CpG promoter methylation of the ALKBH3 alkylation repair gene in breast cancer

DNA samples were derived from primary breast tumors (n = 265) and adjacent normal breast tissue (n = 30). The normal breast tissue was obtained from non-tumorous regions of the breast. The DNA was previously isolated from freshly frozen tissue following a standard protocol based on phenol-chloroform (+proteinase K) extraction. RNA samples were available for a subset of the tumor (n = 36) and normal breast (n = 10) tissue samples isolated using Tri-Reagent (Thermo Fisher Scientific). Clinical parameters, including tumor size, nodal status, histological grade along with disease-specific follow-up times were obtained from the nationwide Icelandic Cancer Registry [22]. This work was carried out according to permits from the Icelandic Data Protection Commission (2006050307) and Bioethics Committee (VSNb2006050001/03–16). Informed consent (written) was obtained from all patients.

The cell lines used in this study were obtained from the American Type Culture Collection. The cells were cultured in DMEM (CAMA-1, MDAMB-468, MCF-7, MCF-10A, MDAMB-231 and SKBr-3) or RPMI (HCC-38, Bt-474) with added 10% serum (+penicillin/streptomycin). DNA and RNA was extracted in parallel from the cell lines using Qiagen’s AllPrep DNA/RNA/miRNA Universal kit (80224; Qiagen).

Bisulfite conversion was carried out using the EZ-96 DNA Methylation-Gold kit from Zymo Research (D5008). We carried out the bisulfite conversion for 16 cycles of {95 °C for 30 s and 50 °C for 1 h} to then hold at 4 °C until samples were added to the DNA columns for completing the conversion following the manufacturer’s guidelines (Zymo Research).

The PyroMark Assay Design 2.0 software was used to design primers for the analysis of ALKBH3 promoter methylation. The following primer sequences were used: 5′-(Btn)-GTGGGATTATTAGGATTGAGGATT-3′ (5-biotin labelled) and 5′-CTCCAACAACTCCCAATCAC-3′. The pre-amplification PCR reaction was carried out using a hot-start PCR polymerase (Immolase DNA polymerase from Bioline; Bio-21,047). The PCR conditions were as follows: 96 °C for 10 min, 45 cycles of (96 °C for 30s, 60 °C for 30s and 72 °C for 30s) followed by 15 min hold at 72 °C and then 4 °C. The PCR products were then captured using streptavidin coated agarose beads (Streptavidin Sepharose High Performance 34 μm beads, GE Healthcare) under denaturing conditions to obtain single-stranded DNA. The pyrosequencing reaction was then carried out using the PyroMark Q24 machine (Qiagen) and PyroMark Gold-Q24 Reagents kit (Qiagen) using the following sequencing-primer: 5′-ACATCAAACACTTCCT-3′.

CpG methylation for three CpG’s were assessed (−58, −53 and −50 bp upstream of the TSS (p1) given the FANTOM5 promoterome database) [23]. The output data (obtained from PyroMark Q24 sequencing reactions), representing percent methylated cytosines over each of these three CpGs, was averaged for each sample analysed. This yielded a single measure representing a proxy for CpG methylation levels over the ALKBH3 promoter region. The statistical analysis of paired tumor and normal breast tissue samples made use of a paired Wilcoxon’s test using the wilcox.test function in R.

RNA was extracted from tumors and normal breast tissues using Tri-Reagent (Thermo Fisher Scientific). The RNA derived from cell lines was isolated in a simultaneous DNA/RNA isolation procedure using the Qiagen’s Allprep kit (Qiagen). Reverse transcription was carried out using High-Capacity cDNA Reverse-Transcription Kit (Thermo Fisher Scientific). ALKBH3 expression was quantitatively analysed by the SYBR green method using a real-time PCR (RT-PCR) machine (Applied Biosystems 7500). HPRT1 expression was used to normalize the expression data by computing the difference in Ct as follows: 2^{-(ALKBH3 Ct – HPRT1 Ct)}. The primers used for ALKBH3 were: 5′-AGCCACGAGTGATTGACAGAG-3′ and 5′- ACAAACAGACCCTAGATACACCT-3′, and for HPRT1: 5′- CCTGGCGTCGTGATTAGTGAT-3′ and 5′-AGACGTTCAGTCCTGTCCATAA-3′. The Spearman’s rank test for correlation was carried out using the cor.test function in R to assess the association between CpG methylation and expression.

Proteins were extracted from cell lines at 80% confluency using the EBC lysis buffer and measured at 490 nm using a spectrophotometer. The samples were denatured and electropherized using a 10% gel followed by transfer to PVDF membrane. The primary antibody (Millipore anti-ALKBH3 rabbit polyclonal 09–882) was used at 1:500 dilution overnight at 4 °C followed by washing with PBS-Tween. The secondary antibody (Santa Cruz donkey anti-rabbit IgG-HRP, sc-2313) was used at 1:10,000 dilution. The membrane was developed with ECL (Pierce ECL Plus Western Blotting Substrate, Thermo Scientific 32,132) and detected in a ImageQuant LAS4000. The β-actin primary antibody (MAB1501R; Millipore) was used at a dilution of 1:20,000 with secondary HRP-rabbit antibody anti-mouse IgG used at 1: 10,000 dilution (61–6020; Invitrogen).

CAMA1 and MDA-MB-468 were grown on coverslips and fixed with freshly prepared 4% para-formaldehyde solution for 15 min. After fixation, cells were treated with 1.5 M HCL for 20 min, to gain access to single stranded DNA, followed by a 2-min treatment with Sodium Borate (pH 8.5) to neutralize the acid. After permabilization (5 min, 0.2% TritonX) and 1 h of blocking (DMEM (Gibco) with 10% FBS (Gibco)) cells were stained with antibodies against 3-methylcytosine (3meC) (rabbit, Active Motif, 61111) and 5-methylcytosine (5meC) (mouse, abcam, ab10805) for 1 h at room temperature. Both antibodies were diluted 1:250 in blocking buffer. Next, samples were incubated with secondary antibodies, Alexa-Fluor 488 goat anti rabbit (A11008, Thermo Fisher Scientific) and Alexa-Fluor 555 goat anti-mouse (A21422, Thermo Fisher Scientific), diluted in blocking buffer (1:1000) for 1 h. Nuclear DNA was stained by DAPI (SIGMA, D9542). The DAPI stain was added directly to the secondary antibody solution (diluted 1:5000). After drying, the coverslips were mounted on glass slides using Fluoroshield (SIGMA, F6182) mounting medium.

Images were acquired using the FV1200 Olympus inverted confocal microscope. Dual colour confocal images were acquired with standard settings using laser lines 488 nm and 543 nm for excitation of Alexa Fluor 488 and Alexa Fluor 568 dyes, respectively. Nuclear DAPI staining was imaged using excitation by the 405 nm laser. For each condition 10 images were randomly acquired with the 20X/0.75 objective and imported into CellProfiler for downstream image analysis. For each data point, 400–600 nuclei (identified by DAPI staining) were analysed for 3meC and 5meC nuclear intensity (mean integrated intensity).

The 3meC and 5meC values presented in Fig. 2d are based on four independent staining experiments. The Wilcoxon’s rank sum hypothesis test was used to assess differences in 3meC and 5meC values (in R 3.1.0).

Estrogen receptor (ER), progesterone receptor (PR) and HER-2 expression were previously analysed on tumors by immunohistochemistry (IHC) on tissue microarrays (TMAs) [24]. The TMAs were constructed as previously described (Stefansson et al. 2009). Immunohistochemistry (IHC) was then applied using 4 μm thick TMA sections using the following anti-bodies: anti-ER (1D5; DAKO), anti-PR (PgR 636; DAKO) and anti-HER2 (HercepTest Kit; DAKO). ER and PR were scored positive given any visible nuclear staining in more than 1% of tumor cell nuclei. HER-2 positivity was defined as score of 3+ according to criteria provided by the anti-body manufacturer.

Information on CpG methylation over the promoter region of ALKBH2 and ALKBH3 was obtained from pre-existing methylome analyses published by Stefansson et al. (GSE52865), Dedeurwaerder et al. (GSE20713) and Teschendorff et al. (GSE69914) available through the Omnibus repository at NCBI’s website (www.​ncbi.​nlm.​nih.​gov/​gds/​) [7, 25, 26]. The normalized data were extracted from the SOFT formatted files using the GEOquery package in R and analysed by comparing normal breast tissue samples and breast cancers. This was carried out using the Student’s t-test on M–values computed using M_i = log₂(B_i / (1-B_i)) where B represents the β-value coupled with the Benjamini-Hochberg adjustment procedure to account for multiple hypothesis testing making use of the p.adjust function in R. The multiple hypothesis adjustment accounted for the total number of CpGs represented on the array platform, i.e. adjusting for the entire >450 thousand CpGs in GSE52865 and GSE69914 and >27 thousand CpGs in GSE20713.

DNA methylation (450 K Infinium) and RNAseq (V2) level 3 data were downloaded from the Cancer Genome Atlas data repository (http://​cancergenome.​nih.​gov/​) [27]. Firstly, the analysis of differential ALKBH3 mRNA expression levels between normal breast tissue and breast cancers was carried out using the Wilcoxon’s rank sum hypothesis test taking into account adjustment for multiple hypothesis testing including the entire set of >20 thousand genes included in the RNAseqV2 dataset. This was carried out using the Benjamini-Hochberg (BH) procedure through the p.adjust function in R. Secondly, the relation between CpG methylation for each site represented over either ALKBH2 and ALKBH3 were studied with respect to ALKBH3 mRNA expression using Spearman’s correlation analysis and, as before, with genome-wide adjustment of the P-values using the BH procedure to account for multiple hypothesis testing.

Information on epigenetic marks for the ALKBH3 promoter region in variant human mammary epithelial cells (vHMEC) was obtained from the Roadmap Epigenomics browser (egg2.​wustl.​edu/​roadmap/​web_​portal/​) [28]. This includes information on ALKBH3 expression based on RNA sequencing and chromatin marks based on ChIP-seq along with data on chromatin accessibility based on DNA sequencing. Data on CpG methylation for the vHMEC cells was derived from methylCRF computational analysis using MeDIP-seq and MRE-seq data to infer whole-genome 5-methylcytosine states as carried out and provided by the Roadmap Epigenomics project [28].

Information on nucleotide positions for ALKBH3 gene structure (introns and exons) was downloaded from Ensembl (GRCh37 browser; HG19). Data on transcriptional start site (TSS) and CpG islands were obtained from the FANTOM5 promoterome [23]. Using the UCSC genome browser, the chromStart/chromEnd fields in the hg19.cpgIslandExt table provided the CpG island positional information. The R statistical software (R 3.1.0) was then used to graphically represent the ALKBH3 promoter with respect to the TSS, 1st Exon and CpG island.

The association between ALKBH3 promoter methylation and subtype-specific markers was assessed using wilcoxon’s rank sum hypothesis testing (wilcox.test in R). The chi-squared test was used to assess the association between tumor subtype classification, histological grade, tumor size and nodal status (chisq.test in R). Differences in breast cancer-specific patient survival with respect to ALKBH3 methylation in tumor tissue was assessed using the log-rank hypothesis test (survdiff function in R). Cox’s proportional hazards regression model was use for multivariate analysis of survival (coxph function in R). The cox.zph function in R was applied to assess the assumptions of the regression model with respect to proportionality over time.

By making use of methylome data for breast cancers and normal breast tissues, we specifically asked whether aberrant CpG methylation events are found over the promoter region of either ALKBH2 or ALKBH3. To achieve this, we used datasets available through Omnibus including those published by Stefansson et al. (GSE52865), Dedeurwaerder et al. (GSE20713) and Teschendorff et al. (GSE69914) [7, 25, 26]. The analysis of these datasets consistently identify aberrant CpG methylation over the ALKBH3 gene promoter in breast cancers. Figure 1a illustrates this finding where statistically significant CpG methylation events are seen over the ALKBH3 promoter region (P_adj < 0.001). In contrast, differential methylation between breast cancer and normal breast tissue was not identified over the promoter region of ALKBH2 (Fig. 1a).

We used the Cancer Genome Atlas dataset to assess the impact of ALKBH3 promoter methylation on mRNA expression levels. Firstly, these data provide additional support for differential methylation over the ALKBH3 promoter region between normal breast tissue and tumours (data not shown). Secondly, significantly lower expression levels of ALKBH3 mRNA were seen in breast cancers compared with normal breast tissue samples. This finding holds statistically significant after adjusting for multiple hypothesis testing taking into account the entire >20 thousand protein-coding genes in the RNAseqV2 TCGA dataset (Wilcoxon’s rank sum test; P _adjusted = 0.018) as shown in Fig. 1b (left panel). Thirdly, the majority of the CpGs identified as differentially methylated are also found significantly associated with down-regulation of ALKBH3 mRNA levels in breast cancers (Fig. 1b, right panel; Additional file 1). The topmost significant CpG derived from this analysis (Spearman’s rho = −0.47; P_adj < 0.00001) is represented by cg12046254 shown in Fig. 1b (right panel).

Information on epigenetic regulation and expression of the ALKBH3 gene in human mammary epithelial cells (vHMEC) is displayed in Fig. 2a demonstrating transcriptionally active chromatin configuration over the promoter region. This can be seen in DNase-seq signal peaks found upstream of the promoter and extending into the first exon together with active histone markings, i.e. H3 lysine 4 tri-methylation (H3K4Me3) localized over the first exon and H3 lysine 36 tri-methylation (H3K36Me3) over the gene body region collectively indicating open chromatin and active transcription (Fig. 2a). Notably, the H3K4Me3 activation marks are found in the absence of repressive H3 lysine 4 mono-methylation (H3K4me1), H3 lysine 27 tri-methylation (H3K27Me3) and H3 lysine 9 tri-methylation (H3K9Me3) as shown in Fig. 2a. Indeed, the expression data (RNAseq track) shows clear signals from all ten exons of the ALKBH3 gene – the first three exons are shown in Fig. 2a.

In agreement with transcriptionally active chromatin configuration, the ALKBH3 gene promoter region and first exon are lacking of repressive CpG methylation marks (DNA methyl track in Fig. 2a). The ALKBH3 promoter is associated with a CpG island (spanning a region from 43,902,254 bp to 43,902,528 bp) extending into the first exon (Fig. 2b) and, indeed, the entire CpG island (CGI) is found unmethylated in the vHMEC breast epithelial cells. The CGI was identified from the UCSC genome table browser defined as regions of at least 200 bp where the GC content is at least 50%. The ALKBH3 promoter associated CGI was found to be 274 base pairs in length.

We designed a DNA methylation pyrosequencing assay to carry out CpG methylation analysis for the ALKBH3 gene promoter region (Fig. 2b). The DNA region assayed, labelled R in Fig. 2b, includes three CpG sites found −50, −53 and -58 bp upstream of the major TSS (FANTOM5 element p1). In this way, the assay was designed to reflect ALKBH3 promoter methylation status which we then applied across 303 DNA samples, i.e. 8 breast-derived cell lines, 30 normal breast tissue samples and 265 primary breast tumors.

Out of the eight breast-derived cell lines analysed with respect to ALKBH3 promoter methylation, seven were derived from breast tumors (CAMA-1, Bt-474, HCC-38, SKBr-3, MDA-MB-231, MCF-7, MDA-MB-468) and one was derived from a fibrocystic breast lesion (MCF-10A). We identified ALKBH3 promoter methylation in two of the seven breast cancer cell lines, i.e. in CAMA-1 and Bt-474 (Fig. 3a). The MCF10A cell line, often used to reflect normal breast epithelial cells, was clearly unmethylated over the ALKBH3 promoter.

ALKBH3 mRNA expression was not detected in either CAMA-1 or Bt-474 whereas all other cell lines showed ALKBH3 mRNA expression (Fig. 3b). The association between ALKBH3 promoter methylation and mRNA expression was found to be statistically significant (Spearman’s rho = −0.73; P = 0.039). Further, complete loss of ALKBH3 protein expression was seen only in the two cell lines showing ALKBH3 promoter methylation, i.e. CAMA-1 and Bt-474 (Fig. 3c).

ALKBH3 has been reported to catalyse the removal of 3-methylcytosine (3meC) on single stranded DNA [29]. In line with that, RNAi mediated knockdown of ALKBH3 has previously been shown to cause an increase in 3meC levels on single-stranded DNA [30]. To determine the functional impact of ALKHB3 promoter methylation, we developed a novel imaging-based assay for the quantification of 3meC on single stranded DNA. Using this method we confirmed previous findings [30] demonstrating increased formation of 3meC damage following siRNA knock-down of ALKBH3 (Additional file 2). Further, we show statistically significant differences in 3meC damages between ALKBH3 deficient cell line CAMA-1 compared with ALKBH3 expressing cell line MDA-MB-468 (P = 0.029), see Fig. 3d and e. For reference, we show that 5meC intensities are not significantly different between CAMA-1 and MDA-MB-468 (P = 0.69). This indicates that, comparable to knockdown of ALKBH3 [30], epigenetic inactivation of ALKBH3 results in a higher burden of 3meC, most likely because of less efficient repair of alkylation damage.

Out of the 265 primary tumors analysed with respect to ALKBH3 promoter methylation, a subset of 30 tumor samples were matched with adjacent normal breast tissue samples from the same patients – thereby enabling assessment of differential methylation in paired samples. This analysis reveals clear differences in CpG methylation over the ALKBH3 promoter region between primary tumors and normal breast tissue from the same 30 individuals (Fig. 4a).

This analysis further shows that, in normal breast tissue, the mean methylation levels over the ALKBH3 promoter region is approximately 1.0% (99%CI: 0.12–1.9%). Deviations from the 99% confidence interval for ALKBH3 promoter methylation levels in normal breast tissue samples can be considered “aberrant”. On the basis of the upper 99%CI for normal breast tissue samples (99%CI upper limit ~2% methylation), the incidence of ALKBH3 promoter methylation in primary breast tumors is approximately 27%, i.e. 72 out of 265 primary breast tumors show aberrant ALKBH3 promoter methylation.

RNA samples available from normal breast tissue and tumor samples were used to assess ALKBH3 mRNA expression. Firstly, this analysis demonstrates statistically significant differences in ALKBH3 mRNA expression between normal breast tissue and tumor samples wherein breast tumors generally show reduced ALKBH3 expression (Fig. 4b). This provides an independent confirmation for the previous observations based on the use of available data from the TCGA project shown in Fig. 1b. Secondly, by using DNA samples available from a subset of the same tumor samples as analysed with respect to ALKBH3 expression, we were able to assess the association between mRNA expression levels and promoter methylation status. This analysis, shown in Fig. 4c, further validates the impact of ALKBH3 promoter methylation on mRNA expression levels.

Table 1 displays clinical and pathological characteristics of the patient cohort (n = 265). ALKBH3 promoter methylation was not found to be significantly associated with the expression of subtype-specific markers, i.e. estrogen-receptor (ER), progesterone-receptor (PR), erb-b2 receptor tyrosine kinase 2 (known as HER2) and MKI67 (known as Ki-67) as shown in Fig. 5a. Additionally, no associations were found for ALKBH3 promoter methylation in relation to discrete subtype classification based on these four subtype-specific markers, histological grade, tumor size or nodal status (Additional file 3).

Nonetheless, we found significantly reduced survival in patients with tumors showing ALKBH3 promoter methylation (Fig. 5b). This was seen in breast tumors with high cytosine methylation levels over the ALKBH3 promoter region, i.e. those showing at least 20% cytosine methylation (Fig. 5b). This level of methylation is indeed substantially higher than the 2% threshold level defining aberrant CpG methylation as was established by looking at the distribution seen in normal breast tissue samples (see section 3.4).

In this study, we show that the ALKBH3 gene promoter region undergoes aberrant epigenetic repression in a significant proportion of primary breast tumours. The ALKBH3 gene is therefore a novel addition to the catalogue of DNA repair genes found inactivated in breast cancer. Previous studies have shown that treatment with the DNA alkylating agent MMS induces the expression of ALKBH3, highlighting the important role of ALKBH3 in repair of alkylation DNA damage [21]. Indeed, our results indicate that inactivation of ALKBH3 is associated with an increased burden of unrepaired alkylation DNA damage involving 3-methylcytosine (3meC) suggesting that our findings has biological consequences. If left unrepaired these modifications can lead to alkylation-induced cell death or can be converted into mutations through the use of error-prone translesion DNA polymerases such as Pol η (eta), Pol ι (iota), and Pol κ (kappa) [31, 32]. Collectively, our observations support the hypothesis that defective repair of alkylation damage occurs in the development of a substantial fraction of breast cancers.

Our results furthermore emphasize the importance of quantification-based methods for DNA methylation analyses in clinical applications. This becomes clear by looking at varied levels of CpG methylation over the ALKBH3 promoter with respect to survival outcomes. Here, increasingly higher levels of promoter methylation were associated with shortened patient survival. Declaring tumours as either ALKBH3 promoter methylated or unmethylated is therefore not a straightforward task. In this study, two different methods are applied for this purpose, i.e. 1) making use of normal tissue samples as reference to then define “abnormally” high levels of ALKBH3 promoter methylation in tumours and 2) by identifying clinically relevant levels of ALKBH3 promoter methylation. Making use of normal breast samples as a reference, does not necessarily provide a means to identify tumours showing ALKBH3 inactivation events. This is because only slightly elevated, but still “abnormal”, promoter methylation might simply reflect passenger events of no relevance to the course of disease progression. The second method refers to making use of clinical parameters involving tumour phenotype or disease survival to then declare a “clinically relevant” threshold for defining ALKBH3 promoter methylated tumours. Given the observed impact on disease progression, this second method is more likely to hold relevant at least as a proxy for identifying tumours affected by ALKBH3 promoter methylation. Using this definition, we report the incidence of ALKBH3 promoter methylation at approximately 5% in our cohort (13 out of 265 primary breast cancers).

Defective DNA repair capacity is frequently observed in various human cancers [14]. In breast cancer, this mostly involves defective repair of DNA double-stranded breaks by homologous recombination (HR), e.g. as is known to occur in tumor cells arising in carriers of either BRCA1 or BRCA2 germline mutations [33]. Other potential sources of DNA repair deficiency in breast cancer, also involving DSB repair processes, include germline mutations in DNA repair genes PALB2, FANCM, ATM, CHEK2 (and possibly also RAD51C) characterized as loss-of-function variants [27, 34, 35]. Additionally, somatic mutations occur in both BRCA1 and BRCA2 as well as the ATM gene accounting for a small fraction of patients [27].

In addition to genetic mutations, our previous studies along with others have shown that the BRCA1 gene undergoes CpG promoter methylation in at least 5–10% of all sporadically arising breast tumors [6]. More recently, RAD51C (a RAD51 paralog involved in double-strand break repair) and the RAD51 recombinase have been identified as targets of epigenetic silencing in breast cancer [8, 36, 37]. Other contributors include deregulated expression of miRNAs including miR-182 and miR-146a/b targeting BRCA1 mRNA transcripts for degradation [38, 39].

In summary, the development of breast cancer is clearly linked to inactivation of genes involved in the repair of DNA double-stranded breaks by HR. The involvement of other DNA repair pathways in breast cancer development is, however, currently unclear. Of these, DNA repair of alkylation damage by direct reversal has been suggested following the identification of CpG promoter methylation of the MGMT gene in breast cancer [17]. Our results support this notion by identifying the ALKBH3 gene as a novel addition to the catalogue of DNA repair genes found inactivated in breast cancer.

According to our results, ALKBH3 represents a candidate tumor suppressor gene. A high burden of mutations, caused by ineffective DNA repair, is generally accepted as an important factor in cancer development. Incomplete removal of alkyl groups on DNA has been shown to cause DNA damage, cell cycle arrest and apoptosis. Indeed, cancer genome sequencing has already established an important role for alkylation damage leading to genetic mutations [20]. Consequently, it seems likely that ALKBH3 mediates its tumor suppressive function via its role in DNA alkylation repair. This is supported by the increased level of alkylation damage in ALKBH3 inactivated cells, reported here and in previous studies [30]. It can, however, not be ruled out that other described functions of ALKBH3 also contribute to its anti-tumor activities – including alkylation repair of RNA and the recently described link to regulation of 1-methyladenine mark in mRNA [29, 40‐42].

While our data suggest that ALKBH3 has tumor suppressor properties in breast cancer, other research has shown ALKBH3 overexpression in various cancer types. ALKBH3 overexpression is found in prostate cancer [43] and other cancer types [44‐48]. In this context, ALKBH3 overexpression likely relates to adaptation of cancer cells to tolerate endogenous alkylation damage to DNA or RNA. This interpretation is in fact relevant in the context of a recent pre-clinical study showing resistance to alkylating drug temozolomide following experimental overexpression of MGMT [49]. Similar results have also been described with respect to ALKBH2 [50].

Precision medicine is aimed at optimizing treatment benefits by looking at each patient in terms of genomic or epigenomic variants and, on the basis of this information, to then select the most appropriate combination of cytotoxic or targeted drugs. This is highly relevant with respect to CpG promoter methylation of the MGMT gene now widely recognized as a predictor for patient response to alkylating agents such as temozolomide [19, 51]. Indeed, temozolomide induces methylation damage in DNA involving O6-methylguanine along with N7-methylguanine. In this way, tumor cells lacking the MGMT repair gene as a result of CpG promoter methylation events are highly sensitive to temozolomide [19]. Similarly, in breast cancer, loss of MGMT was recently linked to temozolomide sensitivity in a pre-clinical study [49]. Whether the same principle can be applied with respect to the ALKBH3 gene, i.e. by inducing 3meC and 1meA in patients having developed breast tumors with ALKBH3 promoter methylation, remains to be determined.