Proteins

Phage T4 polynucleotide kinase and E. coli exonuclease III (Xth) were purchased from New England Biolabs (Ipswich, MA). E. coli ArcticExpress (DE3) cells were from Merck Biosciences (Darmstadt, Germany). The purified BER proteins including truncated version of human uracil–DNA glycosylase (UNGΔ84), E. coli endonuclease IV (Nfo) and human AP endonuclease 1 (APE1) were from laboratory stocks [51]. Expression and purification of TDG and MBD4 is described below. The activities of DNA glycosylases were verified using their canonical substrates immediately before use.

Oligonucleotides

Sequences of the DNA and RNA oligonucleotides used in this work are shown in Table 1. All oligonucleotides were synthesized by Eurogentec (Seraing, Belgium). To follow enzymatic cleavage, the oligonucleotides were 5′-end labelled with γ[³²P]-ATP (PerkinElmer, Shelton, CT) or carried a fluorescent dye Cy5 conjugated to the 5ʹ-end. To obtain duplex substrates, they were annealed with corresponding complementary strands as described previously [52]. The resulting duplexes are referred to as X*•N, where X is a residue in the labelled strand and N is a base opposite to X in the complementary non-labelled strand. It should be noted that the majority of data in this work were obtained with ³²P-labelled DNA substrates, while both Cy5-labelled and radioactively labelled oligonucleotides were used to perform kinetic measurements.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Sequences of the oligonucleotides used in the study.

https://doi.org/10.1371/journal.pone.0304818.t001

Expression and purification of TDG and MBD4

His-tagged human TDG^FL (wild-type and the inactive N140A mutant), TDG^cat and MBD4 proteins were produced as described previously [13, 51]. Briefly, E. coli ArcticExpress (DE3) cells were transformed with the expression vectors pET28c-TDG, pET28c-TDG^N140A, pET28c-TDG^cat or pET6H-MBD4 and grown with vigorous shaking at 37°C in LB medium supplemented with 50 μg/ml of kanamycin or ampicillin to OD₆₀₀ = 0.6–0.8. Then the temperature was reduced to 12°C, the expression was induced by 0.2 mM isopropyl-β-D-1-thiogalactopyranoside, and the cells were further grown for 15 h. All purification procedures were carried out at 4°C. The bacteria were harvested by centrifugation and the cell pellets were lysed using a French press at 18,000 psi in a buffer containing 20 mM HEPES–KOH (pH 7.6), 40 mM NaCl and 0.025% Nonidet P-40 supplemented with Complete^TM protease inhibitor cocktail (Roche Diagnostics, Basel, Switzerland). The lysates were cleared by centrifugation at 40,000×g for 1 h at 4°C, the supernatant was adjusted to 500 mM NaCl and 20 mM imidazole and loaded onto a HiTrap Chelating HP column (GE Healthcare, Chicago, IL). The column was washed with Buffer A1 consisting of with Buffer A1 consisting of 20 mM HEPES–KOH pH 7.6), 500 mM NaCl, 20 mM imidazole, and the bound proteins were eluted with a linear gradient of 20–500 mM imidazole in Buffer B1(20 mM HEPES–KOH pH 7.6), 500 mM NaCl, 500 mM imidazole). The fractions containing the proteins of interest were pooled, diluted tenfold with Buffer A2 consisting of 20 mM HEPES–KOH pH 7.6), 50 mM NaCl and 0,0125% Nonidet P-40 and loaded to a 1-ml HiTrap Heparin column (GE Healthcare, Chicago, IL). Proteins bound to the column were eluted with a 50–800 mM NaCl gradient in Buffer B2 (20 mM HEPES–KOH pH 7.6), 1000 mM NaCl and 0,0125% Nonidet P-40). The eluted fractions were analysed by polyacrylamide gel electrophoresis (PAGE) in the presence of sodium dodecyl sulfate (SDS), and the fractions containing the pure TDG, TDG^cat and MBD4 proteins were stored at −80°C in 50% glycerol. The concentration of purified proteins was determined by the Bradford assay.

DNA repair activity assays

Unless indicated otherwise, the standard reaction mixture (20 μl) for DNA repair assays contained 20 nM ³²P-labelled oligonucleotide duplex, 20 mM Tris–HCl (pH 8.0), 1 mM ethylenediaminetetraacetate (EDTA), 1 mM dithiothreitol (DTT), 100 μg/ml bovine serum albumin (BSA) and 200 nM TDG^FL, TDG^cat or MBD4. The reaction mixtures were incubated at 37°C for 5 to 60 min when measuring TDG-specific activities (T*•G and U*•G duplexes; the asterisk denotes the labelled strand containing the specified base in a defined position) or 1 h and 18 h when measuring TDG-catalysed futile activity (T*•A, T•A*, C*•G and C•G* duplexes). The reaction mixture for Nfo contained 50 mM KCl, 20 mM HEPES–KOH (pH 7.6), 100 μg/ml BSA, 1 mM DTT and 1 nM Nfo. The reaction for Xth was performed in NEBuffer 1 (New England Biolabs) with 1 nM enzyme. The reaction mixture for UNGΔ84 contained 20 mM Tris–HCl (pH 8.1), 100 mM NaCl, 1 mM EDTA, 1 mM DTT, 100 μg/ml BSA, and was incubated for 30 min at 37°C. All reactions were stopped by adding 10 μl of a stop solution consisting of 0.5% SDS and 20 mM EDTA. After incubation, unless indicated otherwise, the samples were treated with 0.1 M NaOH for 3 min at 95°C to cleave AP sites left after base excision by DNA glycosylases, after which the mixtures were neutralized with HCl. In some experiments the samples were treated either with light piperidine (10% v/v piperidine at 37°C for 30 min) or with 10 nM APE1 in TDG buffer supplemented with 5 mM MgCl₂. To analyse the reaction products, the samples were desalted using Sephadex G-25 DNA grade SF column (GE Healthcare) equilibrated in 7.5 M urea and separated by electrophoresis in a denaturing 20% (w/v) polyacrylamide gel (7 M urea, 0.5×TBE, 42°C). The gels were exposed to a Fuji FLA-3000 Phosphor Screen (Fujifilm, Tokyo, Japan), scanned with Typhoon FLA 9500 imager (GE Healthcare) and quantified using Image Gauge v4.0 software (Fujifilm). At least three independent experiments were conducted for all kinetic measurements.

To generate size markers, 20 nM ³²P-labelled duplexes were incubated either with 10 nM Nfo for 5–60 min at 37°C, or with 1 U of Xth for 10 min at room temperature. The reactions were stopped by adding stop solution (10 mM EDTA, 0.25% SDS) and heating at 95°C for 5 min. Alternatively, size markers were generated by using Fast Bisulfite Conversion kit (Qiagen, Venlo, the Netherlands) and UNG. For this, 20 nM ³²P-labelled single-stranded oligonucleotides was treated with sodium bisulfite according to the manufacturer’s instructions and then incubated with 10 nM UNG at 37°C for 10 min followed by hot alkaline treatment to cleave at AP sites.

Single-turnover kinetics

To obtain rate constants (k_obs) not affected by enzyme–substrate association or by product inhibition, such that k_obs reflects the maximal base excision rate (k_obs ≈ k_max), we used single-turnover kinetics under a large excess of the enzyme over the substrate ([E] >> [S] > K_d). The time courses were performed in a large volume with 200 nM TDG and 20 nM substrate at 37°C. At each time point, a 20-μl aliquot was withdrawn and treated and analysed as described above. The data were fitted by non-linear regression to one-phase exponential rise to maximum Eq S1 using GraphPad Prism v5 software (GraphPad Software, Boston, MA): (Eq S1) where A is the maximum fraction product, k_obs is the rate constant, and t is the reaction time.

Bioinformatic analysis

For residue coevolution analysis, sequences of all TDG homologs from chordates were retrieved from the NCBI RefSeq Protein database using human full-length TDG (NP_003202.3) as a query. After excluding incomplete and low-quality entries, the remaining 402 sequences were aligned with Clustal Omega [59] and analysed using CRASP [60] for correlations in physicochemical parameters within amino acid substitutions. Protein disorder and tendency to undergo phase separation was analysed using ParSe v2 [61].

Molecular modelling

The model of full-length TDG generated by AlphaFold2 (AF-Q13569-F1-model_v2) [62, 63] with unfolded N- and C-tails was subject to coarse-grained Monte Carlo molecular simulation in CABS-flex v2.0 [64, 65]. Restraints (linear potential with slope 1 beyond 3.8–8.0 Å) were applied to pairs in which both residues are located in secondary structure elements (helices or sheets) separated by > 3 positions; the temperature factor was set at 1.4. Ten trajectories of 50,000 models were generated; for each, 1,000 models were extracted. After clustering, ten centroid structures for each of the ten trajectories were converted to all-atom representation and analysed using MDTRA [66].

Native TDG excises regular pyrimidines from non-damaged duplex DNA and generates abasic sites

Previously, in our unpublished studies, we observed weak activity of native TDG towards regular DNA oligonucleotide duplexes upon long incubation at 37°C, which we tended to attribute to trace contaminating non-specific nucleases. In the present work, we decided to examine whether TDG^FL can act on regular DNA substrates using the catalytically inactive TDG^FL-N140A mutant as a control for the presence of nuclease contamination from the overproducing bacterial cells. Furthermore, to avoid degradation of a DNA substrate by putative nuclease activities, we constructed a 41-mer dumbbell-shaped duplex (dmbDNA) composed of a long 55-mer strand forming hairpins at both ends (“55-mer hairpin”, Table 1) and a complementary 27-mer oligonucleotide annealing in the middle (14–03; Fig 1 and Table 1). This type of DNA construct was used in our previous study of APE1 [54]. The 27-mer fragment, which contained either A or G at position 16 or U at position 11, was annealed to the 5′-³²P-labelled 55-mer containing a regular T at position 26 (i. e., opposite to A¹⁶ or G¹⁶ mentioned above), to obtain the dmbDNA substrate in which the A¹⁶ or G¹⁶ residues of 14–03 are located in a CpR sequence context (R, purine base; Fig 1B). To examine DNA glycosylase activity on both strands of a duplex, we also used dmbDNA with the 5′-³²P-labelled 27-mer strand annealed to the non-labelled 55-mer hairpin oligonucleotide. The resulting dmbDNA substrates G16•T*, T•A*, A•T* and T•U11* (asterisk denotes the labelled DNA strand with the residue of interest) were incubated in the presence of TDG^FL or TDG^FL-N140A for 1 h or 18 h at 37°C, followed by either light piperidine or hot alkaline treatment to cleave AP site generated by the DNA glycosylase action.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Action of native TDG^FL and catalytically dead mutant TDG^FL-N140A on regular 41-mer 14–03 A*•T and A•T* dmbDNA substrates.

(A) Denaturing PAGE analysis of cleavage products. Lane 1, control 41-mer G•T* dmbDNA containing T at position 26; lane 2, same as lane 1 but treated for 5 min with TDG^FL. Lanes 3–7, 41-mer A*•T dmbDNA containing A at position 26 (27-mer top strand labelled): lane 3, no enzyme; lanes 4–5, TDG^FL, 1 h and 18 h; lane 6, TDG^FL-N140A, 18 h; lane 7, same as lane 5 but treated with NaOH instead of light piperidine. Lanes 8–14, 41-mer A•T* dmbDNA containing A at position 26 (55-mer bottom strand labelled): lanes 8 and 10, no enzyme; lanes 9, 11 and 12, TDG^FL, 5 min, 1 h and 18 h; lane 13, TDG^FL-N140A, 18 h; lane 14, same as lane 12 but treated with NaOH instead of light piperidine. Lane 15, 5′-³²P-labelled 27-mer top strand containing U at position 11; lane 16, same as lane 15, but with UNG. Arrows mark the size of the DNA substrate and the cleavage fragments. (B) Schematic representation of dmbDNA sequence with arrows pointing to the pyrimidines excised by TDG^FL. For details, see Materials and Methods.

https://doi.org/10.1371/journal.pone.0304818.g001

As shown in Fig 1, 5-min incubation of the 41-mer dmbDNA G16•T* duplex with a limiting amount of TDG^FL expectedly resulted in the excision of T opposite to G and the generation of a 25-mer cleavage product from the 55-mer hairpin (Fig 1A, lane 2). Note that the 25-mer product (Fig 1A, lane 2) migrates slower than the 27-mer strands 14–03 of the A*•T duplex and 14-03-U11 of the T•U11* duplex (Fig 1A, lanes 3 and 15), possibly due to the presence of a hairpin structure in the former. Unexpectedly, long 18-h incubation of TDG^FL with A*•T dmbDNA, in which the 27-mer strand 14–03 with A at position 16 is labelled, yielded 24-, 22-, 14-, 10- and 4-mer cleavage fragments (Fig 1A, lane 5), suggesting futile excision of regular C opposite to G in a CpA context (Fig 1B, dashed arrows above the top strand). Moreover, 1-h and 18-h incubation of TDG^FL with A•T* dmbDNA, in which the 55-mer strand with T at position 26 was labelled, produced 38-, 35-, 29-, 25-, 17- and 15-mer cleavage fragments of varying intensities (Fig 1A, lanes 11 and 12), implying the excision of regular T residues opposite to A at positions 39, 36, 30, 26, 18 and 16 in the hairpin oligonucleotide (Fig 1B, arrows below the bottom strand). Notably, in the absence of mismatches, TDG^FL preferentially excised T from positions 30 and 16 of the 55-mer rather than from position 26 as in the control G16•T* substrate. Note that two samples treated by hot alkali instead of light piperidine were not freshly prepared and contained less radioactivity than the rest of the samples because of ³²P decay (Fig 1A, lanes 7 and 14). Mutant TDG^FL-N140A cleaved neither A*•T nor A•T* dmbDNA after 18-h incubation at 37°C (Fig 1A, lanes 6 and 13), indicating that the futile excision of T is not due to a bacterial host contamination but is an intrinsic property of the human DNA glycosylase.

To inquire into the nature of the observed activity, we analysed the reaction products by denaturing PAGE before and after hot alkali treatment. As shown in Fig 2, in the absence of NaOH, incubation of TDG^FL with 14–03 A•T* dmbDNA for 18 h produced only faint smeared cleavage fragments (Fig 2, lane 7), confirming that the enzyme cannot directly cleave the DNA backbone. As expected, when TDG^FL reaction products were treated by hot alkali, distinct cleavage fragments appeared (Fig 2, lanes 8 and 9), suggesting that the observed products are not of an AP lyase or a nuclease origin, but result from TDG^FL excising regular pyrimidines to generate AP sites in duplex DNA. To further substantiate the appearance of AP sites after TDG-catalysed futile excision, we compared the products of several well-established post-treatment methods used to cleave DNA at AP sites. For this we incubated ³²P-labelled dmbDNA G•T* and A•T* substrates with TDG^FL for 1 h and 18 h at 37°C, respectively, and treated the reaction products with hot NaOH, light piperidine, or APE1. Consistent with the respective mechanisms of AP site cleavage, hot NaOH generated a fast-migrating 31-mer fragment with a 3′-phosphate terminus, light piperidine treatment yielded the slowest 31-mer fragment with 3′-terminal α,β-unsaturated aldehyde residue, and APE1 produced an intermediate-mobility 31-mer cleavage fragment with a 3′-OH end (S1 Fig). We observed no significant difference in the quantity of the products between the treatments, indicating that TDG demonstrates G/T-specific and futile activities irrespective of the method used to reveal the nascent AP sites (S1 Fig). Strikingly, TDG^cat (residues 111–308) exhibited dramatically lower futile activity towards 14–03 A•T* dmbDNA than did its full-length counterpart (Fig 2, lane 5 vs lane 9). After 18 h, TDG^cat excised only 0.5–0.8% of T at positions 39, 36, 30, 26, 18 and 16 in the 55-mer strand of 14–03 A•T* dmbDNA, whereas TDG^FL excised 3.3–30.1% of T at the same positions. These results suggest a possible role of non-catalytic N-terminal (residues 1–110) and C-terminal (residues 309–410) tails of human TDG in the futile DNA excision.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Denaturing PAGE analysis of the cleavage products following the action of TDG^cat and TDG^FL on the regular 41-mer 14–03 A•T* dmbDNA substrate.

After the reaction, samples were either treated or not with hot alkali. Lane 1: control 14–03 dmbDNA with 5′-³²P-labelled bottom 55-mer strand; lanes 2–3, same as lane 1 but with TDG^cat, no hot alkali treatment; lanes 4–5, same as lanes 2–3, but with hot alkali treatment; lanes 6–9, same as lanes 2–5 but with TDG^FL. Arrows mark the size of the DNA substrate and the cleavage fragments. The numbers next to the bands in lanes 5, 8 and 9 correspond to the percentage of cleavage products. For details, see Materials and Methods.

https://doi.org/10.1371/journal.pone.0304818.g002

In addition, we examined whether MBD4, a functional analog of TDG, is capable of similar futile T excision from DNA duplexes. We incubated human MBD4 with G•T* and A•T* duplexes for 1 h and 18 h, respectively. As expected, MBD4 efficiently excised mismatched T opposite to G but failed to exhibit detectable activity towards a regular DNA duplex (S2 Fig), suggesting the difference in the mechanism of substrate recognition between these G/T-specific DNA glycosylases.

Kinetics of TDG^FL-catalysed futile thymine excision

We further characterized the substrate specificity of TDG^FL by comparing the kinetic parameters of cleavage of short blunt-ended 27-mer 14–03 and 24-mer 10–05 oligonucleotide duplexes containing either G•T* or A•T* base pair at defined positions in a TpG sequence context. Reactions were performed under single-turnover conditions using a 10-fold molar excess of the enzyme over the substrate, which provides the maximal rate of base excision (k_obs) for a given substrate. As anticipated, product accumulation time courses showed that G•T is preferred by TDG^FL; >80% of T at position 16 in the mismatched 14–03 G•T* duplex was excised already in 5 min at 37°C (Fig 3A and 3B) in contrast to only 21% and 18% of T at positions 16 and 12, respectively, in a regular 14–03 A•T* duplex after 18 h at 37°C (Fig 3C and 3D). Similar large differences between the conventional and futile DNA glycosylase activities were observed in the kinetics of excision of T by TDG^FL from 10–05 G•T* and A•T* duplexes (S3 Fig). From these time courses, we calculated k_obs values for TDG-catalysed canonical and futile activities. As shown in Table 2, when using the 14–03 duplex as a DNA substrate, the k_obs values of TDG^FL-catalysed excision of T opposite to A were, respectively, 330- and 270-fold lower than that of mismatched T (0.47 min⁻¹). When comparing the k_obs values obtained on 10–05 DNA duplexes, the futile activity towards A•T base pair was 2.5–3-fold greater than for the 14–03 duplex while the activity on a mismatched T was only 1.3-fold higher, resulting in 85- and 100-fold differences between the canonical and futile activities on 10–05 substrates (Table 2). These results suggest that the efficiency of both mismatched and normal base excision by TDG^FL markedly depends on the DNA sequence context. Moreover, given a huge excess of normal DNA in the chromatin, the measured rate constants are consistent with the possibility of futile action of TDG in vivo.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Time courses of TDG^FL-catalysed excision of T from G•T and A•T base pairs in the 27-mer 14–03 duplex.

(A, C) Denaturing PAGE analysis of the reaction products. Time courses were performed using 200 nM TDG and 20 nM 5′-Cy5-labelled 27-mer 14–03 duplex containing G•T or A•T base pairs at position 16. Arrows mark the size of the DNA substrate and the cleavage fragments. Percentage of cleavage products is indicated under the gel images. (B, D) Plots of pre-steady-state single turnover kinetic of TDG^FL-catalysed cleavage of 14–03 duplexes. Mean ± SD of three independent experiments is shown. For details, see Materials and Methods.

https://doi.org/10.1371/journal.pone.0304818.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Pre-steady-state kinetic parameters of TDG-catalysed excision of T opposite to A or G in regular DNA duplexes.

https://doi.org/10.1371/journal.pone.0304818.t002

Previously Morgan et al. reported that TDG excises T from a 19-mer T*•A duplex with an exceedingly low rate, ~17,600-fold slower than the activity on a T*•G duplex (k_max = 1.3×10⁻⁵ min⁻¹ vs 0.22 min⁻¹) [49]. In a later study by Maiti et al. from the same laboratory, the futile activity of TDG was measured using a ³²P-labelled 28-mer T*•A duplex and again observed a ~20,000-fold kinetic preference in favour of mismatched T excision [50]. As noted in the Introduction, in order to avoid enzyme inactivation over the long time course obligatory for futile activity measurements, the reaction was run at 22°C in these studies, which may explain the deviation from only 100-fold preference for T*•G vs T*•A substrates observed in our work. It should also be noted that the 28-mer duplex used in [50] had a very particular sequence context with the cleaved strand highly enriched in purines with a single T in a TpG context.

To benchmark our results directly against the published data, we examined the futile activity of TDG^FL on ³²P-labelled 19-mer and 28-mer A•T* duplexes (19AD and 28Adr, respectively; Table 1) used in [49, 50] at 22°C and 37°C under our experimental conditions (S4 and S5 Figs). As expected, after 18 h at 37°C, TDG^FL was able to excise 3–10% of T from A•T base pairs at several positions in the top and bottom strands of 19AD/c19AD and 28Adr/c28Adr duplexes (S4 Fig), indicating that TDG exhibits futile activity in a wide range of sequence contexts. Treatment of a 27-mer 14–03 G•T* duplex with TDG^FL at 22°C and 37°C for 1 h did not show significant difference in the excision of mismatched T (S5A Fig). On the contrary, incubation of a 27-mer 14–03 A•T* duplex with TDG^FL at 22°C and 37°C for 18 h revealed a ~5-fold lower excision of T opposite to A at 22°C (S5A Fig). When TDG^FL was presented with 19AD and 28Adr duplexes in which either strand was labelled, we observed >10-fold less excision of T opposite to A at 22°C than at 37°C (S5B Fig). The futile activity measured on the same 28Adr duplex at 37°C here and at 22°C in [50] revealed more than an order of magnitude difference: 9.9% of T excised in 1080 min (18 h) at 37°C vs 3% in 4000 min (66.7 h) at 22°C (S4 Fig). Yet at 22°C little if any (~1.4-fold) difference was observed between our data and [50] on 28Adr: 0.6% T excised in 1080 min here (S5B Fig) vs 3% in 4000 min in [50]. Taken together, these results suggest that the nature of the enzyme (full-length or truncated), DNA substrate sequence contexts, reaction temperature and incubation time are critical for the detection and characterization of the futile activity of TDG.

Long-term stability of TDG

Time courses shown in Fig 3 and S2 Fig indicate that, even after 6, 9 and 12 h at 37°C, TDG^FL continues to excise T from a regular DNA duplex, suggesting that the enzyme somehow maintains its activity despite the problem of long-term stability. Native TDG contains largely disordered N- and C-terminal tails [46, 67], which were suggested to promote protein aggregation and loss of enzyme activity upon long incubation at 37°C [48]. To examine whether the tails have an impact on the stability of TDG, we kept TDG^cat and TDG^FL in the reaction buffer at 37°C for up to 48 h before adding a ³²P-labelled 24-mer 10–13 T*•G duplex (a conventional TDG substrate) and incubating for another 1 h at 37°C to measure the remaining activity. As shown in Fig 4A and 4B, both truncated and native variants of TDG lost 90% of their activity after 12 h under these conditions (Fig 4A, lanes 4 and 11). Of note, TDG^FL exhibited somewhat higher stability with 13% and 3% activity remaining after 3 h and 12 h, respectively, while TDG^cat retained only 9.7% and 2.0% of its activity at these time points (lanes 10–11 vs lanes 3–4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Loss of the G/T-mismatch-specific DNA glycosylase activity of TDG^cat and TDG^FL upon their prolonged incubation at 37°C.

(A, C, E) Analysis of the TDG-catalysed cleavage products of the 24-mer 10–05 G•T* duplex by denaturing PAGE. Prior to the activity assay, 200 nM TDG was incubated at 37°C in the presence of (A) 0 nM (C) 20 nM, or (E) 200 nM non-labelled regular 10–13 DNA duplex. Percentage of cleavage products is indicated under the gel images. Arrows mark the size of the DNA substrate and the cleavage product. (B, D, F) Plots of the remaining G/T–DNA glycosylase activity of TDG^cat and TDG^FL pre-incubated in the presence or in the absence of the non-labelled regular duplex. Mean ± SD of three independent experiments is shown. For details, see Materials and Methods.

https://doi.org/10.1371/journal.pone.0304818.g004

Previously, Maiti et al. reported that addition of non-specific DNA prevents TDG from inactivation at 37°C for 2 h [48]. Therefore, we verified whether the presence of a regular DNA duplex would maintain the enzymes’ catalytic proficiency at 37°C. Adding 20 nM unlabelled 24-mer 10–13 duplex to 200 nM TDG^FL or TDG^cat slightly stabilized both variants (compare Fig 4C, lanes 3–7, 10–14 and Fig 4D with Fig 4A, lanes 3–7, 10–14 and Fig 4B). Increasing the duplex concentration to 200 nM (i. e, equimolar with the enzyme) did not affect the stability of TDG^cat, as its activity quickly dropped after 3 h and was completely lost after 24 h at 37°C (Fig 4E, lanes 3 and 5). Unexpectedly, supplementing 200 nM TDG^FL with equimolar unlabelled 10–13 regular DNA duplex dramatically prolonged the enzyme survival at 37°C (Fig 4E, lanes 9–14, Fig 4F). TDG^FL maintained full activity after 3 h at 37°C and was able to excise 12% of mismatched T even after a 48-h pre-incubation (Fig 4E, lanes 10 and 13). Overall, these results suggest that the presence of a non-specific DNA duplex preserves the catalytic proficiency of a significant fraction of native TDG but not TDG^cat in the reaction.

Single-stranded oligoribonucleotides can modulate the futile activity of TDG

Recently, it has been shown that TDG binds G-rich single-stranded RNAs with high affinity. As a result, short and long RNA molecules with specific sequences can inhibit TDG-catalysed excision from U•G and T•G mismatches in duplex DNA [58]. We examined the effect of short RNA oligonucleotides with varying sequences taken from [58] on the cleavage of 24-mer 10–05 T*•G and T*•A duplexes by TDG^FL. Consistent with the literature, RNAs unable to form stable complexes with TDG, such as a G-rich RNA hairpin (GC-RNA, Table 1) or single-stranded A-rich RNA (A30-RNA, Table 1) [58], had little or no effect at 0.5–1.0 μM on the G/T-specific TDG activity (Fig 5A, lane 2 vs lanes 3–4 and 7–8; Fig 5B). On the other hand, the presence of 0.5–1.0 μM modified G-rich RNA hairpin (Mut-RNA) containing an internal loop and a 3′-single-stranded tail significantly inhibited TDG activity on the T*•G duplex down from 95% DNA cleavage in the control to 68–74% in the presence of RNA (Fig 5A, lane 2 vs lanes 5–6; Fig 5B). As shown in Fig 5C, the presence of 0.5–1.0 μM high-affinity Mut-RNA also strongly inhibited futile excision of T from a 24-mer 10–05 T*•A duplex down from 54% in the control to 40% and 12% in the presence of 0.5 μM and 1 μM RNA, respectively (Fig 5C, lane 2 vs lanes 5–6). Mut-RNA inhibited the futile activity of TDG to a higher degree than its activity on a T•G mismatch, perhaps because of the differences between TDG binding affinities for RNA vs specific and non-specific DNA. Comparing published TDG binding constants for a T•G mismatch (K_d = 18 nM), a T•A duplex (290 nM), Mut-RNA (140 nM), and GC-RNA (2900 nM) [58, 68], one can see that the affinity of TDG for Mut-RNA is 8-fold lower than for the T•G duplex but 21-fold higher than for the T•A duplex. Therefore at the same concentration Mut-RNA will more strongly compete with TDG futile activity than with its G/T mismatch activity. Interestingly, unlike Mut-RNA, low-affinity GC-RNA and A30-RNA moderately stimulated TDG futile activity, up from 54% cleavage of a 24-mer 10–05 T*•A duplex in the control to 67–71% in the presence of RNA (Fig 5C, lane 2 vs lanes 3–4 and 7–8). These results suggest that non-specific, low-affinity binding of native TDG to dsRNA and G-poor ssRNA may stabilize the protein conformation and thus increase its stability at 37°C, similar to what was observed with non-specific DNA duplexes. The presence of an equimolar concentration of unlabelled non-specific DNA did not influence the futile activity of TDG in a statistically significant manner (Fig 5D).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Impact of RNA oligonucleotides on the G/T-mismatch-specific and futile DNA glycosylase activities of TDG^FL.

(A, C) Analysis of the TDG-catalysed cleavage products of 24-mer 10–05 G•T* (A) and A•T* duplex (C) by denaturing PAGE. Percentage of cleavage products is indicated under the gel images. (B, D) Plots of the amounts of the products of cleavage of 10–05 G•T* (B) and A•T* (D) duplexes. Statistical significance of the differences with the reaction in the absence of RNA was evaluated using two-tailed Student’s t-test (ns, not significant; *, p < 0.05; **, p < 0.01). For details, see Materials and Methods.

https://doi.org/10.1371/journal.pone.0304818.g005

Structural basis of TDG stabilization by non-specific nucleic acids

Intrigued by the ability of supposedly disordered N- and C-terminal tails of TDG to entail stabilization on the protein upon interaction with DNA, we explored possible reasons for this phenomenon. To figure out if there might be any communication between the tails, we have analysed the coevolution of amino acid properties in these regions. Correlated changes in physicochemical properties of residues located far apart in the sequence often reflect functional interactions or dynamic physical contacts between them, which may be not evident from a static structure [69, 70]. Taking all TDG homologs from chordates and discarding those lacking the tails, we have used the CRASP tool [60] to search for correlations in the isoelectric point, local flexibility, hydrophobic index (Fig 6A–6C) and several other parameters reflecting hydrophobicity/hydrophilicity, chain flexibility, surface accessibility and tendency to form secondary structure (S6 Fig). As expected, no coevolution could be observed within the TDG catalytic core due to a high level of conservation and a small number of amino acid changes. The C-terminal tail was also almost free of either internal or cross-domain correlations. This is perhaps not surprising since the C-terminal tail provides a wide-area interface for sumoylation, which regulates many of TDG activities [44, 71, 72], and thus might coevolve instead with SUMO proteins or SUMO E3 ligases. However, multiple coevolving residue pairs were observed within the N-terminal tail, which also often showed correlations with residues at the junction of TDG^cat and the C-terminal tail (Fig 6A–6C and S6 Fig). This suggests that the N-terminal tail might be at least partially ordered and involved in some structural and/or functional cross-talk with other TDG regions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Analysis of structure and properties of TDG disordered tails.

(A–C) Correlation between the isoelectric point [73], local flexibility [74], and hydrophobic index [75] of TDG residues. Red circles under the diagonal, positive correlation; blue circles above the diagonal, negative correlation; circle radii are proportional to the absolute value of correlation coefficient (r_ij) for the property for the given position pair. Only the pairs with p < 0.0001 (|r_ij| > 0.1947) are shown. The rectangles delimit the catalytic domain [44] (residues 111–308; solid lines) and the UDG-F2_TDG_MUG conserved domain as defined in the NCBI Conserved Domain Database [76] (residues 124–293; dashed lines). (D) Color-coded contact frequency map of TDG residues averaged over all coarse-grained simulations. (E) Ten centroid structures from one coarse-grained simulation illustrating conformational variability of the TDG tails. The TDG^cat domain is shown in green as a single experimental structure (PDB ID 5HF7 [46]); the possible phase separation-prone fragment (residues 186–205) is colored cyan. The cores of the simulated structures are superimposed over TDG^cat but omitted from the representation for clarity, leaving only the N-terminal tail (magenta) and the C-terminal tail (red). (F, G) Positions of TDG^cat bound to the DNA minor groove (F) and to the duplex end (G); the latter is obtained by superimposing TDG^cat over the structure of Mug bound to the DNA end (PDB ID 1MTL, chain A [77]. (H) Box plots showing the steric clashes between DNA and the N-terminal tail (N, magenta), TDG^cat (cat, green) and C-terminal tail (C, red) of the simulated structures docked in the groove and end positions. Clashes were counted as occurrences of a protein atom within 2.5 Å of any DNA atom. (I) ParSe predictions for TDG: cyan, residues intrinsically disordered and prone to undergo phase separation; yellow, intrinsically disordered but do not undergo phase separation; black, may or may not be intrinsically disordered but can fold to a stable conformation; white, no prediction. Bar shows TDG fragments with disorder predictions assigned based on the presence of 20 or more contiguous residues that are at least 90% of the same type. (J) Area of the 186–205 fragment occluded by the N-terminal tail (magenta), TDG^cat (green) and C-terminal tail (red) of the model centroid structures in individual simulations.

https://doi.org/10.1371/journal.pone.0304818.g006

At the next step, we sought to probe the conformational space accessible to full-length TDG. To do this, we started with an AlphaFold-generated model, in which both tails are present as extended chains with low confidence score (S7A Fig). The catalytic core is modeled by AlphaFold with a good confidence, with r.m.s.d. = 0.393 over the backbone atoms from the X-ray structure (PDB ID 5HF7 [46]). Since regular full-atom molecular dynamics is poorly suited to sample the huge number of conformations available to long unfolded protein chains, we resorted to coarse-grained Monte Carlo simulations. We used CABS-flex, a modeling tool that implements a knowledge-based force field optimized to provide the best convergence with molecular dynamics of proteins in explicit water at 300 K [64, 65]. Ten independent simulations with random seeds were run, and 1000 models were extracted and clustered for each one. Expectedly, the catalytic core was quite stable during the simulation (r.m.s.d. from the starting model 2.3–3.3 Å) whereas the tails coalesced from the fully unfolded state into better-defined structures, which still demonstrated much higher conformational freedom than the core (Fig 6D, 6E, and S7B–S7K Fig). Interestingly, antiparallel hairpin-like structures were often observed in the tails, especially the N-terminal one (Fig 6D, 6E, and S7B–S7K Fig). Superimposing the models over the structure of TDG^cat bound to substrate DNA in the minor groove (Fig 6F) [46] clearly reveals that a substantial conformational rearrangement of both tails would be required to avoid clashing with DNA (Fig 6H). It is known, however, that Mug, the E. coli homolog of TDG lacking the tails, can bind ends of a DNA duplex in a non-productive mode [77]. When we docked the models to the position of Mug occupying a DNA end (Fig 6G), it significantly relieved clashes by the N-terminal tail (p < 0.005, Student’s paired t-test) and completely abolished them for the C-terminal tail (Fig 6H). In the former, the atoms escaping from the clash were now more frequently found at distances between 2.5 Å and 4 Å from DNA (p < 0.05, Student’s paired t-test), which is permissive for binding interactions. Also, the hairpin-like structures in the N-terminal tail in many cases could be fit into DNA grooves with slight conformational adjustment.

Some, but not all intrinsically disordered protein regions are prone to undergo liquid–liquid phase separation, which is crucial for spatial organization of many processes, including DNA repair, inside the cell [78, 79]. TDG was reported to induce phase separation and condensation in chromatin both in vitro and in cells’ nuclei [80, 81]. To see whether the disorder in the TDG structure may assist phase separation, we used ParSe, a family of physics-based classifiers trained on a set of intrinsically disordered proteins to distinguish between condensate-forming and non-separating polypeptide regions [61]. While both tails were predicted to contain disorder-promoting residues of both types, only the N-terminal tail carried a contiguous stretch of such residues long enough to be classified as intrinsically disordered but not prone to phase separation (Fig 6I and S8 Fig). Interestingly, both the basic ParSe algorithm and two extended classifiers with more physicochemical parameters considered robustly identified a fragment of TDG^cat (residues 186–205) as intrinsically disordered and prone to undergo phase separation. This region of the core was among the least stable in the simulations (r.m.s.f. plots in S7B–S7K Fig) and, even more intriguingly, is partly disordered in the structure of SUMO3-conjugated free TDG [72]. In the models, the 186–205 fragment contacts the N-terminal tail but not the C-terminal one (Fig 6J), raising the possibility of induced disorder that may trigger a conformational change in TDG^cat. Moreover, this fragment also interacts with the end of the duplex in the Mug-like binding mode.

The modeling and bioinformatic exercises allow us to propose a mechanism of TDG^FL protection by non-specific DNA. We suggest that TDG^FL binds ends of a DNA duplex (or, perhaps, even single-stranded DNA or RNA). In the free state, the conformationally relaxed, disorder-prone N-terminal tail might eventually induce disorder in the 186–205 fragment, possibly followed by TDG condensation. The core of isolated TDG^cat where the 186–205 fragment is directly exposed to solution, might also be prone to disorder. However, in the unproductive binding mode, the N-terminal tail is partially diverted to interactions with DNA whereas the 186–205 fragment is protected by its contacts with the duplex end. Thus, while the observed TDG^FL protection by non-specific DNA is clearly an in vitro phenomenon, it could reflect physiologically pertinent features of TDG related to order/disorder transitions in this important epigenetic regulator and guardian.

The presence of G•T and G•U mispairs in DNA inhibits futile excision of neighbouring pyrimidines by TDG

TDG is known as a slow-turnover enzyme that lingers at the AP site left after the mismatched base excision for hours and has to be stimulated by APE1, NEIL1 glycosylase or sumoylation to facilitate product release [82–85]. Therefore, we decided to investigate whether the pattern of futile action is affected in the presence of T•G and U•G mispairs in the middle of a long 63-mer oligonucleotide duplex. As shown in Fig 7, 18-h incubation of ³²P-labelled regular 63-mer T*•A and T•A* duplexes with TDG^FL generated multiple cleavage fragments (55-, 44-, 38-, 20-, 16-, 13- and 7-mers in the T*•A and 57-, 54-, 45-, 29-, 25-, 23-, 10- and 7-mers in the T•A* substrate) which can be mapped to sites of pyrimidines (T and C) excision from TpG, CpA, CpG and TpA contexts (Fig 7, lanes 2 and 4). As expected, 63-mer T*•G duplex was cleaved by TDG^FL in a specific manner, producing a major 31-mer band (Fig 7, lane 6). Consequently, longer futile excision products such as 55- and 44-mer fragments disappeared (since the cleavage is dominated by the mismatched site closer to the labelled 5′-end) while shorter ones (20-, 16-, 13- and 7-mer products) were not affected (Fig 7, lane 6 vs lane 2). The generation of short cleavage fragments suggests that tight binding of TDG^FL to the AP site product of excision of T opposite to G does not suppress the futile activity at distances of 11 nt or more 5′ of the mismatch. However, when we labelled the complementary strand and compared cleavage of T•G* and T•A* 63-mer duplexes, we observed a pronounced decrease in the generation of the 23-, 25-, 29- and 45-mer, but not of the 7-, 10-, 54- and 57-mer cleavage fragments (Fig 7, lane 8 vs lane 4). Thus, the futile excision if strongly inhibited at distances of 14 and 17 nt 3′ of the mismatched G, and 2, 6 and 8 nt 5′ of the mismatched G (Fig 7, lane 8 vs lane 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Action of TDG^FL on 63-mer A*•T, A•T* G•T*, T•G*, G•U*, U*•G, G•C*, and C•G* oligonucleotide duplexes in which either the top or the bottom DNA strand is ³²P-labelled.

Denaturing PAGE analysis of the cleavage products is shown. The DNA bands framed in red boxes denote the difference in the cleavage patterns between regular duplexes vs T•G and U•G duplexes. Arrows mark the size of the DNA substrate and the cleavage fragments. Percentage of cleavage products is indicated under the gel images. The sequence of the 63-mer duplex is shown right to the gel image with red arrows pointing to the pyrimidines excised by the enzyme. For details, see Materials and Methods.

https://doi.org/10.1371/journal.pone.0304818.g007

In addition, we compared TDG^FL-catalysed excision patterns using 5′-³²P-labelled 63-mer U*•G, U•G*, C*•G and C•G* duplexes (Fig 7, lanes 10, 12, 14 and 16). TDG^FL excises mismatched U more efficiently than T (Fig 7, lane 10 vs lane 6), resulting in a stronger inhibition of the futile activity towards neighbouring pyrimidine bases in a U•G duplex compared to a regular C•G duplex (Fig 7, lane 10 vs lane 14 and lane 12 vs lane 16). Again, due to the cleavage of U at position 32, all downstream cleavage fragments (54, 44 and 38 nt long) observed with a C*•G duplex disappeared completely (Fig 7, lane 10 vs lane 14). Moreover, some upstream products of futile excision such as the 27- and 20-mer cleavage fragments also disappeared completely with the U*•G duplex (Fig 7, lane 10 vs lane 14), evidencing suppression of the futile activity on the U-containing strand at distances of 4 and 11 nt 5′ of the mismatched U. In the opposite strand, the products of futile excision 29, 25 and 23 nt long, observed in the C•G* duplex, disappeared completely in the U•G* duplex (Fig 7, lane 12 vs lane 16), showing strong inhibition of the futile activity at the sites 2, 6 and 8 nt 5′ of the mismatched G. At the same time, the presence of a U•G mispair did not suppress futile excision at longer distances, e. g., 15 nt 5′ of U (16-mer cleavage fragment) in the top strand of the U*•G duplex, and at 23 nt 3′ of the mismatched G (54-mer fragment) and 21 nt (10-mer fragment) 5′ of the mismatched G in the bottom strand of the U•G* duplex.

Taken together, these results suggest that TDG^FL tightly bound to DNA after highly efficient excision of T and U opposite to G interferes with the TDG^FL-catalysed futile activity in both strands at short distances (8–11 nt) both 3′ and 5′ of T•G and U•G mismatches in duplex DNA. The most obvious explanation is physical interference of the TDG^FL globule residing on DNA with the association of another enzyme molecule nearby. As the physical footprint of TDG^cat on DNA (1 nt 5′ of the mismatch and 3 nt 3′ of the mismatch in the T/U-containing strand; 9 nt 5′ of the mismatch and 4 nt 3′ of the mismatch in the G-containing strand [86]) is shorter than the interference distance, it is feasible that some of the suppression might be effected through the tails of TDG^FL.

Futile activity of TDG towards enhancer sequences containing C, 5mC and 5hmC residues in a CpG context

Recently, it has been demonstrated that specific sets of enhancers in post-mitotic neurons are hotspots for single-strand breaks [27]. We hypothesized that TDG could generate single-strand breaks in enhancers via its futile activity. To examine this possibility, we measured TDG^FL activity in two known mouse enhancer DNA sequences, neuronal enhancer element mm876 [56] and retinoic acid response element (RARE) involved in the regulation of the Hic1 (Hypermethylated in Cancer 1) gene [57]. As with other regular DNA duplexes, we observed significant futile excision of T and C in TpG, CpA and CpG contexts from both DNA strands of a 63-mer mm876 duplex and a 27-mer RARE duplex (S9 and S10 Figs). These results may suggest that TDG could introduce SSBs in enhancers in vivo if stably bound to DNA for a sufficiently long time.

As epigenetic cytosine modifications are common in regulatory DNA sequences, we then investigated whether TDG futile activity can excise the products of cytosine methylation and demethylation, 5mC and 5hmC. Two 27-mer oligonucleotide duplexes, RARE and 14.CpG (an artificial sequence context rich in CpG), containing C, 5hmC, or 5mC at position 15 were used as models. As shown in Fig 8, 18-h incubation of TDG^FL with unmodified RARE C*•G duplex generated 14- and 11-mer cleavage fragments corresponding to the excision of C residues at positions 15 and 12, respectively (Fig 8, lane 4; 9.8% at both sites). Interestingly, when TDG^FL was incubated with RARE 5hmC*•G and 5mC*•G duplexes for 18 h, the enzyme excised 18% of 5hmC (1.8-fold increase over C) and only 0.8% of 5mC residues (12-fold decrease relative to C) at position 15 (Fig 8, lanes 7 and 10). Similar patterns of excision of C (20%), 5mC (1%) and 5hmC (27%) by TDG^FL were observed for the 14.CpG duplex (S11 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Action of TDG^FL towards the 27-mer RARE oligonucleotide duplex containing C, 5hmC or 5mC at position 15.

(A) Denaturing PAGE analysis of the cleavage products after incubation of TDG^FL with 5′-³²P-labelled 27-mer RARE C*•G, 5hmC*•G and 5mC*•G duplexes for 1 h and 18 h at 37°C. Bisulfite-treated single-stranded 27-mer RARE oligonucleotide was incubated with UNG to generate size markers corresponding to cytosine positions. The numbers next to the bands in lanes 3, 4, 6, 7, 9 and 10 correspond to the percentage of cleavage products. For details, see Materials and Methods. (B) Schematic representation of the RARE sequence with arrows pointing to the pyrimidines excised by the enzyme.

https://doi.org/10.1371/journal.pone.0304818.g008

To quantitatively characterise TDG^FL-catalysed excision of C and its derivatives, we measured the rate constants of base excision (k_obs) at a defined position in ³²P-labelled 27-mer RARE and 14.CpG T*•G and X*•G (X = C, 5mC or 5hmC) duplexes under single-turnover conditions. As shown in Table 3, the k_obs values for excision of C and 5hmC at position 15 opposite to G in the RARE duplex was 110- and 168-fold lower, respectively, than that of mismatched T excision (0.72 min⁻¹). On the 14.CpG substrate, both the regular excision of T and the futile excision of C and 5hmC were ~2-fold higher compared to the RARE duplex, resulting in similar 118- and 145-fold differences between the G/T mismatch and futile activities in 14.CpG duplexes. Under the conditions used in the single-turnover experiments, in both RARE and 14.CpG sequence contexts TDG^FL excised C and 5hmC opposite to G with statistically similar efficiencies (Table 3). We were not able to measure the rate constant of 5mC excision from either duplex because of exceedingly low activity of TDG^FL showing less than 1% of cleavage product formation after 18 h of incubation (Fig 7 and S8 Fig). Comparison of the k_obs values of TDG^FL-catalysed futile excision of T, C and 5hmC shows that, despite some sequence context-dependent variation, the rate constants for T•A, C•G and 5hmC•G base pairs were within the same order of magnitude (Tables 2 and 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Pre-steady-state kinetic parameters of TDG-catalysed excision of T, C, 5mC and 5hmC opposite to G in 27-mer RARE and 14.CpG duplexes.

https://doi.org/10.1371/journal.pone.0304818.t003

Based on these observations, it is plausible that in post-mitotic cells TDG could generate AP sites further converted to SSBs via the BER pathway in the enhancer DNA sequences containing 5hmC residues even without their conversion to 5fC or 5caC. Intriguingly, DNA methylation not only prevents transcription factors binding to enhancers but also strongly inhibits the futile excision of cytosines, thus avoiding appearance of persistent SSBs. It may also follow that TET-mediated generation of 5hmC in enhancer and promoter regions in post-mitotic cells could provide for slow replication-independent DNA demethylation and generation of programmed DNA strand breaks via TDG-initiated BER.