Differential Effect of Three Base Modifications on DNA Thermostability

Article pubs.acs.org/ac

Differential Effect of Three Base Modifications on DNA Thermostability Revealed by High Resolution Melting Carlos M. Rodríguez López,† Amanda J. Lloyd, Kate Leonard, and Mike J. Wilkinson*,† Aberystwyth University, IBERS, Institute of Biological, Environmental and Rural Sciences, Aberystwyth, Wales, SY23 3DA S Supporting Information *

ABSTRACT: High resolution melting (HRM) can detect and quantify the presence of 5-methylcytosine (5mC) in DNA samples, but the ability of HRM to diagnose other DNA modifications remains unexplored. The DNA bases N6-methyladenine and 5-hydroxymethylcytosine occur across almost all phyla. While their function remains controversial, their presence perturbs DNA structure. Such modif ications could affect gene regulation, chromatin condensation and DNA packaging. Here, we reveal that DNA containing N6-methyladenine or 5-hydroxymethylcytosine exhibits reduced thermal stability compared to cytosine-methylated DNA. These thermostability changes are sufficiently divergent to allow detection and quantification by HRM analysis. Thus, we report that HRM distinguishes between sequence-identical DNA differing only in the modification type of one base. This approach is also able to distinguish between two DNA fragments carrying both N6-methyladenine and 5-methylcytosine but differing only in the distance separating the modified bases. This finding provides scope for the development of new methods to characterize DNA chemically and to allow for low cost screening of mutant populations of genes involved in base modification. More fundamentally, contrast between the thermostabilizing effects of 5mC on dsDNA compared with the destabilizing effects of N6-methyladenine (m6A) and 5hydroxymethylcytosine (5hmC) raises the intriguing possibility of an antagonistic relationship between modification types with functional significance.

T

putative adenine DNA methyltransferase in the genome of Arabidopsis thaliana7 and the detection of m6A residues in the DRM2 gene5 implies that it is at least possible. In mammals, 5hmC has been found both in nuclear9 and mitochondrial10 DNA, with 5hmC levels being highly tissue especific.11 Although initially proposed to act as an intermediate in either active or passive demethylation of 5mC,12 recent evidence suggests an important functional role for this epigenetic mark.13,14 Previously, 5mC has been shown to elicit a stabilizing effect under HRM15 conditions; an observation that correlates with its reported effect on the fine structure of DNA,16 which causes alterations to gene expression.17 More recently, spectroscopic and calorimetric analyses have suggested that 5hmC introduction reverses the stabilizing effect of 5methylcytosine.18,19 Whether and how the observed changes to DNA thermostability imparted by 5hmC could have any functional impact on gene expression is still a matter of controversy. Some authors support the view that 5mC hydroxylation is more likely to be associated with transcriptional activation than with repression,18,20,21 while others argue that it can contribute to both transcriptional activation and repression in a context-dependent manner.14 Before either

he genomic DNA of almost all known phyla includes at least one form of chemically modified base,1 such as N6methyladenine (m6A), N4-methylcytosine (m4C), 5-methylcytosine (5mC), and 5-hydroxymethylcytosine (5hmC). Whereas m4C is confined to prokaryotes, m6A, 5mC, and 5hmC are found in both prokaryotes and eukaryotes.1 Such modifications can be inheritable epigenetic marks and have been implicated in many biological and pathological processes.2 To date, most attention has focused on the localization and function of 5mC.1,2 In comparison, relatively little is known about the distribution and function of 5hmC and m6A in higher organisms (for a detailed review see Iyer et al., 20113). The methylated modification of adenine, m6A, was first discovered in Escherichia coli4 and has since been found in a wide range of other organisms including ciliates (Tetrahymena thermophila) and Plasmodium species, chlorophyta,1 and higher plants.5 There is even indirect evidence that m6A may also be present in mammals.5 m6A lowers the thermodynamic stability of DNA and changes its conformation.6 It is thought that the projection of the methyl group at the N6 position of adenine into the major groove is partly responsible for the conformational changes to the dsDNA.7 In prokaryotes, it appears that the DNA conformational changes that arise from the presence of m6A alter protein−DNA interactions, particularly interactions between RNA polymerase and promoters containing m6A sites.7 Although it is unclear whether m6A is essential for regulation of any eukaryotic gene, the presence of a gene for © 2012 American Chemical Society

Received: March 15, 2012 Accepted: July 16, 2012 Published: July 16, 2012 7336

dx.doi.org/10.1021/ac301459x | Anal. Chem. 2012, 84, 7336−7342

Analytical Chemistry

Article

In the second experiment, we created a dilution series of the first amplicon templates to produce variation in PCR product concentrations used for HRM (5 dilutions [0.1−0.5 μL template in 20 μL reaction mixtures] x two replicates). Replicated RT-qPCR runs data were then exported to Excel and plotted together. De Novo Synthesis of DNA Containing Methylated Adenine and Cytosine. HPLC purified methylated and nonmethylated DNA representing two oligonucleotide sequences were synthesized by Sigma Aldrich (UK) (Supporting Information Table S-1b). Both fragments comprised complementary 84bp oligonucleotides that represent a section (See Supporting Information Figure S-1 for localization in the gene) of the domains rearranged methyltransferase (DRM2) gene (NCBI-GI 30684984) or a 77 base long oligonucleotide representing a hypothetical region of a typical eukaryotic gene and contained one cytosine and two adenine residue sites that could be methylated (Supporting Information Table S-1b). In the methylated version, one cytosine or one adenine were replaced by 5-methylcytosine (5mC) and N6-methyladenine (m6A) respectively on both strands. For the hemimethylated versions, only one of the DNA strands contained methylated bases. In all cases, dsDNA fragments were created by mixing complementary oligonucleotides in equal proportions as described on Supporting Information Table S-2, heating to 95 °C for 5 min and allowing to cool slowly to room temperature for at least 15 min to ensure complete reannealling. HRM Conditions. For HRM analysis, 0.2 μM dsDNA was added to a reaction mixture containing 2.5 mM MgCl2, 1.6 mM (NH4)SO4, 6.7 mM Tris-HCl (pH 8.8 at 25 °C), 0.001% Tween-20, and the intercalating fluorescent DNA dye SYTO9 (Invitrogen, U.K.) to create a final concentration of 0.5 μM. All meltings were performed in 20 μL volumes. Melting curve analysis was conducted on the Rotor-Gene 6000 (software version 1.7, Qiagen, U.K.) using the Cycling A-Green channel. During the melting, temperature was increased from 50 to 90 °C in 0.1 °C incremental steps, with each step held for 2 s. Using Rotor-Gene 6000 software, the melting curves were normalized by calculation of the ‘line of best fit’ between two normalization regions selected before and after the major decrease in fluorescence (representing the ‘fragment melting’). Comparisons were made between methylated and nonmethylated DNA in terms of Tm or in a combination of Tm and altered curve shape. Comparative Effect of 5mC and 5hmC on Fragment Thermostability. The ability of HRM to provide a qualitative measure of the type of cytosine methylation was assessed in two experiments comprising of replicated HRM reactions of 5mC, 5hmC, and nonmethylated dsDNA generated as described above either from a 76 bp oligonucleotide template or an 85 bp oligonucleotide template (replicating part of a region 2322 bases upstream of the tomato CNR gene). The resultant PCR products were unmethylated, methylated or hydroxymethylated in six positions (76 bp fragment) or eight positions (CNR fragment) (Supporting Information Table S-1; Figure S-1). Raw and normalized replicated HRM runs data were exported to Excel and plotted together. Effects of 5mC and m6A on DNA Thermostability. The purposes of this experiment were to (1) assess the capacity of HRM to directly detect the presence of 6-methyl-adenine (m6A) in a dsDNA fragment and to evaluate any impact on thermostability, (2) compare the individual effects of 5mC and

hypothesis can be tested, there is a need for a simple but effective means of differentiating between 5mC and 5hmC, and an effective forward genetic tool to identify genes involved in conversion between states. Several techniques have been developed for the detection and localization of DNA methylation, but all have limitations.15 Some works have detected m6A using gel mobility shift analysis,6 and methylation sensitive enzymes.5 Sodium bisulfite treatment of DNA does not affect m6A and is unable to distinguish between 5hmC and 5mC,22 but a recent modification of the method does allow for this.23 Robustness of the method has yet to be stablished. In contrast, techniques based on immunoprecipitation with anti-5mC antibody or proteins that bind to methylated CpG sequences22 combined with anti-5hmC16or with the selective glucosylation of 5hmC residues followed by precipitation using J-binding protein124 could be used to detect genomic regions containing each type of cytosine modification but generally lack resolution and are relatively time-consuming. More recent developments in highthroughput sequencing technologies (i.e., single-molecule, realtime (SMRT) sequencing,25 nanopore sequencing19) offer the enticing long-term possibility of directly assaying for the various forms of DNA base modification across the genome. However, these technologies are still some way from widespread availability.26 Previously the feasibility of using HRM technology to detect changes to the cytosine methylation status of DNA has been demonstrated.15 The fact that spectroscopic and calorimetric analyses indicate the thermodynamic stabilization conferred by 5mC is reversed by 5hmC18 and by m6A27 gives rise to the (as yet) untested expectation that HRM profiles will be similarly reversed. Here, we therefore describe the first evaluation of HRM technology to detect and study the influence of 5hmC and m6A on the thermostability of dsDNA.

■

EXPERIMENTAL SECTION Target DNA. Generation of DNA Containing 5-Methyland 5-Hydroxymethylcytosine. Synthetic DNA fragments containing modified bases were produced by polymerase chain reaction (PCR) using modified deoxycytidine triphosphates, 5-methyl-2′-deoxycytidine 5′-triphosphate (5mdCTP) (Fermentas; Glen Burnie, MD) and 5-hydroxymethyl-2′deoxycytidine 5′-triphosphate (5hmdCTP) (Bioline; Taunton, MA) as described previously22 but with some modifications. In brief, starting templates of 0.5 ng of a single-stranded 76-mer or an 85-mer oligonucleotide (Supporting Information Table S-1) were used to generate PCR amplicons containing six and eight modified CpG sites respectively. RT-qPCR conditions were: 94 °C for 4 min, followed by 45 cycles of 94 °C for 20 s, 55 °C for 25 s, 68 °C for 30 s. Reaction buffer contained 1.5 mM MgCl2, 0.1 mM of each dNTP (or 5mdCTP or 5hmdCTP in place of dCTP), 0.6 μL forward and reverse primers (10 μM) (Supporting Information Table S-1), SYTO9 (Invitrogen, U.K.) to a final concentration of 0.5 μM and 1U Taq DNA polymerase (Bioline, Taunton, MA). A second identical PCR was then performed using 0.1−0.5 μL of initial amplicons to maximize the removal of unmodified DNA templates from the final products.22 Two experiments were performed to characterize the reproducibility of HRMs. In the first experiment, we created variability in the concentration of PCR products used for HRM analysis through simple replication of the second amplification step (5 replicates of 0.5 μL template in 20 μL reaction mixture). 7337

dx.doi.org/10.1021/ac301459x | Anal. Chem. 2012, 84, 7336−7342

Analytical Chemistry

Article

generated significantly different HRM curves. The software also provides the probability that each sample belongs to the assigned (and other) cluster(s) and the typicality of samples or how consistent a sample is within its own group which indicates how well the sample fits within its allocated cluster (for details of the calculation of posterior probabilities and typicalities see Reja et al., 2010).29

m6A presence on the thermostability of dsDNA, as detectable by HRM analysis, (3) use HRM analysis to assess the interaction/dominance of the possible positively or negatively cooperative effects of the presence of m6A and 5mC on DNA thermostability. For these, we performed three sets of HRM profile experiments using differentially methylated (but sequenceidentical) 84bp fragments of the A. thaliana DRM2 gene (See Supporting Information Table S-1 for sequence and methylation modifications). First, we compared fragments carrying one m6A on each strand with those carrying no m6A on either strand and to two forms of hemimethylated DNA in which a m6A resided on either the forward or complementary strands. We then compared fragments containing only nonmethylated bases or one 5mC or one m6A on each strand. Finally, we compared fragments containing one 5mC and one m6A on each strand but with the modified bases separated by different distances. We then repeated all experiments using a 77 base long oligonucleotide representing a hypothetical region of a typical eukaryotic gene (Supporting Information Table S-1). Four replicates were measured for each condition. Detecting the Loss of 5hmdCTP during Bacterial Growth. We next tested the capacity of HRM to detect differences in hydroxymethylation content of DNA extracted from living material by comparing the HRM profiles of native and modified plasmid DNA before and after bacterial growth. Here, we first generated a modified plasmid containing 5-hydroxymethyl-2′deoxycytidine 5′-triphosphate (5hmdCTP; Bioline, Taunton, MA) by Ligation During Amplification (LDA) PCR as described previously.28 We likewise generated an equivalent version of the plasmid containing only deoxycytidine triphosphate (dCTP) using the same approach.28 In brief, the reaction mixture (50 μL) contained 10 ng of native plasmid, 10 pmol of each primer (Supporting Information Table S-1), 5 nmol of each dNTP, 5 nmol of ATP, 2.5 U Pfu DNA polymerase (Stratagene, La Jolla, CA), 4 U Pfu DNA ligase (Stratagene), in 1× cloned Pfu DNA polymerase reaction buffer. The mixture was preincubated at 70 °C for 10 min and then subjected to thermal cycling at 94 °C for 10 s, 50 °C for 2 min, 72 °C for 4 min, 94 °C for 10 s and 72 °C for 10 min for 20 cycles. Then OmniMAX 2 -T1R cells (Invitrogen) were transformed with the PCR generated plasmids. OmniMAX 2 -T1R cells lack the E. coli K12 restriction systems (mcrA Δ(mrr hsdRMS-mcrBC)) which permits its transformation with highly methylated DNA. Negative control samples (not transformed) all failed to grow in the selective medium because of the absence of the plasmid. However, we were able to extract plasmid DNA from bacteria transformed with plasmids containing dCTP, or 5hmdCTP after 16 h of incubation using the QIAprep Spin Miniprep Kit (Qiagen, U.K.). Isolated plasmid DNA was diluted in nanopure water to produce working stocks of 100 ng μL−1. The isolated DNA was then sonicated at 120 W for 3 min to obtain fragments in the 100− 600 bp size. We then subjected 500 ng of these fragmented DNAs to HRM as described above. Statistical Analysis. The significance of observed differences between treatments in Tm and altered curve shape was calculated by principal component analysis (PCA) and sample cluster analysis (SCA) using Rotor-Gene ScreenClust HRM Software (Qiagen)29 as described previously.15 The software calculates the optimal number of clusters and allocates each sample into the most appropriate cluster, indicating in this case which methylation contents and reagent concentrations

■

RESULTS AND DISCUSSION Base Modifications and DNA Thermostability. Previous work demonstrated the capacity of HRM to detect and quantify the presence of 5mC directly from DNA samples.15 However, the capability of HRM to similarly detect and differentiate between two other important DNA modifications (5hmC and m6A) remains untested. Of these, greatest recent interest has focused on 5hmC,18,20,21 partly because of the unknown relationship between this modification and 5mC,12,16−18,20,21 but also because of the inability of conventional bisulfite treatment to distinguish between them.22 HRM of sequenceidentical DNA templates that differed only in the form of cytosine they contained (i.e., C, 5mC, and 5hmC) revealed considerable variance in the shape of the resultant melting profiles and in the Tm values between types of DNA. These differences were conserved regardless of the initial concentration of DNA used for the HRM (Supporting Information Figure S-2), and with high reproducibility between replicates and experiments (Figure 1). In comparison to nonmethylated templates, 5mC-containing dsDNAs invariably showed significantly raised Tm values and divergent melting curve profiles

Figure 1. Effect of 5mC and 5hmC on DNA thermostability. Normalized melting profiles representing two HRM experiments of a 76 bp oligonucleotide containing 6 CpG sites obtained using software on the Corbett Rotor-Gene 6000. Inset: PCA plot and SCA of the same samples using Rotor-Gene ScreenClust HRM Software. DNA HRM templates containing nonmethylated, methylated, or hydroxymethylated cytosines were generated by PCR using dCTP (dark and light blue), 5mdCTP (red and orange), and 5hmdCTP (dark and light green) (Supporting Information Figure S-2). Red, dark blue, and dark green samples were generated from a single PCR Mastermix containing 0.5 μL of first amplification products. Orange, both shades of light blue and light green were generated from a dilution series of first amplicon templates (0.1−0.5 μL template). The melting curves were normalized by calculation of the “line of best fit” between normalization regions before and after the major fluorescence decrease using the Rotor-Gene 6000 software. 7338

dx.doi.org/10.1021/ac301459x | Anal. Chem. 2012, 84, 7336−7342

Analytical Chemistry

Article

Figure 2. Effect of m6A of DNA thermostability. Normalized melting profiles from 4 replicates of an 84 bp oligonucleotide (Supporting Information Table S-1) representing the A. thaliana Domains Rearranged Methyltransferase (DRM2) gene containing 1 adenine residue that could be methylated, analyzed using software on the Corbett Rotor-Gene 6000. Inset, PCA plot and SCA of the same samples using Rotor-Gene ScreenClust HRM Software. The melting curves were normalized by calculation of the “line of best fit” between normalization regions before and after the major fluorescence decrease using the Rotor-Gene 6000 software.

bacterial growth. For this, we first established that plasmids generated by LDA-PCR to contain 5hmdCTP yielded divergent HRM profiles from those containing dCTP (Supporting Information Figure S-5). These plasmids were transformed into OmniMAX 2-T1R bacterial cells and cultured for 16 h before being recovered by plasmid extraction. As expected, the HRM profiles of the “recovered dCTP” plasmids were identical to those of the original templates used for transformation (Supporting Information Figure S-5), implying no change had occurred during in vivo plasmid replication. Conversely, the recovered 5hmdCTP plasmids were highly divergent from their original templates and were now indistinguishable from the template and recovered dCTP plasmids (Supporting Information Figure S-5), suggesting that the 5hmdCTP modifications had been lost through in vivo plasmid replication. Furthermore, the consistency of results between biological replicates implies that matrix differences did not have a large influence on HRM profiles in this instance. We note that it is nevertheless important to consider the potential for such effects when performing experiments of this kind. Effect of m6A of DNA Thermostability. Subjection of de novo synthesized DNA templates to HRM revealed that DNA samples containing only nonmethylated adenine could be distinguished from those incorporating m6A, and that it was even possible to separate hemimethylated templates from those carrying methylated adenine on both strands (Figure 2). The two alternate forms of hemimethylated DNA showed intermediate Tm values, between nonmethylated and fully m6A DNA, but also exhibited subtle differences in their Tm values (e.g., Figure 2). For templates carrying a single m6A, melting curve shape was largely similar to the nonmethylated controls, and in stark contrast to 5mC-containing DNA exhibited significantly reduced Tm values. PCA and SCA showed that differences in m6A content generate significantly

(Figure 1), as reported previously.15 Conversely, Tm was significantly increased in DNA containing hydroxymethylation but melting curve shape of fragments carrying this modification were largely indistinguishable from the nonmethylated controls (Figure 1 and Supporting Information Figure S-3). PCA and SCA revealed melting curves of individual samples consistently clustered according to the three modification types used (Figure 1 and Supporting Information Figure S-3). Analysis also showed high probabilities (P) and typicalities (T) of samples belonging to their respective cluster (P = 1; T > 407) (Supporting Information Tables S-3 and S-4). Thus, melting analyses performed here support recent research suggesting that 5-methylcytosine hydroxylation may change dsDNA structure in solution and demonstrates that it reduces DNA thermostability18,19 by margins that are sufficiently large to allow detection by HRM. One plausible hypothesis put forward by Wanunu and colleagues19 to explain the increase in Tm associated with the presence of 5mC was that water molecules around the least polar nucleoside 5mC leads to increased rigidity of dsDNA, and that this feature increases its thermostability. Likewise, Zubay and Doty30 proposed that the preferential stabilizing effect of Mg2+ over the hydrophobic repulsions of the solvent in the methylated dsDNA may be abolished by the presence of 5hmC, implying that the presence of this modification should reduce the Tm of dsDNA when compared to methylated DNA. Detecting the Loss of 5hmdCTP during Bacterial Growth. HRM can thus detect small differences in Tm induced by the presence of 5hmC in de novo synthesized DNA samples but was untested on DNA recovered from living organisms. There is therefore the possibility that matrix effects could mask such differences when using DNA extracted from live material. We therefore tested whether the approach could detect the loss of 5hmdCTP from introduced plasmids by dilution during 7339

dx.doi.org/10.1021/ac301459x | Anal. Chem. 2012, 84, 7336−7342

Analytical Chemistry

Article

Figure 3. Effect of relative position of 5mC and m6A on fragment thermostability. Normalized melting profiles representing 4 replicates using the software incorporated on the Corbett Rotor-Gene 6000 of an 84 bp oligonucleotide from the A. thaliana Domains Rearranged Methyltransferase (DRM2) gene containing 1 cytosine and 2 adenine residues that could be methylated. Inset, PCA plot and SCA of the same samples using RotorGene ScreenClust HRM Software. The melting curves were normalized by calculation of the line of best fit between normalization regions before and after the major fluorescence decrease using the Rotor-Gene 6000 software. For fragment sequences see Supporting Information Table S-1.

and nonmethylated) generated significantly different melting patterns, with replicates clustering according to their methylation status (Figure 3 and Supporting Information Figure S-4). Analysis also showed high probabilities (P) and typicalities (T) of samples belonging to the assigned cluster (P > 0.999; T > 0.486) (Supporting Information Table S-5). When DNA fragments containing only one of the three DNA modifications were compared on the same HRM, as expected, the Tm values differed significantly and followed the same order in the previous experiments, with the m6A DNA having a lower Tm than the nonmethylated control which itself was lower than the fragment containing 5mC. Moreover, the shape of the melting curves was largely conserved between cytosine methylation and control (as described previously15), but the m6A DNA melting curve shape differed in showing a much more concave profile, indicative of a rapid denaturation process. Collectively, these results indicate that m6A significantly destabilizes dsDNA thereby favoring denaturation, whereas 5mC has the opposite effect. Given that dsDNA must be partially melted to enable transcription, it has been suggested that DNA replication and gene expression in bacterial genomes could be facilitated by a lowered melting temperature at the oriC region induced by m6A at GATC sites.39 It is tempting to speculate that these changes may also influence gene expression in eukaryotes. In a study of transient transgene expression in barley, Rogers38 reported the presence of m6A enhanced transcription to an extent that was greater in magnitude than the suppression to transcription caused when all CpG dinucleotides contained 5mC. If this observation applies more generally, it raises the issue of how these modifications with differential effects on DNA stability interact when they are in close proximity to each other. Both 5mC and m6A have been found in the same A. thaliana DRM2 gene5 illustrating that both DNA modifications can

different melting patterns, with samples clustering according to their methylation status (Figure 2 and Supporting Information Figure S-4). Analysis also showed high probabilities (P) and typicalities (T) of samples belonging to the assigned cluster (P = 1; T > 0.522) (see Supporting Information Table S-4). The presence of m6A has been associated to temperature-dependent conformational changes to DNA that are thought to be related to the initiation of transcription at the trp promoter.31 This conformational change was reported to decrease Tm on the m6A compared to that of the unmethylated sequence.32 This is supported by the HRM results obtained here, with the presence of a single m6A seemingly destabilizing the dsDNA such that there was a significant decrease in Tm. Two explanations have been put forward for the decreased DNA thermostability conferred by the presence of m6A. First, the hydrophobic nature of the relatively bulky methyl group could prevent the rotation of the amino group,32 thereby reducing the exposure of the remaining exchangeable proton to the solvent. Second, the normally solvent accessible (non-hydrogenbonded) and exchangeable N6-amino proton can be replaced by a methyl group.31 For a long time, m6A was believed to be specific to prokaryotes. It is now known that it is also present in eukaryotes, although it appears to be specific to plants,8,33−35 where it seems to have positive effect on transient gene expression,36,37 which may be mediated by expression of a m6A DNA binding protein.37 Cooperative effects of 5mC and m6A on DNA Thermostability. Having demonstrated the capacity of HRM to differentiate between dsDNA containing methylated, hemimethylated and nonmethylated adenine residues (Figure 2), with m6A decreasing DNA thermostability, we next tested the relative strength of this effect in comparison to the previously reported stabilizing effect of 5mC.15 PCA and SCA confirmed that the different DNA templates (i.e., m6A, 5mC, 7340

dx.doi.org/10.1021/ac301459x | Anal. Chem. 2012, 84, 7336−7342

Analytical Chemistry

Article

opposing effects of m6A and 5mC on DNA stability are sufficiently large for detection by HRM and provide a method enabling further studies on the relationship between DNA thermostability and transcriptional activity. We have further shown that m6A has the same destabilizing effect as 5hmC on DNA stability under HRM conditions. It is tempting to speculate that both modifications (5hmC and m6A) may therefore have the same antagonist function to 5mC of enhancing gene expression by easing the process of DNA denaturation. However, of greater practical significance is the possibility that global HRM analysis provides the capacity to rapidly and cheaply assess for gross changes in the hydroxymethylation/methylation status of entire genomes. This capability in turn provides a convenient tool for the large-scale screening of mutant populations to identify genes responsible for the conversion between these two DNA states. Viewed from a more technical perspective, this feature may open the possibility of using the HRM approach to develop new techniques (possibly based on chip technology) to enhance the study of base modification within the genome and to conduct further empirical studies to test the relative importance of methylation mark interactions on the thermostability of loci known to be under direct methylation control.

coexist in close proximity in biological material. Interaction between the effects of m6A and 5mC on DNA thermostability was therefore assessed experimentally by generating HRM profiles of de novo synthesized DNA containing both modifications separated by one or seventeen nucleotides (Supporting Information Table S-1) and comparing these to those produced by sequence-identical DNA containing no methylation or only m6A or 5mC. DNA fragments containing both modifications showed intermediate Tm when compared to those presenting only one modified base (i.e., m6A or 5mC) (Figure 3), reinforcing our earlier finding that the two modified bases have antagonistic effects. Again, PCA and SCA showed that the observed differences in the melting profiles shapes and Tm were significantly divergent (P > 0.996; T > 0.478) (Supporting Information Table S-5). Of greater interest were the differences between the DNA fragments carrying both types of base modification. In the DRM2 sequence context, the Tm of DNA templates carrying adenine methylation was slightly but significantly increased when carrying an adjacent methylated cytosine, and there was a further increase of similar magnitude when the methylated cytosine was more distally positioned. In both cases, the observed Tm were lower than those of the nonmethylated DNA, implying that in this sequence context the destabilizing effect of m6A is stronger than the potentially antagonistic influence of 5mC DNA. In the hypothetical eukaryotic gene sequence context, adenine and cytosine methylation showed the same antagonistic effect but in this case the increase in Tm was higher when the methylated cytosine was situated adjacent to the methylated adenine. These observed differences between the two sequence contexts might be associated to the fact that long DNA fragments have multiple melting domains that are highly dependent on the fragment’s sequence.40 Further research is needed to understand the synergistic/antagonistic effects of base modifications and sequence during HRM. Whether or not these opposing influences have functional significance in a biological context is questionable, although the apparent absence of strong DNA− protein interactions linking m6A to enhanced transcription in tobacco36 at least suggests that the mechanisms involved probably differ from those acting in prokaryotes.6 Speculating still further; if m6A has a functional role in plants tied to an ability to enhance expression (as demonstrated at least transiently38) by reducing dsDNA thermostability, then the dominance of this effect over the antagonistic stabilization conferred by 5mC observed on the DRM2 sequence context, would provide a natural explanation for its lower abundance in eukaryotic genomes.1,5 However, further work is required to test such possibilities.

■

ASSOCIATED CONTENT

S Supporting Information *

Oligonucleotide sequences, average typicalities and probabilities of all sample cluster analyses, DRM2 nucleotide sequence, plasmid DNA HRM profiles, and RT-qPCR profiles. This material is available free of charge via the Internet at http:// pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Present Address †

School of Agriculture, Wine and Food, The University of Adelaide, Waite Campus, Australia 5005 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors thank the Aberystwyth University Commercialization and Consultancy Services (CCS) for the funding provided for A.J.L. This project was supported by the Early Stage Development Fund (ESDF) in support of the Knowledge Exploitation Fund and Patent Proof of Concept Fund (PPOC) from the Welsh Assembly Government.

■

■

CONCLUSIONS We have shown that it is possible to use HRM not only to distinguish between DNA fragments containing 5mC as we showed before15 but also between those containing other modifications such as 5hmC and m6A (Figures 1−3). HRM’s ability to detect all modifications is associated to the effect that each type has on the structure and thermostability of the DNA. While 5mC has a stabilizing effect when compared to nonmethylated cytosine, 5hmC and m6A reduce DNA stability. In mammals, even if still controversial,14,18,20,21 5mC and 5hmC seem to have antagonistic effects on gene expression. Although 5mC has the same function on plant gene expression as it does on mammals (i.e., down-regulation) to our knowledge 5hmC has yet to be reported in plants. Here we demonstrate that the

REFERENCES

(1) Hattman, S. Biochemistry (Moscow) 2005, 70 (5), 550−558. (2) Henderson, I. R.; Jacobsen, S. E. Nature 2007, 447, 418−424. (3) Iyer, L. M.; Abhiman, S.; Aravind, L. Prog. Mol. Biol. Transl. Sci. 2011, 01, 25−104. (4) Dunn, D. B.; Smith, J. D. Nature 1955, 175, 336−337. (5) Ashapkin, V. V.; Kutueva, L. I.; Vanyushin, B. F. FEBS Lett. 2002, 532 (3), 367−372. (6) Polaczek, P.; Kwan, K.; Campbell, J. L. Mol. Genet. Genomics 1998, 258, 488−493. (7) Sternberg, N. J. Bacteriol. 1985, 164, 490−493. (8) Shorning, B. Y.; Vanyushin, B. F. Biochemistry (Moscow) 2001, 66, 753−762. (9) Kriaucionis, S; Heintz, N. Science 2009, 324, 929−930. 7341

dx.doi.org/10.1021/ac301459x | Anal. Chem. 2012, 84, 7336−7342

Analytical Chemistry

Article

(10) Shock, L. S.; Thakkar, P. V.; Peterson, E. J.; Moran, R. G.; Taylor, S. M. Proc. Natl. Acad. Sci. U. S. A. 2011, 108, 3630−3635. (11) Muenzel, M.; Globisch, D.; Carell, T. Angew. Chem., Int. Ed. 2011, 50 (29), 6460−6468. (12) Huang, Y.; Pastor, W. A.; Shen, Y.; Tahiliani, M.; Liu, D. R.; Rao, A. PLoS One 2010, 5 (1), e8888. (13) Robertson, J.; Robertson, A. B.; Klungland, A. Biochem. Biophys. Res. Commun. 2011, 411 (1), 40−43. (14) Wu, H.; D’Alessio, A. C.; Ito, S.; Wang, Z.; Cui, K.; Zhao, K.; Sun, Y. E.; Zhang, Y. Genes Dev. 2011, 25, 679−684. (15) Rodríguez López, C. M.; Guzmán Asenjo, B.; Lloyd, A. J.; Wilkinson, M. J. Anal. Chem. 2010, 82, 9100−9108. (16) Heinemann, U.; Hahn, M. J. Biol. Chem. 1992, 267, 7332−7341. (17) Jones, P. A.; Laird, P. W. Nat. Genet. 1999, 21, 163−167. (18) Thalhammer, A.; Hansen, A. S.; El-Sagheer, A. H.; Brown, T.; Schofield, C. J. Chem. Commun. (Cambridge, U. K.) 2011, 47, 5325− 5327. (19) Wanunu, M.; Cohen-Karni, D.; Johnson, R. R.; Fields, L.; Benner, J.; Peterman, N.; Zheng, Y.; Klein, M. L.; Drndic, M. J. Am. Chem. Soc. 2011, 133, 486−492. (20) Ito, S.; D’Alessio, A. C.; Taranova, O. V.; Hong, K.; Sowers, L. C.; Zhang, Y. Nature 2010, 466, 1129−1133. (21) Zhang, H.; Zhang, X.; Clark, E.; Mulcahey, M.; Huang, S.; Shi, Y. G. Cell Res. 2010, 20 (2), 1390−1393. (22) Jin, S.-G.; Kadam, S.; Pfeifer, G. P. Nucleic Acids Res. 2010, 38 (11), e125. (23) Booth, M. J.; Branco, M. R.; Ficz, G.; Oxley, D.; Krueger, F.; Reik, W.; Balasubramanian, S. Science 2012, 336 (6083), 934−937. ́ (24) Robertson, A. B.; Dahl, J. A.; Vågbø, C. B.; Tripathi, P.; Krokan, H. E.; Klungland, A. Nucleic Acids Res. 2011, 39 (8), e55. (25) Flusberg, B. A.; Webster, D. R.; Lee, J. H.; Travers, K. J.; Olivares, E. C.; Clark, T. A.; Korlach, J.; Turner, S. W. Nat. Methods. 2010, 7, 461−465. (26) Gupta, P. K. Trends Biotechnol. 2008, 26 (11), 602−611. (27) Quignard, E.; Fazakerley, G. V.; Teoule, R.; Guy, A.; Guschlbauer, W. Eur. J. Biochem. 1985, 152 (1), 99−105. (28) Chen, Z.; Ruffner, D. E. Nucleic Acids Res. 1998, 26 (4), 126− 1127. (29) Reja, V.; Kwok, A.; Stone, G.; Yang, L.; Missel, A.; Menzela, C.; Bassam, B. Methods 2010, 504, S10−S14. (30) Zubay, G.; Doty, P. Biochim. Biophys. Acta 1958, 29, 47. (31) Lingbeck, J.; Kubinec, M. G.; Miller, J.; Reid, B. R.; Drobny, G. P.; Kennedy, M. A. Biochemistry 1996, 35, 719−734. (32) Rinkel, L. J.; van der Marel, G. A.; van Boom, J. H.; Altona, C. Eur. J. Biochem. 1987, 163, 275−286. (33) Pinter-Toro, J. A. Biochem. Biophys. Res. Commun. 1987, 147, 1082−1087. (34) Ngemprasirtsiri, J.; Chollet, R.; Kabayashi, H.; Sugiyama, T.; Akazawa, T. J. Biol. Chem. 1989, 264, 8241−8248. (35) Mtiller, E.; Brow, P. T. H.; Hartke, S.; Lorz, H. Theor. Appl. Genet. 1990, 80, 673−679. (36) Torres, J. T.; Block, A.; Hahlbrock, K.; Somssich, I. E. Plant J. 1993, 4, 587−592. (37) Brodzik, R.; Hennig, J. Plant Physiol. Biochem. 1998, 36 (5), 401−406. (38) Rogers, J. C.; Rogers, S. W. Plant J. 1995, 7, 221−233. (39) Yamaki, H.; Ohtsubo, E.; Nagai, K.; Maeda, Y. Nucleic Acids Res. 1988, 16 (11), 5067−5073. (40) Wittwer, C. T. Hum Mutat. 2009, 30 (6), 857−859.

7342

dx.doi.org/10.1021/ac301459x | Anal. Chem. 2012, 84, 7336−7342