Defining the impact of sumoylation on substrate binding and catalysis by thymine DNA glycosylase

Abstract

Thymine DNA glycosylase (TDG) excises thymine from mutagenic G·T mispairs generated by deamination of 5-methylcytosine (mC) and it removes two mC derivatives, 5−formylcytosine (fC) and 5−carboxylcytosine (caC), in a multistep pathway for DNA demethylation. TDG is modified by small ubiquitin-like modifier (SUMO) proteins, but the impact of sumoylation on TDG activity is poorly defined and the functions of TDG sumoylation remain unclear. We determined the effect of TDG sumoylation, by SUMO-1 or SUMO-2, on substrate binding and catalytic parameters. Single turnover experiments reveal that sumoylation dramatically impairs TDG base-excision activity, such that G·T activity is reduced by ≥45-fold and fC and caC are excised slowly, with a reaction half-life of ≥9 min (37°C). Fluorescence anisotropy studies reveal that unmodified TDG binds tightly to G·fC and G·caC substrates, with dissociation constants in the low nanomolar range. While sumoylation of TDG weakens substrate binding, the residual affinity is substantial and is comparable to that of biochemically-characterized readers of fC and caC. Our findings raise the possibility that sumoylation enables TDG to function, at least transiently, as reader of fC and caC. Notably, sumoylation could potentially facilitate TDG recruitment of other proteins, including transcription factors or epigenetic regulators, to these sites in DNA.

INTRODUCTION

Thymine DNA glycosylase (TDG) initiates base excision repair (BER) by removing modified bases from DNA, including those resulting from deamination or oxidation of 5-methylcytosine (mC), and it has important functions in DNA repair and epigenetic regulation (1). TDG was discovered for its ability to excise thymine from G·T mispairs (2,3), as needed to preclude C→T transitions arising via mC deamination. It also functions in a multistep pathway for active DNA demethylation, which likely explains findings that depletion of TDG causes embryonic lethality in mice (4,5). The established DNA demethylation pathway begins with a ten-eleven translocation (TET) enzyme, followed by TDG and then BER (6). TET enzymes oxidize mC to give 5-hydroxymethylcytosine (hmC), 5-formylcytosine (fC) and 5-carboxylcytosine (caC); the latter two are excised by TDG and follow-on BER yields unmodified cytosine (7–13). Human TDG (410 residues) has a central catalytic domain (∼195 residues) flanked by two disordered regions that mediate various functions and are subject to post-translational modifications (PTMs) including acetylation, phosphorylation, ubiquitination, and SUMO (small ubiquitin-like modifier) conjugation (Figure 1A) (14–23). TDG also has a SUMO-interacting motif (SIM) that binds non-covalently to SUMO proteins, including conjugated (intramolecular) SUMO domains (Figure 1B), and SUMO–SIM binding likely mediates many of the effects of SUMO conjugation on TDG function (18,24–26). However, the impact of sumoylation on TDG activity is poorly defined and the functions of TDG sumoylation remain unclear. In the studies reported here, we investigated the effects of SUMO conjugation (sumoylation) on the biochemical activities of TDG.

Figure 1.

SUMO modification of TDG. (A) Primary structure of TDG including the catalytic domain and two flanking disordered regions that mediate various functions. Shown are post-translational modification (PTM) sites, the SUMO-interacting motif (SIM), and the PIP degron motif that enables TDG interaction with PCNA, ubiquitin modification, and degradation. The site(s) of ubiquitin modification is currently unknown. (B) Crystal structure (2.1 Å) of the TDG catalytic domain (residues 112–339) modified by SUMO-1 (green) (PDB ID: 1WYW); residues 301–331 of TDG (magenta) exhibit some secondary structure (β-strand, α-helix) for sumoylated TDG but are likely disordered in unmodified TDG. Because no structure is available for sumoylated TDG bound to DNA, the DNA shown here (orange) was positioned by aligning a structure (1.54 Å) of DNA-bound TDG (PDB ID: 5HF7, hidden) with that of TDG∼SUMO-1 (shown), using PyMol (RMSD = 0.573 Å for 136 Cα atoms). While this model cannot be accurate, because the SUMO-induced helix of TDG overlaps with the DNA, it is useful for illustrating the location of the SUMO–SIM interface relative to the DNA binding region and the active site.

It is informative to briefly review what is known about TDG sumoylation. In an initial study, Schär et al. showed (by western blots) that TDG is modified by SUMO proteins, in multiple types of human cells, and that recombinant TDG is modified similarly in cell extracts (18). Subsequent studies confirmed these findings and demonstrated sumoylation of TDG in vitro, using recombinant forms of the SUMO-activating (E1) and SUMO-conjugating (E2) enzymes and a SUMO protein, or by co-expression of these factors with TDG in bacterial cells (20,21,25,27–29). Sumoylation occurs exclusively at a single Lys residue in a consensus site (VKEE) (30) located in the disordered C-terminal region of TDG (Figure 1A). TDG is modified by SUMO-1, SUMO-2 or SUMO-3 (18,25,26); the latter two are virtually identical (referred to as SUMO-2/3) and they share ∼48% amino acid sequence identity with SUMO-1 (31). SUMO domains that are tethered to TDG can bind the SIM of the same TDG molecule via intramolecular SUMO·SIM binding, as shown by crystal structures of a sumoylated form of TDG that includes the catalytic domain and the SUMO modification site (residues 112–339) (Figure 1B) (25,26). Notably, the structures indicate that SUMO–SIM binding induces the formation of secondary structure (α-helix, β-strand) in a C-terminal region of TDG (D316-K330) that is otherwise likely to be disordered (19). While there are no reported structures of sumoylated TDG (TDG∼SUMO) bound to DNA, it has been proposed that the SUMO-induced α-helix could impair DNA binding (Figure 1B) (25,26). Although the crystal structures suggest a stable intramolecular SUMO–SIM complex, it is possible that SUMO–SIM binding is dynamic in aqueous solution, particularly for DNA-bound TDG∼SUMO and/or for sumoylated full-length TDG.

Initial studies reported that sumoylated TDG does not bind to DNA containing a G·T or G·U mispair, a G·C pair (nonspecific DNA), or even to a G·AP product site, suggesting that sumoylation abolishes detectible DNA binding (18,25,26). The effects on catalytic turnover (k_cat) of TDG were also studied, with the finding that sumoylation completely abrogates detectible processing of G·T substrates while it enhances k_cat for G·U substrates (18). Multiple groups had shown that TDG, like many DNA glycosylases, binds tightly to its AP-DNA product, severely impeding catalytic turnover in vitro (32–35). It was proposed that sumoylation regulates TDG product release and thereby enables efficient catalytic turnover (k_cat). More specifically, the model holds that sumoylation occurs selectively for product-bound TDG and that SUMO is removed from TDG after product release to enable processing of other substrates, such that SUMO modification and deconjugation occurs for each catalytic cycle of TDG (18). While this paradigm seems to be broadly accepted (25,31,36–38), it is not directly supported by experimental evidence and remains to be substantiated. Moreover, the model has been challenged by recent findings (20,21), and many studies have shown that follow-on BER proteins, which act on glycosylase-generated AP sites, can dramatically enhance the catalytic turnover of TDG, as discussed below (21,39,40).

To understand the potential role(s) of TDG sumoylation it is important to define the effects of SUMO modification on discrete activities of TDG. While the general perception may be that much is known about how sumoylation impacts TDG activity, in fact, the opposite is true. For example, in previous studies the kinetics experiments were performed under multiple turnover conditions, often without a determination of K_m (hence with unknown [S]/K_m) (18,28,29). As such, the reported rate constants were influenced by many reaction steps, including enzyme-substrate association, product release, and product inhibition (Figure 2) and are therefore of limited value for understanding the effect of sumoylation on TDG activity. To date, no studies have examined the impact of sumoylation on TDG glycosylase activity using single turnover experiments, which yield the maximal rate of base excision when collected under saturating enzyme conditions (41). In addition, previous studies into the impact of TDG sumoylation on DNA binding were performed with qualitative methods (electrophoretic mobility shift assay or EMSA) and a truncated form of the enzyme (TDG^112–339) that lacks the N-terminal region, which has been shown to enhance DNA binding (16,28,42,43) and might interact with the SUMO domain of modified TDG (28). No prior studies have used a quantitative method that yields a dissociation constant (K_d) or used full-length TDG that is uniformly modified with a single SUMO isoform (e.g., SUMO-1 only). Thus, the effect of sumoylation on two critical phases of the TDG reaction, substrate binding and base excision, remain unknown for G·T substrates. Moreover, the effect of sumoylation on TDG activity for the epigenetic substrates G·fC and G·caC has not been investigated. We addressed these problems in the studies described below.

Figure 2.

Minimal kinetic mechanism for the TDG reaction. Association of TDG (E) and DNA substrate (S) gives a collision complex (E·S), and the reversible nucleotide flipping step (K_flip), involving conformational changes in E and S, gives the reactive enzyme-substrate complex (E′·S′). Cleavage of the N-glycosyl bond and addition of the (water) nucleophile in the chemical step (k_chem) gives the ternary product complex (E′·B·P′). Dissociation of E′·B·P′ likely involves rapid release of the excised base (B) and much slower release (k_dis) of abasic DNA (P). TDG is severely inhibited by abasic DNA but does not bind to the nucleobases it removes from DNA (51). Solid lines denote reaction steps that contribute to enzymatic rate constants (k_max, k_cat), and the dissociation constant (K_d) for the reactive (E′·S′) complex. Note that catalytic turnover (k_cat) is influenced by multiple steps, including dissociation of the initial product complex (k_dis) and subsequent product inhibition.

MATERIALS AND METHODS

Materials

Full length human TDG (410 residues) and the N140A variant (TDG^N140A) (44) were expressed in Escherichia coli growing in Terrific Broth (TB) and purified as described (45). TDG (or TDG^N140A) conjugated by SUMO-1 or by SUMO-2 was produced in E. coli that was co-transformed with two expression plasmids, one for TDG (or TDG^N140A) and another for human SUMO-1 or SUMO-2 (mature form) together with the activating E1 (SAE1-SAE2) and conjugating E2 (Ubc9) enzymes that mediate sumoylation (46), as described (20). Cells were grown in TB and expression was induced with 0.4 mM IPTG at 22°C for ∼16 h; sumoylated TDG was purified as described for unmodified TDG (45), with the final chromatographic step performed using a Mono Q anion exchange column (GE Healthcare). Protein preparations were >99% pure as determined by SDS-PAGE. The SUMO-conjugated protein was free of detectible unmodified protein, as shown by Western blots (Supplementary Figure S1). Purified protein was flash frozen and stored at −80°C. Protein concentration was determined by absorbance at 280 nm (47), using extinction coefficients of ϵ²⁸⁰ = 31.5 mM⁻¹cm⁻¹ for TDG (48), ϵ²⁸⁰ = 37.8 mM⁻¹cm⁻¹ for TDG∼SUMO-1 and ϵ²⁸⁰ = 34.8 mM⁻¹cm⁻¹ for TDG∼SUMO-2; ϵ²⁸⁰ is unchanged by the N140A mutation.

Standard oligodeoxynucleotides (ODNs) were obtained from IDT. ODNs that contained fC, caC or fluorescein (FAM) were synthesized by the Keck Foundation Biotechnology Research Laboratory at Yale University. ODNs containing 2′-fluoroarabinodeoxythymidine (T^F) were also synthesized at Yale using a phosphoramidite from Link Technologies (44). The subtle 2′-F substitution precludes TDG cleavage of dT, dU, 5-formyl-dC and 5-carboxyl-dC, due to transition-state destabilization (44,49–51). Crystal structures show that the 2′-F analogs of dU, 5-formyl-dC, and 5-carboxyl-dC flip completely into the TDG active site, forming key E·S interactions (43,50–52). ODNs modified with a 5′ sulforhodamine (Texas Red, or TR) were synthesized by Midland Certified Reagent Company (Midland, TX). ODNs were purified by reverse phase HPLC as described (53), and purity was confirmed by denaturing anion-exchange HPLC (54). Purified ODNs were exchanged into 0.02 M Tris–HCl pH 7.5, 0.04 M NaCl, and their concentration was determined by absorbance (45). Duplex DNA included a 28mer target strand, 5′-GTG TCA CCA CTG CTC AxG TAC AGA GCT G-3′, where x is the target base (fC, caC, T, U, T^F), and a complementary strand, 5′-CAG CTC TGT ACG TGA GCA GTG GTG ACA C-3′, such that the target base (x) is paired with G and located in a CpG context (underlined). FAM and TR labels were attached to the complementary strand.

Glycosylase assays

Glycosylase (base excision) activity of sumoylated and unmodified TDG was determined using kinetic experiments performed using either single turnover conditions (saturating enzyme) to obtain the maximal rate of base excision (k_max), or multiple turnover experiments (saturating substrate) to obtain the maximal rate of catalytic turnover (k_cat) (21,41,55). Assays were performed at 37°C in HEN.1 buffer (0.02 M HEPES pH 7.5, 0.1 M NaCl, 0.2 mM EDTA) with 0.1 mg/ml BSA. The 28 bp duplex DNA substrates contained a single G·fC, G·caC or G·T base pair. Single turnover assays included 0.5 μM DNA substrate and 1.5 μM enzyme, while multiple turnover assays used 1.0 μM DNA substrate and 0.05 μM enzyme. Kinetics reactions were initiated by adding concentrated enzyme to substrate in reaction buffer, then aliquots were removed at selected time points, rapidly quenched with 3× quench buffer (0.3 M NaOH, 0.03 M EDTA), and heated at 85°C for 3 min to quantitatively cleave the target DNA strand at TDG-generated abasic sites. The resulting DNA fragments were resolved by denaturing anion-exchange HPLC, using a DNAPac PA200 column (Thermo Fisher), and peak areas were used to determine fraction product (54). For single turnover experiments, progress curves (fraction product versus time) were fitted by non-linear regression to Equation (1) using Grafit5 (56).

(1)

where A is the amplitude, k_obs is the rate constant, and t is reaction time. Experiments were performed with saturating enzyme ([E] >> K_d; [E] > [S]) such that the observed rate constant approximates the maximal rate of product formation (k_obs ≈ k_max) and is not influenced by enzyme–substrate association or by product release or product inhibition (41). Saturating conditions were confirmed by observation of identical rate constants for experiments performed at other enzyme concentrations. For multiple turnover experiments, the linear (steady-state) portion of the progress curve was used to obtain the initial velocity (v₀). Under these saturating substrate conditions ([S] ≫ K_m), the Michaelis-Menten equation simplifies such that v₀ approximates the maximal rate of catalytic turnover (k_cat = v₀/[E]).

Equilibrium binding monitored by fluorescence anisotropy

Fluorescence anisotropy experiments were used to monitor binding of enzyme (TDG^N140A or sumoylated derivatives) to DNA containing a single G·fC, G·caC or G·T^F pair and Texas Red (TR) at the 5′-end of the non-target strand (48,57). Individual samples (125 μl) contained DNA at a final concentration of 1 nM (G·fC, G·caC) or 10 nM (G·T^F) and a varying enzyme concentration (0.01 nM to 10 μM) in binding buffer (20 mM HEPES pH 7.5, 0.1 M NaCl, 0.2 mM EDTA, 0.1 mg/ml BSA). Samples were equilibrated at 22°C for 1 h and then transferred to a quartz cuvette (Starna Cells). Anisotropy data were collected using a PTI Quantamaster 40 spectrofluorometer configured in T format, where one of the two PMTs is connected directly to the sample compartment, with no monochromator, and with wavelength selection provided by a 628-nm band pass filter (Semrock, Inc.) (48,57). The excitation wavelength was 590 nm (6-nm band pass), and the single-emission monochromator was set to 615 nm (8-nm band pass). Wavelength selection of the excitation and emission monochromators was enhanced by 586 nm and 624 nm band-pass filters, respectively (Semrock, Inc.). Dissociation constants (K_d) for enzyme-DNA complexes were determined by fitting the anisotropy data to binding models using DynaFit 4 (58,59); a representative DynaFit script is provided in the Supplementary Information. We previously described the rationale and benefits of using Dynafit for fitting anisotropy data in this system (TDG binding DNA) (48). The reported parameters were derived from global fitting of at least three independent experiments. The fitted parameters include the dissociation constant for binding of one enzyme subunit to one specific site of the DNA (K_d) and binding of a second TDG subunit to a nonspecific site (K_d2) at higher enzyme concentrations. Fitted parameters also include the anisotropy values for free DNA (r_D) and for 1:1 and 2:1 complexes with enzyme (r_ED, r_EED). As described in the figure legends, selected parameters were in some cases constrained to a fixed value if needed for proper fitting of the other parameters.

RESULTS

Effect of sumoylation on the rate of base excision by TDG

As noted above, previous studies of how sumoylation impacts the activity of TDG were limited to multiple turnover experiments, which give little insight into the effects on discrete elements of the enzymatic reaction. Moreover, the effect of sumoylation on TDG activity has been investigated for G·T and G·U mispairs but not for G·fC and G·caC substrates. Additionally, it is unknown whether conjugation by different SUMO isoforms might exert differential effects on the glycosylase (base-excision) activity of TDG and whether this might be substrate dependent. Unlike previous studies, we used single turnover kinetics under saturating enzyme conditions to obtain rate constants that approximate the maximal rate of base excision (k_obs ≈ k_max) and are not influenced by enzyme-substrate association or steps after base excision (Figure 2). The experiments were performed using unmodified TDG or TDG modified uniformly by SUMO-1 or SUMO-2 (Figure 3, Table 1).

Figure 3.

Single turnover kinetics experiments (37°C) give the maximal glycosylase (base excision) activity of TDG∼SUMO-1 (squares) and TDG∼SUMO-2 (triangles). (A) Activity of sumoylated TDG (5 uM) and G·T substrate (1 uM); fitting to Equation (1) gives k_max = 0.020 ± 0.001 min⁻¹ for TDG∼SUMO-1 and k_max = 0.019 ± 0.001 min ⁻¹ for TDG∼SUMO-2. (B) Activity of sumoylated TDG (1.5 uM) and G·fC substrate (0.5 uM); k_max = 0.079 ± 0.007 min⁻¹ for TDG∼SUMO-1 and k_max= 0.078 ± 0.008 min⁻¹ for TDG∼SUMO-2. (C) Sumoylated TDG (1.5 uM) acting on a G·caC substrate (0.5 uM); k_max= 0.069 ± 0.002 min⁻¹ for TDG∼SUMO-1 and k_max = 0.074 ± 0.003 min⁻¹ for TDG∼SUMO-2.

Table 1. Glycosylase activity for unmodified and sumoylated TDG.

Enzyme	Substrate	k max (min−1)	Fold change in kmax	t 1/2 (min)
TDG	G·T	0.89 ± 0.03	-	0.78
TDG∼SUMO-1	G·T	0.020 ± 0.001	1/45	35
TDG∼SUMO-2	G·T	0.019 ± 0.001	1/47	36
TDG	G·fC	1.69 ± 0.05	-	0.41
TDG∼SUMO-1	G·fC	0.079 ± 0.007	1/21	8.7
TDG∼SUMO-2	G·fC	0.078 ± 0.008	1/22	8.8
TDG	G·caC	0.50 ± 0.02	-	1.4
TDG∼SUMO-1	G·caC	0.069 ± 0.002	1/7.2	10
TDG∼SUMO-2	G·caC	0.074 ± 0.003	1/6.8	9.3

Fold change k_max gives the ratio of k_max values for unmodified and SUMO-modified TDG (k_max^TDG∼SUMO / k_max^TDG).

The largest effect of sumoylation on TDG base-excision activity is observed for G·T mispairs, where k_max is reduced by 45-fold and 47-fold upon modification by SUMO-1 or SUMO-2, respectively (Figure 3A). Turning to the epigenetic substrates, sumoylation also greatly impairs TDG glycosylase activity for G·fC pairs, where k_max is reduced 21- and 22-fold upon modification by SUMO-1 or SUMO-2, respectively (Figure 3B). Thus, sumoylation increases the reaction half-life from 25 s to 9 min. Sumoylation also exerts a large effect on TDG excision of caC, where modification with either SUMO-1 or SUMO-2 causes a 7-fold reduction in k_max (Figure 3C). Thus, the half-life for TDG excision of caC increases from 1.4 to 10 min. Notably, we find that sumoylated TDG exhibits nearly equivalent activity (k_max) for fC and caC substrates, while unmodified TDG has three-fold higher activity for fC relative to caC. It is also notable that for each substrate examined, the impact of sumoylation on TDG activity is nearly equivalent for the two different SUMO isoforms.

Sumoylation has little effect on TDG catalytic turnover for fC and caC substrates

The effect of sumoylation on the catalytic turnover (k_cat) of TDG has been investigated for G·U and G·T (18,28,29) but not for G·fC and G·caC substrates. We performed multiple turnover experiments using a saturating (1.0 uM) concentration of G·fC or G·caC substrate and a limiting (0.05 uM) concentration of enzyme (TDG, TDG∼SUMO-1 or TDG∼SUMO-2) (Figure 4, Table 2). Under these conditions, the observed steady-state velocity (v_o) approximates the maximal level (v_o ≈ v_max) and yields the maximal catalytic turnover (k_cat = v_max/[E]). For unmodified TDG, catalytic turnover is vastly lower than the maximal rate of base excision (k_cat << k_max) for G·fC and G·caC substrates (Table 2), indicating that k_cat is severely limited by steps after base excision, which could include product release and/or product inhibition. Previous studies reported similar results (k_cat << k_max) for TDG acting on other substrates (including G·T, G·U) (21,33,35). Modification of TDG by SUMO-1 or SUMO-2 gives a 1.4-fold elevation in k_cat for G·fC but has no impact for G·caC substrates. Previous studies reported that modification by SUMO-1 generated k_cat enhancements of 2-fold and 4-fold for G·T and G·U substrates, respectively (29). As noted above, k_cat reports on multiple steps of the TDG reaction, including nucleotide flipping, base excision, and product release and product inhibition (Figure 2). Observation that sumoylation greatly reduces k_max (Table 1) but has little effect on k_cat seems likely to reflect offsetting effects on different steps of the reaction. For example, the impairment in base excision (k_max) could be offset by faster product release and/or diminished product inhibition, such that k_cat is similar for sumoylated and unmodified TDG.

Figure 4.

Multiple turnover kinetics experiments for unmodified and sumoylated TDG (50 nM) acting on G·fC or G·caC substrates (1000 nM), performed at 37°C. (A) G·fC; linear fitting of data in the steady-state region gives velocities of v₀ = 0.236 ± 0.011 nM·min⁻¹ for TDG, v₀ = 0.340 ± 0.036 nM·min⁻¹ for TDG∼SUMO-1, and v₀ = 0.315 ± 0.022 nM·min⁻¹ for TDG∼SUMO-2. (B) G·caC; linear fitting of data in the steady-state region gives velocities of v₀ = 0.363 ± 0.023 nM·min⁻¹ for TDG, v₀ = 0.356 ± 0.040 nM·min⁻¹ for TDG∼SUMO-1, and v₀ = 0.354 ± 0.037 nM·min⁻¹ for TDG∼SUMO-2.

Table 2. Steady-state catalytic activity for unmodified and sumoylated TDG.

Enzyme	Substrate	v o (nM·min−1)	k cat (min−1)	k max/kcat
TDG	G·fC	0.236 ± 0.011	(4.7 ± 0.3) × 10−3	359
TDG∼SUMO-1	G·fC	0.340 ± 0.036	(6.8 ± 0.7) × 10−3	12
TDG∼SUMO-2	G·fC	0.315 ± 0.022	(6.3 ± 0.5) × 10−3	12
TDG	G·caC	0.363 ± 0.023	(7.3 ± 0.5) × 10−3	68
TDG∼SUMO-1	G·caC	0.356 ± 0.040	(7.1 ± 0.8) × 10−3	10
TDG∼SUMO-2	G·caC	0.354 ± 0.037	(7.1 ± 0.8) × 10−3	10

Effect of sumoylation on TDG substrate binding

While dissociation constants (K_d) were reported for unmodified TDG binding to G·T and G·U mispairs (48,57), binding to G·fC and G·caC has only been studied for the TDG catalytic domain (residues 111–308) (50), which lacks N-terminal residues that can greatly enhance substrate binding (28,43,50,60). Regarding sumoylated TDG, binding has been studied only qualitatively for G·T mispairs and not at all for G·fC or G·caC pairs. Accordingly, we used fluorescence anisotropy experiments to define the K_d for TDG and TDG∼SUMO binding to three different substrates (G·T, G·fC, G·caC) (48,57). The studies were performed using N140A-TDG (TDG^N140A), a variant that binds substrates like wild-type TDG but has greatly reduced catalytic activity due likely to disrupted binding of the water nucleophile (44,50–52,61). Indeed, TDG^N140A has no detectable activity on G·T substrates (over 48 h) (44) and has exceedingly slow activity for G·fC (k_obs = 4 × 10⁻⁵ min⁻¹; t_1/2 = 329 h) (51) and G·caC substrates (k_obs = 6 × 10⁻⁶ min⁻¹; t_1/2 = 1977 h) (Supplementary Figure S2). As such, enzyme-substrate binding experiments (<3 h) can readily be performed in the absence of base excision. Many previous studies show that TDG binds DNA with a stoichiometry of 1:1 or 2:1 (TDG:DNA), where a single TDG subunit binds a specific site of the DNA (e.g. a G·fC pair) and a second subunit binds to a non-specific site as the concentration of TDG increases (48,57,62). However, previous studies show that 1:1 binding is fully sufficient for catalysis and 1:1 complexes are observed in recent crystal structures, even though the DNA is long enough to accommodate two TDG subunits (48,51,53,63). Thus, our focus is on the affinity of TDG for binding the target site (G·T, G·fC, G·caC), that is, the K_d for 1:1 binding.

We first consider binding of unmodified TDG^N140A to a G·T^F mispair, where T^F (2′-fluoroarabino-dT) is a dT analogue that is not cleaved by TDG and was used in our previous studies for unmodified TDG (48). We find that unmodified TDG^N140A binds to a G·T^F site with K_d = 16 ± 5 nM (Figure 5, Table 3). Notably, this is equivalent to the value determined previously for wild-type TDG binding to the same G·T^F DNA (K_d = 18 ± 3 nM) (48), indicating that the N140A mutation does not substantially alter substrate binding, as previously reported (44). Regarding the impact of sumoylation, TDG affinity for G·T mispairs is weakened by 6- and 7-fold upon modification by SUMO-1 or SUMO-2, respectively (Figure 5, Supplementary Figure S3). While the impact of sumoylation on G·T binding is substantial, the impact on glycosylase activity (k_max) is much larger (≥45-fold; Table 1). We next consider the affinity of TDG^N140A and its sumoylated derivatives for binding to DNA containing a single G·fC or G·caC pair. TDG^N140A binds with high affinity to DNA containing a G·fC site (K_d = 4.0 ± 0.8 nM), and this binding is weakened by 8- or 9-fold upon modification of TDG by SUMO-1 or SUMO-2, respectively (Figure 6, Supplementary Figure S4). Unmodified TDG^N140A binds with very high affinity to DNA containing a G·caC pair (K_d = 2.3 ± 0.6 nM), and sumoylation by either SUMO-1 or SUMO-2 weakens this binding by ∼6-fold (Figure 7, Supplementary Figure S5).

Figure 5.

Equilibrium binding of unmodified and sumoylated TDG^N140A to G·T^F DNA (10 nM) monitored by fluorescence anisotropy. (A) Data for TDG^N140A binding to G·T^F DNA, fitted to a two-site model (using Dynafit), gives K_d = 16 ± 5 nM and K_d2 = 1800 ± 2000 nM, and anisotropy values of r_D = 0.184 ± 0.002, r_ED = 0.253 ± 0.007, r_EED = 0.292 ± 0.014. (B) Binding of TDG^N140A∼SUMO-1 to G·T^F DNA yields K_d = 95 ± 21 nM, K_d2 = 186 ± 68 nM, r_D = 0.183 ± 0.002, r_ED = 0.249 (fixed), and r_EED = 0.281 ± 0.003. (C) Binding of TDG^N140A∼SUMO-2 to G·T^F DNA, fitted to a two-site model, yields K_d = 118 ± 22 nM, K_d2 = 135 ± 38 nM, r_D = 0.190 ± 0.001, r_D = 0.255 (fixed), and r_EED = 0.285 ± 0.002. Fitting of the TDG∼SUMO-1 and TDG∼SUMO-2 data was performed using a fixed value for rED (anisotropy of 1:1 complex); fitting this parameter yielded unreasonably high values. The value used was determined using the relationship rED = rD + Δr_1:1^ave, where rD was fitted as noted above (B, C) and Δr_1:1^ave is the average anisotropy change associated with 1:1 binding (Δr_1:1 = rED – rD) observed for TDG binding to three substrates (G·T, G·fC and G·caC).

Table 3. Dissociation constants for enzyme-substrate complexes.

Enzyme	Substrate	K d (nM)	Fold change in Kd
TDG	G·TF	16 ± 5	-
TDG∼SUMO-1	G·TF	95 ± 21	5.9
TDG∼SUMO-2	G·TF	118 ± 22	7.4
TDG	G·fC	4.0 ± 0.8	-
TDG∼SUMO-1	G·fC	30 ± 4	7.5
TDG∼SUMO-2	G·fC	36 ± 3	9.0
TDG	G·caC	2.3 ± 0.6	-
TDG∼SUMO-1	G·caC	14 ± 2	6.1
TDG∼SUMO-2	G·caC	13 ± 2	5.7

All of the equilibrium binding studies were performed using the N140A variant of TDG.

Figure 6.

Equilibrium binding of TDG^N140A, unmodified and sumoylated, to G·fC DNA (1 nM) monitored by fluorescence anisotropy. (A) Data for TDG^N140A binding to G·fC DNA, fitted to a two-site model gives K_d = 4.0 ± 0.8 nM and K_d2 = 1190 ± 370 nM, and anisotropy values of r_D = 0.185 ± 0.002, r_ED = 0.248 ± 0.003, r_EED = 0.293 ± 0.003. (B) Binding of TDG^N140A∼SUMO-1 to G·fC DNA yields K_d = 30 ± 4 nM, K_d2 = 560 ± 130 nM, r_D = 0.186 ± 0.002, r_ED = 0.252 (fixed), and r_EED = 0.283 (fixed). (C) Binding of TDG^N140A∼SUMO-2 to G·fC DNA, fitted to a two-site model, yields K_d = 36 ± 3 nM, K_d2 = 2370 ± 490 nM, r_D = 0.187 ± 0.002, r_ED = 0.253 (fixed) and r_EED = 0.284 (fixed). Fitting of TDG∼SUMO-1 and -2 data was performed using a fixed value for rED and rEED (anisotropy of 1:1 and 2:1 complexes, respectively), which were otherwise poorly fitted. The values used for rED were determined using the relationship rED = rD + Δr_1:1^ave, as described above (Figure 5). The rEED value was determined using the relationship rEED = rD + Δr_2:1^ave, where Δr_2:1^ave is the average total anisotropy change associated with forming 2:1 complex TDG∼SUMO-1 and -2 binding to G·T DNA.

Figure 7.

Equilibrium binding of TDG^N140A, unmodified and sumoylated, to G·caC DNA (1 nM) monitored by fluorescence anisotropy. (A) Data for TDG^N140A binding to G·caC DNA, fitted to a two-site model (using Dynafit), gives K_d = 2.3 ± 0.6 nM, K_d2 = 820 ± 500 nM, r_D= 0.184 ± 0.0033, r_ED = 0.251 ± 0.004, and r_EED = 0.284 ± 0.004. (B) Binding of TDG^N140A∼SUMO-1 to G·caC DNA fitted to a two-site model (black line) gives K_d = 14 ± 2 nM, K_d2 = 24000 ± 16000 nM, r_D= 0.189 ± 0.002, r_ED = 0.255 (fixed), and r_EED = 0.285 (fixed). Fitting to a one site model (red line) yields essentially the same affinity, K_d = 16 ± 3 nM, with r_D = 0.190 ± 0.002, and r_ED = 0.257 ± 0.002. (C) Binding of TDG^N140A∼SUMO-2 to G·caC DNA fitted to a two-site model (black line) gives K_d = 13 ± 2 nM, K_d2 = 5000 ± 2000 nM, r_D = 0.185 ± 0.003, r_ED = 0.251 (fixed), and r_EED = 0.281 (fixed). Fitting to a one site model (red line) yields K_d = 20 ± 3 nM, with r_D = 0.187 ± 0.002, and r_ED = 0.260 ± 0.002. Fitting for TDG∼SUMO-1 and -2 employed fixed values for rED and rEED, using the same approach described for fitting binding of these enzymes to G·fC DNA (Figure 6).

DISCUSSION

Unmodified TDG binds tightly to G·fC and G·caC pairs in DNA

While our studies focus largely on the effects of sumoylation on TDG activity, we also report that unmodified TDG possess high affinity for G·fC and G·caC pairs (in a CpG context), with K_d of 4.0 nM and 2.3 nM, respectively (Table 3). By comparison, TDG has greater affinity for G·U mispairs (K_d = 0.6 nM) but lower affinity for G·T mispairs (K_d = 16 nM), an unmodified CpG site (K_d = 63 nM), and nonspecific DNA (K_d = ∼300 nM) (48). Thus, TDG has high specificity for G·fC and G·caC pairs, binding these sites with ΔΔG_bind of 2.5 kcal/mol and 2.9 kcal/mol, respectively, compared to nonspecific DNA (or ΔΔG_bind of 1.6 kcal/mol and 2.0 kcal/mol relative to an unmodified CpG site). Because fC and caC impart only small, localized changes to DNA structure (64,65), the specificity of TDG for these sites is likely explained largely by specific interactions that it forms with the formyl and carboxyl groups, as observed in crystal structures (50,51). Although the affinity of TDG for G·hmC pairs has not been reported, evidence that TDG does not form specific interactions with the hydroxyl of hmC is provided by findings that the TDG catalytic domain (residues 111–308; TDG^111–308) lacks detectable affinity for G·hmC (or G·mC) pairs while it binds specifically to G·fC (K_d 130 nM) and G·caC (K_d 70 nM) pairs (50). Importantly, our results reveal that the affinity of TDG for G·fC and G·caC pairs is much tighter than that reported in two previous studies. One study reported that TDG binds G·fC and G·caC with K_d values that are weaker by 11- and 41-fold, respectively, compared to those reported here (40). Another study, which used the catalytic domain (TDG^111–308) (50), reported K_d values for G·fC and G·caC that are 33- and 30-fold weaker than those reported here for full-length TDG. This latter discrepancy could indicate that some portion of the disordered N-terminal region, which is absent in TDG^111–308, enhances TDG binding to G·fC and G·caC pairs, as demonstrated for G·T mispairs (28,43).

Impact of TDG sumoylation on substrate binding and glycosylase activity

Our findings reveal the quantitative effect of TDG sumoylation on two key catalytic parameters, substrate binding (K_d) and the maximal rate of glycosylase activity (k_max). These are the first studies to define the effect of sumoylation on K_d and k_max for a G·T substrate, and the first studies to investigate the impact of sumoylation on TDG activity for G·fC and G·caC substrates. Moreover, this is the first investigation of whether modification by different SUMO isoforms causes differential effects on TDG substrate binding and catalysis. Remarkably, the observed parameters (k_max, k_cat, K_d) varied by less than 10% for modification of TDG by SUMO-1 versus SUMO-2. This is significant, given that SUMO-1 and SUMO-2 share only 48% amino acid sequence identity, but perhaps not surprising given the high similarity of crystal structures of TDG modified with SUMO-1 or SUMO-3 (RMSD of 1.09 Å over 269 Cα atoms) (25,26). Notably, SUMO-2 and SUMO-3 are nearly identical (referred to as SUMO-2/3), thus the effects of TDG modification by SUMO-2 reported here are likely indicative of those that would be observed for SUMO-3.

Although sumoylation weakens TDG substrate binding by 6- to 9-fold, the modified enzyme retains substantial affinity for G·fC (K_d ≤ 36 nM) and G·caC (K_d ≤ 14 nM) pairs (Table 3). Indeed, the affinity of sumoylated TDG for G·caC pairs is greater than that of unmodified TDG for G·T substrates. Our studies also reveal that TDG∼SUMO exhibits weakened though readily measurable binding to G·T mispairs (K_d ≤ 118 nM). Similarly, we showed previously that sumoylated TDG retains substantial albeit weakened affinity for abasic sites in DNA (20). Thus, our results provide an important advance because in previous studies, substrate and product binding had not been observed for sumoylated TDG (18,25,26).

Importantly, our studies also reveal that sumoylation imparts large adverse effects on the base excision activity of TDG, with reductions in k_max of 46-, 22-, and 7-fold for G·T, G·fC, and G·caC substrates, respectively. Together, these results directly challenge the current paradigm that the predominant effect of sumoylation is on the binding affinity of TDG for DNA substrates and product, with minimal effect on catalytic activity (18,28). On the contrary, our results show that the adverse effects of sumoylation are greater for catalytic activity relative to those for substrate binding.

Our discovery that sumoylation dramatically impairs k_max reveals a corresponding decrease in the upper limit of catalytic turnover (k_cat) for sumoylated versus unmodified TDG, even if k_cat is enhanced by other factors (e.g., APE1), because k_cat cannot exceed k_max (Figure 2). Thus, even if sumoylation potentiates the stimulatory effect of APE1 on TDG catalytic turnover, as reported for G·U substrates (18), the upper limit of APE1-stimulated k_cat is far lower for sumoylated versus unmodified TDG. We note that k_cat for sumoylated TDG is probably not particularly relevant anyway, because if sumoylation does serve to enhance TDG catalytic turnover, the mechanism seems likely to involve modification of product-bound TDG, with unmodified TDG handling the base excision step (18).

How might sumoylation impact substrate binding and glycosylase activity of TDG?

It is of interest to consider potential mechanisms by which sumoylation impacts substrate binding and base excision by TDG. Previous studies indicate that the adverse effects of sumoylation on DNA binding and glycosylase activity are likely mediated by non-covalent binding of the tethered SUMO domain to the SIM of TDG (Figures 1B and 8) (25,26). Crystal structures indicate that SUMO–SIM interactions lead to the formation of a β-strand and an α-helix in a C-terminal region of TDG (residues 307–331) (25,26), which is likely disordered in the unmodified enzyme (19,25). The nascent β-strand of TDG contacts a β-strand of SUMO to form an intermolecular β-sheet, while the α-helix forms few interactions with SUMO or other regions of TDG and protrudes away from the protein surface. The affinity of TDG for binding free SUMO proteins was shown to be reduced by mutation of TDG residues that mediate SUMO–SIM interactions, including R281, E310, Y313, and F315 (Figure 8), and these mutations also reverse (at least partially) the adverse effect of sumoylation on TDG binding to abasic DNA (25,26). Together, these previous findings suggest that non-covalent SUMO–SIM interactions mediate the adverse effects of sumoylation on TDG activity, but what is the molecular mechanism?

Figure 8.

Closeup view of the SUMO–SIM interface and the catalytic ‘insertion’ loop, using the same crystal structures and coloring scheme as described above for Figure 1B. Some of the residues that mediate SUMO–SIM binding, and residues of the catalytic ‘insertion’ loop are shown. Interactions between these residues could potentially link SUMO–SIM binding to impairment of productive nucleotide flipping. Because no structure is available for sumoylated TDG bound to DNA, the DNA shown (orange) was positioned as described for Figure 1B.

The current paradigm involves a model whereby the SUMO-induced α-helix of TDG perturbs DNA binding by clashing with the phosphodiester backbone (Figures 1B and 8) (25,26), which may seem consistent with previous reports that sumoylation fully abrogates detectable binding of TDG to DNA substrates and abasic DNA product (18,25,26,28). However, our studies reveal that the effects of sumoylation on TDG substrate binding are relatively modest (ΔΔG_bind ≤ 1.3 kcal/mol), and sumoylated TDG retains substantial binding affinity for G·fC and G·caC pairs in DNA. Moreover, observation that sumoylated TDG retains some base excision activity requires that it can bind productively to DNA substrates. These results suggest that the SUMO–SIM-induced α-helix might be destabilized or displaced upon binding to DNA substrates or abasic product (20). Otherwise, binding would require substantial deformation of the DNA backbone (to avoid steric clash with the α-helix), which seems unlikely. Addressing the question of how the putative steric clash is resolved will require a structure of sumoylated TDG bound to DNA.

Our findings, together with structural observations, suggest an alternative mechanism for the effects of sumoylation on TDG substrate binding and base excision. Because our single turnover experiments were performed under saturating enzyme conditions, the rate constants (k_max) report on steps of the TDG reaction that occur after formation of the initial enzyme-substrate complex and up to formation of enzyme-bound product (Figure 2), which include the enzyme-substrate conformational changes associated with nucleotide flipping (E·S to E′·S′) and the chemical step(s) (E′·S′ to E′·B·P′). One plausible explanation for the dramatic SUMO-induced reductions in k_max is that sumoylation alters nucleotide flipping and thereby reduces the fraction of substrate-bound enzyme that forms a productive conformation (E′·S′). Notably, an adverse effect on nucleotide flipping could also account, at least in part, for the observed sumoylation-induced weakening of substrate binding (K_d) (Figure 2). The possibility that sumoylation perturbs nucleotide flipping is supported by structural observations. An important catalytic loop of TDG, termed the ‘insertion loop’, contains many conserved residues including Arg275, which penetrates the DNA helix to fill the void created by nucleotide flipping (Figure 8) (43,51,53). The Arg ‘plug’ is highly conserved in TDG and other G·T mismatch glycosylases and likely helps to stabilize nucleotide flipping (44,66). Residues of the insertion loop (Phe279, Arg281) interact with residues in the SIM or with SUMO, linking these regions of sumoylated TDG and suggesting a mechanism by which sumoylation could alter nucleotide flipping. Additional structural and biochemical studies will be needed to test this and other potential mechanisms for SUMO-induced reductions in substrate binding and catalysis.

What are the functions of TDG sumoylation?

It is of interest to consider the potential implications of our results regarding the role(s) of TDG sumoylation. The current paradigm is that sumoylation of TDG regulates product release and enhances its catalytic turnover (k_cat) (18). TDG binds very tightly to its reaction product, abasic DNA, which severely impedes catalytic turnover in vitro (32–35). The current model holds that sumoylation occurs selectively for product-bound TDG, triggering product release, and that SUMO deconjugation reactivates TDG for processing additional substrates, such that each catalytic cycle of TDG requires sumoylation and desumoylation (18). While this model seems widely accepted (25,31,36–38) it is not directly supported by experimental evidence and has been challenged by recent studies. For example, in vitro studies showed that E2-mediated sumoylation of TDG is efficient but not specific for product-bound TDG (20). Moreover, sumoylation of TDG was shown to be dispensable for efficient removal of TET-generated caC from genomic DNA in human cells (21). In addition, catalytic turnover of TDG is strongly enhanced by follow-on BER enzymes, including AP endonuclease 1 (APE1), which process AP sites generated by DNA glycosylases. APE1 can enhance TDG catalytic turnover to a level that approaches the theoretical maximum (such that k_cat approaches k_max) (21,39). The bifunctional enzymes NEIL1 and NEIL2 also stimulate TDG catalytic turnover in vitro, and they contribute to the efficient removal of fC and caC by TDG in cells (40). Additionally, the XPC repair complex (XPC, RAD23B, possibly CENT2) interacts with and stimulates catalytic turnover of TDG in vitro and in mammalian cells (67,68). Taken together, these observations indicate that other potential roles for TDG sumoylation should be considered.

Previous studies suggest sumoylation of TDG could serve to alter its subcellular localization and modulate its interactions with other proteins (15,27), which would be consistent with the roles of sumoylation that have been observed for most other proteins that undergo SUMO modification. Sumoylation might also help to suppress potentially deleterious activity of TDG in S phase of the cell cycle. TDG is depleted in S phase via ubiquitination and proteasome degradation (22,23,69), for reasons that remain unclear. One possibility is that degradation precludes TDG processing of G·T (or G·U) mismatches that arise from replication errors, which could potentially yield tight TDG-product complexes and disrupt DNA replication. Sumoylation might provide a safeguard in S phase, to suppress the glycosylase activity of TDG molecules that escape degradation. One study reported that the low residual level of TDG detected in S phase was predominantly sumoylated, suggesting that ubiquitin-mediated degradation is less efficient for sumoylated versus unmodified TDG (22).

Our findings raise the possibility of an intriguing new role for TDG sumoylation - it might enable TDG to function, at least transiently, as a reader rather than just an eraser of fC and caC. This idea stems from our findings that sumoylation dramatically reduces the glycosylase activity of TDG such that it excises fC and caC much less efficiently (t_1/2 >9 min, 37°C). Meanwhile, sumoylated TDG retains relatively high affinity for these substrates, with K_d ≤36 nM for G·fC and K_d ≤14 nM for G·caC. Indeed, these affinities are similar to that of unmodified TDG for binding G·T mispairs in DNA. Thus, sumoylation yields a form of TDG that binds specifically to fC and caC but is far less likely excise them compared to the unmodified enzyme. TDG interacts with a broad range of proteins including transcription factors such as p300 (14), retinoic acid receptors (RAR, RXR) (70), and the estrogen receptor α (71), among others, and epigenetic regulators such as DNA methyltransferases (72) and the SIRT1 deacetylase (17). Many of these proteins are sumoylated and/or contain a SIM (20). Sumoylation could potentially afford TDG with an enhanced ability - and more time - to recruit other proteins to DNA sites containing fC and caC before it excises these bases. Notably, the SUMO domain of sumoylated TDG might participate in protein recruitment, depending on its accessibility for DNA-bound TDG∼SUMO and the presence of a SUMO-interacting motif in other proteins. Previous studies indicate that 20–50% of TDG is sumoylated in human cells, as determined from western blots of nuclear extract (15,18,27). However, the fraction of TDG that is sumoylated in vivo will likely depend on many factors, including the abundance and activity of enzymes that mediate SUMO conjugation (E1, E2, possibly E3) and deconjugation (SENPs) (21). Notably, TDG can be rapidly modified by the SUMO-conjugating enzyme (E2∼SUMO thioester), with maximal rate constant of k_max = 1.6 min⁻¹ (t_1/2 of 0.4 min) (20). It has been shown that SENP1 deconjugates modified TDG, but the rate constant has not been reported (21).

It is of interest to consider how the affinity of TDG∼SUMO for fC and caC compares to that of other potential readers of these bases. Mass spectrometry studies uncovered many putative readers of fC and caC (73,74) but only a few have been biochemically characterized. N-methylpurine DNA glycosylase (MPG, aka AAG, ANPG) binds (but does not excise) fC in DNA with a reported affinity of 13 nM (using an ELISA assay) (74). If accurate, this affinity is about twofold tighter than that of TDG∼SUMO for fC. MAX, a binding partner of the MYC transcription factor, binds enhancer-box (E-box) elements and has some specificity for caC; MAX binds with a K_d of ∼30 nM when the CpG site in these elements contains caC or unmodified C (75). Another putative caC reader is the CXXC domain of TET3, which binds unmodified CpG sites, and, with 3-fold higher affinity, CpG sites with a caC modification (76). Thus, our results reveal that sumoylated TDG binds fC and caC with similar affinity to previously-characterized readers of these bases in DNA.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

National Institutes of Health [R01-GM72711 to A.C.D.]; Support for procuring the imaging system (GE Typhoon FLA 9500) used in these studies was provided by the National Institutes of Health Grant [S10-OD011969]. Funding for open access charge: National Institutes of Health.

Conflict of interest statement. None declared.