An autoinhibitory role for the GRF zinc finger domain of DNA glycosylase NEIL3

The NEIL3 DNA glycosylase maintains genome integrity during replication by excising oxidized bases from single-stranded DNA (ssDNA) and unhooking interstrand cross-links (ICLs) at fork structures. In addition to its N-terminal catalytic glycosylase domain, NEIL3 contains two tandem C-terminal GRF-type zinc fingers that are absent in the other NEIL paralogs. ssDNA binding by the GRF – ZF motifs helps recruit NEIL3 to replication forks converged at an ICL, but the nature of DNA binding and the effect of the GRF – ZF domain on catalysis of base excision and ICL unhooking is unknown. Here, we show that the tandem GRF – ZFs of NEIL3 provide affinity and specificity for DNA that is greater than each individual motif alone. The crystal structure of the GRF domain shows that the tandem ZF motifs adopt a flexible head-to-tail configuration well-suited for binding to multiple ssDNA conformations. Functionally, we es-tablish that the NEIL3 GRF domain inhibits glycosylase activity against monoadducts and ICLs. This autoinhibitory activity contrasts GRF – ZF domains of other DNA-processing enzymes,

NEIL3 plays a critical role in protecting cells during DNA replication. NEIL3 is expressed in early S phase, and the protein is found primarily in proliferating cells and co-localizes with Replication Protein A (RPA) (13)(14)(15)(16). NEIL3 is the only DNA glycosylase of the Nei/Fpg superfamily to have a strong prefer-ence for lesions in ssDNA and at branched structures expected to be found at replication forks (4,7,8,11,17). In addition, NEIL3 repairs damaged telomeric DNA during S phase and excises thymine glycol, spiroiminodihydantoin, and guanidinohydantoin lesions from G-quadruplexes (18)(19)(20). NEIL3 is highly expressed in various human cancer cells and in primary malignant melanomas associated with metastasis (21,22), exhibits a high frequency of loss of heterozygosity in hepatocellular carcinomas (23), and is frequently mutated in colorectal and breast cancer (24,25). Loss of NEIL3 was shown to increase deleterious strand breaks during DNA replication and to enhance sensitivity to Ataxia telangiectasia and Rad3-related protein (ATR) and poly(ADP-ribose) polymerase inhibitors (26).
Consistent with its role in DNA replication, NEIL3 participates in the repair of interstrand DNA cross-links (ICLs) (27)(28)(29). ICLs are highly toxic lesions that covalently link opposing DNA strands and consequently inhibit replication (30). ICL repair is initiated by unhooking the tethered strands, either by dual incisions on one strand by endonucleases associated with Fanconi anemia and nucleotide excision repair, or alternatively by cleavage of the cross-linked nucleotide by a specialized DNA glycosylase (5,31). NEIL3 unhooks ICLs derived from the natural product psoralen in both Xenopus cell extracts and in human cells in a replication-dependent manner (27)(28)(29). Moreover, NEIL3 unhooks ICLs derived from highly abundant endogenous AP lesions (17,27). AP-ICLs are formed between a ring-opened AP aldehyde on one strand and the exocyclic N 6 amine of adenine on the opposite strand (32)(33)(34)(35). NEIL3 cleaves the nonnative N-glycosidic bond from the AP site (27) and preferably acts on AP-ICLs residing on the leading-strand template at a forked DNA structure (17). This substrate specificity is unique to vertebrate NEIL3, because an unrelated bacterial ICL glycosylase does not cleave AP-ICLs but recognizes ICLs independent of DNA structure (36)(37)(38).
NEIL3 is a distinct member of the Nei DNA glycosylases, which contain a helix-two-turn-helix glycosylase domain (GD) (1). The structure of the NEIL3-GD explains the enzyme's preference for ssDNA, because it lacks two of the three residues important for intercalating into the dsDNA during excision of the damaged nucleobase (9). Unlike NEIL1, NEIL2, and the other Nei members, NEIL3 further contains a 323-residue C-terminal extension with three zinc finger (ZF) motifs-an internal RanBP/ NPL4-type (NZF) and two glycine-arginine-phenylalanine (GRF) ZFs (10,12 required for recruitment of NEIL3 to replication forks that have converged at an ICL in Xenopus extracts (28). The NZF does this through direct interaction with ubiquitylated CMG helicase found at convergent forks, whereas the GRFs do so presumably through their interaction with ssDNA (28). GRF-ZFs are ssDNA-binding elements present at the C-terminus of several DNA-processing enzymes (39,40). In several cases these motifs have been shown to enhance enzymatic activity (39,(41)(42)(43). For example, disrupting the GRF-ssDNA interaction in AP endonuclease 2 (APE2) inhibits nuclease activity in vitro (39) and is required for full functional complementation of APE2 synthetic lethality with BRCA-deficient cell lines (44). Despite our understanding of the role of the GRF domain on recruitment of NEIL3 to converged forks at an ICL (28), how the GRF domain cooperates to bind DNA and regulate NEIL3 DNA glycosylase activity is unknown. Here, we report the first crystal structure of the NEIL3 GRF domain and probe the role for GRF DNA binding on NEIL3 ICL unhooking and base excision. We find that ssDNA binding by NEIL3's GRF motifs inhibits NEIL3 enzymatic activity. In addition, the structural flexibility between tandem GRF motifs suggests that this domain can adapt to a variety of ssDNA conformations.

GRF zinc fingers bind ssDNA and fork-like structures
The individual NEIL3 GRF motifs have been shown to bind ssDNA (28), but the DNA binding specificity of the tandem GRF domain (GRF12), has not been investigated. We purified GST fusions (Fig. S1) of the Mus musculus (m)NEIL3 individual GRF1 and GRF2 and tandem GRF12 proteins and used electrophoretic mobility shift assays (EMSAs) to probe DNA binding to 40-mer ssDNA, dsDNA, and fork (splayed arm) DNA structures, the latter of which contained a 20-nt dsDNA region with 20-nt ssDNA arms. Each GRF construct bound ssDNA, but not dsDNA ( Fig. 1, A-C), consistent with previous results from Xenopus laevis (x)NEIL3 (28). We also observed that the individual and tandem GRF motifs bound fork DNA with an affinity similar to that of ssDNA. Notably, the tandem GRF12 protein displayed 5-10-fold higher affinity for ssDNA and fork DNA (apparent K d = 10 nM) compared with the isolated GRF1 or GRF2 constructs (K d,app = 50-100 nM). In addition, the GRF12 domain showed only one major protein-DNA complex band in EMSAs (Fig. 1A), whereas the individual GRF1 and GRF2 motifs exhibited multiple protein-DNA bands indicative of nonspecific binding at higher protein concentrations (Fig. 1, B and C). Despite the appearance of the plotted data shown in Fig. 1 (B and C), there is no statistically significant difference between the affinities of GRF1 and GRF2 for either ssDNA or forked DNA. In control experiments, neither purified recombinant GST (Fig. 1D) nor a NEIL3 GST-NZF domain fusion (Fig.  1E) bound DNA with appreciable affinity. These results indicate that the NEIL3 GRF12 domain binds ssDNA with no apparent specificity for a fork structure and that the tandem GRF motifs bind with higher affinity and specificity than either GRF motif alone.

Structural basis for ssDNA binding by the NEIL3 GRF domain
To shed light on the basis for cooperative DNA binding by the GRF12 domains, we crystallized and determined a 2.6 Å crystal structure of human (h)NEIL3 GRF12 domain using single-wavelength anomalous dispersion at the zinc edge ( Fig. 2A and Fig. S2). The final model, which contains two GRF12 protomers per asymmetric unit, was refined to a crystallographic residual of 22.7% (R work ) and 26.6% (R free ) with reasonable statistics ( Table 1). The structure resembles a butterfly, in which two side-by-side GRF motifs are joined by a three-residue flexible linker and oriented in a head-to-tail fashion, with the zincbinding loops on opposite corners of the molecule ( Fig. 2A). Each GRF forms a crescent-shaped structure composed of a three-stranded b-sheet with a CHCC-type zinc finger on one end and a 7-10-residue loop on the other. The concave surfaces of both GRF1 and GRF2 contain a preponderance of basic residues (Fig. 2B), several of which have been shown to bind DNA in xNEIL3 and xAPE2 (28,39). In NEIL3 GRF12, the concave surfaces face outward and form a continuous surface that likely binds DNA (Fig. 2B). The relative orientation between the GRF1 and GRF2 motifs is different in the two protomers in the asymmetric unit, indicating a large degree of flexibility between them that could accommodate different DNA conformations (Fig. 2C).
The structures of GRF1 and GRF2 are virtually identical, with an RMSD of 0.89 Å for all atoms (Fig. 2D). In addition, both NEIL3-GRFs are similar to that of xAPE2, with RMSD values of 0.86 Å (GRF1) and 0.60 Å (GRF2) for all atoms (Fig. 2E). The positions of GRF1/GRF2 residues Arg 517 /Lys 563 , Lys 521 / Lys 567 , and Phe 530 /Phe 576 correspond to the DNA binding residues in xAPE2 (Arg 473 , Lys 477 , and Phe 486 ) (39) (Fig. 2, D-F). We validated the importance of these residues for DNA binding by mutating each of the corresponding residues in mNEIL3-GRF proteins to glutamate and quantifying DNA affinity in solution by fluorescence anisotropy. Individual glutamate substitutions were made in GST-GRF1 or GST-GRF2 fusion proteins. WT GST-GRF1 and GST-GRF2 bound to 40mer ssDNA with K d values of 0.3 6 0.04 and 0.4 6 0.06 mM, respectively, and the GST tag alone showed no binding (Fig.  2G). Glutamate substitution of either mNEIL3 Arg 518 or Lys 522 , which correspond to hNEIL3 Arg 517 and Lys 521 , or mNEIL3 Lys 568 , which corresponds to hNEIL3 Lys 567 , completely abrogated DNA binding in the individual GST-GRF1 and GST-GRF2 proteins (Fig. 2G). The lack of binding by the GRF1 K522E mutant is consistent with that shown previously by the corresponding mutant (K500E) in an xNEIL3 MBP-GRF1 construct (28). We next tested binding of a K522E/K568E double mutant in GRF12 that had the GST tag removed (Fig. 2H). WT GRF12 bound the 40-mer ssDNA with a K d value of 0.2 6 0.01, whereas the double mutant showed at least 100-fold reduced DNA binding affinity. Thus, the continuous concave surface of the GRF12 domain is likely the main DNA-binding site.
A Dali search for structural homologs identified the GRFlike zinc ribbon (ZR) motifs in Escherichia coli topoisomerase I (TopI) in addition to that of APE2 (45,46). NEIL3, APE2, and TopI all contain similar GRF-ZF or GRF-like ZR motifs at their extreme C terminus (Fig. 3A). TopI contains five ZR motifs, four of which bind ssDNA in linear fashion (45) (Fig. 3B). Like NEIL3-GRF12, the TopI-ZR1-5 motifs are highly similar to one another, are joined by flexible linkers, and present a contiguous DNA-binding surface. This similarity suggests that NEIL3 will bind to ssDNA across the two GRF motifs in a similar way. The NEIL3-GRFs are most similar to TopI-ZR1 (RMSD = 1.86 Å for backbone atoms), although all five TopI-ZRs have the same general GRF-like architecture (Fig. 3C). TopI-ZR1 interacts with ssDNA by an arginine that corresponds to Arg 517 / Lys 563 in hNEIL3, as well as several aromatic residues from the b strands, including Phe 616 , which corresponds to Phe 530 / Phe 576 in hNEIL3 (Fig. 3D). TopI-ZR3, -ZR4, and -ZR5 show similar interactions, whereby the arginine side chains contact the deoxyribose-phosphate backbone and a phenylalanine or tyrosine at the center of the concave cleft stacks against the nucleobases (Fig. 3, C and D). Despite the wide variety of relative orientations between consecutive ZRs, the ssDNA binds TopI in a linear fashion with two or three nucleotides per ZR.

GRF1 GRF2
To gain insight into the footprint of ssDNA across NEIL3-GRF12, we used fluorescence anisotropy to monitor binding to ssDNA of various lengths from 10 to 40 nucleotides (Fig. 3E).  (40-mer). A stoichiometric binding experiment with 40-mer ssDNA revealed saturation of binding at a GRF12:DNA ratio of 8:1, which corresponds to five nucleotides per GRF12, consistent with the two or three nucleotides bound to each TopI ZR motif (Fig. 3F).

The GRF domain inhibits NEIL3 glycosylase activity
The GRF-ZF of APE2 and GRF-like ZR domains of TopI and human topoisomerase III have been shown to enhance catalytic activity (39,(41)(42)(43). We therefore investigated whether the tandem GRF motifs have an impact on NEIL3 DNA glycosylase activity. We first tested the effect of the GRF DNA binding mutations on ICL unhooking activity by full-length mNEIL3 under single-turnover conditions. We compared WT protein against a K522E/K568E double mutant (GRF mut ) and a GRF12 deletion mutant (DGRF) (Fig. 4A). We previously found that the mNEIL3-GD is able to unhook AP-ICLs at forks only when the AP site is on the strand corresponding to the leading-strand template (17). Here, we found that the full-length protein has the same specificity. We saw no unhooking of the laggingstrand AP-ICL substrate (Fig. 4B) but observed both unhooking and AP lyase activity when NEIL3 was incubated with the leading-strand AP-ICL substrate (Fig. 4C). Interestingly, both the GRF mut and DGRF mutants showed significantly faster ICL unhooking kinetics relative to the WT protein (Fig. 4D), suggesting an autoinhibition of ICL unhooking activity by the GRF domain.
Because the mNEIL3-GD has base excision activity on its own (8,17), we tested the mechanism by which the GRF-ZF inhibits catalytic activity by adding the purified GRF12 domain in trans to mNEIL3-GD enzymatic reactions containing ssDNA and fork substrates harboring a single DHT residue (Fig. 5A). Compared with reactions that contained no GRF12 protein, the presence of an equimolar amount of GRF12 caused a severe reduction in DHT excision kinetics (Fig. 5B). To test whether this reduction of enzymatic activity was caused by competition of GD and GRF binding to the DNA substrate, we performed the same reaction with the GRF12 K522E/K568E double mutant (GRF12 mut ). The addition of GRF12 mut did not inhibit the reaction (Fig. 5B), indicating that the presumed DNA-binding surface of GRF12 inhibits base excision by the glycosylase domain.
The same preference that exists for AP-ICLs on the leading template strand of forks also exists for DHT monoadducts (17). Unlike with ICL forks, however, mNEIL3-GD retains a low level of activity for monoadducts on the lagging-strand template (17). We therefore tested whether GRF DNA binding would affect this specificity by altering the relative activities for forks containing a monoadduct on the leading versus the lagging strand (Fig. 5C). Similar to the result with ssDNA, the addition of GRF12, but not GRF12 mut , caused the same inhibition of glycosylase activity against both DHT fork substrates, indicating that the GRF domain has no effect on the preference for leading-strand damage at forks (Fig. 5C).

Discussion
GRF-ZFs are ssDNA-binding elements present at the C-terminus of several DNA repair enzymes (39,45). We show here that the GRF domain of NEIL3 is similar in structure and ssDNA-binding properties to the GRF-like motifs of metazoan APE2 and bacterial TopI. However, unlike APE2 and TopI, in which GRF ssDNA binding enhances activity of the catalytic domains (39, 41-43), we find that the NEIL3 GRF-ssDNA    interaction inhibits both base excision and ICL unhooking activity in vitro. Based on the nanomolar ssDNA binding affinity of the tandem GRFs and the requirement for ssDNA in NEIL3 substrates (8,17,27), we hypothesize that the GRF and glycosylase domains compete for the ssDNA substrate. Indeed, when we introduced point mutations that ablate GRF-ssDNA binding, we rescued glycosylase activity. An alternative possibility is that the GRF12 domain interacts in cis with the NEIL3-GD to modulate its activity. Indeed, protein structures that mimic ssDNA binding to APE2 have been observed in the X-ray structure of the APE2 helix-GRF domain region, suggesting that GRF-ssDNA binding could be regulated by GRF protein-protein interactions (39,(41)(42)(43). We propose that the GRF domain endows NEIL3 with a specificity for ICL-stalled replication forks and a variety of other complex environments through its interactions with DNA and other macromolecules. The C-terminal ZF domains of NEIL3 aid in the recruitment of NEIL3 to replication forks that have converged at an ICL (28). The NZF domain interacts with ubiquitinylated CMG helicase, and the GRF domain enhances the interaction, likely by interacting with ssDNA on the lagging-strand template (28). In those studies, disrupting the GRF-ssDNA interaction via mutagenesis resulted in loss of ICL unhooking by NEIL3 in Xenopus egg extracts, but not in vitro, consistent with the GRF-DNA interaction being important for the recruitment of NEIL3 to converged forks. Because we now find that disruption of the GRF-ssDNA interaction increases glycosylase activity on a simple ICL fork substrate in vitro, it is likely that at converged forks in the cell, the GRF domain is bound to ssDNA away from the lesion so it does not interfere with glycosylase domain access. The other interactions between NEIL3 and the replisome, such as NZF and ubiquitylated CMG, would direct the manner in which NEIL3 is positioned. The autoinhibition of NEIL3 activity in vitro also suggests that the GRF domain may regulate catalytic activity prior to recruitment to converged forks, as a means of avoiding spurious strand cleavage on ssDNA present during replication.
We find that two tandem GRF motifs provide enhanced and more specific binding to ssDNA than each individual GRF motif alone. In addition, the different relative configurations of GRF1-GRF2 motifs observed between the two crystallographically unique molecules in the structure suggest an overall flexibility of the tandem GRF12 domain that may promote binding to ssDNA in different contexts or broaden the conformations that can be sampled upon binding. For example, the C-terminus of NEIL3 plays a role in recruitment of the protein to telomeric damage (20), which is a different DNA environment from converged replisomes. This flexibility may also provide opportunities for the GRF to interact with protein partners. Interestingly, GRF2 contains putative ubiquitylation sites that may aid in such protein-protein interactions. NEIL3 interactions with PCNA, FEN1, TRF1, and CMG have been reported (20,28). There are likely others, especially given that NEIL3 has been associated with roles outside of replication, including cell signaling, immunity, pulmonary function, myocardial infarction, and ischemic stroke (1,(47)(48)(49)(50)(51)(52)(53)(54)(55)(56)(57)(58)(59)(60).
Nucleotides encoding mNEIL3-GD (residues 1-282) and containing a C-terminal His 6 tag was expressed from pET30a in E. coli BL21(DE3) Star cells by autoinduction as previously described (12). The cells were lysed in buffer A (20 mM Tris-HCl, pH 8.0, 300 mM NaCl, 10% glycerol, 5 mM bME) containing 20 mM imidazole, 1 mM leupeptin, and 1 mM pepstatin A. Lysate was incubated with Ni-NTA (Thermo Scientific) beads for 30 min at 4°C. Protein was eluted using a 20-500 mM imidazole gradient in Buffer A. Fractions were pooled, diluted to 75 mM NaCl in buffer B (40 mM HEPES-NaOH, pH 7.0, 10% glycerol, and 5 mM bME), and dialyzed at 4°C overnight. Protein was loaded onto a heparin-Sepharose (GE Healthcare) column, washed, and eluted with an NaCl gradient (0-1 M) in Buffer B. The fractions were pooled, diluted to 500 mM NaCl in buffer B, and loaded onto a HiLoad 26/600 Superdex 200 sizeexclusion column (GE Healthcare). The protein was eluted in 20 mM HEPES-NaOH, pH 7.0, 100 mM NaCl, 1 mM TCEP, concentrated, and frozen. Nucleotides encoding mNEIL3 GRF1 (residues 506-549), GRF2 (residues 550-595), tandem GRF12 (residues 506-595), and NZF (residues 319-353) proteins were cloned into pBG101 (Vanderbilt Center for Structural Biology), which produces an N-terminal His 6 -GST fusion protein that can be cleaved by rhinovirus 3C protease. Proteins were expressed in E. coli BL21 RIL cells in LB medium containing 10 mM ZnSO 4 by induction with 250 mM IPTG overnight at 16°C. The cells were lysed in buffer containing 50 mM Tris-HCl (pH 8.0 for GRF1, pH 7.0 for GRF2, or pH 7.5 for GRF12 and NZF), 500 mM NaCl, 15% glycerol, 0.5 mM TCEP, 0.02% NP-40, 1 mM leupeptin, and 1 mM pepstatin A. Lysate was incubated with Ni-NTA (Thermo Scientific) beads for 30 min at 4°C. The beads were washed with 20 mM imidazole, and protein was eluted with 500 mM imidazole in buffer C (50 mM Tris-HCl, 500 mM NaCl, 15% glycerol, 0.5 mM TCEP) supplemented with 0.02% NP-40. Imidazole was diluted to less than 200 mM, and protein was incubated overnight at 4°C with GSH-Sepharose resin (Thermo Scientific) in buffer C with 0.02% NP-40. GSH beads were washed with buffer C and 0.02% NP-40, and protein was eluted with 10 mM reduced GSH in buffer C with 0.01% NP-40. The protein fractions were pooled and concentrated. At this stage the purified His 6 -GST fusions were either stored for use in EMSAs or dialyzed into buffer C and cleaved by 3C protease. The His 6 -GST tag was removed with Ni-NTA in buffer C containing 15-20 mM imidazole. GRF protein was concentrated and stored in buffer C. GRF1-R518E, GRF1-K522E, GRF2-K568E, and GRF12-K522E-K568E (GRF mut ) point mutants were generated using a QuikChange (Agilent) mutagenesis kit and purified the same as the WT proteins.
Human NEIL3 GRF12 for crystallization (residues 501-605) was cloned into pMCSG9 by ligation-independent cloning to produce a His 6 -MBP-TEV-GRF construct. For expression, plasmid was transformed into BL21AI cells and grown in terrific broth to an A 600 of 0.8. The cells were induced with 20% Larabinose, 50 mM ZnSO 4 , and 100 mM IPTG and grown overnight at 16°C. For purification, the cells were resuspended in 50 mM Tris, pH 8.0, 500 mM NaCl, 1 mM TCEP, 0.1 g of lysozyme, and protease inhibitor tablet (Roche). The cells were lysed and centrifuged. Lysate was added to amylose resin, flow-through was collected, and resin was washed with the above buffer followed by 50 mM Tris, pH 8.0, 1 M NaCl, and 1 mM TCEP. Salt was dropped to 150 mM NaCl, and protein was eluted with 50 mM Tris, pH 8.0, 150 mM NaCl, 1 mM TCEP, and 10 mM maltose. Elutions were collected, incubated with TEV protease, and subjected to heparin-Sepharose chromatography with a 0.05-1.0 M NaCl gradient in 20 mM Tris, pH 8, and 1 mM TCEP. Fractions were pooled and run over a HiLoad 16/600 Superdex 75 (GE Healthcare) size-exclusion column in 10 mM Tris pH 8, 100 mM NaCl, and 1 mM TCEP, and protein was concentrated to 10 mg/ml.

X-ray crystallography
Crystals of hNEIL3 GRF12 were obtained by sitting-drop vapor diffusion at 20°C by mixing the protein solution with 0.1 M Tris, pH 8.5, and 25% PEG 6000. The crystals were cryoprotected in 20% ethylene glycol. X-ray diffraction data from a single crystal were collected at 105 K on Beamline 22-ID of the Advanced Photon Source and processed with HKL2000 (61). Phasing and refinement was carried out using the PHENIX suite of programs (62). The structure was determined by SAD phasing from 2.8 Å anomalous data collected at the zinc edge (1.27059 Å) using AutoSol and Phaser. An initial model was built using side chain-truncated APE2 GRF residues 461-508 from PDB 5U6Z as a guide and the Zn-SAD electron density, followed by manual building of the remainder of the model in Coot (63). The model was refined against 2.6 Å native data collected at 1.0000 Å wavelength. The final model was validated with MolProbity (64) and contained no residues in disallowed regions of the Ramachandran plot. Refinement and validation statistics are shown in Table 1. Structures were analyzed and figures made using PyMOL (Schrödinger). All software was curated by SBGrid (65).

Fluorescence anisotropy
DNA binding of WT and mutant GRF constructs were monitored by fluorescence anisotropy. For K d determination, proteins at varying concentration were incubated with 25 nM FAM-labeled ssDNA (Table S1) in 20 mM HEPES, pH 7.0, 100 mM NaCl, 5 mM MgCl 2 , 3% glycerol, and 0.5 mM TCEP at 4°C in the dark for 30 min. To determine stoichiometry of binding, 50 nM FAM-labeled and 5 mM unlabeled 40-mer ssDNA was used so that the total DNA concentration was over 20-fold excess of the K d . Fluorescence anisotropy data at excitation and emission wavelengths of 485 and 528 nm were collected at room temperature in 96-well plates using a BioTek Synergy H1 plate reader. The data were fit to a two-state binding model using GraphPad Prism 8.

Glycosylase activity assays
DNA substrates containing an AP-ICL or DHT were prepared as previously described (17). Both AP-ICL unhooking and DHT excision assays were performed under single-turnover conditions, which was verified by confirming that the reaction rates remained constant at higher enzyme:DNA ratios. AP-ICL unhooking reactions were performed at 25°C and contained 250 nM full-length mNEIL3 or deletion mutants and 25 nM FAM-labeled DNA in glycosylase buffer (20 mM HEPES-NaOH, pH 7.0, 100 mM NaCl, 5% glycerol, 1 mM DTT, and 10 mg/ml BSA). The reactions were stopped by the addition of 10 mM EDTA, 80% (v/v) formamide at 70°C for 5 min and run on precast 10% TBE/urea gels (Invitrogen) at 180 V in 0.53 TBE buffer. Base excision activity of mNEIL3-GD in the presence of purified GRF12 proteins was carried out by mixing 10 mM mNEIL3-GD with 10 mM or either GRF12 or GRF12 mut , followed by incubation at 25°C with 50 nM FAM-labeled DHTcontaining DNA in glycosylase buffer. Reactions were stopped by the addition of 0.1 N NaOH and 10 mM EDTA, 80% (v/v) formamide and heated at 70°C for 5 min. Band intensities were quantified using gel analyzer and plotted with one-phase association exponential fit using GraphPad Prism 8.

Data availability
Atomic coordinates and structure factors for the hNEIL3-GRF12 crystal structure have been deposited in the Protein Data Bank under accession code 7JL5. All other data are contained within the article.
Acknowledgments-We thank Sylvie Doublié for the NEIL3/pET-30a vector. We thank Lars Pedersen of the NIEHS, National Institutes of Health collaborative crystallography group, and the Advanced Photon Source SE Regional Collaborative Access Team staff for assistance with crystallographic data collection. We thank Lars Pedersen and Bill Beard for comments. The X-ray diffraction data were collected at the Southeast Regional Collaborative Access  Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.