INTRODUCTION

One-carbon unit transfer reactions, including methylation reactions catalyzed by methyltransferases, play an important role in cell metabolism. S-Adenosyl-L-methionine (AdoMet) is a main donor of the methyl group and is the second to ATP in terms of the diversity of biochemical transformations that involve the compound [1]. DNA, RNA, proteins, and low-molecular-weight compounds are substrates of methylation reactions, and AdoMet may be considered as a second reaction substrate. DNA methylation is one of the factors that determine the epigenetic regulation of gene expression in eukaryotes [2, 3]. In mammals, methylation affects the C5 position of cytosine in CpG sequences of DNA, and the distribution of methylated bases forms a methylation pattern, which is established by de novo DNA-(cytosine C5)-methyltransferase (MTase) Dnmt3a [4, 5]. Aberrant DNA methylation is often observed in cancer: overall hypomethylation of the genome is accompanied by local hypermethylation of the promoter region in tumor suppressor genes. A decrease in MTase activity reactivates the genes and suppresses the growth of cancer cells [6].

Modification of the sulfonium center or the amino or carboxyl group in AdoMet usually leads to a partial or complete loss of its affinity for the enzyme [7]. MTase Dnmt3a is distinct from other methyltransferases in being capable of utilizing AdoMet analogs in place of AdoMet. In the case of AdoMet analogs that have the ethyl or propyl group in place of the methyl group of the sulfonium center, the alkyl group transfer rate dramatically decreases with the increasing alkyl group size [8, 9]. Modification of the amino group in the methionine moiety abolishes the substrate properties of AdoMet. Methyl and ethyl esters of AdoMet act as methyl group donors in the reaction catalyzed by prokaryotic CpG-recognizing MTase M.SssI, but are somewhat less efficient than AdoMet [10]. A similarity of catalytic mechanisms between prokaryotic cytosine MTases [11, 12] and the respective eukaryotic enzymes [4] discloses the benefits to study Dnmt3a with the use of AdoMet analogs with a modified HOOC group. It should be noted that the analogs must have a free amino group and the sulfonium center, which are both important for their productive binding to the active center of Dnmt3a [13].

Organophosphorous analogs of AdoMet and S-adenosyl-L-homocysteine (AdoHcy) were used as chemical tools in this work. In these analogs, the carboxyl group was substituted with phosphorus-containing group (Fig. 1). The phosphonous analog of AdoMet (AdoMet-РН) was used as a main instrument of the study. AdoMet-РН is known to be a methyl group donor in tRNA-methyltransferase reactions (with total tRNA and total methyltransferases of Escherichia coli [14]). AdoMet-РН is synthesized in L1210 cells grown in the presence of a phosphonous methionine analog (Met-PH, Fig. 1) [15]. In contrast to AdoMet, AdoMet-РН is chemically stable [16]. Intramolecular alkylation of the carboxyl group producing homoserine lactone and 5’-methyladenosine underlies the mechanisms of chemical hydrolysis of AdoMet and the AdoMet-lyase reaction [17]. Therefore, AdoMet-РН may be considered not only as a functionally active, but also as a catabolically stable AdoMet mimetic. The interactions of the AdoMet and AdoHcy phosphorus-containing analogs (Fig. 1) with DNA methyltransferases have not been investigated so far. Here, we studied whether AdoMet-РН and AdoMet-Р5 (Fig. 1) are methyl group donors in the DNA methylation reaction catalyzed by Dnmt3a and whether the phosphorus-containing AdoHcy analogs AdoHcy-РН and AdoHcy-Р5 (Fig. 1) inhibit the reaction.

Fig. 1.
figure 1

Structures of AdoMet, AdoHcy, their respective phosphonous (AdoMet-PH and AdoHcy-PH) and phosphonic (AdoMet-P5 and AdoHcy-P5) analogs, and the phosphonous methionine analog (Met-PH).

Full size image

EXPERIMENTAL

Materials. The phosphorus-containing analogs of AdoMet and AdoHcy were synthesized as in [18]. AdoMet, AdoHcy, and other reagents and salts were purchased from Sigma (United States), Merck (Germany), and Helicon (Russia). The catalytic domain of Dnmt3a MTase (Dnmt3a-CD) was isolated as described previously [19]. The protein concentration was measured by the Bradford protein assay [20]. Restriction endonuclease Hin6I was purchased from SibEnzyme (Russia). We used the following FAM-labeled 30-bp DNA duplex (fGCGC/CGCGf):

$$\begin{gathered} 5{\kern 1pt} '{\text{-}}\text{FAM}{\text{-}}\text{CTGAATACTACTTG}\underline {\text{CG}} \text{CTCTCTAACCTGAT} \hfill \\ \,\,\,\,\,\,\,\,\,\,\,3{\kern 1pt} '{\text{-}}\text{GACTTATGATGAAC}\underline {\text{GC}} \text{GAGAGATTGGACTA}{\text{-}}\text{FAM} \hfill \\ \end{gathered} $$

which contained a target CpG dinucleotide within the Hin6I recognition site G*CGC (cleavage sites are indicated with asterisk). Oligonucleotides with the fluorescent dye 6(5)-carboxyfluorescein (FAM) attached to their 5' termini through the –NH–(CH2)6– amino linker were obtained from Sintol (Russia). Buffer A contained 20 mM HEPES-NaOH (pH 7.5), 100 mM KCl, 1 mM EDTA, and 1 mM 1,4-dithiothreitol.

Analysis of the DNA·(Dnmt3a-CD)·AdoHcy ternary complex formation. The dissociation constant Kd of the complex was determined by fluorescence polarization, directly titrating a solution of the DNA duplex fGCGC/CGCGf with a Dnmt3a-CD solution. A mixture (120 µL) containing 10 nM fGCGC/CGCGf and 100 µM AdoHcy or its phosphorus-containing analog in buffer A was incubated for 2 min, and fluorescence polarization was measured. Then the mixture was titrated with 2 µM Dnmt3a-CD by adding aliquots (0.5–2.0 µL) to a final enzyme concentration of 700 nM; fluorescence polarization was measured 2 min after each addition. Enzyme concentration dependences of fluorescence polarization were fitted as in [19], using the Hill equation:

$$\theta = \frac{{\left( {P - {{P}_{0}}} \right)}}{{\left( {{{P}_{m}} - {{P}_{0}}} \right)}} = \frac{{{{{\left[ E \right]}}^{n}}}}{{{{{\left[ E \right]}}^{n}} + {{{[{{K}_{d}}]}}^{n}}}},$$

where [E] is the Dnmt3a-CD concentration; P0 and Pm are the fluorescence polarziation values of FAM-labeled DNA in the free and fully bound states, respectively; and n is the Hill coefficient. Fluorescence polarization measurements (λex = 495 nm, λem = 520 nm) were performed using a Cary Eclipse fluorescence spectrophotometer (Varian, United States). Experimental data were fitted using OriginPro 2015.

Substrate properties of the phosphorus-containing AdoMet analogs in DNA duplex fGCGC/CGCGf methylation catalyzed by Dnmt3a-CD. The methylation efficiency of the DNA duplex fGCGC/CGCGf was assessed using inability of Hin6I restriction endonuclease to cleave the methylated site GCmGC, following [19]. The reaction mixture contained 0.3 µM DNA duplex fGCGC/CGCGf, 0.6–10 µM AdoMet or its phosphorus-containing analog (AdoMet-РН or AdoMet-Р5), and 2 µM Dnmt3a-CD in buffer A. The mixture was incubated at 37°C for 60 min. The reaction was terminated by heating the mixture at 95°C for 1 min. DNA cleavage was carried in the same mixture, which was supplemented with 3 µM MgCl2 and Hin6I (2 units per 10 µL of the mixture) and incubated at 37°C for 60 min. The reaction products were analyzed by 20% PAGE with 7 M urea. The DNA fluorescence intensity in the gel was determined using a Typhoon FLA 9500 scanner (GE Healthcare Life Science, Australia) and GelQuantNET software. A mixture of AdoMet or its phosphorus-containing analog with the DNA duplex fGCGC/CGCGf and the same mixture, but with Hin6I were used as controls. The extent of duplex methylation was inferred from the ratio of cleaved and uncleaved DNAs according to the following equation:

$$M = \frac{{{{\omega }_{{{\text{R}}{\text{.Hin6I}}}}} - {{\omega }_{{{\text{Dnmt3a}}}}}}}{{{{\omega }_{{{\text{R}}{\text{.Hin6I}}}}}}} \times 100\% ,$$

where ωDnmt3a is the extent of DNA cleavage after treatment with Dnmt3a-CD and ωR.Hin6I is the extent of DNA cleavage without preliminary methylation.

In the case of reactions with the AdoMet analogs, the extent of methylation was determined relative to that in the respective reaction with AdoMet.

Inhibitory activity of the phosphorus-containing AdoHcy analogs in DNA duplex fGCGC/CGCGf methylation catalyzed by Dnmt3a-CD. Inhibitory properties (IC50) of the AdoHcy analogs were estimated in comparisons with AdoHcy. The reaction mixture contained 0.3 µM DNA duplex fGCGC/CGCGf, 0.6 µM AdoMet, 0–50 µM AdoHcy or its analog, and 2 µM Dnmt3a in buffer A. The mixture was incubated at 37°C for 60 min. The reaction was terminated by heating the mixture at 95°C for 1 min. The reaction products were analyzed as above. IC50 was calculated using OriginPro 2015 as in [19].

RESULTS

The active Dnmt3a form is a tetramer of two Dnmt3a molecules and two molecules of the Dnmt3L regulatory factor. Four Dnmt3a subunits form the active enzyme in the absence of Dnmt3L [21]. A Dnmt3a subunit comprises an N-terminal regulatory region and a C-terminal catalytic domain (Dnmt3a-CD) [22], which is capable of DNA methylation in the absence of the regulatory region [23]. The catalytic properties of the C-terminal domain are identical to those of the full-size enzyme [22]. Murine Dnmt3a-CD was used in this work. Its primary structure is the same as that of human DNMT3A-CD [4, 24], and murine Dnmt3a-CD thus provides an adequate model of human DNMT3A. The FAM-labeled 30-bp DNA duplex fGCGC/CGCGf was used as a substrate [25]. The duplex contains the GCGC site, which is recognized by restriction endonuclease Hin6I. When CpG is methylated in the site, Hin6I does not cleave the methylated site in DNA. The fluorescent FAM labels at the 5' ends of the duplex make it possible to assess Dnmt3a activity via PAGE, by detecting both intact duplex and its cleavage products, the amount of which depends on the extent of duplex methylation [25].

Phosphorus-Containing Analogs of AdoMet and AdoHcy Do Not Affect Hin6I Restriction Endonuclease Activity

The method used to measure Dnmt3a activity is based on the two consecutive enzymatic reactions; substrate DNA methylation and cleavage with Hin6I restriction endonuclease. To study the effects of the AdoMet and AdoHcy phosphorus-containing analogs on Hin6I activity, the analogs were added to a reaction mixture without Dnmt3a-CD to a final concentration of 25 µM and the DNA duplex fGCGC/CGCGf was digested with Hin6I. The extent of duplex cleavage was approximately 90% (Fig. 2) in the presence of AdoMet or either of its analogs (Fig. 1). Similar results were obtained for AdoHcy and its analogs (data not shown).

Fig. 2.
figure 2

Effects of AdoMet and its phosphorus-containing analogs on Hin6I restriction endonuclease activity. (a) Lanes: (1) DNA duplex fGCGC/CGCGf; (2) fGCGC/CGCGf + Hin6I; (3) fGCGC/CGCGf + Hin6I + 25 µM AdoMet. (b) Lanes: (1) DNA duplex fGCGC/CGCGf; (2) fGCGC/CGCGf + Hin6I + 25 µM AdoMet-PH; (3) fGCGC/CGCGf + Hin6I + 25 µM AdoMet-P5. Digestion products were analyzed by 20% PAGE with 7 M urea.

Full size image

Endonuclease digestion is therefore possible to perform in the methylation reaction mixture after adding Mg2+. Based on the findings, we studied the substrate properties of the phosphorus-containing AdoMet analogs and the inhibitory properties of the phosphorus-containing AdoHcy analogs in the DNA methylation reaction catalyzed by Dnmt3a.

Phosphorus-Containing AdoMet Analogs Act as Methyl Group Donors in the DNA Methylation Reaction Catalyzed by Dnmt3a

The nonmethylated DNA duplex is a substrate of Dnmt3a, and the methyl group donor AdoMet is converted to AdoHcy in the enzymatic reaction. The phosphorus-containing AdoMet analogs have a reactive sulfonium center (Fig. 1) and may be considered as methyl group donors although the carboxyl group of AdoMet is replaced with an acidic phosphorus-containing group. Methylation of the DNA duplex fGCGC/CGCGf with the two phosphorus-containing AdoMet analogs (Fig. 1) was studied at four analog concentrations (0.6–10 µM); the reaction was carried out for 60 min because further accumulation of methylated DNA was almost not observed after 60 min. Both AdoMet-PH and AdoMet-P5 were found to be half as efficient methyl group donors as natural AdoMet (Fig. 3).

Fig. 3.
figure 3

Methylation of the DNA duplex fGCGC/CGCGf by MTase Dnmt3a-CD with AdoMet, AdoMet-PH, and AdoMet-P5 used as methyl group donors. (a) Cleavage of fGCGC/CGCGf with restriction endonuclease Hin6I before and after methylation (60 min) as detected by 20% PAGE with 7 M urea. Lanes: (1) fGCGC/CGCGf; (2) fGCGC/CGCGf + Hin6I; (3–5) methylated fGCGC/CGCGf (0.6 µM AdoMet, AdoMet-PH, or AdoMet-P5, respectively) + Hin6I. (b) Comparative methylation efficiency of the DNA duplex fGCGC/CGCGf in the presence of AdoMet, AdoMet-PH, or AdoMet-P5 (0.6, 2, 5, and 10 µM) as a methyl group donor. The degree of methylation was measured relative to that observed with AdoMet; the results were averaged over six independent experiments. The error bars show the standard deviation.

Full size image

Thus, in spite of the substantial structural difference of the phosphorus-containing groups, both AdoMet-PH and AdoMet-P5 acted as methyl group donors in DNA methylation and were comparable in activity with AdoMet.

AdoHcy-РН and AdoHcy-Р5 Form Ternary Complexes with the DNA Duplex fGCGC/CGCGf and Dnmt3a-CD

AdoHcy is a coproduct of methylation reactions and acts as an efficient natural inhibitor of many methyltransferases [26]. AdoHcy interacts with Dnmt3a and the DNA duplex to form a ternary complex, fGCGC/CGCGf⋅Dnmt3a⋅AdoHcy. A possible formation and stability of fGCGC/CGCGf⋅Dnmt3a-CD⋅AdoHcy-РН(Р5) ternary complexes was studied in comparison with the AdoHcy-containing complex by fluorescence polarization. The dissociation constant Kd was calculated for each ternary complex from the results of direct titration. A solution containing the DNA duplex fGCGC/CGCGf (10 nM) and AdoHcy or its phosphorus-containing analog (100 µM) was titrated with a Dnmt3a-CD solution to a final enzyme concentration of 700 nM (Fig. 4). The Hill coefficient n was higher than unity in all cases (Table 1), suggesting cooperative DNA–Dnmt3a-CD binding. Similar Kd values were obtained for the three complexes (Table 1). Thus, the structure of the acidic fragment of AdoHcy only slightly affected the stability of the DNA duplex fGCGC/CGCGf·Dnmt3a-CD·AdoHcy/AdoHcy-РН(Р5) ternary complex.

Fig. 4.
figure 4

Isotherms of the formation of the ternary complexes DNA duplex fGCGC/CGCGf–Dnmt3a-CD–AdoHcy/AdoHcy-РН/AdoHcy-Р5 obtained by fluorescence polarization. The reaction mixture contained 10 nM DNA duplex fGCGC/CGCGf and 100 µM AdoHcy, AdoHcy-PH, or AdoHcy-P5 in buffer A.

Full size image
Table 1.   Dissociation constant Kd and Hill coefficient n estimated for the ternary complexes DNA–Dnmt3a-CD–AdoHcy/AdoHcy-РН/AdoHcy-Р5 in three independent experiments.
Full size table

Phosphorus-Containing AdoHcy Analogs Inhibit DNA Methylation Catalyzed by Dnmt3a

The finding that the phosphorus-containing AdoHcy analogs are as efficient as AdoHcy in stabilizing the Dnmt3a-CD complex with DNA makes it possible to assume Dnmt3a-inhibiting activity for the analogs. Inhibitory activities of the AdoHcy analogs were assayed by varying their concentrations (0–50 µM) in the reaction mixture containing 0.6 µM AdoMet. To calculate IC50 for the three inhibitors, the extent of DNA duplex fGCGC/CGCGf methylation by MTase Dnmt3a-CD was studied as dependent on the AdoHcy, AdoHcy-PH, and AdoHcy-P5 concentrations (Fig. 5). The two phosphorus-containing analogs were found to act as Dnmt3a-CD inhibitors, and their IC50 values were similar to each other and approximately 4 times higher than AdoHcy IC50 (Fig. 5).

Fig. 5.
figure 5

The efficiency of inhibition of DNA duplex fGCGC/CGCGf methylation by MTase Dnmt3a-CD within 60 min as a function of the logarithm of the AdoHcy, AdoHcy-PH, or AdoHcy-P5 concentration. The inset shows the IC50 values calculated for AdoHcy, AdoHcy-PH, and AdoHcy-P5 from the results of three independent experiments. Standard deviations are given.

Full size image

DISCUSSION

Replacement of the carboxyl group with an acidic phosphorus-containing group in amino acids yields various analogs, including aminophosphonic (I) and aminophosphonous (II) acids (Fig. 6).

Fig. 6.
figure 6

Amino acids and their α-aminophosphonic (I) and α-aminophosphonous (II) analogs.

Full size image

Aminophosphonic acids (I) have been studied most comprehensively and act usually as weak competitive inhibitors of the respective amino acid-metabolizing enzymes or, in far rarer cases, as substrates with affinities lower than those of natural amino acids [27, 28]. The differences are explained by the fact that the phosphonic group is a double-charged tetrahedron and thus poorly models the planar carboxyl group. However, several exceptions are known. The phosphonic alanine analog L-Ala-P5 (Fig. 6, R = СН3), which lacks reactive groups, irreversibly inhibits alanine racemase [29], and the phosphonic AdoHcy analog AdoHcy-Р5 (Fig. 1) irreversibly inhibits AdoHcy hydrolase [30]. It should be noted that both S- and R‑isomers are active in the latter case, but S-isomer activity is far higher than R-isomer activity: KI = 345 ± 3 µM, kinact = 0.0049 min–1 and KI = 1094 ± 49 µM, kinact = 0.0145 min–1, respectively [30]. Aminophosphonates are commonly used as transition state analogs of the HOOC group. Efficient enzyme inhibitors have been found among the respective derivatives, and some of them have found application in medicine and agriculture [27].

Aminophosphonous acids (II) have been studied to a far lesser extent as compared with aminophosphonic acids (I) [28]. The single-charged phosphonous group has a flattened tetrahedral geometry and models the planar carboxyl group sufficiently well in this case, and aminophosphonous acids act as substrates of certain enzymes in contrast to aminophosphonic acids. A phosphonous alanine analog (Fig. 6, II, R = СН3) is efficiently converted to a phosphonous pyruvate analog by pyridoxal 5'-phosphate (PLP)-dependent alanine aminotransferase [31, 32], and this pyruvate analog acts as an irreversible nanomolar inhibitor of pyruvate dehydrogenase [33]. Proximal phosphonous analogs of aspartic and glutamic acids (Fig. 6, II, R = HO(O)CCH2– and R = HO(O)CCH2CH2–, respectively) are converted to the respective α-ketophosphonous acids by PLP-dependent aspartate aminotransferase [34]. PLP-dependent tyrosine phenol-lyase cleaves the phosphonous tyrosine analog to produce the phosphonous pyruvate analog [35, 36]. Phosphonous acids have been found to act as substrates of certain NAD-dependent enzymes; i.e., the phosphonous analogs of pyruvate and lactate are readily converted to each other by lactate dehydrogenase [31], and a phosphonous succinic semialdehyde analog is converted to a phosphonous succinate analog by succinic semialdehyde dehydrogenase [37]. Met-PH and AdoMet-PH (Fig. 1) are not the exceptions. Early studies have already shown that Met-PH acts as a substrate in ATP-PPi exchange catalyzed by aminoacyl-tRNA synthetase [38]. Met-PH is a substrate of PLP-dependent methionine γ-lyase and is cleaved to a phosphonous α-ketobutyrate analog [39]. Like natural methionine, Met-PH induces methionine γ-lyase biosynthesis in Citrobacter intermedius cells [40]. These data demonstrate the structural and functional similarity of Met-PH and methionine. Finally, AdoMet-PH production in L1210 cells grown in the presence of nontoxic Met-PH concentrations seems to be one of the factors responsible for their growth inhibition because exogenous AdoMet-PH showed a somewhat greater efficiency as compared with Met-PH [15].

In this work, we studied for the first time whether the phosphonous (AdoMet-PH) and phosphonic (AdoMet-P5) analogs of AdoMet act as methyl group donors in the reaction catalyzed by Dnmt3a. The substrate properties of the analogs were found to be independent of the structure of their phosphorus-containing moiety; each of the analogs was approximately half as active as natural AdoMet (Fig. 3b). AdoHcy-PH and AdoHcy-P5 (Fig. 1) are the analogs of AdoHcy, which is a natural methyltransferase inhibitor. Both phosphorus-containing analogs similarly inhibited Dnmt3a, but their efficiencies were somewhat lower than the AdoHcy efficiency (Fig. 5). It should be noted that preincubation of the enzyme with AdoHcy-PH or AdoHcy-P5 did not affect the inhibition efficiency (data not shown), suggesting that enzyme inhibition is reversible. A comparison of the substrate properties of the AdoMet analogs and inhibitory activities of the AdoHcy analogs between Dnmt3a and total E. coli tRNA methyltransferases shows that, in the latter case, AdoMet-PH is one or two orders of magnitude more active than AdoMet-P5 and its affinity for the enzyme is similar to that of natural AdoMet [14]. A similar pattern has been observed for inhibition of E. coli tRNA methyltransferases by the AdoHcy analogs [14]. The differences between the activities of phosphonous and phosphonic analogs towards Dnmt3a and tRNA methyltransferases are most likely due to the peculiarities of substrate binding to active sites of these enzymes. The binding of the methionine moiety of AdoMet to the MTase active site only slightly depends on the ionic state of the carboxyl group because both methyl and ethyl AdoMet esters are capable of acting as methyl group donors [10].

Studies of the interactions of AdoMet-PH with methyltransferases are important for the design of the regulators of the enzymatic activity because AdoMet-PH will apparently fail to act as a methyl group donor in all methyltransferase reactions. It should be noted that AdoMet-PH, but not AdoHcy-PH, has been detected in L1210 cells grown in the presence of Met-PH [15]. It is possible to assume that AdoMet-PH synthesized de novo provides a poor methyltransferase substrate in the presence of natural AdoMet. It cannot be excluded that AdoMet-PH synthesized de novo is recognized as AdoMet in L1210 cells like Met-PH is recognized as Met in C. intermedius [40]. This assumption is supported by the substrate properties of the phosphonous analogs of amino acids toward the respective enzymes involved in amino acid metabolism (see above) and the phosphonous group may be considered as a bioisostere of the HOOC group. If AdoMet-PH is indeed recognized as AdoMet in L1210 cells, such a recognition may change the AdoMet/AdoHcy ratio, which is a critical factor for efficient methylation reaction, and thus may inhibit the cell growth.

It is certainly of interest that AdoMet-PH is capable of acting as a methyl group donor in the DNA methylation reaction catalyzed by Dnmt3a. Along with the above substrate-like Met-PH transformations, including AdoMet-PH biosynthesis in L1210 cells, our results provide a basis for utilizing the AdoMet analog as a tool to study the reactions catalyzed by methyltransferases. AdoMet-PH is chemically and enzymatically stable and is thus more convenient to work with as compared with AdoMet, which is insufficiently stable under physiological conditions.