Pre-incision structures reveal principles of DNA nucleotide excision repair

Summary

Nucleotide excision repair (NER) removes bulky adducts from genomic DNA and prevents the ultraviolet light (UV)-sensitivity disease xeroderma pigmentosum (XP), cancer and pre-mature-aging ¹. After initial lesion recognition by XPC in global genome repair (GGR) or by stalled RNA polymerases in transcription-coupled repair (TCR), a lesion and surrounding DNA duplex are unwound by TFIIH, which includes ATPases XPB and XPD, and additional NER factors XPA, XPF, XPG and RPA, to form a ~27-nt bubble ². The ds-ss junction-specific endonucleases XPF and XPG cleave DNA on the 5ʹ and 3ʹ sides of the lesion, respectively. We report here the functional steps and atomic structures of the ATPase-driven and lesion-dependent DNA bubble formation and arrangement of the complete NER factors for dual incision. Unwinding of nearly 30 bps DNA depends mainly on the dsDNA translocase XPB and the duplex dividers XPA and XPF. XPD binds the lesion strand with XPF at the 5ʹ ds-ss junction. XPF cuts the lesion strand only after XPG binds the 3ʹ ds-ss junction. The ERCC1 subunit of XPF facilitates DNA strand separation and recruitment of RPA to the non-lesion strand. These findings shed light on the causes of human diseases and potential targets for enhancing chemotherapeutic efficacy.

Introduction

DNA base-dimers induced by ultraviolet light (UV) or cisplatin, and bulky lesions caused by various aromatic adducts impede replication and transcription ^3,4. These lesions need to be removed promptly by NER. Defective NER is manifested as XP, an autosomal recessive disorder characterized by extreme sensitivity to UV light, skin lesions and cancers. Mutations in genes encoding NER factors are also implicated in Cockayne syndrome (CS) and trichothiodystrophy (TTD), which are associated with impaired development, mental retardation and premature death ^5,6. After initial lesion recognition, GGR and TCR both depend on XPA, RPA ⁷ and the seven-subunit core complex of TFIIH (Core7) ^8,9 to unwind 22–29 base pairs of DNA and verify the presence of a bulky DNA lesion ¹⁰. XPF and XPG then make dual incisions near the borders of the DNA bubble.

The core NER process has been dissected genetically and biochemically ^11,12. Human GGR has been reconstituted in a cell-free system with purified NER factors ^13–15. To date, however, only lesion recognition by XPC and recruitment of Core7 and XPA to the lesion-containing DNA site have been structurally characterized ^16,17. After flipping out damaged base(s), XPC binds the non-lesion strand and DNA duplex downstream of the lesion ¹⁸. Core7 is recruited by XPC with XPB binding the dsDNA upstream of the lesion ¹⁷. Binding of XPA to the Core7, XPC and DNA complex (C7CD) induces extensive conformational changes in Core7 that shifts the XPC-lesion DNA a helical turn away from XPB and bends dsDNA between XPB and XPD in the C7CAD complex (for Core7, XPC, XPA and DNA) ¹⁷. In contrast, without XPC and a lesion, XPA binds the ds-ss junction of a Y-shaped DNA between XPB and XPD, which binds the ssDNA 3ʹ to the duplex (Extended Data or ED Fig. 1a-d) ¹⁶. Extensive mapping of the protein-interactome among NER factors and AlphaFold predictions of the full NER complex structure ¹⁹ have established a framework for understanding NER. The atomic details of XPF, XPG and RPA assembling on a bulky DNA lesion, as well as the dynamic process of DNA bubble formation leading to dual incision, have yet to be determined.

By combining mutagenesis with in vitro and in-cell GGR assays and cryoEM analyses, we have determined the pathway and intermediate structures involved in lesion-DNA bubble formation and strand cleavage during NER. We have elucidated the sequential roles of XPF and XPG in DNA strand separation and displacement of XPC and visualized the dynamic movements of two ATPases (XPB and XPD) and two endonucleases (XPF and XPG), which are orchestrated by XPA and ERCC1.

Reconstitution of global genome repair

We successfully reconstituted GGR in vitro using purified NER factors (Fig. 1a-c). A 94 bp DNA containing an internal Cy5, which mimic a bulky adduct was subjected to dual incision and produced lesion-containing oligonucleotides. In this reconstituted GGR reaction, 80% of substrate was converted to product within 60 min, and the main products were 27–28 nt long. When XPF was omitted, XPG cleaved the lesion strand ~5 nt downstream of Cy5, resulting in Cy5 oligonucleotides of ~70 nt (Fig. 1c). When XPG was omitted, however, XPF had no detectible nuclease activity (Fig. 1c). In the presence of a nuclease-inactive XPG, XPF cleaved ~20-nt upstream of Cy5 and produced Cy5 fragments of ~50 nt. XPG cleaves the lesion strand independent of XPF binding and plays an architectural role in activating XPF for dual incision, as reported previously ^20,21. These XPF and XPG cleavage sites are consistent with the NER dual incision sites using genuine UV and bulky adduct lesions ^22–24. In all cases the lesion is much closer to the 3ʹ than the 5ʹ end of excised DNA (Fig. 1c).

Figure 1.

Activity and structure of dual incision in GGR. a. Primary structure of NER factors. Well defined structural domains are in solid rectangles, and untraced in cryoEM structures are indicated by dotted lines. Interacting domains in RPA and XPF are boxed in dashed lines. b. Purity of NER factors are shown in SDS gel. c. Results of dual incision shown in a DNA denaturing gel. Cy5-labeled oligonucleotide substrate is diagrammed below the gel. Time course of WT proteins, results of omitting XPF (-F) or XPG (-G) and using nuclease-inactivated XPG mutant (+Gm) are shown. For gel source data of b and c, see Supplementary Figures 1 and 2. d. ATP- and RPA-dependent alternative assembling of GGR with addition of XPF or XPG first. Both lead to an 18-nt lesion-containing bubble. e. cryoEM map of C7CAD and No-F in the presence of ATP·Mg²⁺. f. The DNA structure with a bubble of 9-nt in No-F. g-k. cryoEM maps of No-G (g), DuI^S (h), and DuI^M (j) and corresponding DNA structures of DuI^S (i), and DuI^M (k). Each protein subunit is colored coded as labeled.

Forming and enlarging lesion-DNA bubble

Recruitment of Core7 and XPA to a lesion site by XPC is independent of ATP ¹⁷, and unpaired DNA is limited to 2–3 bp at the lesion where XPC binds. Yet the same structure of Core7 and XPA can bind to a preformed Y-shaped DNA (PDB: 6RO4) ^16,17. It is thus predicted that upon addition of ATP, a DNA bubble of ~10 bp would form between XPA and XPC ¹⁷ (Fig. 1d). Indeed, after incubation of C7CAD with ATP·Mg²⁺ or ATP·Mg²⁺ and RPA, cryoEM analysis (ED Fig. 2) confirmed that a DNA bubble is formed, with XPA at the 5ʹ ds-ss junction and XPD binding 4–9 nt of the unpaired lesion strand (Fig. 1e, 2a; Methods, and ED Table 1). In the 3.5 Å C7CAD·ATP structure, Core7 is the same as in C7CAD and as when bound to the Y-shaped DNA. Interestingly, the C-terminal region of XPC (aa 889–940) remains anchored in Core7 (XPB/p52/p8), while the main body of XPC, along with the lesion and downstream DNA duplex, is mobile and poorly defined (ED Fig. 2). Neither RPA nor ATP or ADP are observed.

Figure 2.

Conformational changes induced by XPF binding. a. A 9-nt DNA bubble formed in C7CAD upon ATP·Mg²⁺ addition. b. XPF is at the 5ʹ ds-ss junction in No-G complex and induces large conformational changes of Core7 and dissociation of XPA from XPD. c. XPF has extensive hydrophobic and charge interactions with the long helix in XPA (LHA). d. XPF and XPA penetrate the DNA duplex from the major and minor groove side, respectively, and separate the two strands. e. XPD in No-G and DuI^S complexes interacts extensively with XPC’s LHN and LHC and stabilizes the trimeric XPC on the lesion DNA. f. The tip of Arch domain (plug) in transcription PIC (PDB: 6NMI) is folded back (pink) and blocks the DNA binding channel. In the NER complexes, the tip is extended (plum color) and interacts with XPC and RAD23 subunits.

Subsequent addition of XPF or XPG had very different consequences. Analysis by cryoEM revealed that the structure of the C7CAD·ATP complex with addition of XPG (D812N mutant nuclease, and named No-F because it lacks XPF remains unchanged (Methods, ED Fig. 2) except for the complete disappearance of XPC, consistent with its displacement by XPG, as previously reported ²⁵, and appearance of a helical patch of XPG (aa 188–257), which interacts with XPD (ED Fig. 2i) exactly as predicted by AlphaFold ¹⁹. The catalytic domain of XPG and the associated DNA are too mobile to be traced in the cryoEM map (ED Fig. 2). In addition, the XPD-bound lesion strand is better ordered than without XPG (Fig. 1f). Efficient DNA nicking in the absence of XPF (Fig. 1c) indicates that XPG has access to the 3ʹ ds-ss junction and cleaves the lesion strand DNA downstream of where XPD binds. In contrast, in the structure following XPF addition to the C7CAD·ATP complex (named No-G as it lacks XPG), which was refined to 4.3 Å (Fig. 1g; ED Fig, 3 and Methods), Core7 undergoes major conformational changes (Video 1), and XPF is inserted between XPA and XPD at the 5ʹ ds-ss junction. The lesion DNA bubble is enlarged to 16 nt with XPA and XPF at the 5ʹ and XPC at the 3ʹ ds-ss junctions (Fig. 1d, 1h). While most of the unpaired non-lesion strand is disordered, XPC retains its anchor in Core7 and its binding of the non-lesion strand opposite the Cy5 and downstream duplex. The RPA trimerization domain (TriCDE, composed of OB domains of RPA70-C, RPA32-D and RPA14-E) ²⁶ appears next to the ZnF (zinc finger) domain of XPA (aa 102–136, XPA^ZnF) (ED Fig. 3), approximately as predicted ¹⁹ (see detail below).

When both XPF and XPG (D812N mutant nuclease) were added along with RPA and ATP·Mg²⁺ for dual incision (DuI), two distinct complexes were revealed by cryoEM analysis (Methods). The structure of the first complex refined to 3.3 Å was nearly superimposable with the No-G complex (Fig. 1d, 1i-j; ED Fig. 4 and Table 1), in which XPA/XPF and XPC frame the 16-nt bubble. Two regions of XPG are ordered — the helical patch (aa 188–257) that interacts with XPD and a lone extended β strand (aa 349–361) that interacts with XPF. We termed this complex as DuI^S for its being the starting point of DuI. The second complex, which is refined to 3.4 Å and we term DuI^M for its being in mid-process, differs dramatically from DuI^S. XPF and ERCC1 together translocate 35Å and rotate 45° relative to its position in DuI^S (Fig. 1h-k) (Video 2). The XPG catalytic domain (XPG^cat, aa 1–93 and 766–987) is well-ordered and binds the 3ʹ ss-ds junction and downstream DNA. The DNA bubble is enlarged to 18 nt, widened to 80Å, and entirely traceable, with the lesion strand bound by XPF and XPD and the non-lesion strand by RPA70’s OB-ABC domains. The five OB domains (A, B, C, D and E) of RPA are located between XPA^ZnF and XPG^cat (Video 2). Perhaps because the DNA bubble is not large enough, the active site of XPF is distant from the DNA substrate. Our structural and biochemical studies (ED Fig. 1e-h) indicate that recruitment of XPG and XPF occurs independently with no particular order, which agree with the previous cell-based analyses ^27,28. Unexpectedly, the DNA bubble enlargement requires XPF, and the resulting No-G complex (like DuI^S) appears much more abundant than No-F.

XPF rearranges Core7, XPA and lesion DNA

Core7 in C7CAD is dramatically rearranged upon XPF binding. Both flex points in Core7 (between p34-p44 and within p52) ¹⁷ are rotated, and the result is that XPD and XPB at the tips of two Core7 arms undergo a relative rotation of 70° and translation of 50 Å (Fig. 2a-b, Video 3). This unexpected rearrangement ¹⁹ is an extension of the shape-change induced by XPA binding to C7CD (PDB: 8EBS) and formation of C7CAD (PDB: 8EBT) ^16,17 (ED Fig. 5, Video 1). The interface shared by XPB and XPD in C7CAD is reduced (464 Ų to 140Ų) in No-G and DuI^S complexes, while XPD and XPA separate completely. These changes allow XPF to access the 5ʹ ds-ss junction.

In No-G and DuI^S, the XPF subunit interacts with the long α-helix of XPA (LHA, aa 198–232) (Fig. 2c, 3a) after displacing CETN2 subunit and LHC of XPC in C7CAD ¹⁷. The RecA2 and helical domains of XPF (Fig. 1a) form a 700Ų interface with LHA by hydrophobic (XPF’s G612+Q617+Y619+L620+L623 with XPA’s F219+V223+L226+V230) and polar interactions (XPF’s E213 and E616 with XPA’s R231 and R227) (Fig. 2c). An XP mutation in XPF G611R ²⁹ at the XPF-XPA interface and several other XP mutations in the helical domain of XPF (I225M, C236R, and E239Q) (Qiagen, Ingenuity Pathway Analysis) likely perturb its recruitment and proper orientation by XPA. XPF and XPA approach the 5ʹ ds-ss junction from the major and minor groove, respectively, and the tip of XPF’s helical domain (C242) and XPA’s DNA separation pin (H171) form van der Waals contact where the DNA strands diverge (Fig. 2d). Following XPF recruitment to lesion sites by the well-documented XPA-ERCC1 interactions (Fig. 4a) ^27,30,31, the XPF subunit is positioned at the 5ʹ ds-ss junction by XPA (LHA) for DNA strand separation and cleavage. Alteration of the LHA-XPF interface, such as by the C236R mutation, results in severe NER deficiencies and persistent XPF at the lesion foci ³². At the other end of the horseshoe-shaped XPF, ERCC1’s NLD (nuclease-like domain, EDCC1^NLD) supports XPA in keeping the non-lesion and lesion strands apart (Fig. 2b, Video 2). This may explain why without XPF, XPG cannot enlarge the DNA bubble in No-F or rearrange Core7 to form DuI-like complexes (Fig. 1d-f). On the other hand, in No-G and DuI^S structures, XPF is auto-inhibited with its active site 40Å from the lesion DNA strand and access to its active site blocked by its own HhH domains (Fig. 1g-h, 2b) ³³.

Figure 3.

XPG re-orients and stabilizes XPF and RPA on lesion DNA. a-b. Top views of the DuI^S (a) and DuI^M (b) structures. XPF (XPF and ERCC1) and TriCDE of RPA move 35Å and rotate 45° relative to XPD, and the DNA bubble is enlarged. The long ⍺ helix of XPA (LHA) is shown as ribbons and covered with molecular surface. c. A front view of DuI^M. XPG binds DNA duplex downstream of the lesion mimic (Cy5) and interacts with both XPD and XPF. d. A zoom-in view of the dashed box in panel c. XPG brings XPF and XPD together. e-f. Zoom-in of the XPD-XPF interface when the two are brought together by XPG in DuI^M (e) and when they are apart in DuI^S (f). The area of XPF in the dashed oval interacts with LHA in DuI^S (f) or with XPG in DuI^M (e). In e and f panels, XPF is superimposed, and XPA, XPD and DNA are shifted due to the XPG binding.

Figure 4.

XPA and ERCC1 assemble the NER complexes. a. Cartoon of XPA and 16-nt DNA bubble in No-G and DuI^S complexes. The bottom insert shows ERCC1 (XPF) recruitment by XPA⁵⁶⁻⁷⁵, which also interacts with XPD. b. Diagram of interacting partners of XPA. c. Interacting partners of XPA and ERCC1 in the DuI^M structure. The non-lesion strand is bound by RPA (OB-ABC). Zoom-in views shown in panels e, f and g are indicated by the dashed boxes and oval. d. The XPA^22–45 helix is sandwiched between RNA32^WH and XPD. e. The extended XPA⁸⁸⁻¹⁰² interacts with TriCDE of RPA. The interaction between XPA’s F100 and R144 of RPA32, reminisces that between TERT and RPA32 ⁵⁰. f. The NLD domain of ERCC1 interacts with the OB-ABC of RPA70, XPD, RPA32^WH and XPA (back view of panel c). The non-lesion strand is colored orange. Location of five ERCC1 mutants (1 to 5) are labeled. g. XPA^169−175 form a hairpin that separates the lesion and non-lesion strand. Replacing the seven residues by one Gly or three Glu causes NER defects. h. Global genome repair (GGR) was assessed by the unscheduled DNA synthesis (UDS) in term of EdU incorporation. EdU intensities were normalized to XPA^WT or ERCC1^WT cells for each condition (n = 3). Statistics were performed using a one-way ANOVA, with multiple testing controlled using the Bonferroni multiple comparison test (* indicating p<0.05, ** p<0.01, *** p<0.005, **** p<0.001). i. Representative UDS images for selected conditions. Scale bar indicates 5 μm.

The changes in Core7 allow XPD to interact with the translocated XPC in No-G and DuI^S when the 16 bp duplex are unpaired (Fig. 2b, 2e). The three subunits of XPC along with the lesion DNA are translocated 55Å away from XPB and XPA. The C-terminal long α-helix of XPC (LHC) and CETN2 subunit contact the base of the Arch domain of XPD instead of LHA as in C7CAD; LHN (N-terminal long α-helix) of XPC and RAD23 are ordered and sandwiched between DNA duplex and the tip of the Arch domain (aa 275–320, known as the plug ¹⁶) (Fig. 2e). In the transcriptional pre-initiation complex, the plug is folded back, protected by the CAK module of TFIIH, and blocks the DNA-binding channel (PDB: 6NMI) ³⁴. When bound to DNA in NER complexes, the plug becomes dislodged and often disordered, allowing ssDNA to pass through XPD ^16,17,35. In the No-G and DuI^S structures, the plug is well-ordered and serves as a landing pad for XPC (Fig. 2f). XPA has been shown to dissociate the CAK module from TFIIH ⁸, but it is achieved indirectly. XPA shifts XPC-lesion DNA relative to TFIIH ¹⁷, and in the presence of XPF the translocated XPC overlaps with and displaces CAK (ED Fig. 3c). While XPC maintained its binding to the lesion DNA, the Cy5 is located between XPC and XPD at the entrance of the ssDNA binding channel of XPD (Fig. 2e-f).

The rearrangement of Core7 in the presence of XPF also exposes the zinc-finger domain of XPA (XPA^ZnF, aa 88–135) to interact with TriCDE of RPA (Fig. 4a-c,4e). TriCDE in our NER complexes are like previously reported structures of RPA alone or complexed with ssDNA ^26,36,37 (ED Fig. 6a-d). However, half of the ssDNA-binding surface of TriCDE interacts with XPA instead. The ssDNA-binding OB domains A and B of RPA70 (OB-AB) (aa 183–431) ³⁸ are more mobile than TriCDE as is the unpaired non-lesion strand in No-G and DuI^S (ED Fig. 3h).

XPA contributes more than XPA^ZnF to RPA recruitment. Its N-terminal helix (aa 22–45, XPA^22–45) is sandwiched between the winged helix-turn-helix (WH) domain of RPA32 (aa 204–270, RPA32^WH) and XPD (Fig. 4d). The XPA^22–45 and RPA32 ^WH interaction has been predicted ³⁹, but the interface includes hydrophobic in addition to the polar residues ⁴⁰. Moreover, XPA mediates XPF and Core7 interactions in No-G and DuI^S. The previously disordered residues 54–75 (XPA⁵⁴⁻⁷⁵) form an extended β-strand (aa 56–66) that interacts with the large β sheet in XPD’s HD2 (helicase domain 2) and then a β-hairpin loop (aa 68–75) that plunges into ERCC1’s NLD (nuclease-like domain) for XPF recruitment (Fig. 4a, insert) ^27,31.

XPG replaces XPC and engages XPF and RPA

In DuI^M, XPG^cat binds the duplex downstream of the lesion and is superimposable with the crystal structure of XPG^cat-DNA complex ⁴¹. Its active site is poised to cleave the lesion strand in the duplex region 3ʹ to the bubble. In addition, XPG tightens the interactions of XPF with Core7 and RPA with DNA (Fig. 3a-b). In DuI^M the helical patch of XPG that interacts with XPD is extended to include a long helix (aa 156–181), which replaces the LHC of XPC and contacts the base of Arch domain over the ssDNA binding channel of XPD (Fig. 3c-d). A newly formed 3-stranded β-sheet of XPG (aa 271–350) in DuI^M interacts with the RecA1 domain of XPF and directly links the lone β strand (aa 351–360) bound to XPF with the helical patch (Fig. 3c-d). Folding of these formerly disordered residues significantly shortens the distance between XPD and XPF and results in a 30–35 Å movement of XPF and ERCC1 relative to XPD (Fig. 3a-b; Video 4). Consequently, a new XPF-XPD interface of 530 Ų is formed in DuI^M, while the XPF-XPA interface in DuI^S (or No-G) is shifted and reduced by 440 Ų (Fig. 3e-f). A well-documented XP mutation in XPG, I290N ⁴² in the hydrophobic core of the 3-stranded β-sheet and helical patch, confirms the relevance of DuI structures and highlights the importance of this previously disordered region for lesion bubble enlargement and dual incision.

Together with XPF-ERCC1, TriCDE of RPA and associated XPA^ZnF also moves 20–30 Å, and the non-lesion strand, which is stretched out between XPA^ZnF and XPG^cat for OB-ABC of RPA70 to bind (Video 2), becomes ordered. In addition, preceding the ZnF domain, XPA has gained a newly formed strand (aa 88–101), which fits in a groove in OB-C domain of RPA70 in a non-intersecting path to the ssDNA-binding groove, and the XPA-RPA interaction is thus solidified (Fig. 4c, 4e). The ssDNA-binding OB-AB domains and TriCDE have no direct interaction and move independently, and each undergoes slight adjustments in domain-domain and protein-DNA interactions compared to the previously reported crystal structures of RPA ^26,38 (ED Fig. 6). The RPA trimer and its position relative to XPA and the ssDNA also deviate significantly from the AlphaFold prediction ¹⁹ (ED Fig. 7d-e). The non-lesion strand is kinked by 90° between OB-AB and TriCDE with RPA outside of the DNA bubble. The relative orientation of TriCDE and OB-AB in DuI^M differs from the structures of RPA alone in complex with ssDNA by a 150° rotation (Video 5, ED Fig. 7a-c) ^36,37. The 15-nt of unpaired non-lesion strand are bound in the ssDNA-binding groove of OB-ABC, but the DNA trajectories in the DuI complexes deviate from crystal structures by 5–7 Å due to the presence of NER factors. In DuI^M ERCC1^NLD is adjacent to the unpaired non-lesion strand and directly interacts with each long β hairpin loop of the OB-ABC domains (Fig. 4f, ED Fig. 6f).

Accompanying the XPF-ERCC1 movement, the horseshoe-shaped XPF opens by 8–10° at its two ends (helical and nuclease domains) in DuI^M (Video 6). At the helical end, HhH domains of XPF and ERCC1 become disordered as they move away from the XPF catalytic center and no longer block its access by DNA substrate (ED Fig. 4j-k). The XPF structures in DuI complexes are like the apo XPF-ERCC1 ³³. In DuI^M the tip of the helical domain of XPF is in contact with the first 4 nt of the unpaired lesion strand, but the 5ʹ ds-ss junction and lesion strand remain 40 Å away from the XPF active site. As the DNA bubble is 18 nt in size, additional unwinding of DNA to reach 25–30 nt and further protein conformational changes are needed for XPF to access the lesion strand. Based on its location in the DuI complexes, XPF cannot access the upstream DNA duplex and likely cleaves the lesion strand in an unpaired form as the DNA footprinting data indicated ⁴³ rather than a duplex as modeled ¹⁹. The product size of dual incision is likely determined by the need for the DNA bubble becoming large enough to allow the lesion strand to access XPF.

XPA and ERCC1 organize the dual incision

Assembly of the seventeen-subunit DuI complex and the sequential conformational changes are coordinated by two small protein subunits, XPA and ERCC1. XPA consists of multiple dispersed structural elements, which interact with the DNA ds-ss junction, XPB, XPC, XPD, XPF and RPA (Fig. 4b). ERCC1, a subunit of XPF, links the two unpaired DNA strands and organizes the ssDNA binding by OB domains of RPA. The H244R mutation in XPA, which occurs at the site that anchors XPA to Core7 (p8 and p52) ¹⁷, has been shown to cause defects in XPA recruitment and NER ⁴⁴. The XPA-ERCC1 interface has been shown to be essential for XPF recruitment ^30,31,45. To investigate the importance of previously unexplored molecular interactions mediated by XPA and ERCC1, fifteen variants of XPA and ERCC were designed and stably expressed at comparable levels in the corresponding knockout RPE1 cells (Fig. 4, ED Table 2). NER efficiency was assessed using unscheduled DNA synthesis (mainly reflecting GGR) (Fig. 4h-i) and transcription recovery (mainly reflecting TCR) following UV exposure (ED Fig. 8).

Each of five ERCC1^NLD variants (#1-#5), designed to perturb the individual interface with XPD (#1), OB-ABC of RPA70 (#2–4), and RPA32^WH (#5) (Fig. 4f), support NER at near wildtype levels (Fig. 4h, ED Fig. 8a). We hypothesize that after mediating XPF recruitment by XPA^27,30,31, ERCC1^NLD acts as a spacer to separate the lesion from the non-lesion strand, and its individual interface with OB-ABC, RPA32^WH or XPD accounts for only a fraction of this function. Indeed, when the five WT-like mutants are combined, the resulting ERCC1-all5 behaves as ERCC1 null in NER assays (Fig. 4h), supporting the essential role of ERCC1 in the dual incision. In parallel, four XPA variants, which delete the DNA separation pin (replaced by a Gly or three Glu, #2 and #2*, Fig. 4g) or perturb the TriCDE interface (#3, #4 and combined 3+4, Fig. 4e), exhibit clear NER defects (Fig. 4h, ED Fig. 8a). Unlike the deletion mutants (#2 and #2*), amino acid-substitutions in the DNA separation pin (#1) has no deleterious effect (Fig. 4g-h). Likewise, amino acid substitutions of the XPA residues at the XPD interface (#6) do not disrupt NER because the interface is formed by hydrogen bonds between mainchain atoms (Fig. 4a, insert). The WT-like NER activities of XPA mutants #5 and #7 at the RPA32^WH and XPD interface (Fig. 4d, 4h), respectively, may be due to the limited and mild Ala substitutions, as extended and more drastic alterations of the same RPA32^WH and XPD interface cause NER deficiency ⁴⁰. Among the NER-deficient XPA mutants, only deletion of the DNA separation pin (mutations #2 and #2*) reduces recruitment of XPA and downstream ERCC1 to sites of local UV damage (Fig. 4i, ED Fig. 8b). The XPA–RPA (TriCDE) interactions (XPA#3 and #4 encompassing aa 91–113 of XPA) appears dispensable for XPF/ERCC1 recruitment (No-G or DuI^S), but essential for dual incision.

ATPase Requirement for opening the DNA duplex

ATP hydrolysis is required to unwind DNA duplex and form a ~30 nt DNA bubble. XPD, which binds ssDNA, is a helicase capable of separating DNA duplexes ^46,47. But XPB, which binds and translocates along DNA duplexes, becomes an effective helicase only when assisted by the base-pair separation pin of XPA, which creates the ds-ss junction 5ʹ to the lesion. To assess whether XPB, or XPD, or both ATPases are required for each step toward lesion-DNA bubble formation and enlargement, the Walker A motif mutations of XPB (K346R) and XPD (K48R) were introduced, and Core7 mutants with XPB^KR or XPD^KR were purified as reported previously ⁹. We assayed the ATPase activity of WT as well as KR mutants of XPB and XPD in assemblies of C7CD, C7CAD, No-G, No-F and DuI complex (Fig. 5a-b). Before XPA recruitment to the lesion DNA, XPB and XPD contribute similarly to the total ATPase activity (WT level). Addition of XPA, which makes the dsDNA translocase XPB an effective helicase, increases both XPB (XPD^KR), and XPD (XPB ^KR) ATPase activities coupled with the DNA bubble formation. In C7CAD and No-G (with XPF and RPA), the total ATPase activity appears to be the sum of XPB and XPD. However, in the presence of XPG (No-F and DuI), the XPB ATPase activity is three to four times that of XPD, and WT Core7 ATPase activity is nearly independent of XPD (Fig. 5b). We suspect that XPF modulates the XPB ATPase activity and balances it with XPD’s. In the absence of XPF, the high rate of ATP hydrolysis may be futile because the DNA bubble is limited to 9 nt in size (Fig. 1).

Figure 5.

XPB and XPD ATPases in DNA translocation and bubble formation. a. Five states of GGR. b. ATPase activities of inactivated XPB^KR (cyan), XPD^KR (green), and WT (orange) Core7 in the five states. For gel source data, see Supplementary Figure 3. c. Structures of the helicase domains of XPB (HD1 and HD2) in the open (apo) and closed (ADP-bound) state. d. Mechanism of DNA translocation by XPB. In each ATPase cycle (following a specific base pair marked by an outlined oval), ATP binding leads to the movement of one strand (top, orange) by 1 nt, and ADP·Pi release leads to the movement of the other strand (bottom, chocolate) also by 1 nt. Thus, ATP binding and release of ADP and Pi after ATP hydrolysis lead to translocation of 1 bp.

To visualize how XPB and XPD ATPase activity contribute to GGR complex formation, XPB^KR and XPD^KR mutant Core7s were examined in DNA bubble formation (C7CAD·ATP and No-F) and enlargement (No-G complex) (Fig. 1d) by gradient centrifugation (GraFix) ¹⁷ or the on-column crosslinking method (see Methods). Analyses by cryoEM revealed that XPB^KR and XPD^KR both supported formation of a 4–9 nt DNA bubble in C7CAD and No-F. However, without XPB’s ATPase activity, Core7 transformation and a 16-nt bubble cannot be detected (ED Fig. 2). In contrast, XPD^KR allowed formation of a 16-nt bubble and XPF recruitment, and the resulting No-G complex was structurally indistinguishable from WT complex (ED Fig. 3), except that the efficiency of bubble formation was reduced. These results suggest that ATP hydrolysis by XPD may aid but is not necessary to translocate XPC, open Core7, and unwind 16 bp. In the dual incision assay, without the XPB ATPase activity there is no cleaved product, while without the XPD a trace amount of excised DNA product can be observed ⁹. The unequal dependence of NER on the XPB and XPD ATPase activity may be rationalized by their different roles of translocating and unwinding ~20 bp DNA (XPB) versus binding ssDNA and detecting bulky lesions (XPD).

Waddling mechanism of DNA translocation

Many ATPases are known to translocate along DNA duplexes, but their translocation mechanism is not as well understood as that of ssDNA translocases. Among the different NER complexes we visualized, two RecA-like helicase domains (HD1 and HD2) of XPB are most often in the open (nucleotide-free) state, but they are closed in No-G and DuI^S complexes and superimposable with the ADP-bound structure in transcription initiation ⁴⁸. Yet no ADP or ATP was observed in the cryoEM structure of No-G and DuI^S. It is possible that the nucleotide dissociated during cryoEM grid preparation. Nevertheless, the conformational changes of XPB and the DNA duplex can be summarized in a two-step model for translocation of 1 bp per ATP hydrolyzed (Video 7) (Fig. 5c-d). Upon ATP binding and closing of the HD1 and HD2 helicase domains, the non-lesion strand is translocated by 1 nt while maintaining base pairing, resulting in tilting of the DNA base pairs. After ATP hydrolysis, release of ADP and Pi drives translocation of the lesion strand by 1 nt, restoring B-form DNA with an overall translocation of 1 bp. In summary, two key phases of each ATPase cycle _— ATP binding and release of ADP and Pi_— result in two power strokes, with each strand of the DNA duplex being translated by 1 nt during each power stroke, while all base pairs remain intact. Similar dsDNA translocation is also observed in the nucleosome remodeler Swi/Snf ATPases ⁴⁹. We refer to this DNA duplex translocation as the “waddling mechanism” due to its resemblance to the way penguins walk.

Conclusion

This study, which includes all the protein factors necessary for dual incision, defines the structural basis of nucleotide excision repair (NER). While mutations in the active sites of helicases or nucleases leading to xeroderma pigmentosum are easily understood, many other mutations occur at the protein-protein and protein-DNA interfaces. The dynamic movements revealed in this study shed light on both interfacial mutations and internal residues that are essential for the integrity of the step-by-step NER assemblies. Furthermore, the dramatic and continuous movement of XPD during the NER process addresses the long-standing question of why adjacent mutations or different amino acid substitutions of a given residue in XPD can lead to Xeroderma Pigmentosum, Cockayne syndrome, or trichothiodystrophy ^5,6. This variation arises because the environment surrounding an individual mutation changes dynamically, which allows nearby mutations or different substitutions at the same site to affect NER assembly to varying extents. Finally, since NER also removes DNA damage caused by chemical agents and reduces the efficacy of chemotherapies, the detailed dual-incision structures reveal specific target sites for modulating NER activity and enhancing the killing of cancer cells.

Methods

Protein Expression and Purification

The recombinant human XPA, XPC trimer (XPC, RAD23B and CETN2), XPF/ERCC1, XPG, and Core7 of TFIIH (XPB, XPD-Flag, p8, p62, p34, p52 and His-p44) including ATPase mutant XPB(K346R) or XPD(K48R) were expressed and purified as previously described ^{9, 17}. The human RPA subunits (RPA70/32/14) were cloned into the expression vector pET11d ⁵¹ and expressed in BL21(DE3) cells in TB medium with 100 mM ZnSO₄ and induction by 1 mM IPTG at OD₆₀₀ 0.8. After 4 hours incubation at 37°C, 1L bacteria culture was harvested and lysed in lysis buffer (50 mM Tris-HCl (pH8.0), 0.25 mM EDTA, 5 mM β-mercaptoethanol, 0.01% NP-40, 10 mM ZnCl₂) supplemented with protease inhibitor cocktail (Roche). Cell lysate was loaded onto 10ml HiTrap Blue HP affinity column (Cytiva) equilibrated with buffer A (lysis buffer plus 10% glycerol). After extensive wash by buffer B (Buffer A plus 800 mM KCl), a second wash step was performed with 6 CV of 34% elution buffer (Buffer A plus 1.5M NaSCN). The recombinant RPA was then eluted by 100% elution buffer and peak fractions were collected and buffer exchanged to Buffer C (Buffer A plus 100 mM KCl and 5 mM TCEP). The eluant was further purified on a 5ml HiTrap Q (Cytiva) column pre-equilibrated with Buffer C without KCl. Elution was performed with a linear gradient from 0% to 60% of 1M KCl in Buffer C. The peak fraction was concentrated and purified on a Superdex 200 column (10/300, Cytiva) with buffer containing 20 mM Tris-HCl pH7.4, 200 mM KCl, 10% glycerol, 5 mM DTT and 25mM ZnCl₂. The peak fractions containing RPA were pooled, concentrated, flash frozen, and stored at −80°C.

The nuclease-dead XPF (XPF2m with E725Q/D713N, pWY3060) and XPG (XPG2m with D812N/D861N, pWY3058) as well as the nuclease attenuated XPG_D812N (pWY3061) were generated using site-directed mutagenesis and confirmed by sequencing. The nuclease dead XPF and XPG were used in NER dual incision assays to determine the effect of pre-incubation of XPF or XPG. XPG_D812N was used in structural analysis. All expression vectors for this study are listed in Supplementary Table 1.

Preparation of DNA oligos and dual incision assay were carried out as previously reported ¹⁷.

Cryo-EM preparation and data collection

The No-G complexes were prepared by mixing 0.6 μM Cy5-DNA with Core7 (WT, XPB^KR or XPD^KR mutant), XPC, XPA, XPF/ERCC1, RPA at a 1:1:1:1:1:2 molar ratio in 25 mM HEPES (pH7.9), 100 mM KCl, 2 mM TCEP and 10% glycerol, and incubating with 4 mM ATP and 10 mM MgCl₂ at 37°C for 20 min. The reaction was stopped by adding 10 mM EDTA and 1 mM CaCl₂. The No-G protein-DNA complexes were stabilized and purified using the GraFix Method ⁵². A glycerol (20–40%) and glutaraldehyde (0–0.04%) gradient was prepared with BioComp Gradient Master 108 and centrifuged at 165,000g for 14 hours at 4°C. The fractions were analyzed on SDS-PAGE gels, and those containing the protein-DNA complexes were pooled and treated with 100mM (final) Tris-HCl (pH8.0) to quench the excess glutaraldehyde. Each combined protein-DNA complex sample was dialyzed against 25 mM HEPES (pH7.9), 100 mM KCl, 2 mM MgCl₂, 2 mM TCEP and 0.5% glycerol in a Slide-A-Lyzer MINI (2ml, 20kDa MWCO) for 12 h and concentrated to 0.6mg/ml by Amicon Ultra Centrifugal Filter (0.5ml, 30kDa MWCO), and used for the cryo-EM grid preparation directly. Quantifoil (R1.2/1.3) holey carbon grids were glow-discharged for 60s in an easiGlow system (TED PELLA, Inc.) before use. A drop of 2.5μl No-G sample was applied to each grid in an FEI VitroBot IV maintained at 4°C with 100% humidity. The grid was blotted for 4s with force 4 and plunge-frozen in liquid ethane cooled by liquid nitrogen. The Titan Krios G4 electron microscope (Thermo Fisher) equipped with a Gatan imaging filter Quantum system and a post-Gatan imaging filter Gatan K3 Summit direct electron detector (Multi-Institute Cryo-EM Facility; MICEF) of the NIH operated at 300kV with SerialEM ⁵³ was used to collect 19,910 movies of WT at 105,000 × nominal magnification in super-resolution mode (corresponding to a calibrated pixel size of 0.415 Å at the sample level). During data collection, each movie was recorded with an exposure time of 2 s and an accumulated dosage of about 55.8 e/Ų, which was fractionated into 27 frames (exposure of 113ms each), and the dose per frame was 2.07. For the XPB^KR and XPD^KR mutant No-G, samples were prepared as for the WT complex and 8,907 and 3,220 movies were collected for each mutant, respectively, on a Glacios microscope (MIEF, NIH).

NER dual incision complexes were assembled by mixing 0.13 μM Cy5-containing DNA (94_iCy5_62/30, as described ¹⁷) with XPC, Core7, XPA, RPA, XPF and XPG_D812N at 1:2:3:2:1:1 molar ratio in 0.5ml buffer containing 25 mM Tris-HCl pH8.0, 50 mM KAc 5 mM DTT, 15% glycerol. After pre-incubation at 37˚C for 5 min, 4 mM ATP and 10 mM MgCl₂ were added and further incubated at 37°C for 5 to 10 min. The reaction was attenuated by adding 10 mM CaCl₂ (final concentration), and the protein-DNA complexes were stabilized and purified using the on-column crosslinking method ⁵⁴. Briefly, 0.5 ml glutaraldehyde (0.2–0.4%) was load onto a Superose 6 increase column (10/300, Cytiva) pre-equilibrated with 25 mM HEPES (pH7.9), 100 mM KAc, 2 mM CaCl₂, 2 mM TCEP, 0.05% octyl β-D-glucopyranoside and 10% glycerol at a flow rate of 0.3 ml/min for 7 min. NER dual incision (DuI) complex was then loaded and fractions containing the protein-DNA complexes were pooled and treated with 50 mM (final) Tris-HCl (pH8.0) to quench the excess glutaraldehyde (ED Fig. 4). For cryoEM grid preparation, the DuI complexes were concentrated to 0.2~0.4 mg/ml in an Amicon 30 kDa centrifugal filter and buffer exchanged to 25 mM Tris-HCl pH8.0, 100 mM KAc, 2 mM CaCl₂, 2 mM TCEP, 0.5% glycerol and 0.05% octyl β-D-glucopyranoside. For C7CAD•ATP (WT, XPD^KR and XPB^KR) complexes, the protein-DNA complexes were prepared as DuI complexes but without RPA, XPF and XPG; for No-F complexes, XPF was omitted. Quantifoil (R1.2/1.3) holey carbon grids (EMS) were glow-discharged for 60 seconds at 15 mA in an PELCO EasiGlow (TED PELLA, Inc.) before use. A drop of 2~3 μl protein-DNA complexes was applied to the grid in an FEI Vitrobot IV at 4°C with 100% humidity. The grid was blotted for 1 s with force 4 and plunge-frozen in liquid ethane cooled by liquid nitrogen. All movies were collected at the Titan Krios G3 electron microscope (Thermo Fisher) equipped with a Gatan imaging filter (GIF) Quantum system and a post-GIF Gatan K3 Summit direct electron detector (Multi-Institute Cryo-EM Facility (MICEF, NIH) operated at 300 kV. 52,926 movies of DuI were collected at 105K magnification in super-resolution mode. A total of 10,516 movies of C7CAD•ATP and 13,842 movies of No-F were collected at 81K magnification in super-resolution mode. During data collection, each movie was recorded with an exposure time of 3.0 s and an accumulated dosage of about 50.0 ~ 52.5 electrons / Å2, which was fractionated into 40 frames (75 ms exposure each). Total 6,516 and 4,256 Movies of C7CAD•ATP (XPD^KR or XPB^KR) were also collected separately at 36K magnification in super-resolution mode on a Glacios microscope (MICEF, NIH).

Structure determination and refinement

Beam-induced movement in each movie was corrected using Motion Cor2 ⁵⁵ with dose-weighting applied, and the pixels were binned 2X to 0.83 Å for 105K dataset and 1.03 Å for 81K dataset to generate a set of dose-weighted and another of dose-unweighted micrographs. Gctf ⁵⁶ was used to estimate CTF (contrast transfer function) parameters on dose-unweighted micrographs. For all data, particles were picked using Gautomatch (www2.mrc-lmb.cam.ac.uk/research/locally-developed-software/zhang-software/) with or without template, and extracted from dose-weighted micrographs, separately. To speed up the data processing, all particles were downscaled to a pixel size of 3.32 Å and a box size of 110X110. 2D classification was carried out in RELION (ver. 3.1) a couple of rounds to remove junk particles. Good 2D classes were selected to generate the 3D initial model using RELION (ver. 3.1) ⁵⁷. Particles belonging to each type of NER complex were selected and re-extracted for auto-refinement. Non-duplicated particles from templated and non-templated picking processes were selected, combined and auto refined. To improve the conformational purity, several rounds of focused 3D classification without alignment within separate local masks over XPD/XPF, XPC/XPB/XPD, XPC or XPB were applied.

For the No-G complex, selected subsets of particles were re-extracted with a pixel size of 1.66 Å and a box size of 220X220, and subjected to CTF refinement, Bayesian polishing and auto-refinement in RELION (ver. 3.1). The final reconstruction of XPC_DNA and Core region in No-G reached 7.7 Å and 4.3 Å, respectively. The composite map of No-G was generated by aligning and merging the focused maps using UCSF Chimera (ver. 1.14) ⁵⁸ and postprocessed to 4.3 Å resolution. For DuI complexes, selected subset of particles containing DuI^S or DuI^M were merged separately and subjected to CTF refinement, Bayesian polishing and auto-refinement in RELION (ver. 3.1). The cryoSPARC 3D variability analysis (v4.6) showed continuous heterogeneity of final DuI^S and DuI^M particles. The reconstructed map of DuI^S was of 3.3 Å and DuI^M 3.4 Å resolution (ED Fig. 4). After improvement using focused refinement, the final composite map was assembled. DeepEMhancer was used to improve map quality for model building. For C7CAD·ATP and No-F datasets, data processing was performed by the same procedure. The No-F sample is intrinsically unstable, and the final reconstructed map is at 5.7 Å resolution.

Coordinates of TFIIH Core7 subunits and XPA (PDB: 6RO4, 8EBT, 8EBS), CETN2 (PDB: 2GGM), XPC/RAD23B (PDB: 8EBS, 8EBT), RPA (PDB: 1L1O, 1JMC), XPF/ERCC1 (PDB: 6XSA) and XPG catalytic domain (PDB: 6TUS) were initially docked into cryoEM maps. The structures of XPD (aa 247–440), XPC (aa 181–630), and XPG (aa 150–360) were predicted using Alphafold3 ⁵⁹ and fitted into the focus-refined cryoEM maps. After initial fitting into the map using Chimera (ver. 1.14) ⁵⁸, all parts were manually adjusted in Coot ⁶⁰. A standard B-form DNA was docked into the Cryo-EM maps of No-G initially and the single stranded DNA region was built manually in Coot. The final models were refined using PHENIX Real Space Refinement ⁶¹, and validated using MolProbity ⁶². The “gold standard” of the 0.143 Fourier shell correlation criterion was applied to determine the resolution of each structure, and ResMap ⁶³ and RELION (ver. 3.1) ⁵⁷ were used for local resolution estimation. Structure figures were made using PyMol (https://pymol.org/3.1). The C7CAD·ATP, C7CAD (XPB^KR)·ATP, and C7CAD(XPD^KR)·ATP complexes were processed and refined to final resolutions of 3.5 Å (WT), 6.3 Å (XPD^KR) and 4.9 Å (XPB^KR) (ED Fig. 2). For all C7CAD·ATP complexes, 4–6 nt bound by XPD were evident in cryoEM maps. Refinement statistics are reported in ED Table 1.

All structure coordinates and cryoEM maps have been deposited with PDB and EMDB.

Dual incision experiment

The dual incision reaction was performed in a 20 μl reaction mixture as previously descripted ^{9, 17}. For XPF or XPG preincubation, NER proteins (XPC, XPA, Core7, and RPA) together with either XPF or XPG and the DNA substrate, were pre-incubated at 37˚C for 20 min in the presence of 1mM CaCl₂. The reaction was then supplement with 1mM EGTA to chelate calcium ions. To initiate the dual incision assay, the second nuclease (either XPF or XPG), along with 3 mM ATP and 7.5 mM MgCl₂ were added. The reactions were incubated at 37˚C and stopped after 1, 3, 5, 10, 20, and 30 min by adding EDTA to the final concentration of 50mM. To denature the DNA oligos, the reaction mixture was heated at 95˚C for 5 min in TBE-Urea sample buffer before loading onto 15% polyacrylamide TBE-Urea gel. The gel and dual incision products were scanned and analyzed by Typhoon FLA 9500 Phosphor Imager.

To resolve the XPG cleavage bands, a large (20 cm x 20 cm) 15% polyacrylamide TBE-Urea gel was pre-run at 30W power for 1 h to reach approximately 60˚C. The reaction mixture was loaded after denaturing at 95˚C for 5 min in TBE-Urea sample buffer, and the gel was run for 3 h 30 min at 30W power. The products were scanned and analyzed by Typhoon FLA 9500 Phosphor Imager. To generate the Cy5 oligo size ladder by the hydroxyl radical cleavage method, 2 μl of 2 mM NER substrate was incubated with a freshly prepared mixture containing 2 μl each of 1 mM Fe²⁺, 2 mM EDTA, 0.6% H₂O₂ and 10mM sodium ascorbate for 5 or 10 minutes. The reaction was quenched with 2 μl of 100 mM thiourea stop solution.

Cell lines

All cell lines are listed in Supplementary Table 2. RPE1 cells were cultured at 37°C in an atmosphere of 5% CO₂ in DMEM, supplemented with antibiotics, 8% fetal calf serum (FCS) and glutaMAX (Gibco).

Plasmids

pX458 encoding Cas9 was purchased from Addgene. pU6-sgXPA-PGK-puro-2A-tagBFP and pU6-sgERCC1-PGK-puro-2A-tagBFP were obtained from the LUMC Sigma-Aldrich library. A region spanning the PGK promoter was amplified by PCR and used to replace the CMV promoter in pEGFP-C1-IRES-PURO. The XPA^WT and ERCC1^WT cDNAs were inserted into pPGK-EGFP-C1-IRES-PURO. All XPA and ERCC1 mutants were synthesized by Twist and listed in ED Table 2. All sequences were verified by Sanger sequencing. All plasmids are listed in Supplementary Table 3, and primers used are listed in Supplementary Table 4.

Generation of knockout cells

Parental RPE1-hTERT cells stably expressing inducible Cas9 (iCas9) that are also knockout for TP53 and the puromycin-N-acetyltransferase PAC1 gene were described previously (referred to as RPE1-iCas9). RPE1 cells were transfected with Cas9–2A-EGFP (pX458; Addgene #48138) together with either U6-sgXPA-PGK-puro-2A-tagBFP or pU6-sgERCC1-PGK-puro-2A-tagBFP using lipofectamine 2000 (Invitrogen). The sgRNAs used are listed in Supplementary Table 5 and plasmids in Supplementary Table 3. Cells were FACS sorted on BFP/EGFP and plated at low density after which individual clones were isolated, expanded and verified by western blot analysis and/or Sanger sequencing.

Stable integration of EGFP-tagged NER cDNAs

RPE1 cells knockout for either XPA (clone 1) or ERCC1 (clone 16) were transfected with either pPGK-EGFP-XPA-IRES-PURO (WT and mutant #1 through #7) or pPGK-EGFP-ERCC1-IRES-PURO (WT and mutant #1 through all5) plasmids. Cells were selected with 1 μg/mL puromycin to generate a polyclonal stable cell-line. Homogenous expression of EGFP-tagged NER proteins was verified by Western blot analysis.

Western blotting

Cell extracts were generated by cell lysis and boiled in Laemmli sample buffer. Proteins were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) on 4–12% Criterion XT Bis-Tris pre-cast gels (Biorad) in NuPAGE™ MOPS SDS running buffer (Thermo Fisher) and transferred to 0.45 μm PVDF membranes (Immobilon). Protein expression was analyzed by immunoblotting with the indicated antibodies in commercial blocking buffer (Rockland) (Supplementary Table 6), followed by detection using the Odyssey infrared imaging scanning system (LI-COR biosciences, Lincoln, Nebraska USA).

Unscheduled DNA synthesis (UDS)

Cells were plated in DMEM supplemented with 8% FCS. Subsequently, cells were placed in DMEM supplemented with 0–1% FCS for at least 24 hours prior to UV irradiation to reduce the excess of available deoxythymidine in the culture medium. Cells were locally UV irradiated through 5 μm pore filters (Millipore) with 30 J/m² and immediately pulse-labelled with 20 μM 5-ethynyl-deoxy-uridine (EdU; VWR) and 1 μM FUDR (Sigma Aldrich) for 1 hour. After medium-chase with DMEM containing 10 μM thymidine for 30 min, cells were fixed with 3.7% formaldehyde in PBS for 10–15 min at room temperature and stored in PBS. Next, cells were permeabilized for 20 min in PBS with 0.5% Triton-X100 and washed 2 times with PBS + 3% BSA (Thermo Fisher) for 5 minutes. The incorporated EdU was visualized by click-it chemistry, labelling the cells for 1 hour with a mix of 60 μM Atto azide-Alexa647 (Atto Tec), 4 mM copper sulphate (Sigma) and 10 mM ascorbic acid (Sigma) in a 50 mM Tris-buffer (pH8). After this, the cells were post-fixed with 2% PFA for 10 min and blocked with 100 mM Glycine. Cells were washed extensively with PBS, DNA was denatured with 0.5 M NaOH for 5 minutes, blocked with 10% BSA (Thermo Fisher) in PBS for 15 minutes and equilibrated in 0.5% BSA and 0.05% Tween20 in PBS (WB buffer). Sites of local UV damage were visualized by labelling the cells for 2 hours at room temperature, or overnight at 4 °C, with mouse anti-CPD (Cosmo Bio; 1:1000 in WB buffer). After primary antibody incubation, cells were washed extensively with WB buffer, stained with goat anti-rabbit IgG-Alexa555 (1:1,000 in WB buffer) for 1 hour, again washed extensively with WB buffer, stained with 0.1 μg/mL DAPI, washed extensively with PBS and mounted in Polymount (Brunschwig). Antibodies used are listed in Supplementary Table 6.

Recovery of RNA Synthesis (RRS)

Cells were plated in DMEM supplemented with 8% FCS. Subsequently, cells were placed in DMEM supplemented with 1% FCS for at least 24 hours prior to UV irradiation to reduce the excess of available uridine in the culture medium. Cells were UV irradiated (9 J/m²), allowed to recover for the indicated time periods, and pulse-labelled with 400 μM 5-ethynyl-uridine (EU; Axxora) for 1 hour. After medium-chase with DMEM without supplements for 15 min, cells were fixed with 3.7% formaldehyde in PBS for 15 min and stored in PBS. The cells were permeabilized for 10 min in PBS with 0.5% Triton-X100, washed with PBS + 1.5% BSA (Thermo Fisher) and then incubated with PBS + 1.5% BSA for 10 minutes. After washing with PBS, nascent RNA was visualized by click-it chemistry, labeling the cells for 1 hour with a mix of 60 μM Atto azide-Alexa594 (Atto Tec), 4 mM copper sulfate (Sigma), 10 mM ascorbic acid (Sigma) and 0.1 μg/mL DAPI in a 50 mM Tris-buffer (pH8). Cells were washed extensively with PBS and mounted in Polymount (Brunschwig).

Local UV damage

Cells were plated in DMEM supplemented with 8% FCS. Subsequently, cells were locally UV irradiated through 5 μm pore filters (Millipore) with 100 J/m². At 1 hour after UV irradiation, cells were fixed with 3.7% formaldehyde in PBS for 10–15 min at room temperature and stored in PBS. Next, cells were permeabilized by incubation with 0.5% TritonX100 in PBS for 5 minutes. Cells were subsequently treated with 100 mM Glycine in PBS for 10 minutes to block unreacted aldehyde groups, washed extensively with PBS and blocked with 0.5% BSA and 0.05% Tween20 in PBS (WB buffer) for 10–15 minutes. The TFIIH subunit p89/XPB (used as damage marker) and ERCC1 were visualized by labeling the cells for 2 hours at room temperature, or overnight at 4 °C, with mouse anti-ERCC1 (1:200 in WB buffer) and rabbit anti-p89/XPB (1:500 in WB buffer). After primary antibody incubation, cells were washed extensively with WB buffer, stained with goat anti-rabbit or anti-mouse IgG-Alexa 488, 555 and/or 647 (1:1,000 in WB buffer) for 1 hour, again washed extensively with WB buffer, counterstained with 0.1 μg/mL DAPI, washed extensively with PBS and mounted in Polymount (Brunschwig). Antibodies used are listed in Supplementary Table 6.

Microscopic analysis of fixed cells

Images of fixed samples were acquired on a Zeiss AxioImager M2 or D2 widefield fluorescence microscope equipped with 63x PLAN APO (1.4 NA) oil-immersion objectives (Zeiss) and an HXP 120 metal-halide lamp used for excitation. Fluorescent probes were detected using the following filters: DAPI (excitation filter: 350/50 nm, dichroic mirror: 400 nm, emission filter: 460/50 nm), Alexa 555 (excitation filter: 545/25 nm, dichroic mirror: 565 nm, emission filter: 605/70 nm), Alexa 647 (excitation filter: 640/30 nm, dichroic mirror: 660 nm, emission filter: 690/50 nm). Images were recorded using ZEN 2012 (blue edition v1.1.0.0) software and analyzed in Image J (1.48v).

Extended Data

ED Fig. 1.

Structures of lesion recognition in GGR and dual incision assays. a. Diagrams of primary structures of Core7, XPC and XPA. b-d. Structures of C7CD (b), C7CAD (c) and Core7 and XPA bound to a Y-shaped DNA (d). The three structures are superimposed based on XPB and proteins are color-coded as in panel a. e-g. Comparison of dual incision (DuI) efficiency of NER (e), NER with pre-incubation of XPF (f) or XPG in the presence of Ca²⁺ (g). The molecular size and products of XPG cleavage and DuI are indicated. The resolved XPG cut and DuI products at single nucleotide resolution are shown in the insert next to panel g. h. A plot of the time courses of averaged triplicate measurements of total DNA cleavage (sum of dual incision and XPG cut) in NER and with XPF or XPG preincubation (PreF or PreG). For gel source data of panels e-h, see Supplementary Figures 4 and 5.

ED Fig. 2.

Data processing and structure refinement of C7CAD·ATP and No-F complexes. a. Workflow of image processing of C7CAD·ATP (WT, XPD^KR and XP^KR Core7) and WT No-F complexes. The coordinates of XPB (green) and XPD (cyan) in Core7 without ATP (PDB: 8EBU) align well with the C7CAD·ATP and No-F map (light grey). b, d, Angular distribution of particles used in the final map calculation of WT C7CAD·ATP (b) and No-F (d). c, e. The composite map of WT C7CAD·ATP (c) and No-F (e) are colored according to the resolution scale bar on the bottom. f, g, FSC analysis of the quality and map resolution. The map-model FSC and half-map FSC of WT C7CAD·ATP (f) and No-F (g) are indicated as (M) and (H), respectively. h, Map volume corresponding to the C-terminal region of XPC (colored blue) in the C7CAD·ATP map (grey) shows its presence, while there is no map volume corresponding to the C-terminal region of XPC in No-F. i. Binding of the helical patch of XPG (purple) to XPD in No-F overlaps with p62 in the pre-initiation complex of transcription (yellow, PDB: 6NMI).

ED Fig 3.

Data processing and structure refinement of No-G complexes. a. Workflow of image processing of No-G complexes containing WT, XPB^KR and XPD^KR Core7. The cryoEM maps of WT and XPD^KR No-G (boxed in dark blue) indicate the presence of XPF, which is associated with the 16-nt DNA bubble. b. FSC analysis of the quality and map resolution of WT No-G. The map-model FSC and half-map FSC are indicated as (M) and (H), respectively. c. XPC and CENT2 subunits in No-G overlaps with MAT1 in the pre-initiation complex of transcription (PDB: 6NMI), when XPD is superimposed. d, e, Angular distribution of particles used in the final map calculation of WT (d) and XPD^KR (e). f, g. The composite map of WT (f) and XPD^KR No-G (g) are colored according to the resolution scale bar on the bottom. h, The cryoEM map of WT No-G, in which RPA fits. i, cryoEM map of XPD^KR mutant No-G is superimposable with the WT model.

ED Fig. 4.

Data processing and structure refinement of DuI complexes. a. Workflow of image processing of DuI complexes. b. Elution profile of the on-column crosslinked DuI complexes. c. SDS gel analysis of the eluted fractions. Proteins were stained with SYPRO ruby, and DNA was detected by Cy5 fluorescence. For gel source data of panel c, see Supplementary Figure 6. d, f, Angular distribution of particles used in the final map calculation of DuI^S (d) and DuI^M (f). e, g. The composite map of DuI^S (e) and DuI^M (g) are colored according to the resolution scale bar on the bottom. h, i, FSC analysis of the quality and map resolution. The map-model FSC and half-map FSC of DuI^S (h) and DuI^M (i) are indicated as (M) and (H), respectively. j, k, Local map of DuI^S (j) and DuI^M (k) corresponding to XPF and ERCC1 subunits. The HhH domains of XPF and ERCC1 are mobile and “disappear” in DuI^M map.

ED Fig. 5.

Change of Core7 structure during the DNA bubble formation. Comparison of XPD position in C7CD (light grey), C7CAD (yellow) and No-G (or DuI, green XPB, cyan XPD and orange and chocolate DNA) when XPB is superimposed. The red arrows follow the movement of the 4Fe4S cluster bound to XPD advancing in the three states.

ED Fig. 6.

Comparison of OB domains of RPA in DuI^M with apo and ssDNA bound states. a-c. Comparison of TriCDE in DuI^M with the structures of apo RPA and RPA-ssDNA complexes. d. OB-C (RPA70C) binds to both ssDNA and XPA (aa 88–102) in a non-intersecting fashion. Binding of XPA precludes ssDNA binding as without NER factors. e-f. Comparison of OB-AB (e) and OB-ABC (f) in DuI^M complex with OB-AB alone bound to ssDNA. The DNA trajectories differ because of the interactions between OB-ABC and ERCC1 in DuI^M complex.

ED Fig. 7.

Comparison of DuI^M with the RPA-ssDNA crystal structure and AI-based NER model. a. RPA and non-lesion strand in DuI^M. b. The crystal structure of RPA-ssDNA (PDB: 4GOP). c. Overlay of the two structures after superposition of OB-AB. d. When the RPA70C domain in the predicted NER models (PDB: 9A89 or 9A88, reference 21) is superimposed with that in DuI^M, the OB-AB (colored blue) are flipped by 180° and located inside of the DNA bubble (grey) instead of outside as in DuI^M (green and yellow/orange). e. With the same superposition (represented by RPA32D), the ZnF domain of XPA in DuI^M (encircled, hot pink) deviates by 10Å and 27° from the predicted model (yellow). RPA32^WH and XPA^22–40 helix are far away between the two as indicated by the double arrowheads.

ED Fig. 8.

Cell-based analyses of fifteen ERCC1 and XPA mutants in TCR. a. Transcription-coupled repair (TCR) was evaluated by the recovery of RNA synthesis (RRS). EU intensities were normalized to XPA^WT or ERCC1^WT cells at 24 hours post-UV irradiation for each condition (n = 3). Statistics were performed as in Fig. 4h. For source data shown in panel a and Fig. 4h, see Supplementary Source Data. b. Local damage analysis of NER factor recruitment in NER-defective ERCC1-all5, XPA mutants #2 and #2* (DNA-binding hairpin), #3 and #4 (TriCDE binding). Scale bar in panel b indicates 5 μm. For validation of stable XPA or ERCC1 knock-down cells, see Supplementary Figures 7 and 8.

ED Table 1.

cryoEM data collection, processing and model refinement

Type of complex PDB accession code EMDB accession code	C7CAD·ATP 9XYU EMD-72341	No-F 9PD5 EMD-71526	No-G 9PCP EMD-71512	DuIS 9PD3 EMD-71524	DuIM 9PD4 EMD-71525
Data collection and processing
Magnification	81k	81k	105k	105k
Voltage (kV)	300	300	300	300
Total electron dose (e−/ Å2)	50	50	55.8	52.5
Defocus range (μm)	−0.8 to −2.4	−0.8 to −2.4	−0.8 to −2.4	−0.8 to −2.4
Pixel Size (Å)	1.03	1.03	0.83	0.83
Symmetry imposed	C1	C1	C1	C1
Initial particle images (no.)	6,24,772	5,94,005	15,39,927	13,14,793
Final particle images (no.)	1,42,798	64,352	81,380	1,61,019	1,40,332
Map resolution (Å)	3.5	5.7	4.3	3.3	3.4
FSC threshold	0.143	0.143	0.143	0.143	0.143
Map resolution range (Å)	3.2 – 6.5	4.5 – 8.0	3.7 – 9.0	3.0 – 6.5	3.0 – 6.5
Refinement
Initial model used (PDB code)	8EBT, 6RO4	8EBT, 6RO4	8EBS, 6XSA	8EBT, 1L1O, 1JMC, 6TUS
Model resolution (Å)	3.9	6.8	4.6	3.6	3.8
FSC threshold	0.5	0.5	0.5	0.5	0.5
Map sharpening B factor (Å2)	−40	−120	−100	−40	−40
Model Composition
Non-hydrogen atoms	23,406	23,740	36,356	35,083	40,040
Protein residues	2,789	2,816	4,319	4,246	4,732
Ligands	7	7	10	8	9
Nucleotides	49	53	97	70	100
CC_mask	0.69	0.72	0.77	0.80	0.82
CC_volume	0.69	0.72	0.77	0.79	0.81
CC_peaks	0.68	0.56	0.63	0.7	0.71
B-factor (Å2)
Protein	83.2	356.1	205.3	181.1	202.9
Nucleotide	96.8	356.8	184.6	213.6	216.6
Ligand	134.6	473.2	197.3	266.4	200.1
R.m.s. deviations
Bond length (Å)	0.002	0.002	0.002	0.002	0.002
Bond angle (°)	0.44	0.44	1.19	0.43	0.44
Ramachandran plot
Most favored (%)	97.0	97.3	92.7	96.5	96.9
Allowed (%)	3.0	2.7	7.3	3.5	3.1
Disallowed (%)	0	0	0.05	0	0
Validation
Clash score	7.8	7.3	8.9	6.8	5.7
MolProbity score	1.6	1.5	1.9	1.6	1.7
Poor rotamer (%)	0.6	0.6	0.03	0.8	1.8

Extended Data Table 2:

Designed mutant XPA and ERCC1 proteins

Name	Interface
XPA#1 (H172A, Q174A, W175A)	Separation pin (DNA)
XPA#2 (D169–175 to G)	Separation pin (DNA)
XPA#2* (D169–175 to 3E)	Separation pin (DNA)
XPA#3 (M113A, F100A, M98E)	RPA70, RPA32
XPA#4 (V91E, P94R, V97D)	RPA70 (OB-C)
XPA#3+4 (M113A, F100A, M98E, V91E, P94R, V97D)	RPA70, RPA32
XPA#5 (A35L)	RPA32-WH
XPA#6 (G57D, G58S, P66S)	XPD loading
XPA#7 (M37A, L38A, A41L)	XPD
ERCC1#1 (N147E, L148A, V175L, K176E, Q180A)	XPD
ERCC1#2 (H154A, Q158A, L191A, A192V)	RPA70 (OB-A)
ERCC1#3 (Q134A+S135D+N163A+A165V)	RPA70 (OB-B)
ERCC1#4 (R117E, P120I)	XPA, RPA70 (OB-C)
ERCC1#5 (P105D, R106A, D125R)	RPA32, XPA, RPA70
ERCC1all5 (#1+#2+#3+#4+#5)	XPD+RPA70+RPA32+XPA

Data Availability

Coordinates and cryoEM maps reported here have been deposited with PDB (accession code 9XYU, 9PD5, 9PCP, 9PD3 and 9PD4 and EMDB (accession code EMD-72341, 71526, 71512, 71524 and 71525) for C7CAD·ATP, No-F, No-G, DuI^S and DuI^M complexes.