Progress in Microbes and Molecular Biology Original Research Article 1 Landscape Of HOXA Genes Methylation in Colorectal Cancer Muhiddin Ishak1, Rashidah Baharudin1, Loh Teng-Hern Tan2, Learn-Han Lee2*, Nurul-Syakima Ab Mutalib1* 1UKM Medical Molecular Biology Institute (UMBI), Universiti Kebangsaan Malaysia, Jalan Yaacob Latif, 56000 Cheras, Kuala Lumpur, Malaysia 2Novel Bacteria and Drug Discovery Research Group (NDBB), Microbiome and Bioresource Research Strength (MBRS), Jeffrey Cheah School of Medicine and Health Sciences, Monash University Malaysia, 47500 Bandar Sunway, Selangor Darul Ehsan, Malaysia Abstract: Colorectal cancer (CRC) is among the most common cancers worldwide and the second leading cause of cancer- related death in Malaysia. The HOXA gene cluster is a family of Homeobox A genes encoding transcriptional regulators that play vital roles in cancer susceptibility and progression. Dysregulated HOXA expression influences various aspects of carcinogenesis processes. Therefore, this study aims to elucidate the methylation landscape of HOXA genes in CRC. Twelve pairs of CRC — adjacent normal tissues were subjected to Infinium DNA MethyEPIC array. Differentially methylated regions were identified using the ChAMP Bioconductor and methylation levels of HOXA genes were manually curated. We identified 100 significantly differentially methylated probes annotated to HOXA genes. HOXA3 has the highest number of differentially methylated probes (n=27), followed by HOXA2 (n=20) and HOXA4 (n=14). The majority (43%) of the probes were located at the transcription start site (TSS) 200, which is one of the gene promoters. In respect to CpG islands (CGI), the probes were equally located in the island and shore regions (47% each) while a minor percentage was in the shelf (6%). Our work gave a comprehensive assessment of the DNA methylation pattern of HOXA genes and provide the first evidence of HOXA2, HOXA3 and HOXA4 differential methylation in Malaysian CRC. The new knowledge from this study can be utilized to further increase our understanding of CRC methylomics, particularly on the homeobox A genes. The prognostic and diagnostic roles of the differentially methylated HOXA genes warrant future investigations. Keywords: Homeobox A genes; colorectal cancer; DNA methylation; HOXA2; HOXA-AS3 Received: 10th April 2020 Accepted: 12th May 2020 Published Online: 18th May 2020 Citation: Ishak M, Baharudin R, Tan LT-H et al. Landscape Of HOXA Genes Methylation in Colorectal Cancer. Prog Microbes Mol Biol 2020; 3(1): a0000085. https://doi.org/10.3687/pmmb.a0000085 Introduction Cancer is a continuous global burden and colorectal cancer (CRC) placed as the fourth most frequently diagnosed cancer worldwide[1] and second in Malaysia[2,3]. Cancer occurs through the accumulation of multiple genetics and epigenetics changes[4]. Somatic mutation in APC, BRAF, KRAS, PIK3CA and TP53[5–8] are identified in CRC at varying frequencies and are perceived as the drivers of CRC formation. Although there are many efforts on investigating the molecular alterations involved in CRC pathogenesis, the existing knowledge remains inadequate for an early diagnosis and prognosis assessment. Therefore, further understanding of epigenetic components involved in CRC carcinogenesis is highly sought after and will unravel new genes which can be utilized as the diagnostic, prognostic and predictive biomarkers to prevent CRC-related mortality. Epigenetics mechanism can be generally categorized into histone modification and DNA methylation[9], with the latter being the most widely investigated. The clinical application of DNA methylation markers to determine at-risk patient populations, improve diagnostic criteria, and provide prognostic factors to guide treatment decisions are becoming increasingly relevant. This Copyright @ 2020 by Ishak M and HH Publisher. This work under licensed under the Creative Commons Attribution-NonCommercial 4.0 International Lisence (CC-BY-NC4.0) *Correspondence: Nurul-Syakima Ab Mutalib, UKM Medical Molecular Biology Institute (UMBI), Universiti Kebangsaan Malaysia, Jalan Yaacob Latif, 56000 Cheras, Kuala Lumpur, Malaysia; syakima@ppukm.ukm.edu.my. Learn-Han Lee, Novel Bacteria and Drug Discovery Research Group (NDBB), Microbiome and Bioresource Research Strength (MBRS), Jeffrey Cheah School of Medicine and Health Sciences, Monash University Malaysia, 47500 Bandar Sunway, Selangor Darul Ehsan, Malaysia; lee.learn. han@monash.edu. 2 is because DNA methylation can be reversed, thereby providing alternative treatment options for patients with methylation phenotype. Clinically, DNA methylation has also been demonstrated to have significant utility owing to its stability and relative ease of testing[9]. For instance, SEPT9 gene is the first US Food and Drug Administration (FDA)–approved diagnostic assay for CRC screening (https://www.epiprocolon.com). However, the clinical utility of the methylated SEPT9 assay is still limited owing to patients’ heterogeneity factor which includes various demographic characteristics and pathological features[10]. The search for a good diagnostic, prognostic and predictive DNA methylation markers in CRC is an active area of research. Homeobox genes play a role as the master regulators of morphogenesis and are aberrantly expressed in cancer[11]. These genes possess a highly conserved DNA sequence that code for the homeodomain proteins, which act as transcription factors that bind specifically to the DNA motifs and regulate the genes involved in cellular processes including adhesion, proliferation and differentiation[11]. Over 200 homeobox genes have been discovered in the human genome and are divided into four HOX gene clusters namely HOXA, HOXB, HOXC, and HOXD, positioned at the chromosome 7p15, 17q21.2, 12q13, and 2q31 loci, respectively. Of these, there are 11 HOXA genes[12] (Table 1). Table 1. The list of 11 HOXA genes under HOXL subclass homeoboxes [11]. HGNC ID (gene) Approved symbol Approved name HGNC:5099 HOXA1 Homeobox A1 HGNC:5103 HOXA2 Homeobox A2 HGNC:5104 HOXA3 Homeobox A3 HGNC:5105 HOXA4 Homeobox A4 HGNC:5106 HOXA5 Homeobox A5 HGNC:5107 HOXA6 Homeobox A6 HGNC:5108 HOXA7 Homeobox A7 HGNC:5109 HOXA9 Homeobox A9 HGNC:5100 HOXA10 Homeobox A10 HGNC:5101 HOXA11 Homeobox A11 HGNC:5102 HOXA13 Homeobox A13 Dysregulated homeobox gene expression is a frequent in cancer and one of the mechanisms causing such dysregulation is DNA methylation. The HOXA cluster is often methylated in the non–small cell lung cancer (NSCLC)[13,14], while HOXA5 and HOXA11 promoter methylation diminishes their tumour-suppressive function through transcriptional silencing[15,16]. HOXA11 is hypermethylated in gastric cancer tissues and is significantly associated with TNM III and IV patients[17]. Hypomethylation HOXA10 and HOXA11 able to discriminate ovarian cancer tissue from the normal tissue[18]. HOXA9 was reported to be hypermethylated in half of the ovarian cancer patients and is significantly associated with endometrioid histological subtype[19]. In bladder cancer, HOXA9 promoter methylation has been linked with cisplatin chemotherapy-resistant and metastatic bladder cancer, and reversing the DNA methylation using decitabine sensitized the cancer cells to cisplatin[20]. Yet, little is known about HOXA methylome in CRC. Only recently, Li and colleagues reported the DNA methylation status of three HOXA genes, which are HOXA2, HOXA5 and HOXA6 in CRC using targeted bisulfite sequencing assay[21]. Nevertheless, a comprehensive, unbiased methylome profile of HOXA genes has not been described. Therefore, this study aims to investigate the methylation landscape of HOXA genes in CRC context. Material and Methods Clinical specimens The archived 12 pairs of tumour-adjacent normal fresh frozen colon tissues (n=24) from CRC patients were retrieved from the UMBI-HCTM Biobank. The tissues were collected according to the procedures approved by UKM Research Ethics Committee. As a quality control procedure, all tissues were cryosectioned, followed by haematoxylin and eosin staining for the pathologist to determine the percentage of tumour cells and normal cells contents. Only tumour samples with ≥ 80% cancerous cells and normal adjacent colon tissues with ≤ 20% necrosis were selected for DNA extraction using Allprep DNA/ RNA/miRNA Universal Kit (Qiagen, USA) according to the manufacturer’s instructions. Then, the integrity of DNA was assessed using agarose gel electrophoresis while the quantity and purity were evaluated using Nanodrop 2000c Spectrometer (Thermo Fisher Scientific, USA). Bisulfite Conversion and Methylation Microarray Five hundred nanogram (500 ng) of DNA was subjected to bisulfite conversion to change all unmethylated cytosine to uracil using the EZ DNA methylation — Gold kit (Zymo Research, USA) according to the manufacturer’s protocol. The effectiveness of bisulfite conversion was determined using Universal Methylated DNA Standard & Control Primers (Zymo Research, USA) according to the manufacturer’s protocol. The Infinium DNA MethylationEPIC assay, covering 850,000 CpG dinucleotides spread over the whole genome, was performed according to the manufacturer’s specifications (Illumina, Inc.). Methylation Microarray Data Analysis The raw idat files obtained from methylation microarray were subsequently analyzed using GenomeStudio V1.9.0 and CHAMP Bioconductor packages[22]. Filters were applied to all datasets where CpG sites with detection p-values ≥ 0.01 in one or more samples were omitted from further analysis. To reduce the technical biases intrinsic to the probe design, the raw intensities were SWAN- normalized prior to statistical analysis[23]. β-values were then extracted and subjected to further statistical analysis. Expression of HOXA genes from The Cancer Genome Atlas (TCGA) study The mRNA expression data of selected HOXA genes in CRC were retrieved using web-based FireBrowse Gene Expression Viewer tool (https://gdac.broadinstitute.org/) Landscape Of HOXA... 3 Ishak M et al. from Broad Institute. This tool provides access to results of various omics analyses involving more than 14,000 cancer cases, from 38 types of cancers based on TCGA data version 2016_01_28. Statistical Analysis A T statistic from the limma Bioconductor package was used to determine the differentially methylated CpG sites[24,25]. The CpG sites were further filtered at an adjusted p-value < 0.05 to identify significant differentially methylated HOXA genes. To substantiate the specificity and accuracy of the differentially methylated probes, the discriminative performance of the probes was evaluated by receiver operating characteristic (ROC) curves, and the area under the ROC curve (AUC), specificity, and sensitivity at the optimal cut-offs were determined using GraphPad Prism V8 (GraphPad Software, Inc., USA). Results Locations of differentially methylated loci in HOXA genes We analysed the differential methylation status of 12 CRC tissue samples with the 12 adjacent cancer-free colonic tissue samples and only differentially methylated regions with adjusted p-value < 0.05 were reported. From the list of differentially methylated probes, we further filtered for HOXA genes. Here, we found that there are 100 CRC- associated differentially methylated probes in 11 HOXA genes, noncoding HOXA-AS3 and HOXA10-HOXA9 readthrough. HOXA3 has the highest number of differentially methylated probes (n=27), followed by HOXA2 (n=20) and HOXA4 (n=14) (Figure 1A). The majority (43%) of the probes were located at the transcription start site (TSS) 200 (Figure 1B), which is one of the gene promoters. In respect to CpG islands (CGI), the probes were equally located in the island region and shore regions (47% each) while a minor percentage was in the shelf (6%) (Figure 1C). Figure 1. Differentially methylated HOXA genes in CRC. (A) The number of differentially methylated probes in each HOXA genes. (B) Distribution of methylated loci in HOXA genes with respect to features. (C) Distribution of methylated loci in HOXA genes with respect to CGI. The genomic and gene-related regions of the significant differentially methylated HOXA genes were distributed differently. Generally, 53 probes (in seven genes) were hypermethylated compared to 47 loci (eight genes) that were hypomethylated. The largest portion of hypomethylated sites (55.3%) were in the shore and subsequently decreased in other categories (island 42.6% and shelf 2.1%). In contrast, more than half (50.9%) of the significantly hypermethylated loci of HOXA genes were on the island, followed by the shore (39.6%), and shelf (9.4%). None of the loci was identified in opensea region. Meanwhile, most of the significantly hypomethylated loci were in the 5’UTR (46.8%), followed by TSS200 (14.9%), gene body (12.8%), and 8.5% in each 1st exon, 3’UTR and TSS1500. Meanwhile, around a quarter (26.4%) of the significant hypermethylated loci were located in TSS200 and TSS1500, while the rest were mainly found in the gene body (20.8%), 1st exon (15.1%), and, to a lesser extent, in the 5’ and 3’ UTR (7.5% and 3.8%, respectively). Differentially methylated HOXA genes All of the 11 HOXA genes are significantly differentially methylated. Due to the power of comprehensive contents in the microarray platform, we also identified significant hypomethylation of the noncoding HOXA-AS3 and HOXA10- HOXA9 readthrough. High resolution, probe-level analyses revealed hypomethylation of 20 loci in HOXA3, with the remaining seven loci were hypermethylated. In HOXA2 and HOXA9, all the probes were hypermethylated, while probes in HOXA4 were hypomethylated. HOXA6 exhibited eight hypermethylated loci and only one was hypomethylated. The summary of hypo- and hypermethylated probes in each gene were summarized in Table 2. 4 Table 2. The number of hypermethylated and hypomethylated probes in the HOXA genes. Genes Number of hypermethylated probes Genes Number of hypomethylated probes HOXA1 2 HOXA-AS3 4 HOXA2 20 HOXA10 4 HOXA3 7 HOXA10-HOXA9 1 HOXA5 5 HOXA11 1 HOXA6 7 HOXA13 2 HOXA7 2 HOXA3 20 HOXA9 10 HOXA4 14 HOXA6 1 The 100 significant probes with the methylation changes (Δβ) are illustrated in Table 3. It is worth mentioning that the loci in HOXA2 were the locations with the highest methylation changes; these probes were hypermethylated in CRC as compared to the normal colon. On the other hand, the loci in HOXA3 were the most hypomethylated probes in CRC as compared to the normal colon. Table 3. The 100 significant differentially methylated probes in HOXA genes. Genes Probes Adjusted p-value Δβ Feature CGI HOXA2 cg06055873 2.66E-05 0.368 1st Exon shore HOXA2 cg24058604 1.35E-04 0.362 TSS200 shore HOXA2 cg05921905 2.49E-04 0.362 TSS200 shore HOXA2 cg04737131 3.75E-04 0.357 TSS1500 shore HOXA2 cg17353412 6.30E-06 0.357 1st Exon shore HOXA2 cg20747380 2.23E-05 0.356 1st Exon shore HOXA2 cg02979457 3.64E-04 0.326 TSS200 shore HOXA2 cg22943986 1.19E-03 0.322 TSS1500 shore HOXA2 cg06786372 5.16E-03 0.319 Body shore HOXA2 cg26069745 1.55E-04 0.317 1st Exon shore HOXA2 cg06769202 8.92E-04 0.312 TSS200 shore HOXA2 cg09871315 1.95E-03 0.293 TSS1500 shore HOXA2 cg23979631 2.88E-04 0.291 TSS200 shore HOXA5 cg03744763 7.31E-03 0.286 TSS1500 island HOXA1 cg07450037 1.02E-03 0.282 Body shore HOXA2 cg20087093 2.13E-03 0.276 TSS1500 shore HOXA2 cg02803819 7.12E-03 0.258 Body shelf HOXA9 cg12600174 1.19E-02 0.249 TSS200 island HOXA3 cg27539480 4.52E-03 0.244 3’UTR shore HOXA2 cg23206851 6.94E-03 0.243 TSS1500 shore HOXA3 cg02627455 5.21E-03 0.240 5’UTR shelf HOXA2 cg13661519 4.09E-03 0.240 Body shelf HOXA3 cg07153966 2.74E-02 0.234 Body island HOXA9 cg21001184 1.42E-02 0.228 TSS200 island HOXA9 cg03217995 3.19E-02 0.226 Body shore HOXA2 cg00188704 3.46E-02 0.222 Body shelf HOXA1 cg03116258 4.38E-04 0.220 1st Exon shore HOXA5 cg14882265 4.94E-02 0.211 TSS1500 island HOXA2 cg02225599 1.51E-02 0.204 TSS1500 island HOXA3 cg14216068 9.16E-03 0.194 3’UTR island HOXA3 cg09591524 3.06E-02 0.186 5’UTR island HOXA6 cg14044640 3.34E-02 0.184 TSS200 island HOXA5 cg03368099 2.16E-02 0.183 TSS1500 island HOXA7 cg20725013 2.91E-02 0.182 Body shore HOXA3 cg02439266 1.46E-02 0.182 5’UTR island HOXA5 cg01748892 3.58E-03 0.180 TSS1500 island Landscape Of HOXA... 5 HOXA3 cg12305431 1.79E-02 0.175 5’UTR shelf HOXA5 cg13694927 1.20E-02 0.168 TSS1500 island HOXA9 cg26476852 4.68E-02 0.167 1st Exon island HOXA9 cg20399871 2.10E-02 0.164 1st Exon island HOXA6 cg03529432 4.98E-02 0.161 TSS200 island HOXA9 cg16104915 1.81E-02 0.154 TSS200 island HOXA6 cg22469274 4.68E-02 0.154 TSS200 island HOXA6 cg09936824 3.98E-02 0.153 TSS1500 island HOXA6 cg19183743 4.34E-02 0.150 TSS1500 shore HOXA7 cg21778348 3.36E-02 0.149 Body island HOXA9 cg16913789 4.62E-02 0.146 Body island HOXA2 cg01217984 1.10E-02 0.134 TSS1500 island HOXA9 cg05065989 1.07E-02 0.129 TSS200 island HOXA9 cg07778029 3.18E-02 0.125 1st Exon island HOXA6 cg04265576 2.56E-02 0.123 TSS200 island HOXA9 cg03698009 4.16E-02 0.108 Body island HOXA6 cg12810523 4.84E-02 0.103 TSS200 island HOXA10 cg08938793 4.30E-02 -0.041 3’UTR shore HOXA6 cg23590202 8.97E-04 -0.052 TSS1500 shore HOXA13 cg01363170 4.33E-02 -0.056 3’UTR shelf HOXA-AS3 cg10374314 4.89E-02 -0.070 Body shore HOXA10 cg05092861 5.66E-03 -0.078 TSS200 shore HOXA4 cg23884241 2.58E-02 -0.078 1st Exon island HOXA4 cg04317399 1.43E-02 -0.084 1st Exon island HOXA13 cg02366798 4.96E-02 -0.089 3’UTR shore HOXA4 cg03724423 2.08E-02 -0.096 TSS1500 shore HOXA4 cg11410718 1.61E-02 -0.108 TSS200 island HOXA4 cg07317062 2.15E-02 -0.114 5’UTR island HOXA10 cg01078824 2.94E-02 -0.118 TSS200 shore HOXA4 cg19142026 1.84E-02 -0.137 5’UTR island HOXA4 cg17591595 4.32E-02 -0.151 TSS1500 shore HOXA4 cg22997113 1.80E-02 -0.170 1st Exon island HOXA3 cg16406967 1.99E-02 -0.196 5’UTR island HOXA3 cg22798849 3.57E-02 -0.198 5’UTR island HOXA3 cg18680977 1.76E-02 -0.209 5’UTR island HOXA3 cg16748008 4.02E-02 -0.220 5’UTR island HOXA4 cg11532431 2.10E-02 -0.231 Body island HOXA11 cg05516617 8.55E-03 -0.231 3’UTR shore HOXA3 cg23403004 2.42E-04 -0.244 5’UTR shore HOXA3 cg04778178 2.48E-04 -0.254 5’UTR island HOXA3 cg16644023 4.88E-02 -0.255 5’UTR shore HOXA3 cg15982700 4.41E-02 -0.259 5’UTR shore HOXA3 cg23806243 2.21E-03 -0.260 5’UTR shore HOXA3 cg00318947 1.81E-02 -0.270 5’UTR shore HOXA4 cg00562553 1.01E-02 -0.272 1st Exon island HOXA4 cg20171892 5.24E-04 -0.273 Body island HOXA3 cg04351734 2.10E-02 -0.274 5’UTR island HOXA10 cg05517976 1.19E-02 -0.275 TSS200 shore HOXA4 cg11227540 1.46E-02 -0.279 Body shore HOXA4 cg09799676 3.54E-03 -0.286 Body island HOXA3 cg21556281 1.04E-02 -0.291 5’UTR shore HOXA-AS3 cg14429861 7.94E-03 -0.299 TSS200 shore HOXA3 cg18430152 7.29E-04 -0.305 5’UTR island Ishak M et al. 6 HOXA10-HOXA9 cg22274074 1.72E-03 -0.307 TSS1500 shore HOXA3 cg14072564 6.32E-04 -0.311 5’UTR island HOXA4 cg17132446 4.36E-03 -0.323 Body shore HOXA3 cg01820751 2.39E-03 -0.328 5’UTR shore HOXA-AS3 cg06188746 3.12E-03 -0.331 TSS200 shore HOXA-AS3 cg18091117 3.05E-03 -0.331 TSS200 shore HOXA3 cg03483713 4.12E-04 -0.346 5’UTR shore HOXA3 cg26297005 1.81E-03 -0.361 5’UTR island HOXA3 cg15725372 6.95E-04 -0.367 5’UTR island HOXA3 cg00431187 7.17E-04 -0.381 5’UTR shore HOXA3 cg09798023 2.84E-04 -0.390 5’UTR shore TSS: transcription start site UTR: untranslated regions Expression of HOXA genes and their correlation with methylation level Using FireBrowse, gene expression of the HOXA genes were retrieved from COAD[5] and COADREAD[26] studies (Figure 2). The data were presented as log2 fold change. The expression of seven HOXA genes was downregulated (HOXA1, HOXA2, HOXA4, HOXA5, HOXA6, HOXA7, and HOXA13) while four of genes (HOXA3, HOXA9, HOXA10, HOXA11) were upregulated. The expression profiles of HOXA1, HOXA2, HOXA5, HOXA6, HOXA7, HOXA10, and HOXA11 are inversely related to the methylation level as predicted, but not the HOXA4, HOXA9 and HOXA13. Figure 2. Gene expression of HOXA genes from TCGA COAD and COAREAD studies. Landscape Of HOXA... 7 Receiver operating characteristics (ROC) curve analysis Lastly, the sensitivity and specificity of the methylation levels were further assessed using receiver-operator curve (ROC) analysis. The methylation levels of 10 topmost hypermethylated CpG sites significantly differentiated the CRCs from the normal colonic tissues (p-value 0.0032 to 0.0002) (Table 4). The highest discriminative accuracy was demonstrated by HOXA2 cg06055873 (AUC = 0.9514, confident interval = 0.8547 to 1.000, p-value = 0.0002). Other candidate probes also reached high diagnostic accuracy (Table 4; Figure 3). Table 4. Receiver operating characteristics (ROC) curve analysis of the top 10 differentially methylated probes in HOXA2 gene. Gene_probe Area Std. Error 95% confidence interval P value HOXA2 cg06055873 0.9514 0.04935 0.8547 to 1.000 0.0002 HOXA2 cg24058604 0.9444 0.04489 0.8565 to 1.032 0.0002 HOXA2 cg05921905 0.875 0.08441 0.7096 to 1.04 0.0018 HOXA2 cg04737131 0.8958 0.07498 0.7489 to 1.043 0.001 HOXA2 cg17353412 0.9444 0.05546 0.8357 to 1.053 0.0002 HOXA2 cg20747380 0.9375 0.06155 0.8169 to 1.058 0.0003 HOXA2 cg02979457 0.8681 0.07918 0.7129 to 1.023 0.0022 HOXA2 cg22943986 0.8611 0.09216 0.6805 to 1.042 0.0027 HOXA2 cg06786372 0.8542 0.08306 0.6914 to 1.017 0.0032 HOXA2 cg26069745 0.9167 0.0691 0.7812 to 1.052 0.0005 Figure 3. Box plot illustrating the comparison of β values and ROC curve-based evaluation of the diagnostic accuracy for the top 10 hypermethylated HOXA2 probes in cancer- ous and normal tissues. Ishak M et al. 8 Discussion In this study, we analyzed in greater detail the genome- wide methylation patterns of HOXA genes from 12 CRCs compared with their adjacent normal tissues. Interestingly, due to the high throughput nature of microarray platform, our analysis revealed that all of the 11 HOXA genes are significantly differentially methylated in CRCs. We are the first to report this new finding and shed new light on the possible role of these genes in colorectal carcinogenesis. Epigenetic alteration of HOXA genes has been widely studied in many cancers, especially non-small cell lung cancer (NSCLC)[27–29], yet, the investigation about this gene cluster in CRC is lacking. To date, there is only a handful of study which investigates the methylome of HOXA genes[21,30–32]. HOXA2 is the most significantly hypermethylated HOX gene in our study, and the methylated loci were mostly located in the promoter regions (TSS200 and TSS1500). Hypermethylation of this gene in CRC has also been recently reported in concordance with our finding[21]. In addition, the significant association between HOXA2 methylation with age, node status (N), stage, metastasis (M), lymphovascular invasion, perineural invasion, as well as the number of lymph node was also demonstrated[21]. Another study has shown that HOXA2 is hypermethylated in the rectal cancer mucosa compared to the nonmalignant rectal mucosa[30]. DNA methylation is known to be inversely correlated with mRNA expression; however, the published data on HOXA2 expression is lacking and the aforementioned two studies did not investigate the gene expression levels. Therefore we attempted to investigate the mRNA expression using the TCGA CRC dataset and the finding is in agreement with our hypothesis. The expression of HOXA2 is indeed downregulated in CRC compared to the normal tissues by 0.68 fold, and thus, could be explained by its hypermethylation status. Nevertheless, HOXA2 promoter is also found to be hypermethylated in nasopharyngeal cancer, whereby it associates with low mRNA expression in the biopsies and cell lines[33]. Moreover, in a study involving 101 patients from stage I–III NSCLC, methylation of HOXA2 was proposed to have prognostic significance in squamous cell carcinoma (SCC) subtypes patients[34]. Due to the limited number of patients and the lack of clinical information in our study, the association with clinical features were not established and warrant further investigation. In addition, ROC curves for HOXA2 gene show exceptional diagnostic ability in differentiating CRC from the normal healthy tissue, especially in Stage I patients with the AUC = 0.9979 [21]. We also observed a similar trend, whereby the loci in HOXA2 are the most significantly hypermethylated and exhibited high discriminative accuracy. HOXA5 and HOXA6 promoters were both hypermethylated in our study and this is supported by recent data by Li and colleagues[21]. The authors went on to demonstrate a significant association between HOXA5 hypermethylation and age, tumour (T), metastasis (M), stage, and patients’ tumour status, while HOXA6 hypermethylation is correlated with age and presence of KRAS mutation. Similarly, with HOXA2 genes, HOXA5 and HOXA6 have not been studied in detail in CRC. In other malignancies, such as NSCLC, HOXA5 is hypermethylated[35] and a separate study showed that low HOXA5 expression indicates unfavourable prognosis and reduces cell proliferation by via p21 expression[36]. Our TCGA analyses revealed downregulation of HOXA5 and HOXA6 by 0.58 and 0.62 fold, respectively. The relationship between downregulation of HOXA5 and HOXA6 with CRCs patients prognosis is the subject for further research. HOXA5 also plays a role in haematopoietic differentiation, whereby HOXA5 is hypermethylated in the development of acute myeloid leukaemia (AML)[37]. Additionally, HOXA6 hypermethylation was reported in oral cancer [38] and more recently in meningiomas[39]. We observed the hypomethylation of HOXA3 and HOXA4 among our patients, which is in disagreement with other CRC studies [31, 32]. It is unclear what causes the discrepancies and it will be worthwhile to reconfirm this finding in a larger cohort. By looking at the gene expression of HOXA3 and HOXA4 in CRC, several studies partly support our findings. For instance, Zhang and colleagues[40] demonstrated that HOXA3 expression is increased in both CRC tissues and cell lines. Their analysis of the relationship between HOXA3 and tumour progression has revealed that elevated HOXA3 expression is linked with poor survival rates in CRC. On the other hand, our finding on HOXA4 hypomethylation is also partially supported by Bhatlekar and colleagues whereby HOXA4 is found to be overexpressed in CRC[41]. They further demonstrated that overexpression of HOXA4 encourages self-renewal, leading to the overabundance of colon cancer stem cell[42], which play an essential role in the metastasis and relapse of this disease. Taken together, HOXA3 and HOXA4 hypomethylation, as identified from our study, may play an important role in CRC. To the best of our knowledge, hypomethylation of the long noncoding RNA HOXA-AS3 and HOXA10-HOXA9 readthrough has never been reported before. Therefore we are the first to notice such observation in CRC. HOXA10- HOXA9 readthrough represents a naturally occurring read-through transcription between the HOXA10 and neighbouring HOXA9 and is a candidate for nonsense- mediated mRNA decay (NMD), which does not produce any protein product. Published literature regarding this readthrough is severely lacking. On the other hand, HOXA-AS3 has been gaining more attention from cancer researchers. In lung cancer, HOXA-AS3 expression was significantly increased and inhibition of HOXA-AS3 impairs cancer cell proliferation, migration, and invasion[43]. In vitro experiment further supported its oncogenic role, where the A549-derived xenografts with silenced HOXA- AS3 had significantly reduced tumour weights and volumes. These findings suggest the potential application of HOXA-AS3 inhibition as an effective targeted therapy for lung cancer patients[43]. The authors also concluded that the upregulated HOXA-AS3 expression was shown to be caused by histone acetylation, and the link between histone deacetylation and DNA methylation has been established[44]. Furthermore, HOXA-AS3 confers resistance towards cisplatin treatment via interaction with HOXA3 Landscape Of HOXA... 9 in NSCLC[45]. In glioma, HOXA-AS3 upregulation promotes tumour progression and predicts poor prognosis[46]. It is probable that the hypomethylation of HOXA-AS3 in our CRC patients could lead to its increased expression. It will be interesting to validate this observation and investigate its function in CRC. An interrogation HOXA-associated oncogenes or tumour suppressors as prospective mechanisms as predictive biomarkers may offer novel therapeutic strategies for treating cancers[47], including CRC. Yet, the obstacle lays in the fact that our knowledge and understanding of HOXA genes in the context of CRC are still insufficient. In this study, we further extended the understanding of CRC pathology by investigating the methylome landscape of HOXA genes. Nevertheless, our study is not without limitation. While our sample size is rather small, the hypo- and hypermethylation of the HOXA genes reported in this study are relevant to carcinogenesis as reported in several studies. For future study, validation of HOXA methylation changes in cancer tissues from a larger cohort is necessary, and the association with survival and other clinicopathological data is warranted. Furthermore, an integrated analysis with gene expression data will be of importance to further establish the correlation between HOXA methylation and gene regulation. Conclusion Using the latest methylation microarray platform, we report a detailed, unbiased landscape of HOXA genes methylome and discovered epigenetically regulated candidate genes in CRC carcinogenesis. Specifically, our results provide the primary evidence that aberrant methylation of HOXA2, HOXA3 and HOXA4 in Malaysian CRC. The new knowledge from this study can be utilized to further increase our understanding of CRC methylomics, particularly on the homeobox A genes. The prognostic and diagnostic roles of the differentially methylated HOXA genes warrant future investigations. Author Contributions MI and RB performed the lab experiments, data analysis and manuscript writing. LT- HT, L-HL and NS-AM provided vital guid- ance for the project and improvement of the writing. The project was conceptualised by NS-AM. Conflict of interest The authors declare that there is no conflict of interest in this work. Acknowledgement This work was fully supported by Universiti Kebangsaan Malaysia under Research University Grant Scheme (GUP-2018-070 Geran Universiti Penyelidikan). References 1. Bray F, Ferlay J, Soerjomataram I, et al., Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2018; 68(6): 394– 424. 2. Azizah AM, Nor Saleha IT, Noor Hashimah A, et al., Malaysian national cancer registry report 2007-2011. 2016, Putrajaya, Malaysia: National Cancer Institute. 3. Ishak M, Baharudin R, Rose IM, et al., Genome-wide open chromatin methylome profiles in colorectal cancer. Biomol 2020; 10(5): 719– 719. 4. Chow YP, Yunos RIM, Rose IM, et al., Characterization of somatic mutations in Malaysian luminal breast cancer. Prog Microbes Mol Biol 2018; 1(1). 5. Atlas TCG, Comprehensive molecular characterization of human colon and rectal cancer. Nature 2012; 487(7407): 330–337. 6. Abdul S-N, Ab Mutalib N-S, Sean KS, et al., Molecular characterization of somatic alterations in Dukes’ B and C colorectal cancers by targeted sequencing. Front Pharmacol 2017; 8: 465–465. 7. Yusof AM, Tieng FYF, Muhammad R, et al., In-depth characterization of miRNome in papillary thyroid cancer with BRAF V600E mutation. Prog Microbes Mol Biol 2020; 3(1). 8. Yunos R-IM, Ab Mutalib N-S, Khor SS, et al., Whole exome sequencing identifies genomic alterations in proximal and distal colorectal cancer. Prog Microbes Mol Biol 2019; 2(1). 9. Werner RJ, Kelly AD, and Issa J-PJ, Epigenetics and Precision Oncology. Cancer J 2017; 23(5): 262–269. 10. Wang Y, Chen P-M, and Liu R-B, Advance in plasma SEPT9 gene methylation assay for colorectal cancer early detection. World J Gastrointest Oncol 2018; 10(1): 15–22. 11. Rodrigues MFSD, Esteves CM, Xavier FCA, et al., Methylation status of homeobox genes in common human cancers. Genomics 2016; 108(5-6): 185–193. 12. Holland PW, Booth HA, and Bruford EA, Classification and nomenclature of all human homeobox genes. BMC Biol 2007; 5: 47–47. 13. Pradhan MP, Desai A, and Palakal MJ, Systems biology approach to stage-wise characterization of epigenetic genes in lung adenocarcinoma. BMC Syst Biol 2013; 7(1): 141–141. 14. Jaffe AE, Murakami P, Lee H, et al., Bump hunting to identify differentially methylated regions in epigenetic epidemiology studies. Int J Epidemiol 2012; 41(1): 200–209. 15. Wang C-C, Su K-Y, Chen H-Y, et al., HOXA5 inhibits metastasis via regulating cytoskeletal remodelling and associates with prolonged survival in non-small-cell lung carcinoma. PLoS One 2015; 10(4): e0124191–e0124191. 16. Hwang J-A, Lee BB, Kim Y, et al., HOXA11 hypermethylation is associated with progression of non-small cell lung cancer. Oncotarget 2013; 4(12): 2317–2325. 17. Bai Y, Fang N, Gu T, et al., HOXA11 gene is hypermethylation and aberrant expression in gastric cancer. Cancer Cell Int 2014; 14(1): 79–79. 18. Cheng W, Jiang Y, Liu C, et al., Identification of aberrant promoter hypomethylation of HOXA10 in ovarian cancer. J Cancer Res Clin Oncol 2010; 136(8): 1221–1227. 19. Wu Q, Lothe RA, Ahlquist T, et al., DNA methylation profiling of ovarian carcinomas and their in vitro models identifies HOXA9, HOXB5, SCGB3A1, and CRABP1 as novel targets. Mol Cancer 2007; 6: 45–45. 20. Xylinas E, Hassler MR, Zhuang D, et al., An epigenomic approach to improving response to neoadjuvant cisplatin chemotherapy in bladder cancer. Biomol 2016; 6(3). 21. Li D, Bai Y, Feng Z, et al., Study of promoter methylation patterns of HOXA2, HOXA5, and HOXA6 and its clinicopathological characteristics in colorectal cancer. Front Oncol 2019; 9(394). 22. Tian Y, Morris TJ, Webster AP, et al., ChAMP: updated methylation analysis pipeline for Illumina BeadChips. Bioinf 2017; 33(24): 3982– 3984. 23. Maksimovic J, Gordon L, and Oshlack A, SWAN: Subset- quantile within array normalization for illumina infinium HumanMethylation450 BeadChips. Genome Biol 2012; 13(6): R44– R44. 24. Ritchie ME, Phipson B, Wu D, et al., limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015; 43(7): e47–e47. 25. Wettenhall JM and Smyth GK, limmaGUI: a graphical user interface for linear modeling of microarray data. Bioinformatics 2004; 20(18): 3705–3706. 26. Giannakis M, Mu XJ, Shukla SA, et al., Genomic correlates of immune-cell infiltrates in colorectal carcinoma. Cell Rep 2016; 15(4): Ishak M et al. 10 857–865. 27. Rauch T, Wang Z, Zhang X, et al., Homeobox gene methylation in lung cancer studied by genome-wide analysis with a microarray- based methylated CpG island recovery assay. Proc Natl Acad Sci U S A 2007; 104(13): 5527–5532. 28. Ooki A, Maleki Z, Tsay J-CJ, et al., A panel of novel detection and prognostic methylated DNA markers in primary non-small cell lung cancer and serum DNA. Clin Cancer Res 2017; 23(22): 7141–7152. 29. Nelson HH, Marsit CJ, Christensen BC, et al., Key epigenetic changes associated with lung cancer development: Results from dense methylation array profiling. Epigenetics 2012; 7(6): 559–566. 30. Vymetalkova V, Vodicka P, Pardini B, et al., Epigenome-wide analysis of DNA methylation reveals a rectal cancer-specific epigenomic signature. Epigenomics 2016; 8(9): 1193–1207. 31. Shiao SPK, Xiao H, Dong L, et al., Genome wide DNA differential methylation regions in colorectal cancer patients in relation to blood related family members, obese and non-obese controls — A preliminary report. Oncotarget 2018; 9(39): 25557–25571. 32. Hongwei Wu QZ, Aberrantly Methylated CpG Island Detection in Colon Cancer. J Proteom Bioinf 2014; s9. 33. Li H-P, Peng C-C, Chung IC, et al., Aberrantly hypermethylated Homeobox A2 derepresses metalloproteinase-9 through TBP and promotes invasion in Nasopharyngeal carcinoma. Oncotarget 2013; 4(11): 2154–2165. 34. Heller G, Babinsky VN, Ziegler B, et al., Genome-wide CpG island methylation analyses in non-small cell lung cancer patients. Carcinogenesis 2013; 34(3): 513–521. 35. Kim D-S, Kim M-J, Lee J-Y, et al., Epigenetic inactivation of Homeobox A5 gene in nonsmall cell lung cancer and its relationship with clinicopathological features. Mol Carcinog 2009; 48(12): 1109– 1115. 36. Zhang M-l, Nie F-q, Sun M, et al., HOXA5 indicates poor prognosis and suppresses cell proliferation by regulating p21 expression in non small cell lung cancer. Tumour Biol 2015; 36(5): 3521–3531. 37. Strathdee G, Sim A, Soutar R, et al., HOXA5 is targeted by cell- type-specific CpG island methylation in normal cells and during the development of acute myeloid leukaemia. Carcinogenesis 2007; 28(2): 299–309. 38. Xavier FCA, Destro MFdSS, Duarte CME, et al., Epigenetic repression of HOXB cluster in oral cancer cell lines. Arch Oral Biol 2014; 59(8): 783–789. 39. Galani V, Lampri E, Varouktsi A, et al., Genetic and epigenetic alterations in meningiomas. Clin Neurol Neurosurg 2017; 158: 119– 125. 40. Zhang X, Liu G, Ding L, et al., HOXA3 promotes tumor growth of human colon cancer through activating EGFR/Ras/Raf/MEK/ERK signaling pathway. J Cell Biochem 2018; 119(3): 2864–2874. 41. Bhatlekar S, Addya S, Salunek M, et al., Identification of a developmental gene expression signature, including HOX genes, for the normal human colonic crypt stem cell niche: overexpression of the signature parallels stem cell overpopulation during colon tumorigenesis. Stem Cells Dev 2014; 23(2): 167–179. 42. Bhatlekar S, Viswanathan V, Fields JZ, et al., Overexpression of HOXA4 and HOXA9 genes promotes self-renewal and contributes to colon cancer stem cell overpopulation. J Cell Physiol 2018; 233(2): 727–735. 43. Zhang H, Liu Y, Yan L, et al., Increased levels of the long noncoding RNA, HOXA-AS3, promote proliferation of A549 cells. Cell Death Dis 2018; 9(6): 1–14. 44. Irvine RA, Lin IG, and Hsieh C-L, DNA methylation has a local effect on transcription and histone acetylation. Mol Cell Biol 2002; 22(19): 6689–6696. 45. Lin S, Zhang R, An X, et al., LncRNA HOXA-AS3 confers cisplatin resistance by interacting with HOXA3 in non-small-cell lung carcinoma cells. Oncogenesis 2019; 8(11). 46. Wu F, Zhang C, Cai J, et al., Upregulation of long noncoding RNA HOXA-AS3 promotes tumor progression and predicts poor prognosis in glioma. Oncotarget 2017; 8(32): 53110–53123. 47. Chen S-L, Qin Z-Y, Hu F, et al., The role of the HOXA gene family in acute myeloid leukemia. Genes 2019; 10(8): 621–621. Landscape Of HOXA...