Introduction
Myelodysplastic syndrome (MDS) with isolated deletion of chromosome 5q (MDS del5q) is a biologically and phenotypically distinct subtype of MDS. Clinically, it is often characterized by severe macrocytic anemia and normal, or elevated, platelet counts. Patients with MDS del5q show a good overall survival and lower risk of evolution to acute myeloid leukemia (AML) compared to most MDS subtypes [
1]. Furthermore, several large studies demonstrated an excellent response to the immunomodulatory drug lenalidomide for patients with MDS del5q [
2‐
4].
The German MDS study group conducted a prospective, open-label, non-randomized, single-arm, multicenter, phase II trial to evaluate the safety of lenalidomide and identify patterns of disease progression [
5]. Due to the strict inclusion criteria and rigid screening process, the study population was very homogenous. As expected, two-thirds of patients achieved sustained transfusion independence. However, progression to AML was observed in 15% of patients. This was not predictable by any clinical marker and raises the question of feasibility for potential combination therapies in addition to lenalidomide alone [
5].
The mechanism of action of lenalidomide in MDS del5q has just recently been better understood [
6]. As MDS del5q is hypothesized to be a disease of haploinsufficiency [
7,
8], the specific sensitivity to lenalidomide is partially explained through its suppression of haplodeficient proteins by a cereblon-dependent increase of ubiquitination and degradation [
9]. Additionally, lenalidomide has direct antiproliferative effects by inhibiting CDC25C and PP2A phosphatase levels which leads to G2/M-phase cell cycle arrest. It also induces T-cell and NK-cell proliferation and cytotoxic activity, resulting in immunomodulation [
10‐
12].
Altered methylation of the genome is a known pathogenetic mechanism of MDS and has been shown to contribute to progression to AML [
13‐
15]. Hypomethylating agents (HMA) are therefore widely used in the therapy of high-risk MDS and AML with myelodysplasia-related changes. HMAs are potential candidates for combination with lenalidomide in MDS del5q and are undergoing clinical evaluation. Whether lenalidomide itself has an effect on the methylation patterns of MDS patients is unknown. One recent study in multiple myeloma cell lines showed that DNA hypermethylation might be a mechanism of resistance to lenalidomide treatment [
16]. This prompted us to evaluate the methylation status of patients with MDS del5q before and after uniform treatment with lenalidomide. We sought to answer the questions whether lenalidomide treatment alters methylation patterns in MDS, and whether methylation status can predict response to lenalidomide treatment and outcome of patients with MDS del5q.
Methods
Patients
Fifty-one patients who were diagnosed and uniformly treated within a German, multicenter, single-arm, open-label, phase II trial investigating the safety of lenalidomide in patients with MDS del5q were included in this study (MDS-LE-MON-5 study, EudraCT number: 2008-001866-10,
www.clinicaltrials.gov: #NCT01081431). In addition to informed consent for the study according to the Declaration of Helsinki, patients had also consented to high-throughput molecular analyses. As specified by the inclusion criteria, all patients were classified as low to intermediate-1 risk according to IPSS with a bone marrow blast cell count below 5%. The cohort consisted of 9 male and 42 female patients with a median age of 69 years (range 40–87 years) at time of enrollment. Details on the study protocol and results have been previously published [
5]. In short, all patients uniformly received lenalidomide at a dose of 10 mg daily for 21 days of every 28-day cycle. Patients received a median of 13 cycles (range: 1–49 cycles). In 20 patients (39%), treatment had to be ended due to toxicity or incompliance before the completion of 6 cycles.
Additionally, for 17 of the 51 patients, we obtained follow-up bone marrow samples 4–6 months after the start of lenalidomide treatment.
HumanMethylation450 bead array sample and data processing
Mononuclear cells from patients’ bone marrow were isolated using Ficoll density gradient centrifugation and DNA was extracted using the Allprep Kit (Qiagen, Hilden, Germany). Following bisulfite conversion, genome-wide DNA methylation analysis was performed using the HumanMethylation450 BeadChip (Illumina, San Diego, USA) according to the manufacturer’s protocol.
Raw data was processed using the minfi R package [
17] as previously described [
18]. In short, the functions preprocessNoob, maptoGenome, and ratioConvert were utilized to produce beta-values and
M-values. Probes that mapped either to regions with known germline polymorphisms (Illumina supplementary SNP list v.1.2 update table v.3), to multiple genomic loci [
19], or to the sex chromosomes were filtered out. This left 295,926 probes for primary analysis.
Hierarchical clustering for sample classification
For the comparison of 17 paired patient samples (pre-lenalidomide and 4–6 months post-treatment), beta-values were hierarchically clustered in both directions (samples and probes) using an unsupervised approach based on Ward’s minimum distance method. Due to the uniformity of the cohorts’ methylation patterns, probes with standard deviations (SD) greater than 0.20 across samples were chosen for clustering. This resulted in 934 probes. For the analysis of all patients at the start of the study, unsupervised hierarchical clustering was repeated using the whole cohort of 51 pre-treatment samples. Here, a cutoff of SD of greater than 0.20 resulted in 1629 probes for clustering.
Pathway analyses
Probes within promoter regions were associated with the corresponding genes based on the Genecode v19 gene annotations [
18]. The most significantly differentially methylated probes between patients clustering in different clusters were identified by multiple
t-tests on the methylation
M-values. A false discovery rate approach with a
q-value set to <0.05 was used following the two-stage step-up method by Benjamini et al. [
20]. Pathway analysis was performed on the genes associated with the identified probes. KEGG analysis was performed using KEGG Mapper (
http://www.genome.jp/kegg/mapper.html). Gene ontology analysis and chromosomal mapping were performed using the GATHER tool (
http://changlab.uth.tmc.edu/gather/).
Statistical analysis
Differences in clinical features between patients clustering in different clusters were tested for significance using the chi-squared test for categorical variables and the Mann-Whitney U test for continuous variables. Overall survival (OS) was defined as time from start of lenalidomide treatment to death from any cause. Progression-free survival (PFS) was defined as time to progression or death from any cause. Both were estimated by the Kaplan-Meier method using a log-rank test for significance. Median follow-up time was estimated using the reverse Kaplan-Meier method. The significance level for all tests was a p-value <0.05. All calculations were performed using Microsoft Excel (v16.16.3), GraphPad Prism software (v8.0), and R (v4.0.2).
Discussion
We present a comprehensive genome-wide DNA methylation analysis on 51 MDS del5q patients. This represents the largest cohort of MDS del5q patients analyzed for methylation so far. In an earlier analysis by Zhou et al. [
15] on 20 MDS patients, only one patient with del5q was included and, interestingly, this patient clustered with the healthy controls rather than the other MDS subtypes. In a larger, more recent study, Reilly et al. [
21] analyzed 10 patients with del5q in their cohort of 141 MDS patients and found genes in the respective cluster were mostly hypermethylated; especially, the genes
WT1,
NBEA (
Neurobeachin), and
AP1M2 (
Adaptor Related Protein Complex 1 Subunit Mu 2) were significantly differentially methylated. All three genes were also found to be significantly differentially methylated in our cohort. Of note,
WT1 hypermethylation was associated with inferior survival in their study.
The main focus of the present study was the comparison of global methylation prior to and after treatment with lenalidomide. For this, we were able to obtain paired samples from 17 patients. As expected in this very homogenous patient population, we found a very homogenous methylation pattern across patients before the start of therapy. Comparing pre- and post-treatment samples, we found no relevant changes in methylation. This observation was independent of whether patients had achieved a cytogenetic remission. On one hand, this highlights the consistency of our sample and data processing. On the other hand, it shows that methylation patterns seemed to be mostly independent of the size of the del5q clone and lenalidomide treatment had little effect on the genome-wide methylation status of the patients’ bone marrow cells in this study. Lenalidomide does not seem to exert its observed effect by inducing DNA methylation changes.
Despite the fact that the study population of the LE-MON-5 study was very homogenous due to the rigid screening process, there was still one third of patients who did not respond to lenalidomide treatment. Therefore, we next wanted to explore whether differences in the baseline methylation patterns between patients could help predict the response to lenalidomide treatment. Unsupervised hierarchical clustering showed a clear distinction of patients into two main clusters. Even though the percentage of responders in cluster B was higher than in cluster A, this difference was not significant and the methylation status not predictive of response. We additionally performed supervised clustering between responders and non-responders to lenalidomide treatment. The resulting heatmap showed no differences in methylation patterns. This was further confirmed by multiple
t-testing on the sample
M-values, which resulted in zero significantly differently methylated CpGs between responders and non-responders (data not shown). In contrast, there was a clear trend towards inferior survival for patients in cluster A. One potential explanation could be the higher number of patients harboring a mutation in
TP53, which has previously been associated with inferior outcome in MDS del5q [
22]. However, even after excluding patients with
TP53 mutations from the analysis, the remaining patients still clustered into the same two clusters and there was still a trend towards inferior survival for patients in cluster A independent of
TP53 mutations. Only half of the deaths in cluster A were due to progressive disease (only one of them in a
TP53 mutant patient; the other three deaths were due to cardiac disease or unknown reasons) and PFS was in fact not significantly different between the two clusters.
To evaluate whether the methylation differences between the clusters were influenced by the significantly higher number of male patients in cluster A, we compared our data to previously published data on the influence of gender on the methylation of autosomes [
23]. In contrast to the reported finding that male gender is associated with less methylation, cluster A was predominantly composed of hypermethylated CpGs. Additionally, none of the genes reported to be significantly altered by gender and only a very small fraction (around 5%) of the genes considered as potentially altered by gender in the study by Liu et al. [
23] were among our most differentially methylated CpGs. Therefore, we conclude that the methylation differences observed between the clusters cannot be simply explained by gender distribution.
Analyzing the significantly differentially methylated genes leading to the clusters, we found that the majority of genes were hypermethylated in cluster A compared to cluster B. Among those genes, there were some interesting candidates whose aberrant methylation had been associated with inferior outcome in MDS in prior studies, e.g.,
WT1 [
21],
DLX6 (
Distal-Less Homeobox 6) [
24], members of the
PI3K-Akt-mTOR and
MAPK pathways [
14] and, importantly, members of the
Wnt signaling pathway. There have been several reports of a role of
Wnt signaling in the pathogenesis of MDS del5q and its potential as a therapeutic target [
25,
26]. Initially, it was observed by Liu et al. [
27] that
CTNNA1 (
Catenin Alpha 1), a tumor suppressor gene encoded on chromosome 5q31, was expressed at lower levels in MDS del5q due to suppressed expression of the second copy by methylation of the gene promotor. Since then, several studies have reported aberrant methylation of
Wnt family members in MDS including del5q underlining the potentially important role of the regulation of this pathway [
14,
28‐
30].
In summary, utilizing a homogenous, uniformly treated cohort of patients with MDS del5q, we could show that lenalidomide treatment does not have a relevant impact on genome-wide DNA methylation in MDS del5q. Still, before start of treatment, methylation analysis was able to identify a distinct subgroup of patients (27%) with a trend towards inferior overall survival but not inferior progression-free survival. This group showed hypermethylation in a number of interesting cancer-associated genes. Whether this translates into a potential benefit through the use of HMAs is a question this study cannot answer and that needs to be further evaluated in lenalidomide + HMA combination studies. Independently, our study provides additional indications towards the potential evaluation of Wnt signaling as a therapeutic target in MDS del5q.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.