N = 17), (3) moderate (15% to <25%, N = 12), and (4) high (>25%, N = 58).

There was no correlation between ABL1 methylation and age or gender. ABL1 methylation increased with advancing stages of CML (Table 3 and Fig 4): ABL1 methylation averaged 23% (N = 82) in ECP, 32% (N = 11) in late chronic phase (LCP), 35% (N = 12) in accelerated phase (Acc), and 42% (N = 5) in blast crisis (BC). Similarly, when treated as a categorical variable, high methylation was more frequent in advanced cases of CML: 37 of 82 (45%) in ECP, 8 of 11 (73%) in LCP, 9 of 12 (75%) in Acc, and 4 of 5 (80%) in BC (P < .01 for ECP v other). In ECP, ABL1 methylation was higher in the group with a white blood cell count (WBC) ≥50 × 10⁹/L (31/55 [56%] compared with 6/27 [27%], P < .01). However, there was no correlation between ABL1 methylation and other significant patient or disease characteristics, including risk group (Table 4). Among the 72 patients in ECP evaluable for risk classification, high methylation levels were found in 13 of 38 with good-risk, 14 of 24 with intermediate-risk, and 5 of 10 with poor-risk disease (34% v 58% v 50%,P = .16). There was no correlation between ABL1methylation and the percentage of Ph⁺ cells at diagnosis (99.2%, 96.7%, 100%, and 99.9% in the negative, low, moderate, and high categories, respectively).

Variable levels of ABL1 methylation in CML, especially in ECP, could potentially be due to variable contamination with nonneoplastic cells in the samples or could be related to the maturation state of the cells studied (with undifferentiated cells having higher levels of methylation). To address the first issue, we determined the relative contamination of the samples with nonneoplastic cells using Southern blots probed with a BCR probe. In this analysis, the relative proportion of the rearranged band (which reflects the purity of the samples) can be estimated by densitometry. We performed this analysis on 16 randomly selected ECP cases, 8 with negative ABL1methylation and 8 with high levels of methylation. To maximize the detection of rearranged alleles, each sample was digested with three different restriction enzymes, BamHI, EcoRI, andBgl II. All 16 cases had detectable rearranged bands using one or more of the three enzymes (13 with BamHI, 11 withBgl II, and 10 with EcoRI). The proportion of rearranged alleles was similar in the ABL1 methylation-negative versus methylation-high groups regardless of the restriction enzyme used: 34% versus 38% using EcoRI (P = .56), 55% versus 47% using Bgl II (P = .48), and 51% versus 47% (P = .72) using EcoRI (examples in Fig 5). The variations in proportion of rearranged alleles using different enzymes are due to the fact that the nonrearranged alleles are of different sizes and transfer/hybridize more or less efficiently in the Southern blot analysis. In particular, with BamHI, the nonrearranged allele is usually smaller in size than the rearranged allele and hybridizes better to the probe. The reverse is true with Bgl II. By averaging the values of all three enzymes, the rearranged alleles represent 47% of the ABL1-negative methylation cases and 44% of the ABL1 high methylation cases. Thus, contamination with nonneoplastic tissues (which appears to be very minimal in these samples) cannot explain the observed differences in methylation status.

To determine whether differences in ABL1 methylation in ECP could be attributed to the maturation state of the cells studied, we compared the bone marrow differential counts in the four methylation categories. The number of blasts was identical in all four categories (1%). Similarly, the number of promyelocytes, myelocytes, metamyelocytes, and polymorphonuclear cells (PMNs) did not vary according to ABL1 methylation. For example, PMNs represented 28%, 28%, 29%, and 26% of the bone marrow cells in the methylation groups negative, low, moderate, and high, respectively. These data also suggest that the high levels of ABL1methylation in advanced stages of CML are not due simply to an expansion of the blast component, because many ECP cases had levels of methylation comparable to accelerated or blast crisis phases of the disease.

We next analyzed the potential impact of ABL1 methylation on outcome measures in patients with ECP CML. Of the 76 evaluable patients in this category, 33 (43%) achieved a major cytogenetic response, 30 (38%) had a minor cytogenetic response or CHR with no cytogenetic improvement, and 13 (17%) did not respond to initial therapy. There was no correlation between ABL1 methylation (analyzed as a categorical variable) and response (Table5): a major response with IFN-based therapy was observed in 14 of 34 patients with high methylation compared with 19 of 42 among the others (41% v 45%; P value not significant). When analyzed as a continuous variable, ABL1 methylation averaged 23% in the group that had a major response, 22% in the group with a minor response, and 27% in the nonresponders (P value not significant).

At a median follow-up of 7 years, the median survival of the 82 patients in ECP was 63 months, and 56% have died. Using ABL1methylation as a continuous variable, there was no correlation with survival. When using ABL1 as a categorical variable, there appeared to be a trend towards improved survival after 5 years among patients in the two low methylation categories when compared with patients in the high methylation categories (Fig 6), but this difference did not reach statistical significance. In the subgroup of 19 patients who achieved a complete response to initial therapy, there also appeared to be a trend towards improved survival among patients in the two low methylation categories when compared with patients in the high methylation categories (1/9 [11%] relapses v 5/10 [50%] relapses), but this difference did not reach statistical significance at the current median follow-up.

Methylation of the ABL1 promoter Pa appears to be an early event in CML. In this study, we found such methylation in more than half of patients in ECP, and more frequent methylation was observed in advanced stages of CML. These results confirm earlier reports ofABL1 methylation in CML, although we find significantly higher rates in ECP than previously reported.5,11 This could be related to differences in patient characteristics or to the technique used to measure ABL1 methylation. The fact that ABL1methylation was present in a significant number of patients early in the course of CML raised the issue of whether it can be useful as a prognostic marker in this disease or whether it could be modulated favorably by therapeutic interventions (eg, IFN-α or decitabine13).

In our study, no correlation was found between ABL1 methylation and most disease characteristics, risk group, or response to IFN-α therapy. This suggests that ABL1 methylation in ECP does not define a unique subgroup of patients and may simply be a stochastic event that follows the 9:22 translocation. ABL1 methylation was also not a major prognostic factor for survival in ECP CML, within the limitations of the number of patients studied. Among patients who achieved a complete response after initial therapy, there was a trend towards improved outcome with low levels of methylation. A possible explanation for this is that ABL1 methylation is simply a molecular clock reflecting the time since formation of the Philadelphia chromosome. Thus, patients in clinical ECP with high levels ofABL1 methylation may have had undetected subclinical disease for several years, which then results in a higher chance of disease progression and a lower overall survival. Alternatively, it is possible that ABL1 methylation itself results in a more aggressive disease, as discussed below. Nevertheless, it is clear that factors other than ABL1 methylation determine the likelihood of response to initial therapy in ECP CML, which explains the lack of impact of this molecular marker on overall survival in this disease. Whether ABL1 methylation could be useful in predicting relapse among those patients who achieve a complete remission to IFN-α therapy deserves further investigation.

The mechanism and biological significance of ABL1 methylation in CML remain to be defined. As previously indicated by studies in cell lines,5 methylation appears to be limited to the rearrangedABL1 allele because (1) in 2 cases in which the chromosome 9 breakpoint was within our flanking cut, the normal (unrearranged)ABL1 allele was unmethylated; and (2) there was little evidence of methylation that significantly exceeds 50% and never any complete methylation at this locus, even in blast crisis CML. Thus, unlike other hypermethylation events in cancer, ABL1 methylation appears to be triggered exclusively by the 9:22 translocation. Further supportive evidence is that ABL1 methylation is not observed in malignancies that lack the 9:22 translocation, such as AML or solid tumors. One possible explanation for ABL1 methylation in CML then, is that it is a passive event that follows chromatin restructuring induced by the chromosomal translocation14and that it does not provide CML cells with a particular growth advantage. Studies of viral integration sites have suggested that methylation after chromatin restructuring may be a progressive event that evolves with time.15 By analogy, ABL1methylation may then merely be a molecular clock timing the occurrence of the 9:22 translocation, with more cells and alleles involved the longer the disease is active. However, it is also possible thatABL1 methylation is contributing to the pathogenesis of CML, perhaps by affecting the expression of the BCR/ABL transcript. Thus, it is conceivable that the translocated ABL1 promoter competes with the native BCR promoter for transcription factors and/or enhancers, as has been described at other loci.16 17 If this is the case, then hypermethylation and silencing of the translocated ABL1 promoter may be required for efficient expression of the BCR/ABL transcript and may then be selected for during CML progression.

In summary, our study confirms that ABL1 methylation is a frequent and early event in CML, but we found no prognostic impact for this molecular marker within the limitations of our studies.ABL1 methylation could still serve as a useful marker of disease in CML and, in patients who have unmethylated cells at diagnosis, ABL1 methylation may be used to monitor for disease progression. Because ECP-CML can be a slowly evolving disease, further studies should also address the long-term (>7 years) impact ofABL1 methylation on survival in this group of patients.

The authors thank Dr Lynne Rosenblum-Voss and Dr Constance Griffith for the BCR cDNA probe and Mutsumi Ohe-Toyota for excellent technical assistance.