CHD6 has poly(ADP-ribose)- and DNA-binding domains and regulates PARP1/2-trapping inhibitor sensitivity via abasic site repair

Abstract

To tolerate oxidative stress, cells enable DNA repair responses often sensitive to poly(ADP-ribose) (PAR) polymerase 1 and 2 (PARP1/2) inhibition—an intervention effective against cancers lacking BRCA1/2. Here, we demonstrate that mutating the CHD6 chromatin remodeler sensitizes cells to PARP1/2 inhibitors in a manner distinct from BRCA1, and that CHD6 recruitment to DNA damage requires cooperation between PAR- and DNA-binding domains essential for nucleosome sliding activity. CHD6 displays direct PAR-binding, interacts with PARP-1 and other PAR-associated proteins, and combined DNA- and PAR-binding loss eliminates CHD6 relocalization to DNA damage. While CHD6 loss does not impair RAD51 foci formation or DNA double-strand break repair, it causes sensitivity to replication stress, and PARP1/2-trapping or Pol ζ inhibitor-induced γH2AX foci accumulation in S-phase. DNA repair pathway screening reveals that CHD6 loss elicits insufficiency in apurinic-apyrimidinic endonuclease (APEX1) activity and genomic abasic site accumulation. We reveal APEX1-linked roles for CHD6 important for understanding PARP1/2-trapping inhibitor sensitivity.

Introduction

Cells—especially cancer cells—often experience oxidative stress and DNA damage caused by rapid proliferation, inflammatory tissue microenvironments, heightened metabolic needs, and/or exposure to reactive oxygen species, including those generated by environmental toxicants such as redox-cycling metalloids or ionizing radiation^1,2. Since oxidative genomic damage is potentially mutagenic and lethal, cells have evolved responses to survive and proliferate. These responses involve DNA repair to resolve existing damage, as well as the upregulation of antioxidant defences to prevent reactive oxygen species from producing further DNA damage. DNA double- and single-strand breaks stimulate the activity of DNA-dependent poly(ADP-ribose) polymerase (PARP) enzymes such as PARP1 and PARP2 that use NAD⁺ to produce chains of poly(ADP-ribose) (PAR) on itself, DNA, chromatin, and other proteins³. In humans, PARP1, the pioneering and most well-studied member of the PARP family, is responsible for over 90% of the cellular PARP activity that occurs in response to DNA damage. PAR formation is required for the recruitment of a wide variety of DNA damage response factors important for cell resilience, as well as certain ‘backup’ pathways of DNA damage repair. Serine modification with mono(ADP-ribose) by HPF1-regulated PARP enzymes has also emerged as a widespread post translation modification important for maintaining cell health after DNA damage^4,5,6 (reviewed in ref. ³). Small molecule inhibition of PARP1 and 2 (PARP1/2) enzymes has proven to be an effective treatment for cancers that cannot effectively complete homologous recombination (HR) repair of DNA double-strand breaks (DSBs) due to their genetics^7,8,9,10. The discovery of PARP1/2 inhibitors (PARPi) such as Olaparib is a great success story of synthetic lethality-based drug design, and they are now used to treat many breast, ovarian, and prostate cancers. PARPi block PAR formation and ‘trap’ PARP1/2 enzymes on DNA, where inactive PARP enzymes then interfere with ongoing repair and/or block DNA replication leading to fork collapse and formation of DSBs that are frequently lethal if HR is not available^3,11. A major challenge in using PARPi as anti-cancer treatments is determining which cancers are sensitive to PARPi such as Olaparib based on their mutation signatures, and how to address the evolution of resistance to PARPi by tumors over time^12,13,14.

Enzyme-mediated, ATP-dependent chromatin remodeling adjusts nucleosome spacing and regulates DNA accessibility, and is an indispensable part of eukaryotic DNA damage responses that are altered in numerous developmental syndromes as well as cancers¹⁵. Multiple members of the nine-member Chromodomain Helicase DNA-binding (CHD) enzyme family are associated with PARP1/2-dependent DNA damage response mechanisms, and are characterized by tandem chromodomains and an ATPase-helicase domain that confers nucleosome re-spacing, removal or exchange activity^16,17,18. Indeed, CHD2, CHD3, CHD4 and CHD7 each have well described roles in the cellular response to DNA damage^{19,20,21,22,23,24,25,26,27}, whilst CHD5 is a tumor suppressor mutated in many cancer types^28,29,30. We previously determined that CHD6 relocates to oxidative DNA damage in a manner dependent on PARP1/2 activity⁸, although whether CHD6 loss influences PARP1/2-trapping inhibitor associated phenotypes in vivo, or how PAR formation leads to CHD6 retention at DNA damage, were unclear.

CHD6 enables cell growth during oxidative stress, and adenocarcinoma-derived A549 cells deleted for CHD6 show elevated DNA damage when exposed to reactive oxygen species, abnormally relaxed chromatin, G2/M cell cycle checkpoint hyper-sensitivity, and increased cell death after chronic oxidative stress⁸. Insertional mutagenesis studies in mice concluded that CHD6 is a cancer driver³¹, and CHD6 gene copy number is increased in human tumors whose formation is associated with chronically inflamed tissues^32,33,34. CHD6 also has roles in interferon kappa antiviral activity³⁵, autophagic flux³⁶, and missense mutations in the CHD6 C-terminal domain 2 region are linked phenotypically with Hallermann-Streiff syndrome, a rare disorder characterized by hair, ocular, craniofacial, dental, and degenerative skin abnormalities^36,37. Germline translocations arising within the human CHD6 gene resulting in haploinsufficiency are linked with neuro-cognitive impairment and cellular aneuploidy^36,38. From a structure-function perspective, the CHD6 domain that has been implicated in Hallermann-Streiff syndrome is designated as a putative SANT or Myb-like domain referred to sometimes as ‘CHD6CT2’, although the disease relevant point mutation of this region had no impact on nucleosome-binding or chromatin remodeling activity in vitro, raising questions as to its molecular purpose²⁰.

Structures of CHD1 and CHD4 bound to nucleosomes show that the motor domain contacts the nucleosomal DNA and histones^39,40. Outside of the motor domain, of the nine CHD enzymes only CHD1 and CHD2 possess verified sub-domains with strong DNA-binding activity, with the CHD6 ‘CHD6CT2’ sub-domain having weak DNA-binding activity. Known roles for sub-domains outside of the motor in other CHD enzymes include interacting with chromatin via histones and/or ancillary factors. We previously identified two regions of the CHD6 polypeptide as being functionally important but mechanistically ambiguous: an N-terminal region (residues 171-231) necessary and sufficient for initial PAR-dependent relocalization of CHD6 to DNA damage, and a central region (residues 1028-1448) required for long term retention of CHD6 at damaged DNA⁸. In this study, we interrogate the relationship between CHD6 function, DNA damage repair, and PARP-trapping inhibitor effects on cells, exploring functional domains to dissect the mechanism by which CHD6 binds to DNA in a PAR-dependent manner. We also explore interplay between CHD6-dependent DNA damage repair processes and demonstrate a molecular route to achieve synthetic lethality with PARP1/2 inhibitors.

Results

CHD6 loss sensitizes cells to PARP-1/2 trapping inhibitor treatment

In our earlier work⁸, we explored the functional consequences of CHD6 deletion in the adenocarcinoma-derived cell line A549, but did not interrogate functional relationships between PARP1/2- and CHD6-dependent pathways. To explore CHD6 and PARP-1/2 interplay in a normal tissue-derived cellular context, we first used CRISPR/Cas9 gene editing to ablate CHD6 expression in the hTERT-immortalized, normal epithelium derived, diploid cell line RPE1^hTERT, one of the most widely used cell models in the contemporary DNA damage response field. Frameshift mutations were introduced into each CHD6 allele, resulting in an RPE1^ΔCHD6 cell line with loss of CHD6 gene expression (Fig. 1a) but no loss of other CHDs such as CHD1, CHD3.1, or CHD4 (Fig. 1a).

Fig. 1: CHD6 loss or introduction of a catalytic site mutation sensitizes cells to PARP1/2-trapping inhibitors and/or depletion of BRCA1.

For all statistical analysis in this figure, 2-way ANOVAs followed by Tukey post hoc tests were carried out. ns = not significant (p > 0.05), *p < 0.05. ** p < 0.01. ***p < 0.001.****p < 0.0001. Source data are provided as a Source Data file. All clonogenic survival assay line graphs represent data normalized to untreated wildtype controls as arithmetic means ± SEM. In all cases, wildtype data are indicated in blue, while CHD6 negative data are in green. Panel a RPE1 were subjected to CRISPR/Cas9-mediated CHD6 gene deletion (ΔCHD6) and immunoblotted for indicated proteins. Panel b WT and ΔCHD6 cells were treated with increasing peroxide (left) or 50 µM H₂O₂ in PBS and/or increasing doses of the PARPi Olaparib in media (right), then scored for colony formation 7 days later. All data is normalized relative to DMSO alone, n = 6. Panel c WT and ΔCHD6 RPE-1 cells were plated and treated with increasing PARPi (left) or 2.5 nM PARPi and increasing H₂O₂. All data is normalized relative to DMSO alone; n = 6. Panel d Sequencing outcomes of successful CRISPR base editing of endogenous CHD6 (in RPE-1 cells) to introduce K492G + T493A = Catalytic Dead = “CHD6^CatDead”. Panel e Immunoblot of CHD6 and loading control KAP-1 in RPE-1 expressing WT CHD6 and mutant from (d). Panel f WT and CHD6^CatDead-expressing cells were treated with increasing PARPi then scored for colony formation 7 days later; n = 3. Panel g WT and ΔCHD6 RPE-1 cells were transfected with siRNA against BRCA1 or a scrambled control (siMock) and, 3 days later, treated with PARPi as in (d) and scored for colony formation; n = 3. Data in line graph is normalized relative to DMSO alone (with **** reflecting 2-way ANOVA comparisons of all lines relative to one another). Panel h Bar charts of resting state normalized to the WT RPE-1 transfected with siMOCK; error bars represent SEM, with statistical analysis being Mann-Whitney t-tests. n = 4. Immunoblot of BRCA1 is shown below to confirm siRNA efficacy.

Consistent with previous observations of CHD6 loss using adenocarcinoma cells⁸, RPE1^ΔCHD6 displayed a modest but significant (p < 0.05) clonogenic sensitivity to acute exposure to H₂O₂ (Fig. 1b). To assess PARP1/2 and CHD6 interplay, RPE1^ΔCHD6 were exposed to increasing doses of the PARP1/2-trapping inhibitor Olaparib (hereafter referred to as ‘PARPi’) ± H₂O₂. Intriguingly, PARPi alone was lethal to CHD6-deleted cells at a significantly (p < 0.05) greater level relative to wildtype, and 2.5 nM PARPi significantly (p < 0.0001) exacerbated lethality caused by H₂O₂ and vice versa, with the PARPi IC₅₀ in H₂O₂ treated (oxidatively-stressed) CHD6 negative cells being in the low nM range (Fig. 1b, c). We previously demonstrated that CHD6 harboring mutations of the highly conserved active site amino acid K492 do not complement the phenotype of CHD6 deleted cells, while having no impact on the ability of CHD6 to relocate to sites of oxidative DNA damage⁸. To test whether the catalytic activity of CHD6 is required for PARPi resistance, we used virus-like particles and CRISPR-mediated base editing to mutate K492 in the active site of the CHD6 ATPase domain, a mutation known to ablate catalytic activity and that fails to complement loss of CHD6 in functional assays⁸. We achieved a K492G + T493A mutation of the endogenous CHD6 gene of RPE-1 cells, confirmed by gene sequencing (Fig. 1d); expression of the catalytically inactivated CHD6 K492G + T493A mutant (CHD6^CatDead) was comparable to wildtype CHD6 (Fig. 1e). CHD6^CatDead cells displayed PARPi sensitivity comparable to RPE1^ΔCHD6 (Fig. 1f), increasing confidence in the specificity of previous observations, and indicates that PARPi sensitivity arises either from loss of CHD6 activity or protein expression. Using siRNA to deplete BRCA1 protein expression, we then determined that CHD6 loss has a greater-than-additive-impact with BRCA1 depletion on resting-state cell survival, and that the PARPi-sensitization phenotype associated with CHD6 loss is intriguingly additive with the larger effect produced by BRCA1 loss (Fig. 1g, h). We therefore suggest that CHD6’s role in PARPi sensitivity may function through a largely BRCA1-independent mechanism.

The CHD6 PAR-binding motif is a lysine-arginine rich patch at amino acids 195-202

Multiple factors, such as XRCC1, that require PAR formation to relocate to DNA damage also elicit PARPi sensitivity when they are mutated^41,42,43,44. To understand this better in the context of CHD6, we used in silico analysis and site-specific mutation to examine domains of interest⁸ contributing to the PARP-dependent relocalization of CHD6 to DNA damage. We identified six lysines and one arginine (K195-198, K200, R201, K202) that had a high probability of contributing to PAR interaction based on sequence conservation and similarity to known PAR-binding motifs (PBM)^45,46,47 (Fig. 2a), as well as being within a region of interest identified in our earlier work⁸. We then mutated all these lysines or arginines to either alanine (A) or glutamine (Q) and monitored CHD6 relocalization to DNA damage produced by 355 nm laser micro-irradiation, per the workflow in Fig. 2b and as in ref. ⁸. While full length CHD6 (K → A) and (K → Q) PBM mutants were nuclear (Fig. 2c), they relocated to DNA damage in a significantly (p < 0.001) impaired manner, reaching only ~50% of maximum wildtype signal and dissipating with faster kinetics (Fig. 2d). PAR glycohydrolase inhibitor (PARGi) treatment (that increases the magnitude and persistence of CHD6 at DNA damage tracks, per⁸ and Fig. 2e) did not significantly (p > 0.05) impact the relocalization of CHD6 with a mutated PBM (henceforth referred to as CHD6^ΔPBM) (Fig. 2f), suggesting that the seven amino acids within aa 195-202 facilitates CHD6 relocalization to DNA damage in a manner proportionate to steady state levels of PAR. To assess whether the CHD6 PBM-containing region could bind PAR directly, we performed PAR overlay (far-western) slot blot assays (as described in ref. ⁴⁸) using purified, Maltose Binding protein (MBP)-tagged fragments of CHD6 aa 1-80, 81-146, 147-210 or 211-280, as well as full length histone H1 as a PAR-binding positive control (Fig. 2g). Regions aa 1-80 and 211-280 of the CHD6 N-terminus showed no PAR-binding at all, while PBM-containing aa 147-210 displayed direct PAR-binding. The aa 81-146 region also showed PAR-binding in vitro, although this could be a consequence of the lysine-rich nuclear localization sequence at aa 93-99 that has a strong potential to mimic a PAR-binding motif in vitro. As previous mutational analysis in cells revealed this region to be entirely dispensable for PAR-dependent recruitment of CHD6 to DNA damage⁸, we did not consider the aa 81-146 region further in the context of PAR-related function. Collectively, this data indicates that the aa 195-202 PBM defined in vivo overlaps with a region that supports PAR binding in vitro.

Fig. 2: The CHD6 N-terminus contains a PAR-interacting motif needed for relocalization to DNA damage.

For all statistical analysis in this figure, 2-way ANOVAs followed by Tukey post hoc tests were carried out. ns = not significant (p > 0.05), *p < 0.05. ** p < 0.01. ***p < 0.001.****p < 0.0001. Source data are provided as a Source Data file. All line graphs of microirradiation data are means normalized to t = 0 min, ± SEM, and full statistical test outcomes are in Supplemental Table 1. Panel a Schematic of CHD6 domains illustrating nuclear localization signal (NLS), tandem chromodomains (CD1 + 2), catalytic subunit (ATPase), SANT, BRK domain, two regions of interest, and Hallerman-Streiff syndrome mutation domain (CHD6CT2). Sequence alignment of putative PAR-binding motif (PBM) and corresponding mutants. Panel b Workflow of laser micro-irradiation and analysis; this figure was created using Microsoft PowerPoint. Panel c Nuclear localization controls for CHD6^{ΔPBM (K→A)-GFP} and CHD6^{ΔPBM (K→Q)-GFP}. Panel d Cells expressing GFP-tagged CHD6^{ΔPBM (K→A)} and CHD6^{ΔPBM (K→Q)} were micro-irradiated, imaged over 20 minutes, and analyzed as in (b); n = 18-30. Panel e–f Cells expressing GFP-tagged wildtype or CHD6^{ΔPBM (K→A)} were treated with DMSO, 2.5 nM PARPi or 5 μM PARG inhibitor (PARGi) for 1 h prior to micro-irradiation, imaged and quantified as in (b); n = 15–30. Panel g Purified, maltose binding protein (MBP)-tagged CHD6 fragments were slot-blotted on a nitrocellulose membrane and probed with 100 nM purified PAR. PAR binding was detected with the anti-PAR antibody 10H. Histone H1 and MBP were used as positive and negative PAR binding controls, respectively, while Sypro Ruby stain visualized all blotted proteins as a loading control. A Sypro Ruby stained SDS-PAGE of all proteins used in slot blot assay is shown below slot blots.

The CHD6 DNA binding domain encompasses an AT-hook motif and a SANT domain at amino acids 1088-1326

In micro-irradiation experiments, our observation that PARGi was unable to hyper-stimulate recruitment of the CHD6^ΔPBM mutant (Fig. 2f) logically argues against the existence of another module within CHD6 that has a direct, stoichiometric relationship with PAR levels. That said, we noted that CHD6^ΔPBM did not recapitulate the ‘full’ phenotype produced by PARPi, which almost entirely suppresses CHD6 recruitment to DNA damage caused by micro-irradiation (compare Fig. 2e, f); this suggested the existence of another, at least ‘semi PAR-responsive’ domain in CHD6 that contributes to its relocalization to DNA damage, and that requires a basal level of PAR formation (perhaps early on) to trigger its function, but otherwise has a non-stoichiometric relationship with PAR levels. Sequence inspection of the other region of interest (aa 1028-1448) in CHD6 identified by our previous work⁸ (Figs. 2a, 3a), together with structural homology modeling using the SWISS-MODEL⁴⁹ server and AlphaFold3⁵⁰, indicated the presence of a previously undescribed AT-hook motif coupled with a SANT domain between aa 1088-1326 (Fig. 3a, b, Supplementary Fig. 1); this was indicative of a putative (and previously uncharacterized) DNA-binding domain (DBD). AT-hooks are well described DNA-binding motifs present in many chromatin-associated factors, and preferentially bind minor grooves of DNA enriched for adenine-thymine⁵¹. The putative AT-hook containing DBD in CHD6 was distinct from the downstream ‘CHD6CT2’ domain studied previously in the context of Hallermann-Streiff Syndrome²¹ (Fig. 3b). We noted that the putative AT-hook of CHD6 is located within an equivalent area of yeast CHD1 DBD known as ‘Helical Linker-1’ per⁵² and that, whereas the SANT domains are well-predicted by AlphaFold3 (and similar to homology models), AlphaFold3 predicted the area of the AT-Hook / Helical Linker-1 with a low to very low confidence score and therefore was less informative for that region compared to the homology models. As the SANT domain linked to the putative AT-hook occurs earlier (in the CHD6 polypeptide) of the other, already characterized²¹ SANT / CHD6CT2 domain, for clarity we will now refer to the upstream one as ‘DBD1’ and the downstream one as ‘DBD2’ (per Fig. 3b). We next subjected DBD1 amino acids important for putative SANT and/or AT-hook function in other proteins^51,52 to mutagenesis (Fig. 3c), followed by expression and laser micro-irradiation per workflow in Fig. 2b. Mutation of DBD1 amino acids R1099 + K1102 (SANT) and/or R1182 + R1184 + K1188 (AT-Hook) to alanine impaired the ability of CHD6 to accumulate at DNA damage significantly (p < 0.001) (Fig. 3d), producing phenotypes comparable to PBM mutation (Fig. 2d). Note that, moving forward, CHD6^ΔDBD1 refers to the combined mutation of R1099A + K1102A + R1182A + R1184A + K1188A (i.e., both SANT and AT-hook, per Fig. 3c, d).

Fig. 3: The central SANT and AT-hook of CHD6 comprises a putative DNA binding domain important for PAR-responsive CHD6 localization to DNA damage.

For all statistical analysis in this figure, 2-way ANOVAs followed by Tukey post hoc tests were carried out. ns = not significant (p > 0.05), *p < 0.05. ** p < 0.01. ***p < 0.001.****p < 0.0001. Source data are provided as a Source Data file. All line graphs of microirradiation data are means normalized to t = 0 min, ± SEM, and full statistical test outcomes are in Supplemental Table 1. Panel a Schematic of CHD6 domains and corresponding conservation score determined by Jalview 2.10 (per¹⁰⁹). Panel b Structural models of Region of Interest 2 (green, referred to as DBD1) and ‘CHD6CT2’ (beige, referred to as DBD2) of CHD6 based on homology modeling to S. cerevisiae CHD1, with superposition (right) to emphasize overlapping SANT domains and unique AT-hook structure in DBD1. Panel c Models mapping DBD1 and DBD2 mutations. Panel d Cells expressing mutants from (c) were micro-irradiated, imaged over 20 minutes; n = 30. Panel e Cells expressing wildtype or ΔDBD1 CHD6 were exposed to DMSO or 5 μM PARGi for 1h before micro-irradiation; n = 30. Panel f Purified, maltose binding protein (MBP)-tagged CHD6 fragments were slot-blotted, probed with 100 nM purified PAR and immunoblotted for PAR. Histone H1 and MBP were used as positive and negative controls, respectively. Panel g Workflow for delayed PARP (or DMSO) inhibition experiment. Panel h Cells were exposed to PARPi and/or H₂O₂, then immunoblotted for PAR and actin. Panel i Cells expressing wildtype and CHD6 ΔDBD1 were treated with DMSO or 2.5 nM PARPi per the workflow in (g), and analyzed as in Fig. 2b; n = 30. Panels j–k Cells expressing wildtype or indicated mutants were assayed for micro-irradiation relocalization; n = 28–30. For panel k, CHD6 ΔDBD1 + ΔPBM was additionally treated with 5 μM PARGi for 1 h prior to micro-irradiation and analyzed as in Fig. 2b; n = 18–30. Panel l Schematic of KillerRed DNA damage recruitment assay. Panel m: mCherry-tagged KillerRed was induced in U2OS 2-6-3 cells expressing indicated proteins and imaged for mCherry (red) and GFP (green).

Mutation of the DNA-binding domain has a limited effect on PAR-dependent CHD6 phenotypes

We next determined if CHD6^ΔDBD1 displayed altered recruitment to DNA damage tracks after PARGi treatment, or if it was non-responsive to elevated PAR levels similar to CHD6^ΔPBM (Fig. 2f). The addition of PARGi significantly (p < 0.001) increased recruitment and retention of CHD6^ΔDBD1 at DNA damage (Fig. 3e), implying that the PBM and putative DBD are functionally distinct in how they respond to the relative abundance of PAR. That said, the PARGi-induced increase in CHD6^ΔDBD1 retention at DNA damage was significantly (p < 0.001) reduced compared to wildtype (compare maximum recruitment in Figs. 3e–2e), indicating that PAR levels matter in how DBD1 exerts its function, just in a less stoichiometric manner. These outcomes raise the possibility that the CHD6 DBD1 domain may support a small amount of PAR-binding. To test this, we generated MBP-fused fragments of CHD6 aa 1090-1619 encompassing DBD1, as well as point mutants of the SANT or AT-hook (as in Fig. 3c), and screened them for direct PAR-binding using the far-western blot approach described in Fig. 2g. The DBD1 did have detectable PAR-binding activity in vitro that was ablated by either SANT or AT-hook mutation, although this was only a small fraction of the PAR-binding activity displayed by the CHD6 PBM (Fig. 3f). We are cautious not to over-interpret this experiment, given that both PAR and DNA are charged polymers that, especially in vitro, may bind to similar protein motifs, but it raises the possibility that the DBD1 of CHD6 may bind to PAR very weakly. The observation is also compatible with the idea that the putative DNA binding activity of DBD1 might bind genomic regions where early-on PAR formation makes DNA more accessible to binding by the DBD, a phenomenon previously documented for bona fide DNA binding proteins containing macrodomains, WWE, HMG boxes, homeodomains, and more⁵³.

The DNA-binding domain supports CHD6 retention at DNA damage sites after early-on PAR formation

To further interrogate functional interplay between the CHD6 DBD1 and PARP1/2 activity, we engineered a scenario of normal PAR formation at early periods of CHD6 relocalization, followed by PAR suppression (loss) during later periods. We did this by adding PARPi five minutes after DNA damage induction (per workflow in Fig. 3g). Under these conditions, additional waves of PARylation are blocked while the early PAR formed at the sites of DNA damage continues to be erased by the action of PARG. Using immunoblot (and H₂O₂ as a DNA damage trigger), we confirmed initial formation and subsequent degradation of PAR after delayed PARPi addition (Fig. 3h), and then monitored CHD6 relocalization dynamics (Fig. 3i). In delayed DMSO addition controls, wildtype CHD6 relocalized to DNA damage rapidly, with > 90% of maximal signal accumulation within 5 minutes, peaking 2 minutes later, and thereafter remaining relatively stable with very slow dispersion (Fig. 3i, left). PARPi addition at the 5 minutes mark arrested wildtype CHD6 accumulation, followed by slightly faster dissipation over the subsequent 15 minutes (Fig. 3i, right). In the delayed DMSO-addition control, CHD6^ΔDBD1 displayed previously observed phenotypes (Fig. 3i), with an overall reduced maximum accumulation (compared to wildtype) occurring within 2-5 minutes, followed by ~40% signal loss over the subsequent 15 minutes (Fig. 3i, left). The delayed addition of PARPi five minutes post micro-irradiation had no additive impact on the already dissipating CHD6^ΔDBD1, which dispersed steadily after the 5-minute mark similar to delayed DMSO controls (Fig. 3i, right). These outcomes suggested that, five minutes after PAR-dependent relocalization to DNA damage, most already recruited CHD6 is retained in a manner independent of PAR and likely more reliant on binding DNA directly. Such a model could explain why DBD1-mutated CHD6 fails to be retained proficiently even under normal PAR formation and turnover conditions, and why delayed PARPi addition has no further effect on the CHD6^ΔDBD1 mutant (Fig. 3i).

The PBM and DBD1, but not DBD2, of CHD6 cooperate to enable relocalization to DNA damage

Mutation of residues (R1460A + R1534A + R1537A) in the CHD6 DBD2 (analogous mutations to DBD1, per Fig. 3c) produced no significant (p > 0.05) impact on CHD6 relocalization dynamics relative to wildtype, and also exerted no additive effect when combined with ΔDBD1 (compare Fig. 3d, j); this excluded DBD2 from playing a meaningful role in CHD6 relocalization. We also considered that molecular events that enable longer-term CHD6 retention might include PAR-responsive chromatin alterations that make DNA accessible to CHD6 binding via histone tail interactions mediated by the tandem chromodomains characteristic of all CHD family enzymes¹⁵. The latter idea is supported by our previous finding⁸ that inactivating point mutations in both CHD6’s tandem chromodomains (CD) (CHD6^ΔCD) elicit a DNA damage localization phenotype comparable to CHD6^ΔDBD1 (Fig. 3j). To explore this further, we combined DBD1 and CD point mutations (CHD6^Δ(DBD1+CD)) and monitored CHD6 relocalization to DNA damage; however, point mutation of the CHD6 tandem chromodomains was not additive with DBD1 mutation (Fig. 3j). By contrast, combined PBM + DBD1 mutation were additive, and suppressed recruitment to DNA damage sites significantly (p < 0.001) (Fig. 3k) and to a comparable extent as PARPi treatment (Fig. 2e). There was no significant (p > 0.05) impact conferred by combined CD mutation (CHD6^{ΔPBM+DBD1+CD}) and, importantly, the CHD6^ΔPBM+DBD1 mutant was unaffected by PARGi treatment (Fig. 3k). To finish this line of investigation, we used a distinct DNA damage relocalization system⁵⁴ whereby ROS are produced at a single genomic LacO array by LacR-fused KillerRed in the presence of light (Fig. 3l). Light-stimulated KillerRed induced selective accumulation of wildtype CHD6 at the array, with the ΔDBD1 mutant displaying weaker relocalization (Fig. 3m); XRCC1 was used as a positive control, as before⁸. The phenotype in the KillerRed assay reflected some observations using micro-irradiation, but the quantitative range of signal accumulation in this assay was insufficient to be more helpful (Supplementary Fig. 2a). Considered together, the outcomes of our cellular analyses suggest that: (i) the PBM enables CHD6 to bind sites of DNA damage in a manner proportionate to steady state PAR levels, and (ii) DBD1 supports CHD6 retention at DNA damage sites via (putative) DNA- binding likely enhanced via an earlier PAR-dependent molecular event.

CHD6 has direct DNA affinity that is comparable to CHD1 or CHD2 and is essential for catalytic activity

An untested assumption of our conclusions so far is that DBD1 is a bone fide DNA binding domain. To address whether DBD1 of CHD6 binds DNA directly, we cloned an MBP-fusion of human CHD6 (hCHD6) aa 1090-1619 that contained only DBD1 and DBD2, to exclude potentially confounding DNA-binding effects of the catalytic domain. We also prepared the ‘ΔDBD1’ and ‘ΔDBD2’ point mutations exactly as in Fig. 3c, expressed everything using Escherichia coli cultures, and purified them (Fig. 4a). Electrophoretic mobility shift assays (EMSA) using double-stranded DNA indicated that wildtype hCHD6 aa1090-1619 displayed clear DNA binding affinity that was reduced by point mutation of DBD1 but not DBD2 (Fig. 4b). To verify DNA affinity observations using a more quantitative approach, we used fluorescence polarization⁵⁵ (per workflow in Fig. 4c) and the MBP-fused DBDs of the well-characterized human CHD2 (hCHD2) and yeast CHD1 (yCHD1) proteins as controls (Fig. 4a). Wildtype hCHD6 aa 1090-1618 had an equilibrium dissociation constant (K_D) of 0.032 µM with double-stranded DNA, comparable or stronger DNA binding affinity than the DBD of hCHD2 (aa 1108-1355, K_D = 0.056 µM) or yCHD1 (aa 1006-1270, K_D = 0.226 µM) under identical experimental conditions (Fig. 4d). An MBP-fusion containing only the CHD6 DBD2 domain (aa 1449-1619) displayed substantially weaker DNA binding affinity of K_D = 1.8 µM (Fig. 4d), consistent with previous findings using this region in isolation³⁶. Mutation of the AT-hook and/or SANT in DBD1 (mutations as in Fig. 3c) functionally ablated CHD6 DNA binding, with K_D rising from 0.032 to 1.26 µM, indicating that the CHD6 DBD1 is a high-affinity DNA binding module within the protein (Fig. 4e). In contrast, point mutation of the DBD2 domain had a negligible impact (K_D rising from 0.032 to 0.037 µM) (Fig. 4e).

Fig. 4: CHD6 possesses direct DNA binding activity.

For all statistical analysis in this figure, 2-way ANOVAs followed by Tukey post hoc tests were carried out. ns = not significant (p > 0.05), *p < 0.05. ** p < 0.01. ***p < 0.001.****p < 0.0001. Source data are provided as a Source Data file. Panel a. Coomassie stain of indicated MBP-tagged purified proteins. Panel b. Increasing amounts of hCHD6 aa 1090-1619 wildtype, ΔDBD1 or ΔDBD2 point mutants were incubated with double-stranded DNA and resolved by electrophoretic mobility shift assay (EMSA). Please note that we attribute the absence of a discrete DNA complex band in the highest concentrations of ΔDBD1 (despite loss of free probe signal) to the overall weak but still measurable affinity of ΔDBD1 for DNA (compared to WT and ΔDBD2) that results in a less stable complex with a higher off-rate causing free probe signal to smear across the middle section of the gel at higher concentrations, rather than form a discrete band. Panel c Workflow for fluorescence polarization-based DNA binding assay. Panel d MBP-fused proteins from (a) were titrated into fluorescently labelled double-stranded DNA and resulting change in depolarization was measured; fluorescence polarization line graphs show geometric means ± CI95%; n = 3. Dissociation constants (K_D) were determined by fitting the data to a logistic function and identifying the point on the model with the largest change in fluorescence polarization. Panel e Wildtype and indicated point mutants of hCHD6 aa 1090-1619 (from a) were assayed as in (d); graph shows geometric means ± CI95%; n = 3.

To investigate the impact of CHD6 DBD1 and DBD2 on CHD6 ATPase and nucleosome binding and remodeling activities, we cloned, expressed using baculovirus-infected SF9 cell cultures, and purified an MBP-fusion of human CHD6 (hCHD6) aa 270-1618, which contained the tandem chromodomains, the ATPase domain, and both DBD1 and DBD2. We then performed EMSA using purified mononucleosomes formed with 147 bp double-stranded DNA and a 50 bp linker DNA (total 197 bp). Whilst wildtype hCHD6 exhibited clear nucleosome binding activity, a truncation mutant (aa 270-1029) lacking both DBD1 and DBD2 displayed no binding at all (Fig. 5a, b). ATPase assays showed that wildtype hCHD6 ATPase activity is stimulated 45% and 350% in the presence of double-stranded DNA and mononucleosomes, respectively (Fig. 5c). This stimulation of hCHD6 ATPase activity by DNA and mononucleosomes is lost by the removal of DBD1 and DBD2, with this mutated hCHD6 having similar ATPase activity as hCHD6 containing a Walker A box K492R mutation designed to block ATP hydrolysis (Fig. 5c). To separate the roles of DBD1 and DBD2 in this context, we performed an in vitro nucleosome sliding assay (Fig. 5d) with either wildtype hCHD6, hCHD6 with mutated DBD1, or hCHD6 with DBD1 and DBD2 removed. Wildtype hCHD6 was effective in sliding a nucleosome with ATP (Fig. 5e). Point mutation of DBD1 reduced CHD6 nucleosome sliding activity significantly (p < 0.0001) (Fig. 5f), although removal of both DBD1 and DBD2 was needed to fully ablate nucleosome sliding activity (Fig. 5g). Altogether, these outcomes favor CHD6 possessing potent DNA binding activity largely conferred via the DBD1 (amino acids 1088-1326), while DBD2 operates in a cooperative manner with DBD1 to enable chromatin binding and ATP-dependent chromatin remodeling. These findings suggest that DBD(1+2)-mediated DNA binding is needed for CHD6 catalytic activity and, when considered together with the outcomes of our CHD6 relocalization analyses (Figs. 2–3), implies that CHD6 chromatin remodeling requires (or is at least stimulated by) PARP1/2-dependent events early on in the DNA damage response that enable CHD6^PBM binding to PARylated protein substrate(s) and CHD6^DBD1 binding to DNA.

Fig. 5: CHD6 DNA binding activity is essential for ATP-dependent nucleosome remodeling.

For all statistical analysis in this figure, 2-way ANOVAs followed by Tukey post hoc tests were carried out. ns = not significant (p > 0.05), *p < 0.05. ** p < 0.01. ***p < 0.001.****p < 0.0001. Source data are provided as a Source Data file. Panel a. Schematic of wildtype and point mutants of hCHD6 aa 270-1618, as well as the hCHD6 aa270-1029 truncation mutant lacking both DBD1 and DBD2 entirely. Panel b Ability of indicated proteins from (a) to bind mononucleosomes (147 bp DNA around the histone octamer and 50 bp linker DNA) was assessed by EMSA. Panel c ATP hydrolysis assay with proteins from (a) in presence or absence of free double-stranded DNA (dsDNA) or mononucleosomes from (b); centre lines of ATP assay data represent arithmetic means ± SD; n = 3. Panel d. Workflow for chromatin remodeling assay; this figure was created with Biorender.com (agreement number UO268HP0OZ) and Microsoft PowerPoint. Panel e–g Wildtype CHD6 and indicated mutants (from a) were assayed with and without ATP using chromatin remodeling assay outlined in (d). All FRET data line graphs are geometric means ± CI95%; n = 3.

The nuclear CHD6 interactome includes PARP-1, and is highly enriched for PARylated proteins

To evaluate whether CHD6 interacts with PARylated or PAR-associated proteins in cells in an unbiased manner, we carried out affinity purification mass spectrometry (AP-MS) to systematically identify nuclear proteins co-purifying with hemagglutinin (HA)-tagged wildtype CHD6 (CHD6^HA), using the workflow in Fig. 6a and methodology validated in ref. ⁵⁶. We confirmed the selective recovery of CHD6^HA from nuclear extracts (Fig. 6b), and performed on-bead tryptic digestion. Eluted peptide analysis by LC-MS/MS identified 279 proteins that were either only in or > 1.5-fold enriched in affinity purified CHD6^HA compared to IgG controls. Hits were filtered to exclude cytoplasmic and common contaminants (such as keratin) to isolate a network of 132 proteins associated with CHD6. Confidence scores were calculated via significance analysis of interactome (SAINT) algorithms⁵⁷ based on fold enrichment compared to IgG controls (see Supplemental Information for full details of the screening outcomes and methodology). Among the highest confidence hits, considered likely to be true biological interactions, were multiple histones, DNA repair proteins, RNA-processing factors, oxidative stress responders, and chromatin regulators, with PARP-1 (the major acceptor of PAR in cells) being the overall highest scoring hit after CHD6 itself (Fig. 6c, d). A combined analysis of CHD6- and PAR-modified and/or PAR-associated proteins determined in refs. ^58,59,60,61 highlighted the convergence of the two protein networks. Approximately 70% of top ranking CHD6-associated proteins have been identified in at least one screen for PAR- modified or associated proteins, with three hits (PARP-1, SUPT16H, and XRCC5) being validated by all four screens (Fig. 6e, f). With the outcomes of this proteomic analysis, we now have multiple lines of evidence indicating that the CHD6 protein interaction network overlaps with those associated with PAR-modified/associated proteins. We note that loss of other PAR- and/or PARP-1 interacting proteins, including XRCC1¹⁰ and the chromatin remodeler ALC1/CHD1L⁶², is also associated with PARPi synthetic lethality and, therefore, these findings are also consistent with the PARPi sensitivity produced by CHD6 loss in inactivation we report in Fig. 1.

Fig. 6: The nuclear CHD6 interactome is highly enriched for PAR-modified and PAR-associated proteins.

For all statistical analysis in this figure, 2-way ANOVAs followed by Tukey post hoc tests were carried out. ns = not significant (p > 0.05), *p < 0.05. ** p < 0.01. ***p < 0.001.****p < 0.0001. Source data are provided as a Source Data file. Panel a Workflow used for identification of interacting partners of CHD6; this figure was created using Biorender (https://BioRender.com/i42e687) and Microsoft PowerPoint. Panel b CHD6^HA was transfected into HEK293T, immunoprecipitated from nuclear cells extracts, and immunoblotted for CHD6 to confirm IP efficacy (with KAP-1 used as a loading control); this data is representative of three biological replicates. Panel c SAINT scores of all identified CHD6-interacting proteins and compared to their fold enrichment above IgG controls. Panel d Partial network of proteins identified exclusively in CHD6^HA IP or that exhibited a spectral count fold change > 2 compared to IgG controls. Protein hits were filtered for spectral count > 3 across three biological replicates and nuclear localization using the STRING database. Connections between protein hits are based on physical interactions annotated in the STRING database. See Supplemental Information for complete network and all protein names. Panel e. Network of the highest confidence interacting proteins of CHD6 (highlighted by yellow shading in (c). High confidence hits were determined based on a minimum fold change of 2 (dashed line) and SAINT score of 0.5. Annotated interactions from the BioGRID database are also shown. Nodes with yellow borders indicate known targets of PARP1 activity or proteins that interact with PAR-modified substrates. Panel f. Stacked Venn diagrams illustrate the number of proteins identified in the CHD6 nuclear interactome that also were listed as significant in four published screens identifying substrates for PARylation or PAR-interactors.

CHD6-deletion causes PARPi-dependent γH2AX foci accumulation in S-phase, but does not impact DSB repair

Genetic alterations (such as XRCC1, ALC1 or BRCA1/2 loss)^3,9,10,62 that lead to synthetic lethality with PARP1/2-trapping inhibitors are typically associated with DSB accumulation during DNA replication, where HR is a primary pathway for repair of ‘one-ended’ DSBs arising from collapsed replication forks. To explore whether this phenomenon applies to CHD6 loss, we used γH2AX signal as a readout for DSB accumulation and analyzed cell populations in G1-phase versus those in mid S-phase, separating populations into cell cycle phases using a software-assisted, DAPI-signal based process described in ref. ⁶³ and Supplementary Fig. 2b–d. Visual inspection (Fig. 7a) and quantitative analysis of γH2AX foci per nucleus or ‘refined γH2AX’ (which accounts for variation in focal intensity, focal number, and nuclear volume, as in ref. ⁶³) indicated that PARPi significantly (p < 0.0001) increased signal in CHD6-deleted cells in mid S-phase, but not in G1-phase (Fig. 7b, c). PARPi-associated differences were also observed using RAD51 as a separate marker for DSBs undergoing active HR in S-phase (Fig. 7d), distinguishing the phenotype of CHD6 loss (in the context of PARPi synthetic lethality) from BRCA1/2 loss where RAD51 filaments cannot form^3,9. In washout experiments using S-phase CHD6-deleted cells treated with PARPi for 2h, both γH2AX and RAD51 decreased to near-background levels within 3 h of PARPi removal (Fig. 7e), consistent with our previous study where reporter assays indicated no innate difference in non-homologous end-joining (NHEJ) or HR-mediated DSB repair capacity in CHD6-deleted cells⁸. By using ionizing radiation (IR) to induce equivalent DNA damage amounts and monitoring recovery over time, we found similar-to-wildtype γH2AX foci disappearance kinetics in CHD6-deleted cells that were either in G1 phase, where NHEJ is the primary means of DSB repair, or in S or G2 phase where both HR and NHEJ are required for normal DSB repair (Fig. 7f). S- and G2-phase cells lacking CHD6 also successfully formed RAD51 foci after IR, further indicating that HR is functional (Supplementary Fig. 3a–b).

Fig. 7: PARPi triggers an S-phase specific accumulation of γH2AX and RAD51 in CHD6-deleted cells, which display highly functional DNA double strand break repair.

For all statistical analysis in this figure, 2-way ANOVAs followed by Tukey post hoc tests were carried out. ns = not significant (p > 0.05), *p < 0.05. ** p < 0.01. ***p < 0.001.****p < 0.0001. All experiments are n = 3, centre lines of all foci data graphs are arithmetic means ± CI95%. Source data are provided as a Source Data file. Panel a Logarithmically dividing WT and ΔCHD6 cells were treated with 2.5 µM PARPi or DMSO for 1h and immunostained for γH2AX and DAPI. Images are representative of cells in mid S-phase based on relative DAPI signal. Scale bars = 5 µm. Panel b For the experiment described in (a), γH2AX foci per nucleus for cells in either G1 or mid S-Phase were enumerated. Panel c For the experiment described in (a), refined γH2AX signal (that accounts for γH2AX focal intensity, focal number, and nuclear volume) in either G1 or mid S-Phase was measured. Panel d S-phase cells from (b) were immunostained for γH2AX and RAD51, and enumerated for the arithmetic mean number of RAD51 foci per nucleus; n = 3. Representative images are from PARPi-treated ΔCHD6 S-phase cells. Scale bars = 5 µm. Panel e Logarithmically dividing ΔCHD6 RPE-1 cells were incubated in media with 2.5 mM PARPi or an equal volume of DMSO for 2h, then placed into fresh media and harvested immediately (0h), or after another 1, 2, and 3h. Cells were immunostained for γH2AX and RAD51, and enumerated as in (b, d). Panel f Logarithmically dividing WT and ΔCHD6 RPE-1 cells were irradiated with 2 Gy gamma ray ionizing radiation (IR), fixed between 0.5-24h later, and immunostained for γH2AX and DAPI. Graphs are γH2AX foci per nucleus for cells predominantly in (left to right) G1-phase, S-phase, or G2-phase. Panel g Logarithmically dividing WT and ΔCHD6 cells were treated with 50 µM H₂O₂ (or PBS), detergent extracted, immunostained for PARP-1, imaged and quantified for nuclear PARP-1 signal; n = 3.

CHD6-deletion increases PARP-1 retention on chromatin and hypersensitizes cells to replication stressors

As HR-mediated DSB repair capacity is seemingly normal in CHD6-deleted cells, another explanation for PARPi sensitivity (and increased S-phase DSBs) is heightened DNA replication fork collapse due to collision with an elevated number PARP enzymes trapped at DNA lesions by Olaparib. Of note, the relative difference in γH2AX induced by PARPi in wildtype versus CHD6-deleted S-phase cells was greatest in cells grown near ambient O₂ levels (as in Fig. 7), but was not evident in experiments conducted at 5% O₂ (Supplementary Fig. 3c). We speculated that this outcome likely reflects already described⁸ oxidative stress that occurs in the absence of CHD6, and that under greater % O₂ more PARP1/2 enzymes become trapped on DNA (by PARPi) as they are ‘already’ interacting with a greater volume of oxidative DNA lesions versus wildtype. To evaluate this hypothesis, we measured the relative amount of chromatin-associated (detergent extraction-resistant) PARP-1 in wildtype and CHD6-deleted cells treated ± H₂O₂, a technique used before to measure DNA-bound PARP-1⁶⁴. We found that loss of CHD6 resulted in a significant (p < 0.0001) and marked ( + 270%) H₂O₂-induced increase in PARP-1 retention on chromatin relative to wildtype ( + 43%) (Fig. 7g). The ability of different PARP1/2 inhibitors to induce lethality has been directly linked to their potency in trapping PARP1/2 enzymes on DNA and eliciting DSBs¹¹. For example, while the PARPi drugs Olaparib and Veliparib are equally effective catalytic inhibitors of PAR formation (Supplementary Fig. 3d), Veliparib has reduced ability to trap PARP1/2 enzymes on DNA and is known to be far less potent than Olaparib at killing cells lacking BRCA2¹¹. To evaluate the relative importance of drug-induced PARP1/2 trapping for PARPi-associated lethality in CHD6-deleted S-phase cells, we monitored clonogenic survival after Veliparib exposure. Unlike Olaparib treatment (Fig. 1c, d), CHD6-deleted cells were not hypersensitive to Veliparib compared to wildtype controls (Fig. 8a, b), indicating that PARP1/2 trapping on DNA (i.e., a physical block that stalls replication forks and increases DNA replication stress) is a key part of the PARP1/2 inhibitor lethality phenotype associated with CHD6 loss. If correct, we speculated that CHD6 loss might also elicit a general sensitivity to DNA replication stressors that either slow forks, or preclude reversal and restart after stalling. Indeed, CHD6-deleted cells showed significant (p < 0.0001) hypersensitivity to hydroxyurea (HU), which increases risk of replication fork collapse by slowing forks via dNTP depletion, as well as inhibitors of ATR, whose protein kinase activity is needed for replication fork restart, as well as HR⁶⁵ (Fig. 8a, b). These data support the idea that CHD6 loss produces lethality under multiple conditions where replication forks slow, stall, and/or cannot reverse.

Fig. 8: CHD6-deletion causes hypersensitivity to replication stress.

For all statistical analysis in this figure, 2-way ANOVAs followed by Tukey post hoc tests were carried out. ns = not significant (p > 0.05), *p < 0.05. ** p < 0.01. ***p < 0.001.****p < 0.0001. Source data are provided as a Source Data file. All clonogenic survival assay line graphs represent data normalized to untreated wildtype controls as arithmetic means ± SEM. Panel a WT and ΔCHD6 RPE-1 cells were plated and treated either with 7.5 μM Veliparib or an equal volume of DMSO, 500 μM hydroxyurea (HU) versus an equal volume of PBS buffer, or 500 nM ATR inhibitor (ATRi) or an equal volume of DMSO, or 2.5 nM PARPi with and without 1 nM camptothecin (CPT), and scored for colony formation 6 days later. Panel b Quantified clonogenic survival data from (a) for WT and ΔCHD6 RPE-1 cells treated with increasing doses of Veliparib, HU or a fixed dose of 2.5 nM PARPi and increasing doses of HU, or increasing doses of ATRi. Panel c WT and ΔCHD6 RPE-1 cells treated with increasing doses of methyl methane-sulfonate (MMS), a fixed dose of 2.5 nM PARPi with increasing MMS, or a fixed dose MMS with increasing PARPi. Panel d Logarithmically dividing WT and ΔCHD6 RPE-1 cells were treated with 1 µM TLSi or an equal volume of DMSO for 1 h, immunostained for γH2AX and DAPI, with centre line of all foci data graphs represent arithmetic mean γH2AX foci per nucleus for cells in either G1 or mid S-Phase ± CI95%; n = 3.

Loss of CHD6 increases dependence on translesion synthesis

We were interested to find that inducing alkylation-associated DNA lesions by methyl methane sulfonate (MMS) was equally lethal to CHD6 wildtype and deleted cells unless PARP1/2 are inhibited, wherein CHD6-deleted cells displayed significant (p < 0.001) hypersensitivity (Fig. 8c). Comparable results were obtained using the topoisomerase inhibitor camptothecin (CPT) (Supplementary Fig. 3e), indicating that CHD6 is less involved in withstanding DNA replication blocking lesions that are not directly related  to PARP1/2 binding, such as CPT-induced poisoning of topoisomerases on DNA. Based on these outcomes, we speculated that CHD6 loss is associated with a subtle increase in replication fork-blocking lesions associated with PARP1/2 recognition, making cells more sensitive to the replication slowing effects of HU or ATRi. We suggest that such subtle increases are proficient at increased trapping of PARP1/2 in the presence of Olaparib, but are otherwise ‘overwhelmed’ by the massive, exogenous damage produced by MMS or CPT. That said, even a modest increase in replication fork blocking lesions in CHD6-deleted cells should, in theory, render these cells more reliant on translesion synthesis (TLS) DNA damage tolerance pathways used to bypass DNA lesions via lower fidelity DNA polymerases⁶⁶. Small molecule TLS inhibitors (TLSi) that target the REV7-REV1 subunits of Pol ζ are considered an emerging class of anti-cancer agents with promise⁶⁷, as TLS is also a mechanism by which some cancers may develop PARPi resistance⁶⁸. Notably, TLSi treatment induced a small but significant (p < 0.0001) increase in γH2AX foci in CHD6-deleted cells in S-phase, but not in G1-phase nor in any wildtype condition (Fig. 8d); this small effect did not, however, correspond to any impact on clonogenic survival (Supplementary Fig. 3f). Collectively, these experiments suggest that CHD6 is important for cells to withstand replication stress, that its loss increases reliance on damage tolerance pathways to prevent additional replication stress, and is especially vital if PARP1/2 enzymes are trapped via inhibition; we suggest this could be explained by the accumulation of one or more replication-blocking DNA lesions recognized by PARP1/2.

CHD6-deleted cells display abnormal induction and repair of lesions detected by Fpg-modified comet assay

DNA lesions that block replication include a variety of damage types, including many substrates of base excision repair, whose accumulation within the genome can serve to increasingly trap (inhibited) PARP1/2 enzymes. To explore this idea, we quantified base lesion formation and resolution using the Fpg enzyme-modified alkaline comet assay per Fig. 9a. Fpg is a bifunctional DNA glycosylase and apurinic/apyrimidinic (AP) lyase that selectively cleaves DNA containing 8-oxoguanine, alkylated purines, and/or AP (also called abasic) sites to produce DNA single strand breaks. The difference in comet tail moment between an undigested sample (only DNA breaks) and Fpg-digested sample (DNA breaks plus base lesions) enables a quantitative assessment of base lesion induction and resolution over time. CHD6 deleted cells displayed a modest increase in DNA break induction per dose of H₂O₂ via standard alkaline comet assay (Fig. 9b, c, darker dots and bars), comparable to our previous observation using CHD6-deleted adenocarcinoma-derived cells⁸ (and demonstrating phenotypic reproducibility between cell lines). However, we were skeptical that these differences (observed primarily at early time points) could explain why CHD6-deleted cells are so PARPi sensitive in clonogenic survival assays. As expected, all Fpg-digested alkaline comet tail moments were higher compared to the standard assay (confirming that the method worked) (Fig. 9b, c, lighter dots and bars). Importantly, the difference between Fpg-digested minus undigested tail moments revealed a 2-fold increase in base lesions in the CHD6 deleted cells relative to wildtype control. Unlike DNA strand breaks alone (Fig. 9c, dark bars), which were resolved to a comparable level in both wildtype and the CHD6 mutant within 7.5 minutes of induction, DNA base lesions and/or abasic sites (Fig. 9c, light bars) remained ~100–500% elevated in CHD6 mutant cells up to 1 h later compared to wildtype levels. Such a substantial difference represents clear evidence that CHD6 supports base lesion elimination and/or suppresses base lesion formation.

Fig. 9: CHD6-deletion causes increased Fpg-enzyme sensitive base damage persistence following H₂O₂ exposure.

For all statistical analysis in this figure, 2-way ANOVAs followed by Tukey post hoc tests were carried out. ns = not significant (p > 0.05), *p < 0.05. ** p < 0.01. ***p < 0.001.****p < 0.0001. Source data are provided as a Source Data file. Panel a A workflow for Fpg-modified and standard alkaline comet assay; this figure was created using Microsoft PowerPoint. Panel b. WT and RPE-1^ΔCHD6 in suspension were treated with 50 µM H₂O₂ and analyzed by Fpg-modified and standard alkaline comet assay, with scatters representing raw data from three independent experimental repeats and centre lines representing geometric means ± CI95%; n = 3. Panel c Geometric means and 95% confidence intervals from (b) were expressed as a bar chart representing, showing the differences between standard and Fpg-modified comet assay tail moments as numbers in parentheses above each time point.

CHD6-deleted cells display insufficiency in abasic site repair, and genomic abasic site accumulation

To investigate the specific mechanism underlying the increase in lesions revealed by the Fpg-modified comet assay in CHD6 deleted cells, we used fluorescence multiplex host cell reactivation to measure base excision repair capacity in a systematic, unbiased manner. Briefly, fluorescence multiplex host cell reactivation is a previously validated^69,70,71 DNA repair capacity screening approach that involves transfecting cells with several fluorescent reporter plasmid substrates (containing known quantities of specific DNA lesions), after which fluorescent signal is measured by flow cytometry and used as a quantitative measure of DNA repair capacity. Using this, we measured the relative ability of RPE1^ΔCHD6 to resolve lesions targeted primarily by OGG1 (8OxoG:C), MUTYH (8OxoG:A), APEX1 (tetrahydrofuran abasic site analogue), UDG (Uracil:G), MGMT (O6-MeG), MPG (hypoxanthine:T), or repair of the mixture of lesions caused by exposure of plasmid DNA to ultraviolet C (UVC) light in vitro (Fig. 10a). RPE1^ΔCHD6 cells displayed normal OGG1, MUTYH, UDG, and MPG enzyme-mediated repair, but a modest and statistically significant (p < 0.05) deficiency in the resolution of abasic sites via APEX1 and also the repair of UVC-induced lesions (which include abasic sites). These outcomes narrowed down the steps within base excision repair (or possibly nucleotide excision repair⁷²) in which CHD6 likely functions, excluding a role in the initial recognition and excision of damaged bases by DNA glycosylases, and pointing towards a need for CHD6 during subsequent processing of abasic sites.

Fig. 10: Loss of CHD6 reduces cellular base excision repair capacity to resolve apurinic/apyrimidinic (AP) sites, leading to AP site accumulation within genomic DNA.

Source data are provided as a Source Data file. Panel a Wildtype and CHD6-deleted cells were analyzed by fluorescence multiplex host cell reactivation (as in refs. ^69,70,71) using reporter plasmids that monitor resolution of the indicated base excision repair substrates. Bar graph data are geometric means ± CI95%; n = 8. 2-way ANOVAs followed by Tukey post hoc tests were carried out between WT and ΔCHD6, with ns = not significant (p > 0.05), and * representing p = 0.0198 for abasic site and p = 0.0197 for UV substrate. Panel b Workflow for the ELISA assay monitoring apurinic/apyrimidinic (AP) sites present within genomic DNA; figure was created using Biorender (https://BioRender.com/y64d242) and Microsoft PowerPoint. Panel c Wildtype and CHD6 cells were treated ± 0.04% (v/v) MMS for 1 h and either harvested immediately (0 h) or placed into fresh media for 3 h and allowed to recover, before being analyzed as in (b). Data are represented as 5–95 percentile box and whisker plots for n = 4. Black dots in middle of plot represent arithmetic means, centre lines represent medians, and + % values represent relative increase in AP sites per 10⁵ bp in CHD6-deleted cells relative to wildtype. 2-way ANOVAs followed by Tukey post hoc tests were carried out. ns = not significant (p > 0.05),****p < 0.0001. Panel d A composite structural prediction of the arrangement of the CHD6 functional domains in relation to a mononucleosome with linker DNA (light grey), with CHD6 tandem chromodomains shown in purple, the ATPase-Helicase motor domain in blue, and the DBD1 (SWISS-MODEL) in green and the nucleosome template in grey (PDB ID 5O9G). Panel e A model for the molecular circumstances leading to PARPi-induced synthetic lethality with CHD6 loss; this figure was created using Biorender (https://BioRender.com/j29p818) and Microsoft PowerPoint.

Unresolved abasic sites are incompatible with life as they interfere with processes such as DNA transcription and replication, increasing mutation rates and stalling replication forks that may lead to collapse and chromosomal breakage. Abasic sites often transition into (and out of being) reactive aldehydes that can lead to Schiff-base intermediates with abundant DNA binding proteins (including PARP-1, glycosylases or Ku70/80), and may even oxidize into 2-deoxyribonoloactone that forms covalent DNA-protein crosslinks with proteins such as DNA polymerase beta or the replication fork ‘proofreader’ protein HMCES. In the context of PARP1/2 enzymes, the Wilson laboratory has demonstrated that PARP1 forms DNA-protein crosslinks with abasic sites in vitro and in vivo, and that this process is enhanced by treatment with PARPi⁷³. DNA-protein crosslinks are potent replication fork blocks, and (a subset of) abasic site resolution that requires CHD6 could mechanistically explain many of the phenotypes we have observed in CHD6-deleted cells exposed to PARPi. We also reflected on the fact that our proteomic screen identified PARP-1, Ku70/80 and APEX1 as high confidence nuclear interactors of CHD6 (Fig. 6), lending further support to this hypothesis. To investigate this, we directly monitored the number of abasic sites per 100 kilobases of genomic DNA in wildtype and CHD6-deleted cells via an ELISA approach outlined in Fig. 10b and⁷⁴. At baseline, genomic DNA from CHD6 deleted cells contained 57.6% greater number of abasic sites per 100 kb relative to wildtype controls, a significant (p < 0.0001) difference that increased proportionately after exposure to a low dose of MMS, indicative of an ongoing issue in repairing some abasic sites irrespective of an endogenous or exogenous origin (Fig. 10c). Collectively, these outcomes indicate a specific role for CHD6 in the elimination of abasic sites (a well-known source of PARPi induced PARP1/2 enzyme trapping leading to replication stress), and present a strong mechanistic explanation for the synthetic lethality between PARPi exposure and CHD6 loss.

Discussion

PAR-dependent responses to DNA damage encompass a wide variety of pathways including base excision repair, nucleotide excision repair, alternative non-homologous end-joining, transcription shutdown, histone H1 relocalization, and CHD1-, CHD2-, CHD4-, and CHD7-dependent processes in DSB repair. In this study, we identify CHD6 as an important factor to understand DNA-damage responsive PARP enzyme function, cellular sensitivity to PARP1/2-trapping inhibitors, and abasic site repair processes involving APEX1. Our work demonstrates that CHD6 binds to PAR-modified chromatin regions during a DNA damage response, and overlaps with a wide range of PAR-modified/associated proteins in cells, including PARP-1 itself. CHD6 binds to PAR via a canonical K/R-rich PBM, as well as a previously undescribed PAR-dependent DNA binding activity conferred by a DBD comprised of a SANT domain and AT-Hook.

The existence of DNA binding domains within CHD6 with the same or greater potency as the DBDs from CHD1 and CHD2 was surprising, as class III CHD enzymes are not categorized as possessing this activity, which has been historically a hallmark of class I CHDs. In previous structures of other CHD family enzymes in complex with the nucleosome^39,75, the yCHD1 DBD is observed bound to an unwrapped linker DNA. Structural modelling of the CHD6 polypeptide (using SWISS-MODEL, I-TASSER, and AlphaFold) indicates both sequence and structural similarity between the CHD6 DBD1 and the well-studied DBD of yCHD1. Given the high similarity between yCHD1 and CHD6’s ATPase motor and tandem chromodomains, we hypothesize that CHD6 is likely to bind unwrapped linker DNA via its DBD1 in a similar orientation to yCHD1 on a nucleosome (Fig. 10d). Such an arrangement would be consistent with how AT-hook and SANT domain modules bind DNA, and is supported by our biochemistry showing that AT-Hook and SANT mutations ablate DNA binding affinity substantially. If correct, the orientation of CHD6 on a nucleosome (where the DBD1 positions towards linker DNA) may help explain why DBD1 function during DNA damage recruitment kinetics is ‘subservient’ to PARP1/2 activity. More specifically, PAR-dependent release of histone H1 from chromatin around DNA lesions⁷⁶ may be prerequisite for the CHD6 DBD to support chromatin retention, following earlier recruitment via the PBM interacting with PAR chains. In our earlier work, we excluded CHD6 as being required for PAR- dependent H1 displacement⁸, and it will be interesting in the future to determine if, in fact, the reverse might be true.

As an overall functional model (Fig. 10e), our data suggests that abasic sites accumulate in the absence of CHD6 due to an insufficiency in APEX1 activity; given the enzymatic nature of CHD6, we suggest this phenotype is likely because certain chromatin environments (in which abasic sites occur) are challenging for APEX1 to access or function in a timely manner. Where CHD6 specifically exerts its chromatin remodeling activity has not yet been systematically mapped, but assessing this via the assay for transposase-accessible chromatin with sequencing (ATAC-Seq) or a comparable technique is likely to be informative. It is likely that specific chromatin locales ‘normally’ regulated by CHD6 are going to be highly nuanced, as in our previous work we found CHD6-deleted cells display an increase in chromatin sensitivity to micrococcal nuclease⁸ (indicative of a net relaxation in heterochromatin^77,78), and CHD6 is also known to be associated with the CFTR locus associated with active transcription⁷⁹. Importantly, there is precedence for abasic site (and base excision) repair proficiency being chromatin-context dependent, with the relative rotational and translational position of the base lesion with respect to the nucleosome dyad influencing whether or how fast APEX1 can cut DNA in vitro⁸⁰. There is evidence that the HECTD1 E3 ubiquitin ligase and ALC1/CHDL1 chromatin remodeling enzyme contribute to resolving some of these barriers^81,82, and we now propose CHD6 may operate with an analogous process or even perhaps the same pathway—an important area of future research.

Importantly, the outcomes of our experiments indicate a role for CHD6 in the timely elimination of abasic sites, a phenotype that has been linked mechanistically to PARPi hypersensitivity in the context of XRCC1 or Pol β mutation^41,83. We suggest the ~60% increase in abasic sites in CHD6-deleted cells is something that these cells can withstand during DNA replication under normal circumstances, due to a combination of BER processes active in other chromatin regions, DNA damage tolerance enabled via TLS, and fork protection via ATR; this idea is supported by our observation that CHD6 cells display a hyper-reliance on ATR for survival and TLS for DSB suppression. However, if DNA damage-responsive PARP enzymes such as PARP1/2 (and potentially other plentiful, chromatin-bound proteins) accumulate at or are trapped via PARPi on the over-abundant abasic sites that occur in a CHD6 negative cell—as our data in Fig. 7g suggests—we suggest that this likely leads to a level of DNA replication blockage that overwhelms DNA damage tolerance, fork protection, and/or BER systems, resulting in fork collapse and an increase in DSBs in S-phase (Fig. 10e). It is also possible that removal of PARP1/2 enzymes trapped at sites of DNA damage, an action carried out by the p97 unfoldase/segregase^12,14, may also be overwhelmed or possibly even could partly depend on CHD6 in certain chromatin regions.

The discovery of factors whose mutation (or manipulation) elicits synthetic lethality in combination with PARPi is of major interest to health researchers globally, with the great challenge(s) being to identify tumors that are likely to respond to PARPi (based on their genetics), and/or how to restore PARPi sensitivity in the event of tumor-evolved PARPi-resistance. The concept of ‘BRCAness’—a phenotype of impaired RAD51-dependent stages of HR mirroring the loss of BRCA1/BRCA2–remains the primary criteria by which synthetic lethal factors to PARPi are identified^9,14. In recent years, however, several PARPi-sensitizing and chromatin-related factors that operate outside of BRCA1/2-dependent HR have been reported, including p97 that remove cytotoxic, trapped-PARP1 from DNA (reducing PARPi effectiveness)^12,14, and the ALC1/CHD1L nucleosome remodeling enzyme that promotes BER intermediate processing following uracil misincorporation or base alkylation⁸¹. Our experiments indicate that CHD6 loss is additive (synthetic lethal) with BRCA1 loss in the context of PARPi sensitivity—a phenotype also supporting separate mechanisms of action for BRCA1/2 and CHD6 in terms of PARPi sensitization. Similar to CHD6, the ALC1/CHD1L chromatin remodeler also binds PAR chains and relocates with equivalent kinetics to DNA damage where PAR is formed, although it does so via a macrodomain⁶². Our observation that CHD6 contributes to abasic site resolution and modifies PARPi sensitivity parallels how Boulton and colleagues⁸¹ envisaged ALC1/CHD1L loss to be synthetic lethal with HR deficiency; taken together with our results and that of others⁸², these studies collectively highlight an emerging understanding that chromatin remodeling is important during APEX1-dependent stages of base excision repair. Perhaps “APEXness”—a phenotype of insufficient APEX1 activity achieved pharmacologically or observed via cancer-related mutation in proteins such as CHD6 or ALC1/CHD1L—is another route towards PARPi sensitivity (or re-sensitization) that will prove beneficial to investigate in a clinical setting.

We emphasize that (despite its name) the ALC1/CHD1L chromatin remodeler is not a CHD-class enzyme, but rather is an ISWI-class remodeler with consequently very different biochemical activity. In fact, ISWI and CHD-class remodelers classically display opposing chromatin remodeling activities⁸⁴, in which the arrangement of nucleosomes (within chromatin) produced by one class of enzyme is reversed by the other, with both classes of activity proven as important to produce chromatin arrangements suitable for processes such as DNA repair^15,85. As one example, the ISWI-class enzyme SNF2H directly ‘opposes’ the CHD-class activity of CHD3.1 during DSB repair in regions of heterochromatin, with both activities needed for productive DSB repair²⁶. It is possible that the activity of CHD6 could cooperate with ALC1/CHD1L at abasic sites to produce a chromatin configuration compatible with subsequent repair stages, with the activity of CHD6 and ALC1/CHD1L each coming into play depending on the chromatin location within which the abasic site arises.

Trapping PARP1/2 enzymes on DNA (and keeping them there) has emerged as a key component of PARPi efficacy as an anti-cancer therapeutic^9,12,14. Under conditions of heightened oxidative stress and/or reduced antioxidant capacity, which we have demonstrated are both consequences of CHD6 loss⁸, abundant free radicals may attack the C1’ position on DNA to generate highly reactive 2-deoxyribonolactones that react with AP lyase enzymes such as DNA polymerase beta, PARP1, or HMCES^86,87,88. In the case of HMCES, it is thought that the formation of the DNA-HMCES crosslink is beneficial to cell health, as it suppresses DSB formation in S-phase by slowing TLS at abasic sites⁸⁹. DNA-HMCES crosslinks form preferentially within single stranded DNA, but are severed in double-stranded DNA via controlled release without proteolysis⁹⁰. PARP1-DNA crosslinks (via abasic site reaction) have been proposed to occur via cysteine addition to 3′-DNA termini produced by abasic site incision⁹¹, and likely pose powerful barriers to DNA replication forks as they accumulate within chromatin. As such, it is logical that the formation and removal of abasic sites, which promote the formation of DNA-to-PARP1 crosslinks enhanced by PARP1/2-trapping via Olaparib⁷³, has emerged as important for PARPi sensitivity. We speculate that the convergence of reduced APEX1-mediated abasic repair capacity (demonstrated here) together with elevated oxidative stress⁸ and demonstrably greater PARP-1 occupancy on chromatin (also demonstrated here) associated with CHD6 loss creates a ‘perfect storm’ for PARP1/2-trapping on DNA via PARPi, and consequently replication fork stalling leading to collapse, DSB formation, and synthetic lethality (between CHD6 and PARPi).

In terms of open questions in need of addressing, further work is needed to resolve relationship between PAR-dependent recruitment of CHD6 to damaged DNA, and whether it may directly enable APEX1 activity, or perhaps also contributes to the removal of AP site-bound or inhibitor-trapped PARP1/2 enzymes. We note that it has been demonstrated^92,93 that PARP1/2 recognize and are able to bind AP sites before they are incised by APEX1, and will be stimulated weakly to produce PAR before being further activated by APEX1 mediated AP site incision and single strand break formation. Therefore, it is mechanistically possible for some PAR-induced CHD6 recruitment to occur that may enable APEX1 activity in otherwise hard-to-incise genomic locations.

Our observations regarding the separate but related function of the PBM, DBD1 and DBD2 of CHD6 highlights specific functional domains within the CHD6 polypeptide that could, with future work, be targeted by small molecules or gene editing interventions to disrupt protein function for therapeutic gain. That the CHD6 DBD2 contributes to nucleosome binding and remodeling activity, albeit in a manner that enhances a DNA-binding function largely conferred by DBD1, provides a biochemical explanation for why the Hallermann-Streiff syndrome point mutation (that occurs within the DBD2) is sufficient to elicit a clinical phenotype. It is also worth noting that the phenotype caused by ablation of the CHD6 catalytic domain (exon 12) in mice impacts brain function and motor coordination, resulting in ataxia⁹⁴. Ataxia is a very common and widely-associated outcome of loss or mutation of base excision repair factors such as XRCC1, Senetaxin, NEIL1, NEIL2, NEIL3 and the alkyladenine DNA glycosylase AAG/MPG^{41,95,96,97,98}. That CHD6 loss in mice produces a comparable outcome is also consistent with the role for CHD6 in base excision repair, that the outcomes of this study suggest.

Methods

Reagents and tissue culture

Phenylmethylsulfonyl Fluoride (PMSF), Wortmannin (WM), Microcystin-LR (MC-LR), Dithiothreitol (DTT), H₂O₂ and hydroxyurea were all from Sigma. Camptothecin, MMS, ATR inhibitor AZD6738/Ceralasertib, PARP1/2 strong trapping inhibitor (PARPi) AZD2281/Olaparib, PARP1/2 weak trapping inhibitor ABT-888/Veliparib, and REV1-REV7 inhibitor (TLSi) JH-RE-06 were obtained from Selleck Chemicals. PARG inhibitor (PARGi) PDD00017273 was from Tocris. RPE-1 hTERT (obtained from ATCC, catalog number CRL-4000) were cultured in 1:1 DMEM/F12 supplemented with 10% (v/v) FCS and 1 mg/mL hygromycin B. RPE-1 cell identity was confirmed by gene sequencing and karyotyping. RPE-1 hTERT, as well as HEK293T and A549 cell lines (ATCC catalog numbers CRL-1573, CRM-CCL-185) used for mass spectrometry and micro-irradiation experiments, respectively (see subsequent sections for details), were all tested regularly for mycoplasma contamination and confirmed to be negative.

Antibodies, Immunoblots and Immunofluorescence

The primary antibodies from this study were as below:

• anti-Actin^Mouse, from Abcam, ab3280, use at 1:4000 in immunoblot

• anti-CHD1^Mouse, from Santa Cruz, sc-271626, use at 1:500 for immunoblot

• anti-CHD3^Rabbit, from Abcam, ab84528, use at 1:500 for immunoblot

• anti-CHD4^Mouse, from Abcam, ab54603, use at 1:400 for immunoblot

• anti-CHD6^Mouse, from Abcam, ab51330, use at 1:200 for immunoblot

• anti-KAP-1^Rabbit, from Abcam, ab10584, use at 1:800 for immunoblot

• anti-γH2AX^Mouse, from Abcam, ab26350, use at 1:800 for immunofluorescence

• anti-RAD51^Rabbit, from Millipore Sigma, PC130, use at 1:1500 for immunofluorescence

• anti-BRCA1^Mouse, from Santa Cruz, sc-6954, use at 1:200 for immunoblot

All secondary antibodies for immunofluorescence = anti-mouse or anti-rabbit IgG coupled to Alexafluor 488 or 594 (Invitrogen Molecular probes; used at 1:800). PBS-washed cells were fixed in 3% (w/v) PFA + 2% (w/v) sucrose for 10 min, permeabilized for 5 min in 0.5% (v/v) Triton X100 (in PBS) and immunostained for 1 h or overnight at 4 °C with primary antibody (diluted in 3% (v/v) newborn goat serum in 1x PBS). Where indicated, cells were counterstained with 0.1 μg/mL DAPI to visualized nuclei and were mounted using Fluoromount G. For RAD51 foci staining, PBS-washed cells were pre-extracted in pre-extraction buffer (25 mM HEPES pH 7.5, 50 mM NaCl, 1 mM EDTA, 3 mM MgCl₂, 300 mM sucrose, 0.5% (v/v) Triton X-100) for 5 minutes on ice before fixing and staining. Images were acquired on a Zeiss AxioObserver Z1 platform microscope with a Plan-Neofluar 40x/0.75 objective and an AxioCam MRm Rev.3 camera. Acquisition software used was Zen Blue and γH2Ax foci were enumerated and analysed in ImageJ using the TANGO plugin as described in ref. ⁶³. RAD51 foci were enumerated using CellProfiler (BROAD Institute) using the Speckle Counting pipeline described in ref. ⁹⁹.

For immunoblotting, cells were harvested by scraping into cold PBS and lysed in 5x the packed cell volume of NETN buffer (150 mM NaCl, 0.2 mM EDTA, 50 mM Tris-HCl pH 7.5 and 1% (v/v) NP-40) or RIPA buffer (150 mM NaCl, 1% NP-40, 0.5% (w/v) sodium deoxycholate, 0.1% (w/v) SDS, 50 mM Tris pH 7.4) supplemented with 1x Roche protease inhibitors, 200 µM PMSF, 1 mM DTT for 10 minutes on ice. Lysates were clarified by centrifuging at 20,000 × g for 10 minutes at 4 °C. Protein concentrations were quantified using DC protein assay (Bio-Rad). For CHD6 blots, samples were incubated in 2X SDS Laemmli sample buffer (6X: 30% (v/v) glycerol, 1 M Tris pH 6.8, 6% (w/v) SDS, 0.01% (w/v) bromophenol blue, 3% (v/v) β-mercaptoethanol) for 5 minutes at room temperature before loading onto an SDS-PAGE gel for separation. For all other blots, samples were boiled at 95 °C then loaded onto an SDS-PAGE gel. Samples were resolved by SDS-PAGE on 8% polyacrylamide (low-bis), 10% or 12.5% polyacrylamide based on molecular weight. Samples were moved through the stacking gel at 75 V for 15 minutes, then the voltage was increased to 135 volts for 75 minutes. Proteins were transferred to nitrocellose at 100 V for 60 minutes at 4 °C in transfer buffer. Nitrocellulose membranes were then blocked with 5% milk powder (w/v) in Tris-buffered saline supplemented with 0.1% (v/v) Tween-20 (TTBS) for 1 h. Membranes were incubated with primary antibodies in TTBS supplemented with 0.1% (w/v) gelatin and 0.05% (w/v) sodium azide overnight at 4 °C with rocking. Membranes were washed 3X with TTBS for 5 minutes and incubated with HRP-secondary antibodies (Bio-Rad) in TTBS and 1% (w/v) milk powder for 1 h. Membranes were washed 3X with TTBS, and visualized using enhanced chemiluminescence (ECL, PerkinElmer) on a ChemiDoc^TM (Bio-Rad). For all immunoblots, uncropped versions are presented in the Source Data File.

Plasmids, mutagenesis and transfection

GFP-tagged CHD6 expression constructs (CHD6^GFP) were made by cloning in full length wildtype human CHD6 cDNA (accession NM_032221; NP_115597) into pEGFP-C2 backbone all under CMV promoters (Clonetech) as described in detail in ref. ⁸. Plasmid transfection was achieved using PolyPlus (VWR) according to the manufacturer’s protocol. GFP-tagged CHD6 plasmid mutations (including amino acid changes, nucleotide positions and codon changes) are indicated in Table 1.

Table 1 CHD6 plasmid mutation summary

BRCA1 siRNA-mediated protein depletion and immunoblot

Wildtype and ∆CHD6 RPE-1 cells were plated at 1 × 10⁵ cells in 6-well plates. 24 h later, cells were transfected with 15 nM ON-TARGETplus SMARTpool BRCA1 or ON-TARGETplus Non-targeting Control Pool siRNA (Horizon Discovery) using Liopfectamine RNAiMAX (Thermo Fisher Scientific) according to the manufacturer’s instructions. 48 h after transfection, cells were then subjected to experimentation per data in Fig. 1. Immunoblotting for BRCA1.

355 nm laser micro-irradiation and live cell imaging

A549 cells were transfected with 2 µg CHD6 expression construct (as indicated) and incubated with 10 μM bromodeoxyuridine (BrdU, from Sigma) for 16-24h prior to micro-irradiation. Small molecule inhibitors were all added to cells 1 h prior to DNA damage induction. DNA damage tracks were induced in live cells (kept at 37 ^oC in a humidified environment at 5% CO₂) using a 355 nm 5 mW self-aligning solid state diode laser (10 µm per sec, 35% power) projected through an EC Plan-Neofluor 40x objective, via a Zeiss PALM MicroBeam laser microdissection module on a Zeiss Axio Observer Z1 platform. Images were captured on an AxioCam MRm Rev.3 camera. Laser irradiation was controlled by RoboSoftware 4.5. Acquisition and analysis software = Zen Pro (Zeiss). CZI image files were captured every 60 seconds for the indicated times, and subsequently analysed using Zen image processing software. For analysis of signal gain, a region of interest was created to cover the area of the laser track and the signal intensity (arbitrary units) within this area was measured over the time course. To measure cell background, a region of interest was cloned and placed next to the region containing the track, providing the signal intensity for the background of an area of identical size within the nuclei. The signal intensity of background is then subtracted from the signal intensity of the track, divided by the sum of both regions of interest and presented as a percent gain in signal intensity with the pre-damage (t = 0) representing baseline.

CRISPR-Cas9 gene editing

CHD6 ablated cell lines were achieved following the protocol outlined in ref. ¹⁰⁰. Here, all sgRNA guides were designed using the BROAD institute CRISPick design tool (https://portals.broadinstitute.org/gppx/crispick/public), and manually selected for estimated ‘on-target efficacy score’ of over 50%, ‘off-target rank’ limited to the top 10–20% of guides generated and having a combined offset (bp distance between the two guides) of no more 40 bp per¹⁰¹. Successful guides were Forward: ACCGTTGGTGCCCGAGGCCTCCCG and Reverse: AAACTCCTGGACTTCGTCCTGGCT produced by Integrated DNA Technologies (https://www.idtdna.com/pages). The oligonucleotides were diluted to 100 µM, then were annealed, and phosphorylated by T4 PNK (NEB, M0201S). AIO plasmids were purchased from Addgene (Plasmid #74120 and #74119), and linearized using BbsI (NEB, R0539S), then the first guide was integrated using a quick ligation kit (NEB, M2200S). Guide integration was confirmed through loss of an internal BamH1 site (NEB, R0136S) and sequencing. Second guide integration was carried out by linearization with Bsa1 (NEB, R0535S), then confirmation through loss of an internal BspD1 site (NEB, R0557S) and sequencing. Guide efficiency was assayed through T7 endonuclease-based mutation detection (NEB, E3321) following the protocol provided. With the guides incorporated into the AIO plasmids, 1 × 10⁶ RPE-1 hTERT cells were electroporated with 4 μg purified plasmid, using a LONZA nucleofection basic epithelial kit (Cat#: VPI-1005). Cells were incubated for 24 h in normal growth media, then single sorted into 96-well plates, gating for the top 1% of the brightest cells. Cells grew for 14 days, then were split and screened for INDEL formation using primers designed around the targeted editing region (Forward: AAGATGGGGCAAAGAAGGCA and Reverse: GCATTAGCACCAAGGGGTTACC with an unedited product size of 282 nt). Here, DNA was extracted from adherent cells using QuickExtract DNA Extraction Solution (Epicentre, QE09050) and using Phusion Hot Start II polymerase (Thermo Scientific, F549L), 96 clones were screened simultaneously. PCR products were separated on a 4% agarose gel and visualized with SafeView (abm, G108) under UV light. Clones identified as having an INDEL were taken forward for further screening by Western blot, qPCR and gene sequencing analysis.

Production of Virus-like particles and Base editing of endogenous CHD6 in RPE-1 cells

Base editing the endogenous CHD6 gene in RPE-1 cells was conducted using virus-like particles (VLPs). A gag-ABE8e SpRY plasmid was constructed from the gag-ABE8e plasmid (Addgene plasmid #181751). The plasmid was digested with BglII, and reconstituted while introducing the SpRY mutations within nCas9 by Gibson Assembly. The following gRNA 5’-GGAAAACCATCCAGTCCAT-3’ was cloned to target endogenous CHD6 as described in¹⁰². For the production of VLPs, Lenti-X 293T cells were plated in a 75 cm flask at 5 × 10⁶ cells/flask in 10 mL complete DMEM. The following day, transfection was performed using Polyplus jetPRIME according to published protocols¹⁰³. Briefly, 400 ng Vsv-g, 3375 ng pBS-CMV-gagpol (Addgene #35614), 1125 ng gag-SpRY-ABE8e, and 4400 ng of gRNA plasmid were combined with 18.6 μL of JetPRIME in OptiMEM for a total volume of 500 μL. The transfection solution was incubated for 10 minutes at room temperature and added dropwise to the cells. The next morning, the media was replaced with fresh complete DMEM. Four days post transfection, media containing VLPs was collected, centrifuged at 500 × g for 5 minutes and filtered. The VLP preparation was concentrated by ultracentrifugation at 4 °C for 2 h at 30,000 × g, and the pellet was resuspended in 100 μL of DMEM F12 media¹⁰³. For the base editing procedure, 3.5 × 10⁴ RPE1 cells were plated in a 48-well plate with 250 μL DMEM F12 media on day 0. On day 1, 50 μL of VLP was added directly to the cells. On day 4, cells were transferred to a larger well, and a small aliquot was collected for genomic DNA extraction and Sanger sequencing. After the initially edited (1 × VLP transduction) RPE1 cells were recovered, a second transduction was performed following the same transduction protocol. Genomic DNA was extracted¹⁰⁴, and editing was verified by Sanger sequencing with the following primers 5’-GGAGTGTGACCCCATCCTTA-3’ and 5’-AGTCCCTGGGTCTCTGATCT-3’.

Protein expression and purification

The region of CHD6, which includes the putative DNA binding domains (aa 1090-1619), was cloned from a pFastBacHT A CHD6 vector, incorporating a C-terminus 6x-His tag. The purified PCR products were subsequently subcloned into a pET MBP TEV expression vector using BrioBrick polypromoter restriction sites (14-Q). The resulting plasmids were propagated and transformed into a Rossetta 2(DE3) E. coli expression cells. To express the CHD6^DBD proteins, an overnight LB starter culture was expanded in a 2 L TB culture supplemented with ampicillin and chloramphenicol antibiotics and grown at 37 °C until reaching an OD₆₀₀ of 0.6. The culture was then induced with 0.4 mM Isopropyl ß-D-1-thiogalactopyranoside (IPTG) and transferred to 16 °C for 16 h. For protein purification, the pelleted cells were resuspended in a solution containing 20 mM HEPES pH 8.0, 750 mM NaCl, 10% (v/v) glycerol, 3% (w/v) sucrose, along with lysozyme and a cOmplete inhibitor tablet (Roche). The solution was subjected to 5 rounds of sonication, with each round consisting of 30 seconds of sonication followed by a 30-second interval on ice. The lysate was then centrifuged at 15,000 × g for 30 minutes at 4 °C. The resulting supernatant was filtered, and 2 mL of pre-equilibrated Ni-NTA beads were added. After an hour of incubation in the 4 °C cold room, the bound protein was separated using a gravity flow column, washed, and eluted with increasing concentrations of imidazole. Fractions containing the protein of interest were pooled together, concentrated using a Vivaspin 6, 10 kDa MWCO concentrator (Cytiva), and subjected to size-exclusion chromatography using a Superdex S200 increase column (Cytiva). Pure fractions were collected, concentrated, and flash-frozen for future use. Mutations were introduced through site-directed mutagenesis, and the proteins were purified following the same protocol described above.

To express the MBP-tagged chromo-ATPase with or without DBD in insect cells, the designated regions were cloned with a C-terminus 6X-His tag into a modified pFastBac His6 MBP TEV cloning vector with BioBrick polycistronic restriction sites (11-C) (N-terminus His tag deleted). The overall plasmid was then transformed into DH10Bac to generate bacmids. To generate the CHD6 baculovirus, ~5 ug of bacmid was initially transfected into approximately 800,000 Sf9 cells in 2 mL SF900 media (Gibco, REF 10902-096) using Cellfectin II (Gibco REF 10362-100) according to manufacturer’s instructions. After 8 h, the transfection reaction was removed, and replaced with fresh 2 mL media, and incubated at 27 ^oC for 7 days. To amplify the baculovirus, the resulting 2 mL of media was then removed and added to 50 mL Sf9 cells at 1.0 × 10⁶ cells/mL, and cultured for 7 days. The culture was then pelleted at 1000 × g, for 10 min, and the 50 mL of supernatant was added to another 50 mL of Sf9 cells, and cultured for another 7 days. The culture was pelleted for the supernatant which contains the P3 amplified CHD6 baculovirus. Optimal expression of CHD6 proteins was achieved by infecting 2.5 mL of P3 CHD6 baculovirus in 50 mL Sf9 cell cultures for 48 hrs. The cells were pelleted and frozen with liquid N₂. To purify the MBP-tagged chromo-ATPase-DBD CHD6 proteins, pellets of insects cells were suspended in a buffer containing 20 mM HEPES pH 8.0, 750 mM NaCl, 10% (v/v) glycerol, 3% (w/v) sucrose, 0.2% Triton-X-100 and a cOmplete inhibitor tablet (Roche). The mixture was sonicated, centrifuged and the supernatant subjected to Ni-NTA pulldown. The eluted fractions containing the protein of interest were pooled together concentrated and applied to a Superdex S200 increase column. Clean fractions were collected, concentrated, and flash-frozen for future use

N-terminal regions of CHD6 corresponding to amino acids 1-80, 81-210, or 211-280 were PCR amplified from a pC2-eGFP-CHD6 vector and subcloned into a pET His10 MBP Asn10 TEV LIC cloning vector (2-CT-10) by Gibson’s Assembly. The plasmid was transformed into BL21(DE3) E. coli expression cells and expanded into a 50 mL TB culture supplemented with ampicillin. Cultures were grown at 37 °C until an OD₆₀₀ of 0.6 was achieved. Cultures were induced with 0.4 mM IPTG and transferred to 16 °C for 16 h. Bacteria were pelleted and resuspended in 20 mM HEPES, 500 mM NaCl, 10% (v/v) glycerol supplemented with 100 µM PMSF, 1 mM DTT, 1 mg/mL lysozyme, cOmplete inhibitor tablet (Roche), 2 mM MgCl₂, 10 µM ZnCl₂ and then subjected to 5 rounds of sonication (with each round consisting of 30 seconds of sonication followed by a 30-second interval on ice). Lysates were clarified by centrifugation at 15,000 × g for 30 minutes at 4 °C. The resulting supernatant was gravity filtered through 0.5 mL of packed, pre-equilibrated Ni-NTA beads (Qiagen). The beads were washed in 20 mM HEPES, 500 mM NaCl, 10% (v/v) glycerol and 10 mM imidazole, and protein was eluted with increasing concentrations of imidazole. Fractions were analyzed by SDS-PAGE and fractions containing protein of interest were pooled and concentrated using a Vivaspin 6, 10 kDa MWCO concentrator (Cytiva). Pooled protein was further purified by size-exclusion chromatography using a Superdex S200 increase column (Cytiva). Fractions were analyzed by SDS-PAGE and pure fractions with protein of interest were concentrated and stored at −80 °C for further use.

PAR binding assay

Purified proteins were slot-blotted on a 0.2 µm pore–size nitrocellulose membrane (Bio-Rad) using a Bio-Dot® microfiltration apparatus (Bio-Rad). Twenty-five picomoles of MBP and MBP-tagged CHD6 fragments diluted in 200 μL of 1x TBS (20 mM Tris-HCl pH 7.5, 150 mM NaCl) were applied to the vacuum manifold sample template. Calf thymus histone H1¹⁰⁵ (Calbiochem) and MBP were respectively used as positive and negative PAR-binding proteins. Because of the high affinity of histone H1 to PAR and to avoid signal saturation, only one pmol of histone H1 was slot blotted for PAR overlay assays. Following complete aspiration of the protein samples, the nitrocellulose membrane was rinsed three times in 50 mL TBST (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.01% (v/v) Tween-20). The membrane was incubated overnight with TBST containing 100 nM PAR purified by dihydroxyboronyl Bio-Rex (DHBB) chromatography⁴⁸. The membrane was extensively washed with TBST and blocked with a PBS-MT solution (PBS supplemented with 5% (w/v) milk and 0.1% (v/v) Tween-20) for 1 h. The membrane was then incubated with the anti-PAR monoclonal antibody clone 10H (Tulip Biolabs) for 1 hour. The membrane was extensively washed with PBS-T and incubated with peroxidase-conjugated anti-mouse secondary antibodies (Jackson ImmunoResearch) for 30 min. The membrane was finally washed three times with PBS-T and signal was detected using the Western Lightning® Plus ECL chemiluminescence substrate (PerkinElmer). SYPRO™ Ruby protein blot stain was used according to the manufacturer's protocol (Bio-Rad) to assess the efficiency of protein transfer to the nitrocellulose membrane. Fluorescence imaging was acquired using a Geliance CCD-based bioimaging system (PerkinElmer). Protein purity was assessed by SDS-PAGE (4-12% Criterion™ XT Bis-Tris gel, Bio-Rad) and visualized using fluorescence imaging of the gels following staining with SYPRO™ Ruby protein gel stain (Bio-Rad).

EMSA assays

All proteins were buffer-exchanged into a DNA-binding buffer containing 20 mM HEPES pH 7.6, 50 mM KCl, 5% (v/v) glycerol, 1 mM DTT, and 2 mM MgCl₂. To generate the indicated concentrations shown in the figure, a serial dilution of the proteins under study was performed. Then, 20 µl of each protein dilution was mixed with 120 nM FAM-labelled 50-mer DNA, resulting in a final DNA concentration of 60 nM. The mixture was incubated at room temperature for 30 minutes. Subsequently, 18 µL of the solution was loaded onto a 1.2% agarose gel. The gel was electrophoresed at 90 V for 40 minutes at room temperature and imaged using Chemidoc (BioRad) imaging systems.

Fluorescence polarization

DNA binding activity of CHD6 DBD wild-type and mutants were quantified with end-point fluorescence polarization assays. A 5’-FAM-labelled 35-mer DNA (Integrated DNA Technologies) was used as a substrate at 5 nM. Increasing concentration of proteins were incubated with fixed concentration of substrate in 10 mM Tris-HCl, pH 7.5, 50 mM KCl, 3% (v/v) glycerol and 1 mM EDTA for 30 min at room temperature. Polarization was measured with a BMGlabtech PHERAstar FSX microplate reader. Curves were fitted with a quadratic binding equation as implemented in GraphPad Prism software.

End-point ATPase assays

ATPase activity assays were conducted using the ADP-Glo^TM kinase kit. The procedure involved initially incubating 0.4 μM of the protein with 500 μM of Ultra-Pure ATP, in a buffer containing 20 mM HEPES (pH 7), 50 mM NaCl, 10% (v/v) glycerol, 5 mM MgCl₂, and 0.01% (v/v) Triton X-100. This mixture, in a total volume of 5 µL, was incubated for 1 h at room temperature. Prior to this, the proteins were mixed with 0.2 μM 601 DNA and 0N50 nucleosomes. Following the incubation, 5 µL of ADP-Glo^TM Reagent was added to the mixture and left for 40 minutes to eliminate any unhydrolyzed ATP. The ADP, a by-product of ATP hydrolysis, was then converted back into ATP. This newly formed ATP was subsequently transformed into light by luciferase reaction upon the addition of 10 µL of Kinase Detection Reagent. The resulting luminescence was measured using a BMGlabtech PHERAstar FSX microplate reader.

Fluorescence resonance energy transfer

To prepare nucleosomes for bulk FRET (Fluorescence Resonance Energy Transfer) analysis, we specifically attached Cy3 maleimide (from Cytiva) to the mutated residue K119C of human histone H2A, following a previously established method¹⁰⁶. These labeled H2A histones were then combined with H2B, H3, and H4 proteins to create an octamer, as per¹⁰⁷. We utilized PCR (Polymerase Chain Reaction) to produce Cy5-tagged DNA, which had 50 nucleotides extending from the exit site. Nucleosomes positioned at this end (termed 0N50) were assembled using the aforementioned method. The reaction mixtures, each 20 µL in volume, contained 0.4 μM of the enzyme, 0.03 μM nucleosome. These were prepared in a solution of 20 mM HEPES pH 7.0, 50 mM KCl, 10% (v/v) glycerol, 10% (w/v) glucose, 5 mM MgCl₂, and 0.01% (v/v) Triton X-100, 5 mM DTT. The remodeling process was initiated by adding 1 mM ATP at 30 °C, using an integrated injector. For assessing the kinetics of nucleosome remodeling, we excited the Cy3 dye at 540 nm and recorded its emission at 575 nm, along with the Cy5 emission at 670 nm, using a BMGlabtech PHERAstar FSX microplate reader.

Site-specific KillerRed DNA damage system

U2OS 2-6-3 cells were co-transfected with 1.5 µg KillerRed-LacR-stop and 1 µg GFP-CHD6 or GFP-XRCC1 as we previously did in ref. ⁸. Immediately after transfection, dishes containing cells were wrapped in aluminum foil to prevent KillerRed activation. The induction of ROS by activation of KillerRed was done essentially as described previously¹⁰⁸. Briefly, 24 h after transfection cells were exposed to a 24 W Sylvania Lynx-L compact white fluorescent bulb for 300 s in a customized stage 15 cm away from the light bulb. The 300 s exposure to light at a rate of 24 W ( = 24 J/s) results in a dose of 7200 J. After exposure, cells were immediately fixed for 10 minutes with 3% formaldehyde and 2% (w/v) sucrose.

FPG-modified comet assay

The day before, 0.8% (w/v) in 1x PBS agarose (Invitrogen) plugs were prepared on frosted slides (VWR). The lysis buffer (10 mM Tris pH 10, 2.5 M NaCl, 0.1 M EDTA) and electrophoresis buffer (50 mM NaOH, 1 mM EDTA) were also prepared the day before. Parental and ΔCHD6 RPE-1 cells were collected, washed once with cold PBS, counted and diluted to 5 × 10⁴ cells per mL. Cells were resuspended in PBS containing the indicated dose of H₂O₂ and incubated on ice for 20 minutes, collected, resuspended in warm media and returned to the incubator. 1 mL of cells was removed per time point. After treatment, cells were washed once with 1x PBS and the pellet collected and resuspended in 1 mL of PBS. Using 1.2% (w/v) low-melting agarose (Invitrogen) in 1x PBS, 200 µL of cells were placed in a fresh tube to which 200 µL of warm low-melting agarose was added, mixed and 200 µL was quickly pipetted on top of pre-set agarose plugs under a coverslip. The plugs containing cells were left in the dark at 4 °C to set for 30 minutes. For lysis, coverslips were removed and slides placed in a black container in a dark cold room and incubated in lysis buffer (supplemented with 1% (v/v) DMSO and 1% (v/v) Triton X-100) for 1 h. Cells were washed 3x in cold Fpg-digest buffer (40 mM HEPES pH 8.0, 100 mM KCl, 0.5 mM EDTA, 0.2 mg/mL BSA), and 75 µL of 2.67 U/mL Fpg in Fpg-digest buffer was added on top of the agarose plugs with coverslips before being placed in the dark at 37 °C. After Fpg digest, slides were washed with cold dH₂O, then placed in an electrophoresis chamber and covered with electrophoresis buffer (supplemented with 1% (v/v) DMSO) for 45 minutes in the dark to neutralize the pH. Electrophoresis was carried out at 25 V for 15 minutes, comets were neutralized by submerging plugs in 0.4 M Tris pH 7.0 until ready to score. Immediately prior to scoring, slides were stained with 1:10,000 dilution of SYBR green supplemented with 4 µg/µL of antifade (stock at 11 µg/µL) for 5–10 minutes at room temperature in dark, removed and scored immediately. Comet scoring was performed on a Zeiss Axiovision microscope with a camera associated with Comet Assay IV software (Perspective Instruments). A total of 150 comets were generally scored per condition, per repeat, and the data was expressed as the average comet tail moment on a scatter plot.

FM-HCR assay

FM-HCR assays were carried out in RPE-1 and RPE-1^ΔCHD6 cells, using the same approach we used in ref. ⁸ and described in refs. ^69,70,71. RPE-1 and RPE-1^ΔCHD6 were maintained in culture at 37 °C under 5% O₂ and 5% CO₂. Cells were seeded 2 days prior to harvesting for transfection (cells were harvested for transfection at ~85% confluency). A Neon Transfection System (Thermo Fisher Scientific) was used to transfect 500,000 cells using a 10 μL tip at 1400 V in one 30 ms pulse. Cells were transfected with reporter plasmids with DNA lesions or corresponding undamaged reporter plasmids^69,70,71. After transfection, RPE-1 and RPE-1^ΔCHD6 cells were seeded into 6-well plates containing 2 mL of media per well, and returned to the incubator. At 24 h post transfection, cells were harvested and analyzed for DNA Repair Capacity (DRC) by flow cytometry. Fluorescent protein signal was quantitated for each respective plasmid for 15,000 cellular events per replicate (n = 3). Percent reporter expression for fluorescent proteins was quantified as previously described^69,70,71. For reporter assays where fluorescent protein expression is inversely related to repair, observed % reporter expression was subtracted from 100%.

Abasic site quantification

RPE-1 and RPE-1^ΔCHD6 were treated with 0.04% (v/v) MMS for 1 h, washed twice with 1x PBS, and allowed to recover in conditioned media for the indicated times. Cells were collected by trypsinization, washed twice with 1x PBS, and genomic DNA was extracted from RPE-1 and RPE-1^ΔCHD6 using Qiagen DNeasy® Blood & Tissue kits. The number of AP-sites/100 kb was determined using Dojindo Nucleostain DNA Damage Quantification Kit – AP Site Counting. Briefly, genomic DNA was labelled with a biotin coupled-aldehyde reactive probe. Labelled DNA was purified by ethanol precipitation overnight and bound to microplates. Samples were washed 5x with 1x TBST, probed with streptavidin-ARP for 1  h at 37 °C, washed 5x with 1x TBST and detected using TMB substrate. Absorbance was measured at 650 nm using a SpectraMax ID3 from Molecular Devices.

Affinity purification coupled mass spectrometry

Samples analyzed were CHD6-3xHA or IgG isotype controls, and both sample and control were analyzed in biological triplicate (n = 3). HEK293T cells expressing CHD6^3xHA were cross-linked with 1% (v/v) formaldehyde in serum-free medium for 8 minutes at room temperature. The reaction was quenched by adding glycine to a final concentration of 0.1 M. Cells were harvested by scrapping into cold PBS and pelleted at 2000 × g, Cells were resuspended in nuclei extraction buffer (50 mM HEPES pH 7.5, 140 mM NaCl, 1 mM EDTA, 10% (v/v) glycerol, 0.5% (v/v) NP-40, 0.25% (v/v) Triton X-100) and left on ice for 10 minutes, then centrifuged at 2000 × g. Nuclei were washed 1X in 10 mM Tris pH 8.0, 200 mM NaCl, 1 mM EDTA and 0.5 mM EGTA. Nuclei were resuspended in nucleus lysis buffer (10 mM Tris pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.1% (w/v) sodium deoxycholate, 0.5% (w/v) sodium lauroylsarcosine) and sonicated for 30 minutes in a Bioruptor® UCD-200 (VWR). 30 µL of 10% (v/v) Triton X-100 were added to each sampled, vortexed, and then lysates were clarified by centrifugation at 20,000 × g.

Anti-HA agarose beads (Pierce) were equilibrated by washing 3x in PBS, then blocked for one hour in 2% (w/v) BSA in 1x PBS. Lysates were applied to blocked beads and immunoprecipitated for 1.5 h at 4 °C. Samples were washed 3x in bead wash buffer (50 mM HEPES pH 7.6, 1 mM EDTA, 0.7% (w/v) sodium deoxycholate, 1% (v/v) NP-40 and 0.5 M LiCl), followed by 3x washes in 50 mM ammonium bicarbonate. Samples were reduced with 10 mM DTT at 56 °C for 30 min, then alkylated with 25 mM iodoacetamide for 30 minutes in the dark. Samples were digested with 1 µg MS grade trypsin (Pierce) for 4 h at 37 °C; initial digests were supplemented with 1 µg of trypsin and digest was continued overnight. Tryptic peptides were collected by centrifugation at 10,000 × g for 1 minute. 20 µL of extraction solution (25% (v/v) acetonitrile (EMD Millipore; v/v), 0.1% formic acid (ThermoFisher) in water (EMD Millipore) was added to beads and gently vortexed. Peptides were retrieved by centrifugation at 10,000 × g and the extraction was performed twice. Peptides were lyophilized and resuspended in 1% formic acid in water.

Peptides were analyzed on an Orbitrap Fusion Lumos Tribrid mass spectrometer (Thermo Scientific) operated with Xcalibur (version 4.4.16.14) and coupled to a Thermo Scientific Easy-nLC 1200 system. Peptides were loaded onto a C18 trap (75 µm x 2 cm, Acclaim PepMap 100, ThermoScientific) at a flow rate of 2 µL/min of solvent A (0.1% formic acid in LC-MS grade H₂O). Peptides were separated on a C18 analytical column (75 µm x 50 cm, PepMap RSLC C18, ThermoScientific) using the following gradient: 5-28% solvent B (0.1% formic acid in 80% LC-MS grade acetonitrile) at a flow rate of 300 nL/min over 40 minutes, followed by an increase to 40% solvent B over 5 minutes. Peptides were electro-sprayed using 2.1 kV into the ion transfer tube of the Orbitrap Lumos operating in positive mode. An MS1 scan was acquired from 375-1575 m/z at a resolution of 120,000 FWHM (full width at half maximum). Precursor ions with an intensity greater than 5,000 were subjected to HCD fragmentation at a collision energy of 30%. The Orbitrap was operated using top speed mode with a 3 second cycle time for precursor ion selection. Orbitrap AGC was set at 4e⁵ charges and maximum injection time was set at 50 ms. Dynamic exclusion of precursors was enabled for 30 s.

RAW files were converted to MGF files using RawConverter (v1.1.0.18, The Scripps Research Institute) operating in a data dependent mode. Monoisotopic precursors with charges between + 2 and + 7 were selected for conversion. MGF files were searched against a database containing the UNIPROT human proteome (02.08.2024) and common contaminants using Mascot (Matrix Sciences; version 2.7). Trypsin was used as the enzyme with a maximum number of missed cleavages set to 1. Peptide charge was set to + 2 or greater and carbamidomethylation of cysteine was set as a fixed modification. Oxidation of methionine was set as a variable modification. Precursor mass error tolerance was set to 10 ppm, and fragment ion mass error tolerance was set to 0.6 Da. Peptides with a confidence score of 95% or greater were kept for further analysis.

The Lumos raw data files (*.raw) were converted into Mascot Generic Format (MGF) using RawConverter (v1.1.0.18; The Scripps Research Institute) operating in a data dependent mode. Monoisotopic precursors having a charge state of + 2 to + 7 were selected for conversion. This MGF file was used to search a human database using Mascot algorithm (Matrix Sciences; version 2.7). Search parameters for MS data included trypsin as enzyme, a maximum number of missed cleavage of 1, a peptide charge equal to 2 or higher, cysteine carbamidomethylation as fixed modification, methionine oxidation as variable modification and a mass error tolerance of 10 ppm. A mass error tolerance of 0.6 Da was selected for the fragment ions. Only peptides identified with a score having a confidence higher than 95% were kept for further analysis. The Mascot data files were imported into Scaffold (v5.3.2, Proteome Software Inc) for comparison of different samples based on their mass spectral counting. Resulting protein hits were compared to the CRAPome and deemed high confidence hits if their SAINT score > 0.5 and fold enrichment > 1.5. Protein interaction networks were generated in Cytoscape using the STRINGapp plugin.

We have updated the description of the UNIPROT human proteome with its unique proteome identifier (UP000005640P) alongside the date we downloaded it. Peptide lengths were unrestricted for the database search.

Statistical analysis

Statistical analysis was performed using Prism software. Data represented as either arithmetic (if raw values contained zeros) or geometric means (if data was log normal) + 95% confidence intervals, as indicated. Significant differences were calculated using Student’s t-test or ANOVA, as indicated. For laser micro-irradiation data, differences in the overall recruitment to damage sites was determined by a two-way ANOVA (multiple conditions with multiple time points) and subsequently analyzed by a Tukey’s test to determine differences at individual time points. Please see Supplementary Table 1 for full details of all comparisons. For all tests, specific significant P-values are indicated on figures and also as follows: ns = non-significant ( > 0.05); * = statistically significant ( < 0.05); ** = statistically significant ( < 0.01); *** = statistically significant ( < 0.001); **** = statistically significant ( < 0.0001). Experiments were repeated a sufficient independent number of times to ensure reproducibility of results, with a minimum of three independent experimental repeats. See specific figure legends and method sections for details.

Graphic design

All figures and images in this study were generated by the authors using Adobe Illustrator or Microsoft PowerPoint, and are not subject to copyright elsewhere. Figure elements created using Biorender.com were done so under valid license held at the University of Calgary by individual laboratories, with license details indicated in each relevant figure.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.