Abstract
Background. Aberrant DNA methylation is an important mechanism by which the normal patterns of microRNA expression are disrupted in human cancers including B-cell precursor acute lymphoblastic leukemia (BCP ALL), the most common pediatric malignancy.
Objectives. To characterize the methylation profile landscape of microRNA genes in BCP ALL patients.
Materials and methods. We employed Infinium® MethylationEPIC BeadChip Arrays to measure the methylation of microRNA genes from bone marrow samples of children with BCP ALL (n = 38) and controls without neoplasms (n = 4).
Results. This analysis revealed differential methylation of the microRNA genes in the pediatric BCP ALL when compared to the control. A subcluster amongst BCP ALL patients with TCF3-PBX1 genetic subtype was also observed. No other differences were observed in association with age, gender or risk group. Several interesting leukemia-related phenotypes are enriched by the genes with hyper- and hypomethylated sites located in promoters as well as gene bodies. The top 3 miRNA genes, promoters of which were the most statistically significantly hypermethylated in BCP ALL were MIR1273G, MIR1304 and MIR663, and the top 3 hypomethylated were MIR4442, MIR155 and MIR3909.
Conclusions. In this study, a different microRNA genes methylation landscape was shown in pediatric BCP ALL compared to children without neoplasms. A visible subcluster among BCP ALL samples consisted of individuals with TCF3-PBX1 genetic subtype. No other differences were observed in association with age, gender or risk group. Several interesting leukemia-connected phenotypes were found, associated with genes with hyper- and hypomethylated sites located on promoters as well as gene bodies.
Key words: children, microRNA, methylation, BCP ALL
Background
B-cell precursor acute lymphoblastic leukemia (BCP ALL) is the most common malignancy in children.1 Understanding the molecular and genetic pathways that affect the development and clinical course of BCP ALL is a key to improving the treatment outcomes and is, therefore, an important focus of current research into BCP ALL.
Normal hematopoietic cell development is highly controlled epigenetic regulation of genes via DNA methylation, the chemical modification of histones, and the expression of noncoding RNAs. Each of these epigenetic factors can become dysregulated during leukemic transformation. The DNA methylation is by far the most well-characterized epigenetic modification and is involved in the regulation of gene expression, the maintenance of genome stability, and cellular differentiation.2 The methylation of cytosine residues in CpG dinucleotides plays a pivotal role in the establishment of cellular identity by influencing gene expression.3
MicroRNAs (miRNAs) are short, noncoding RNA molecules that regulate gene expression by forming complexes with their mRNA counterparts in order to cause translational repression, either by mRNA degradation or cleavage by deadenylation.4, 5 Furthermore, miRNAs are key regulators of hematopoiesis and are also involved in leukemogenesis.6, 7 To date, at least 32 dysregulated miRNAs are known to be associated with ALL prognosis.8 Interestingly, aberrant DNA methylation seems to be a major mechanism by which the normal patterns of miRNA expression are disrupted in human cancers,9 including ALL.10 Many tumor suppressor miRNAs appear to be downregulated by DNA hypermethylation, and various oncogenic miRNAs (onco-miRNAs) are known to be upregulated via DNA hypomethylation.9
We have previously shown significant differences in genomic methylation profiles in the bone marrow of BCP ALL and healthy control patients.11 Nevertheless, not enough attention was given to the methylation of miRNA genes, which is an important factor affecting oncogenic processes.
Objectives
In this study, we attempted to perform the analysis of methylation profiles of miRNA genes in BCP ALL patients in order to enhance the results of the previous analysis and to shed some light on a potential role of epigenetic regulation of miRNA expression in pediatric leukemia. We have used genome-wide methylation data obtained in our previous work11 and performed an in-depth analysis of the methylation differences in functional elements of miRNA genes in healthy and leukemic bone marrow samples.
Materials and methods
Patients and samples
The approval of the ethics committee for this study was obtained from the Institutional Review Board of the Medical University of Lodz, Poland (approval No. RNN/226/11/KE). Informed consent was obtained from the parents/legal guardians of all participating children. Forty-two samples of bone marrow were obtained in 2015–2016 from 38 patients (male/female 21/17; median age 5.0 years, age range 1.5–17.0 years) with pediatric BCP ALL at the time of diagnosis. The control samples of bone marrow were collected from children in whom other types of cancer and other genetic diseases had been previously excluded. The patients were stratified into prognostic groups according to the ALL IC-BFM 2009 protocol.12 This stratification is based on the initial clinical features including patient age, white blood cells count at time of diagnosis, presence of specific genetic aberrations, the response to steroids at day 8 of therapy, the cytomorphological response in bone marrow at day 15 and 33, and the minimal residual disease level at day 15. Finally, 5 patients were included into the high-risk group. Various genetic aberrations associated with ALL were detected among most of the patients. Most frequently, hyperdiploidy (>50 chromosomes) (13 patients) and t(12;21) with fusion ETV6-RUNX1 (7 patients) were revealed. Subsequently, t(1;19) with fusion TCF3-PBX1 (3 patients), hyper/hypotriploidy (3 patients) and IGH rearrangement (3 patients) were detected. Seven patients had other genetic aberrations, which were different than those mentioned above. In 2 cases, the normal karyotype was confirmed.
Samples and DNA methylation profiling
The DNA samples were analyzed with the Infinium® MethylationEPIC BeadChip approach (Illumina, San Diego, USA), according to the manufacturer’s protocol. It allowed for the analysis of 850,000 methylation sites per sample. These sites include those within known CpG islands and outside CpG islands, as well as non-CpG methylated sites identified in human stem cells and differentially methylated (DM) sites identified in tumors compared to normal samples and across several tissue types. The assay also includes probes for 7084 CpG sites associated with miRNA genes, out of which 4188 probes were retained for further analysis following quality filtering and extensive annotation analysis. These probes are associated with 1008 different miRNA genes, covering 52.5% of the current version of miRBase (v. 22).13
Data quality control and statistical analysis
Our previous study describes the steps taken for data quality control and the identification of DM sites.11 In brief, after initial normalization and removal of batch effects, the ratio of methylated (C) to unmethylated (T) DNA (also called the methylation beta-value (β)) was calculated for each CpG site. A β-value of 0 represents a completely unmethylated CpG site and a β-value of 1 represents a fully methylated CpG site. The differential methylation for individual probes between groups was calculated using the Chip Analysis Methylation Pipeline (ChAMP). The champ.DMP() function of ChAMP package pipeline14 was applied. It uses the limma package15 to calculate the p-value by a linear model. The DMP-determined t-test p-values were corrected for multiple testing using the Benjamini–Hochberg procedure.16 Adjusted p-values (adjPs) <0.05 were considered statistically significant.
The differential methylation analysis was performed with several discrete goals: 1) the identification of general differences between all BCP ALL patients and controls, and 2) the identification of DM sites between standard/intermediate and high-risk group of patients. Additional comparisons have been made within BCP ALL patients group to detect sites associated with confounding factors, such as age (≤6 years compared to >6 years) or gender (males compared to females). The differential methylation analysis was also performed between patients with different chromosomal aberrations characteristic of leukemia (ETV6-RUNX1 (n = 7), TCF3-PBX1 (n = 3), IGH (n = 3), hyperdiploidy (n = 13), and triploidy (n = 3)), using pairwise comparisons between specific cytogenetic subtype and all other leukemia patients with known cytogenetic status (n = 8). Cytogenetic diagnosis was performed as previously described.11
Functional gene annotation and target genes prediction
The miRNA genes associated with specific DM sites were analyzed in terms of their disease phenotypes annotated in the available databases. The phenotype enrichment analysis was performed using the WEB-based GEne SeT AnaLysis (WebGestalt) toolkit (www.webgestalt.org),17 exploiting information from the Human Phenotype Ontology (www.hpo.jax.org)18 and PharmGKB (www.pharmgkb.org)19 databases. The WebGestalt analysis was performed using all annotated human genes (genome option), and it was limited to identifying enriched phenotypes with the Fisher’s exact test false discovery rate (FDR)-adjP <0.01 and with at least 4 genes within phenotype categories. The prediction of miRNA target genes was done with the use of miRDB tool. Only genes with a prediction score ≥95 (implying the highest confidence of gene as a target) were analyzed.20
Results
Differences in miRNA genes
methylation profiles between
BCP ALL and control samples
General assay performance was evaluated based on control probes. The performance was satisfactory across all studied samples, with all samples passing filtering criteria implemented in the BeadArray Controls Reporter (Illumina) software. After initial filtering, β-values for 800,619 probes were retained and normalized. Subsequently, batch effects were identified and removed by evaluation of components of variation and by the singular value decomposition (SWD) method.21 Among all the analyzed probes, 4188 probes associated with 1008 different miRNA genes were identified and analyzed (data available on reasonable request).
Comparison of methylation levels between BCP ALL and control samples resulted in the detection of 578 DM probes. Out of 578 probes, 377 were annotated to promoters of 242 different miRNA genes. Promoter sites were defined as: TSS200 (0–200 nt upstream of transcription start site (TSS)) and TSS1500 (200–1500 nt upstream of TSS). The remaining 201 DM sites were annotated as located within the gene bodies of 74 different miRNAs.
The principal component analysis based on all DM probes reveals visible differences in the miRNA genes methylation profiles between BCP ALL and control samples, with greater variation observed among leukemic samples (Figure 1). Unsupervised hierarchical clustering (based on Euclidean distance and the DM probes β-values) showed the clear separation of the leukemic and non-leukemic methylation profiles into 2 distinct clusters (Figure 2). A visible subcluster among BCP ALL samples was created by 3 cases of TCF3-PBX1 genetic subtype (Figure 2), suggesting that it possesses a distinct miRNA genes methylation profile. No clustering of samples with age, gender or risk group was observed, suggesting a lower influence of these factors on global miRNA gene methylation profiles.
The initial analysis of differential methylation of CpG sites between BCP ALL and control samples showed a similar pattern of methylation differences for sites located in promoters and gene bodies of miRNA genes (Figure 3, Figure 4). However, taking into account that the methylation in these regions can have the opposite effect on transcription process, DM CpGs were analyzed separately, depending on their location within the functional element of a given gene.
Differential methylation of miRNA gene promoters in BCP ALL
Changes in methylation profile of miRNA gene promoters in leukemia were studied based on 377 DM (t-test FDR < 0.05) CpG loci between BCP ALL and ccontrols, located within TSS200 and TSS1500 of 242 different miRNA genes (data available on reasonable request). These sites were rarely associated with known CpG islands (17.8% of all) or shores (23.3%), and about half of them (54.9%) were hypermethylated in the BCP ALL group. When comparing the distribution of hyper- and hypomethylated sites across the genome, we found that hypermethylated sites are more commonly situated in CpG islands than hypomethylated (29.5% compared to 3.5%), whereas hypomethylated sites were more common in “open sea” (64.7% compared to 53.8%). The lowest average methylation level was observed for CpGs located within known CpG islands in both the control (17.7%) and BCP ALL samples (43.2%) (Table 1).
A high level of methylation was observed for CpGs located in sites distant from a CpG island (“open sea”; 65.8% in control samples) and was the highest for sites that were hypermethylated in BCP ALL (82.9%) (Table 2). The average absolute Δβ (average difference in methylation level) was similar for hyper- (0.197) and hypomethylated (0.213) sites.
The DM sites located in promoters were associated with 242 different miRNA genes. The phenotype enrichment analysis showed that the genes were enriched in several cancer-related disease phenotypes, including hematologic neoplasms, carcinoma, B-cell lymphoma, or leukemia itself (Table 3). Moreover, several interesting leukemia-connected phenotypes were found, when genes associated with hyper- and hypomethylated sites were analyzed separately (data available on reasonable request).
The top 3 miRNA genes whose promoters were the most statistically significantly hypermethylated in BCP ALL were MIR1273G, MIR1304 and MIR663, and the top 3 hypomethylated were: MIR4442, MIR155 and MIR3909. The analysis of the most probable target genes (top score from miRDB software) for the miRNAs being a product of those genes allowed for the detection of 44 different coding genes, among which we found e.g., pre-B-cell leukemia transcription factor 2 (PBX2) and metastasis associated 1 family member 2 (MTA2) genes (Table 4).
Differential methylation of miRNA gene bodies in BCP ALL
In total, 201 DM sites located in the gene bodies of 74 different miRNA genes between BCP ALL and control samples were detected (data available on reasonable request). As expected, the DM sites were mainly located outside known CpG islands (80.1%). No sites that were hypomethylated and located within known CpG islands were detected in BCP ALL. Most of the gene body-associated sites were characterized by a high level of methylation in both groups (45–85% on average in control samples). Only sites positioned within CpG islands were characterized by lower methylation level (15.1% in controls (CTR)) (Table 5).
Sites were classified according to their location in open sea, shore, shelf, and island of known CpG islands. The number and percentage of probes in each location is given. Additionally, the average methylation in both study groups and the difference in methylation level between them is provided.
Out of the gene body-associated sites DM between BCP ALL and control groups, 102 (50.7%) were hypermethylated in BCP ALL. The average difference in methylation level (average absolute Δβ) between BCP ALL and control samples was higher for hypo- (0.27) than hypermethylated (0.14) sites. The hypermethylated sites were associated with 51 different miRNA genes, out of which 3 most significantly DM were MIR1273H, MIR5096 and MIR5095. The sites hypomethylated in BCP ALL were connected with bodies of 45 different miRNA genes. Among the top 3 genes with hypomethylated bodies, we found MIR548Q, MIR3163 and MIR181A1HG (the host gene). The analysis of the top 3 hyper- and hypomethylated miRNAs target genes showed 483 potential targets (data available on reasonable request), among which we found few having annotation to leukemia phenotype in PharmGKB database (e.g. TRIM72, SPI1, MARVELD3, TOPORS, SSBP2, GAB2). Altogether, the miRNA genes with DM sites in gene body significantly enriched 6 disease phenotypes, including hematologic neoplasms, lymphoma, carcinoma, or viral infections (Table 6).
miRNA genes differentially methylated in specific leukemia genetic subtypes
To detect miRNA genes potentially associated with separate leukemia genetic subtypes (ETV6-RUNX1, TCF3-PBX1, IGH, hyperdiploidy, triploidy), patients with a specific subtype were compared against all the other patients with known cytogenetic status. This analysis allowed us to detect 21 DM sites in patients with ETV6-RUNX1, corresponding to promoters and gene bodies of 12 different miRNA genes (Table 7). Only 7 of these sites were located in promoters and were mainly (71.4%) hypomethylated with an average absolute Δβ of 0.27. Nearly 66.7% of DM sites in ETV6-RUNX1 cases were located in gene bodies and were predominantly associated with hypomethylation (64.2% of sites). For TCF3-PBX1 subtype, we found 56 DM sites associated with 37 miRNA genes (Table 7). Most of the sites were located in promoters (n = 38, 67.8%). Moreover, most of them (89.5% of sites) were hypomethylated in TCF3-PBX1 cases. The DM sites (between TCF3-PBX1 cases and other leukemia subtypes) located in gene bodies (n = 18) also showed high ∆β values (around 0.39) and were predominately (61.1%) hypomethylated. In patients with IGH rearrangements, only 3 probes (2 in promoters and 1 in gene body) associated with 3 different miRNA genes were hypomethylated with a low average Δβ of 0.18.
The occurrence of hyperploidy in BCP ALL patients was associated with differential methylation of 34 sites (25 different miRNA genes), the majority of which (61.7%) was located in gene bodies of 14 different genes. The sites located in promoters showed slightly higher absolute Δβ (0.27) than those located in gene bodies (0.21). Both promoter and gene body-associated sites predominantly showed hypermethylation in patients with hyperploidy, with 71.4% of promoter sites hypomethylated and all sites (100%) located in gene bodies hypomethylated in hyperploidy cases.
Only a single probe, associated with the body of MIR548H4 gene, was DM (hypermethylated) in samples obtained from patients with triploidy (Table 7).
Discussion
Aberrant expression of some miRNA genes may be a contributing factor in the oncogenesis of many cancers,22, 23 including acute leukemias.24, 25, 26, 27 Epigenetic modifications, such as methylation of miRNA genes, regulate the expression of genes.28 Therefore, altered miRNA gene methylation may be regarded as a causal factor for leukemogenesis and it can determine the clinical course of acute leukemia as well.
In this study, 377 sites located in the promoters of 242 different miRNA genes were DM in BCP ALL compared to control samples. Many of these miRNAs participate in cell cycle and differentiation control (Table 4). This research was focused on determining the methylation landscape of miRNA genes only; therefore, the expression level of the examined miRNA remains unknown. Despite the lack of evidence from this study to link the particular miRNA methylation level with its expression, the results obtained by others may indicate such a relationship.29, 30, 31, 32, 33, 34 Stumpel et al.30 identified 11 miRNAs that were downregulated in t(4;11)-positive infant ALL, as a consequence of CpG hypermethylation. Seven of those miRNAs were reactivated after the exposure to the demethylating agent, zebularine. In our study, 4 out of these 7 genes (MIR200B, MIR429, MIR10A and MIR432) were hypermethylated as in the previously mentioned study.30 Their impact on the development and clinical course of leukemia may be significant. For example, the zinc finger E-box binding homeobox 2 (ZEB2) gene is the best-known, validated target gene of the miR-200b/a-429 cluster.31 Homeobox A3 (HOXA3) gene has been described as a potential MIR10A targeted gene.32 For MIR432, more than 100-fold downregulated expression was observed in t(4;11)-positive infant ALL, as compared with normal bone marrow.30 The MIR432 is located within the large MIR127 cluster which is silenced in various malignancies by CpG island hypermethylation and aberrant histone modifications.33
Schotte et al.34 showed the reduced methylation level at CpG islands in the promoter regions of MIR196B, yet it was limited to MLL-rearranged BCP ALL cases. It corresponded to an upregulation of MIR196B, suggesting an epigenetic origin for its overexpression. This is in line with the results of our study, where MIR196B gene is hypomethylated in all BCP ALL cases.
The results of other studies27 revealed that the genes of at least 5 miRNAs (MIR326, -200c, -125B, -203, and -181A) have a significantly different expression in BCP ALL cases compared to healthy controls. According to our study, the gene promoters of 3 out of the abovementioned miRNA (MIR125B, -203 and -181A) are DM in BCP ALL patients. The comparison between their methylation status and their expression (according to the available references) was shown in Table 8. The expression of MIR125B gene, opposite to MIR203 and MIR181A, was higher in BCP ALL patients, although its promoter was hypermethylated.
Top score target genes for miRNA whose gene promoter sites are most statistically significantly hyper- and hypomethylated in BCP ALL are listed in Table 4. Some of these genes are known to be associated with carcinogenesis. For example, MIR1273 expression is increased in the pancreas of mouse model of pancreatic cancer.35 The MIR1304 is a tumor suppressor and HO-1 is its direct target in non-small cell lung cancer.36 The downregulation of MIR1304 is related to early stage breast cancer.37 The overexpression of MIR663 significantly suppressed the proliferation and invasion of glioblastoma cells in vitro as well as in vivo.38 The MIR663 may act as an oncogene in nasopharyngeal carcinoma.39 Interestingly, in our study on childhood BCP ALL, the promoter of MIR663 was hypermethylated in pediatric acute myeloid leukemia (AML), with significantly lower expression compared to normal bone marrow.40
The MIR4442 was amidst the genes with the top hypomethylated promoters. Although its role in leukemogenesis is unclear, its predicted targets include GTPase activating Rap/RanGAP domain-like 3 and zinc finger protein 765.41 The next gene with hypomethylated promoter is MIR155, which plays a complex role in AML.42 The levels of MIR155 significantly influence the set of genes involved in biologic processes related to antiapoptotic, proliferative, and inflammatory activities.43 The increased expression of MIR155 causes the downregulation of SPI1 and CEBPB, and consequently may block myeloid differentiation in AML.44 Moreover, the expression of HSA21-encoded MIR155 is altered in B cells of Down syndrome individuals and may play a role in Down syndrome-associated leukemia.45 There is also some evidence linking the significantly upregulated MIR155 expression level to the high levels of minimal residual disease and poor prognosis in ALL patients.46 According to the results from other study,47 MIR155 was upregulated in the HCV-4 associated ALL group. Hence, the increased MIR155 level may be related to acute leukemia development. Our results, where MIR155 promoter was hypomethylated, are consistent with this observation.
To sum up, there was a slight predominance of differentially hypermethylated (207/377) over hypomethylated (170/377) promoter-associated sites in miRNA genes in BCP ALL patients. However, the potential impact of miRNAs genes methylation level on their expression has to be established in future studies.
There were also 201 DM gene body associated sites with nearly equal (102:99) distribution between their hyper- and hypomethylated status. They corresponded to 74 different miRNA genes. The function of gene body methylation is not well understood. Typically, DNA promoter methylation is believed to be a marker of transcriptional repression. However, the DNA methylation within the gene body appears to serve a different function than DNA methylation within the promoter region. While there is some evidence that intragenic DNA methylation is related to transcriptional repression,48 the bulk of evidence suggests that it is associated with gene activation.49 Characteristics of DM in BCP ALL miRNA gene bodies-associated sites are presented in Table 6. Interestingly, a few of the 483 potential target genes of the most hyper- and hypomethylated miRNAs are annotated to leukemia phenotype.
Among different genetic subtypes, only 1 subtype (3 patients) with the translocation t(1;19) generating the TCF3-PBX1 fusion gene, had a distinct miRNA gene methylation profile. For this subtype there were 56 DM sites found, associated with 37 miRNA genes. Most of the sites were located in promoters and were hypomethylated. According to the previously published data, several miRNAs are downregulated or upregulated in TCF3/PBX1-positive ALL.50, 51, 52
Herein, we have shown that the miRNA genes of pediatric BCP ALL patients are DM compared to controls. This epigenetic dysregulation seems to play an important role in controlling miRNA expression, therefore affecting the clinical course of BCP ALL. To date, several miRNAs have been observed to have altered expression in patient cohorts compared to healthy individuals, while several studies have identified specific miRNAs that can be used as biomarkers to diagnose ALL, classify it into subgroups and predict the prognosis.27, 53, 54, 55 Moreover, hypermethylated genes can be targeted by hypomethylating agents, such as cytosine analogs, azacitidine or decitabine, which may open up new potential treatment options for this type of leukemia.56, 57
Limitations
In this study, we presented a methylation profile of genes for miRNAs in the bone marrow of BCP ALL and healthy subjects; however, due to major constraints, we were not able to determine the expression of DM miRNA genes. Thus, we had to refer to data regarding their expression level from other studies. The second limitation is using only one method for determining DNA methylation level. Additionally, for some of the analyzed miRNAs (i.e., the ones encoded in introns of mRNA genes) we were unable to determine if the observed methylation level of DNA is associated with the regulation of the miRNA or its host mRNA gene expression. However, this does not affect the main goal of this study, which was to select miRNA genes with the most altered methylation as targets for future studies. The last limitation is the small control group. This was due to the limited availability of the bone marrow samples from healthy children. The bone marrow aspiration is an invasive procedure, so performing it without clear medical indications would be highly unethical. Because of that, we were not able to obtain bone marrow samples from completely healthy children. On the other hand, only individuals with any cancer and genetic disease were ruled out, as well as with the normal bone marrow smears, were included in the control group.
Conclusions
In this study, a different were genes methylation landscape was shown in pediatric BCP ALL compared to children without neoplasms. A visible subcluster among BCP ALL samples consisted of individuals with TCF3-PBX1 genetic subtype. No other differences were observed in association with age, gender or risk group. Several interesting leukemia-connected phenotypes were found to be enriched in genes associated with hyper- and hypomethylated sites located on promoters as well as gene bodies.
