Research Article For reprint orders, please contact: [email protected]

Aberrations in DNA methylation are detectable during remission of acute lymphoblastic leukemia and predict patient outcome Aim: Aberrant DNA methylation patterns are a hallmark of cancer, although the extent to which they underlie cancer development is unknown. In this study, we aimed to determine whether acute lymphoblastic leukemia (ALL) patients in clinical remission retained abnormal DNA methylation patters and whether these were associated with patient outcome. Materials & methods: We investigated CpG island methylation of genes known to exhibit hypermethylation in leukemia using quantitative pyrosequencing analysis. Results: Although methylation levels were reduced in remission samples, they remained significantly higher than those seen in healthy controls. This retained methylation was not related to low levels of residual leukemia cells still present at remission. Methylation levels were also stable (or increased) during continuous remission and significantly correlated with long-term survival in adult ALL patients. Conclusion: This study determined that abnormalities in DNA methylation are retained during ALL remission and may represent a novel prognostic marker for adult ALL patients. Keywords:  CpG island • DNA methylation • leukemia • prognostic marker • remission

While great advances have been made in the survival of childhood acute lymphoblastic leukemia (ALL) patients in recent years [1] , ALL in the adult population remains difficult to treat successfully. In young adults (25–60 years old), five-year-survival rates of ALL patients are only 30–40% and long-term survival in patients over 60 years old is typically 10% or less [2] . Thus, novel treatments and improved methods for patient stratification, to help optimize current therapies, are urgently needed. Alterations in DNA methylation are a central feature of cancer development. These alterations include genome-wide hypomethylation, in association with dramatic increases in methylation of promoter-associated CpG islands [3] . Hypermethylation of CpG islands leads to loss of expression of the associated gene and many important tumor suppressors are inactivated by this mechanism [4] . Therapeutic strategies aimed at reversing tumor-specific alterations in methylation are

10.2217/EPI.14.78 © 2015 Future Medicine Ltd

being actively explored [5] and inhibitors of DNA methylation are now used as standard therapy for patients with myelodysplastic syndrome  [6] . Alterations in DNA methylation are also of potential clinical use as prognostic markers [7] , due to their high tumor specificity and comparative ease of detection. For example, the methylation status of the DNA repair gene MGMT is used to predict response to therapy in glioblastoma patients [8] and our group has recently shown that methylationbased markers are superior to currently used molecular markers for prediction of outcome in chronic lymphocytic leukemia (CLL) [9] . While the extent to which alterations in DNA methylation are early events that drive the process of leukemia development or are secondary changes is unclear [10] ; recent studies suggest that alterations in DNA methylation are likely to be very early events in leukemia development. For example, a recent study of CLL found that the abnormal patterns of DNA methylation observed at diagnosis were

Epigenomics (2015) 7(1), 35–45

Sanne D van Otterdijk1, Jean Norden2, Anne M Dickinson2, Mark S Pearce3, Caroline L Relton4,5, John C Mathers6 & Gordon Strathdee*,1 Northern Institute for Cancer Research, Newcastle University, Newcastle Upon Tyne, UK 2 Institute of Cellular Medicine, Newcastle University, Newcastle Upon Tyne, UK 3 Institute of Health & Society, Newcastle University, Newcastle Upon Tyne, UK 4 Institute of Genetic Medicine, Newcastle University, Newcastle Upon Tyne, UK 5 MRC Integrative Epidemiology Unit, School of Social & Community Medicine, University of Bristol, Bristol, UK 6 Human Nutrition Research Centre, Institute for Ageing & Health, Campus for Ageing & Vitality, Newcastle University, Newcastle Upon Tyne, UK *Author for correspondence: Tel.: +44 0191 282 1327 gordon.strathdee@ ncl.ac.uk 1

part of

ISSN 1750-1911

35

Research Article  van Otterdijk, Norden, Dickinson et al. remarkably stable during disease progression, suggesting that CLL-specific aberrant DNA methylation patterns were fixed relatively early in the disease process [11] . Furthermore, in healthy volunteers, ageing is associated with alterations in methylation in DNA from peripheral blood that are reminiscent of those seen in leukemia [12] . In this study, we investigated DNA methylation changes during the disease course of ALL using matched diagnostic, remission and relapse samples from ALL patients. We observed that remission-specific patterns of altered CpG island methylation can readily be detected and that methylation levels in remission samples correlated significantly with survival of adult ALL patients. Patients & methods Sample collection

DNA was purified from bone marrow collected from 24 childhood ALL patients and 21 adult ALL patients (Table 1) . Samples were taken at diagnosis, remission and where applicable, relapse. All samples were collected with appropriate ethical approval. Childhood ALL samples were taken from patients between 1 and 12 years of age, with an average age of 4.2 years. Remission samples were taken at approximately 12 weeks and, where available, 2-years after initiation of treatment. The adult ALL samples were taken from patients between 17 and 65 years of age, with an average age of 38.2 years. 17/21 adult ALL patients subsequently relapsed, from which two were long-term survivors of the disease (> 5 years). Minimal residual disease (MRD) information was available for five of the adult samples and was defined using PCRbased detection of clonal re-arrangements of the IGHV or the T-cell receptor loci, as described previously [13] . Peripheral blood DNA was collected from healthy control individuals. These consisted of 98 adult control samples, with an age range from 17 to 50 years of age, with an average age of 39, used as a control group for the adult samples and cord blood DNA form 49 neonate samples, used as a control group for the childhood samples. All the controls were collected as part of the North Cumbria Community Genetics Project (NCCGP) [14] and The Newcastle Thousand Families Study [15] . DNA extractions were carried out using standard techniques as previously described [16] . DNA modification & PCR

200 ng of genomic DNA was modified with sodium bisulfite using the MethylampTM One-Step DNA Modification Kit (Epigentek, cat# P-1010–2) as per the manufacturer’s instructions. All samples were resuspended in 15 μl of TE (tris-EDTA) and 1 μl of this was used for subsequent PCR reactions. The samples

36

Epigenomics (2015) 7(1)

were amplified in 25 μl volumes containing 1× manufacturer’s buffer, 1 unit of FastStart taq polymerase (Roche, cat# 04738403001), 1–4 mM MgCl2, 10 mM dNTPs and 75 ng of each primer. PCR was performed with one cycle of 95°C for 6 min, 40 cycles of 95°C for 30 s, 57–63°C for 30 s and 72°C for 30 s, followed by one cycle of 72°C for 5 min. For each set of primers (Supplementary Table 6 see online at: www.futuremedicine.com/doi/full/10.2217/EPI.14.78), one of the forward or reverse primers included a 5’-biotin label to allow for subsequent analysis by pyrosequencing. Quantitative DNA methylation analysis using pyrosequencing

Pyrosequencing was performed with Pyromark Gold Q96 Reagents (cat# 972804) on a PyroMark Q96 MD pyrosequencer (Qiagen, Hilden, Germany) according to manufacturer’s recommendation. Pyrosequencing analyses were performed in duplicate. If the duplicates showed a difference < 2% of methylation, the average methylation level was used for further analyses. If the difference between repeats was > 2%, then a third repeat was performed and all three repeats averaged. All runs, which did not pass the pyrosequencer quality control (i.e., labeled as ‘pass’), or which did pass quality control but had peak heights below 100, were excluded and the analysis repeated. Primer design was performed using the manufacturer’s proprietary PyroMark software. For initial assessment of all assays, pyrosequencing was carried out on samples of known methylation status (5, 10, 15, 20, 50, 70 and 100% methylated), which were produced by diluting peripheral blood-derived DNA (from a young adult volunteer) into in vitro methylated DNA (Millipore, cat# S7821). Statistical analyses

Nonparametric Mann–Whitney U (Wilcoxon ranksum) tests were used to test for differences between groups, as methylation data were not normally distributed. Paired Wilcoxon-signed ranks tests were used to examine the methylation levels in the paired diagnostic, remission and relapse samples. Pearson’s correlation was used to examine correlations. All statistical tests were done using SPSS. Results Methylation levels are significantly increased during ALL remission compared with healthy controls

CpG island methylation levels were quantified by pyrosequencing at eight loci, seven of which are known to exhibit very frequent methylation changes in ALL, and at the imprinted IGF2 locus which is

future science group

Aberrations in DNA methylation during remission of acute lymphoblastic leukemia 

not believed to be involved in ALL and acted as a control gene. For the seven ALL-methylated loci, we have previously found them to be frequently methylated in both childhood and adult ALL [16–18] , [Unpublished Data, ssd van otterdijk, he gautrey, m hili & g strath-

and this was confirmed in the current study (Supplementary Table 1; see online at: www.futuremedicine.com/doi/full/10.2217/EPI.14.78). We have also determined that they exhibit two different patterns during the normal process of ageing; either with low levels of methylation that increase steadily across the life course (EPHA10, HAND2, HOXD4, TUSC3 and TWIST2) or intermediate levels that stay constant across the life course (HOXA4 and HOXA5) [17] . Pyrosequencing provides quantification of methylation at the individual CpG level of resolution, which enables the detection of even very small differences in DNA methylation levels. All loci were assessed in 45 sets of matched samples from ALL patients (Table 1), including diagnostic, remission and relapse samples (where applicable). As expected, methylation levels of the seven loci known to exhibit frequent methylation changes in ALL were very high in diagnostic samples, were reduced in remission samples and then returned to levels similar to diagnostic levels in samples from patients who relapsed. In contrast, the nonleukemia-associated gene IGF2 exhibited no significant reduction in methylation during disease course (Supplementary Figure 1) . However, for patients in remission, methylation levels were not equivalent to those seen in leukemia-free healthy controls. Methylation levels in remission samples from adult ALL patients remained elevated relative to agematched controls for all seven loci and were increased by about 70% for TWIST2, EPHA10, HAND2,

dee]

Research Article

HOXD4 and TUSC3 (Supplementary Table 2, examples in Figure 1A) . All five of these loci exhibit age-related changes in DNA methylation [17] ; however, there was neither significant correlation between patient age and remission methylation for any of the five genes individually, nor between patient age and the average across the five genes (r2 = 0.01; p = 0.62; ages listed in Supplementary Table 2). In childhood ALL remission samples, increases in methylation were less pronounced but were still statistically significant for all loci, except EPHA10 and HOXA5 (Supplementary Table 3, examples in Figure 1B) . These results demonstrate that even during remission increased DNA methylation is still detectable in ALL patients. Increased methylation levels in remission are not likely to reflect surviving leukemia cells

A potential explanation for the increased methylation seen in remission samples could be the presence of low levels of remaining leukemia cells. Clinical remission is defined as less than 5% blasts present in the bone marrow, equivalent to the levels seen in a healthy individual [19] . However, a low level of leukemia-derived blasts can still be present and these residual leukemic cells could influence the level of DNA methylation in the remission samples. However, multiple lines of evidence demonstrate that residual leukemia cells cannot explain the increased methylation that we detected in the remission samples. First, the extent of the increased methylation is too large to be explained by a low level (i.e., < 5%) of residual leukemic blasts. Indeed for several of the remission samples, with higher levels of methylation, the level of residual leukemia cells would have to be greater than 30% to account for the increased methylation (Supplementary Table 4), clearly well in excess of

Table 1. Characteristics of patient population. Clinical characteristics  

Childhood ALL patients (n = 24)

Adult ALL patients (n = 21)

Age in years (median)

< 1–12 (3)

17–65 (40)

Gender†

 

 

Male

13 male

14 male

Female

10 female

7 female

Immunophenotype†

 

 

B-cell

20

11

T-cell

3

8

Null

1

1

Survival

 

 

Died

3

15

Alive

21

6

Information on gender was unavailable for 1 childhood patient and immunophenotype data on 1 adult patient. ALL: Acute lymphoblastic leukemia. †

future science group

www.futuremedicine.com

37

Research Article  van Otterdijk, Norden, Dickinson et al.

A

B 40

Methylation levels (%)

Methylation levels (%)

40

30

20

10

30

20

10

0

0

Child Child Control ALL remission ALL remission TWIST2 TUSC3

Control

Adult Control Adult ALL remission ALL remission TWIST2 TUSC3

Control

D

C

40

TUSC3 methylation (%)

TWIST2 methylation (%)

40

30

20

10

0

30

20

10

0 Control

Survivor

Non-survivor

Patient outcome

Control

Survivor

Non-survivor

Patient outcome

Figure 1. Methylation levels in acute lymphoblastic leukemia patients in remission are increased compared with healthy controls and are associated with patient outcome. (A) A significant increase in both TWIST2 and TUSC3 methylation was observed for adult ALL remission samples compared with healthy age-matched controls (p