Our genetic information is constantly challenged by endogenous and exogenous perturbances. Highly efficient preservation mechanisms have evolved to retain the original genetic information with one of them being nucleotide excision repair (NER). Deficiencies in NER can lead to different genetic disorders, such as xeroderma pigmentosum (XP), trichothiodystrophy (TTD) and Cockayne syndrome (CS) (or a combination of XP and CS). Whereas the hallmark of XP and XP/CS is an increased risk in cancer, TTD and CS patients can suffer from neurological impairments and photosensitivity [1] thus exemplifying the importance of NER for human health. NER is a template based repair mechanism and requires an intact DNA strand to work efficiently [2]. NER is unique in comparison to other DNA repair mechanisms due to its broad substrate specificity, addressing structurally very diverse DNA lesions like acetyl amino fluorene adducts, cisplatinum DNA crosslinks, and lesions induced by UV irradiation such as cyclo butane pyrimidine dimers (CPD) or 6,4 photoproducts (6,4-PP) [3]. To achieve this variability, more than 30 different proteins participate in a multi-step pathway in eukaryotic NER ensuring that lesions are efficiently detected and removed [4, 5, 6]. Initial lesion detection (Figure 1) follows two entry pathways. In global genome repair (GG-NER) the XPC complex comprising xeroderma pigmentosum group C (XPC), the human Rad23 homolog B (RAD23B), and Centrin-2 constantly scans the accessible genome for lesions. For CPDs and 6,4-PPs it requires assistance by the UV-DNA damage binding (UV-DDB) proteins. Once a distorted DNA structure has been confirmed, the general transcription factor II H (TFIIH) is recruited. In transcription coupled repair (TC-NER) RNA polymerase II (RNA Pol II) is stalled when it encounters an anomaly in the template DNA. This stalling initiates the recruitment of the Cockayne syndrome B (CSB), Cockayne syndrome A (CSA), and UV stimulated scaffold protein A (UVSSA) proteins. Their combined actions also lead to the recruitment of TFIIH [7]. With the arrival of TFIIH, the two pathways merge into a common mode of action. TFIIH is a multi-subunit complex consisting of 10 proteins in which xeroderma pigmentosum group D (XPD), xeroderma pigmentosum group B (XPB), p8, p34, p44, p52, and p62 constitute the core TFIIH complex, whereas the cyclin dependent kinase 7 (CDK7), Cyclin-H, and menage a trois (MAT1) proteins build the cyclin activating kinase complex (CAK) [8]. After recruitment, TFIIH utilizes its translocase and helicase activity (facilitated by the XPB and XPD super family 2 (SF2) helicases, respectively) to widen the bubble around the lesion and perform damage verification [4,6,9]. In addition, it has been proposed that this activity might also be utilized to push back RNA Pol II in TC-NER and to support the dissociation of the XPC complex in GG-NER. While the XPB mediated translocase within TFIIH is constantly activated [10], helicase activation of XPD requires the dissociation of the CAK. The latter process is initiated by the interaction of the xeroderma pigmentosum group A (XPA) protein and RPA with TFIIH. Upon the release of the CAK, XPD helicase activity leads to a widening of the bubble around the lesion and possibly also to lesion verification [11, 12, 13]. Since the initial encounter with DNA only requires the presence of thermodynamic destabilization in GG-NER and a lesion, an unusual sequence or the presence of the nucleosome that can stall RNA Pol II in TC-NER, lesion verification is vital to prevent excision of undamaged DNA [4]. Lesion verification and subsequent 5′ incision demark the point of no return in the cascade and lead to the excision of the damaged fragment. The two nucleases involved in this step are the xeroderma pigmentosum group F (XPF)/ERCC1 complex and xeroderma pigmentosum group G (XPG). Both are scaffolded by the central TFIIH/XPA/RPA complex, with XPF/ERCC1 being recruited and positioned via XPA and XPG is interacting with XPD early after TFIIH recruitment prior to XPF/ERCC1 [14]. The first incision is performed by the XPF/ERCC1 complex at the 5′ end with respect to the lesion. The 5′ incision then triggers the gap filling machinery consisting of PCNA, RCF, and DNA polymerase δ (or others) to initiate DNA synthesis based on the undamaged template strand [15]. The progression of this complex triggers the 3’ incision carried out by XPG [4] with the damaged DNA fragment leaving the scene bound to TFIIH [16]. The nicked DNA is finally sealed by DNA ligase I or IIIa.

Although progress has been made in other steps of NER, for example structures of the endonucleases XPG and XPF [17, 18, 19] have been solved and provide detailed information on their interaction with DNA, we will focus in this review on the NER core reactions, namely the initial damage detection and subsequent verification steps. These early processes define the overall efficiency and accuracy of the entire pathway and there has been tremendous progress on their structural elucidation.

While crystal structures of damage recognition complexes harboring RNA pol II [20] and the yeast XPC homolog Rad4 [21,22], or UV-DDB [23,24] complexes have been available for quite some time, exciting studies revealed recently new insights into the mechanism of initial damage detection and the subsequent recruitment of downstream factors on the two initial branches of NER, TC- and GG-NER. In TC-NER a lesion stalled RNA Pol II recruits CSB replacing the elongation factor DSIF and thus initiating the repair process [25]. CSB in turn recruits CSA which recruits UVSSA rendering CSB the important central player for TC-NER [7]. In addition, CSA complexes with the DNA damage binding protein 1 (DDB1), cullin 4A (CUL4A), and the RING box protein 1 (RBX1), building the ubiquitin E3 ligase CRL4CSA have been solved. Kokic and co-workers obtained 5 individual structures of RNA Pol II with different factors that shed unprecedented insights into the molecular requirements of factor recruitment and how processes like RNA Pol II pushing and ubiquitylation are facilitated and coordinated [25]. Of the five structures two represent initial recognition and recruitment complexes that will be discussed here (Figure 2). RNA Pol II can not only be stalled by DNA damages but also due to specific sequences or bound nucleosomes [20]. The SF 2 translocase CSB, when recruited upstream of RNA Pol II, aims to push Pol II forward and if the obstacle can be bypassed, RNA synthesis resumes. First structural evidence on these molecular principles has been observed in yeast [26] using Rad26, the yeast ortholog of CSB and RNA Pol II. Kokic et al. could show further how CSB with its two ATPase lobes engages productively with DNA, resulting in translocation in the context of RNA Pol II, CSA and UVSSA via a feature called pulling hook in a pre- and post-translocational state (Figure 2b,c) [25]. If the encountered obstacle cannot be bypassed CRL4CSA ubiquitylates RNA Pol II [27] and TFIIH is recruited via UVSSA [28] aided by ELOF1 [29,30]. Once the damage has been repaired, CSB is ubiquitylated again by CRL4CSA and RNA synthesis can be resumed. The missing piece in this puzzle is how TFIIH is recruited and what actions it takes subsequently to recruitment. In a recent study, one of TFIIHs major recruitment factors was investigated, the p62 protein. P62 engages with the XPD, p44, and p34 subunits of TFIIH. The N-terminus of p62 is highly flexible and interacts with different factors such as UVSSA and XPC but also with TFIIE and p53 thus managing the functional fate of TFIIH [31]. The high flexibility of p62 is likely a key component of the differential recruitment and positioning of TFIIH in TC-NER, GG-NER, and transcription.

One of the major differences between GG-NER and TC-NER is substrate accessibility. Whereas the damaged region can be readily approached in TC-NER, the GG-NER pathway faces obstacles due to packing constraints of the DNA into the chromatin structures. A structural glimpse of how this problem is tackled was recently provided by structures of UV-DDB bound to damages in a nucleosome complex [32]. Damages located in the accessible minor groove were bound without altering the nucleosome structure, analogously to what has been observed for UV-DDB on standard DNA substrates [23]. However, buried damages could also be located by UV-DDB using a mechanism that was called slide assisted exposure, essentially utilizing regularly occurring nucleosome register dynamics to stabilize the lesion in an accessible environment.

After initial detection, structural information on how TFIIH can be recruited to lesions in GG-NER was provided through the analysis of the XPC complex from yeast (Rad4/Rad23/Rad33) probing a lesion containing DNA in complex with yeast core TFIIH where XPB is engaged with a region of the double stranded DNA 5’ to the lesion (Figure 3a) [33]. The structure likely represents an early intermediate in TFIIH recruitment since XPD has not yet engaged with the DNA to achieve further DNA opening and damage verification (Figure 4). The quality of the XPC complex density did not warrant atomic modeling from the cryo EM map, necessitating the placement of a previously solved crystal structure that matched the density [34]. The β-hairpin emerging from the BHD3 domain probes for the damage and, despite the low resolution, a further DNA opening of the lesion bubble dependent on TFIIH (specifically XPB) could be observed. The positioning of XPB on the DNA is similar to other DNA bound XPB structures obtained from transcriptional complexes (reviewed by Kuper et al. [35]), suggesting that XPB acts in a comparable way in NER and transcription [4,33,35].

In this early intermediate it is unclear, however, how additional factors will further progress the NER cascade and how they interact with this complex. To address this problem AI based approaches may be of significant importance. Since the development of alphafold [36] AI based predictions of protein structures and complexes have become increasingly accurate and more reliable. In a recent study, a large number of binary complexes for the yeast proteome were predicted and it was also suggested that the yeast orthologs for Centrin-2 and XPA might form a complex [37]. This analysis rationalizes that XPA and the XPC complex can be present at the lesion at the same time [38]. Although the superposition of the structures in which XPA is bound to TFIIH [39] and the XPC- TFIIH complex [33] leads to some clashes of the molecules involved, the individual subunits, specifically XPA (alphafold model) provides sufficient flexibility to overcome this issue (Figure 3b) [37]. This analysis provides an excellent example how progression and handover in complex cascades could be tackled by a combination of available biochemical data, structural data and AI based structure predictions to advance our knowledge on these pathways.

The major factor in damage verification is the TFIIH complex. It harbors the activities to increase the size of the lesion bubble and provides the means for damage verification. Since TFIIH is involved in transcription initiation and NER, structures from both fields have greatly enhanced our molecular understanding on the inner workings of TFIIH. The first glimpse on a repair competent TFIIH complex was caught by a structure of TFIIH engaged with a Y-fork DNA substrate in complex with XPA [39] (Figure 4a). This structure indicates remarkable flexibility of the subunits within TFIIH as they undergo major conformational changes in comparison to apo TFIIH [40], TFIIH involved in transcription [41], or early in recruitment during lesion recognition [33](Figure 4b). Since the conformation of XPB in TFIIH is retained in NER, transcription, or the apo form [35] it can be viewed as an anchor point. The stability of the XPB conformation could be mainly mediated by the tight interaction with the p52/p8 subunits of TFIIH that also regulate XPB activity [10,40,42]. In contrast, the position of XPD differs in all three structures, thus reflecting the different properties of XPD in transcription and NER. In transcription, XPD merely acts as a scaffold, whereas its helicase activity is essential for NER [43]. The activity of the XPD helicase is regulated by the differential influence of three proteins. Whereas the TFIIH subunits p44 and p62 are activators [44,45], MAT1 from the CAK complex acts as a repressor [46,47]. Engagement with the single stranded DNA of the Y-fork substrate used in the analysis by Kokic et al. requires a repositioning of the XPD protein in the NER complex. The DNA is thereby threaded via the XPD RecA like domains towards a pore like feature closed by the Arch domain (Figure 4c). The Arch domain is likely highly mobile and its movement may be linked to ATP hydrolysis and could be involved in substrate unwinding [46]. The structural information obtained so far provides a glimpse towards the enigmatic damage verification step. After engagement of TFIIH with the XPC complex [33], XPB may use its translocase activity to widen the bubble around the lesion in analogy to transcriptional bubble opening. The XPC complex, connected to TFIIH via p62, could serve as the anchor point that enables torsion-based DNA opening via the XPB translocase. Intriguingly, DNA opening could be boosted by the presence of XPA. The Y-fork probing β-hairpin of XPA is directly located at the bubble opening towards XPB. Threading of the DNA towards this β-hairpin may be causative for strand separation, providing the β-hairpin with the function of a helicase wedge (Figure 4c). This would render the XPB/XPA pair an unusual helicase complex that could explain early observations of XPB based TFIIH helicase activity [48]. Upon increase of the bubble, XPD can engage with the DNA but requires the dissociation of the CAK for its activity, which is achieved through the interaction of XPA/RPA with TFIIH [46]. The strand separation activity of XPD is then used for further opening and damage verification [11, 12, 13] through a yet unknown mechanism. A study using all atom molecular dynamic simulations on XPD in the presence of undamaged ssDNA and ssDNA containing a 6,4-PP suggests how XPD locks down the pore in the presence of a damage thereby inducing stalling [49]. However, further simulations and structural studies on XPD in the presence of damaged DNA are required to substantiate these initial findings. Similar events could be envisioned for the TC-NER pathway, however, it is still unclear how further repair bubble opening takes place. RNA Pol II might not function as an anchor for XPB translocase mediated opening since it is pushed back. On the other hand, further opening might not be necessary if the bubble created by RNA Pol II is of sufficient size for XPD to engage. It is assumed that XPD damage verification requires the protein to be stalled on the DNA and the now “locked” TFIIH/XPA/RPA complex provides a stable scaffold for the nucleases to proceed with the incision process [50,51].

Although tremendous progress has been made in the context of NER, there are still many open questions that need to be addressed to obtain a complete understanding of the events of the core NER processes. How does XPD verify the lesion? What are the structural requirements of full TFIIH activation? And ultimately what re-arrangements are required for all players in the game forming the pre-incision complex? It will be fascinating to witness how these questions will be approached in future studies by using structural biology in combination with the ever increasing power of AI predictions.