HOXA4 Gene Promoter Hypermethylation as an Epigenetic Mechanism Mediating Resistance to Imatinib Mesylate in Chronic Myeloid Leukemia Patients

Development of resistance to imatinib mesylate (IM) in chronic myeloid leukemia (CML) patients has emerged as a significant clinical problem. The observation that increased epigenetic silencing of potential tumor suppressor genes correlates with disease progression in some CML patients treated with IM suggests a relationship between epigenetic silencing and resistance development. We hypothesize that promoter hypermethylation of HOXA4 could be an epigenetic mechanism mediating IM resistance in CML patients. Thus a study was undertaken to investigate the promoter hypermethylation status of HOXA4 in CML patients on IM treatment and to determine its role in mediating resistance to IM. Genomic DNA was extracted from peripheral blood samples of 95 CML patients (38 good responders and 57 resistant) and 12 normal controls. All samples were bisulfite treated and analysed by methylation-specific high-resolution melt analysis. Compared to the good responders, the HOXA4 hypermethylation level was significantly higher (P = 0.002) in IM-resistant CML patients. On comparing the risk, HOXA4 hypermethylation was associated with a higher risk for IM resistance (OR 4.658; 95% CI, 1.673–12.971; P = 0.003). Thus, it is reasonable to suggest that promoter hypermethylation of HOXA4 gene could be an epigenetic mechanism mediating IM resistance in CML patients.


Introduction
Chronic myeloid leukemia (CML) is a myeloproliferative disorder that comprises 14% of all leukemias. e molecular pathogenesis of CML involves the clonal expansion of pluripotent haematopoietic stem cells containing the BCR-ABL fusion oncogene. BCR-ABL gene results from a reciprocal translocation between chromosome 9 and 22 to form the Philadelphia chromosome [1]. is BCR-ABL fusion gene codes for a p210 kD protein with increased tyrosine kinase activity. Imatinib mesylate (IM) or Glivec (NOVARTIS Pharma) is a selective molecular inhibitor of the BCR-ABL oncogene protein and permits long term disease control in about two thirds of chronic phase CML patients [2]. IM has dramatically improved the treatment of CML and is generally considered as frontline therapy for CML patients. Despite its striking efficacy, development of resistance in signi�cant proportion of CML patients on IM therapy has emerged as a major clinical problem affecting both patients and treating physicians.
Various mechanisms of resistance and suboptimal response to IM have been described, involving BCR-ABL1-dependent and BCR-ABL1-independent pathways [3,4]. BCR-ABL1-dependent mechanism usually involves point mutations in the tyrosine kinase domain (TKD) and ampli�cation of BCR-ABL gene, with mutations in the BCR-ABL tyrosine kinase domain being better characterized [5]. Our previous study on BCR-ABL TKD mutation analysis showed that BCR-ABL mutations accounted for IM resistance in only 21.7% of Malaysian CML patients on IM therapy (communicated separately; in Press). is indicated that BCR-ABL mutations are not the only cause for relapse and resistance. It is presumed that the mechanisms of IM resistance in CML patients who do not have TKD mutation might be mediated through BCR-ABL-independent pathways. However, the exact mechanism in BCR-ABLindependent pathway still remains unclear, despite several genetic and epigenetic mechanisms postulated to be involved in the BCR-ABL-independent pathway.
It is largely known that DNA in cancer cells is very unstable. Epigenetic silencing is a phenomenon whereby gene transcript maybe suppressed through DNA methylation. Gene expression can be strongly modi�ed through epigenetic alteration such as DNA hypo or hypermethylation. DNA methylation at cytosine residues in gene promoter CpG sequences is known to inhibit gene transcription, resulting in decreased protein expression. Genomic instability and DNA modi�cations certainly confer to the cancer cells, a higher capability of becoming resistant [6]. e human Homeobox (HOX) gene network encodes master regulators in haematopoiesis and DNA methylation has been implicated to have an important role in aberrant control of HOX gene expression [7]. Inappropriate expression of HOX gene has been implicated in development of hematopoietic malignancies. Methylation of a HOX gene, HOXA4 has been strongly associated with progression to blast crisis and poor response to treatment in other types of leukemia patients [7].
In CML, increased epigenetic silencing of potential tumor suppressor genes has been found to be correlated with disease progression in a small proportion of patients treated with Imatinib [8]. is suggests the possibility of a relationship between epigenetic silencing and development of IM resistance. Few studies have suggested that hypermethylation might play a role in disease progression in CML. It could be plausible that changes in gene silencing by DNA methylation might play a role in developing alternative routes for cells to circumvent the effects of IM. We hypothesized that promoter hypermethylation of HOXA4 could be an epigenetic mechanism which mediate resistance to IM in CML patients. is study was designed to test this hypothesis. were Philadelphia chromosome positive CML patients in chronic, accelerated, or blast phase, treated for at least 12 months, with IM (400 mg and 600 mg, resp.) on frontline treatment. ese CML patients were categorized into IM resistant and IM good responders based on their molecular and/or cytogenetic response. IM-resistant patients were de�ned as those CML patients showing less than complete cytogenetic response by 12 months and/or lack of attainment of major molecular response by 18 months aer initiation of therapy. Secondary resistance was de�ned as loss of complete cytogenetic response and/or loss of major molecular response.

Methodology
ree millilitres of peripheral blood from each patient was collected in EDTA tube. Additionally, blood samples from 12 normal healthy controls were also collected and included for analysis. Universal methylated DNA and unmethylated DNA (ZYMO research, USA) were used as 100% and 0% methylation DNA control, respectively. Both types of the DNA were modi�ed with bisul�te treatment and was subsequently mixed according to the ratio of 10%, 25%, 50%, and 75%. is serial methylation percentage was included in each experimental run.

�enomi� D�A ��tra�tion and Sodi�m Bis�l�te �reatment.
e genomic DNA of all patients and controls was isolated using the GENTRA PUREGENE Blood Kit (Qiagen, Germany) according to the supplier's recommendation. DNA quantity was identi�ed spectrophotometrically by using NanoQuant In�nite M200 (Tecan, Switzerland) and the quality of the DNA was con�rmed by agarose gel electrophoresis using 1% agarose gel.
Aer extraction of genomic DNA, 500 ng of the DNA was subjected to bisul�te treatment utilizing the EZ DNA Methylation-Gold Kit (ZYMO Research, USA) following manufacturer's recommendation. Besides the patient samples, universal methylated DNA and unmethylated DNA were also treated with bisul�te using the same kit. Before mixing the methylated and unmethylated controls into 10%, 25%, 50%, and 75% percentages, the concentration of bisul�te treated DNA control samples were carefully measured at a value of 40 g/mL for Ab 260 = 1.0 (the wavelength used corresponds to RNA wavelength as the recovered bisul�tetreated DNA was single stranded with limited non-speci�c base-pairing at room temperature) (ZYMO research, USA). e concentration of the eluted bisul�te treated DNA samples of all patients was also measured and the �nal concentration used was 20 ng for MS-HRM analysis.

Primer Design.
Primers were designed based on criteria stated by Wojdacz et al. [9] with some additional mod-i�cations using the Methyl Primer Express v1.0 Soware (Applied Biosystem, USA). While designing the primers, the following points were considered. e primers should amplify 100 to 150 bp PCR product with only one CpG dinucleotides each, as more CpG dinucleotide in the primer sequence was found to promote bias ampli�cation towards the methylated template. e primers also should amplify both methylated and unmethylated sequence simultaneously F 1: Derivative Melt peak of the serial percentage of methylation control produced two speci�c peaks which represent the unmethylated (approximately 77 ∘ C) and methylated (approximately 80 ∘ C) PCR product. Fully unmethylated sample produced only unmethylated peak, 100% methylated sample produced only methylated peak and samples with mixture of unmethylated, and methylated displayed both peaks.
( Figure 1). e CpG in the primers should not be at 3 ′end, and preferably, it should be placed as close as possible to the 5 ′ -end of the primer. However, CpG situated in the middle (at least 5th nucleotide from the 3 ′ -end) could also be acceptable provided it could produce good melt curve differentiation [10�. To ensure that the bisul�te converted DNA speci�c ampli�cation and to prevent ampli�cation of unconverted DNA template, the 3 ′ -end of primers should contain one or more Ts derived from the non-CpG aer bisul-�te treatment. Accordingly the primers sequences designed for our study was 5 ′ -TTTTGAAGGATA GAAGTTTGA-3 ′ (forward primer) and 5 ′ -TCCTCTC AAAACCCTCTAC-3 ′ (reverse primer) for HOXA4 promoter.

Validation of Designed
Primer. Subsequent to primer design, the forward and reverse primer sequences were tested for their possible secondary structure, self dimer and hetero dimer formation using OligoAnalyzer 3.1 Soware (http://eu.idtdna.com/analyzer/Applications/OligoAnalyzer/ Default.aspx). As Δ less than −7 may form a very stable primer dimer, primer sequence with Δ higher than −7 was chosen. e higher the Δ (more than −3.5) the better it seemed, as it could subordinate the primer dimer problem.
e computational prediction of the melting curve as well as the derivative melting curve shape was also derived on the sequence of the PCR product generated, using algorithm like the uMelt v2.0.2 (http://www.dna.utah.edu/umelt/um.php). By using this algorithm, the expected melting temperature of the PCR product was of help in forecasting the melting curve temperature adjustment during the optimization of the laboratory work. Care was taken to see that the derivative melting peak also had only one speci�c peak without any shoulder at the adjacent slope. PCR amplicon with several melting peaks would be showing the presence of multiple melting domains and may produce complex melting pro�le that maybe hard to interpret.
A sequence similarity search program designed to explore in silico bisul�te modi�ed DNA (either methylated or not at its CpG dinucleotides) was used to con�rm the ampli�cation speci�city of the designed primer. e primers were blast before synthesised, using the methBLAST soware (http://medgen.ugent.be/methBLAST/). All samples and DNA percentage controls were performed in triplicate. e PCR condition was started at 95 ∘ C for 2 minutes for initial denaturation, followed by 50 cycles of 10 seconds at 95 ∘ C for denaturation, 30 seconds at 50 ∘ C for annealing and another 30 seconds at 72 ∘ C for extension. e PCR ampli�cation was then followed by heteroduplex formation at 95 ∘ C for 30 seconds and subsequently 60 ∘ C for 1 minute. e high-resolution melting analysis was performed immediately aerwards by increasing the temperature from 65 ∘ C to 95 ∘ C for 10 seconds at each step with the 0.2 ∘ C increments. For each run, a no template control (NTC) and serial percentage control (0%, 10%, 25%, 50%, 75%, and 100%) in triplicate were included ( Figure 2). e annealing temperature during the PCR ampli�cation was gradiently optimized as it could create ampli�cation bias in MS-HRM. Higher annealing temperature could introduce bias towards the ampli�cation of methylated template. e most preferable annealing temperature would be the one that could differentiate between the serial percentage controls. Hence, multiple annealing temperatures with mixtures of methylated controls were tested and the best �t standard melting curve was selected.

Statistical Analysis.
Unconditional logistic regression analysis was used to assess the relationship between HOXA4 promoter methylation percentage and the response of CML patients to IM by calculating the Odd Ratios (ORs) and 95% Con�dence Interval (CI). e test was conducted by SPSS soware with all values as two-sided.

Results
A total of 95 samples including both IM resistant ( 57) and IM good response ( ) CML patients and 12 samples from normal control donors were tested for methylation percentage employing the methylation-speci�c   38 IM good response CML patients, 20 were males and 18 were females with mean age of 36 years. Among the IMresistant CML patients, 48 patients were categorized into primary resistance group and 9 patients were categorized into secondary resistance group.
Methylation percentage of the promoter region of HOXA4 gene in the normal controls was in the range of 10% to 49%. In the case of whole group of 95 CML patients, the HOXA4 promoter methylation was in the range of 10% to 100% with most of them showing dense range of more than 50% methylation. Table 1 shows the methylation percentage frequencies of HOXA4 gene promoter in normal controls and CML patients, in which the methylation percentages were subdivided into four categories. Except for the low level category (0-24%), the percentages of methylation levels in other 3 categories were signi�cantly higher in CML cases in comparison to control.
When the HOXA4 methylation pro�le among CML patients showing good response and resistance to IM was evaluated separately, hypermethylation was found to be signi�cantly less dense in IM good response CML patients, compared to IM-resistant CML patients. However, when the methylation percentages of HOXA4 were categorized into two classes, 1-49% as methylated and 50-100% as hypermethylated and the values were compared, HOXA4 hypermethylation was signi�cantly higher among IM-resistant CML patients ( ) than IM good response CML patients. Furthermore, when the risk association of the two methylation categories (methylated and hypermethylated) with IM resistance was evaluated, HOXA4 hypermethylation was found to be associated with a signi�cantly higher risk for IM resistance with OR, 4.658 (95% CI, 1.673-12.971; 3) as shown in Table 2.

Discussion
DNA promoter hypermethylation is a powerful mechanism of tumor-suppressor gene silencing that mediates neoplastic transformation [11]. Despite CML starts as a genetically homogeneous disease, it has been hypothesized that disease progression and clinical heterogeneity in CML are related to epigenetic factors including DNA hypermethylation. Hypermethylation in several tumor-suppressor genes (i.e., TFAP2A and EBF2) had been reported in CML patients on disease progression [12]. Recently, Jelinek et al., [8] observed a higher frequency of hypermethylation in OSCP1 and NPM2 genes among CML patients who were resistant or intolerant to IM. However, there are still no reports available on the involvement of HOX gene family hypermethylation in mediating resistance to IM. e HOX gene family consisting of 39 genes are a large family of homeodomain containing transcription factors which regulate developmental process, haematopoietic differentiation, and leukemogenesis. HOX gene translocations are observed frequently in leukemia. Majority of the HOX genes have CpG islands at their transcription start site (TSS) regions. Silencing of HOX genes by DNA methylation are thought to disrupt normal development of blood cells and thus to be involved in leukemic transformation [13]. Hence, compared to other protooncogenes, hypermethylation of HOX genes might affect the CML transformation.
By utilizing MS-HRM, the HOXA4 promoter methylation quanti�cation showed a distribution pro�le of 10% to 100% methylation, with none of the samples showing 0% methylation. Samples from normal individuals showed methylation of 10% to 49% whereas samples from CML patients showed methylation of 10% up to 100%.
However, based on our experience as well as from the literature, designing of the primer was found to be the most crucial part in ensuring that the methylation percentage be clearly differentiated [10]. Our experience prompt us to suggest that, in order to amplify HOXA4 promoter for methylation pro�ling, the methylation-speci�c primers should have only one CpG site in the forward and reverse primer, respectively. By considering this kind of factor in primer designing, methylation percentage ranging from 0% to 75% could be clearly differentiated. However, samples that show 75%-100% hypermethylation cannot be clearly differentiated among themselves.
In leukemia-free normal population, few researchers showed absence of methylation, whereas few other studies showed a very low/absence of methylation [14]. However, in the present study, normal samples showed a range of 10% to 49% methylation level at the promoter region of HOXA4. Because of this, samples that showed methylation level of 50%-100% only were considered as HOXA4 hypermethylated samples.
ere are reports correlating hypermethylation of HOXA4 with the development of leukemia. Zangenberg et al. [15] reported that 77% of their acute myeloid leukemia (AML) patients exhibited hypermethylation of HOXA4 promoter region. Apart from AML, another study had demonstrated the contribution of HOXA4 promoter hypermethylation in chronic lymphoid leukemia (CLL) [7]. Furthermore, HOXA4 hypermethylation has been demonstrated to be usually associated with the progression of CML to blast phase and play an important role in the development of leukemia [7].
In the current study, we further evaluated whether HOXA4 hypermethylation induced gene silencing could be an alternative mechanism of CML cells to circumvent the effects of IM and thereby develop resistance to IM. To the best of our knowledge, no previous reports are available on the involvement of HOXA4 in mediating IM resistance among CML patients, ours being the �rst of its kind. Interestingly, in our study, HOXA4 hypermethylation level of 50-100% was signi�cantly higher ( ) among IM-resistant CML patients compared to IM good response CML patients. When the association between HOXA4 hypermethylation and IM resistance was examined, patients with HOXA4 promoter hypermethylation level in 50-100% showed a signi�cantly higher risk for IM resistance (OR � 4.658; 95% CI, 1.673-12.971; value, 0.003). us, hypermethylation of HOXA4 may be a marker of resistance to IM. However, mechanistic studies are still needed to con�rm if hypermethylation of HOXA4 is indeed causing poor response to IM.
e treatment mechanism of IM involves the arrest of BCR-ABL dynamic activity. No known mechanism of IM activity towards HOXA4 has been literally reported so far. Hypermethylation of HOXA4 has been found to promote inactivation of gene expression [15,16]. As HOXA4 protein is a DNA-binding transcription factor which may regulate gene expression, morphogenesis, and differentiation, it is reasonable to suggest that the suppression of HOXA4 protein production by hypermethylation induced gene silencing could be one of the potential mechanisms in BCR-ABL-independent pathway that promote IM resistance in CML patients. Moreover, Fournier et al. demonstrated the potential of HOXA4 retrovirus-mediated over expression of haematopoietic stem cell to give rise to mature myeloid progeny [17]. In ovarian cancer, several studies reported that HOXA4 plays an important role in cell motility, spreading and cell-cell adhesion [18,19]. Based on those reports, it is reasonable to suggest that suppression of HOXA4 protein might be impairing the normal development as well as proliferation of myeloid progeny and could be a potential epigenetic mechanism in BCR-ABL-independent pathway in promoting IM resistance among CML patients.
It is reasonable to suggest that hypermethylation of HOXA4 gene might be circumventing the clinical response to IM and thus playing an important role as inhibitor to normal leukemogenesis. is data contributes to a new understanding of epigenetic mechanism also as a mediator in resistance development to IM in CML patients. Inhibition of this process may have potential as better therapy and warrants the need of utilizing hypomethylating agents for CML patients showing this epigenetic mechanism of resistance. us, hypermethylation pro�le of HOXA4 gene also could be considered as an epigenetic biomarker, in addition to the BCR-ABL gene mutations analysis, for prediction of response to IM treatment among CML patients.