Part1 ERCC1 Mutations Impede DNA Damage Repair And Cause Liver And Kidney Dysfunction in Patients

Contact: ali.ma@wecistanche.com

Katja Apelt et al

Click to Cistanche extract benefits for kidney function

Introduction

The six billion base pairs of the human genome are continually exposed to DNA-damaging agents, causing a wide variety of genomic DNA lesions, including UV light-induced photoproducts, intra- and interstrand cross-links (ICLs), and DNA double-strand breaks (DSBs). To ensure genomic integrity, these DNA lesions need to be repaired in a specific and efficient manner. Different DNA repair mechanisms have evolved that each act upon a specific subset of DNA lesions. While many DNA repair proteins are involved in a single pathway, the heterodimeric endonuclease ERCC1-XPF is shared between several mechanistically distinct DNA repair pathways (Manandhar et al., 2015).

XPF is unstable and rapidly degraded, which is prevented by its heterodimerization with ERCC1 (Biggerstaff et al., 1993). The interaction between these two proteins is mediated by a double helix-hairpin helix (HhH)2 motif located in their respective C-termini, as well as through the central domain of ERCC1, and the nuclease domain of XPF (de Laat et al., 1998b; Jones et al., 2020; Tripsianes et al., 2005; Tsodikov et al., 2005). The binding to double-stranded DNA is facilitated by ERCC1, while XPF mediates the association with single-stranded DNA (ssDNA) through its (HhH)2 domain (Tsodikov et al., 2005). Once positioned, XPF employs its catalytic activity through its highly conserved nuclease motif to cleave different DNA substrates at the 59 junctions of bubbles and single-stranded 39 overhangs extending from a DNA double helix (de Laat et al., 1998a; Enzlin and Sch¨ are, 2002; Matsunaga et al., 1996). A key role of ERCC1- XPF was first demonstrated in nucleotide excision repair (NER), where it cleaves the damaged DNA 59 to the lesion (Sijbers et al., 1996a).

NER recognizes a wide variety of structurally unrelated DNA lesions, including UV-induced photoproducts, via its two subpathways: global genome repair (GGR) and transcription-coupled repair (TCR). The bulk of the NER activity relies on GGR, where the XPC protein recognizes photoproducts throughout the genome and recruits the transcription factor IIH (TFIIH) complex (Sugasawa et al., 1998; Volker et al., 2001). DNA lesions that cause stalling of RNA polymerase II (RNAPII) during transcription trigger the recruitment of TCR-specific proteins, such as CSB, CSA, and UVSSA, which in turn recruit the TFIIH complex (Nakazawa et al., 2020; van der Weegen et al., 2020). Following TFIIH recruitment to the DNA lesion, GGR and TCR funnel into a common molecular mechanism involving the association of XPA and XPG (Marteijn et al., 2014), which stimulate the helicase activity of TFIIH and form a stable precision complex together with RPA (Kokic et al., 2019; Riedl et al., 2003; Wakasugi and Sancar, 1998). The ERCC1-XPF heterodimer is recruited to the NER complex by XPA through the XPA-binding domain of ERCC1 (Tsodikov et al., 2007; Volker et al., 2001). A dual incision by the coordinated activity of the endonucleases XPF (59 from the lesion) and XPG (39 from the lesion) releases a stretch of ssDNA containing the DNA lesion (Li et al., 1995; Matsunaga et al., 1995; O’Donovan et al., 1994; Staresincic et al., 2009), after which the generated gap is filled in and ligated to complete repair.

In addition to NER, it was shown that ERCC1-XPF also plays an essential role in ICL repair (Klein Douwel et al., 2014). This repair pathway is initiated by stalled replication and involves recognition of the ICL-stalled replication fork by the multi-protein core Fanconi anemia (FA) complex, which serves as an E3 ubiquitin ligase complex for FANCD2. The mono-ubiquitylation of FANCD2 triggers the recruitment of SLX4, which in turn targets ERCC1-XPF to ICLs (Abdullah et al., 2017; Klein Douwel et al., 2014; Walden and Deans, 2014; Wood, 2010). Once recruited, ERCC1-XPF mediates unhooking of the ICL by incising the parental DNA strand on either side of the lesion to enable subsequent lesion bypass and repair (Kuraoka et al., 2000). Finally, ERCC1-XPF has also been implicated in DSB repair by homologous recombination (Ahmad et al., 2008), during which it can efficiently cleave different recombination intermediates (Wyatt et al., 2017). Alternatively, cells use the error-prone RAD52-dependent single-strand annealing pathway, which involves annealing of regions with microhomology and subsequent removal of the nonhomologous 39 ssDNA tails by ERCC1-XPF (Adair et al., 2000; Li et al., 2013, 2019). However, the precise role of ERCC1-XPF in these DSB repair pathways is not fully understood.

The importance of DNA repair pathways is underscored by the clinical phenotype of individuals with inherited DNA repair–deficiency disorders. Inactivation of GGR results in Xeroderma pigmentosum (XP; MIM 278700, 610651, 278720, 278730, 278740, 278760, and 278780; DiGiovanna and Kraemer, 2012), which is characterized by photosensitivity and a 2,000- fold increased risk of skin cancer. Neurodegeneration, growth retardation, and multisystem disease without skin cancer are observed in Cockayne syndrome (CS; MIM 216400 and 133540). CS patients have a selective genetic defect in TCR that prevents the processing of DNA damage–stalled RNAPII (Karikkineth et al., 2017; Nakazawa et al., 2020; Nakazawa et al., 2012; Ribeiro et al., 2018; Tufegdˇzi´ c Vidakovi´ c et al., 2020). Mutations in essential ICL repair genes cause FA (MIM 227650, 300514, 227645, 605724, 227646, 600901, 603467, 614082, 609053, 609054, 614083, 614087, 610832, 613390, and 613951), which is characterized by anemia, skeletal abnormalities, organ malformations, and genomic instability (Auerbach, 2009). The role of ERCC1-XPF in ICL repair is underscored by the existence of mutations in XPF that give rise to the FA phenotype (Bogliolo et al., 2013; Ceccaldi et al., 2016; Kashiyama et al., 2013). Inherited defects in DSB repair cause a wide range of disorders associated with radiosensitivity, cancer predisposition, immunodeficiency, and neurodegeneration (Helfricht et al., 2020; McKinnon and Caldecott, 2007).

DNA repair–deficiency disorders are rare and occur with an estimated frequency of ∼1 per 200,000 live births worldwide for XP, CS, and FA. However, only two reports of individuals with bi-allelic ERCC1 (MIM 126380) mutations have been described to date (Jaspers et al., 2007; Kashiyama et al., 2013). Both individuals displayed features consistent with CS and died at 14 mo and 2.5 yr. The first and most severely affected individual (165TOR) carried a premature stop codon (Q158X) on one allele and an F231L missense variant on the other (Jaspers et al., 2007). The second individual (CS20LO) was homozygous for the F231L missense variant (Kashiyama et al., 2013). Cells from both individuals showed sensitivity to UV-induced DNA damage, and 165TOR, in particular, was also mildly sensitive to ICL-inducing agents. Both individuals died in early childhood (1–2 yr; Jaspers et al., 2007; Kashiyama et al., 2013). A third individual (XP202DC) with bi-allelic ERCC1 mutations who died at 37 yr of age is cited in a meeting abstract (Imoto, K., et al. 2007. Patients with defects in the interacting nucleotide excision repair proteins ERCC1 or XPF show xeroderma pigmentosum with late-onset severe neurological degeneration. [Abstract] J. Invest. Dermatol. 127:S92). This individual was compound heterozygous for a nonsense mutation (K226X) and a splice mutation in ERCC1. A detailed phenotypic description is not available, and the impact of these ERCC1 mutations is unknown.

The phenotype of the two aforementioned individuals is more severe than that expected from NER deficiency alone, consistent with the involvement of ERCC1-XPF in multiple DNA repair pathways. In line with this, either ERCC1 or XPF-KO mice displayed high embryonic lethality and decreased life span and died due to severe liver failure (McWhir et al., 1993; Tian et al., 2004; Weeda et al., 1997), which is not observed in NERdeficient XPA-KO mice (de Vries et al., 1995; Nakane et al., 1995).

In the current study, we describe two siblings with bi-allelicERCC1 mutations who have a unique phenotype of short stature, photosensitivity, progressive cholestatic liver disease, and renal tubulopathy. Both individuals developed progressive liver impairment and required liver transplantations before the age of 10yr. Functional studies show that the steady-state protein level of ERCC1 and XPF were dramatically reduced in inpatient cells and knock-in epithelial cells carrying the missense variant found in the patients. Additionally, the mutant ERCC1 protein interacted only weakly with other NER and ICL repair proteins. We report a new case of ERCC1 deficiency that strongly affects NER and has a considerable impact on ICL repair, which together result in a unique phenotype combining short stature, photosensitivity, and progressive liver and kidney dysfunction.

Results

Two siblings with photosensitivity, short stature, and progressive liver and kidney dysfunction

Sibling 1 (PV50LD; Fig. 1 A) is 13 yr of age, the eldest of three siblings to healthy, unrelated parents of mixed ethnicity including Indigenous Australian, Maltese, and Anglo-Celtic heritage. During the pregnancy, her mother had a transient dilated cardiomyopathy that resolved postpartum. She was born at 39 wk of gestation, and her birth weight was 1.9 kg (Z = −3.8), length 44 cm (Z = −2.44), and head circumference 29.5 cm (Z = −3.8). She had poor growth in infancy, and at 18 mo of age, she was noted to have liver dysfunction with a predominantly cholestatic pattern. Liver ultrasound and magnetic resonance cholangiopancreatography were normal. Liver function progressively declined, as evident from progressive increases in γ-glutamyltransferase (GGT), alanine transaminase (ALT), and bilirubin levels (Fig. S1, A–C). At a liver biopsy (at age 3.5 yr), the lobular parenchyma showed variation in hepatocyte nuclear morphology, with some much larger nuclei and cells with double nuclei (Fig. S1 D) to areas where the cells have small, unremarkable nuclei. There was mild portal fibrosis and mild fibrous portal expansion and mild focal interface inflammation with no ductopenia or periductal fibrosis (Fig. S1 E).

She experienced a large number of recurrent infections including tonsillitis, chickenpox, hand-foot-and-mouth disease, pneumonia, bronchitis, and recurrent episodes of fever, abdominal pain, and pale stools with no cause found. Immuno-logical investigations did not identify any immunodeficiency. She developed episodes of ocular and skin photosensitivity. At age 6 yr, renal dysfunction was detected, with features suggestive of proximal tubular dysfunction characterized by albuminuria (sub-nephrotic range) and hypercalciuria. Her renal function fluctuated, with intermittent episodes of acute kidney injury, progressive kidney impairment with increasing creatinine levels (Fig. S1 F), and minimal response to acetylcholinesterase inhibition. Renal ultrasound showed small kidneys with increased echogenicity and reduced corticomedullary differentiation. Lung function tests showed mild to moderate restrictive lung disease.

Developmental milestones were normal, but some learning difficulties were evident at school age with no sign of regression. Vision and hearing were normal. Growth remained very slow despite supplemental feeding. She was relatively stable until 9.5 yr of age when she displayed evidence of liver decompensation with a progressively rising bilirubin level (Fig. S1 C) and international normalized ratio. She underwent orthotopic liver transplantation at age 9 yr 10 mo. Following a liver transplant, her tubulopathy has stabilized, although her serum creatinine continues to increase at a slower rate (Fig. S1 F). She is clinically stable. At age 12 yr, ovarian insufficiency was diagnosed. Brain magnetic resonance imaging (MRI) at age 12 yr showed mild cerebral atrophy with moderate cerebellar atrophy and mild brainstem atrophy. At last assessment (aged 13.5 yr), growth was slow (weight 22.4 kg [Z = −6.08] and height 134.7 cm [Z = −3.72]). She had a very slim build with poor muscle bulk and a paucity of subcutaneous fat. She had developed freckling on sun-exposed areas (Fig. 1 A). She had mildly deep-set eyes, and her scalp hair was thin. There were no radial ray abnormalities. Neurological examination showed mild weakness, minimal ataxia, and depressed reflexes.

Sibling 2 (PV46LD; Fig. 1 A) is 11 yr of age, a younger sister to sibling 1. The pregnancy was complicated by placenta previa and transient maternal cardiomyopathy, and sibling 2 was born at 35 wk of gestation with a birth weight of 1.79 kg (third centile), a length of 45 cm (50th centile), and an ahead circumference of 29 cm (less than the third centile). She had a paucity of subcutaneous fat and similar facial features to her older sibling. She exhibited a failure to thrive in the first year of life and liver impairment from age 2 yr, with significantly increased GGT, ALT, and bilirubin levels (Fig. S1, A–C). Liver biopsy at age 6 yr showed damage to the intrahepatic bile ducts, resulting in periductal fibrosis. As with the liver biopsy from her sister, a moderate number of double-nucleated hepatocytes was seen, some with large nuclei and large nucleoli.

She had episodes of ocular and skin photosensitivity. She had findings of a renal tubulopathy with mild renal impairment, with progressively increasing creatinine levels (Fig. S1 F). Renal ultrasound showed small kidneys with nephrocalcinosis. Developmental milestones were normal, but a mild intellectual disability (IQ 66) was diagnosed at school age. There has been slow forward progress with neurodevelopmental milestones and no definite regression. Vision and hearing were normal. The liver impairment was progressive, and at age 8 yr, she underwent liver transplantation. Brain MRI at age 5 yr was normal, but a repeat MRI at 10 yr showed moderate cerebellar atrophy and mild cerebral atrophy. At age 11 yr, measurement of estrogen, follicle-stimulating hormone, and luteinizing hormone showed a pattern suggestive of ovarian insufficiency. At last assessment at age 11 yr, growth was slow (weight 20 kg [Z = −3.89] and height 120.7 cm [Z = −3.17]). She had a very thin build with minimal subcutaneous fat and poor muscle bulk, familial facial features, mildly deep-set eyes, and freckling on sun-exposed areas (Fig. 1 A). She had no radial ray abnormalities. Neurological examination showed a mild weakness with mild ataxia and absent reflexes.

The molecular analysis identifies novel bi-allelic ERCC1 mutations

Exome sequencing revealed that both siblings harbor a novel missense variant in exon 4 of the ERCC1 gene (p.R156W; c.466C>T) on the paternal allele (Fig. S2 A). The presence of the c.466C>T missense variant was confirmed by Sanger sequencing in both affected siblings (Fig. 1 B). Whole-genome sequencing revealed a deletion in exon 4 on the maternal allele (hg19: chr19:45,922,224- 45,924,375; Fig. S2 B), which was confirmed by comparative quantitative PCR (qPCR) of exon 4 and exon 5 of ERCC1 in both siblings and the mother (Fig. 1, C and D). The deletion, which is expected to be a null allele, was detected in neither the father nor in four negative control samples (Fig. 1 D).

While a previously described pathogenic ERCC1 missense variant (p.F231L; c.693C>G) is located in the (HhH)2 domain (Fig. 1 E), the missense variant (p.R156W) is located within the central domain of ERCC1 and is in close proximity to the XPAbinding pocket (110–154 aa) of ERCC1 (Fig. 1 E). Structural analysis of the ERCC1 central domain revealed a narrow V-shaped hydrophobic pocket that binds a short motif (67–80 aa) present in XPA (Fig. 1 F; Tsodikov et al., 2007). Amino acids that line the XPA-binding pocket in ERCC1 (Fig. 1 F) are important for NER but dispensable for other ERCC1-dependent repair pathways (Orelli et al., 2010). The R156 residue that is substituted in the patients is located just below the XPA-binding pocket and forms a salt bridge with the opposing amino acid D129 (Fig. 1 F). The R156W substitution therefore likely affects the stability of the XPA-binding pocket. It is possible that this substitution also weakens the interaction between the central domain of ERCC1 and the nuclease domain of XPF (Jones et al., 2020), and we would predict that this could lead to a mild to moderate destabilization of the heterodimer interface, resulting in reduced stability of the protein.

PV46LD and PV50LD fibroblasts display a severe NER defect A hallmark of NER deficiency is a strong sensitivity to UV-C irradiation. To address the impact of the ERCC1 deficiency on NER function, we obtained primary fibroblasts from skin biopsies of both affected siblings. To allow clonogenic survival assays, we immortalized PV50LD cells by introducing hTERT (Fig. S3 A). As a control, we generated full ERCC1-KO cells in RPE1-hTERT cells by CRISPR-Cas9 (Fig. 2 A). Clonogenic UV-C survival assay showed that PV50LD-hTERT cells are hypersensitive to UV-C irradiation, although not to the same extent as full ERCC1-KO cells (Fig. 2 B), suggesting a severe defect in NER.

UV-induced DNA lesions encountered during transcription are repaired by TCR, which can be measured by the recovery of RNA synthesis (RRS) assay (Nakazawa et al., 2010). We measured RRS in 48BR fibroblasts (WT), patient fibroblasts, and included XPA-deficient (XP1PD) primary fibroblasts as a control. Labeling of nascent transcripts by 5-ethynyl-uridine incorporation showed a strong UV-induced decrease in nascent transcripts in all cell lines at 3 h after UV-C. While 48BR fully recovered transcription within 18 h after UV-C irradiation, both patient fibroblasts recovered even less than the NER-deficient XP-A patient cells (Fig. 2, C and D), suggesting that TCR is virtually absent under these conditions.

To examine GGR activity, we measured unscheduled DNA synthesis (UDS) in nondividing cells (Nakazawa et al., 2010). To this end, primary fibroblasts were locally irradiated with UV-C light followed by pulse-labeling with the thymidine analog 5-ethynyl-deoxyuridine (EdU) to measure repair. Clear EdU incorporation was detected in 48BR cells at sites of local UV-induced DNA damage. However, only very low levels of EdU incorporation were detectable in the patient fibroblasts at similar levels as in XPA-deficient fibroblasts (XP1PD; Fig. 2, E and F). The UDS levels in PV46LD and PV50LD cells were comparable to, or even lower than, the levels measured in the previously described ERCC1-deficient 165TOR (Jaspers et al., 2007) and CS20LO (Kashiyama et al., 2013) cells, which we included in parallel (Fig. 2 G and Fig. S3 B). Importantly, reexpression of menus-tagged ERCC1 in both fibroblasts fully rescued EdU incorporation (Fig. 2, E and F; and Fig. S3 C), confirming that the strong UDS defect is due to ERCC1 deficiency. We also obtained primary fibroblasts from the parents, which displayed normal UDS at sites of local UV damage (Fig. 2 H), suggesting that the respective heterozygous ERCC1 defect in the parents is fully compensated by the WT allele.

PV46LD and PV50LD fibroblasts have very low ERCC1 and XPF protein levels

NER involves the highly coordinated and sequential assembly of DNA repair complexes, during which XPA recruitment occurs downstream of TFIIH but upstream of ERCC1-XPF (Volker et al., 2001). To address to which extent the patient fibroblasts still support NER complex assembly, we locally irradiated fibroblasts with UV-C and monitored the recruitment of several core NER proteins by immunofluorescent labeling (Fig. 3 A, quantification, and Fig. S3 D). Clear recruitment of TFIIH, XPA, and ERCC1 at sites of local UV-induced DNA damage was detected in 48BR cells, while XPA-deficient XP1PD cells failed to recruit ERCC1 as previously described (Fig. 3 A and Fig. S3 D; Volker et al., 2001). Both PV46LD and PV50LD cells showed normal recruitment of TFIIH and XPA, while ERCC1 recruitment to sites of local UV-induced DNA damage was undetectable (Fig. 3 A and Fig. S3 D).

The loss of ERCC1 localization could be due to a failure of ERCC1 recruitment, a general reduction in ERCC1 protein levels, or a combination of the two. Western blot analysis indeed revealed that the steady-state levels of both ERCC1 and XPF were dramatically reduced in PV46LD and PV50LD cells (Fig. 3 B). Residual ERCC1 expression was, however, still detected when compared with full ERCC1-KO cells (Fig. 3 B). The reduction in ERCC1 and XPF protein levels in PV46LD cells was highly similar to the reduced levels detected in either CS20LO or 165TOR cells (Fig. 3 C and Fig. S3, E and F), suggesting that the more severe phenotype in the individual from which these cells were derived is not due to a stronger impact on protein stability. In line with our UDS results, we detected normal ERCC1 and XPF expression in cells from the parents, indicating compensation by the WT allele (Fig. 3 D). Immunofluorescent labeling confirmed that ERCC1 and XPF protein levels were severely reduced, while cells from the parents were indistinguishable from WT cells (Fig. 3 E). Quantification of Western blot data shows that ERCC1 protein levels were reduced to ∼20% inpatient fibroblasts, also resulting in similarly reduced levels of XPF levels (∼20%; Fig. 3 F).

Knock-in of the R156W missense variant leads to ERCC1-XPFprotein instability and a severe NER defect

It is tempting to speculate that the low steady-state protein levels of ERCC1-XPF and the strong NER defect in the patient fibroblasts are caused by the R156W amino acid substitution inERCC1. However, considering that the affected individuals are siblings, we cannot exclude that other genetic features shared between them may also contribute to this phenotype.

To directly assess this, we decided to generate knock-in lines (KIs) carrying the R156W amino acid substitution in RPE1- hTERT cells using CRISPR-Cas9 technology. We obtained homozygous KI clones carrying the patient mutation in ERCC1 (p.R156W; c.466C>T), which was confirmed by Sanger sequencing in two individual clones (Fig. 4 A). The missense mutation caused a severe reduction of ERCC1 and XPF protein levels in both KI clones (clones 2–17 and 2–51; Fig. 4 B and Fig. S4, A and B). Quantification of the Western blot data showed that the ERCC1 and XPF protein levels were almost as low as in ERCC1-KO cells (Fig. 4 C). These findings establish that the R156W amino acid substitution has a strong impact on the stability of ERCC1.

To further establish that the NER defect in patient fibroblasts is caused by the R156W amino acid substitution in ERCC1, we measured TCR activity by monitoring transcription restart after UV irradiation. While parental RPE1-hTERT cells showed normal restart at 18 h after UV, both R156W-KI clones showed a strong TCR defect comparable to full ERCC1-KO cells (Fig. 4, D and E). Similarly, UDS experiments revealed normal GGR activity in parental RPE1-hTERT cells, which was severely reduced below 10% in both R156W-KI clones and ERCC1-KO cells (Fig. 4, F and G). Finally, we measured the UV sensitivity of one R156W-KI clone (2–51) in comparison to the hTERT-immortalized patient fibroblast PV50LD. The R156W-KI cells showed a strong UV-sensitive phenotype, although not as sensitive as full ERCC1-KO cells (Fig. 4 H). Interestingly, the UV-sensitive phenotype of the R156W-KI was very similar to that of the PV50LD-hTERT patient fibroblasts. These findings establish that the low ERCC1-XPF protein levels and the strong NER defect are a direct consequence of the R156W amino acid substitution in ERCC1.

ERCC1R156W substitution causes partial cytoplasmic localization

The patient fibroblasts and the KI cells display dramatically reduced protein levels and strongly reduced NER activity. We next wished to explore whether the NER defect is only caused by reduced ERCC1-XPF protein levels, or whether the ERCC1R156Wmutant protein has reduced activity as an endonuclease or in NER.

To address this, we generated ERCC1-KO cells by CRISPRCas9 in U2OS cells equipped with the Flp-In/T-Rex system. Using Flp-based site-directed recombination, we targeted cDNAs encoding GFP-tagged ERCC1WT or ERCC1R156W to the genomic Flp recognition target (FRT) site to enable inducible expression from a strong viral promoter. Western blot analysis showed that ERCC1WT and ERCC1R156W were expressed at similar levels and that expression of either protein rescued the reduced protein levels of XPF observed in ERCC1-KO cells (Fig. 5 A). These cells are therefore an excellent model system to disentangle the specific impact of the ERCC1R156W substitution when expressed at WT levels from the impact of lower ERCC1 expression in the patient fibroblasts and KI epithelial cells.

Previous studies have shown that missense mutations in XPF can cause mislocalization of ERCC1-XPF in the cytoplasm (Ahmad et al., 2010). In line with this, we also detected that ∼40% of the ERCC1R156W protein pool mislocalized to the cytoplasm, which was only ∼15% in ERCC1WT-expressing cells (Fig. 5, B and C). This mislocalization of ERCC1 also increased the fraction of cytoplasmic XPF (Fig. 5, B and C), indicating that ERCC1R156W may be partially misfolded.

To test this possibility, we performed immunoprecipitation experiments on ERCC1WT-GFP and ERCC1R156W-GFP followed by quantitative label-free mass spectrometry (MS) to map differential interactors of the two ERCC1 proteins. Importantly, quantitative MS confirmed that equal amounts of ERCC1WT-GFP and ERCC1R156W-GFP protein were pulled down (Fig. 5 D). Strikingly, the interactome of ERCC1R156W was enriched for heatshock proteins (HSPs), in particular of the HSPA family, while the interaction with XPF was quantitatively reduced compared with ERCC1WT (Fig. 5 D). The HSPA proteins are protein-folding chaperones that prevent aggregation of misfolded proteins (Stetler et al., 2010). Co-immunoprecipitation (Co-IP) experiments indeed confirmed that ERCC1R156W-GFP strongly interacted with HSPA4 compared with ERCC1WT (Fig. 5 E), suggesting partial misfolding of ERCC1R156W.

To address to which extent ERCC1R156W-XPF is still active as an endonuclease, we purified the recombinant complex from Sf9 insect cells (Fig. S4 C). Our previously established procedure involves a gel-filtration step that allows us to assess whether ERCC1-XPF is in a dimeric or aggregated state (Fig. 5 F, fractions 2 and 1, respectively). The yield of the recombinant ERCC1R156WXPF complex was lower and showed a higher level of aggregated protein. Moreover, we also noticed a reduction in the fraction of heterodimeric protein compared with ERCC1WT-XPF (Fig. 5 F). Nonetheless, the heterodimeric recombinant ERCC1R156W-XPF in fraction 2 was fully active when incubated with a stem-loop model DNA substrate, while the catalytically inactive ERCC1-XPFD687A, included as a negative control, was devoid of incision activity (Fig. 5 G). These findings suggest that ERCC1R156W-XPF is partially misfolded and mislocalized, but the protein fraction that folds properly is still active as an endonuclease.

The ERCC1R156W mutant protein fails to efficiently interact with core NER factors

We next addressed how the ERCC1R156W substitution affected NER using the reconstituted ERCC1-KO cells. Because the R156W substitution is located just underneath the XPA-binding pocket in ERCC1 (Fig. 1 F), we asked whether ERCC1R156W could still interact with NER proteins in response to UV irradiation. To test this, we immunoprecipitated ERCC1WT-GFP, ERCC1R156W-GFP, or GFP fused to a nuclear localization signal (GFP-NLS) as control from UV-irradiated cells and subsequently performed label-free MS. By comparing ERCC1-GFP with GPF-NLS, we identified constitutive as well as UV-induced interactors of ERCC1. Our analysis revealed that both ERCC1WT and ERCC1R156W interacted with XPF and the ICL repair-specific scaffold proteins SLX4 and SLX4IP, albeit quantitatively reduced in ERCC1R156W-expressing cells (Fig. 6, A and B). Strikingly, the UV-induced interaction with the TFIIH subunit XPB in ERCC1WT cells was completely lost in ERCC1R156W cells (Fig. 6, A and B). Co-IP experiments confirmed that ERCC1R156W failed to associate with either TFIIH or XPA in response to UV irradiation, while robust interactions were detected after pull-down of ERCC1WT (Fig. 6 C and Fig. S4, D and E).

The recruitment of ERCC1-XPF to NER complexes is fully dependent on XPA (Volker et al., 2001), raising the question of whether ERCC1R156W is still recruited to sites of UV-induced DNA damage. To address this, we monitored the recruitment of ERCC1-GFP to UV-induced DNA lesions using live-cell imaging. To this end, cells were irradiated with a UV-C (266-nm) laser, which triggered the rapid recruitment of ERCC1WT-GFP within the first 60 s following irradiation, reaching a plateau around 120 s (Fig. 6, D and E). In contrast, only very weak recruitment of ERCC1R156W-GFP could be detected upon UV-C irradiation using identical conditions (Fig. 6, D, and E). These findings suggest that the inability of ERCC1R156W to interact with XPA severely limits its association with the NER complex.

Our findings in reconstituted ERCC1-KO cells suggest that ERCC1R156W still supports residual (∼40%) repair activity when expressed at similar levels as ERCC1WT. To attempt to measure residual repair in the PV46LD and PV50LD primary fibroblasts, which showed severely reduced expression of the ERCC1R156W mutant protein, we performed UDS experiments in which we allowed cells to incorporate EdU for extended periods of time (4 h) to capture residual repair. While UDS was still below ∼10% in XPA-deficient XP1PD cells, in PV46LD and PV50LD cells UDS increased from ∼10% to ∼20% UDS under these conditions (Fig. 6 J and Fig. S5 A). These findings show that the ERCC1R156W mutant protein, expressed at low levels in patient cells, does support residual NER activity, likely explaining the mild XP-like clinical features seen in both siblings. Under these conditions, we also detected residual UDS in previously described ERCC1- deficient 165TOR (Jaspers et al., 2007) and CS20LO (Kashiyama et al., 2013) cells at even higher levels than detected in PV46LD and PV50LD cells (Fig. 6 J and Fig. S5 A).

The impact of ERCC1R156W in ICL repair

During ICL repair, the SLX4 scaffold protein recruits ERCC1-XPF to perform the unhooking incisions that enable subsequent repair. Our MS analysis revealed that ERCC1R156W still interacted with SLX4, albeit at reduced levels compared with ERCC1WT (Fig. 6, A and B). Co-IP experiments indeed confirmed a strong interaction of ERCC1WT with SLX4, which was not affected by UV irradiation, while the interaction with ERCC1R156W was strongly reduced (Fig. 7 A). To address if this reduced interaction affected the recruitment of ERCC1 to ICLs, we locally irradiated cells with a UV-A laser (365 nm) in the presence of trioxsalen, which is a psoralen derivative that forms ICLs upon UV irradiation (Velimezi et al., 2018). Local irradiation with the UV-A laser triggered the recruitment of endogenous FANCD2 only in cells sensitized with trioxsalen, demonstrating that ICLs were induced under our conditions (Fig. S5 B). We could detect strong recruitment of GFP-ERCC1WT-GFP to local ICLs, while the recruitment of ERCC1R156W was much weaker (Fig. 7, B and C), suggesting that the decreased interaction with SLX4 also lowers the efficiency with which ERCC1R156W is recruited to ICLs. However, clonogenic survival assays after exposure to the ICLinducing agent mitomycin C (MMC) showed that ERCC1R156W rescued the hypersensitivity of ERCC1-KO cells to MMC, although slightly less than ERCC1WT (Fig. 7 D). These experiments show that the ERCC1R156W mutant protein supports ICL repair close to WT levels when the mutant protein is expressed near-normal levels.

To address whether PV46LD and PV50LD patient fibroblasts, which showed substantially reduced expression of ERCC1R156W, still supported efficient ICL repair, we performed an MMC-induced chromosome breakage assay. This method measures MMC-induced chromosome breaks in metaphase cells caused by a deficiency in ICL repair. While 48BR (WT) cells accumulated very few breaks after MMC, we measured significant MMC-induced chromosome breakage in Fanconi patient-derived WK8103 cells (Poll et al., 1984; Fig. 7 E). Both PV46LD and PV50LD showed an intermediate phenotype with increased break formation in response to MMC, although not to the same extent as Fanconi cells (Fig. 7 E). In line with this, PV50LD-hTERT cells and RPE1-hTERT R156W-KI cells were sensitive to MMC in clonogenic survival assays, while FANCF-deficient VU121F-hTERT cells (Joenje et al., 1997) and especially ERCC1-KO cells were hypersensitive to MMC (Fig. 7 F). In conclusion, the lower expression of ERCC1R156W and the decreased interaction with SLX4 have a considerable impact on ICL repair, although patient cells are not nearly as sensitive as Fanconi or full ERCC1-KO cells. This likely explains why no FA-like features are apparent in the two siblings.

The impact of ERCC1R156W in DSB repair

Given the role of ERCC1-XPF in DSB repair, we also addressed to what extent ERCC1R156W still supports this repair process. To this end, we performed live-cell imaging after UV-A laser irradiation in cells sensitized with BrdU to generate local DSBs (Lukas et al., 2003). Recruitment of endogenous XRCC4 could be detected at sites of local UV-A laser irradiation only after BrdU sensitization (Fig. S5 C), showing that DSBs were induced under our conditions. Recruitment of ERCC1WT-GFP to sites of DSBs was clearly detected, whereas this was much weaker for ERCC1R156W (Fig. 8, A and B). However, clonogenic survival assays after ionizing radiation (IR) showed no difference between ERCC1-KO cells reconstituted with either ERCC1WT or ERCC1R156W (Fig. 8 C), suggesting that the mutant ERCC1 protein supports DSB repair.

To also address this in the patient cells, we performed clonogenic survival assays in hTERT-immortalized fibroblasts after exposure to increasing doses of IR. Fibroblasts deficient in XRCC4 (CS16NG-hTERT; Guo et al., 2015) were clearly sensitive to IR already at the lowest dose (2 Gy). However, PV50LD-hTERT cells were not very sensitive at 2 Gy and only showed an increased sensitivity at higher doses (Fig. 8 D). In fact, full ERCC1-KO cells and R156W-KI cells only displayed marginal sensitivity to IR (Fig. S5 D), suggesting a minor contribution of ERCC1-XPF to protect cells against IR. Another method to assess DSB repair capacity is monitoring the resolution of IR-induced γH2AX foci in time. Both 48BR and patient fibroblasts showed a normal clearance of γH2AX foci within 24 h after irradiation with a physiological dose of 2 Gy (Fig. 8, E and F; and Fig. S5 E). However, treatment of 48BR cells with DNA-dependent protein kinase (DNA-PK) inhibitor fully suppressed γH2AX foci resolution at all time points analyzed (Fig. 8, E and F). These findings suggest that PV46LD and PV50LD cells are fully proficient in DSB repair under low damage load.