Structural and Functional Characterization of the Human DNA Repair Helicase XPD by Comparative Molecular Modeling and Site-directed Mutagenesis of the Bacterial Repair Protein UvrB*

A molecular model for the human nucleotide excision repair protein, XPD, was developed based on the structural and functional relationship of the protein with a bacterial nucleotide excision repair (NER) protein, UvrB. Whereas XPD does not share significant sequence identity with UvrB, the proteins share seven highly conserved helicase motifs that define a common protein structural template. They also have similar functional roles in their ATPase activity and the ability to unwind DNA and verify damaged strands in the process of NER. The validity of using the crystal structure of UvrB as a template for the development of an XPD model was tested by mimicking human disease-causing mutations (XPD: R112H, D234N, R601L) in UvrB (E110R, D338N, R506A) and by mutating two highly conserved residues (XPD, His-237 and Asp-609; UvrB, H341A and D510A). The XPD structural model can be employed in understanding the molecular mechanism of XPD human disease causing mutations. The value of this XPD model demonstrates the generalized approach for the prediction of the structure of a mammalian protein based on the crystal structure of a structurally and functionally related bacterial protein sharing extremely low sequence identity (<15%).

Nucleotide excision repair (NER) 1 is a process by which damaged nucleotides, from UV light and chemical carcinogens, are removed from DNA. NER is one of the most highly conserved biochemical pathways, and enzymes that mediate this process appear in prokaryotes, archaea, and eukaryotes (1,2). NER can be viewed in the following five continuous steps: 1) damage recognition and verification, 2) incision, 3) excision, 4) repair synthesis, and 5) ligation. During the damage recognition step, a protein complex first identifies a structural perturbation in the helical DNA, which is then verified by additional damage processing and strand opening proteins. Once the lesion has been verified, endonucleases are recruited to the damaged strand performing the initial incision 3Ј followed by a 5Ј incision facilitating the excision of an oligonucleotide containing the damage. Repair synthesis fills in the resulting gap. Finally DNA ligase seals the newly completed repair patch. The evolutionary conservation of proteins involved in NER is exemplified by human XPD, which maintains 56 and 52% (protein) sequence identity, respectively, with the equivalent Schizosaccharomyces pombe, Rad15, and Saccharomyces cerevisiae, RAD3, proteins over the length of the entire protein (BLAST alignments). Human and mouse XPD retain 98% (protein) sequence identity (3).
XPD is one of the constituent proteins forming the TFIIH complex that is responsible for damage processing and strand verification (4,5). XPD and XPB have intrinsic helicase activities that are absolutely required for NER (4). TFIIH is also required for initiation of RNA polymerase II transcription. Mutations in the XPD gene can lead to one of the following three disease syndromes in humans: xeroderma pigmentosum (XP), trichothiodystrophy (TTD), and Cockayne syndrome (CS) (6). Analysis of specific point mutations associated with each disease has revealed that a majority of patients with XPD have mutations in one of seven helicase motifs (6).
Helicases from bacteria to humans are believed to have evolved from a common ancestor (7). Helicase motifs may be so highly conserved because they represent a fundamental mechanism for DNA processing necessary for transcription, replication, repair, and recombination (8). Mutations in several helicase proteins have been linked to cancer-related and accelerated aging disorders (Werner's syndrome, Bloom's syndrome, trichothiodystrophy, Cockayne's syndrome, and xeroderma pigmentosum) (9 -11).
Helicases have been classified into three large superfamilies (designated SF1, SF2, and SF3) and two smaller families (F4 and F5) (12). Proteins belonging to SF1 and SF2 are characterized as such based on possessing a set of seven conserved motifs ranging in length from about 20 to 40 amino acids. These helicase motifs demonstrate both a sequence and three-dimensional structure conservation. SF1 and SF2 helicase proteins were predicted to have a conserved ␤-pleated sheet structure based on other solved NTPase structures sharing these motifs (12). Structural studies have confirmed this prediction. Helicases contain the Walker box A and B motifs found in many NTP-binding proteins (in helicase motif I and II). Individual specificity of the protein function is conferred by the insertion of other domains not shared between the proteins. Crystal structures of three known monomeric/dimeric helicases and one repair protein that shares a common helicase fold have been determined as follows: 1) Escherichia coli Rep with ssDNA alone, and a ternary complex with ADP (13, 14); 2) Bacillus stearothermophilus PcrA with ADP-PNP, ssDNA, and Mg 2ϩ (15)(16)(17); 3) hepatitis C virus NS3 with single-stranded oligonucleotide (14, 18 -20); and 4) UvrB, from Thermus thermophilus (21,22), and Bacillus caldotenax with ATP and Mg 2ϩ (23,24).
In general, structural homology between related proteins is much more highly conserved than sequence homology (25). Thus the helicase motifs possess similar folds and relative three-dimensional relational positions within the proteins despite their low overall sequence identity. Over the entire length of the protein, XPD shares 12.8% sequence identity and 61% sequence similarity with PcrA, 15% identity and 59% similarity with Rep, 15% identity and 62% similarity with UvrB ( Fig. 1), and 12.5% identity and 56% similarity with NS3. Both UvrB and XPD belong to SF2 and exhibit greater similarity in sequence as compared with the three other helicases whose structures have been solved. In addition, UvrB serves a homologous function as TFIIH in bacterial NER (24,26,27) in that it assists in verifying a damaged site and processing the lesion into an open conformation apparently by inserting a ␤-hairpin between the two DNA strands (23,24,26). Finally, both XPD (as part of TFIIH) and UvrB serve as a scaffold on which to recruit the nucleases that perform the dual incisions. Structurally UvrB is folded into four domains, 1a, 1b, 2, and 3 ( Fig. 2A). Domains 1a and 3 contain helicase motifs I-III and IV-VI, respectively, sharing a common fold to other monomeric helicases such as NS3 (23).
The goal of this study was to create a molecular model of the human NER protein XPD, based on the solved crystal structure of the related bacterial NER protein, UvrB (23), as a template. These proteins are closely related based on their common helicase structural motifs and biological function, to open doublestranded DNA during NER (28). Accuracy of the XPD model was confirmed experimentally by expressing B. caldotenax UvrB proteins containing mutations corresponding to known disease-causing mutations in XPD. In addition, two residues highly evolutionarily conserved throughout all bacterial UvrB species and also RAD3 and XPD (mouse and human) were also mutated. The behavior of these mutant UvrB proteins was examined through incision, DNA binding, and oligonucleotidedestabilizing assays in a reconstituted system containing purified UvrA and UvrC from B. caldotenax. The XPD model serves as a guide for structural-functional studies of the protein and assists in gaining a molecular understanding of the NER process in humans, XPD disease-causing mutations, and XPD protein-protein and protein-DNA interactions.

Molecular Modeling of XPD
Multiple Sequence Alignment-A structure-based multiple sequence alignment was developed for the DNA helicase proteins, using the location of domains and helicase motifs, as these sequences do not possess high sequence identity. The alignment included sequences of all the solved monomeric helicase structures (Rep, PcrA, NS3, and UvrB), five bacterial UvrB sequences, human and mouse XPD, and RAD3 the yeast XPD homologue sequences. Part of this multiple alignment, the alignment of UvrB and XPD sequences, is included in Fig. 1. The alignment method used was based on the work of Gorbalenya et al. (29) and Gorbalenya and Koonin (30) using the programs DIAGON (31) and OPTAL (32) and the Sankoff algorithm (33) for multiple sequence alignment. The XPD sequence was aligned by hand to the multiple sequence alignment described for SF2 helicases by Gorbalenya et al. (29) and Gorbalenya and Koonin (30) using the RAD3 sequence as a guide.
XPD Model Construction-The XPD protein model was constructed using the bacterial B. caldotenax UvrB crystal structure coordinates ( Fig. 2A (Protein Data Bank code 1D9Z) (23)) as the template. The sequence alignment indicated that XPD shared closer sequence similarity within the helicase motifs with UvrB, and the alignment resulted in the fewest deletions and insertions throughout the protein sequence as compared with Rep, PcrA, and NS3. UvrB and XPD also share an analogous function as nucleotide excision repair proteins. The XPD homology model was constructed using the Accelrys Homology software package (MSI Homology Users Guide, San Diego). There were 15 loop insertions constructed using the Molecular Simulations homology modeling loop generation software. Because the structure for these loops is less well determined than the rest of the model, as they are not built from experimentally determined structure, they are indicated in black in the model structure in Fig. 2B. UvrB residues 180 -301 were deleted from the XPD model structure, as there are no corresponding residues in the XPD helicase. XPD contains an additional domain, XPD residues 283-448, not present in UvrB. This domain is not similar to any other helicase protein domains and does not possess sequence homology to any other solved protein or protein domain structures. A suggestion of a possible structure for this domain is made from threading and inverse folding algorithms in combination with secondary structure prediction methods (as described in the Supplemental Material). The structural model of this domain is shown in Fig. 3.
Modeling the Inserted (Non-homologous) XPD Domain (Residues 283-448)-Three protein-threading algorithms were applied independently to identify a protein fold to model the XPD domain represented by residues 283-447 (not present in UvrB) as follows: the GenThreader algorithm (34,35), the 3D-PSSM algorithm (36,37), and the Bioinbgu algorithm (42). The top protein folds for this XPD domain given by all three algorithms independently were all ␣-helical proteins, and all shared a fold with nucleic acid-binding proteins. The consensus from these three methods is that this domain has a protein fold similar to that of the AraC family transcriptional activator (Protein Data Bank entry 1BL0) or DNA-binding domain of the centromere binding protein B (CENP-B) (Protein Data Bank entry 1BW6). Both of these proteins are members of the same SCOP data base protein fold class characteristic of all ␣-helical DNA/RNA binding, three ␣-helical bundle proteins. The structural model of this domain based on this fold is shown in Fig. 3.
Modeling the XPD C Terminus (Residues 713-761)-The NMR and crystal structure of the UvrB C terminus were solved independently of the rest of the protein (38,39). This part of the protein was found to have a helix-turn-helix fold and to interact with a homologous domain in UvrC. There was no homology or sequence similarity between the XPD and UvrB C termini.
As the C-terminal XPD sequence does not exhibit high sequence identity to any previously solved structure, threading was used to search for a "structural" similarly folded homologue. Consensus results from seven protein threading algorithms (LOOPP (40), The Sausage Machine (41), 3D-PSSM (36,37), Bioinbgu (42), GenThreader (34,35), COBLATH (43), and Fugue (44)) were used in concert with secondary structure prediction algorithms (PSIPRED (45), PHD (46), SSPRO (47), PROF (48), and PREDATOR (49)) to develop a threaded model for the XPD C terminus (49 residues, 713-761). The top structural hits from all threading methods were all ␣-helical proteins (all members of the CATH 1 10 10 60 class), and secondary structure prediction methods also predicted the 49 C-terminal residues would form an ␣-helical bundle. Similar protein structures that were selected by more than one of the seven different threading methods are as follows: 1OCTC, the oct-1 POU domain (a DNA binding protein); 1BH9A (TBP associated factor); and 4HB1, a designed four-helical bundle.
A structural model was developed for the XPD C terminus based on the Protein Data Bank structure entry 4HB1, which is the structure of a designed 4-helical bundle. This structure consists of two 2-helical bundles related by a 2-fold axis of symmetry through the center. Only the coordinates from two of the helices were used to model the XPD C terminus. The fold of the XPD C terminus is shown to be similar to that of UvrB. The threaded model structure developed for the XPD C terminus is shown and compared with the UvrB experimentally determined C-terminal structure (38,39) in Fig. 3. A significant number of frequently occurring XPD disease-causing mutations of XP or TTD patients have mutations occurring in the C terminus of the protein (50). C-terminal XPD residues 658, 675, 683, 713, and 722 have been shown experimentally to bind to the N-terminal domain of another component TFIIH protein, p44 (50). p44 residues Asp-66, Glu-166, Asp-178, and Gly-200 have been shown experimentally to interact with XPD (51).

Experimental Assays
Construction and Expression of UvrB Mutants-UvrA, UvrB, and UvrC proteins from B. caldotenax were purified by standard procedures (New England Biolabs IMPACT TM T7 system manual) with some modifications for each protein, which will be published elsewhere. T4 polynucleotide kinase was purchased from Invitrogen. Pfu DNA polym- erase was purchased from Stratagene. All UvrB mutants were constructed by PCR using QuikChange site-directed mutagenesis method (Stratagene) with Pfu Turbo DNA polymerase, pUC18uvrB Bca as a template DNA and two oligonucleotide primers (Genosys), each complementary to opposite strands of the template DNA, containing the desired mutation. Resulting mutant plasmids were verified by sequencing for the absence of additional PCR-induced mutations in the uvrB gene sequence. Mutated uvrB gene was finally subcloned into pTYB1 expression vector.
DNA Substrates-Fluorescein-containing DNA substrates were synthesized by Sigma Genosys. The DNA sequence of a 50-bp dsDNA substrate containing a single internal fluorescein adduct (F 26 -50 dsDNA) is shown in Fig.2A. For 5Ј labeling 10 pmol of 50-mer fluorescein-containing top strand was incubated with 25 units of T4 polynucleotide kinase in 70 mM Tris/Cl (pH 7.6), 10 mM MgCl 2 , 100 mM KCl, 1 mM 2-mercaptoethanol, and 15 pmol of [␥-32 P]ATP (3000 Ci/mmol). After incubation at 37°C for 1 h, the reaction was terminated by incubation at 80°C for 10 min in the presence of 20 mM EDTA. Annealing of the top and the bottom strand was performed in the presence of 50 mM NaCl, followed by purification through Bio-Spin P-30 polyacrylamide gel column (Bio-Rad) for removal of unincorporated nucleotides. The double-stranded character and homogeneity of the 50 bp substrate were examined by a restriction assay (38) and analyzed on a 10% polyacrylamide sequencing gel under denaturating conditions. The cholesterol containing substrate was a generous gift from N. Goosen (Leiden, The Netherlands) (52).
The DNA sequence of the helicase substrate (HS1F-M13mp19) was described previously (26). Five pmol of a 26-mer containing an internal fluorescein adduct (HS1F) were labeled at its 5Ј terminus under the same conditions as the F 26 -50 top strand. The helicase substrate was constructed by hybridizing 0.4 pmol of 5Ј-labeled HS1F oligonucleotide with equimolar amounts of M13mp19 (ϩ) strand (Invitrogen) and purified as described above.
Gel Mobility Shift Assay-Binding reactions were performed with 2 nM DNA substrate (5Ј-32 P-labeled fluorescein-50 dsDNA) with 20 nM B. caldotenax UvrA and 60 nM B. caldotenax UvrB in 20 l of UvrABC buffer (50 mM Tris/Cl (pH 7.5), 10 mM MgCl 2 , 50 mM KCl, 1 mM ATP, 5 mM DTT) for 20 min at 55°C. Glycerol was then added to the reaction (8% v/v), and the reaction mixture was loaded onto a 4% native polyacrylamide gel (80:1). The gel and the running buffer (89 mM Tris, 89 mM boric acid, 2 mM EDTA) contained 1 mM ATP and 10 mM MgCl 2 . The electrophoresis was performed for 3 h at 100 V at room temperature. The gel was dried and exposed against Storage Phosphor Screen (Amersham Biosciences) overnight at room temperature.
UvrABC Incision Assay-The 5Ј-terminally labeled cholesterol-50 or fluorescein-50 dsDNA substrate (2 nM) was incised by UvrABC (20 nM UvrA, 60 nM UvrB, and 50 nM UvrC) in 20 l of UvrABC buffer at 55°C for 1 h. The reaction was terminated by ethanol precipitation. The samples were denatured with formamide and heated to 90°C for 5 min and then quick-chilled on ice. The incision products were analyzed by electrophoresis on a 10% polyacrylamide sequencing gel under denaturating conditions at 400 -500 V with TBE buffer. In the case of the helicase substrate incision, the reaction mixture (15 l) contained ϳ8 fmol (in ssDNA circles) of DNA substrate, 50 nM UvrA, 100 nM UvrB, and 100 nM UvrC in buffer A2 (50 mM Tris/Cl (pH 7.5), 100 mM KCl, 15 mM MgCl 2 , 1 mM EDTA, 5 mM ATP, 2 mM DTT) and was incubated at 42°C for 1 h. The reaction was quenched with 5 l of stop solution (25% (v/v) Ficoll, 1% SDS, 0.1 mM EDTA, 0.25% orange G) and heated for 2 min at 85°C, and the entire sample was then loaded onto a 15% denaturating polyacrylamide gel equilibrated with TBE running buffer. Electrophoresis was carried out at 400 -500 V for 1-2 h. The gels were processed as described above.
ATP Hydrolysis Assay-The conversion of ATP to ADP by the UvrABC system was determined by a coupled enzyme assay system consisting of pyruvate kinase and lactate dehydrogenase to link the hydrolysis of ATP to the oxidation of NADH. The assay mixture consisted of 50 mM Tris/Cl (pH 7.5), 50 mM NaCl, 4 mM MgCl 2 , 1 mM DTT, 20 units/ml lactate dehydrogenase, 20 units/ml pyruvate kinase, 2 mM phosphoenolpyruvate, 0.15 mM NADH, and 200 nM Uvr proteins in the presence or absence of 50 ng of UV-irradiated DNA substrate. The continuous assay was performed at 55°C for 60 min; the ATPase activity was determined from the slope of the turnover of NADH versus time.

The XPD Model
The structure of the molecular model developed for the human XPD protein, based on the x-ray crystal structure of B. caldotenax UvrB (23,24), is shown in Fig. 2B. This model includes all the defined SF2 helicase motifs, domains 1a, 1b, 2, and 3, and structure for XPD residues 10 -282 and 448 -712 based on the UvrB structure coordinates as the template. The model of the additional XPD domain, residues 283-448, absent in UvrB, and location relative to the helicase motifs are indicated in Fig. 3. The XPD C-terminal model structure (residues 713-761) and location relative to the helicase motifs are shown in Fig. 3. Fig. 4 indicates the location of all the currently known disease (XP, TTD, and CS)-associated mutations found in XPD on the structural model. It is interesting to note that the majority of the disease-associated mutations are found in domains 1a and 3 whose structure is highly conserved across all helicases. Other frequently occurring disease-causing muta- tions are located in helicase motifs III, V, and VI (3,6,53,54). No disease-associated mutations have been observed in the XPD-inserted threaded domain (residues 283-447).
Overall comparison of the modeled XPD structures with other solved monomeric helicase structures illustrates the striking conservation of the central ␤-pleated sheet structure of domain 1a with surrounding ␣-helices common to NTPases as first evidenced in the structure of the nucleotide binding domain of e1F4A (55). This domain contains the Walker A and B motifs common to NTPases and is involved in ATP binding and hydrolysis. Domain 3 of the XPD model also conserves the central ␤-pleated sheet structure with surrounding ␣-helices of FIG. 3. The location of the helicase motifs in the UvrB structure and XPD model and the relative location of protein domains present in XPD that are absent in UvrB. Helicase motifs are colored as follows: I, dark blue; Ia, light blue; II, orange; III, red; IV, yellow; V, green; and VI, magenta. The threaded models developed for the XPD-inserted domain and the C terminus are shown above their relative locations in the protein sequence and relative to the helicase motis in the bar view of the protein. UvrB structure highlighting helicase motifs and the five residues selected for experimental mutation are as follows: Glu-110, Asp-338, His-341, Arg-506, and Asp-510. XPD model highlighting the residues corresponding to the experimental mutations made in UvrB as follows: Arg-112, Asp-234, His-237, Arg-601, and Asp-609. The orientation of UvrB and XPD has been rotated 90°from Fig. 2, such that the "top view" is presented.

FIG. 4. Location of many diseasecausing mutations giving rise to XP, CS, and TTD syndromes in patients.
Notice most disease-causing mutations occur in protein domains 1a and 3 and in regions of the XPD model that are highly conserved with UvrB. The orientation of XPD is the same as in Fig. 2. the other solved SF1 and SF2 family helicases. Significant differences between the XPD modeled structure and UvrB structure are observed in domain 2, with the insertion of a large loop in place of the antiparallel ␤-sheets in UvrB, and domain 1B where the ␣-helices of UvrB are absent in XPD. In addition, the ␤-hairpin believed to play an essential role in DNA binding in UvrB (26) is not as distinct and extended in the XPD model structure as the two proteins exhibited low sequence similarity in this region. The overall root mean square deviation between the backbone UvrB structure and XPD model (superposition of 3392 atoms) is 1.89 Å (excluding the XPD-inserted domain and C terminus).

Validation of XPD Model through Site-directed Mutagenesis of UvrB
Experimental Strategy, Selection of Mutants, and Biochemical Analyses-The validity of our modeling approach was tested through biochemical experiments with mutant UvrB proteins containing homologous changes as some of the disease-causing mutations in the XPD gene. Two categories of experimental mutations were made (Table I). One group of UvrB mutants was selected as they represent the residues analogous to the XPD disease-causing mutations (E110A/ E110R, D338N, and R506A). The second group of mutants was selected based on their high conservation in UvrB, RAD3, and XPD. This second group of UvrB mutants (H341A, D510N/ D510A) can be thought of as "predictive" in that they are based on their conservation to have functional significance. The structural locations of the UvrB mutants constructed and purified are shown in Fig. 3. The biochemical properties of these UvrB mutants were studied by incision (Fig. 5), gel mobility shift, oligonucleotide destabilization (Fig. 6), and ATP hydrolysis assays (Table II).
Mutations in Helicase Motif II-The D338N and H341A mutants were selected due to their location in the "DEAD" box Walker B highly conserved helicase II motif found in many NTPases. In addition, His-341 (motif II) is believed to interact with residues in motif VI and III (56). Asp-338 corresponds in sequence alignment with XPD to a high prevalence D234N XP disease-causing mutation. Overall, the UvrB mutations in the helicase motif regions identical to XPD disease-causing mutations had reduced DNA repair capacity. Previously, XPD disease-causing mutations (XPDG675R, R722W) have been isolated and purified and have been shown experimentally to perform poorly in helicase activity assays (50). The RAD3 K48R ATP-binding site mutant, which is part of the GXGKT Walker A Box, helicase I motif, has been shown to lose ATPase and DNA helicase activities but still binds ATP (57). This same mutation has been made previously in E. coli UvrB and shown to reduce greatly its ATPase activity (58). Helicase SF2 members contain the conserved consensus "DEAD/H" box in helicase motif II. The first (Asp) residue of this motif binds the Mg ϩ2 ion; the second residue (Glu) is proposed to bind water during ATP hydrolysis. Studies in other helicases have shown mutations in motif II to be ATPase-deficient and incapable of DNA unwinding (7). The UvrB D338N mutant 1) is strongly impaired in its ability to support incision of damaged fluorescein or cholesterol-containing DNA (Fig. 5); 2) did not form a complex with DNA (Fig. 6A); 3) is defective in the destabilization of a damaged oligonucleotide (Fig. 6B); and 4) displayed significantly reduced ATP hydrolysis activity (Table II). The D338A mutation has been reported previously (59) in E. coli UvrB, although it was made under the assumption that the protein was functioning as a nuclease and not to test unwinding activity. In contrast mutating the nearby His-341 to alanine (conserved in 29/29 bacterial species, Table I) in helicase motif II resulted in a milder defect in the four functional assays.
Mutations in Helicase Motif V-The UvrB Arg-506 mutation corresponds to the high prevalence R601L/R601W XP diseasecausing mutation. Mutating this residue to alanine had a mild phenotype in the four biochemical assays. However, mutating the nearby Asp-510 residue to alanine (corresponding to XPD Asp-609), in the highly conserved helicase V motif, resulted in drastic reduction in activity across all the biochemical assays. The Asp-510 residue was mutated because it is part of a highly conserved motif found in helicase motif V, EG(I/L)D that is  conserved from UvrB to RAD3 (S. cerevisiae) and XPD (human). The Glu residue of this motif is believed to be involved in long distance interactions across the protein with motif Ia (56). This mutation was reported previously by Lin et al. (59) in E. coli UvrB and shown to reduce its activity. Other residues in helicase motifs V and VI have been shown to be important for ATP hydrolysis; UvrB cannot hydrolyze ATP when Arg-540 and Arg-543 are mutated (24). It is interesting to note that the D510N mutation has a mild phenotype in most biochemical assays but has a significantly higher UvrA/UV DNA-stimulated ATPase activity than the wild type UvrB.
TTD-like Mutation in the ␤-Hairpin-The E110A/E110R UvrB mutant was selected for analysis due to its correspondence (in location according to the sequence alignment) to the R112H high prevalence TTD disease-causing mutant and location in the UvrB ␤-hairpin proposed to participate in DNAbinding interactions (23,24,26,52). The Glu-110 to arginine or alanine UvrB mutations behaved the same as wild type protein in all assays and did not exhibit any deficiency in the ability to bind, cleave, or destabilize oligomeric DNA. Surprisingly, this mutation which lies outside the ATP binding domain leads to a significant increase in the UvrA/UV DNA-stimulated ATPase activity. Although this mutation was selected because it corresponded to the XPD R112H TTD-causing mutation, this region of the protein, however, shows little sequence conservation between the UvrB and XPD proteins. The ␤-hairpin turn has already had been proposed to be one of the structural loop regions of the XPD model predicted with lower reliability than the rest of the structure and where differences in behavior between UvrB and XPD were anticipated. Additionally, the positively charged arginine residue would be expected to exhibit different biochemical behavior than the negatively charged glutamic acid. The Arg-112 XPD mutation is a TTD disease-causing mutation unlike the others selected which are XP disease-causing mutations. It is possible because TTD is viewed as a disease of transcription and not NER that some other mechanism is involved in the TTD disease syndrome for which the bacterial NER enzyme UvrB is an inappropriate model (6).

CONCLUSIONS
The goal of these studies was to construct a molecular model of human DNA repair protein XPD based on the experimen-  tally determined crystal structure of the bacterial DNA repair protein, UvrB. Model building was validated by experimental verification of XPD mutations in corresponding residues in the UvrB protein. The XPD model helps to give structural insights into some of the mutations in the XPD gene which when mutated yield the XP phenotype. For example, one prediction of the XPD model is that Arg-683, which is located on the "backside" of XPD on domain 3, is part of a binding surface for interaction with p44. This residue is in close proximity to another mutation, D681N, which is the most common XPD mutation. These two mutations help define a p44 interaction site on XPD that is essential for XPD to be tightly bound to TFIIH (50). This XPD model in conjunction with the future biochemical analysis of these mutations will be useful formulating structure-function explanations for the various XPD mutations leading to TTD, XP, or CS phenotypes.
In summary, we found that XP disease-like mutations could be successfully mimicked experimentally in UvrB (D338N, H341A, and R506A) and had a greater impact on the activity of UvrB than a TTD disease-like mutation (E110A/E110R). This model, and the concept that XPD and UvrB both play a vital step in damage-strand verification, is supported by recent DNA cross-linking experiments with XPD (60). Reardon and Sancar (60) show that XPD is making intimate contact with a psoralen monoadduct. Similar studies have shown UvrB can also be cross-linked to psoralen (61)or bis-platinum adducts (62). This general approach, the development and validation of a structural model for other mammalian proteins, should be generally applicable where structural information exists for closely functional related bacterial, archaeal, or yeast proteins, which share significant protein structural motifs or a structural protein fold.