Identification and characterization of putative methylation targets in the MAOA locus using bioinformatic approaches

Epigenetics. Author manuscript; available in PMC 2011 September 8.

Published in final edited form as:

Epigenetics. 2010 May 16; 5(4): 325–342.

Published online 2010 May 5.

PMCID: PMC3169210

NIHMSID: NIHMS318203

Identification and characterization of putative methylation targets in the MAOA locus using bioinformatic approaches

Elena Shumay and Joanna S. Fowler

Brookhaven National Laboratory, Medical Department, Upton, NY 11973, USA

Corresponding author.

Elena Shumay: eshumay/at/bnl.gov; Joanna S. Fowler: fowler/at/bnl.gov

The publisher's final edited version of this article is available free at Epigenetics

See other articles in PMC that cite the published article.

Abstract

Monoamine oxidase A (MAO A) is an enzyme that catalyzes the oxidation of neurotransmitter amines. A functional polymorphism in the human MAOA gene (high- and low- MAOA) has been associated with distinct behavioral phenotypes. To investigate directly the biological mechanism whereby this polymorphism influences brain function, we recently measured the activity of the MAO A enzyme in healthy volunteers. When found no relationship between the individual's brain MAO A level and the MAOA genotype, we postulated that there are additional regulatory mechanisms that control the MAOA expression. Given that DNA methylation is linked to the regulation of gene expression, we hypothesized that epigenetic mechanisms factor into the MAOA expression. Our underplaying assumption was that the differences in an individual's genotype play a key role in the epigenetic potential of the MAOA locus and, consequently, determine the individual's level of MAO A activity in the brain. As a first step towards experimental validation of the hypothesis, we performed a comprehensive bioinformatic analysis aiming to interrogate genomic features and attributes of the MAOA locus that might modulate its epigenetic sensitivity. Major findings of our analysis are the following: (1) the extended MAOA regulatory region contains two CpG islands (CGIs), one of which overlaps with the canonical MAOA promoter and the other is located further upstream; both CGIs exhibit sensitivity to differential methylation. (2) The uVNTR's effect on the MAOA's transcriptional activity might have epigenetic nature: this polymorphic region resides within the MAOA's CGI and itself contains CpGs, thus, the number of repeating increments effectively changes the number of methylatable cytosines in the MAOA promoter. An array of in silico analyses (the nucleosome positioning, the physical properties of the local DNA, the clustering of transcription-factor binding sites) together with experimental data on histone modifications and Pol 2 sites and data from the RefSeq mRNA library together suggest that the MAOA gene might have an alternative promoter. Based on our findings, we propose a regulatory mechanism for the human MAOA according to which the MAOA expression in vivo is executed by the generation of tissue-specific transcripts initiated from the alternative promoters (both CGI-associated) where transcriptional activation of a particular promoter is under epigenetic control.

Keywords: human MAOA gene, epigenetic regulation, DNA methylation, epigenetic potential, computational analysis

Introduction

Monoamine oxidase A (MAO A) is a flavin-containing enzyme that resides in the outer mitochondrial membrane, it catalyzes the oxidation of the neurotransmitter amines including norepinephrine, serotonin, and dopamine, thereby regulating biogenic amine tone ¹. The enzyme's medical- and biological-importance is well supported by the proven efficacy of the MAO inhibitor drugs in treating depression ² and recent evidence suggesting that patients with major depressive disorder (MDD) have significantly elevated levels of brain MAO A ³. Moreover, it was demonstrated that deleting the MAOA gene entails profound biochemical (decreased biogenic amine metabolism) and behavioral consequences (increased aggression) in a rare human family ⁴ and in the Maoa knockout mice ⁵ substantiating the relevance of this gene in psychiatric disorders. The landmark discovery of a functional uVNTR polymorphism upstream of the MAOA ⁶ stimulated many human studies reporting association with distinct structural- and behavioural phenotypes (reviewed in ⁷, though there are exceptions ^{8, 9}). Some studies also presented evidence for gene × environment-interactions, showing that the MAOA polymorphism apparently influences the risk for development of antisocial behaviour by altering susceptibility to social (childhood maltreatment ^{10, 11}) or chemical stressors (prenatal nicotine exposure, ¹²). Though the MAOA gene is being considered as a biomarker for certain behavioural- and psychiatric-phenotypes ¹³, our understanding of the biological mechanisms by which the variations in the gene sequence might affect gene transcription and the formation of the gene product (MAO A) and thus, to modulate individual susceptibility to environmental stressors (and exposures) remains incomplete.

The uVNTR polymorphism comprises of 2-, 3-, 3.5-, 4-, and 5-copies of a repeated consensus of 30 nucleotides ⁶. Alleles of the 4 and 3 repeats are the most common; in Caucasian males, they occur in approximately a 2:1 ratio and are referred to as the high- and low-MAOA genotypes, respectively. The uVNTR polymorphism is considered as a marker of the MAOA functional regulation because gene fusion and transfection experiments demonstrated its effect on transcriptional activity ⁶. Mechanistically, the MAOA polymorphism might influence behavioral manifestations if this sequence variation influences the production of the MAO A enzyme and, consequently, affects the rate of metabolic oxidation of biogenic amines -neurotransmitters in the brain. To test this postulate, we recently measured the levels of the MAO A enzyme in the brains of healthy male volunteers using positron emission tomography (PET) with a radiotracer specific for MAO A ([¹¹C]clorgyline). We found that while brain MAO A activity vary greatly among individuals showing normal distribution in the population sample tested, it does not show correlation with the MAOA polymorphism ¹⁴. Our in vivo finding was in line with previously reported lack of correlation between the MAOA polymorphism and expression levels or enzyme activity revealed in the study of post-mortem human brain samples ⁹.

Given the discordance between the MAOA genotype and brain MAO A activity and considering the growing evidence for gene-environment interactions, we hypothesized that the epigenetic factors, specifically DNA methylation, might be involved in controlling this gene expression. The role of DNA methylation in the gene expression is well established ¹⁵; however, the methylation changes often are viewed as causing diseases or malfunctions, and consequently, most epigenetic studies focus on detecting aberrant methylation events or epimutations. In contrast to this traditional view, we hypothesize that the variations in the DNA methylation of the MAOA gene might be a factor contributing to its regulation. Under this assumption, genetic and epigenetic factors interact to regulate the MAOA expression whereby the methylation pattern is influenced by individual sequence thus, differentially impact the MAOA's transcription rate. Thus, normal distribution of brain MAO A activity that we saw in our healthy subjects ¹⁴ reflects normal distribution of the individual MAOA transcription rates and further, considerable epigenetic heterogeneity of individual genomes.

There are indirect- and direct-evidence that suggests an involvement of epigenetic mechanisms in the MAOA regulation. For example, MAO A activity changes during the human lifespan; it is high in embryos and low in adults ^{16, 17}, suggesting that its transcription rate is modulated by age-specific factors. The MAOA gene localized on the X-chromosome, and therefore in females, one of the MAOA alleles has to be silenced to ensure dosage compensation ¹⁸; it is well established that the X-chromosome inactivation relay on the DNA methylation ¹⁹. Epigenetic mechanisms are likely to be a major force that drives abundant MAO A activity in the placenta ²⁰, whereas erasure of methyl groups from the DNA of placental cells ^{21, 22} lifts the transcription constraints that could be imposed by methylation in somatic cells. The same mechanism can explain high MAO A activity in mature sperm cells: hypomethylated state of the MAOA gene is expected, as these cells display the lowest methylation level compared to all other tissues in all genomic categories ²³. Thus, the proposed model of the MAOA regulation agrees well with observation that the “null” methylation pattern associates with active transcription of the respective gene. A link between the methylation of the MAOA promoter and the abundance of transcripts was demonstrated by analysis of mRNA-expression from autopsy samples of human brains ²⁴. This study provided first clinical evidence that the extent of allelic MAOA methylation is highly variable (in females) and that the DNA methylation regulates the MAOA ²⁴. Another clinical study corroborated this evidence by establishing a correlation between the MAOA methylation status with nicotine- and alcohol-dependence in women ²⁵.

Considering the above arguments and the notion that the local DNA methylation status referred to as an indicator of epigenetic state of a genomic region ²⁶ we chose to analyze the MAOA promoter methylation. We sought to make use of existing evidence and to further it by evaluating the methylation quantitatively, using recently developed advanced techniques. As a first step towards testing of our hypothesis; we conducted a comprehensive bioinformatic analysis aiming to identify methylation-sensitive regions in the MAOA locus, to assess their methylation status and to evaluate possible functional consequences of differential methylation.

In the mammalian genome DNA methylation predominantly occurs on cytosine in context of the 5′-CpG-3′ dinucleotides ²⁶. Sparsely distributed genomic CpGs are mostly methylated (about 80%); whereas CpG sites accumulated within relatively small stretches of DNA called CpG islands (CGIs) are predominantly hypomethylated or unmethylated. CGIs are frequently colocalizing with gene's transcription start sites (TSSs) where they are associated with transcriptionally competent open chromatin structure ^27,28. The fundamental feature of the epigenetic marks is their dynamics which is a basic mechanism of environmental adaptability ²⁷. The changes in methylation status of the CpG islands (CGIs) overlapping gene promoter enable epigenetic control of transcription ^{29, 30}. At the same time, established methylation pattern has to be maintained with high fidelity since aberrant methylation events (hypo- and hyper-methylation) lead to dys-regulation of the downstream gene.

An assessment of the methylation potential of the locus ultimately entails an identification of CGI's boundaries. The CGIs have been a subject to rigorous investigation for several decades, however, some of the basic characteristics of CGIs remain unknown and there is no common definition for the CGIs. Initially discovered experimentally as “methylation-free zones of DNA” ³¹, CGIs are currently defined by sequence-based-, or mathematical-criteria ^{31, 32,33}. At this point experimentally-derived methylation maps are still fragmental ³⁴. Active investigations aiming to complete such maps are ongoing ³⁵, but progress in this direction is relatively slow. A significant drawback impeding the progress is that depending on the analytical tool or experimental approach used for the analysis, estimates for the number and character of methylated CpG islands markedly vary from study to study ³⁶. Moreover, the results of experimental mapping of the CpG islands are strongly contingent upon specific experimental conditions, such as properties of the cell line and its origin.

The demand to determine genomic regions sensitive to differential methylation was met by numerous bioinformatic approaches that were developed to predict CGIs ³⁷. Most epigenetic studies employ the CpG annotation of the UCSC Genome Browser (CpGIS) that is based on Gardiner and Frommer's algorithms³¹ and is using modified Takai and Jones criteria ³³ (G+C content ≥ 55%, CpG O/E (observed versus expected) ≥ 0.65 and length ≥ 500 bp). The CpGIS algorithm is very conservative and its parameters and constrains are highly stringent, for that reasons it likely is missing some functional CGIs, for example those which span is shorter and CpG O/E (observed/expected) ratio and G+C content is lower than the ad hoc thresholds ³⁸. Novel computational methods to predict CGIs differ from the traditional ones in their use of cut-offs for length, GC %, and CpG density, therefore, aiming to improve confidence in predicting CGI's boundaries, we applied several different computational approaches to analyze the sequence of the MAOA promoter.

Because methylation directly targets the DNA molecule ²⁷, a CpG's susceptibility or resistance to methylation might be determined by sequence characteristics. Several computational approaches infer the methylation status of the CGI from combinatorial analysis of particular genomic attributes ³⁷. Reliably predicting whether a particular CGI is constitutively methylated or unmethylated ³⁹, these programs can not adequately estimate non-binary heterogeneous methylation-patterns ⁴⁰ in human genome where DNA methylation patterns are exceedingly heterogeneous, and vary from tissue-to-tissue and depending on developmental stage ⁴¹.

According to our hypothesis, we expect that non-binary methylation of the MAOA's CGIs manifests in continuous distribution of the MAO A's level in population. To gain a theoretical support for the hypothesis, we analyzed a number of CGI's attributes apposite to their functional state, including DNA sequence patterns, repeats and transcription-factor binding sites (TFBS) and evaluated co-occurrence of regulatory moieties such as well-positioned nucleosomes, genomic insulators, and the physical properties of the local DNA structure. We assessed available empirical data pertinent to the epigenetic state of the region, such as data on histone modifications and DNase I Hypersensitive sites (DNase I HS) maps arguing that this experimentally derived information would complement the in silico analyses.

Our analysis supports the hypothesis on epigenetic sensitivity of the MAOA locus. We specified several targets that could be differentially affected by methylation, including: two bona fide CGIs, a potentially polymorphic GC-rich region of tandem repeats, a GC-rich insulator located in 5′ (in trans) and two DNAse 1 hypersensitive regions in 3′ to the MAOA. We also present theoretical evidence that brings to life a novel perspective on complexity of the MAOA regulation.

Results

Genomic region under analysis (ROI)

Basic assumption of our hypothesis is interaction of genetic-and epigenetic factors to control the MAOA expression, therefore we consider the uVNTR region of the MAOA gene as a pivotal component of the ROI that adds to the methylation potential of the locus. This approach contrasts the conventional ones used by most methylation studies in that most studies exclude repetitive DNA elements ²¹. Indeed, analysis of the repeats is technologically challenging ²¹, but given the established connections between repeats, chromatin environment and gene regulation in general ⁴² and the functional significance of the uVNTR polymorphism of the MAOA in particular ⁶, we thought that our approach is most appropriate. For these reasons, instead of analysing 500 bp or 1 kb 5′ flanks to transcription start sites (TSSs) as elsewhere ⁴³, in this study we defined the ROI (the extended MAOA promoter) as a sequence that spans for ~2 kb upstream of the annotated MAOA TSS. Additional arguments support this rationale: Firstly, CG-dense sequences upstream of the TSS are not confined to its immediate vicinity but stretch throughout the putative region of the extended promoter and secondly, the statistical analysis of overrepresented motifs in human promoters ⁴⁴ defined human putative promoter regions as sequences that correspond to -2000 to +1000 bp relative to the TSS. It is also important to mentioned that under this strategy, the uVNTR polymorphism which resides ~1200 bp (-) to TSS is within the boundaries of the extended MAOA promoter (Figure 1).

Figure 1

The human MAOA gene

Identification of CGIs in the MAOA locus

Bona-fide CGIs

We first used a strategy that affords high specificity screening for identifying transcriptionally competent CpG islands, “bona-fide CGIs” ⁴⁵. Using traditional sliding window method, this software accounts for a number of the sequence features pertinent to its epigenetic state. Integrating genomic and epigenetic information this program aims to identify functional CGIs and to estimate their strength, that is, a score assigned by the program to the CGI measures its propensity to resist methylation. Using this program, we found that two bona fide CGIs reside within the MAOA regulatory region (Figure 2A, CGI_1059 and CGI_1060 shown as blue bars). The bona fide CGIs demonstrated precise positional overlap with the two CpG islands predicted by the traditional conservative approach ³³ (Figure 2A, green bars), albeit their boundaries are broader. The refinement of CGI's boundaries is important since the length of the CGIs is a determining parameter of its functionality ³⁸ and ensures that the functionally important sequences featuring epigenetic properties are not missed.

Figure 2

CpG islands mapping

Detecting CpG islands based on the distance between neighboring CpGs

Another algorithm, the CpGcluster (CGC), analyzes the physical distance between neighboring CpGs dinucleotides on the chromosome to identify the positions of CpGs-dense regions ⁴⁶. The CpGcluster's determined regions exhibit statistically significant overlap with the TSS of known genes ⁴⁶, affording high accuracy and low rate of false-positive predictions. In contrast to the sliding window methods^{31, 32, 35} computational efficiency of CpGcluster is independent from the window and step sizes, as well as from the other traditional parameters. Because a minimum-length threshold is not required, CpGcluster can find short but fully functional CGIs usually missed by other algorithms. We retrieved data on an upstream sequence of the MAOA from the CpGcluster repository (Figure 2B). The program predicted four CGC, three of which (19. 48 and 18) are projected to the bona fide CGI regions, while the forth (CGC 7) was mapped farther upstream. The epigenetic properties of the CGC 7 (multiple CCGC and CACC patterns and 13 occurrences of CpG nucleotides) render this region as a possible target for the methylation analysis.

CG dinucleotide clusters

The CG-clustering algorithm is based on the statistical properties of a sequence ⁴⁷. This mathematical approach is not constrained by conventional criteria for CpG islands (G+C content and observed/expected CG dinucleotide ratio); rather, its predictions rely on the genome-wide decay of CG dinucleotide content, and the preservation of CG-density in restricted regions, wherein the number of CG dinucleotides and the length of the genomic fragment are linearly related. Comparative analysis of two resources, CG clusters versus CpG islands, revealed the former's superior performance as a predictive algorithm for identifying promoters and regions of hypomethylated or incompletely methylated DNA ⁴⁷. Employing CG clusters annotation to assess the 5′- MAOA region (Figure 2C), we found that the CG cluster spans over 2000 bp (chrX: 43,269,861-43,272,055).

Prediction of CpG islands associated with promoters (CGIs)

The structure of the CGIs located over the TSS (start CGIs) differ from that of the no-start CGIs ⁴⁸: the start CGIs are longer and display a greater CpGo/e ratio and G+C level than no-start CGIs. The CpGProD program focuses on prediction of promoter-overlapping CGIs; for our ROI the CpGProD computation produced a single hit: a start-CGI, 2522 nucleotides long.

Prediction of biologically functional CGIs

Based on cumulative mutual information of physical distances between two neighboring CpGs, the CpG_MI tool aims to identify functional CGIs ³⁸. This program's predictions are in good correspondence with experimental detection of unmethylated CGIs ⁴⁹ and active histone modification marks. The CpG_MI's scan of the MAOA promoter predicts one long functional CGI (chrX:43398527-43400764).

Despite different methodologies and assumptions, the results of the various tests that we applied to specify the CGIs in the MAOA promoter are largely concordant and correspond well with our initial postulation; thus ensuring that at least (-) 2 kb from the MAOA TSS region carry epigenetic regulatory potential. Herein, we are using two bona fide CGIs as reference in terms of genomic coordinates.

Estimation of the CGI's methylation status

We began this constituent of the study from predicting the methylation status of the MAOA CGIs using computational means. The “bona fide” algorithm is commonly used by programs predicting CGIs methylation status as a reference ^{50, 38}. The score assigned by the program to each predicted CGI accounts for quantitative differences in characteristics determining methylation sensitivity, and thus, reflects the intrinsic tendency of the CGI to maintain the unmethylated state ⁴⁵. All genomic CGIs are thus grouped into four sets as B1 (0–0.33), B2 (0.33–0.50), B3 (0.50–0.67) and B4 (0.67–1), whereby CGIs with combined epigenetic scores >0.5 represent the ‘bona fide’ CGIs that are strongly associated with epigenetic regulatory function. The two MAOA's CGIs have scores 447 and 621 for CGI_1060 and CGI_1059 respectively, suggesting that both CGIs are “bona fide” (since the scores are not absolute, we consider the 447 score for the CGI_1060 as approaching 500). We analyzed several different features referred to as germane to methylation ⁴⁵ querying the CGIs sequences (Table 1) and found that epigenetic features of these CGIs are apparently different suggesting their distinct propensity to constitutional methylation.

Table 1

Genomic Features of CGIs in the regulatory region of the MAOA gene

The MethCGI classifier that predicts methylation status of CGIs using a support vector machine (SVM) based on nucleotide sequence contents and TFBSs, performs best estimating a CGI status for brain tissue ⁵¹. This program renders the promoter-overlapping CGI+1059 methylation-prone and the CGI_1059 methylation-resistant.

Another indicator of methylation-resistant CGI is over-representation of the 4-mer CCGC in its sequence ⁴⁵. We found multiple occurrences of this 4-mer in the sequences of both CGIs (11 and 9 motifs in the CGI_1059 and CGI_1060, respectively). Some additional information might also be taken into consideration, for example, it was demonstrated that the occurrence of particular transcription factor binding sites (TFBS) protects CpG islands from de novo methylation ⁵². One of such TF is Sp1. A sensitivity of the MAOA transcription to Sp1 stimulation is well-established ⁵², and, indeed, in addition to the two Sp1 sites in the vicinity of the core promoter, we found an array of the Sp1 consensuses in the distal to the TSS region corresponding to the CGI_1059 which suggests that the formal is methylation-resistant. (The role of the TFBSs in the MAOA regulation will be discussed below in more details).

Even though there are inconsistencies in details of predictions by different computations, overall, the results are congruent in that the CGI_1060 is likely methylation-resistant. However, medium scores in prediction of the CGIs methylation status is suggestive of their sensitivity to various different factors determining methylation and thus, we expect that in vivo, in individual genomes, the MAOA's methylation is heterogeneous.

We next assessed available experimental data on methylation status of the MAOA locus. Based on MeDIP technology (methylated DNA immunoprecipitation) ²¹ information on the tissue-specific DNA methylation is integrated into the Ensembl Browser. These data provide an estimate of absolute methylation levels based on MeDIP signals. For the MAOA region, however, the data is limited to methylation profile of sperm cells. The sperm cells are generally hypomethylated, therefore this source of these data poses significant limitation to their value. Indeed, it is rather expected that the region upstream of the MAOA TSS (~-2 kb - +50) in sperm cells was tested as either unmethylated or intermediately methylated. This finding however, could not be extrapolated to predict the methylation pattern in cells of different origin.

Recent paper reported results of genome-scale DNA methylation mapping of clinical samples ³⁴. The accompanied the paper dataset contains data on bisulfite sequencing measures of DNA methylation in clinical samples. Using this resource, we procured the methylation profiles for our ROI in normal tissues (Figure 2D). As evident from the comparison of the methylated cytosines detected in normal colon and blood cell DNA methylation in the MAOA promoter region is tissue-specific: there are several sites that were tested positive in the colon cells and negative on the blood cells (Figure 2D, the red circles show methylation-free sites in blood cells). The technology used in the study affords detection of epimutations on a genomic level, but has limited specificity that is needed for the analysis of a particular genomic target. In addition, the assay's coverage does not include repetitive regions and, consequently, there is no data for the uVNTR region (Figure 2D, green box).

Potential impact of the repeated sequences on the MAOA sensitivity to epigenetic regulation

The CGI_1060 encompass canonical MAOA TSS thus, fulfils the definition of the promoter-overlapping CGI. The second CGI of the MAOA regulatory region, the (CGI_1059) stretches over the uVNTR, which modulates the MAOA promoter activity ⁵³. We noted that the repeated increment of this polymorphism is a 30 bp unit (ACCGGCACCGGCACCAGTACCCGCACCAGT), wherein the three occurrences of the CG dinucleotides represent three potential methylation targets. Thus, depending on the VNTR allele, combined methylation potential of the CGI_1059 in individual genome will vary as a function of the VNTR length where an exponent physically adds three CpGs to the number of methylatable sites.

In addition, we located another region of tandem repeats located immediately upstream of the CGI_1059 (Figure 1, circled red). Its 140 bp sequence (chrX:43398629-43398768) is exceptionally GC--dense (123 out of 140 nucleotides are C+G!), and contains 18 CpG dinucleotides. Inherent instability of repeated genomic elements is well-recognized ⁵⁴, therefore we thought to find out whether this repeating sequence exhibits propensity to polymorphism. The VARScore server affords such analysis ⁵⁴. The score assigned by the program to the sequence (0.9897) reflects its high instability and, consequently, there is a high probability that this region is polymorphic. Considering both, the location of this repeating region (within the CG cluster) and its potential polymorphism, it seems plausible that the sequence variability of the region might affect the total number of methylatable CpGs in the MAOA regulatory region, thus emphasizing the need to include this region in analysis of the MAOA methylation.

Chromatin-based regulators in the MAOA locus

The DNA's methylation status and the chromatin-regulatory machinery are interconnected and influence each other ⁵⁵.

Histone- modification variants detected for the MAOA locus

The chromatin maps (ENCODE Consortium) are built based on the results of in vitro ChIP seq experiments using two cell lines (GM12878 and K562). We found that in both cell lines the positive for acetylated histones sequences located upstream of the core MAOA promoter (Figure 3A). These two cell lines, however, differ in distribution of other active chromatin marks, such as H3K4me2 and H3K4me3 as it is evident from the histone-methylation profiles (Figure 3B). While in the K562 cells the H3K4me2 and H3K4me3-positive marks are mirroring the acetylated histones, in the GM12878 cells significantly longer DNA sequence covering two CGIs was positive for H3K4me2 and H3K4me3 marks. The difference in the histone marks boundaries may be due to the distinct roles the specific types of histone modifications play in transcription initiation depending on the cellular environment. Alternatively, they can reflect different rate of the MAOA expression in specific tissues.

Figure 3

Empirical data on histone modifications in the MAOA promoter region

The formation of an active transcription-initiation complex entails sequence-specific recruitment of RNA polymerase II ⁵⁶ and accordingly, the experimental ChIP data have proven that RNA-Polymerase II is often present at the major transcriptional start-site ⁵⁷. Our inspection of the MAOA locus revealed that the Pol II sites were mapped only to the distal CpG island (CGI_1059) (Figure 3A, B, red arrow).

Nucleosome-exclusion regions

Genomic DNA is wrapped around histone proteins and packaged into repeating subunits of chromatin, called nucleosomes that are linearly arrayed along the DNA ⁵⁸. In vivo, DNA interaction with histone proteins governs chromatin condensation ^59-61. Constitutively active genes have nucleosome-free promoters where open state of the DNA molecule makes it accessible for transcription factors to initiate the transcriptional event ⁶². In contrast, transcription of inducible genes require eviction of well-positioned nucleosomes from the promoters ⁶³. A potential interconnection between the epigenetic regulators and nucleosome occupancy stems from the observations that the methylation status of CpG dinucleotides in the genomic region modulates nucleosome strength ⁶⁴ and nucleosome density; and equally, nucleosome occupancy in the gene-regulatory region inversely correlate with an expression of a cognate gene ⁶⁵.

Several computational algorithms based on DNA-sequence properties and information on in vivo nucleosome locations have been developed to predict nucleosome positioning ⁵⁹. We used a custom track which annotates nucleosome-exclusion regions (NXRegions, ⁶⁶) to extract information for our ROI (Figure 4A). NXScores for the human gene promoters are positively correlated with gene-expression level, although this trend is reversed at very high NXScores (above 540) ⁶⁶. The peaks that are mapped to the CGIs (dashed boxes) have nearly identical scores that correspond to the established maximum for human TSSs. We note that the initiation sites of both, canonical- (AK223499) and long- (BC044797) MAOA transcripts are projected to the middle of NXScore peaks.

Figure 4

Regulatory sequences in the MAOA locus

Next, we wanted to know whether the length of the uVNTR will ultimately affect chromosome positioning in the region. This analysis is afforded by the sequence-based prediction of nucleosome mapping (FineStr 1.0, ⁶⁷). Querying of the sequences corresponding to the uVNTR with 3, 3.5 and 4 repeated increments revealed that not only the nucleosome centre is differentially placed in the cognate sequences, but the actual sequence composition affects nucleosome strength as well (Figure S1). Further, we noted that the novel region of tandem repeats upstream of the CGI_1059 that we described earlier, resides precisely in the middle of the NX peak (scores at 846). The score this high, as mentioned, points to a low probability for constitutive transcriptional initiation from this site due to the reversed trend; at the same time, it emphasizes that this region is highly sensitive to stimulation-induced transcription initiation.

DNase I hypersensitive sites (DNase I HS)

On the genomic scale, DNA methylation correlates with DNase 1 hypersensitivity ⁶⁸. DNase I HS result from the loss or remodeling of one or more nucleosomes, (see above) which renders DNA fragments susceptible to enzymatic cleavage. In the MAOA promoter positions of the DNase I HS sites vary depending on a cell line (Figure S2A). Notably, for the most cell lines, DNase I HS sites are mapped to the CGI_1059 whilst nuclease accessible sites in the CGI_1059 were detected only in CD34(-) lymphocytes. Even though in most human genes, DNase I HS site(s) are confined to the promoter region, our inspection of the MAOA locus revealed that two DNase I HS sites are also mapped to the 3′-end of the gene (Figure S2B). The DNAse 1 HS sites in 3′ of the MAOA, similarly to those in the promoter, are CG-dense and contain CpG dinucleotides (intragenic DNAse HS site has one CpG and the DNAse 1 HS located in 3′ has five CpGs) (Figure S2C). Importantly, the DNAse 1 HS site located downstream of the MAOA gene also contains an insulator site (see below). It was demonstrated that aberrant DNA methylation occurring in 3′ of genes results in an abrogation of the blocking activity of the insulator sites ⁶⁹ and is accompanied by dysregulation of gene expression. Considering this observation, we suggest that the methylation status of the MAOA 3′ the DNAse 1 HS sites might be an important determinant of the combined MAOA's epigenetic potential.

Cis- and Trans-regulatory elements of the MAOA gene

Methylation-sensitive transcription-factor binding sites (TFBS) in the MAOA regulatory region

Identification of the TFBSs in the gene regulatory region is important component of the methylation analysis because of two reasons. First, as we mentioned earlier, a distinct characteristic of constitutively unmethylated CGIs is enrichment in representation of specific TFBS ³⁶. Second, the most frequent consequence of DNA methylation is to change the affinity of TF binding-site sequences and, thus, to attenuate sensitivity to stimulation-induced transcription. Thus, to estimate a potential impact of the methylation state of the promoter on the MAOA expression, we set up to identify among the TFBSs that reside in the MAOA regulatory sequence those which motifs contain the methylatable sites. We used the JASPAR database ⁷⁰, to analyse the ROI sequence and found that along with the others, the Sp1, MZF1_1-4, CREB1, NFIC, YY and GATA2 motifs are over-represented. The above-mentioned TFs were selected as putatively functionally significant: the consensuses of these TFs contain at least one methylatable CpG dinucleotide (Additional File 1). We noticed that the Sp1 sites are not confined to the core promoter region as was reported previously ^{71, 72} but scattered over the entire ROI forming major cluster about at (-) 1.5 -1.8 kb to TSS that is projected to the CGI_1060 (Figure S3). We next used the CisRed program ⁷³ to determine the boundaries of the regulatory region. This software computes the nominal ‘search region’, or the CisRed region, accounting for TFBSs motifs distribution. The CisRed region predicted for the MAOA, spans over both CGIs (Figure S4).

An important parameter of TF-mediated gene expression is combinatorial interactions of TFs where the underlying requirement for the TF interactions is genomic clustering of TF binding sites. We thought that the query for enrichment of predicted transcription-factor binding sites would help to validate the boundaries of the genomic region with regulatory potential. To detect statistically significant clusters of TF binding sites in the ROI, we applied the Cluster-Buster program ⁷⁴, and identified a cluster in the region spanning (-) 1591 bp to (-) 1178 bp relative to the MAOA TSS (Figure S5 shows details of the cluster). The PReMod database ⁷⁵ affords prediction of a transcription-regulatory module within the region under evaluation by detecting TFBSs clusters and assigning of a “module score”, while accounting for several other factors including the length and CG content of the ROI. Analysis of the MAOA sequence using PReMod program produced results similar to that of the CisRed. According to the PReMod, the regulatory module of the MAOA is a region overlapping the distal end of the CGI_1059 (chrX:43269871-43270446, Figure S5B).

Structural properties of the DNA in the 5′ region of the MAOA gene

The physical properties of DNA, such as its deformability and propensity to form structures other than double-helical B-DNA DNA, are important genomic characteristics. Locus-specific three-dimensional organization of DNA into higher-order structures function determines local molecular interactions whereby directly controls expression of a cognate gene. The tendency of a DNA molecule to fold into an unusual structure is sequence-dependent; for instance, G-rich sequences can adopt a G-quadruplex (G4) structure both in vivo and in vitro ⁷⁶. The functional significance of these motifs elicits from the findings that G4s are enriched in key chromosomal regions, and that they are highly conserved ⁷⁶. Located in the promoter region, G4 structures act as cis-regulatory elements ⁷⁷, modulating gene regulation bidirectionally. That is, G-quadruplex formed upstream of the gene can either enhance its transcription ⁷⁸ or silence it ⁷⁹. We reasoned that the high density of CG nucleotides in the MAOA promoter might affect physical properties of the local DNA and lead to formation of quadruplex motifs. Using the nucleotide-sequence-analysis module of the QGRS Mapper server ⁸⁰, we evaluated our ROI for its propensity to form G4 structures and found that at least 16 non-overlapping QGRS (Quadruplex-forming G-Rich Sequences) are predicted for this region (Figure S6). A graphic representation of the G4 regions shows that they form two clusters that positionally overlap with two respective CGIs. We noted that the uVNTR region is located between two G-quadruplex cluster regions; the tandem repeats immediately follow the G4 motif (GGCGGCACCGGCACCGG) with the highest (18) score, suggesting that the number of repeats as well as it methylation state may affect the quadruplex generation.

G-quadruplex motifs or QGRSs can be formed by the RNA molecule as well; indeed, such structures are detected in vivo⁷⁶. Occurring in the UTR of the RNA, G-quadruplex motifs add to regulation of gene expression ⁷⁶. Furthermore, while propensity to adapt non-B conformation in DNA is counteracted by the presence of the complementary strand ⁸¹, such counterforce does not act in case of the single-stranded RNA molecule and therefore influence of quadruplex motifs on gene expression manifests stronger on the level of transcripts ⁸¹. Hypothesizing that this mechanism is likely engaged in the MAOA regulation, we tested sequences of several MAOA mRNAs for propensity to form QGRSs with QGRS Mapper and found that the number of predicted QGRSs is largely determined by the length of the transcript, such that mRNAs BC044787.1 (long), AK223499.1 (medium) and M68840.1 (short) are to form 19, 12 and 11 QGRS, respectively.

Insulator sites

Insulators prevent distal regulatory elements, such as enhancers, from activating non-target genes, thus protecting them from negative positional effects and the influences of the chromosomal environment ⁸². In vertebrates, the versatile transcription-regulator CCCTC-binding factor (CTCF) is the only trans-acting factor that confers enhancer-blocking insulator activity. CTCF-bound insulators separate transcriptionally active and silent chromatin domains, whilst their function depends heavily on the local status of DNA methylation and chromatin modifications ⁸³. Position of the MAOA on the X-chromosome merits importance of functional insulators in the locus since this gene belongs to the 20% of X-linked genes that escaped inactivation (dosage compensation in females). Insulators support the active status of specific genes on the X-chromosome while maintaining the silencing of adjacent ones. At the level of chromosomal domains, those boundary elements demarcate transcriptionally active regions and block heterochromatization during X-inactivation. Investigation of the genes escaping inactivation revealed that their 5′ ends contain a DNA fragment with insulator properties and CTCF-binding sites ⁸³.

Considering relevance of this mechanism to the MAOA regulation, we set up to identify specific-sequence elements that could promote the establishment of discrete insulator moieties. First, we screened the MAOA locus for experimentally validated insulator-sites using the CTCFBSDB, i.e., - a CTCF-binding site database; three such sites are mapped in trans with respect to the MAOA gene (Figure 4B, yellow circle). We noticed that H3K27me3-positive spikes mark the margins of the insulator region (Figure 4B, enlarged fragment), suggesting that this region readily is accessible to interactions with regulatory elements, such as TFs. We then used another feature of the CTCFBSDB database, a CTCFBS prediction tool, to identify possible insulator motifs within the extended promoter region of the MAOA gene; this query produced two additional targets (Figure 4B, lower panel). Contained within the MIT_LN23 sequence, CG dinucleotides (highlighted in red) indicate that the methylation of these cytosine residues potentially could attenuate the function of this insulator site. To further clarify the role of insulators in MAOA regulation we analyzed the details of this genomic region using the experimental data on the CTCF that are integrated in the Ensemble Genome Browser. Visualization of the CTCF peaks revealed clusters of high-scoring peaks within the ~2000 bp of the annotated TSS of the MAOA, with a major cluster mapped to a sequence between the CGI_1059 and CGI_1060, indicating that it might function to block transcription initiated from the distant TSS (Figure S7).

Phylogenetic analysis of the MAOA regulatory sequence: GC-rich transcription elements emerged in primate lineage

Highly conserved genomic elements are represented by the sequences shared by different species; therefore, evolutionary conservation analysis has been used to predict genome's regulatory regions ⁸⁴. High conservation of the coding sequences of the MAOA gene over large evolutionary distances reflects the ubiquitous biological function of the MAO A enzyme. In sharp contrast, the regulatory sequences of the MAOA gene exhibit very limited conservation (Vertebrate Multiz Alignment and Conservation, 17 species, UCSC Browser). It is important to emphasise here that the changes in the sequence are accompanied by increase in the GC content (Maoa promoter has GC content 0.49 (mouse genome, mm9), whereas in human MAOA promoter has 0.59 (human genome, hg18). This observation raises a possibility of emergence of evolutionally recent GC-rich functional and structural transcriptional elements. Intrigued by the observation, we wanted to get more insight on potentially conserved functional cis-regulatory elements of the MAOA implementing advanced computational means.

The PAZAR database integrates several data collections, transcription-factor databases and annotations of regulatory sequences ⁸⁵. One of the PAZAR's annotations, the ‘Pleiades genes’, utilizes the Pleiades Promoter Project, i.e., the library of short regulatory sequences that drive gene expression in defined brain regions, hence, it is particularly relevant for our analysis. The stretch of about 2 kb in 5′ to the MAOA TSS was identified as “a Pleiades gene region” (Figure 5A, top) emphasizing the pivotal role of the entire region in the brain-specific expression of the MAOA and suggesting that the extended promoter region represents a regulatory block.

Figure 5

Conservation of the MAOA regulatory region

Another PAZAR's annotation, the ORKA toolkit, enables to compute a conservation profile for the ROI based on phylogenetically retained functionally important sites. First, we applied a multiple species analysis and built a TFs binding profile for the MAOA regulatory region. This analysis identified several high conservation regions (Figure 5A, ORKA conservation spikes). Therein, the conservation score of these peaks accounts for the number of orthologous conserved regions across genomes. Mapped to the promoter-overlapping CGI, a dense cluster of high-conservation peaks projecting to the MAOA coding sequence; this region was also identified as a conserved region (turquoise block, indicated by blue arrow). Contrasting this, the core MAOA promoter show very limited conservation and the distal regulatory region corresponding to the CGI_1059 has only four single peaks. Limited phylogenetic conservation of the human MAOA regulatory region is not absolute - the human MAOA promoter resembles the one of primate ⁸⁶. The MAOA loci in genomes of these two species display pronounced sequence similarities and, consequently, have similar nucleotide content (hg18-GC is 0.59; panTro2-GC is 0.58). To compare regulatory features of the primate and the human MAOA promoters, we aligned the orthologous sequences of the MAOA gene of humans and primates for a pair-wise comparison (Figure 5B). High concordance of these two sequences is evident from the broad conservation spikes and blocks reflecting high conservation scores. However, we noted that the two regions of tandem repeats (uVNTR and a novel region we described earlier) are projected to the conservation gaps (blue arrows), suggesting that these repeating sequences are unique for human genome.

The MAOA mRNA isoforms

Transcription of the MAOA gene generates multiple different mRNAs; nine out of ten known mRNAs are alternatively spliced variants, whereas one form is unspliced (AceView ⁸⁷). These mRNAs differ in the truncation of their 5′ – and the 3′-ends, the presence or absence of a cassette exon, and overlapping exons with different boundaries. GeneBank/dbEST contains data on almost five hundred transcripts of different origin, verifying the expression of the MAOA gene in most normal- tissues and pathological ones, as well as throughout developmental stages. The existence of multiple transcripts raises the question whether the unique biological features of the particular transcript serve to fulfil distinct functional needs. The functional diversity of the RNA transcripts is furthered at the post-transcriptional level; the specifics of this diversification are determined by both the nucleotide sequence and the secondary structure of the mRNA. Variations in the transcripts' length is a key factor that determines their stability; therefore, we calculated the thermodynamic ensemble for three MAOA mRNAs using the RNAfold web server, finding that the transcripts' stability is positively correlated with their length (Figure S8).

As we mentioned earlier, G-rich DNA sequences prone to non-B-conformation. This property is also known for RNA molecule, and the formation of four-stranded RNA structures with stabilities resembling their DNA counterparts is well established ⁸¹. In human mRNAs, RNA quadruplexes are prevalent in 5′-UTRs ⁸¹. Running QGRS Mapper for three MAOA's mRNAs, we found that the transcript length is a primary predictor of potential RNA quadruplex formation in vivo, hence, their potential stability (Figure S8).

In silico prediction of the MAOA promoter

Different computational approaches that we used to analyze genomic features of the regulatory region of the MAOA consistently pointed to the presence of an alternative promoter overlapping with the CGI_1059. To gather more evidence reinforcing this finding, we next accessed the prediction-based positions of putative promoters for the MAOA via UCSC Browser interface (Figure S9). Most of the programs that use an initio and de novo predictions (such as the AUGUSTUS where a generalized hidden Markov model (GHMM) is used to model coding and non-coding sequences, translation start and end) consistently hint to a promoter with TSS within the boundaries of the CGI_1059. In contrast, the programs that are based on conservation (such as SGP gene prediction that relay on mouse/human homology) detected only one promoter corresponding to the canonical MAOA promoter. This, however, is rather expected, since the region predicted by the SGP as promoter projects to a peak of phylogenetic conservation (Figure S9). Next, we used several programs to analyze the ROI sequence, which consistently predicted multiple promoters, TSS and enhancer sites. Thus, prediction by the Promoter Scan ⁸⁸ produced two hits and the Promoter 2 ⁸⁹ predicted four putative TSS, all with the score corresponding to the marginal prediction category. Based on recognition of human PolII promoter region and start of transcription predictions by SoftBerry ⁹⁰ resulted in four and five hits, respectively. Oddly, a single hit obtained by the FPROM (Human promoter prediction) application (SoftBerry) falls in the CGI_1060.

The absence of the TATA-box in the MAOA promoter suggests importance of alternative functional promoter elements that could compensate for the TATA-box enabling the recruitment of basic transcriptional machinery and initiation of the MAOA transcription. The CCAAT motif (a C/EBP (CCAAT enhancer-binding protein) binding site) represent the functional elements that can engage Pol II; the CCAAT consensus often found about 75-80 bases upstream of the respective TSS ⁹¹. By scanning the sequences of the MAOA-associated CpGs, we noticed that CGI_1059 contains two C/EBR binding sites at the functional position related to the proposed alternative promoter activity, whereas CGI_1060 lacks them. We also found that the regulatory sequence of the MAOA contains one M22 (TGCGCANK) motif immediately upstream of the distal CpG island (Figure 6A, red arrow); this atomic motif was recently identified as characteristic regulatory feature of TATA-less human genes ⁹².

Figure 6

Promoter prediction with CoreBoost_HM

The CAAT-box represents GC-rich structural genomic element recognizes by NF1 transcription factor. NF1 sites distribution in human genome correlates with gene distribution ⁹³, where these sites are accumulating around TSS ⁹³. Mapping of these sites to our ROI revealed that multiple NF1 sites in the MAOA promoter form detectable cluster around canonical TSS as well as within the CGI_1060 boundaries (not shown).

Lastly, we took advantage of a novel computational method for high-resolution promoter prediction - CoreBoost_HM ⁹⁴. This program integrates DNA sequence features with epigenetic information to identify RNA polymerase II core-promoters in the human genome. The CoreBoost analyzes CpG and non-CpG promoters separately because of their distinct sequence features and detected histone modifications ⁹⁵. Hence, we run the program two times, where consecutive queries treated the MAOA promoter first as CpG and then as non-CpG. We used the UCSC interface to visualize the results (Figure 6). The characteristic profile of predicted promoter (two major spikes with a gap between them corresponding to the putative TSS) was projected to the CGI_1059 (Figure 6B, green dashed box), whereas only minor spike corresponds to the MAOA core promoter region (Figure 6B, blue dashed box). Even though it was shown that in case when these two types of promoters are mixed together, the pattern specific to CpG-related promoters dominates ⁹⁵, in our case, both set ups of the query produced similar results (spikes in the regions on the predicted alternative MAOA promoter are higher) suggesting that potential role of this region as transcription initiator is not contingent on MAOA expression level ⁹⁴.

The juxtaposition of the CoreBoost-predicted promoter with the putative initiator motif M22 (Figure 6A and Figure 6B, red arrows) demonstrated the precise positional overlap of the M22 motif with the first spike of the predicted promoter strengthening putative functionality of the former.

Discussion

A comprehensive in silico analysis of epigenetic properties of the human MAOA gene revealed new aspects of the epigenetic potential of this gene, previously unknown associations between different regulatory moieties and lead to some unexpected findings. This analysis was conceived to test our hypothesis suggesting that epigenetic mechanisms are involved in the control of the activity of the human MAOA gene. Unarguably, the validation of the hypothesis necessitates experimental testing that affords direct measurement of the methylation state of this genomic region. There are numerous methods to detect and to quantify methylation; nevertheless all of the currently available assays have some limitations and drawbacks. For these reasons, we sought that the sequence-based analysis might add to comprehensive assessment of epigenetic potential of the region. In fact, the thorough bioinformatic analysis has its own merits; for example, a notable advantage of the in silico analysis in that it is independent of the detection methods used ⁴⁵. Sequence-based analysis is neither contingent upon the nature of the biological sample used for the analysis, specifically, the cell type, tissue of origin, differentiation state etc., nor is influenced by transient nature of methylation state in vivo.

We performed this analysis as a precede to the experimental testing of the MAOA methylation in primary genomic samples arguing that its results would help to properly design experiments, to minimize possible experimental bias, and thus, to improve validity of the experimental data.

Our findings support the original hypothesis of involvement of epigenetic mechanisms in regulation of the human MAOA gene. The principal findings of the analysis are the following: Firstly, the MAOA locus features multiple genomic targets exhibiting sensitivity to differential DNA methylation. Secondly, the uVNTR polymorphic region contributes to the epigenetic potential of the MAOA allele. Thirdly, the methylation status of the MAOA regulatory region in individual human genomes likely varies, whereby individual variations in the gene sequence might affect variations in methylation profile. Lastly, site-specific conditions, modulated by the epigenetic marks, can favor transcriptional initiation from one of multiple TSSs of the MAOA, so generating transcripts of different lengths, and thus adding a factor to the tissue-specific levels of the MAO A enzyme.

Proposing pivotal role of epigenetic factors in regulation of the MAOA expression, we are constrain this paradigm to the human and the primate MAOA gene. The comparative genomics approach is widely used to identify functional sequence elements and regulatory networks ^{96, 97}. Accordingly, in our study we employed both comparative and evolutionary analyses but their findings were unexpected. We found that the regulatory sequences of the MAOA lack conservation (Figure 5). Moreover, it turned out that the evolution of the regulatory MAOA sequence was accompanied by gene conversion and results in considerable accumulation of G+C nucleotides (GC content in the mouse Maoa promoter is 0.49 contrasting the 0.59 in human MAOA). These findings require some explanation. Analysis of phylogenetic patterning suggested that variations in the CG content are associated with the emergence of the evolutionally novel GC-rich structural and functional transcription elements ⁹⁸. In this view, GC-bias conversion of the MAOA promoter enables the action of additional regulatory mechanisms (epigenetic by nature) in human and primates. In addition, whilst the mouse Maoa promoter is non-CGI, the chimp MAOA promoter is CGI-overlapping which speaks to the possible involvement of epigenetic mechanisms in the MAOA regulation in primates and human lineage.

We established that the regulatory region of the MAOA gene spans over (-) 2 kb relative to the annotated TSS. This whole region was consistently marked as methylation-sensitive and contains CGIs. Interestingly, that the concept of “CpG Island shores” that was introduced in a recent paper ⁹⁹ proposed the region about 2 kb in 5′ of TSS as functionally significant.

Established role of gene promoter methylation in expressional regulation ²⁶, prompted us explore properties of the MAOA promoter in more details. Housekeeping genes ¹⁰⁰, including the MAOA gene, commonly have a CGI-associated promoter and a ‘spike’ of GC content in the vicinity of the TSS. The characteristics of the CGI_1060 fulfil these criteria. Intrigued by detection of an additional CGI_1059 further upstream and by the fact that this CpG island encompasses uVNTR polymorphism with known effect on transcriptional rate ⁶, we explored genomic features of this region and found that it might represent an additional promoter for the MAOA. Implementation of several computational approaches including analysis of TF clusters, refinement of the boundaries of CisRed region (Figure S4) and nucleosome positioning (Figure 4A), and analysis of physical properties of DNA (Figure S2) – all these analyses consistently indicated the presence of an additional region with promoter characteristics upstream of the canonical MAOA promoter.

We performed several computational tests using promoter-searching programs that implement sequence-based predictions and predictions based on integration of different genomic annotations. Various experimental data, namely, ENCODE data on histone modifications and RNA polymerase II- binding sites collaborate our finding. The fact that the human MAOA transcript BC044787 initiates about 1.5 kb upstream of the MAOA TSS (initiation site is within the CGI_1059, Figure S6, red arrow) demonstrates in vivo utilization of the alternative MAOA promoter.

Most mammalian genes have multiple promoters containing multiple TSSs ¹⁰¹, thus the possibility of alternative promoters for the MAOA gene does not put this gene apart. An important aspect of our finding is that it offers new regulatory paradigm for the MAOA gene since the putative promoters carry distinctly different information about gene expression, and therefore, may exhibit different sensitivity to the regulatory signals. Contrasting most human promoters where only one of alternative gene promoters is CGI-associated ¹⁰², both putative alternative promoters of the MAOA are CGI-promoters, suggesting ultimate role of methylation in regulation of the MAOA. Intriguingly, the larger size of the distal CGI_1059 taken in the context of the observation that the CGIs overlapping with TSS are much longer comparing with these in other genic environments ¹⁰³, point to the dominance of this site as transcription initiation site.

The alternative MAOA promoters have distinguishing properties, including occurrence and distribution of TFBSs, profiles of histone variants, and regional DNA folding, which can reflect their functionally distinct characteristics. For instance, we found dense nucleosome deposition directly upstream of the core TSS, while, in contrast, a large region upstream of the distal promoter is nucleosome-depleted (Figure 4A), pointing to the potential of these regions to initiate transcription under basic conditions. Also, Pol2 was detested only at the distal promoter region (Figure 3, red arrows), suggesting that this promoter might maintain gene expression in a poised state ¹⁰⁴.

The concept of the alternative promoter for the MAOA gene offers additional possible biological mechanisms to explain functional effect of the uVNTR polymorphism. The uVNTR region directly affects mRNA transcripts initiated from the alternative MAOA determining their length as a function of increment. Thus, we suggest that compositional characteristics of the MAOA promoter that encompasses uVNTR polymorphic region profoundly influence stability of the cognate mRNAs, for example due to differential stability (higher number of G4 structures was predicted for the longer transcripts). A similar regulatory mechanism has been demonstrated for the human insulin gene (ILPR). The minisatellite (INS VNTR [MIM 125852]) upstream to the insulin gene promoter is related to gene's transcriptional activity and to the susceptibility for insulin-dependent diabetes mellitus (IDDM) (reviewed in ¹⁰⁵ and ¹⁰⁶). Studies of the mechanisms which drive VNTR's effect on the gene expression had shown that the G-rich strand of the VNTR ILPR adopts an intra-molecularly folded hairpin G-quartet structure where the number of repeats defines the nature of DNA folding and regional stability ¹⁰⁶.

The concept of the alternative MAOA promoter has several implications. For example, it is apparent that the transcription initiation from the distal MAOA promoter would result in production of longer mRNAs that are thermodynamically more stable than shorter ones. Stability of the longer transcripts entails their effective translation into protein ¹⁰⁷. In addition, since most predicted localization motifs that enable intracellular trafficking of the mRNAs are mapped to 5′ UTR of the transcripts ¹⁰⁸, it is feasible that the longer transcripts would be more efficient in enabling targeting of the translational process to specific subcellular compartments, for instance, to the vicinity of neuronal terminals. In addition, well-established modulating effect of environment on the MAOA expression might be exerted by the switch in transcription initiation, given that the alternative promoter is CGI-overlapping.

Conceptually, our hypothesis on the involvement of epigenetic mechanisms in the MAOA regulation differs from the ones investigated in the other studies. Currently, most of the experimental and clinical studies investigate the role of DNA methylation in human diseases (see, for example a review ¹⁰⁹). As a result, a solid link between the aberrant epigenetic marks and the diseases has been established (reviewed in ¹¹⁰). Less numerous research report an association of genomic polymorphisms and altered epigenetic status of a gene and only single ones present evidence supporting a role of DNA methylation in the regulation of normal gene expression ^{111, 108}. For example of the former is uncovered regulatory mechanism of the glucocorticoid receptor gene: Turner and others ¹¹² showed that individualized epigenetic pattern of glucocorticoid receptor promoter which orchestrates the expression of this gene is influenced by the sequence variations. Akin, Candiloro and Dobrovic discovered that particular genotypes of the MGMT gene are associated with its methylation in healthy individuals ¹¹³. We foresee that, in a similar way, genetic variations in the MAOA promoter (including uVNTR) might be a contributing factor to the methylation state of the region. Considering these observations, we strongly suggest that the practical methylation analysis ultimately have to include sequencing of the primary genomic samples. Knowing the genetic variations in the samples will enable control for confounding effects due to heterogeneity of individual's genomes.

The emerging complex model of MAOA regulation offers a new perspective on the developmental dynamics and individual variations in the MAOA gene-expression, and ultimately, on regulation of the MAO A enzymatic activity in the human brain. We propose a novel regulatory paradigm for the MAOA gene in which its expression is governed in a highly coordinated fashion by the epigenetic mechanisms and the local transcriptional machinery, and is carried out by generation of tissue-specific alternative transcripts initiated from the alternative promoters. We will use the results of the analysis to guide our experimental investigation of the methylation status of the MAOA locus in individual genomes. Based on the findings, we will then explore an association between the MAOA methylation and the phenotype assessed by the level of the MAO A enzyme in the brain.

Conclusions

In this study we have applied an array of in silico analyses which suggest that MAOA expression in vivo is executed by the generation of tissue-specific transcripts initiated from the alternative promoters (both CGI-associated) where transcriptional activation of a particular promoter is under epigenetic control. However, since the study is based on an in silico analysis, the validation of its results ultimately requires experimental analysis. A key point here is that even if experimental analysis will establish differential methylation of the MAOA promoter in individual DNA samples, this would not be sufficient to assert a causative relationship with MAOA expression in vivo. Thus the merits of using the methylation status of the MAOA as a biomarker for the MAOA expression necessitate the establishment of its correlation with the phenotype.

This study was conceived to help appropriately design experiments on methylation analysis of the MAOA gene, and we believe that its results and conclusions allow us to do so. Our strategy for experimental methylation analysis will include deep sequencing of the entire ROI - about 2 kb (-) MAOA TSS which will allow to customise design of methylation analysis depending on allele composition and to avoid confounding effect of genetic variations; consideration the regions with potential differential methylation that are outside of the promoter, i.e., insulator (in trans, 5′) and DNase I HS (3′ end of the MAOA); complementing the analyses based on bisulphate conversion of DNA (current “golden standard”) with assays used other techniques, such as enzymatic restriction and MeDIP to ensure detection of low levels of methylation and to improve validity of the results. We will use the results of experimental methylation profiling to explore the association between the epigenetic status of the MAOA gene and its expression in vivo.

Methods

Table 2 lists the software and programs used in this study. For most analyses we queried 2.7 kb sequence (chrX:43,398,150-43,400,850 in hg17/18 Human Genome Assembly 2006) and used default settings of the respective programs. The use of older versions of the reference sequence is stated. We customized settings and used filters to minimize false-positive predictions of the TFBSs with the JASPAR database ⁷⁰ : only binding sites on the lagging strand were considered and the sites with score below 5 were discounted.

Table 2

Software and programs used in this analysis

Supplementary Material

Additional Table 1

Click here to view.^{(141K, pdf)}

Additional Table 2

Click here to view.^{(19K, pdf)}

Supplementary Figures

Click here to view.^{(262K, pdf)}

Acknowledgments

This work was performed at Brookhaven National Laboratory with infrastructure support for the Department of Energy, Office of Biological and Environmental Research and funded by the National Institute on Drug Abuse, grants KO1 DA025280-01A1 (ES) and K05 DA20001 (JSF). We thank Dr. Ian Craig for helpful discussions, Dr. Jacob Hooker for discussion and his suggestions for the organization of the manuscript and Dr. Avril Woodhead for critical reading and discussion.

Abbreviations


MAO A	monoamine oxidase A

the MAOA	monoamine oxidase gene

VNTR	variable number of tandem repeats

CGI	CpG islands

PET	positron emission tomography

TSS	transcription start site

NXRegions	nucleosome-exclusion regions

QGRS	G-quadruplex structure

Pol 2	polymerase 2

DNase 1 HS	DNase I Hypersensitive sites

ChIP	chromatin immunoprecipitation assay

Footnotes

Competing interests: The authors declare that they have no competing interests.

Authors' contributions: ES initiated the study and performed the analysis and interpretation. JSF participated in writing the manuscript.

References

1. Shih JC, Thompson RF. Monoamine oxidase in neuropsychiatry and behavior. American journal of human genetics. 1999;65:593–8. [PubMed]

2. Stahl SM, Felker A. Monoamine oxidase inhibitors: a modern guide to an unrequited class of antidepressants. CNS spectrums. 2008;13:855–70. [PubMed]

3. Meyer JH, Ginovart N, Boovariwala A, Sagrati S, Hussey D, Garcia A, Young T, Praschak-Rieder N, Wilson AA, Houle S. Elevated monoamine oxidase a levels in the brain: an explanation for the monoamine imbalance of major depression. Archives of general psychiatry. 2006;63:1209–16. [PubMed]

4. Brunner HG, Nelen M, Breakefield XO, Ropers HH, van Oost BA. Abnormal behavior associated with a point mutation in the structural gene for monoamine oxidase A. Science. 1993;262:578–80. [PubMed]

5. Cases O, Seif I, Grimsby J, Gaspar P, Chen K, Pournin S, Muller U, Aguet M, Babinet C, Shih JC, et al. Aggressive behavior and altered amounts of brain serotonin and norepinephrine in mice lacking MAOA. Science. 1995;268:1763–6. [PMC free article] [PubMed]

6. Sabol SZ, Hu S, Hamer D. A functional polymorphism in the monoamine oxidase A gene promoter. Human genetics. 1998;103:273–9. [PubMed]

7. Buckholtz JW, Meyer-Lindenberg A. MAOA and the neurogenetic architecture of human aggression. Trends in neurosciences. 2008;31:120–9. [PubMed]

8. Rosenberg S, Templeton AR, Feigin PD, Lancet D, Beckmann JS, Selig S, Hamer DH, Skorecki K. The association of DNA sequence variation at the MAOA genetic locus with quantitative behavioural traits in normal males. Human genetics. 2006 [PubMed]

9. Balciuniene J, Emilsson L, Oreland L, Pettersson U, Jazin E. Investigation of the functional effect of monoamine oxidase polymorphisms in human brain. Human genetics. 2002;110:1–7. [PubMed]

10. Caspi A, McClay J, Moffitt TE, Mill J, Martin J, Craig IW, Taylor A, Poulton R. Role of genotype in the cycle of violence in maltreated children. Science. 2002;297:851–4. [PubMed]

11. Kim-Cohen J, Caspi A, Rutter M, Tomas MP, Moffitt TE. The caregiving environments provided to children by depressed mothers with or without an antisocial history. The American journal of psychiatry. 2006;163:1009–18. [PubMed]

12. Wakschlag LS, Kistner EO, Pine DS, Biesecker G, Pickett KE, Skol AD, Dukic V, Blair RJ, Leventhal BL, Cox NJ, Burns JL, Kasza KE, Wright RJ, Cook EH., Jr Interaction of prenatal exposure to cigarettes and MAOA genotype in pathways to youth antisocial behavior. Molecular psychiatry. 2009 [PMC free article] [PubMed]

13. Singh I, Rose N. Biomarkers in psychiatry. Nature. 2009;460:202–7. [PubMed]

14. Fowler JS, Alia-Klein N, Kriplani A, Logan J, Williams B, Zhu W, Craig IW, Telang F, Goldstein R, Volkow ND, Vaska P, Wang GJ. Evidence That Brain MAO A Activity Does Not Correspond to MAO A Genotype in Healthy Male Subjects. Biological psychiatry. 2007;62:355–8. [PMC free article] [PubMed]

15. Kagami M, Sekita Y, Nishimura G, Irie M, Kato F, Okada M, Yamamori S, Kishimoto H, Nakayama M, Tanaka Y, Matsuoka K, Takahashi T, Noguchi M, Tanaka Y, Masumoto K, Utsunomiya T, Kouzan H, Komatsu Y, Ohashi H, Kurosawa K, Kosaki K, Ferguson-Smith AC, Ishino F, Ogata T. Deletions and epimutations affecting the human 14q32.2 imprinted region in individuals with paternal and maternal upd(14)-like phenotypes. Nature genetics. 2008;40:237–42. [PubMed]

16. Vitalis Tania, A C, C K, S JC, G P, C O. Developmental expression pattern of monoamine oxidases in sensory organs and neural crest derivatives. 2003:392–403. [PubMed]

17. Nicotra A, Falasca L, Senatori O, Conti Devirgiliis L. Monoamine oxidase A and B activities in embryonic chick hepatocytes: differential regulation by retinoic acid. Cell biochemistry and function. 2002;20:87–94. [PubMed]

18. Gartler SM, Riggs AD. Mammalian X-chromosome inactivation. Annu Rev Genet. 1983;17:155–90. [PubMed]

19. Heard E, Disteche CM. Dosage compensation in mammals: fine-tuning the expression of the X chromosome. Genes & development. 2006;20:1848–67. [PubMed]

20. Auda GR, Kirk SH, Billett MA, Billett EE. Localization of Monoamine Oxidase mRNA in Human Placenta. J Histochem Cytochem. 1998;46:1393–400. [PubMed]

21. Rakyan VK, Down TA, Thorne NP, Flicek P, Kulesha E, Graf S, Tomazou EM, Backdahl L, Johnson N, Herberth M, Howe KL, Jackson DK, Miretti MM, Fiegler H, Marioni JC, Birney E, Hubbard TJ, Carter NP, Tavare S, Beck S. An integrated resource for genome-wide identification and analysis of human tissue-specific differentially methylated regions (tDMRs) Genome research. 2008;18:1518–29. [PubMed]

22. Mann MR, Lee SS, Doherty AS, Verona RI, Nolen LD, Schultz RM, Bartolomei MS. Selective loss of imprinting in the placenta following preimplantation development in culture. Development (Cambridge, England) 2004;131:3727–35. [PubMed]

23. Eckhardt F, Lewin J, Cortese R, Rakyan VK, Attwood J, Burger M, Burton J, Cox TV, Davies R, Down TA, Haefliger C, Horton R, Howe K, Jackson DK, Kunde J, Koenig C, Liddle J, Niblett D, Otto T, Pettett R, Seemann S, Thompson C, West T, Rogers J, Olek A, Berlin K, Beck S. DNA methylation profiling of human chromosomes 6, 20 and 22. Nature genetics. 2006;38:1378–85. [PMC free article] [PubMed]

24. Pinsonneault JK, Papp AC, Sadee W. Allelic mRNA expression of X-linked monoamine oxidase a (MAOA) in human brain: dissection of epigenetic and genetic factors. Human molecular genetics. 2006;15:2636–49. [PubMed]

25. Philibert RA, Gunter TD, Beach SR, Brody GH, Madan A. MAOA methylation is associated with nicotine and alcohol dependence in women. Am J Med Genet B Neuropsychiatr Genet. 2008;147B:565–70. [PubMed]

26. Suzuki MM, Bird A. DNA methylation landscapes: provocative insights from epigenomics. Nature reviews. 2008;9:465–76. [PubMed]

27. Bird AP. CpG-rich islands and the function of DNA methylation. Nature. 1986;321:209–13. [PubMed]

28. Saxonov S, Berg P, Brutlag DL. A genome-wide analysis of CpG dinucleotides in the human genome distinguishes two distinct classes of promoters. 2006:1412–7. [PubMed]

29. Galm O, Herman JG, Baylin SB. The fundamental role of epigenetics in hematopoietic malignancies. Blood reviews. 2006;20:1–13. [PubMed]

30. Esteller M. Cancer epigenomics: DNA methylomes and histone-modification maps. Nature reviews. 2007;8:286–98. [PubMed]

31. Gardiner-Garden M, Frommer M. CpG islands in vertebrate genomes. Journal of molecular biology. 1987;196:261–82. [PubMed]

32. Takai D, Jones PA. Comprehensive analysis of CpG islands in human chromosomes 21 and 22. Proceedings of the National Academy of Sciences. 2002;99:3740–5. [PubMed]

33. Takai D, Jones PA. The CpG island searcher: a new WWW resource. In silico biology. 2003;3:235–40. [PubMed]

34. Gu H, Bock C, Mikkelsen TS, Jager N, Smith ZD, Tomazou E, Gnirke A, Lander ES, Meissner A. Genome-scale DNA methylation mapping of clinical samples at single-nucleotide resolution. Nature methods. 7:133–6. [PMC free article] [PubMed]

35. Eckhardt F, Lewin J, Cortese R, Rakyan VK, Attwood J, Burger M, Burton J, Cox TV, Davies R, Down TA, Haefliger C, Horton R, Howe K, Jackson DK, Kunde J, Koenig C, Liddle J, Niblett D, Otto T, Pettett R, Seemann S, Thompson C, West T, Rogers J, Olek A, Berlin K, Beck S. DNA methylation profiling of human chromosomes 6, 20 and 22. Nature genetics. 2006;38:1378–85. [PMC free article] [PubMed]

36. Straussman R, Nejman D, Roberts D, Steinfeld I, Blum B, Benvenisty N, Simon I, Yakhini Z, Cedar H. Developmental programming of CpG island methylation profiles in the human genome. Nature structural & molecular biology. 2009;16:564–71. [PubMed]

37. Han L, Zhao Z. CpG islands or CpG clusters: how to identify functional GC-rich regions in a genome? BMC bioinformatics. 2009;10:65. [PMC free article] [PubMed]

38. Su J, Zhang Y, Lv J, Liu H, Tang X, Wang F, Qi Y, Feng Y, Li X. CpG_MI: a novel approach for identifying functional CpG islands in mammalian genomes. Nucleic acids research. :38–e6. [PMC free article] [PubMed]

39. Previti C, Harari O, Zwir I, del Val C. Profile analysis and prediction of tissue-specific CpG island methylation classes. BMC bioinformatics. 2009;10:116. [PMC free article] [PubMed]

40. Bock C, Walter J, Paulsen M, Lengauer T. Inter-individual variation of DNA methylation and its implications for large-scale epigenome mapping. Nucleic acids research. 2008;36:e55. [PMC free article] [PubMed]

41. Law JA, Jacobsen SE. Establishing, maintaining and modifying DNA methylation patterns in plants and animals. Nature reviews. advance online publication. [PMC free article] [PubMed]

42. Chapman RD, Heidemann M, Albert TK, Mailhammer R, Flatley A, Meisterernst M, Kremmer E, Eick D. Transcribing RNA polymerase II is phosphorylated at CTD residue serine-7. Science. 2007;318:1780–2. [PubMed]

43. Tabach Y, Brosh R, Buganim Y, Reiner A, Zuk O, Yitzhaky A, Koudritsky M, Rotter V, Domany E. Wide-scale analysis of human functional transcription factor binding reveals a strong bias towards the transcription start site. PLoS ONE. 2007;2:e807. [PMC free article] [PubMed]

44. Marino-Ramirez L, Spouge JL, Kanga GC, Landsman D. Statistical analysis of over-represented words in human promoter sequences. Nucleic acids research. 2004;32:949–58. [PMC free article] [PubMed]

45. Bock C, Walter J, Paulsen M, Lengauer T. CpG island mapping by epigenome prediction. PLoS computational biology. 2007;3:e110. [PubMed]

46. Hackenberg M, Previti C, Luque-Escamilla PL, Carpena P, Martinez-Aroza J, Oliver JL. CpGcluster: a distance-based algorithm for CpG-island detection. BMC bioinformatics. 2006;7:446. [PMC free article] [PubMed]

47. Glass JL, Thompson RF, Khulan B, Figueroa ME, Olivier EN, Oakley EJ, Van Zant G, Bouhassira EE, Melnick A, Golden A, Fazzari MJ, Greally JM. CG dinucleotide clustering is a species-specific property of the genome. Nucleic acids research. 2007;35:6798–807. [PMC free article] [PubMed]

48. Ponger L, Mouchiroud D. CpGProD: identifying CpG islands associated with transcription start sites in large genomic mammalian sequences. Bioinformatics (Oxford, England) 2002;18:631–3. [PubMed]

49. Illingworth R, Kerr A, DeSousa D, Jorgensen H, Ellis P, Stalker J, Jackson D, Clee C, Plumb R, Rogers J, Humphray S, Cox T, Langford C, Bird A. A novel CpG island set identifies tissue-specific methylation at developmental gene loci. PLoS biology. 2008;6:37–51. [PMC free article] [PubMed]

50. Carson MB, Langlois R, Lu H. Mining knowledge for the methylation status of CpG islands using alternating decision trees. Conf Proc IEEE Eng Med Biol Soc. 2008;2008:3787–90. [PubMed]

51. Fang F, Fan S, Zhang X, Zhang MQ. Predicting methylation status of CpG islands in the human brain. Bioinformatics (Oxford, England) 2006;22:2204–9. [PubMed]

52. Boumber YA, Kondo Y, Chen X, Shen L, Guo Y, Tellez C, Estecio MR, Ahmed S, Issa JP. An Sp1/Sp3 binding polymorphism confers methylation protection. PLoS genetics. 2008;4:e1000162. [PMC free article] [PubMed]

53. Guo G, Ou XM, Roettger M, Shih JC. The VNTR 2 repeat in MAOA and delinquent behavior in adolescence and young adulthood: associations and MAOA promoter activity. Eur J Hum Genet. 2008;16:626–34. [PMC free article] [PubMed]

54. Legendre M, Pochet N, Pak T, Verstrepen KJ. Sequence-based estimation of minisatellite and microsatellite repeat variability. Genome research. 2007;17:1787–96. [PubMed]

55. Lindahl Allen M, Koch CM, Clelland GK, Dunham I, Antoniou MN. DNA methylation-histone modification relationships across the desmin locus in human primary cells. BMC molecular biology. 2009;10:51. [PMC free article] [PubMed]

56. Wright JE, Mais C, Prieto JL, McStay B. A role for upstream binding factor in organizing ribosomal gene chromatin. Biochemical Society symposium. 2006:77–84. [PubMed]

57. Tshuikina M, Nilsson K, Oberg F. Positive histone marks are associated with active transcription from a methylated ICSBP/IRF8 gene. Gene. 2008;410:259–67. [PubMed]

58. Henikoff S. Nucleosomes at active promoters: unforgettable loss. Cancer cell. 2007;12:407–9. [PubMed]

59. Jiang C, Pugh BF. Nucleosome positioning and gene regulation: advances through genomics. Nature reviews. 2009;10:161–72. [PubMed]

60. Tirosh I, Barkai N. Two strategies for gene regulation by promoter nucleosomes. Genome research. 2008;18:1084–91. [PubMed]

61. Gal-Yam EN, Jeong S, Tanay A, Egger G, Lee AS, Jones PA. Constitutive nucleosome depletion and ordered factor assembly at the GRP78 promoter revealed by single molecule footprinting. PLoS genetics. 2006;2:e160. [PubMed]

62. Choi JK, Kim YJ. Epigenetic regulation and the variability of gene expression. Nature genetics. 2008;40:141–7. [PubMed]

63. Li B, Carey M, Workman JL. The role of chromatin during transcription. Cell. 2007;128:707–19. [PubMed]

64. Wang JP, Widom J. Improved alignment of nucleosome DNA sequences using a mixture model. Nucleic acids research. 2005;33:6743–55. [PMC free article] [PubMed]

65. Lin JC, Jeong S, Liang G, Takai D, Fatemi M, Tsai YC, Egger G, Gal-Yam EN, Jones PA. Role of nucleosomal occupancy in the epigenetic silencing of the MLH1 CpG island. Cancer cell. 2007;12:432–44. [PubMed]

66. Radwan A, Younis A, Luykx P, Khuri S. Prediction and analysis of nucleosome exclusion regions in the human genome. BMC genomics. 2008;9:186. [PMC free article] [PubMed]

67. Gabdank I, Barash D, Trifonov EN. Nucleosome DNA bendability matrix (C. elegans) Journal of biomolecular structure & dynamics. 2009;26:403–11. [PubMed]

68. Crawford GE, Davis S, Scacheri PC, Renaud G, Halawi MJ, Erdos MR, Green R, Meltzer PS, Wolfsberg TG, Collins FS. DNase-chip: a high-resolution method to identify DNase I hypersensitive sites using tiled microarrays. Nat Meth. 2006;3:503–9. [PMC free article] [PubMed]

69. Hark AT, Schoenherr CJ, Katz DJ, Ingram RS, Levorse JM, Tilghman SM. CTCF mediates methylation-sensitive enhancer-blocking activity at the H19/Igf2 locus. Nature. 2000;405:486–9. [PubMed]

70. Sandelin A, Alkema W, Engstrom P, Wasserman WW, Lenhard B. JASPAR: an open-access database for eukaryotic transcription factor binding profiles. Nucleic acids research. 2004;32:D91–4. [PMC free article] [PubMed]

71. Chen K. Organization of MAO A and MAO B promoters and regulation of gene expression. Neurotoxicology. 2004;25:31–6. [PubMed]

72. Shih JC, Chen K. Regulation of MAO-A and MAO-B gene expression. Current medicinal chemistry. 2004;11:1995–2005. [PubMed]

73. Robertson G, Bilenky M, Lin K, He A, Yuen W, Dagpinar M, Varhol R, Teague K, Griffith OL, Zhang X, Pan Y, Hassel M, Sleumer MC, Pan W, Pleasance ED, Chuang M, Hao H, Li YY, Robertson N, Fjell C, Li B, Montgomery SB, Astakhova T, Zhou J, Sander J, Siddiqui AS, Jones SJ. cisRED: a database system for genome-scale computational discovery of regulatory elements. Nucleic acids research. 2006;34:D68–73. [PMC free article] [PubMed]

74. Frith MC, Li MC, Weng Z. Cluster-Buster: Finding dense clusters of motifs in DNA sequences. Nucleic acids research. 2003;31:3666–8. [PMC free article] [PubMed]

75. Ferretti V, Poitras C, Bergeron D, Coulombe B, Robert F, Blanchette M. PReMod: a database of genome-wide mammalian cis-regulatory module predictions. Nucleic acids research. 2007;35:D122–6. [PubMed]

76. Paeschke K, Simonsson T, Postberg J, Rhodes D, Lipps HJ. Telomere end-binding proteins control the formation of G-quadruplex DNA structures in vivo. Nature structural & molecular biology. 2005;12:847–54. [PubMed]

77. Yadav VK, Abraham JK, Mani P, Kulshrestha R, Chowdhury S. QuadBase: genome-wide database of G4 DNA--occurrence and conservation in human, chimpanzee, mouse and rat promoters and 146 microbes. Nucleic acids research. 2008;36:D381–5. [PMC free article] [PubMed]

78. Schonhoft JD, Bajracharya R, Dhakal S, Yu Z, Mao H, Basu S. Direct experimental evidence for quadruplex-quadruplex interaction within the human ILPR. Nucleic acids research. 2009 [PMC free article] [PubMed]

79. Ou TM, Lu YJ, Zhang C, Huang ZS, Wang XD, Tan JH, Chen Y, Ma DL, Wong KY, Tang JC, Chan AS, Gu LQ. Stabilization of G-quadruplex DNA and down-regulation of oncogene c-myc by quindoline derivatives. Journal of medicinal chemistry. 2007;50:1465–74. [PubMed]

80. Kikin O, D'Antonio L, Bagga PS. QGRS Mapper: a web-based server for predicting G-quadruplexes in nucleotide sequences. Nucleic acids research. 2006;34:W676–82. [PMC free article] [PubMed]

81. Halder K, Wieland M, Hartig JS. Predictable suppression of gene expression by 5′-UTR-based RNA quadruplexes. Nucl Acids Res. 2009;37:6811–7. [PMC free article] [PubMed]

82. Ciavatta D, Kalantry S, Magnuson T, Smithies O. A DNA insulator prevents repression of a targeted X-linked transgene but not its random or imprinted X inactivation. Proceedings of the National Academy of Sciences. 2006;103:9958–63. [PubMed]

83. Filippova GN, Cheng MK, Moore JM, Truong JP, Hu YJ, Nguyen DK, Tsuchiya KD, Disteche CM. Boundaries between chromosomal domains of X inactivation and escape bind CTCF and lack CpG methylation during early development. Developmental cell. 2005;8:31–42. [PubMed]

84. Levy S, Hannenhalli S, Workman C. Enrichment of regulatory signals in conserved non-coding genomic sequence. Bioinformatics (Oxford, England) 2001;17:871–7. [PubMed]

85. Portales-Casamar E, Arenillas D, Lim J, Swanson MI, Jiang S, McCallum A, Kirov S, Wasserman WW. The PAZAR database of gene regulatory information coupled to the ORCA toolkit for the study of regulatory sequences. Nucleic acids research. 2009;37:D54–60. [PMC free article] [PubMed]

86. Inoue-Murayama M, Mishima N, Hayasaka I, Ito S, Murayama Y. Divergence of ape and human monoamine oxidase A gene promoters: comparative analysis of polymorphisms, tandem repeat structures and transcriptional activities on reporter gene expression. Neurosci Lett. 2006;405:207–11. [PubMed]

87. Thierry-Mieg D, Thierry-Mieg J. AceView: a comprehensive cDNA-supported gene and transcripts annotation. Genome Biol. 2006;7 1:S12 1–4. [PMC free article] [PubMed]

88. Prestridge DS. Computer software for eukaryotic promoter analysis. Methods in molecular biology (Clifton, NJ. 2000;130:265–95. [PubMed]

89. Knudsen S. Promoter2.0: for the recognition of PolII promoter sequences. Bioinformatics (Oxford, England) 1999;15:356–61. [PubMed]

90. Solovyev V, Kosarev P, Seledsov I, Vorobyev D. Automatic annotation of eukaryotic genes, pseudogenes and promoters. Genome Biol. 2006;7 1:S10 1–2. [PMC free article] [PubMed]

91. Linhoff MW, Wright KL, Ting JP. CCAAT-binding factor NF-Y and RFX are required for in vivo assembly of a nucleoprotein complex that spans 250 base pairs: the invariant chain promoter as a model. Molecular and cellular biology. 1997;17:4589–96. [PMC free article] [PubMed]

92. Yang C, Bolotin E, Jiang T, Sladek FM, Martinez E. Prevalence of the initiator over the TATA box in human and yeast genes and identification of DNA motifs enriched in human TATA-less core promoters. Gene. 2007;389:52–65. [PMC free article] [PubMed]

93. Khuu P, Ho PS. A rare nucleotide base tautomer in the structure of an asymmetric DNA junction. Biochemistry. 2009;48:7824–32. [PMC free article] [PubMed]

94. Wang X, Xuan Z, Zhao X, Li Y, Zhang MQ. High-resolution human core-promoter prediction with CoreBoost_HM. Genome research. 2009;19:266–75. [PubMed]

95. Barski A. High-resolution profiling of histone methylations in the human genome. Cell. 2007;129:823–37. [PubMed]

96. Gregory SG, Barlow KF, McLay KE, Kaul R, Swarbreck D, Dunham A, Scott CE, Howe KL, Woodfine K, Spencer CC, Jones MC, Gillson C, Searle S, Zhou Y, Kokocinski F, McDonald L, Evans R, Phillips K, Atkinson A, Cooper R, Jones C, Hall RE, Andrews TD, Lloyd C, Ainscough R, Almeida JP, Ambrose KD, Anderson F, Andrew RW, Ashwell RI, Aubin K, Babbage AK, Bagguley CL, Bailey J, Beasley H, Bethel G, Bird CP, Bray-Allen S, Brown JY, Brown AJ, Buckley D, Burton J, Bye J, Carder C, Chapman JC, Clark SY, Clarke G, Clee C, Cobley V, Collier RE, Corby N, Coville GJ, Davies J, Deadman R, Dunn M, Earthrowl M, Ellington AG, Errington H, Frankish A, Frankland J, French L, Garner P, Garnett J, Gay L, Ghori MR, Gibson R, Gilby LM, Gillett W, Glithero RJ, Grafham DV, Griffiths C, Griffiths-Jones S, Grocock R, Hammond S, Harrison ES, Hart E, Haugen E, Heath PD, Holmes S, Holt K, Howden PJ, Hunt AR, Hunt SE, Hunter G, Isherwood J, James R, Johnson C, Johnson D, Joy A, Kay M, Kershaw JK, Kibukawa M, Kimberley AM, King A, Knights AJ, Lad H, Laird G, Lawlor S, Leongamornlert DA, Lloyd DM, Loveland J, Lovell J, Lush MJ, Lyne R, Martin S, Mashreghi-Mohammadi M, Matthews L, Matthews NS, McLaren S, Milne S, Mistry S, Moore MJ, Nickerson T, O'Dell CN, Oliver K, Palmeiri A, Palmer SA, Parker A, Patel D, Pearce AV, Peck AI, Pelan S, Phelps K, Phillimore BJ, Plumb R, Rajan J, Raymond C, Rouse G, Saenphimmachak C, Sehra HK, Sheridan E, Shownkeen R, Sims S, Skuce CD, Smith M, Steward C, Subramanian S, Sycamore N, Tracey A, Tromans A, Van Helmond Z, Wall M, Wallis JM, White S, Whitehead SL, Wilkinson JE, Willey DL, Williams H, Wilming L, Wray PW, Wu Z, Coulson A, Vaudin M, Sulston JE, Durbin R, Hubbard T, Wooster R, Dunham I, Carter NP, McVean G, Ross MT, Harrow J, Olson MV, Beck S, Rogers J, Bentley DR. The DNA sequence and biological annotation of human chromosome 1. Nature. 2006;441:315–21. [PubMed]

97. Di Rienzo A, Hudson RR. An evolutionary framework for common diseases: the ancestral-susceptibility model. Trends in Genetics. 2005;21:596–601. [PubMed]

98. Khuu P, Sandor M, DeYoung J, Ho PS. Phylogenomic analysis of the emergence of GC-rich transcription elements. Proceedings of the National Academy of Sciences of the United States of America. 2007;104:16528–33. [PubMed]

99. Irizarry RA, Ladd-Acosta C, Wen B, Wu Z, Montano C, Onyango P, Cui H, Gabo K, Rongione M, Webster M, Ji H, Potash JB, Sabunciyan S, Feinberg AP. The human colon cancer methylome shows similar hypo- and hypermethylation at conserved tissue-specific CpG island shores. Nature genetics. 2009;41:178–86. [PMC free article] [PubMed]

100. Vinogradov AE. Dualism of gene GC content and CpG pattern in regard to expression in the human genome: magnitude versus breadth. Trends Genet. 2005;21:639–43. [PubMed]

101. Valen E, Pascarella G, Chalk A, Maeda N, Kojima M, Kawazu C, Murata M, Nishiyori H, Lazarevic D, Motti D, Marstrand TT, Tang MH, Zhao X, Krogh A, Winther O, Arakawa T, Kawai J, Wells C, Daub C, Harbers M, Hayashizaki Y, Gustincich S, Sandelin A, Carninci P. Genome-wide detection and analysis of hippocampus core promoters using DeepCAGE. Genome research. 2009;19:255–65. [PubMed]

102. Cheong J, Yamada Y, Yamashita R, Irie T, Kanai A, Wakaguri H, Nakai K, Ito T, Saito I, Sugano S, Suzuki Y. Diverse DNA Methylation Statuses at Alternative Promoters of Human Genes in Various Tissues. DNA Res. 2006;13:155–67. [PubMed]

103. Ioshikhes IP, Zhang MQ. Large-scale human promoter mapping using CpG islands. Nature genetics. 2000;26:61–3. [PubMed]

104. Muse GW, Gilchrist DA, Nechaev S, Shah R, Parker JS, Grissom SF, Zeitlinger J, Adelman K. RNA polymerase is poised for activation across the genome. Nature genetics. 2007;39:1507–11. [PMC free article] [PubMed]

105. Kelly MA, Mijovic CH, Barnett AH. Genetics of type 1 diabetes. Best practice & research. 2001;15:279–91. [PubMed]

106. Breen G, Collier D, Craig I, Quinn J. Variable number tandem repeats as agents of functional regulation in the genome. IEEE Eng Med Biol Mag. 2008;27:103–4. 8. [PubMed]

107. Brockschmidt FF, Nothen MM, Hillmer AM. The two most common alleles of the coding GGN repeat in the androgen receptor gene cause differences in protein function. Journal of molecular endocrinology. 2007;39:1–8. [PubMed]

108. Mignone F, Grillo G, Licciulli F, Iacono M, Liuni S, Kersey PJ, Duarte J, Saccone C, Pesole G. UTRdb and UTRsite: a collection of sequences and regulatory motifs of the untranslated regions of eukaryotic mRNAs. Nucleic acids research. 2005;33:D141–6. [PMC free article] [PubMed]

109. Mueller WC, von Deimling A. Gene regulation by methylation. Recent results in cancer research Fortschritte der Krebsforschung. 2009;171:217–39. [PubMed]

110. Herranz M, Esteller M. New therapeutic targets in cancer: the epigenetic connection. Clin Transl Oncol. 2006;8:242–9. [PubMed]

111. Cheong J, Yamada Y, Yamashita R, Irie T, Kanai A, Wakaguri H, Nakai K, Ito T, Saito I, Sugano S, Suzuki Y. Diverse DNA methylation statuses at alternative promoters of human genes in various tissues. DNA Res. 2006;13:155–67. [PubMed]

112. Turner JD, Pelascini LP, Macedo JA, Muller CP. Highly individual methylation patterns of alternative glucocorticoid receptor promoters suggest individualized epigenetic regulatory mechanisms. Nucleic acids research. 2008;36:7207–18. [PMC free article] [PubMed]

113. Candiloro IL, Dobrovic A. Detection of MGMT promoter methylation in normal individuals is strongly associated with the T allele of the rs16906252 MGMT promoter single nucleotide polymorphism. Cancer prevention research (Philadelphia, Pa. 2009;2:862–7. [PubMed]