Modified Nucleotides for Discrimination between Cytosine and the Epigenetic Marker 5‐Methylcytosine

Abstract 5‐Methyl‐2′‐deoxycytosine, the most common epigenetic marker of DNA in eukaryotic cells, plays a key role in gene regulation and affects various cellular processes such as development and carcinogenesis. Therefore, the detection of 5mC can serve as an important biomarker for diagnostics. Here we describe that modified dGTP analogues as well as modified primers are able to sense the presence or absence of a single methylation of C, even though this modification does not interfere directly with Watson–Crick nucleobase pairing. By screening several modified nucleotide scaffolds, O6‐modified 2′‐deoxyguanosine analogues were identified as discriminating between C and 5mC. These modified nucleotides might find application in site‐specific 5mC detection, for example, through real‐time PCR approaches.

Epigenetic modifications,c aused by methylation of cytosine residues (5methyl-2'-deoxycytosine,5 mC; Figure 1), have been proven to have an impact on av ariety of cellular processes that affect development [1,2] and gene expression [3] as well as the development of various diseases. [4] Genes are frequently found to be silenced [3] if CpG dinucleotides in the corresponding promoters exhibit significant levels of 5mC;thus,5mC is known to regulate gene transcription and thereby affect tumorigenesis. [5] Thel evel of epigenetic methylation has to be precisely regulated in eukaryotic genomes,s ince changes of the methylation pattern lead to severe genetic malfunctions. [6] Therefore,t he detection of the occurrence and distribution of 5mC in the genome holds the potential to serve as an important biomarker for diagnosis as well as disease therapy. [7] This requires efficient strategies for the detection of 5mC.Different concepts for discriminating between cytosine (C) and 5mC have been described which rely on endonuclease digestion, [8] affinity enrichment, [9] nanopore sequencing, [10] different chemical behavior concerning redox reactivity, [11] or selective deamination of Cusing sodium bisulfite. [12] Bisulfite sequencing hasbecome routine forthe detectionof 5mCw iths ingle-nucleotide resolution. [13] This method is based on thes electiveb isulfite-mediatedd eamination of Ct ou racil (U) in thepresence of 5mC,which remains unaffectedasaresult of slower deamination. [14] Thesitesofepigenetic markers canbe revealed by comparison of the outputofconventional sequencingm ethodsb efore anda fter bisulfite treatment, as Cw ill be sequenced as thymine( T),a nd 5mCasC . [15] Although bisulfite sequencing can be used for the genome-wide detection of 5mC,i tp ossesses several drawbacks.Since many steps are required and two sequencing runs are needed for comparison, the method is time consuming and prone to contaminations. [16] In addition, the conditions used for the bisulfite treatment are harsh and destroy approximately 95 %o fthe genomic DNA, and thus al arge amount of DNAisrequired. [17] Furthermore,deamination of Ca nd 5mC after bisulfite treatment is incomplete,t hereby leading to an error-prone output. [18] Nucleotides that are able to discriminate between Ca nd 5mC would be highly interesting tools for the development of new approaches for the site-specific detection of 5mC. However,s ince the methylation of cytosines at C5 does not directly affect Watson-Crick base pairing,a ttempts along these lines are challenging and have not yet been described.
Here we present ac lass of modified nucleotides that can be used for discrimination between Ca nd 5mC in reactions catalyzed by DNAp olymerases.F or this purpose,w e screened and investigated several purine-based 2'-deoxynucleotides for their ability to sense 5mC as aresult of diverging incorporation efficiencies opposite atemplate containing Cor 5mC by different DNAp olymerases ( Figure 2). We found that several O 6 -modified 2'-deoxyguanosine derivatives (Figure 3a)a re incorporated opposite 5mC and extended from 5mC with significantly different efficiencies compared to the unmodified counterparts.
First, we screened av ariety of different modified nucleotides ( Figure 2) in combination with the thermostable DNA polymerases KlenTaq and KOD exo À [19] in primer extension experiments,f ollowed by analysis through denaturing polyacrylamide gel electrophoresis (PAGE) and visualization by autoradiography.B oth DNAp olymerases were able to incorporate differently modified nucleotides opposite Co r 5mC,a lthough with decreased incorporation efficiencies compared to the unmodified dGTP (1;F igure 2). KOD exo À showed the highest potencyf or the desired application, with the most pronounced differences in the incorporation efficiencies being observed during the processing of the nucle- otides O 6 -methyl-dGTP (3), dATP (10), and 5-nitro-1-indolyl-2'-deoxyribose-5'-triphosphate (21). Since modified dATP analogues (nucleotides 11-16)s howed decreased incorporation efficiencies as well as decreased discrimination between both templates,and the synthesis of derivatives of nucleotide 21 seems to be tedious and challenging, we decided to focus on O 6 -methyl-dGTP (3)f or further derivatization. KlenTaq DNAp olymerase showed no significant difference in incorporation efficiencies for either dGTP or the modified nucleotides (see Figure S1 in the Supporting Information).
Next, we focused on optimizing the system by synthesis and functional evaluation of O 6 -alkylated dGTPs. O 6 -Alkylated dGTP derivatives were synthesized starting from commercially available 2'-deoxyguanosine,w hich was acetylated and subsequently chlorinated at the 6-position by known procedures (see Figure S2 in the Supporting Information). [20] Thedifferent alkoxy groups were introduced in position 6by the reaction of 22 with the respective sodium alkoxide solutions.T he obtained nucleosides 23 a-c were converted into the corresponding 2'-deoxynucleoside-5'-triphosphates ( Figure S2). [21,22] ThepotencyofO 6 -methyl-, O 6 -ethyl-, O 6 -propyl-, and O 6isopropyl-dGTP towards diverging incorporation efficiencies opposite Ca nd 5mC was further investigated. Both DNA polymerases were able to incorporate all four nucleotide analogues opposite Ca nd 5mC,a lthough the incorporation efficiency decreased with increased steric demand of the introduced modifications (Figure 3b and see Scheme S4). KOD exo À showed the highest potencyf or the desired application, with the most pronounced differences in incorporation efficiencies being observed during the processing of the nucleotides dGTP, O 6 -methyl-, O 6 -ethyl-, O 6 -propyl-, and O 6 -isopropyl-dGTP opposite Cc ompared to 5mC (Figure 3b). Although the processing efficiencies of those modified nucleotides were decreased compared to natural dGTP, the discrimination between the two templates was increased markedly (Figure 3b). Of note,a sr eported before, O 6methyl-dGTP was also processed opposite T [23] and we could show that the same holds true for the other O 6 -alkyl-dGTP derivatives ( Figure S5). In contrast, KlenTaq DNA polymerase again showed no significant difference in incorporation efficiencies for either dGTP or the modified nucleotides ( Figure S4).
To further investigate this observation, we determined the steady-state kinetics [24] for the processing of the nucleotides dGTP, O 6 -methyl-, O 6 -ethyl-, O 6 -propyl-, and O 6 -isopropyl-dGTP by KOD exo À opposite Ca nd 5mC (Table 1; see also Scheme S2 and Figure S7). Comparison of the catalytic efficiencies (k cat /K m )o bserved during processing of dGTP and the modified nucleotides opposite Co r5 mC in the template strand confirms all the tendencies observed in the   (Table 1a nd see Scheme S2). Ther atio of the catalytic efficiencyo bserved during incorporation of the respective nucleotide opposite C compared to 5mC varied. Theb est discrimination was observed for the processing of O 6 -ethyl-dGTP,w ith the catalytic efficiencies k cat /K m for incorporation opposite C and 5mC differing by af actor of 4.2 (Table 1a nd see  Scheme S2).
Next, we aimed at investigating the effect of the modifications when they are embedded at the 3'-terminus of primer strands and placed directly opposite Co r5 mC. [25] Thus,t he modified nucleosides were converted into phosphoramidites, which were employed in DNAs olid-phase synthesis of the different primer strands (see Figure S3 in the Supporting Information). Them ethyl and ethyl modifications in O 6methyl-dGTP and O 6 -ethyl-dGTP showed the most promising results,a nd so we focused on these modifications ( Figure 3, Table S1).
After their synthesis,t he primer strands were first employed in primer extension studies (Figure 4b). Again, both modifications were well tolerated by both DNA polymerases tested, with all three primers elongated, although the modified primers to as maller extent than the unmodified primer (Figure 4b and see Figure S6). Interestingly,n os ignificant discrimination between Ca nd 5mC was observed with KOD exo À ,e ither when the unmodified or when the modified primer strands were employed (Figure S6). In contrast, significant discrimination was observed with the KlenTaq DNApolymerase for both modified primers (Figure 4b). Am arginal difference in the incorporation efficiencies between extension from Co r5 mC was observed when the unmodified primer was extended with dCTP.I n contrast, elongation of the modified primers bearing either O 6 -methyl-G or O 6 -ethyl-G at the 3'-end showed extensive discrimination between both templates (Figure 4b).
Interestingly,inthese primer extension reactions,elongation from the modified primer is more efficient when paired with 5mC.T oc onfirm these results,w ed etermined steady-state kinetics [24] using KlenTaq DNAp olymerase in combination with the unmodified primer and the two modified primers paired with both templates (Table S3, Figure S8). These experiments confirm that both modified primers are elongated with lower efficiencyt han the unmodified primer and that elongation of the primers from 5mC is more efficient. We found that the primer bearing O 6 -methyl-G showed the best discrimination between Ca nd 5mC (Table 1a nd Scheme S3). Thec atalytic efficiencies (k cat /K m )f or the two reactions differ by afactor of 6.7 (Table 1).
Thedifferent behavior of the DNApolymerases KlenTaq and KOD exo À when processing the modified nucleotides in the context of Cand 5mC is intriguing. Structural data of both enzymes bound to ap rimer and at emplate reveals significantly different interaction patterns of the respective protein with its substrate. [26] This difference might be the origin of the observed effects.
Next we investigated whether we were able to exploit these findings in polymerase chain reaction (PCR) experiments.W eu sed KlenTaq DNAp olymerase and ap rimer modified at the 3'-terminus.T he modified nucleotide forms ab ase pair with am ethylated or unmethylated cytosine in human genomic DNA( gDNA; Figure 5). We designed af orward primer that should lead to a1 55 bp PCR product and analyzed as pecific CpG site in the promotor region of NANOG in gDNA. [1,27] We used gDNAp urified from HeLa cells because it was shown that this CpG site is unmethylated. [27b] However,a ne nzymatically methylated HeLa gDNA was used as afully methylated control. We found in real-time PCR experiments that discrimination between both un-and methylated cytosine in gDNAw as improved by using the primer bearing O 6 -methyl-G at the 3'-end relative to the unmodified primer ( Figure 5). Theu se of the HeLa gDNA leads to ad elayed amplification of 2.95 AE 0.80 (mean AE standard deviation (SD), n = 5) compared to the fully methylated gDNAw hen paired with the primer modified at the 3'terminus.I nc ontrast, the employment of the unmodified primer leads to adelayed amplification of 0.8 AE 0.23 (mean AE SD, n = 5) in the case of the wild-type HeLa gDNA( Figure 5).
In summary,w efound that O 6 -modified 2'-deoxyguanosine analogues are able to sense the presence or absence of  as ingle methylation in Ci nt he template strand, despite this modification not interfering directly with Watson-Crick nucleobase pairing. Remarkably,o pposing trends were observed when the modified nucleotides were incorporated and extension from the modified nucleotides studied. While in the first case KOD exo À DNAp olymerase discriminates best between Ca nd 5mC by efficient incorporation opposite C, in the latter case KlenTaq DNApolymerase discriminates between both cytosines by more efficient extension from 5mC. We were further able to extend our findings to real-time PCR-based systems for the analysis of the methylation status at asingle Cresidue in human gDNA. These results show the potential of using modified nucleotides for application in sitespecific 5mC detection by real-time PCR, which can be run in parallel for higher throughput without the need for bisulfite treatment. Future investigations will aim at engineering DNA polymerases in combination with modified nucleotides to achieve higher selectivity and elucidate the underlying mechanism in greater detail.