Quantitative analyses of DAPK1 methylation in AML and MDS

Abstract

Aberrant DNA methylation and concomitant transcriptional silencing of death-associated protein kinase 1 (DAPK1) have been demonstrated to be key pathogenic events in chronic lymphocytic leukemia (CLL). In acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS), however, the presence of elevated DNA methylation levels has been a matter of continued controversy. Several studies demonstrated highly variable frequencies of DAPK1 promoter methylation by the use of methylation-specific PCR (MSP). By quantitative high resolution assessment, we demonstrate that aberrant DNA methylation is an extremely rare event in this region. We observed elevated levels just in one out of 246 (0.4%) AML patients, all 42 MDS patients were unmethylated. In conclusion, we present a refined DAPK1 methylation analysis in a large representative patient cohort of AML and MDS patients proofing almost complete absence of elevated DNA methylation. Our results highlight the importance of quantitative measurements particularly for translational research questions on primary patient specimens.

INTRODUCTION

DNA hypermethylation is a frequent event in pathogenesis of malignant diseases and associated with transcriptional gene silencing¹. Particularly in hematopoietic malignancies, genome-wide screening approaches and regional candidate assays have revealed numerous genes that are affected by aberrant DNA methylation^2–4. Hypermethylation and concomitant transcriptional silencing of the tumor suppressor gene death-associated protein kinase 1 (DAPK1) was demonstrated in virtually all patients with chronic lymphocytic leukemia (CLL) using quantitative high-resolution analysis of DNA methylation⁵. In acute myeloid leukemia (AML), DAPK1 was initially reported to be hypermethylated in up to 33% of patients⁶. Levels were correlated with clinical parameters and increased frequencies in therapy related AML/MDS were reported⁷. These findings, however, started a vivid dispute in literature about DAPK1 methylation status in acute myeloid neoplasia. Several subsequent studies reported variable degrees and partly contradictory levels of DAPK1 gene locus methylation in AML and myelodysplastic syndrome (MDS) (summarized in Table 1). The majority of these studies used methylation-specific PCR (MSP), a non-quantitative, very sensitive, and cost-effective assay developed in the 1990s⁸. High resolution analyses for the detection of quantitative DNA methylation have now emerged demonstrating their superiority, particularly with regard to biomarker development. The majority of quantitative assays are based on bisulfite conversion of genomic DNA⁹. Direct sequencing using pyrosequencing technique has become an accurate and reliable approach for analysis of shorter DNA stretches (<150 bp)¹⁰. Matrix-assisted laser desorption ionization – time of flight (MALDI-TOF) mass spectrometry provides another rather new flexible tool for quantitative DNA methylation assessment¹¹. Major advantages of this technology are high sensitivity and reproducibility of quantitative measurements, high resolution of target CpG sites and high-throughput capability. In this study, we present a refined quantitative and high resolution DNA methylation analysis of the DAPK1 5’ gene region in a large representative cohort of AML and MDS patients. Using DAPK1 as a biologically and potentially clinically relevant example we highlight the importance of accurate quantitative assessment of DNA methylation in translational studies particularly.

Table 1.

Summary of reports about DAPK1 methylation analyses in AML and MDS

entity	n	sample source	method	methylation frequency	reference
adult AML/pediatric AML	6/26	NA	MSP6	33/4%	6
de novo adult AML/de novo pediatric AML	45/4	NA	MSP6	42/0%	12
de novo adult AML/de novo pediatric AML	45/4	NA	bisulfite-DGGE13	2/0%	12
MDS/AML	34/160	BM MNC	MSP6	47/27.5%	14
MDS	73	BM	MSP6	43%	15
MDS	73	BM	qPCR-based assay16	<5%	15
AML	32/28	BM MNC/PB MNC	MSP6	3.3%	17
MDS	59	BM MNC	MSP6	62.7%	18
AML (of the elderly/secondary AML/refractory&relapsed AML)/MDS	12/18(29)	BM MNC/PB MNC	MSP6	28%	19
de novo MDS/de novo AML/secondary MN	105/208/72	BM MNC	MSP6	15.3/24.4/39%	7
MDS	78	BM MNC	MSP6	42.3%	20

Abbreviations: NA, not available; MN, myeloid neoplasia; BM MNC, bone marrow mononuclear cells; PB MNC, peripheral blood mononuclear cells, MSP, methylation-specific PCR; bisulfite-DGGE, bisulfite denaturing gradient gel electrophoresis.

MATERIAL AND METHODS

Patient samples and cell lines

All patients signed informed consent, and institutional review board approval in accordance with the Declaration of Helsinki from participating centers was present. Patient characteristics are given in Table 2. Bone marrow (BM) specimens from AML patients were enriched for myeloblasts by density gradient, whole BM was used from MDS patients and CD19+ B-cells from peripheral blood (PB) of CLL patients. As reference, buffy coat was obtained from PB of nine healthy donors and granulocytes from 16 healthy volunteers. CD34+ hematopoietic progenitor cells were MACS-isolated from BM of six healthy donors. AML cell lines were grown under standard conditions.

Table 2.

AML and MDS patient characteristics

		AML			MDS
total	n	246		n	42
age at diagnosis (years)	n	212		n	41
	median(range)	56.4	(18.3–82.0)	median(range)	71	(60–79)
gender	n	212		n	41
	f	104	49.1%	f	28	68.3%
	m	108	50.9%	m	13	31.7%
disease type	n	169		n (FAB subtype)	41
	de novo	151	89.3%	RA	4	9.7%
	secondary	18	10.7%	RARS	2	4.9%
	treatment-related	9	5.3%	RAEB	17	41.5%
karyotype	n	203		RAEBt	16	39.0%
	normal	66	32.5%	CMML	2	4.9%
	t(8;21)	4	2.0%	n (cytogenetic risk)	37
	inv(16)	10	4.9%	good	9	24.3%
	t(15;17)	2	1.0%	intermediate	7	18.9%
	t(11q23)	7	3.4%	poor	21	56.8%
	inv(3),t(3;3)	4	2.0%	n (IPSS risk)	41
	del5(q)/del(7(q)	23	11.3%	low	0	0.0%
	+8	7	3.4%	intermediate 1	3	7.3%
	complex	60	29.6%	intermediate 2	16	39.0%
	other	20	9.9%	high	22	53.7%
BM blasts (%)	n	195		n	41
	median(range)	77	(20–100)	median(range)	11	(0–30)
leukocyte count (Giga/l)	n	206
	median(range)	18.4	(0.5–243.6)

Abbreviations: f, female; m, male; FAB, French-American-British classification; RA, refractory anaemia; RARS, refractory anaemia with ringed sideroblasts; RAEB, refractory anaemia with excess of blasts; RAEBt, refractory anaemia with excess of blasts in transformation, CMML, chronic myelomonocytic leukaemia; IPSS, International Prognostic Scoring System.

Quantitative DNA methylation analyses

One µg of genomic DNA was isolated with the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) and consecutively sodium bisulfite-modified using the EZ DNA Methylation Kit (Zymo Research, Irvine, CA, USA) according to the manufacturer’s instructions. DNA methylation was quantitatively assessed at single CpG units (consisting of one or more CpG dinucleotides) using the MassCleave assay (Sequenom, San Diego, CA, USA)¹¹. Briefly, target regions (MassCleave A and B, Figure 1 and Supplemental Table 1) were PCR-amplified and in vitro transcribed. Primers were designed not to contain CpG dinucleotides in order to avoid a bias towards methylated or unmethylated alleles. Defined fragments were generated by base-specific RNase A cleavage and subsequently analyzed by matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry. Relative fragment amounts represent methylation levels at single CpG units. Alternatively, pyrosequencing was performed on PCR products of bisulfite-modified DNA using the semiautomatic Pyromark MD (Qiagen). In brief, the sequencing primer (Supplemental Table 1) was hybridized to the single-strand DNA of the PCR product. During synthesis of the complementary DNA strand, a luciferase-catalyzed lightpulse was generated upon incorporation of the new nucleotides. This photometrically measured lightpulse is in proportion to the number of incorporated nucleotides and thereby quantitatively displays the methylation of the investigated CpG dinucleotides.

Figure 1. Quantitative DNA methylation analysis of the DAPK1 gene 5’ region.

a) Schematic representation of the DAPK1 transcriptional start site (TSS, “+1”) and exon 1 (black box). A CpG island (CGI) around the transcriptional start site (TSS) is displayed as black bar. Two amplicons (A and B) covering the DAPK1 promoter and exon 1 region were quantitatively analyzed by the MassCleave assay (−267 to +336 bases relative to TSS). The region analyzed by pyrosequencing is labeled PyroSeq (−265 to −153 bases relative to TSS). The amplicon generated by the methylation-specific PCR⁶ is indicated by the label MSP.

b) Quantitative DNA methylation results are displayed as heatmap, columns represent single CpG units, rows represent separate samples. The top panel (AML) in red shows 204 AML samples, 31 healthy controls (N) are displayed in different shades of green (dark green, CD34+ selected hematopoietic progenitor cells [CD34+]; medium green, buffy coat cells [BC]; bright green, peripheral blood granulocytes [GRAN]). Nine AML cell lines of different maturation stages (HL-60, OCI-AML3, KG-1a, Kasumi-1, NB-4, U-937, HEL, MEG-01, and K-562) are shown in black (CL) and 61 CLL samples serving as methylated reference are depicted in blue (CLL). Bright green encodes for low methylation levels, dark blue for high methylation levels, grey indicates missing data.

c) Heatmap visualizing quantitative DNA methylation results as assessed at the region indicated in figure 1a by pyrosequencing. The top panel (AML) in red represents 42 AML samples, 42 MDS samples (MDS) are displayed in bright red. The color coding is identical to figure 1b.

d) The upper panel shows a plot of expected versus observed DNA methylation (^mC) for the amplicon A at the DAPK1 locus performed on a 6-point standard (0, 20, 40, 60, 80 and 100% in vitro methylated genomic DNA). Light grey dots represent all single CpG units of amplicon A for a given methylation level. The equation describes the best fitted curve to the DNA methylation standard. The lower panel represents a control plot of expected versus observed DNA methylation after mathematical correction. The equation represents the respective fitted curve which is identical with the ideal relation between expected and observed levels after correction.

e) Summarizing distribution of average DNA methylation (^mC) values across investigated regions as indicated in figure 1a. The upper panel shows MassCleave-derived quantitative DNA methylation data for AML sampls (AML), healthy controls (CD34+, BC and GRAN), AML cell lines (CL) and CLL samples (CLL). The lower panel depicts pyrosequencing derived average methylation levels for AML and MDS.

DNA methylation standard preparation and bias correction

For each amplicon a DNA methylation standard with defined ratios of in vitro methylated whole genome amplified DNA was included (0, 20, 40, 60, 80 and 100%). The standard was used to correct methylation values for PCR bias by robust non-linear regression models implemented in the R statistical computing environment.

RESULTS AND DISCUSSION

Intrigued by the findings of epigenetic silencing of DAPK1 in CLL and the previous reports about differential methylation in acute myeloid neoplasia (Table 1), we performed a systematic analysis of DAPK1 methylation in regions relevant for transcriptional control in AML and MDS. We conducted quantitative high-resolution DNA methylation analysis in 246 AML patient samples encompassing all major subgroups, 42 MDS patients (Table 2) and 31 healthy controls using the MassCleave assay and pyrosequencing. By the MassCleave assay, the DNA methylation status of 31 CpG dinucleotides located in 19 CpG units in a stretch of 557 bases (subdivided into amplicons A and B) around the DAPK1 transcriptional start site (TSS) was investigated (Figure 1a). Amplicon A covers the core promoter region of DAPK1 as previously determined⁵. Its 5’ part is identical with the region analyzed by pyrosequencing. Amplicon B harbors the region located in exon 1 used for MSP-based methylation detection. MassCleave-based DNA methylation analysis clearly indicated absence of DNA methylation in all tested CpG units of amplicon A (Figure 1b) in all except one AML patient sample. This patient showed an average amplicon DNA methylation of 44.4% and had a primary AML with an isolated translocation t(11;19). In amplicon B, DNA methylation levels were mostly below 10%, however, in some patients a pattern of variable low-level DNA methylation in 5 of the investigated CpG units could be observed. There was no overlap of these CpG units with the MSP primers by Katzenellenbogen et al.⁶. Comparing average DNA methylation levels across all investigated CpG units from position −242 to +315 (Figure 1e), no significant differences between different types of controls (selected CD34+ hematopoietic progenitors, buffy coat, granulocytes) and AML samples could be noted. The findings in AML are in pronounced contrast to a set of CLL samples that exhibited markedly increased DNA methylation over the entire region investigated in more than 80% of the samples (Figure 1 b, e). Different AML-related cell line models showed strong methylation differences at the same regions. NB-4, HL-60 and KG-1a had >80% DNA methylation, whereas Kasumi-1 had intermediate levels and U-937, HEL, OCI-AML3, MEG-01, and K-562 were unmethylated at the DAPK1 5’ region. Complementary analyses with pyrosequencing demonstrated concordance with our MassCleave based analysis. Each out of 42 AML samples investigated was virtually unmethylated (median region methylation 1.7%, range 0.2–8.6%). Absence of DNA methylation could also be noted in 42 MDS samples (median region methylation 1.6%, range 0.6–7.5%)(Figure 1 c, e). To exclude an over- or particular underestimation of DNA methylation at the DAPK1 5’ region we included a 6-point standard with the indicated ratios in the MassCleave analysis for both amplicons (shown for amplicon A in Figure 1d). A pronounced underrepresentation of methylated alleles for DNA methylation levels ranging from 40–80% could be observed. Curves were fitted to the 6-point standard measurements and a best fitting equation was determined using robust non-linear regression. This equation was then used to correct quantitative methylation data and to obtain linear DNA methylation measurements.

Taken together, we could demonstrate in large patient cohorts that encompass the major (cytogenetic) subgroups that DAPK1 methylation in the 5’ regulatory region is a very rare and most likely biologically not significant event in primary myeloblasts of AML and BM cells of MDS patients. The false positive detection of methylation by MSP in previous reports could be attributable to several different factors. Firstly, MSP is highly sensitive and can pick up even small fractions of methylated alleles. This is of considerable interest particularly for heterogeneous clinical samples where even small contaminating cell populations can pretend presence of DNA methylation. Quantitative measurements, however, allow to estimate the proportion of methylated alleles and thereby open the possibility to control for tissue–specific methylation biases. Secondly, MSP-derived methylation results show a pronounced variability between different settings and are strongly dependent on sample and PCR conditions. Furthermore, high resolution analyses of larger regions are needed to generate a comprehensive view on functionally important regions that are not influenced by punctual fluctuations of DNA methylation at single sites. In conclusion, thoroughly clarifying the absence of DNA methylation at the DAPK1 5’ region in AML and MDS, we demonstrate the importance of high-resolution, quantitative DNA methylation assessment in translational epigenetic research.