XPB and XPD helicases in TFIIH orchestrate DNA duplex opening and damage verification to coordinate repair with transcription and cell cycle via CAK kinase

Excision repair pathways in eukaryotes and bacteria are valuable comparative systems

A powerful understanding of eukaryotic NER came from the breakthrough ability to reconstitute the required proteins in vitro, so the key players of NER are known [5, 16, 17]. Furthermore, a key concept for understanding NER is the importance of sequential assembly rather than having all components present as a preassembled repairosome [18, 19]. NER that occurs globally in the genome (GG-NER) or during transcription (transcription coupled or TC-NER) differ only in their initial recognition of helix-distorting DNA damage by XPC either with or without DDB1/DDB2 for GC-NER and by RNA polymerase for TC-NER [20, 21] (Fig. 1 and considered in more detail below). Following initial damage recognition, XPA, RPA, XPG, and the TFIIH complex including the XPB and XPD helicases are recruited. TFIIH is a ten subunit complex (containing XPB, p62, p52, p44, p34, TTDA (p8), and XPD) plus the Cdk-activating-kinase (CAK) (cdk7, cyclinH and MAT1) [22]. XPB, with 3’-5’ polarity, and XPD, with 5’-3’ polarity, open the DNA helix to form a 27-nucleotide bubble asymmetrically flanking the damage (22 nts in 5’ and 5 nts in 3’) [23], verify the damage [24, 25], and interact with the CAK kinase to signal transcription and repair status [26, 27]. Following helix opening, XPF in complex with ERCC1 makes an incision 5’ to the damage, and the structure-specific endonuclease XPG makes an incision 3’ to the damage [28]. This dual incision by XP nucleases leads to the removal of ~30 nucleotides (nts) resulting in a single strand gap [29]. This gap, which is the same size for GG-NER and TC-NER, is filled by DNA polymerase δ and κ, or ε [30], which use the undamaged strand as a template, and rejoined by ligase. Although the patch is relatively short compared to DNA replication, proliferating cell nuclear antigen (PCNA) coordinates DNA synthesis and ligation and replication protein A (RPA) protects the ssDNA strand from degradation. This mechanism of eukaryotic NER roughly resembles excision repair in prokaryotes—removal of diverse bulky and helix-distorting lesions from DNA by the coordinated actions of damage detection, helix opening, and gap excision followed by repair synthesis (Fig. 1). In fact, functional similarities between NER in bacteria and eukaryotes prompt us to herein compare the structural biochemistry for the key helicases in bacteria and humans.

In prokaryotes, UvrA, UvrB and UvrC orchestrate the NER damage recognition and incision steps (Fig. 1) [31, 32]. UvrA, as of an A2B or A2B2 complex, scans DNA to recognize defects, allowing UvrB to find and verify the damage site. ATP hydrolysis is required for UvrA disassociation and UvrB becomes stably bound to the lesion promoting local destabilization of the dsDNA. In this complex, UvrB C-terminal domain is exposed to form a UvrBC-complex that makes dual incisions first on the 3’-side and then on the 5’-side of the lesion. UvrB evidently employs its ATPase activity to place the DNA in a strained conformation suitable to be recognized and incised by UvrC. The UvrD helicase unwinds the 12-nucleotide tract containing the lesion with 3’-5’ polarity after the incisions have occurred. Pol I then fills the gap and removes UvrB from the DNA and ligase seals the gap. The structures and biochemistry of these prokaryotic systems provide valuable comparative results for a mechanistic foundation critical for understanding and interventions.

NER diseases, cancer, and aging

Tragic failures in NER are associated with extreme sun sensitivity, cancer risks, neurodegeneration, and premature aging; however, these defects also provide deep insights into cell biology and hope for future knowledge-based interventions [13, 33-35]. XP helicase mutations associated with disease evidently result in deficiencies of the TFIIH complex, and disease phenotypes are complicated by TFIIH’s dual role in both NER and transcription [36, 37]. Emerging roles for XPD that are independent of TFIIH further complicate but enlighten our understanding of human disease [38]. However, considerable insight into human NER processes has come from examination of multiple rare, autosomal disorders that result from defects in NER recognition, duplex opening and excision processes: xeroderma pigmentosum (XP), Cockayne Syndrome (CS), trichothiodystrophy (TTD), and cerebro-oculo-facio-skeletal syndrome (COFS) [34, 39-42] (see also articles by Diderich et al. and Gregg et al. in this issue of DNA Repair). XP is characterized by solar hypersensitivity, extreme predisposition for cancers on areas exposed to sunlight, and neurological abnormalities in some patients. XP can be caused by mutations in one of eight genes (XPA-G and XPV). Most CS patients have defects in the CSA or CSB genes, but in rare cases, some XP-B, XP-D, and XP-G patients have XP combined with CS. CS patients also have sun sensitivity, but without increased skin cancer risk. Also in contrast to XP, CS patients have reduced stature, mental retardation, and segmental aging features such as cachexia (loss of subcutaneous fat tissue), sensorineural deafness, retinal degeneration, and calcification of the central nervous system [39]. COFS shares clinical similarities with CS but is of prenatal onset [41] and is caused by defects in XPG, XPD, or CSB. Defects in XPB, XPD, and TTDA (p8) cause TTD, which has the hallmark characteristic of sulfur-deficient brittle hair that shows a tiger tail pattern under polarized light. TTD patients may have sun sensitivity but do not have increased cancer risk and have short stature, progressive cognitive impairment, and abnormal face shape [39]. Like CS, TTD is considered a segmental premature aging syndrome. Although seen in some patients, premature aging features such as reduced lifespan, cachexia, osteoporosis, kyphosis (hunchback posture), and early graying are most clearly seen in TTD mouse models [43]. In extremely rare cases, mutations in XPD can cause XP combined with TTD [44].

Eukaryotic NER presents puzzles for structural biology

GG-NER is most efficient at targeting bulky lesions that disrupt or distort the DNA double helix, such as pyrimidine dimers (cyclobutane pyrimidine dimers and 6-4 photoproducts) caused by the UV component of sunlight [45]. Yet, NER can also be initiated during transcription by stalled RNAPII [46] or even by other proteins such as ATL that recognizes and stays bound to alkylation damage [47], thus expanding the spectrum of lesions that NER can act upon. In fact, TC-NER can repair abasic (AP) sites generated by glycosylases [48] as the RNA polymerase is arrested by AP sites on the transcribed strand [49]. Other NER substrates include bulky chemical adducts, oxidative damage [50, 51], and DNA intra-strand cross-links [52, 53]. Nevertheless the hallmark of lesions preferentially recognized by NER is that they cause both a modification of DNA chemistry and a distortion of DNA double helical structure, which places special requirements on the structural biochemistry of NER damage recognition and opening for excision. These initial NER steps present six key puzzles for the structural biology that we consider throughout this review: 1) how damage is recognized, 2) why the need for two SF2 helicases (XPB and XPD) to open a bubble around the damage, 3) why the repair bubble is 27 nt and asymmetric to the lesion 4) how damage is verified, 5) how the presence of this damage is signaled to downstream processes, and 6) how repair enzyme products are handed off to avoid release of toxic and mutagenic intermediates. We address these puzzles by integrating structural, cellular, and biochemical results into a unified bind-pry-unwind model for the initial steps of NER.

Damage recognition – insights from XPC, DDB, RNAPII, and ATL

Damage recognition for NER involves proteins that bind to either the damaged site or to the strand opposite the damage (Fig.1). Different proteins act in the recognition of the DNA damage in GG-NER and TC-NER. In GG-NER, the XPC-RAD23B complex acts in distortion recognition, and DDB1 and DDB2 (XPE) can also recognize lesions caused by UV light [54]. In TC-NER, RNA polymerase II acts in initial damage recognition instead of XPC-RAD23B. If RNA polymerase II encounters damage and is arrested, CSB recognizes the stalled polymerase and recruits CSA, chromatin remodelers, and NER proteins including XPG and TFIIH for efficient repair [20]. In vitro evidence suggests that recognition of the stalled polymerase is coordinated by both CSB and XPG [55]. Recent evidence suggests that NER factors may be recruited to certain kinds of weakly distorting lesions such as O⁶-alkylG by the presence of bound proteins such as ATL [47, 56].

The identification of XPC as one of the earliest acting GG-NER factors gives insights into the order of NER component actions [18]. XPC likely acts as the initial sensor for unpaired bases for the global NER pathway. XPC forms a heterotrimeric complex with RAD23B and centrin 2 that binds NER-specific lesions, bubbles, and loop structures. DNA is continually scanned for lesions that cause helix distortions by XPC-RAD23B [57] and by the UV-DNA-damage binding (UV-DDB or DDB1-DDB2) complex [58, 59]. The crystal structure of the yeast XPC orthologue Rad4 with yeast Rad23 bound to DNA containing a single CPD lesion surprisingly revealed that Rad4/Rad23 do not bind the damaged strand, but recognize local destabilization of base pairing and inserts a hairpin motif into the DNA helix [60] (Fig. 1). This insertion causes the two damaged nucleotides to flip out of the double helix in a conformation that places the CPD lesion too far from Rad4 to make any direct contacts and is disordered in the crystal structure [60]. This is distinct from base flipping in BER by DNA glycosylases that flips a single base into the enzyme active site [61, 62]. Rather, Rad4 binds the DNA through three beta hairpin domains (BHD1-3) and a transglutaminase-homology domain (TGD). BHD1 and TGD bind an 11-bp region of fully paired, undamaged double-stranded DNA. Several nucleotides away, BHD3 inserts a β-hairpin feature through the Watson-Crick helix to flip the damaged nucleotides out, while a groove in BHD2 and BHD3 holds the corresponding undamaged nucleotides like the palm of a hand [60]. By binding the strand opposite the lesion, Rad4 (XPC) can accommodate the wide range of bulky lesions that are targets of NER (for a review focusing on the damage recognition in NER see review by Naegeli and Sugasawa in this issue of DNA Repair).

In contrast to Rad4/Rad23, structures of DDB1–DDB2 bound to DNA containing a 6-4 pyrimidine-pyrimidone photodimer (6-4PP) lesion or an abasic site mimic show that DDB2 mainly binds the damaged strand (Fig. 1) and flips the damaged bases into a shallow binding pocket [63]. DDB2 (XPE) is a WD40 repeat β propeller protein and the full diameter of the β propeller is used to bind DNA [63]. The larger DDB1 protein does not participate in DNA binding and binds the DDB2 β propeller on the side opposite from the DNA. DDB2 inserts a hairpin into the minor groove, extrudes the photodimer into a binding pocket, and kinks the duplex. DDB2 also stabilizes the nucleotides opposite the damage resulting in a slight unwinding of the DNA around the damage [63]. The localized probing of photolesions combined with the proofreading by the photodimer pocket, allows DDB2 to detect CPD lesions that make smaller perturbations in DNA than most bulky lesions and explains the more limited range of lesions that DDB2 binds versus XPC [63]. In BER, structures of the Endonuclease V–DNA complex reveal a wedge of four residues (PYIP) that inserts into the minor groove of the helix. DDB2 [64] flips the photodimer somewhere between the 90° rotation of the base lesion in EndoV and the precision extrusion of a single nucleotide of almost 180° by DNA glycosylases [65].

In some bacteria and eukaryotes alkyltransferase-like (ATL) proteins provide a non-classical means to initiate NER. ATL protects DNA from alkylation damage, and shares functional motifs with chemotherapy target O⁶-alkylguanine DNA-alkyltransferase [47, 66, 67]. Yet, ATLs lack alkyltransferase activity but employ non-enzymatic nucleotide flipping plus DNA distortion to convert an NER invisible lesion into a visible one by forming a stable, distorted ATL-DNA base lesion complex suitable for functional interaction with NER proteins. DNA-protein contacts in ATL1-DNA complex structures are with the damaged strand (Fig. 1), analogously to DDB2, but would allow simultaneous binding by XPC-RAD23B which binds the opposite undamaged DNA strand [60]. These results suggest that the functional connection of base damage to NER requires only two features: 1) stable specific binding to base damage, and 2) the ability to interact functionally with NER excision partners, such as ERCC1-XPF and XPG. In fact, the ATL C-terminal loop resembles the loop that recruits the nuclease heterodimer XPF-ERCC1 in XPA, the classical NER damage binding protein [47].

During transcription, damage on the transcribed strand stalls the RNA polymerase and initiates TC-NER that preferentially repairs the transcribed strand of active genes over bulk DNA [46]. Crystal structures of yeast RNA pol II with four different CPD-containing DNA and RNA scaffolds reveal that the polymerase stalls after nucleotide incorporation opposite both CPD thymines such that the damage is found deep within the protein [68] (Fig. 1). Surprisingly, the conformation of the polymerase is the same in all of the CPD-containing structures solved, arguing that the signal for TC-NER is not an altered conformation of the polymerase itself. However, the downstream DNA was found in different positions in two of the RNAPII-CPD structures. Since XPB binds the downstream DNA during transcription initiation in both human and yeast systems [69, 70], XPB may recognize this altered downstream DNA conformation during repair and position TFIIH [68]. Furthermore, Kim et al. [69] proposed that XPB promotes DNA melting and promoter escape by rotating DNA downstream of the transcription bubble relative to the upstream DNA that is rotationally fixed by transcription factors and RNAPII. We propose to extend this model to DNA repair by suggesting that the stalled RNAPII in TC-NER and XPC-RAD23B (aided by DDB1-DDB2 or ATL when needed) in GG-NER serve to rotationally fix the DNA so that XPB can wedge open the DNA and allow NER pathway progression. Therefore, DNA damage recognition proteins not only recruit TFIIH through protein-protein interactions [71, 72], but they also provide DNA binding activity that promotes DNA opening by XPB, ensuring that DNA opening is properly linked to damage recognition.

Rad51/RecA domains provide a unifying structure for NER helicases

DNA helicases are macromolecular motors that catalyze the separation of the DNA duplex by moving along DNA powered by ATP binding and hydrolysis. Structurally, NER helicases are united by their conserved Rad51/RecA-like ATPase domains [73-75], which contain the ATPase and helicase motifs that are characteristic of superfamily 2 helicases. Helicases have seven conserved “helicase motifs” (Walker motif I, Ia, II, III, IV, V and VI) [76] and are grouped into three superfamilies (SF), SF1, SF2 and SF3, based on similarity and organization of these conserved motifs [77]. That XPB and XPD are both SF2 family helicases creates puzzles regarding why two SF2 helicases are required in TFIIH (For a review on functional aspects of TFIIH, see review by Egly and Coin in this issue of DNA Repair).

Crystal structures of Archaeoglobus fulgidus XPB (AfXPB) and archaeal XPDs reveal that in XPB and XPD, these Rad51/RecA domains, connected by a flexible hinge, share an α-β fold and are called helicase domains 1 and 2 (HD1 and HD2) [78-81](Fig. 2). HD1 and HD2 pack against each other to form an interface cleft that brings together motifs I, II, V, and VI to form a composite ATP-binding site. This architecture is consistent with an inchworm model where cycles of ATP binding, hydrolysis, and release are coupled to opening and closing of HD1 and HD2, which drives the enzyme along DNA [82]. The structural relationships of such motifs involved in converting ATP binding to conformation change are critical; for example, the structural placement of the signature motif for the ABC ATPases provided key understanding of their structure-function relationships [83]. Crystal structures of Bacillus caldotenax UvrB and Escherichia coli UvrD reveal that UvrB and UvrD also have Rad51/RecA domains with α-β folds [84-86] (Fig. 2). In all of the NER helicases, the helicase domains that drive helix unwinding are not sufficient for their biological roles in opening DNA bubbles for excision. Rather they require the presence of accessory domains that play roles in duplex splitting or damage recognition (Fig. 2).

Accessory domains have diverse structures to enhance specific functions

All of the NER helicases have two functional parts: two helicase domains that use ATP binding and hydrolysis to drive conformational change and two accessory domains that help transmit helicase domain changes to the DNA and either interrupt the linear sequence of the helicase domains or are found at the N- or C-termini (Fig. 2). Their positions and structures are united only by their diversity.

In XPD, the most striking feature that emerges from first examining the new structures is that HD1 is interrupted by two additional domains, a domain containing a [4Fe-4S] cluster and an Arch-shaped domain [79] (Fig. 2). The 4FeS domains in DNA repair proteins were originally characterized in BER enzymes endonuclease III and MutY [87, 88] and identified as a DNA binding FRCL motif [89]. Despite the recognition that many proteins contain metal ions for functions [90], the discovery of the 4FeS domains in XPD, the Zn hook in Rad50, and Ni in superoxide dismutase (SOD) highlight the fact that metal sites not only play critical roles in most cellular pathways but are still often poorly characterized or even unknown [15, 91-93]. Metal ions can play direct roles in DNA binding, DNA conformational change, as well as catalysis, as seen for FEN1-DNA complexes [94]. Furthermore, loss of metal ions can promote degenerative diseases as implicated for human SOD [95]. Moreover, growing evidence supports the role of 4FeS clusters in DNA damage detection and in possible electron transfer along DNA [96, 97]. Given the emerging role of metals in diverse cellular processes, it will be important to more fully understand metals in NER.

In NER, the 4FeS domain is distinct to XPD and XPD-like helicases including FancJ and is structurally defined by the presence of a [4Fe-4S] cluster [79]. Sulfolobus acidocaldarius (SaXPD) and Thermoplasma acidophilum (TaXPD) 4FeS domains share the same overall fold with slight differences in the number of helices versus loops. In SaXPD, the 4FeS domain contains four helices connected by loops containing four Cys residues that stabilize the 4 Fe ions of the cluster [79]. The TaXPD FeS domain has six helices and three of four Fe-coordinating Cys residues are located in loops with the fourth Cys located in the middle of a helix [81]. Although the structure of Sulfolobus tokodaii (StoXPD) could be solved without an ordered 4FeS domain [80], the [4Fe-4S] cluster of SaXPD provides structural stability for the 4FeS, Arch, and HD1 domains, as chemical oxidation of the cluster in the crystal caused not only the 4FeS domain to become disordered but also part of the Arch domain and residues at the N-terminus [79]. The Arch domain forms a small interface with the 4FeS domain to form a small tunnel of between 8 and 13 Å that is large enough for ssDNA to pass [79, 81, 98] (Fig.2). The Arch represents a novel fold that consists of either a three- or four-stranded β sheet with either four or five helices, depending on the organism [79-81]. The 4FeS and Arch domains are both closely linked to HD1 by β sheet linkages. Due to the interwoven nature of the main chain hydrogen-bonding pattern, β linkages provide interstrand stability not possible with α helices. The 4FeS domain is inserted into adjacent β strands of the central HD1 β sheet while the β sheet of the Arch domain bridges between HD1 and Arch domain α helices. By physically interrupting the linear sequence of HD1, these accessory domains are closely tied to the enzyme’s ATP state, providing a mechanism linking accessory domain function to ATP binding and hydrolysis.

UvrD has an insertion into each of its helicase domains (Fig.2). Domain 1B inserts into helicase domain 1A and domain 2B inserts into helicase domain 2A [84]. Like XPD, both domains 1B and 2B are linked to the helicase domains by β strand linkages, providing a strong structural link between ATP state and the accessory domains. Both domains are inserted into the central β sheets of their respective helicase domain.

By linear sequence, the bacterial UvrB protein seems to share a similar overall architecture with XPD--two helicase domains (1a and 3) and two accessory domains (1b, 2) that are insertions in domain 1a [85] (Fig. 2). Although the two helicase domains both have α/β folds and pack up against one another with the ATP binding site shared between them, the accessory domains are topographically different from XPD. Domains 2 and 1b insert differently into domain 1a. The polypeptide chain leaves domain 1a, forms domain 2, then part of domain 1b before diving back into 1a, and then finishing domain 1b. Neither domain 2 nor 1b contact domain 3 and are on the opposite side of the protein (Fig. 2). These differences in accessory domain connections suggest that there are different ways to connect the ATP-driven actions of the helicase domains in opening DNA duplex to recognizing damage and holding open a DNA bubble (for a discussion of how differences in UvrB and XPB/XPD may affect the substrate specificity in NER, see article by Liu et al in this issue of DNA Repair). These differences and similarities imply a degree of convergent evolution for the bacterial helicases versus XPB and XPD, where domain differences may reflect alternative means to achieve similar functions and differences in their protein partners.

In XPB, the helical ThM domain is an insertion in HD2 while the DRD domain is N-terminal to HD1 (Fig. 2). The XPB ThM domain structurally resembles thumb domains of both T7 and Taq DNA polymerases that grip the minor groove of dsDNA, suggesting a DNA binding role for this domain in XPB [78]. The XPB DRD domain contains five β sheets connected by a single α helix and has structural similarity to the mismatch recognition domain of MutS, suggesting a damage recognition role for XPB [78]. Furthermore, AfXPB unwinds blunt-ended DNA containing a single UV lesion, suggesting that AfXPB can functionally interact with damaged DNA and therefore may play a role in damage verification during NER [78].

DNA binding and conformational change

Surprisingly, the NER helicases are also divergent in the structural elements used to bind DNA and whether or not DNA binding induces large structural changes. Structures of UvrB have been solved both with and without DNA, revealing that the DNA is threaded around a critical β-hairpin motif while the conformation of the rest of the protein is unchanged [86]. In contrast, UvrD undergoes a substantial domain swiveling upon DNA binding and uses a diverse set of motifs to make contacts with the DNA [84, 99]. The crystal structure of UvrB was solved with single-stranded DNA that fortuitously formed a hairpin with 3-base pair (bp) of dsDNA and a 3-bp 3’ overhang. The dsDNA is bound by domains 1a and 1b while the 3’ overhang threads behind the β-hairpin to also make contacts with domains 1a and 1b. UvrB bound to DNA has the same overall conformation as unbound UvrB with the exception of the base of the β-hairpin. The base of the β-hairpin consists of four Tyr residues (92, 93, 95, and 96 in BcUvrB) that make contact with the DNA. Tyrosine hydrogen bonding and aromatic ring structures can provide critical interaction specificity, as seen for SOD [100, 101]. In BcUvrB, Tyr96 makes π stack interactions with the first base of the 3’ overhang at the ss-ds junction [86]. Tyr96Ala mutants cannot form a stable pre-incision complex on cholesterol-containing DNA suggesting that this absolutely conserved residue stabilizes UvrB on DNA [102]. DNA filter binding assays reveal that a double Y92A/Y93A mutant does not discriminate between undamaged and damaged DNA like the wild-type protein [103]. These results imply a stacking role for Tyr in damage recognition that remains to be mechanistically defined.

Crystal structures of UvrD with DNA show that UvrD adopts a closed conformation as seen with PcrA-DNA complexes [104]. This conformation is called closed based on the two different conformations, open and closed, observed in crystal structures of the Rep helicase with DNA [99]. An open to closed conformational change is a common feature of damaged DNA binding, as first seen in UDG during BER [105, 106] and can direct macromolecular pathways and outcomes as shown for example by RFC-PCNA interactions [107]. However the open to closed conformations for NER can be much larger. The difference between the open and closed conformations in UvrD is a 130° swivel of the 2B domain to close it against the 1B domain. Although the open conformation was not observed in the crystal structures of UvrD with DNA, biochemical data supports the idea that both conformations are active for DNA unwinding, albeit by different mechanisms [84]. The closed conformation is proposed to unwind DNA by a “wrench-and-inchworm” mechanism where first the GIG motif of domain 2B binds dsDNA and motif III of domain 1A anchors the ssDNA allowing 1 bp to unwind. ATP hydrolysis and release then drives ssDNA translocation to unwind another bp. The open conformation may unwind DNA by a “strand displacement” mechanism that only requires ssDNA translocation [84].

A computational model of XPB bound to DNA similarly proposes a large conformational rotation of 170° but this change is entirely different from the UvrD swivel. In UvrD, the domain swivel brings two accessory domains together to change the mechanism of DNA unwinding. In XPB, the rotation of the hinge is necessary to bring the two helicase domains together to form the composite ATP binding site [78] (Fig. 3). An initial DNA binding event by HD1 and the DRD domain likely induces this rotation, while ATP hydrolysis facilitates the positioning of the RED motif and ThM domain to stabilize XPB on DNA [78]. This hypothesis was tested in vivo using both host-cell reactivation and UV-survival assays in CHO cells [108]. Mutations in either the ThM domain, RED motif, or the Walker A ATPase motif were defective in these assays, establishing a critical role for these domains and motifs in DNA repair [108]. Furthermore, GFP tagged mutants were defective in recruitment to sites of UV damage in CHO cells, suggesting that recruitment of XPB to damage is an active process requiring ATP hydrolysis coupled with structural stabilization of XPB on DNA by the ThM domain and RED motif [108]. In vitro experiments using mutant Sulfolobus solfataricus XPB (SsoXPB) proteins, found that the RED motif is involved but not essential for SsoXPB function while both the ThM and DRD domains were essential for XPB function [109], confirming the importance of these accessory domains and motifs for XPB function.

Although a crystal structure of XPD with DNA has not been published, three independent computational models of XPD from three different organisms agree on the path of ssDNA binding (Fig. 2). ssDNA passes through a hole formed by the Arch, FeS, and HD1 domains, travels through a basic channel across the top of HD2, and exits near a helical wedge in HD2 [79-81]. These models require conformational flexibility at the hinge between HD1 and HD2 and between the Arch and 4FeS domains to allow for DNA loading (Fig. 3). Conformational flexibility would also explain the different hole sizes observed in the SaXPD (8 Å) and TaXPD (13 Å) structures [98]. All of these models were based on the structure of Hel308 with DNA, which is another SF2 helicase [110]. In two cases, the model was supported by crystallographic evidence of electron density in the modeled ssDNA binding channel. In the SaXPD structure, bound glycerol, isopropanol, and citrate ions from the crystallization buffer appear to mimic the DNA backbone phosphate and sugar moieties [79], while residual bound DNA seems to explain extra electron density peaks in the TaXPD structure [81]. Patches of positively charged residues on both the flat back side and on HD2 suggest the location of dsDNA interactions on either side of the molecule. In SaXPD, the most energetically favorable docking sites for dsDNA were at the HD2-Arch domain junction where a helix-loop-helix on HD2 could act as a wedge to hold open dsDNA [79]. This wedge was also observed in the TaXPD structure [81]. Although the position is different, both TaXPD and SaXPD have basic patches on the flat back suggesting a second site for dsDNA. Such positively charged patches are hallmarks for DNA binding sites in DNA repair enzymes as seen in base-excision repair, mismatch repair and double-strand break repair enzyme-DNA complexes [62, 111-115]. Such electrostatic complementarity can increase the interaction kinetics and orient macromolecules for efficient interactions [116, 117]. In XPD, the positive patches suggest a configuration consistent with XPD interacting with both sides of a DNA bubble during NER and a means for XPD to efficiently interact with DNA within the TFIIH assembly.

Active site ATP binding and hydrolysis

ATP binding and hydrolysis is critical for the function of the NER helicases. For XPB, ATPase activity seems to be more critical for its functions in transcription and NER than helicase activity [108, 118, 119]. ATP is bound between the two Rad51/RecA helicase domains through several conserved motifs. Structures of both UvrB and UvrD have been solved with Mg²⁺ and ATP or ATP analogs, providing a detailed picture of ATP binding and hydrolysis.

UvrB and UvrD bind both the adenine base and the phosphate moieties. The Walker A motif (motif I) coordinates the phosphates in both proteins (Fig. 4). In BcUvrB, the phosphates are bound by hydrogen bonds from backbone nitrogens of residues T41, G42, T43, and the side-chain of K45 [85]. The analogous motif I Lys in EcUvrD is K35 and it also coordinates the phosphate moieties. In addition, residues R73, R284, and R605 also bind the triphosphate [84]. In both of these structures, the adenine is sandwiched between two residues, one of which is a tyrosine (Fig. 4). Y11 and P414 in BcUvrB position the N6 and N7 of adenine for hydrogen bonding while Y283 and R37 position the adenine in EcUvrD. The Walker B (motif II) Asp and Glu residues (D338 and E339 in BcUvrB, D220 and E221 in EcUvrD) as well as the ATP β- and γ-phosphates serve to coordinate the single Mg²⁺ ion necessary for catalysis (Fig. 4).

In XPB and XPD, Walker A (motif I) mutants disrupt cellular functions, sometimes leading to disease. K45 of BcUvrB and K35 of EcUvrD coordinate the ATP phosphate moieties. When the analogous K346 residue is mutated in XPB, this mutant cannot support transcription or DNA repair in vitro, nor can it properly localize to DNA damage in the cell [108]. A similar mutation in XPD, K48R, supports transcription in vitro and localization in vivo, but it does not support DNA repair in vitro [108]. A mutation in the adjacent residue, G47R, leads to XP/CS in humans, supporting the importance of this motif in DNA repair in humans [120]. The Walker B (motif II) Asp residue that is important in coordinating the Mg²⁺ ion in UvrB and UvrD is also critical in human XPD since patients with a D234N substitution have XP. Conserved mutants in SaXPD, G34R (G47R) or D180N (D234N) (Fig. 4), are defective in ATPase and helicase activity, but can still bind DNA strongly [79].

ATP-induced conformational changes and DNA opening

ATP-binding energy is stored as conformational changes in the helicase motifs and domains described above. More specifically, the result of γ-phosphate binding and hydrolysis drives piston-like shifts in helicase motifs and rotations in domains that push the DNA strand. ATP is stabilized by specific interactions within its binding site to achieve efficient hydrolysis. An excellent model for understanding this process comes from the structure of the Hel308 bound to DNA compared to other SF2 helicases [110]. During processive translocation in the ATP-free state, phosphates of the ssDNA backbone are bound at motif Ia (position +4) on HD1 and motif IV on HD2 (position +1) and are separated by 2 intermediary bases. However, in three independent structures of SF2 DEAD-box enzymes in complex with ATP analogs and single-stranded nucleic acids, phosphates bound at motifs Ia and IV are separated by only 1 intermediary base [121-123]. Thus the implication is that ATP binding promotes a conformational change that pushes a single nucleotide across HD1. In Hel308, this ATP-driven closure of HD1 and HD2 may be triggered by binding of the conserved motif VI arginine (R369 in Hel308; R514 in SaXPD; XP/CS mutant R666W in XPD) to the ATP γ-phosphate. Upon ATP binding, XPD HD1 motif 1a may push the DNA toward HD2 motif VI. HD1 movement also necessarily pushes on the inserted 4FeS and Arch domains via its covalent connections with these domains. These ATP-driven domain movements likely result in the different hole sizes observed in the SaXPD and TaXPD structures and suggests that the hole may act as a “gripper” during DNA translocation [98]. Thus, ATP binding and hydrolysis by HD1 and the HD1-HD2 interface cause ratchet-like conformational changes that drive the directional movement of the helicase on DNA.

ATP binding and conformational change involve conserved motifs and analogous HD1-HD2 opening and closing. Yet, ATP-driven conformational changes have different impacts in the different helicases, consistent with their distinct accessory domains, divergent structural elements used to bind DNA, and distinct roles. For example, the mechanism of strand separation differs among helicases and ranges from destabilizing dsDNA by mainly binding and limiting strand separation activity (UvrB, XPB), strand displacement by translocating along ssDNA (UvrD, XPD), and unwinding by making contacts with the DNA duplex (UvrD, XPD) [78, 84, 86, 124]. Structural comparisons among helicases including NS3 and Hel308 suggest the β-hairpin is the strand separation element for these helicases [110]. In the UvrB complex with DNA, the beta hairpin is implicated in both damage recognition and strand separation. In XPB, the RED motif is proposed to introduce a wedge in the double stranded DNA that is then gripped by the ThM domain [23, 78]. UvrD has been proposed to unwind DNA by both a “wrench-and-inchworm” mechanism that involves a separation pin motif and a strand displacement mechanism that only requires translocation along ssDNA with DNA being split around a domain rather than a pin [84]. XPD may unwind DNA by such a strand displacement mechanism with the 4FeS domain playing a key role in duplex splitting [79, 81, 125]. Mutations in XPD’s FeS domain, either at sulfur-coordinating cysteines or at an arginine residue mutated in TTD patients, abolish helicase activity, establishing a role for this domain in DNA unwinding [79, 126].

The differences in the proposed DNA unwinding mechanisms and properties of XPB and XPD provide clues as to the need for two helicases in TFIIH. XPB can open bubbles from dsDNA at sites with less stable DNA duplex [78]. This may aid XPD loading by localizing TFIIH at the damaged site. The DNA binding channel of XPD extends ~10nts whereas the binding channel of XPB is shorter and ~5 nts (Fig. 2). The arch and 4FeS domains regulate DNA access to the motor, so these domains limit ssDNA movement with the probability that 4FeS sensing of damage may play a role in arresting helicase movement [127]. Assuming XPB and XPD bind to different strands, the two helicases together give TFIIH a strong hold on the bubble for repair. This model predicts that XPD is positioned to interact at the end where XPG would cut and XPB is positioned to act at the end where XPF would cut. XPD product release therefore controls repair synthesis, as it has the larger DNA binding site. This suggests a strategy for intervention by blocking XPD release, which is discussed below.

CAK structural biology and XP helicase connections

CAK is both a pivotal activator of cyclin dependent kinases (CDKs) and a component of TFIIH [128, 129]. CAK is also required for the transactivation of transcription factors and the regulation of RNA Polymerase II. CAK is itself a complex consisting of Cdk7 (Cyclin dependent kinase 7), CycH and Mat1 [130]. When CAK is bound to TFIIH, it phosphorylates the C-terminal domain of RNAPII plus various transcription factors [131]. When CAK is dissociated from TFIIH, it phosphorylates CDK1, CDK2, CDK4, CDK6, and other targets [131, 132]. Indeed, the association and dissociation of CAK from TFIIH may be an important regulator of DNA repair. Chromatin immunoprecipitation experiments have recently shown that CAK is released from core TFIIH during NER and this release is catalyzed by XPA [27]. CAK release stimulates incision and repair of the damaged DNA, while CAK reassociation with TFIIH correlates with resumption of transcription [27]. Furthermore, the CAK kinase inhibitor H-8 improved repair efficiency, indicating that CAK can negatively regulate NER by phosphorylation [16]. Both XPG and XPD are important for the association of CAK with core TFIIH as mutations from XP-G/CS patients that prevent the XPG-TFIIH association cause both CAK and XPD to dissociate from TFIIH [133]. However, in XP-G and XP-G/CS cells, XPD recruitment to damage was normal while CAK was not [134], suggesting that XPG is not required for XPD-TFIIH association. XPD may play a role in coordinating CAK activity as XPD overexpression in Drosophila negatively regulates the cell cycle function of Cdk7 and downregulation of XPD results in increased CAK activity and cell proliferation [135, 136]. During transcription, the XPB helicase is required to open the promoter around the start site; in contrast, XPD helicase activity is dispensable but XPD stimulates transcription and anchors the CAK complex to TFIIH [137], supporting the importance of connections between XPB, XPD, and CAK. MAT1 activates Cdk1 kinase activity and both XPB and XPD interact directly with the central coiled-coiled domain of the Mat1 subunit of CAK [26]. We therefore propose that MAT1 may monitor the conformational state of both helicases in TFIIH.

Structural information is available for at least part of all of the CAK subunits. Human CDK7 exhibits a typical kinase fold resembling the inactive conformation of CDK2 with two lobes—an N-terminal lobe comprised of β sheets and a C-terminal lobe of mostly α helices—with ATP bound between the two lobes [138] (Fig. 5). Human cyclin H is a regulatory subunit of CDK7 and has two α helical domains, similar to cyclin A of the CDK2 complex [139, 140]. The N-terminal RING finger domain of human MAT1 was solved by NMR and exhibits a typical RING finger βαββ topology with two zinc-binding sites [141]. This domain is important for transcriptional activation and RNAPII CTD phosphorylation, while the central coiled-coiled domain interacts with XPB and XPD, and the hydrophobic C-terminal domain stimulates CDK7 kinase activity [26]. These structures, along with the structures of other TFIIH components, have allowed us to develop a model for how the ten members of TFIIH come together and consider what insights emerge from such a composite structural model.

The Tao of TFIIH structure: resolving puzzles for XP helicases and CAK functions

Lao Tzu’s Tao Te Ching (4^th Century B.C.) suggests deep insights come from considering paradoxes [142]. For TFIIH the need for two SF2 helicases, the conflicting evidence for their roles in transcription and damage verification, the mechanism for generating the asymmetric bubble flanking the DNA lesion, the mysterious connections between XPD and both incision and CAK signaling, and the opposing phenotypes of excess cell death and excess cancer from adjacent XPD mutations are among the paradoxes to consider. As key insights often emerge from combining electron microscopy (EM) results with those from high-resolution X-ray and NMR structures and detailed biochemical analysis [74, 143, 144], we use this approach here to address TFIIH paradoxes.

The breakthrough EM structure of human TFIIH showed that the TFIIH core components (XPB, p62, p52, p44, p34, TTDA, and XPD) fold into a ring structure with a bulge formed by the bound CAK [145]. Since high-resolution structural information is available for eight of the ten subunits, we now have a valuable opportunity to combine EM with atomic resolution results to propose a nearly complete TFIIH assembly (Fig.5). Computational docking of our archaeal XPB and XPD structures [78, 79] found the best fit for XPB to be integrated into the ring opposite the CAK bulge while XPD fitted best into the ring between XPB and CAK [79] (Fig. 5). As XPD helps bridge the CAK complex to core TFIIH, XPD is positioned to allosterically communicate TFIIH state to the CAK kinase, which can in turn signal to partners for transcription, repair, and cell cycle regulation. Both XPB and XPD interact with the coiled-coiled domain of MAT1 [26], so we have placed the coiled-coil to contact both XPD and XPB (Fig. 5). We have positioned the hydrophobic MAT1 C-terminal domain, for which there is no structural information, towards Cdk7/Cyclin H as yeast four-hybrid experiments mapped the interaction domain there [146]. Since CAK can negatively regulate XPD unwinding activity in vitro and addition of p44 can overcome this inhibition [146], we located p44 between XPD and CAK. Thus, we propose that the coiled-coil protein MAT1, which supports CAK interaction with XPD, may be analogous to those enforcing interactions for the NHEJ pathway [147] and in the WRN helicase-nuclease and its interactions with the DNA-PK kinase, where the coiled-coil and kinase interactions stimulate processivity [148]. Such coiled-coil interactions may provide robust allosteric connections from ATPases to nucleases and kinases in many repair complexes.

Moving around the TFIIH ring (Fig. 5), p52 is known to interact with XPB and stimulate its ATPase activity [118]. Both the N- and C-terminal domains of p52 interact with XPB, suggesting a large interaction region between these two proteins [118]. The tiny TTDA (p8) protein is also known to interact with p52 and participates in p52-XPB ATPase stimulation without directly interacting with XPB [149]. The crystal structure of yeast TTDA (Tfb5) was solved with the C-terminal domain of yeast p52 (Tfb2) revealing a pseudosymmetric heterodimer that protects a hydrophobic surface of p52, suggesting a possible explanation for why TTDA patient mutations destabilize TFIIH [150]. The p62 subunit interacts with many transcription factors and tumor suppressors including TFIIE, Rb, and p53 [151, 152] as well as NER factors XPC-RAD23B and XPG [71, 153]. The interaction of XPG and p62 is mediated by the N-terminal domain of p62 that is not required for TFIIH assembly [154]. The NMR structure of the p62 N-terminal domain revealed a β-sandwich fold with seven antiparallel β strands capped by a single α helix that has structural similarity to pleckstrin homology (PH) domain found in many cytoplasmic signaling proteins [154]. Structural information is not available for p34 nor for the XPB and XPD domains that are unique to humans [78, 79].

It is important to consider that TFIIH is not static, but rather a dynamic complex where individual components shuttle in and out of the complex or undergo large conformational changes. As discussed above, the disassociation and association of the CAK complex is an important regulator of DNA repair—CAK disassociation promotes NER and downstream signaling while reassociation of CAK correlates with the resumption of transcription [27]. XPD and TTDA may also shuttle in and out of TFIIH since two different populations of XPD and TTDA have been found in cells using GFP-tagging and fluorescence recovery of photo-bleaching (FRAP) methods: a slow moving population with dynamics similar to XPB that is associated with TFIIH and another faster moving population that is thought to represent an unbound pool of these proteins [155, 156]. After UV-irradiation, both XPD and TTDA have similar dynamics to XPB, suggesting that they become more stably integrated into TFIIH during DNA repair [155]. TFIIH must also accommodate conformational changes during transcription and repair. XPB DNA binding and ATP hydrolysis promote a large conformational change that is necessary to stabilize XPB, and therefore TFIIH, on DNA [78, 108]. Furthermore, both XPB and XPD are proposed to change conformation during ATP hydrolysis and DNA translocation (Fig. 3). Therefore, accessory factors such as p52, p44, or MAT1 may act as scaffolds to maintain TFIIH architecture while catalytic domains or subunits swing or shuttle in and out of TFIIH. Flexible domains found in human but not archaeal XPB and XPD [78, 79] may act as flexible tethers that allow for catalytic movements while maintaining TFIIH connections. Such flexible tethers are seen in other DNA repair assemblies such as the Mre11-Rad50-Nbs1 complex [157, 158] or Rad51 filament interactions [74] and can pull to effect multi-protein and multi-domain allosteric signaling in response to changes such as kinase phosphorylation [159, 160]. We expect that advances in technologies such as small angle X-ray scattering (SAXS) that allow for analyses of dynamic structures in solution including flexible domains and tethers [161-165] will provide dynamic models of TFIIH that further our understanding of TFIIH function.

Consideration of overall TFIIH architecture combined with known high-resolution structures allowed us to generate a composite model that provides many insights for a unified framework to resolve TFIIH puzzles and paradoxes. We suggest that the TFIIH core can usefully be considered primarily in terms of the XP helicases and the CAK kinase enzyme conformations and activities. Other components can be viewed as scaffolding and connectors; these act analogously to the helicase accessory domains, which provide scaffolding and mechanical coupling of ATP-states to protein and DNA interactions. In these terms, allosteric changes of the enzyme components are transferred throughout TFIIH via connecting domains such that pathway coordination and biological outcomes reflect the multi-protein allostery of TFIIH, which is driven by the enzymes and their interactions with substrates and products. This more detailed view of TFIIH as a dynamic, multi-protein, allosteric machine connecting helicases and signaling kinase with XPG and XPF partner nucleases provides a foundation for a unified model for TFIIH mechanisms and functions in NER. Testable hypotheses for the integrated functions of dynamic multi-protein systems, such as TFIIH, are often best and most efficiently defined in terms of such models.

A unified model for NER

The three enzyme components of TFIIH are XPB, XPD, and CAK: these proteins are essential for the coordinated actions of damage recognition, verification, repair and signaling. Based upon the combined structural biochemistry and genetics for these enzymes, we propose here a unified, testable model of their functions in the context of TFIIH (Fig. 6). Furthermore by analogy with the MRN complex in double-strand break repair [166], these three TFIIH enzymes can be considered a keystone complex connecting damage detection, signaling, and repair. We call this a bind-pry-unwind model because it integrates DNA damage recognition (bind) with DNA opening (pry and unwind by XPB and XPD) and solves the puzzles that NER presents for structural biology we outlined at the beginning of this review.

XPB engagement at promoters with its 3’ > 5’ helicase activity helps to initiate transcription (Fig. 6). XPB helicase activity (without XPD helicase activity per se but stimulated by XPD) is a critical signal for transcription: CAK remains bound to TFIIH and phosphorylates the RNAPII C-terminal domain. For DNA repair, the initial recognition of damage is by XPC-RAD23B (with or without DDB1-DDB2 or ATL) for GG-NER or RNAPII for TC-NER (Fig. 6). The damage recognition proteins not only recruit TFIIH and other downstream factors, but they also serve to rotationally fix the DNA, which we predict facilitates DNA opening by XPB analogous to the role of XPB in promoter melting and escape in transcription [69]. In this model, XPB acts as a molecular pry bar that works against the grip provided by the damage recognition proteins, thus linking damage recognition to DNA opening. This is consistent with the limited unwinding activity of XPB and its large conformational change upon DNA binding and ATP hydrolysis. Furthermore, it allows TFIIH loading directly adjacent to either XPC-RAD23B or RNAPII (during transcription initiation or stalled at damage), thus providing a shared XPB mechanism for TC-, GG-NER, and transcription. Like UvrB, XPB has a short DNA binding channel (Fig. 2) that when bound would position TFIIH perpendicular to the DNA (Fig. 6), either 5’ to XPC-RAD23B during GG-NER or downstream of RNAPII during TC-NER or transcription.

After TFIIH loading, XPB would then pry open the DNA, allowing TFIIH to tilt against the DNA and positioning XPD to bind the damaged strand 5’ to the lesion, approximately 22 nts from the lesion as defined by TFIIH architecture (Fig. 6). This binding event is monitored by MAT1 of CAK, which signals to CAK to release from TFIIH, allowing repair to proceed. XPD helicase unwinds the dsDNA until it is stopped by chemical damage reaching the 4FeS domain gateway [24, 25] (Fig. 2). We therefore propose that the span of the TFIIH ring from XPB to XPD sets the size and asymmetry of the NER bubble, as it creates the correct spacing to create 27nt bubbles (22 nts 5’ and 5 nts 3’ to the damage). After its arrest, XPD, together with XPB, anchor TFIIH and the associated NER-damage-response machinery to the repair site, providing an unambiguous platform for nuclease binding and damaged strand excision. XPD bound to the damaged DNA interacts with XPG and promotes DNA incision by XPG and XPF. Finally, removal of TFIIH is necessary for re-synthesis of the incised damaged stand and reassociation of CAK with TFIIH is necessary for resumption of cellular processes such as transcription.

After XPB binding, XPD is the anchor point for TFIIH on the bubble, so controlling XPD conformation and activity controls pathway outcome for NER. Combined structural and biochemical analyses provide a molecular framework that begins to explain how different XPD mutations lead to three different diseases, XP, XP/CS or TTD. As XP mutations (Fig. 4C red spheres) cluster along the DNA- and ATP-binding channels, they decrease helicase activity. Besides losing helicase activity as expected for XP mutations, XP/CS mutations (Fig. 4C gold spheres) also replace small flexible amino acids with larger ones and are predicted to reduce conformational flexibility. These mutations are largely found at the HD1-HD2 interface and therefore affect the conformational crosstalk between these domains to impact DNA-binding and helicase activities. TTD mutations (Fig. 4C purple spheres) are found throughout the protein and disrupt stabilizing interactions between residue side-chains and/or protein main-chains. These changes disrupt the XPD framework to reduce its structural integrity, which can affect XPD activities and/or partner interactions.

Experiments in S. cerevisiae XPD are also yielding insights into XPD disease. Surprisingly, the rad3-102 mutant (Fig. 4C blue sphere) exhibits only moderate sensitivity to UV damage, but rather causes replication-dependent double strand breaks (DSBs) that require the Mre11-Rad50-Xrs2 system for repair [167]. Evidently NER incision occurs in these cells, but postincision events are blocked due to retention of TFIIH at the site of damage, leaving a ssDNA gap. Arrival of the replication fork at the ssDNA gap causes replication fork breakage, which leads to the formation of one-ended DSBs [168]. Such one-ended breaks cannot be processed for NHEJ by the DNA-PK pathway [169] but must be processed by Mre11 complex-dependent recombination [114]. This yeast mutant evokes comparison to an XP-CS mutant (G675R) that causes aberrant DNA breaks upon UV irradiation in human cells [170]. In SaXPD, this mutant (C523R) has increased ssDNA-binding affinity [79], suggesting that increased XPD binding to DNA may be playing a role in the phenotype of these patients.

These integrated structural and biochemical insights therefore provide a framework to begin to predict why differences in single amino acids in one protein result in large phenotypic changes for an entire organism. Mutations that affect activity without affecting conformation or signaling lead to XP and cancer (Fig. 6). XP-D patients have reduced DNA repair capacity and thus cannot repair all of the damage that occurs on sunlight-exposed areas of the skin, leading to gene mutations and cancer. XP/CS patients have the same cancer susceptibility as XP patients with the added complication of having a mutation that reduces the functionally critical flexibility of XPD. This conformational restriction is predicted to alter XPD functions, partnerships, and downstream events, such a signaling to CAK (Fig. 6). This change thus compromises the cell’s ability to survive, leading to cell death premature aging [171]. TTD mutations also lead to cell death and premature aging since TTD mutations increase protein flexibility, which should both disrupt XPD interactions with other proteins and decrease stability of the TFIIH complex [172]. These changes also block downstream events (Fig. 6) despite sometimes retaining XPD activities [79].

Although specific aspects will likely need modification, this structure-based model has seven attributes we hope will prove valuable: 1) it integrates the structural biochemistry and genetics for TFIIH and its components into a single unified model, 2) it helps researchers worldwide design experiments to test separate enzymatic functions of TFIIH, 3) it provides a mechanism for creating the asymmetric 27nt bubble for repair, 4) it suggests the molecular basis for aging and cancer-causing TFIIH defects, 6) it identifies specific means and targets for therapeutic interventions, and 7) it suggests how the NER-damage response is orchestrated by the three TFIIH enzymes to connect detection of double-helix defects and chemical modifications by the XP helicases to CAK-mediated signaling.

When viewed in terms of individual TFIIH interactions, a bewildering matrix of combinations confronts the researcher. In general, the unified model proposed here (Fig. 6) implies TFIIH ring activities are controlled by allosteric conformations acting in kaleidoscopic effects that create and break specific assemblies and contacts depending upon TFIIH enzyme conformations. For example, the unequal opening flanking the DNA damage has been a mechanistic puzzle. Our TFIIH NER model suggests a simple mechanism for generating the asymmetric bubble that is entirely based upon the TFIIH architecture and the differential XPB and XPD helicase structural biochemistry. We trust the proposed unified model is sufficiently specific to allow testing of its various aspects by researchers who can bring to bear their expertise, systems, and technologies to test its merits.

Implications -- strategies for interventions

Inhibiting repair can cause genome instability [173]; yet, if done strategically then selectively inhibiting repair may allow improved interventions [174]. Indeed creating or blocking a specific protein interaction can control repair progression as shown for Rad51-BRCA2 interactions [74]. For new approaches in cancer interventions, the well-regulated NER pathway may provide a powerful strategy to target the Achilles’ heel of many cancers--defects in DNA repair and cell cycle regulation. Small molecule inhibitors that block XPD release, analogous to the yeast rad3-102 mutant that blocks NER and causes DSBs [167, 168], could potentially work for interventions in some cancers because to process such breaks requires the double-strand break machinery mutated in these cancers, such as the BRCA1 and BRCA2 breast cancers [175, 176]. In contrast to an XPD-deletion, which cannot remove the lesion and thus activates a checkpoint response, disrupting NER by blocked XPD release exemplifies a knowledge-based strategy to fool cancer cells for interventions. For example, the yeast rad3-102 mutant progresses faster through S into M-phase than normal cells and requires replication to convert NER damage to DSBs [167], causing cell death by mitotic catastrophe. This dependence on replication brings specificity for cancer cells with repair and cell cycle checkpoint defects. So small molecule ligands that bind and inactivate XPD without disrupting its ability to bind DNA and partner proteins may provide a successful cancer intervention strategy in combination with chemotherapies generating bulk lesions. This proposal suggests XPD binding is a potential target for cancer therapeutics because keeping XPD bound will block repair progression, but release CAK to allow cell cycle progression and generation of a dsDNA break at the next round of DNA replication.

TFIIH architecture enforces the spatial localization XPD, XPB and CAK such that XPD conformations can influence the binding and activities of these other enzymes. XPD conformations driven by ATP and DNA damage interactions can stabilize or destablize complexes. Such conformational changes are of great interest for controlling biological outcomes because they can also make interfaces accessible or impossible. In principle, the most extreme specificity can be achieved by sterically blocking a site so that subsequent interactions are impossible. The ability to make some interfaces impossible offers extreme specificity and a novel approach to drug design. By developing ligands that bind to and lock specific conformations for TFIIH enzymes, we expect researchers may be able to develop the ability to control interactions specifically and therefore control whether repair, signaling, or apoptosis will take place, thus limiting toxicity due to the essential role of TFIIH in transcription. As NER is more specific and less redundant than many repair responses, targeting this pathway promises to provide unique advantages for at least some cancers.