Nucleotide excision repair in Human cell lines lacking both XPC and CSB proteins

Abstract Nucleotide excision repair removes UV-induced DNA damage through two distinct sub-pathways, global repair and transcription-coupled repair (TCR). Numerous studies have shown that in human and other mammalian cell lines that the XPC protein is required for repair of DNA damage from nontranscribed DNA via global repair and the CSB protein is required for repair of lesions from transcribed DNA via TCR. Therefore, it is generally assumed that abrogating both sub-pathways with an XPC−/−/CSB−/− double mutant would eliminate all nucleotide excision repair. Here we describe the construction of three different XPC−/−/CSB−/− human cell lines that, contrary to expectations, perform TCR. The XPC and CSB genes were mutated in cell lines derived from Xeroderma Pigmentosum patients as well as from normal human fibroblasts and repair was analyzed at the whole genome level using the very sensitive XR-seq method. As predicted, XPC−/− cells exhibited only TCR and CSB−/− cells exhibited only global repair. However, the XPC−/−/CSB−/− double mutant cell lines, although having greatly reduced repair, exhibited TCR. Mutating the CSA gene to generate a triple mutant XPC−/−/CSB−/−/CSA−/− cell line eliminated all residual TCR activity. Together, these findings provide new insights into the mechanistic features of mammalian nucleotide excision repair.


INTRODUCTION
UV damage in the form of cyclobutane pyrimidine dimers (CPDs) and 6-4 p yrimidine-p yrimidone photoproducts [ PPs] is removed by nucleotide excision repair (excision repair) in humans. The hereditary condition called Xeroderma Pigmentosum (XP) is characterized by increased sensitivity to sunlight and high incidence of skin cancers ( 1 ), and genetic analyses have identified seven genes, XPA to XPG , that are involved in the pathogenesis of this condition due to their role in nucleotide excision repair ( 2 ). In addition, two genes, CSA and CSB , have been shown to participate in a related genetic disorder, Cockayne Syndrome (CS), with photosensitivity resulting from deficient excision repair of UV-induced damage specifically within acti v ely transcribed DNA (3)(4)(5). In the years following identification of the XP and CS genes, the corresponding proteins were purified, and nucleotide excision repair was characterized in some detail ( 6 , 7 ). There are two mechanistic pa thways: global repair tha t depends on XPA to XPG and transcription-coupled repair (TCR) that depends on these same factors, with the exception of XPC, and also r equir es CSA and CSB and se v eral other proteins ( 8 ). Global excision repair has been reconstituted with XPA, RPA, XPC, TFIIH, XPF-ERCCI and XPG in vitro (9)(10)(11), and in this system dual incisions are made to excise the damage in a ∼26-nt-long oligomer. TCR, which acts solely on the transcribed strand independently of the presence or absence of the XPC protein, has not yet been reconstituted in vitro and thus the mechanism is less well understood ( 12 , 13 ).
We embarked on this study because our earlier analysis of the repair maps from an XPC-deficient patient-deri v ed cell line (XP-C) re v ealed that although global r epair was gr eatly diminished, it was not eliminated ( 14 ). TCR, which is defined as gr eater r epair of the tr anscribed str and (TS) than the non-transcribed strand (NTS) of genes acti v ely transcribed by Polymerase II (RNAPII), was clearly observed in the XP-C cell line. Although low in abundance, the presence of excised oligonucleotides mapping to intergenic regions and the NTS of genes did not agree with the univ ersally accepted vie w that XPC mutants do not perform global repair. Since RNA-seq efforts have led to the view that 'it appears that almost the entire genome is expressed as RNA' ( 15 ), it is concei vab le that the 'global repair' seen in the NTS of genes and intergenic regions in XP-C cells was due to TCR from unannotated non-coding or spurious RNAPII transcription. To test this e v entuality, we constructed an XPC −/ − / CSB −/ − double mutant cell line with the expecta tion tha t this combina tion would elimina te all repair. We unexpectedly found out tha t, albeit a t grea tly diminished le v els compared to wild-type counterparts, this ne wly generated doub le knockout cell line performs TCR in the absence of CSB and global repair in the absence of XPC.
We proceeded to construct two additional XPC −/ − / CSB −/ − cell lines (Supplementary Figure S1), and tested the repair activity in all the cell lines by UV survival, slot blot analysis of CPD and (6-4)PP remov al, and b y the in vivo excision assay and e X cision R epair -seq uencing (XRseq), which produces a single-nucleotide resolution map of r epair genome-wide. In agr eement with a pr evious r eport ( 16 ), we found that XPC −/ − / CSB −/ − cells were extremely sensiti v e to UV and had essentially undetectable repair activity by the slot blot and in vivo excision assays. However, w hen anal yzed for repair with the sensiti v e XR-seq assay, w hich directl y ca ptures and identifies the excised oligomers to measure repair throughout the genome ( 14 , 17 ), we found that these double mutant cells carry out TCR comparable to XPC mutant cells in terms of the TS / NTS r epair ratio, and T CR was completely eliminated in a triple knockout XPC −/ − / CSB −/ − / CSA −/ − cell line. Quantitati v e spike-in XR-seq experiments allowed us to determine that nucleotide excision repair in XPC −/ − / CSB −/ − cells was a pproximatel y 300-fold less efficient than in wildtype cells and near the limits of detection in the triple knockout cells. Thus, it appears from a mechanistic standpoint that human cells can perform both global repair and TCR in the absence of XPC and CSB. While the physiological relevance of this drastically reduced repair activity remains to be demonstrated, these findings, ne v ertheless, provide a different perspecti v e to the molecular mechanism of human nucleotide excision repair.

Biological r esour ces
The normal human skin fibroblast (NHF1) and patientderi v ed mutant human skin fibroblast cell lines, XP-C (GM15983) and CS-B (GM16095) cell lines were obtained and maintained as previously described ( 14 ). C lustered R egularly I nterspaced S hort P alindromic R epeats (CRISPR)-Cas9 technology was used to generate mutant XPC, CSB, and CSA cell lines (Supplementary Figure S2), and the primers and lentivirus constructs used for the gene-targeting can be found in Supplementary  Table S1.

Survival, slot-blot, e x cision and XR-seq assays
Survival and slot blot assay procedures have been described previously ( 18 ). For all in vivo excision and XR-seq experiments, cells were harvested 2h after treatment with 20 J / m 2 of UVC. XR-seq was performed as previously described ( 14 ), except one hundred times more starting material was used for the XPC −/ − / CSB −/ − cell lines which was achie v ed by combining fiv e batches of twenty 15 cm plates. qXRseq was conducted by adding 5ul of S2 Hirt lysate [diluted 1:1000 from one 15 cm plate of S2 cells, previously described ( 19 ), irradiated with 20 J / m 2 of UVC and harvested after 1h] to the Hirt lysate from the human cells. All XR-seq and qXR-seq experiments were performed at least two times and r epr esentati v e results are shown.

Statistical and data analyses
Analysis of sequencing reads and data visualization was performed as described previously ( 20 ). Reads were trimmed to remove flanking adapter sequences by cutadapt ( 21 ), and then duplicate r eads wer e r emoved by fastx toolkit / 0.0.14 (hannonlab.cshl.edu / fastx toolkit / index.html). Trimmed r eads wer e aligned to hg38 UCSC by using bowtie2 with arguments -f -v ery-sensiti v e ( 22 ). Oligonucleotide lengths and dinucleotide distributions were plotted by R. Only the reads of 24-29-nt in length which contained TT, TC, CC or CT at the expected site of damage were used in the following analysis. For plotting average repair profiles as a unit gene, we chose genes with lengths > 5 kbp, and the distance between genes > 5 kb. The genome distribution of XR-seq uniquely mapped reads was determined by CEAS ( cis-regula tory element annota tion system) ( 23 ) with command line options ceas -g hg38 -b -w -name.

Data availability / sequence data resources
The raw data have been deposited in the Sequence Read Archi v e (SRA) of the National Center for Biotechnology Information (NCBI) under accession number PRJNA933687.

DNA repair in XPC −/ − / CSB −/ − cells
Based on numerous studies using assays of various resolution, it has been generally accepted that XPC mutant cell lines carry out TCR but not global repair (24)(25)(26)(27). Similarly, it has been reported by se v eral groups that CSB mutant cell lines perform global repair but not TCR ( 3 , 4 , 28 ). These conclusions, mostly arri v ed at by using low resolution assays, were more recently confirmed by the XRseq method (Supplementary Figure S3) that directly measures nucleotide excision repair at a genome-wide scale and at single nucleotide resolution ( 14 ). Despite this background, we wanted to find out whether a double mutant of XPC −/ − / CSB −/ − would be completely defecti v e in excision repair as extrapolated from the properties of the individual m utants. Previousl y, an xpc −/ − / csb −/ − mouse strain was constructed and fibroblasts from this strain were tested for UV sensitivity ( 16 ). This double mutant was much more sensiti v e than the individual mutants as predicted. Howe v er, the repair of UV photoproducts was not measured in tha t stud y and ther efor e the pr esence of a low le v el of residual XPC and CSB-independent repair activity remained a possibility.
We decided to investigate whether repair activity could be detected in the double mutant XPC −/ − / CSB −/ − cells using the v ery sensiti v e XR-seq assay to address this e v entuality. Figure 1 A shows genome-wide repair of UV-induced CPDs in the TS and NTS of wild-type (WT) normal human fibroblasts (NHF1) and XP-C and CS-B patient cells as determined by XR-seq. As can be seen from the figure, and in agreement with previous results ( 14 ), there is TCR in WT which is much more pronounced in the XP-C cell line that is lacking global repair; and the CS-B patient cell line showed no TCR. Thus, we decided to knock-out XPC in the CS-B patient cell line (Supplementary Figure S2A Figure S5A).
Since both double mutant CSB −/ − / XPC −/ − cell lines described her e ar e patient-deri v ed and may hav e unstab le genomes due to SV40 immortalization, we considered the  Figure S5B).

Properties of the e x cision products in the repair of mutant cell lines
In normal human cells, nucleotide excision repair removes UV-induced CPDs and (6-4)PPs in the form of 26-29-nt oligomers by dual incisions 19-21 nt to the 5 and 5-6 nt to the 3 of photoproducts ( 14 , 29 ). To find out whether the excision products in the XPC −/ − / CSB −/ − cell lines were produced by the same dual incision mechanism, we analyzed the length distribution (Figure 3 A, Supplementary Figure  S6A) and sequence composition (Figure 3 B, Supplementary Figure S6B) of the XR-seq reads that are mapped to either nuclear or mitochondrial genomes. The mitochondrial DNA fragments (right) and random genomic DNA fragments from unirradiated cells (WT no UV), which are presumable nonspecifically immunoprecipitated (IP) during the XR-seq procedure, do not exhibit the excision repair size distribution or base distribution seen in the excised oligos that map to the nuclear genome (left). The size distribution of excised oligomers in the range of 26-29-nt oligomers and sequence composition of Pyr-Pyr 19-21 nt from the 5 and 5-6 nt from the 3 termini show similar patterns in the WT and mutant cell lines. Taken together our data lead to the conclusion that the XPC −/ − / CSB −/ − cells excise UV photoproducts by the same dual incision mechanism as wild type cells. Figure S7 we analyzed the survival of the cell lines after exposure to different UV doses and found that the double mutant lines are more sensiti v e to UV than the single mutants, as was observed with the mouse csb −/ − / xpc −/ − fibroblasts ( 16 ). To compare the rate of UV-adduct removal in the four NHF1 cell lines, the slot blot method with damage-specific antibodies was used to measure the dynamic loss of total genomic DNA damage. The cells were irradiated with 5 J / m 2 of UVC, and as seen in Figure 4 B, we observed that about half of the CPDs ar e r emoved within 8h in WT NHF1 and within 16h in the NHF1 / CSB −/ − cells. In contrast, both NHF1 / XPC −/ − and NHF1 / XPC −/ − / CSB −/ − cells r equir ed longer than 48h for half of the CPDs to be r epair ed, with the caveat that measurements a t la te timepoints are confounded by dilution because of cell division or cell death. Removal of (6-4)PPs is much faster, with essentially all being removed by 1 h in both WT NHF1 and NHF1 / CSB −/ − cell lines (Supplementary Figure S8A). Only half is removed within 16 h in NHF1 / XPC −/ − cells, and the rate is e v en slower in the double m utant cells. To gether, these results indica te tha t very little repair is occurring in XPC −/ − / CSB −/ − cells.

In Figures 4 A and Supplementary
Another way to dir ectly compar e nucleotide excision repair between cell lines is to radiolabel the excised oligos purified from an equivalent number of cells ( in vivo excision assay). As seen from the gel in Figure 4 C and from the quantitation in Figure 4 D, all three cell lines lacking both XPC and CSB (lanes 5, 7, 9) had le v els of e xcised CPDcontaining oligos close to the background signal seen in the unirradiated control (lane 1). A similar result was observed when (6-4)PP-containing oligos were analyzed (Supplementary Figure S8B, C). We performed titration experiments (Supplementary Figure S9) and determined that diluting WT NHF1 ∼100-fold gi v es a signal equivalent to the double knock out cell lines, which also correlates with the 100-fold more double knockout cells r equir ed to perform XR-seq than for WT cells.

Quantitative XR-seq method
Since nucleotide excision repair in the CSB −/ − / XPC −/ − cell lines was nearly undetectable by standard assays (slot blot and excision assays) but was detectable by XR-seq, we decided to enhance the XR-seq protocol to make it quantitati v e (qXR-seq, Supplementary Figure S10A) in order to measure the relati v e differences in repair between WT and CSB −/ − / XPC −/ − cells. To this end, we spiked in an equal quantity of Hirt extract DNA from Drosophila S2 cells (that had been UV-irradiated and allowed to repair for 1h) into the extracts from UV-irradiated human cells bef ore perf or ming the IP steps of XR-seq. Deter mining the ratio of human to Drosophila excision products allows for direct comparison between samples since the Drosophila extract is constant in all samples. We had to switch from using anti-TFIIH antibodies that were used in the previous experiments to using anti-damage-specific antibodies for the IPs to capture both human and fly excised oligos since the TFIIH antibodies do not recognize the Drosophila protein.
This switch also allowed us to use the denaturing Hirt extraction procedure which is much more efficient than the nondenaturing lysis procedur e r equir ed for the TFIIH IPs, thus having the added benefit of requiring much less starting material. In addition, we also discovered that the XRseq method is generally more efficient for (6-4)PPs than CPDs, which allowed for e v en less starting material, so with all these changes we were able to use 100-fold less starting material for qXR-seq than we had used in the previously  Figure S10B shows the results from a serial dilution of NHF1 using the qXRseq method. Since the ma pped human / fly oligo ratio relati v e to cell number is essentially linear, the amount of excised oligos from different cell lines can be determined relati v e to NHF1, and thus, we were able to determine that nucleotide excision repair in XPC −/ − / CSB −/ − cells is approximately 0.3% of wildtype NFH1 cells. Figure 5 shows the qXR-seq results for repair of UVinduced (6-4)PPs in WT (left) and double knockout XPC −/ − / CSB −/ − cells (middle). The lack of obvious (6-4)PP repair by TCR in WT NHF1 cells was as expected, since (6-4)PPs are very efficiently removed by global repair ( 14 ). In contrast, NHF1 / XPC −/ − / CSB −/ − cells excised (6-4)PP by TCR, similar to what was observed for CPDs (Figure 2 A). Since nucleotide excision repair in the double knockout cells was unexpected, as it is XPC-and CSB-independent, yet exhibits TCR, contrary to consensus view of mammalian excision repair in general and the r equir ement f or CSB f or TCR in particular, we decided to knock out another component of the TCR pathway to assess whether the observ ed TCR acti vity could occur via a novel mechanism. The generally established human TCR mechanism ( 30 , 31 ) consists of damage-stalled RNAPII recruiting CSB translocase and CSA complexed with CRL4 (cullin-ring type E3 ubiquitin ligase 4). This complex then recruits the UVSSA scaffold and ELOF1 which act in concert to ubiquitylate residue K1268 on the largest subunit of RNAPII to ultimately recruit TFIIH and then the other excision repair factors. Thus, we decided to knock out CSA from the NHF1 / XPC −/ − / CSB −/ − cells to generate a triple mutant cell line (Supplementary Figure S2F, S4B). We performed qXR-seq on these cells and found that the le v el of repair was near the bottom limit of the quantitati v e assay at a pproximatel y 0.05% of wildtype NHF1 cells, and that more importantly, the NHF1 / XPC −/ − / CSB −/ − / CSA −/ − cells were totally defecti v e in T CR (Figur e 5 , right). Taken together, these r esults lead us to conclude that, like Drosophila ( 19 ), humans can perform TCR in the absence of CSB, albeit it is very inefficient in human cells. For physiolo gicall y relevant levels of TCR, human cells require the CSA and CSB proteins.

DISCUSSION
In light of the findings reported in this study, we propose the following model for global and transcription-coupled repair in human cells lacking XPC and CSB ( Figure 6 ). In the standard model for global repair, the dual incision comple x is assemb led at the damage site by binding of XPA, XPC and RPA to damage and recruitment of TFIIH which results in unwinding of the duplex around the damage. This is then accompanied by recruitment of the XPG and XPF nucleases to enable the classic dual incision pattern generating 26-27-nt long oligomers. In the standard model of transcription-coupled repair, the dual incision complex is assembled at the damage site after being encountered by RN APII w hich recruits CSB and CSA, and thus initiates a series of e v ents ultimately resulting in the recruitment of TFIIH and the other excision repair factors except XPC.
Our findings at face value seem contradictory to the consensus models for 6-factor (XPA, RPA, XPC, THIIH, XPG, XPF-ERCC1)-dependent global repair and CSBdependent transcription-coupled repair in humans. Howe v er, taking into account the drastically reduced global repair with just fiv e of the six canonical excision repair factors and of TCR in the absence of both CSB and XPC, our data can be reconciled with the standard models when prior findings on excision protein-DNA and protein-protein interactions are taken into account. It has been previously reported that RPA and XPA act cooperati v el y to reco gnize DNA damage ( 32 ) and that XPA interacts with TFIIH ( 33 ).  The two nucleases are recruited by interactions of XPG with RPA ( 32 ) and TFIIH ( 9 ) and XPF-ERCC1 with XPA ( 34 ). Thus through these interactions an XPC-independent dual incision complex assembles at a damage site, similar to the Preincision Complex 3 (PIC3) which assembles with the aid of XPC but does not contain XPC ( 35 ), and carries out the excision reaction. This is consistent with the random assembly and kinetic proofreading model for damage recognition ( 36 , 37 ). Howe v er, this assemb ly and e xcision is not as efficient as the excision complex that forms with the aid of the scaffold XPC protein and consequently, the repair rate is drastically reduced.
With respect to TCR in the absence of CSB, we note two previous in vitro studies: one which showed that RNAPII stalled at a dimer does not inhibit excision repair nor is CSB r equir ed ( 38 ), and a second which showed that a dimer in a 10-nt pseudo transcription bubble (no RNAPII but a CPD at the 5 end of 10 mispaired base pairs) was excised by the conventional dual incision mode (26-27-mer) by 5 repair factors (XPC omitted) ( 39 ). These studies indicate that repair could occur at RNAPII-stalled dimers and that the 'transcription bubble' could replace the XPC-damage In global repair (left), in the absence of XPC, which is r epr esented as a gray dashed oval, the dual incision complex is assembled at a damage site by binding of XP A-RP A to damage and recruitment of TFIIH by its interaction with XPA. Recruitment of the XPG and XPF nucleases enables dual incisions generating a 26-27-nt long oligomer. In TCR (right), in the absence of CSB, r epr esented as a gray dashed oval, the transcription bubble replaces the XPC damage recognition-function and enables the assembly of the 5 excision factors (XPA, RPA, TFIIH, XPG, and XPF). Excision occurs in the absence of XPC and CSB, but it is drastically reduced because assembly of the 5-factor nuclease is inefficient without these two proteins. Adapted with permission from ( 43 ). recognition function and enable the assembly of the 5 excision repair factors (XPA, RPA, TFIIH, XPG, XPF). Thus, it is reasonable to assume that similar reactions could occur at low frequency and manifest as TCR in the absence of XPC and CSB. Howe v er, our observa tion tha t the elimination of both CSB and CSA pr oteins abr ogates all residual TCR indica tes tha t e v en though a low le v el of TCR occurs in the absence of CSB, removal of both proteins eliminates the ability of the cell to perform TCR and thus, the TCR mechanism appears to be more complex than currently appreciated.
A plausib le e xplana tion for the observa tion of TCR in the XPC −/ − / CSB −/ − cell lines but not in the CSB −/ − cell lines is that in the latter there is still substantial global repair mediated by the 6-factor excision nuclease that obscures the drastically reduced TCR in the CSB −/ − cell lines, whereas in the XPC −/ − / CSB −/ − cell lines the 'global repair' is drastically reduced making it possible to observe TCR in the background of very low global repair. We did not observe obvious differences in the pattern of TCR between WT cells and those lacking CSB, which is different than what was observed in Sacchar om y ces cer evisiae , which exhibit r esidual T CR at transcription start sites in the absence of Rad26 / CSB ( 40 ). Of note, yeast have CSB and CSA orthologs ( 41 ), but Drosophila do not, yet still per-form robust TCR ( 19 , 42 ). Yeast and flies are also unique in that they are the only eukaryotes known to r equir e XPC for TCR ( 19 , 41 ). Taken together, these findings provide a unique perspecti v e to the molecular mechanisms of global and transcription-coupled nucleotide excision repair.

DA T A A V AILABILITY
The raw data have been deposited in the Sequence Read Archi v e (SRA) of the National Center for Biotechnology Information (NCBI) under accession number PR-JNA933687.

SUPPLEMENT ARY DA T A
Supplementary Data are available at NAR Online.