Selective excision of 5-carboxylcytosine by a thymine DNA glycosylase mutant

Abstract

The mammalian thymine DNA glycosylase (TDG) excises the mismatched base, uracil, thymine, or 5-hydroxymethyluracil (5hmU), as well as removes 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC) when paired with a guanine. In the previously solved structure of TDG in complex with DNA containing 5caC, the side chain of asparagine 157 (N157) contacts the 5-carboxyl moiety of 5caC via a weak hydrogen bond. We examined the role of N157 in recognition of 5caC by mutagenesis. The asparagine-to-alanine (N157A) mutant has no detectable base excision activity for a G:T mismatch, and its excision activity is reduced for other substrates including G:5caC. Unexpectedly, the asparagine-to-aspartate (N157D) mutant has a comparable base excision rate for G:5caC substrate to that of wild type, but it only has residual activity for G:U and no detectable activity for other substrates. We further show that the N157D mutant has higher activity for 5caC at a lower pH (6.0), suggesting that increased protonation of the carboxylate of 5caC and the aspartate facilitates base excision. The N157D mutant remains highly specific for 5caC even in the presence of large excess of genomic DNA, a property that can potentially be used for mapping the very low amount of 5caC in genomes.

Mammalian DNA cytosine modification is a dynamic process and occurs by converting cytosine (C) to 5-methylcytosine (5mC), established by specific DNA methyltransferases, and then to 5-hydroxymethylcytosine (5hmC) by ten eleven translocation (Tet) proteins. Tet proteins can further oxidize 5hmC to 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC). The mammalian thymine DNA glycosylase (TDG) involves in active DNA demethylation through the removal of deamination products of 5mC (G:T mismatches) or its oxidized derivatives (G:5hmU mismatches) via the base excision repair pathway. Of particular interest are recent reports that TDG can excise 5fC and 5caC (but not 5hmC and 5mC) from DNA. This new specificity of TDG suggests a deamination-independent active DNA demethylation pathway through Tet-mediated oxidation of 5hmC.

Human TDG catalytic domain (residues 111–308) has been crystallized in complex with DNA containing an abasic analog (tetrahydrofuran) ¹¹, a uracil analog (2’-deoxy-2’-fluoroarabinouridine) ¹², an A:5caC mismatch or a modified 5caC (with a 2’-fluoro substitution on the deoxyribose of 5caC) paired with G ¹³, and a post-reactive complex containing a G:5hmU mismatch with the cleaved 5hmU base remaining in a binding pocket of the enzyme ¹⁴. Among the structures of TDG examined so far, the side chain of Asn157 interacts with the 5’ phosphate of the target nucleotide (Fig. 1a-c), which is flipped out from the double stranded DNA. In the structure of TDG with A:5caC ¹³, the 5-carboxyl moiety of 5caC forms a weak hydrogen bond with the side chain amino group of Asn157 (Fig. 1a). Here we study the role of Asn157 in recognition of 5caC by mutagenesis.

Figure 1. N157 of human TDG is involved in DNA binding.

(a) Structure of TDG in complex with an A:5caC mismatch (PDB 3UO7) ¹³. The side chain amino group of Asn157 forms a weak hydrogen bond with the 5-carboxyl moiety of 5caC, which is flipped out from the double stranded DNA. (b) Structure of TDG in complex with a uracil analog (2’-deoxy-2’-fluoroarabinouridine) (PDB 3UFJ) ¹². (c) A post-reactive complex structure of TDG containing a G:5hmU mismatch with the cleaved 5hmU base remaining in the active site pocket (PDB 4FNC) ¹⁴. The side chain of Asn157 interacts with the 5’ phosphate of the target nucleotide. (d-f) The activities of TDG wild type (WT) (panel d), mutants N157A (panel e) and N157D (panel f) on G:X under the single turnover condition ([S_DNA]=0.25 µM and [E_TDG]=2.5 µM) after 10 min reaction at 30°C in 0.1% BSA, 1 mM EDTA, 100 mM NaCl and 10 mM BisTris HCl, pH 6.0. Various 32 bp oligonucleotides labeled with 6-carboxy-fluorescein (FAM) were used as substrates: (FAM)-5’-TCG GAT GTT GTG GGT CAG XGC ATG ATA GTG TA-3’ (where X=C, 5mC, 5hmC, 5fC, 5caC, U, T and 5hmU) and 5’-TAC ACT ATC ATG CGC TGA CCC ACA ACA TCC GA-3’. Human TDG residues 111–308 (pXC1056) and its mutants N157A (pXC1155) and N157D (pXC1138) were prepared as described ¹⁴. The excision reaction was monitored, as described ¹⁴, by denaturing gel electrophoresis following NaOH hydrolysis of the abasic site.

N157A does not affect substrate specificity

We first mutated Asn157 to alanine (N157A) and compared its glycosylase activity to that of wild type (WT) using various 32-base-pair (bp) DNA oligonucleotides, each containing a single modified base X (=C, 5mC, 5hmC, 5fC, 5caC, U, T and 5hmU) within a G:X pair in a CpG sequence. As expected, glycosylase activity was not observed with TDG on oligonucleotides bearing the ‘natural’ G:C, G:5mC and G:5hmC base pairs, while substrates bearing G:T, G:U and G:5hmU mismatches and G:5fC and G:5caC ‘wobble’ pairs ¹⁴ were efficiently cleaved by the WT enzyme (Fig. 1d). Instead of losing activity on G:5caC as we had expected, N157A retains reduced but measurable activity on G:U, G:5hmU, G:5fC and G:5caC and no detectable activity on G:T substrate (Fig. 1e). G:T mismatch is known to be a weak substrate even for WT TDG ¹⁵ (see Fig. 1d, lane T). Further reduction of this weak activity could render the N157A inactive on G:T substrate.

N157D selectively excises 5caC

Mutation of Asn157 to aspartate (N157D) resulted in loss of detectable activity on G:T and G:5hmU and only residual activity on G:U and G:5fC (Fig. 1f). Surprisingly, the N157D mutant remains fully active on G:5caC substrate (Fig. 1f and 2a). The activity of N157D mutant on G:5caC substrate is very sensitive to pH (Fig. 2b) and salt (Fig. 2c).

Figure 2. N157D mutant selectively cleaves 5caC.

(a) The time course (0–60 min) of N157D mutant reaction on seven oligonucleotides with various modifications under the single turnover condition at 30°C, pH 6.0 and 100 mM NaCl. For G:5caC substrate, more time points between 0 to 10 min were also taken. (b) N157D has enhanced activity on G:5caC at lower pH. Four different pH values between 6.0 and 7.2 with 0.4 increments were used at 10 mM BisTris HCl, 100 mM NaCl and 30°C. The intensities of the FAM labeled DNA were determined by Typhoon Trio+, and quantified by the image-processing program ImageJ. The data were fitted to non-linear regression using software GraphPad PRISM 5.0d (GraphPad Software Inc.): [Product] = P_max(1-e^−kt), where P_max is the product plateau level, k(min⁻¹) is the observed rate constant, and t is the reaction time. (c) The activities of N157D on G:5caC at 30°C were measured at pH 6.0 with varying salt concentration between 0–250 mM NaCl with 25 or 50 mM increments. The K_obs values (min⁻¹) were calculated for reactions between 0–10 min. Plot of activity as a function of NaCl concentration after 2.5 min reactions. (d) The log plots of activities of WT TDG (lines 1, 2 and 3) and N157D mutant (line 4) as a function of pH. For lines 1 and 2, the data were measured from a mixture of three buffers ¹⁴, whereas lines 3 and 4 were measured from 10 mM BisTrisHCl as indicated in the legend of panel b.

We measured the mutant N157D activities on G:5caC substrate under single turnover conditions (i.e. [E_TDG] ≫ [S_DNA]) at a pH range between 6.0 and 7.2 with an interval of 0.4 units. We found that the mutant TDG has higher activity by a factor of approximately 2 for the G:5caC at pH 6.0 (k_obs=0.85 min⁻¹) as compared to the activity at pH 6.4 (k_obs=0.4 min⁻¹) (Fig. 2b). Changing the pH to 6.8 weakened the activity by another factor of 2.5 with k_obs=0.16 min⁻¹. Further increasing the pH to 7.2 resulted in accumulative loss of activity by a factor of 17 (k_obs=0.05 min⁻¹), as compared to the activity at pH 6.0. Like WT TDG ¹⁴, the continuing enhancement of activity on G:5caC at lower pH values suggests that the chemical nature of the target nucleotide and the aspartate side chain can contribute to the mutant being active on 5caC. We reasoned that the protonated forms of 5caC and/or the aspartate side chain at lower pH could maintain the interaction between the DNA phosphate and N157D as well as the interaction between N157D and 5caC.

We note that TDG activity on the G:U substrate is relatively constant between pH 5.5–9 ¹⁶, with a slope of 0.1 in a log plot of activity against pH (Fig. 2d, line 1). The enhancement of WT activity on G:5caC at lower pH suggests that the increasing protonation of the carboxylate group of the target nucleotide, as indicated by the linear fits under two different buffer systems (Fig. 2d, lines 2 and 3), can contribute to catalysis. The slope is even steeper for N157D mutant (1.0 vs. 0.5 for WT; Fig. 2d, line 4), suggesting that the protonation of the target nucleotide 5caC and the aspartate side chain both contribute to catalysis. However, the exact values of the slopes do not exactly match the proposed number of protonation events (one for WT and two for N157D). This could be complicated by other residues in the active site, particularly His151 ¹², whose protonation might negatively impact catalysis. Additional study will be required to settle this point.

N157D maintains its specificity for 5caC in the presence of large excess of genomic DNA

Genomic 5fC and 5caC contents are very low, approximately 2×10⁻⁵ and 3×10⁻⁶ per cytosine, respectively, compared to 1.3×10⁻³ of 5hmC, in embryonic stem cells ⁵. A separate study estimated the ratio of 5fC to be about 10 fold higher (2×10⁻⁴), although the amount of 5caC was not measured ¹⁷. One potential application of the N157D mutant is in mapping genomic 5caC sites. The product of TDG reaction is abasic nucleotide, which can be readily selected and analyzed from the rest of the DNA using biotin containing Aldehyde Reactive Probe (ARP) ¹⁸.

We tested whether N157D protein can maintain its specificity for 5caC in the presence of large excess of genomic DNA. An excision reaction (200 µl) was performed with 2.5 nM of 5caC in FAM-labeled oligonucleotide in the presence of 100 µg of salmon sperm DNA (with the ratio of 5caC to C to be ~7×10⁻⁶). When 25 µM of N157D was used, 5caC cleavage was near completion after 10 min, while little cleavage of U had occurred (Fig. 3a), confirming that the specificity of N157D was not affected in the presence of large amount of genomic DNA. We noted, however, large excess of N157D enzyme was required to rapidly cleave 5caC (Fig. 3b), likely due to nonspecific DNA binding by N157D while searching for its target.

Figure 3. Detection of 5caC in the presence of large excess of genomic DNA.

(a) Specificity of N157D (5caC vs. U) in the presence of large access of genomic DNA. The reaction (200 µl) contains 25 µM of N157D, 2.5 nM of 32-bp FAM-labeled G:X (X=5caC or U) oligonucleotides and 100 µg (or 0.5 mg/ml) of salmon sperm genome DNA (Rockland Inc.). The G+C contents of salmon DNA are approximately 44.4% ²⁴, and thus the salmon DNA used in the reaction contains 757 µM of base pair and 360 µM G:C pairs. The 5caC/C ratio (2.5 nM/360 µM) was estimated to be 7×10⁻⁶. (b) Excision reactions (20 µl) contain 25 µM (top panel) or 2.5 µM (bottom panel) of N157D, 25 nM of 32-bp FAM-labeled G:5caC oligonucleotides and 10 µg of salmon sperm genome DNA.

Discussion

Of the three oxidized forms of 5mC generated by Tet proteins, 5fC and 5caC are present in much lower amounts than that of 5hmC. The low levels of 5fC and 5caC are likely due to the observation that both bases can be excised by TDG and repaired by a base excision process. However, a TDG-knockout shows only a minor increase of 5caC in mouse embryonic stem cells ⁶. Another scenario is that only a subset of 5hmC is oxidized to 5fC and even a smaller amount of 5fC becomes 5caC, due to a much slower rate of conversion from 5hmC to 5fC and 5fC to 5caC than that of 5mC to 5hmC. In vitro analysis of Tet2 suggested that this is indeed the situation ⁵. To understand the function and regulation mechanism that control the levels and genomic distribution of 5fC and 5caC, we need to know the genomic locations of 5fC and 5caC accurately, preferably at single-base resolution.

Asn157 of TDG was initially identified as the residue making a specific weak interaction with 5caC base from a low-resolution structure, which was determined by X-ray diffraction data to 3 Å resolution along the a and b axes and asymmetrically to 4 Å along the c axis ¹³. Unexpectedly, the N157D mutant became specific for 5caC among the substrates examined, a property that can potentially be used for mapping the very low amount of 5caC in genomes.

To successfully map 5caC, genomic DNA should be first treated with uracil-specific DNA glycosylase (UDG) to remove any G:U mismatches that can be substrate for the TDG N157D mutant, albeit very slowly. Subsequently, the DNA must be treated with NaBH⁴ (0.1%) to eliminate traces of potentially reactive aldehyde including endogenous abasic sites and product of UDG, before using N157D protein to convert 5caC to abasic sites. DNA fragments containing abasic sites will bind to Aldehyde Reactive Probe (ARP), be selected by avidin bead ¹⁸ and subjected to sequencing analysis. One must note that abasic site will somewhat hinder amplification by DNA polymerase, resulting in incorporation of dATP at the abasic site and occasionally cause one or two base pairs deletion ¹⁹, both of which can be used as an indication of 5caC presence in the original DNA. Using wild type TDG by the same method, both 5fC and 5caC sites could be mapped.

TDG belongs to the family of DNA glycosylases with similar architectures including UDG, mismatch-specific uracil glycosylase (MUG), and single-strand specific monofunctional uracil DNA glycosylase (SMUG1) (reviewed in). Like TDG, Escherichia coli MUG possesses an Asn35, which corresponds to Asn157 of TDG ¹⁴, and is capable of excising G:5caC in addition to G:U mismatch. In addition, SMUG1 excises U and 5hmU ²³, but like UDG and MUG is inactive against thymine. This difference of substrate specificity within the family members has yet to be studied structurally, (bio)chemically and thermodynamically.

Highlights.

• Human TDG N157D mutant has a selective base excision for G:5caC substrate

• N157D mutant has higher activity for 5caC at a lower pH

• N157D remains specific for 5caC under the ratio of 5caC/C of ~7×10^−6>