Analysis of DNA binding by human factor xeroderma pigmentosum complementation group A (XPA) provides insight into its interactions with nucleotide excision repair substrates

Xeroderma pigmentosum (XP) complementation group A (XPA) is an essential scaffolding protein in the multiprotein nucleotide excision repair (NER) machinery. The interaction of XPA with DNA is a core function of this protein; a number of mutations in the DNA-binding domain (DBD) are associated with XP disease. Although structures of the central globular domain of human XPA and data on binding of DNA substrates have been reported, the structural basis for XPA's DNA-binding activity remains unknown. X-ray crystal structures of the central globular domain of yeast XPA (Rad14) with lesion-containing DNA duplexes have provided valuable insights, but the DNA substrates used for this study do not correspond to the substrates of XPA as it functions within the NER machinery. To better understand the DNA-binding activity of human XPA in NER, we used NMR to investigate the interaction of its DBD with a range of DNA substrates. We found that XPA binds different single-stranded/double-stranded junction DNA substrates with a common surface. Comparisons of our NMR-based mapping of binding residues with the previously reported Rad14-DNA crystal structures revealed similarities and differences in substrate binding between XPA and Rad14. This includes direct evidence for DNA contacts to the residues extending C-terminally from the globular core, which are lacking in the Rad14 construct. Moreover, mutation of the XPA residue corresponding to Phe-262 in Rad14, previously reported as being critical for DNA binding, had only a moderate effect on the DNA-binding activity of XPA. The DNA-binding properties of several disease-associated mutations in the DBD were investigated. These results suggest that for XPA mutants exhibiting altered DNA-binding properties, a correlation exists between the extent of reduction in DNA-binding affinity and the severity of symptoms in XP patients.

Nucleotide excision repair (NER) 3 is a DNA damage repair pathway specialized for removing bulky lesions arising from exposure to various types of endogenous and exogenous toxic agents (1)(2)(3)(4)(5). Human NER is a multistep process involving coordinated action of over 30 proteins (4,6). Two NER pathways exist, one repairs lesions in actively transcribed DNA (transcription-coupled repair), and the other processes lesions more generally throughout the genome (global genome repair). These differ only in the mechanism by which the presence of damage is recognized. The subsequent steps, destabilization and unwinding of the DNA by transcription factor II H (TFIIH) to create a DNA structure termed the NER bubble, excision of the damaged nucleotide, and gap-filling synthesis, are understood to be the same (6 -22).
XPA is recruited to the damage site by TFIIH once the duplex is unwound (6,7). Although XPA does not have any enzymatic activity, it acts in concert with replication protein A (RPA) as a critical scaffolding protein through its interactions with DNA and other NER proteins (23).
Defects in NER result in the genetic disorder xeroderma pigmentosum (XP), which is characterized by hypersensitivity to sunlight and increased incidence of skin cancer (24 -27). In severe cases, neurological defects are also observed (24, 28 -30). XP arises from mutations in eight genes, seven of which (XPA-G) are directly involved in NER (4,31). Among these, XPA mutations are usually associated with most severe disease phenotypes, many of which map to the DNA-binding domain (DBD) (28 -30). Different XPA mutations lead to different disease phenotypes (24,29,30,32); complete loss, substantial truncation, and unfolding of the DBD are known to lead to severe XP disease phenotypes. However, genotype-phenotype correlations and mechanisms behind missense mutations remain poorly understood (23,30,(33)(34)(35).
Two structures of the globular core of the Saccharomyces cerevisiae homolog of XPA, Rad14, have been determined for complexes with a DNA duplex containing a cisplatin 1,2-deoxydiguanosine intrastrand cross-link or a N-(deoxyguanosin-8-ly)-2-acetylaminofluorene (AAF) lesion (36). Rad14 has remarkably higher affinity for these two modified duplexes than for unmodified duplexes or duplexes containing other lesions (36). The structure of the Rad14 globular core is quite similar to the previously reported structures of the globular core of human XPA obtained in the absence of DNA (37,38). However, we and others have shown that the globular core of human XPA (XPA 98 -219 ) does not bind NER model DNA substrates with appreciable affinity and that additional C-terminal residues are required for full binding affinity (39,40). Consequently, the high binding affinity for the two lesion-containing duplexes of the Rad14 globular core that lacks additional C-terminal residues is confounding and suggests that in the two studies, the Rad14 and XPA DBDs are not functioning in an equivalent manner biochemically and structurally. Notably, evidence has accumulated suggesting that XPA may have roles outside of NER (23,(41)(42)(43)(44)(45)(46). Nevertheless, the difference is puzzling in the context of NER because, although XPA was first thought to be involved in damage recognition (47)(48)(49), it is now understood that XPA is not recruited to the NER preincision complex until after the presence of damaged DNA is recognized and unwound (6,7,50).
Using the XPA 98 -239 construct (XPA DBD) that we previously reported to exhibit full DNA-binding affinity (39), we report here an investigation of the interaction of human XPA with model NER bubble substrates. We used NMR and DNA affinity measurements to define the binding site of XPA DBD for model NER substrates and compared our findings to the structures of Rad14-DNA complexes (36). To confirm our find-ings, we determined the effect on DNA binding of selected structure-based and disease-associated mutants. Our results show human XPA and yeast Rad14 use similar DNA-binding surfaces, but XPA interactions in the context of NER are different from those observed for Rad14 interactions with high affinity lesion-containing duplexes (36). This study also provides insight into the molecular basis for the disease association of certain XPA mutations.

Human XPA binds a range of ss-dsDNA junction substrates
It had been established previously that XPA binds to a ss-dsDNA junction in the NER bubble (51), although controversy remains over whether it is the 5Ј or 3Ј junction. Moreover, the footprint of XPA on the junction remains unclear. To address these issues, we measured the affinities of the XPA DBD for different junction structures and lengths. We have previously used a DNA fluorescence anisotropy assay to characterize affinities, but turned instead here to microscale thermophoresis (MST) to avoid the tendency of XPA-DNA complexes to aggregate in the 384-well plates during the course of fluorescence anisotropy experiments. Comparisons of the values obtained by the two methods revealed the same trends among substrates but higher dissociation constants (K d ) for MST by ϳ5-fold. The systematically weaker binding in the MST experiments is attributable to the higher ionic strength of the buffer.

DNA binding of human XPA
complexes, RPA coats the undamaged ssDNA opposite the damage-containing strand, so Y-shaped junctions with overhangs on both sides are not likely to represent the relevant substrate in a NER complex. Therefore, we next sought to find a junction DNA substrate that is optimized in length and shape so that it interacts with XPA in a manner that avoids nonspecific secondary binding to the substrate. To systematically screen the types of junction and the length of overhangs, a series of DNA substrates were designed containing a GC-rich duplex (sealed with a 4-nt hairpin (HP) for stability) and oligo(dT) overhangs ( Fig. 1, diagrams 2-9). We found a modest but clearly significant difference in affinity for different nucleotide sequences (supplemental Fig. S1), as has been observed for other sequence nonspecific DNA-binding proteins such as RPA (52).
The data show that XPA DBD binds both 3Ј and 5Ј overhang substrates with approximately the same affinity as Y-shaped substrates ( Fig. 2A and Table 1). It is interesting that the length of the overhang could be shortened to 4 nt without any significant effect on the affinity for substrate ( Fig. 2B and Table 1). The shorter overhang was advantageous for NMR studies, so we selected an 8/4 5Ј overhang substrate for detailed analysis.

NMR analysis defines the DNA-binding sites of human XPA
The binding of DNA substrates by XPA DBD (XPA 98 -239 ) was investigated using NMR spectroscopy. The first step in any detailed NMR analysis is the assignment of the signals (resonances) to specific atoms within the molecule. The NMR backbone resonance assignments for XPA DBD were obtained using a standard series of double and triple resonance 2D and 3D experiments (53). This analysis produced assignments for 98, 98, 94, 98, and 97% of the 15 N, 1 H, 13 CO, 13 C␣, and 13 C␤ resonances, respectively (supplemental Table S1). Fig. 3 shows the 2D 15 N-1 H HSQC spectrum labeled with the corresponding backbone resonance assignments used for the titration analyses of the binding of DNA substrates. Because of the extensive overlap and some highly dynamic peak intensities in the spectrum, assignment of the stretch of residues between Gln-208 to Arg-228 was challenging. In this region, there are groups of residues whose connectivity is unambiguously established, however, there are breaks in between that create uncertainty in exactly how they fit to the sequence. The final assignments were made to maximize internal consistency, in conjunction with the NMR titrations, and the mutational analysis described below.
The results from NMR titrations of 15 N-enriched XPA DBD with different DNA substrates were used to map the interaction surface of XPA DBD (supplemental Fig. S2). Overall, the trends in chemical shift perturbations (CSPs) were the same for the different substrates. Major perturbations were found in the globular core between residues 130 and 210 and in residues extending C-terminally from the core between residues 215 and 232 (supplemental Fig. S2). The N-terminal region of the globular core containing a zinc finger remained mostly unaffected. Although the overall patterns are similar, each substrate had some unique features. For example, perturbation of Glu-156 and Trp-175 was only observed for the Y-shaped junction, whereas Ala-229 was only perturbed by the 8/12 5Ј overhang substrate. These observations reflect the adaptation of XPA DBD to the substrate.
An in-depth analysis of NMR chemical shifts was performed for the titration of 15 N-labeled XPA DBD with the 8/4 5Ј overhang substrate. Significant CSPs were observed primarily in the C-terminal portion of XPA DBD including ␤3, ␣1, the hairpin between ␤4 and ␤5, the C-terminal end of ␣3, and a number of residues in the C-terminal extension (Fig. 4). Of the 34 residues perturbed by addition of the substrate, 16 exhibited fast exchange on the NMR time scale, whereas the 18 others exhibited intermediate exchange and were broadened beyond detection. Making the logical assumption that all CSPs arise from the  Fig. 1, respectively). B, plot of MST data for XPA DBD binding DNA substrates with 8-nt duplex and different lengths of 5Ј overhangs (substrates 6 -9 in Fig. 1). All measurements were made at room temperature in a buffer containing 50 mM Tris-HCl, pH 7.8, 150 mM NaCl, 10 mM MgCl 2 , 0.05% Tween 20, and 1 mM DTT. The error bars indicate standard deviations from at least three measurements. Table 1 The affinity of XPA DBD for ss-ds junction DNA substrates Dissociation constants were extracted from the MST data plotted in Fig. 2. The 8/12 HP splayed-arm substrate is composed of a Y-shaped ss-dsDNA junction with 8-nt duplex and 12-nt ssDNA overhangs on both 5Ј and 3Ј ends and a 4-nt HP on the opposite end. The nomenclature for other substrates follows the same system. The numbers in parentheses identify each substrate as shown in Fig. 1, which also includes the chemical structures. The errors are standard deviations from at least three measurements. The less than 3-fold difference between the K d values indicates that the protein does not have substantial junction substrate specificity. (4) 8.2 Ϯ 0.5 8/10 HP 5Ј overhang (6) 5.9 Ϯ 0.2 8/8 HP 5Ј overhang (7) 4.6 Ϯ 0.1 8/6 HP 5Ј overhang (8) 4.2 Ϯ 0.2 8/4 HP 5Ј overhang (9) 3.5 Ϯ 0.2

DNA binding of human XPA
same DNA-binding phenomenon, the residues with broadened signals have the largest chemical shift differences between the free and bound state and are presumed to be centered in the binding site. This group of residues is highlighted in Fig. 4B, which shows the CSPs mapped on the NMR structure of the human XPA globular core.

Comparative structural analysis reveals similarities and differences between the binding of DNA by XPA and Rad14
To obtain insight into whether the NMR analysis of DNA binding to the XPA DBD was consistent with the structure of the Rad14-DNA complexes, the XPA CSPs induced upon binding DNA were mapped onto the structure. Fig. 5 shows ribbon diagrams of the XPA homology model constructed from a Rad14 crystal structure (Protein Data Bank code 5A3D), with the residues highlighted that are involved in DNA binding as identified on the basis of the crystal structures of Rad14 (red, panel A) or by NMR (blue, panel B). Unlike Fig. 4, Fig. 5C shows the summary of all perturbed residues from all five ss-ds junction DNA substrates analyzed by NMR.
Most of the residues involved in contact with DNA in the Rad14 crystal structures exhibit CSPs in the NMR titrations of the human XPA DBD. However, the NMR study revealed many additional residues with significant CSPs induced by the binding of DNA. The most important differences were the CSPs observed in the extra 20 C-terminal residues extending beyond the globular core, which are not present in the Rad14 construct (Fig. 5C). These CSPs are fully consistent with our previous analysis showing that the C-terminal extension beyond the globular core is essential to recapitulating the DNA-binding affinity of full-length XPA (39). Two Rad14 residues, Thr-239 and His-258, contact the DNA, but the corresponding residues in XPA DBD (Lys-151 and His-171) did not exhibit a CSP in the DNA titrations. The absence of an effect on K151 is likely due to a structural adjustment in the flexible loop in which it is found; both neighboring residues around Lys-151, Leu-150, and Asp-152, showed significant CSPs (Fig. 5C), suggesting that one of these residues replaces the Thr-238 contact. As for His-171, involvement of this residue in DNA binding remains unclear because we were unable to assign its resonances.

DNA binding of human XPA
because this residue directly engages DNA but because the reagent was recruited to the vicinity via the interaction with the adjacent residues. The origin of biotinylation but lack of a CSP for Lys-236 is more puzzling. We have established that progressive truncation of the C-terminal extension from the XPA globular core results in progressively reduced DNA-binding affinity (see below), consistent with a general electrostatic contribution from basic residues such as Lys-236 in the C-terminal extension. Although the NMR chemical shift is very sensitive to even weak interaction, we cannot rule out the possibility that a very weak interaction with DNA was canceled out by an induced structural change, and therefore no NMR CSP was observed. This highlights the need for a high resolution structure of an XPA-DNA complex. In the Rad14 crystal structure, Lys-168 does not contact the DNA and Lys-179 does, but further comparison is limited because the construct is truncated and does not contain Lys-221, Lys-222, and Lys-224.
The F262A mutation in Rad14 was reported to cause complete loss of binding to damage-containing duplexes, suggesting the critical importance of this residue in DNA binding by yeast Rad14 (36). The equivalent residue in XPA, Trp-175, showed a significant CSP (Fig. 4), consistent with its involve-   Table 2). In addition, among the five ss-ds junction DNA substrates tested by NMR, a Trp-175 CSP was observed only in the titration of 8/12 splayed-arm substrate.

Mutation of residues in the C-terminal extension of XPA DBD inhibits binding of DNA
To investigate the contribution of residues 220 -239 to DNA binding more closely, we prepared a series of point mutations of basic residues in this region. Three charge reversal, single-site mutations (K221E, K222E, and R228E) resulted in mild reduction in DNA-binding affinity, whereas the double mutation K221E/R228E had a much more dramatic effect (Fig. 6B and Table 2). Truncations of the DBD also caused significant reductions in DNA-binding affinity ( Fig. 6C and Table 2). Although both truncation mutants retained some DNA-binding activity, the shorter construct XPA 98 -227 had weaker DNA-binding activity than the longer XPA 98 -234 . These results confirm that the cluster of basic residues in the C-terminal of XPA DBD contribute significantly to DNA binding (36,40).

Biophysical and structural studies enhance the general understanding of genotype-phenotype correlations for XPA mutations
The available data for generating genotype-phenotype correlations are limited, especially for many missense mutations discovered in cancer patients (23), simply because there have been few investigations of phenotypes or biochemical malfunctions of specific XPA mutants. The results reported here enable some genotype-phenotype correlations to be made.
Insights were obtained for three disease-associated mutants, V166A, L191V, and R207Q. These mutations were prepared in the XPA DBD construct, expressed, and purified, and their structural integrity and DNA-binding affinity were characterized. In WT XPA DBD, Val-166 did not have significant CSPs upon binding of DNA but is between residues that did (Ile-165 and Lys-167), Leu-191 is distant from residues affected by DNA binding, and Arg-207 has significant CSPs (Fig. 5). As anticipated, L191V did not cause any significant effect on DNA binding, whereas R207Q caused a dramatic decrease in DNA-binding affinity. Interestingly, we were unable to concentrate V166A sufficiently to conduct the DNA-binding assay, suggesting reduced stability of this mutant. We were, however, able to establish that the structural integrity of this mutant was maintained, because both the NMR and CD spectra were the same as  (Fig. 1, 9). The error bars indicate standard deviations from at least three measurements.

Table 2
The affinity of XPA mutants for substrate 9 Dissociation constants (K d ) were extracted from the MST data plotted in Fig. 6. The errors are standard deviations from at least three measurements. Dashes indicate the binding was too weak to extract an accurate K d value.

DNA binding of human XPA
WT (supplemental Fig. S3). To directly determine whether the mutation did indeed alter the stability of the domain, thermal denaturation was monitored by CD for the mutant and the WT protein (supplemental Fig. S3B). These data show that the V166A mutation results in a 4°C lower apparent thermal denaturation midpoint than the WT. Hence, although the V166A mutation does not grossly alter folding, it does significantly reduce the stability of the globular core. In the Rad14 crystal structures and homology models of XPA, the Val-166 residue is a part of the ␤-hairpin that intercalates into the ss-dsDNA junction. The decrease in stability could be due to the loss of hydrophobic interactions that support the cross-strand interaction in the ␤-hairpin. Given its important role in XPA binding of DNA, destabilization of the ␤-hairpin may well prove to be the biochemical malfunction at the origin of the disease-association of the V166A mutant.

Discussion
XPA is central scaffold for human NER machinery, and although loss of DNA-binding activity of XPA is anticipated to lead to severe XP symptoms (35), XPA binding to the NER bubble and its effect on NER efficiency or correlation to XP symptoms have not been investigated. One contributing factor was that the DNA-binding domain had been incorrectly assigned to the XPA globular core, which has only weak affinity for DNA (39,54). Important new insights were obtained recently when the first high-resolution structures in the presence of DNA were reported for the yeast homolog of XPA, Rad14 (36). However, whether or not these structures serve as accurate models for human XPA in the context of NER remains unclear because the DNA substrates in the crystal structures were 15-or 16-bp DNA duplexes modified at the center with specific lesions. These substrates are not representative of canonical NER, because XPA is not recruited to the site of damage until TFIIH unwinds the damaged DNA (6,7). The specific lesion-containing duplexes appear to be very unique substrates because in general, XPA binds ss-ds junctions much more tightly than duplexes (51). Indeed, Rad14 binds unmodified duplexes and other lesion-containing duplexes with substantially lower affinity than duplexes modified specifically with either cisplatin and AAF (36). Although these observations are consistent with the observations that purified XPA preferentially binds to damage containing DNA in vitro (55)(56)(57), it is widely accepted that these lesions are first recognized by XPC in global genome repair or by stalling of RNA polymerase in transcription-coupled repair (2,50,58). XPA has been shown to interact with proteins not involved in NER (e.g. XPA binding proteins (41)(42)(43)(44) and proliferating cell nuclear antigen (23,45)), and it is conceivable that the Rad14 structures reveal the molecular basis of DNA interactions in the other pathways for cellular processing or bypass of damaged DNA. Regardless, structural information on human XPA in complex with NER substrates is required to better understand the molecular basis of XPA function in NER.
We have investigated the interaction of human XPA with DNA using a DBD construct that exhibits the same DNA-binding affinity as the full-length protein and with ss-dsDNA junction substrates that model the NER bubble created after the damaged duplex is unwound by TFIIH. NMR backbone resonance assignments allowed identification of the residues affected by binding to DNA substrates. Analysis of different DNA substrates resulted in similar CSPs, suggesting that the same set of residues is involved in DNA binding for variety of DNA structures with different binding affinities. For the most part, there was a good correlation between NMR CSPs of residues in the globular core and the corresponding DNA binding site in the yeast Rad14 crystal structures. Thus, the two proteins are found to utilize the same general mode of binding DNA. This result is consistent with the conservation of key DNAbinding residues in the globular core (23,36).
Nevertheless, uncertainties remain as to whether the Rad14 structures provide an accurate model for DNA binding by XPA in the context of NER. In particular, the NMR analysis revealed significant DNA contacts with residues in the C-terminal residues that extend beyond the globular core and are missing from the yeast construct, consistent with our finding that the XPA globular core lacking these residues does not bind DNA with appreciable affinity (39). Our mutational analysis of basic residues at the C terminus of XPA DBD shows that these residues collectively contribute to the high affinity for DNA. It is possible that the absence of C-terminal residues in the Rad14 construct used for crystallization lead to differences in the mode of binding of differences substrates.
The observation that Rad14 binds with high affinity only to duplexes containing cisplatin or AAF lesions, and not to unmodified duplexes or duplexes with different lesions, further supports the idea that the crystal structure may reveal a unique mode of DNA binding (36). It is also conceivable that the yeast and human proteins do not completely correspond; for example, the F262A mutant in the Rad14 globular core completely abolished binding to DNA (36), whereas the corresponding mutation in human XPA DBD (W175A) did not significantly alter DNA-binding activity (Fig. 6A). However, it is not possible to rule out the possibility that this difference is the result of the differences in the constructs used in these two studies.
It is also difficult to rationalize that the crystal structures of Rad14 have two molecules bound, one on either end of the lesion-containing duplex (36). That said, it is conceivable that the end-on mode of interaction of Rad14 is in some way mimicked when XPA is engaged at the ss-dsDNA junction within the NER bubble. In fact, characterization of XPA in vitro revealed that it can form a dimer, although the nature of this dimer is not known (57,59). Nevertheless, considering the extensive network of protein-protein interactions of XPA in NER complexes (60 -65), it is difficult to imagine how two molecules of XPA could interact simultaneously within the preincision and incision complexes. Indeed, all current models of NER complexes incorporate only a single XPA molecule (23,66).
Our studies also shed light on genotype-phenotype relationships for certain XPA mutants; in particular, we believe a correlation exists between the effect of mutation on DNA-binding affinity and severity of symptoms in XP patients. For example, XPA patients expressing truncation mutant XPA 1-227 are known to exhibit mild neurological disorders, whereas those with XPA 1-220 have more severe neurological symptoms (30).

DNA binding of human XPA
Our current and previous data show that XPA DBD truncated at Arg-227 retains weak DNA binding ( Fig. 6C and Table 2), whereas truncation at Phe-219 barely retains any DNA-binding activity (39). These observations suggest a correlation exists between the degree of inhibition of XPA DNA-binding activity and XP disease phenotype. Hence, the observation of only very small reductions in DNA-binding affinity for the single-site charge reversal mutations of Lys-221 and Arg-228 ( Fig. 6B and Table 2) suggests that the K221Q and R228Q mutations will result in relatively mild disease symptoms.
DNA-binding affinity is but one of several factors that can contribute to biochemical malfunction of disease-associated mutations. For example, we found that the V166A mutation is destabilized compared with the WT protein, which could affect not only DNA binding but also protein interactions and cellular turnover. Arg-207 presents another interesting example. We found that the disease-associated R207Q mutation has significantly weakened DNA-binding affinity, whereas a host reactivation assay showed that the R207E mutant is capable of the repair of UV lesions (67). The assay is not sensitive to effects on the rate of repair, which may be compromised when DNAbinding affinity is diminished. Moreover, this residue is also important for interactions with the NER factor XPE (68).
The remarkably rapid progress in recent years in structural analysis of large multiprotein complexes, such as those assembled along the trajectory of NER, holds promise for dissecting such complex relationships. Systematic functional analyses, for example of UV lesion repair efficiency, will also be required. The combination of mechanistic and functional data will be critical to attaining the ultimate objectives of a more complete understanding of the relationship of XPA DNA-binding affinity to NER activity and of the genotype-phenotype correlation of disease-associated mutations.

XPA DBD mutant construction
We previously reported the construction and purification of XPA DBD (XPA 98 -239 ) (39). Using the pBG100 XPA DBD plasmid as the template, single-site mutations were introduced using the Q5 site-directed mutagenesis kit (New England Bio-Labs, Inc.) following the manufacturer's protocols. The K221E/ R228E double mutant was created using K221E primers and pBG100 XPA DBD R228E plasmid as the template. Primers for mutagenesis are listed in supplemental Table S2. For the truncation mutants R228X (XPA 98 -227 ) and W235X (XPA 98 -234 ), XPA constructs were amplified by PCR using the oligonucleotides indicated in supplemental Table S2 to introduce 5Ј BamHI and 3Ј NotI cleavage sites. All mutants were cloned into the pBG100 in-house expression vector (L. S. Mizoue, Center for Structural Biology, Vanderbilt University), which incorporates an N-terminal human rhinovirus 3C (HRV3C) protease cleavable His 6 tag. Fig. 1 shows the structures of DNA substrates used in this study. The position of the FITC tag is indicated if applicable. Desalted oligodeoxynucleotides were purchased from Sigma-Aldrich. Y-shaped ssDNA-dsDNA junctions, and duplexes were prepared by mixing an equimolar amount at 0.5-1 mM of each strand in the buffers listed below for NMR or DNA binding experiments. Then the mixture was heated in a boiling water bath and allowed to cool to room temperature for annealing. Hairpin-containing DNA substrates were dissolved in TNE buffer (10 mM Tris, pH 7.0, 50 mM NaCl, 1 mM EDTA) at 2 M concentration and annealed by heating in a boiling water bath followed by immediate cooling on ice.

XPA DBD production
WT and mutant XPA DBD proteins were expressed and purified as described previously (39). The XPA DBD used for NMR titrations was expressed in minimal medium containing 0.5 g/liter of 15 NH 4 Cl (Cambridge Isotope Laboratories, Inc.). The preparation of samples for backbone resonance assignments also contained 2 g/liter 13 C 6 -glucose (CIL, Inc.).

Measurement of DNA binding affinities
The protocol for the fluorescence anisotropy DNA-binding assay was described previously (39). For measurements using MST (69 -72), proteins were dialyzed into MST buffer (50 mM Tris-HCl, pH 7.8, 150 mM NaCl, 10 mM MgCl 2 , 0.05% Tween 20, and 1 mM DTT). Fluorescein-labeled DNA stocks were also diluted in the MST buffer. All samples and buffer were passed through a 0.2-m filter. For each experiment, 16 dilutions of the protein were prepared to varying concentrations. DNA was added to each of the sample to the final concentration of 40 nM in the tubes provided in the Monolith NT.115 series standard treated capillaries kit (NanoTemper, Inc.). All experiments were carried out at room temperature using the capillaries in a Monolith NT.115 Blue/Red instrument (NanoTemper, Inc.) at 20% LED power and 40% MST power. The data were analyzed using MO.Affinity software (NanoTemper, Inc.).

Circular dichroism
Samples of XPA DBD WT and the V166A mutant were dialyzed into a buffer containing 20 mM Tris, pH 7.0, 150 mM KCl, and 1 mM DTT. All samples and buffer were passed through a 0.2-m filter before data collection. The concentration of protein was adjusted to 11 M, and far-UV CD data were collected at 220 nm over the range 15-65°C using a Jasco J-810 CD spectropolarimeter (Easton, MD).

Generation of the XPA DBD homology model
A homology model of human XPA 102-214 was generated using the X-ray crystal structure of S. cerevisiae Rad14 (Protein Data Bank code 5A3D) as template for calculations using Modeler 9.14 (73).

H 1D NMR analysis of XPA DBD V166A
XPA DBD WT and V166A mutant were concentrated to 30 M in a buffer containing 20 mM Tris, pH 7.0, 75 mM KCl, and 1 mM TCEP to which 5% 2 H 2 O was added prior to performing the experiment. The 1 H 1D spectrum of each sample was acquired in a 3-mm tube at 25°C using a Bruker Avance 600-MHz spectrometer equipped with a cryoprobe. DNA binding of human XPA NMR backbone resonance assignments 13 C, 15 N-enriched XPA DBD was concentrated to 460 M in a buffer containing 20 mM Tris, pH 7.0, 500 mM KCl, 1 mM TCEP, and 5% 2 H 2 O. 15 N-1 H HSQC and a series of standard heteronuclear triple resonance 3D experiments (HNCO, HNCA, HNCACB, HNCOCA, and CBCACONH) were acquired in a shaped tube at 25°C using a Bruker Avance-III 600-MHz spectrometer equipped with a QCI cryoprobe. All 3D data sets were collected using flip-back pulses and a Watergate sequence to suppress the water signal (74,75), as well as 25 and 50% nonuniform sampling (76 -81) for the backbone and side chain experiments, respectively. Reconstruction and processing of the raw data were performed with the standard Topspin (Bruker) software, using compressed sensing, one time 0 filling, and shifted squared sine bell (60 -90 degree) apodization. Further data processing and figure preparation were carried out using SPARKY (82). Resonance assignments were obtained and their reliability evaluated using NMRView software, utilizing the RunAbout tool (One Moon Scientific).

NMR titration of XPA DBD with DNA
All samples for DNA titrations were concentrated to 50 M in the NMR buffer containing 20 mM Tris, pH 7.0, 150 mM KCl, 1 mM TCEP, and 5% 2 H 2 O. Titrations were performed at 35°C using a Bruker Avance-III 800-or 900-MHz spectrometer equipped with a TCI cryoprobe following the procedure described previously (39). A second titration with 8/4 5Ј overhang substrate was performed in the NMR buffer with 250 mM instead of 150 mM KCl at 25°C to enable transfer of the assignments from the conditions used for obtaining resonance assignments.
Author contributions-N. S. and W. J. C. designed research. N. S. conducted most of the experiments and analyzed data. M. W. V. performed resonance assignment and provided assistance for NMR experiments. M. S. R. subcloned, expressed, purified, and generated preliminary DNA-binding data for some of the disease-associated XPA DBD mutants. A. M. T.-W. assisted preparation of NMR samples, as well as some MST experiments. N. S. and W. J. C. wrote the paper.